Modelling and Design of Local Energy Systems Incorporating Heat Pumps, Thermal Storage, Future Tariffs, and Model Predictive Control

Modelling and Design of Local Energy
Systems Incorporating Heat Pumps,

Thermal Storage, Future Tariffs, and
Model Predictive Control
Andrew Lyden B.Sc. M.Sc.
A thesis submitted for the degree of Doctor of Philosophy
Energy Systems Research Unit

Department of Mechanical and Aerospace Engineering
University of Strathclyde, Glasgow
June 5, 2020
Declaration
This thesis is the result of the author’s original research. It has been composed by the
author and has not been previously submitted for examination which has led to the award
of a degree.
The copyright of this thesis belongs to the author under the terms of the United
Kingdom Copyright Acts as qualified by University of Strathclyde Regulation 3.50. Due
acknowledgement must always be made of the use of any material contained in, or derived
from, this thesis.
Signed:
Date: 05/06/20
2|
Abstract
The planning-level design of local energy systems requires sufficiently capable modelling
tools which incorporate heat pumps, thermal storage, future electricity markets, and
predictive control strategies. Gaps were identified in a review of existing local energy
system tools: (i) ability to adapt and access source code; (ii) temperature dependence for
heat pump models; (iii) stratification model for thermal storage models; (iv) modelling of
evolving electricity markets; and (v) ability to explore predictive controls.
A novel modelling tool, PyLESA, has been developed to tackle these gaps and to
explore predictive and non-predictive controls, and existing and future electricity tariffs.
PyLESA possesses the following modelling capabilities: resources, and electrical and heat
demands; electricity production; heat pump; hot water tank; electricity tariffs; fixed order
control (FOC); model predictive control (MPC); and KPIs.
A sizing study for a proposed design of a district heating network was devised to
showcase an application of PyLESA. Aims were to compare control strategies and
electricity tariffs, and to identify an optimal size combination of heat pump and hot water
tank. Comparisons between control strategies found that MPC offers savings over FOC.
The lowest levelized cost of heat for the existing electricity tariffs was for the time-of-use
tariff with MPC, 750kW heat pump and 500m3 hot water tank.
A wind tariff, with a 1000kW heat pump and 2000m3 hot water tank, benefits from
using MPC over the FOC: levelized heat costs reduce by 41.1%, and heat demand met by
RES increases from 52.8% to 70.2%. It is shown that the proposed design can be sized
using existing electricity tariffs, and additional hot water tank capacity added later to
benefit from future tariffs.
The results convey the advantage of combining flexible tariffs with optimally sized
thermal storage and showcase PyLESA as capable of usefully aiding the design of local
energy systems.
3|
Table of Contents
1. Introduction ................................................................................................... 19
1.1. General Context......................................................................................................... 19
1.2. Transitioning Energy System ................................................................................... 20
1.3. Local Energy Systems ............................................................................................... 22
1.4. Heat Pumps ................................................................................................................ 23
1.5. Thermal and Electrical Storage................................................................................ 24
1.6. Electricity Markets ..................................................................................................... 25
1.7. Control Strategies ...................................................................................................... 25
1.8. Modelling Local Energy Systems ............................................................................ 27
1.9. Uncertainty in Energy System Modelling............................................................... 30
1.10. Related Projects ..................................................................................................... 32
1.11. Problem Statement ................................................................................................ 33
2. Research Question, Aims, and Methodology ................................................ 34
2.1. Research Question..................................................................................................... 34
2.2. Research Aims............................................................................................................ 34
2.3. Research Methodology ............................................................................................. 35
3. Review of Existing Modelling Tools ............................................................... 37
3.1. Initial Screening Process ........................................................................................... 38
3.2. Categorisation of Modelling Tool Capabilities ...................................................... 42

3.2.1. Input Data Requirements and Input Support Capabilities .................................................... 42
3.2.2. Electrical and Thermal Supply Technology Modelling Capabilities ..................................... 45
3.2.3. Design Optimisation and Output Capabilities ......................................................................... 47
3.2.4. Control Modelling Capabilities including DSM ....................................................................... 49
3.2.5. Storage Modelling Capabilities and Underlying Models ......................................................... 53
4|
3.2.6. Practical Considerations ............................................................................................................... 60
3.2.7. Discussion ....................................................................................................................................... 63
3.3. Tool Selection Process .............................................................................................. 65

3.3.1. Determination of Requirements ................................................................................................. 65
3.3.2. Scoring of Tools Against Requirements .................................................................................... 66
3.3.3. Example Application of the Modelling Tool Selection Process ........................................... 66
3.3.4. Discussion ....................................................................................................................................... 69
3.4. Gaps to be Addressed ............................................................................................... 70
4. Introducing PyLESA: Modelling Methodology, Capabilities, Framework, and

Applications .......................................................................................................... 71
4.1. Brief Introduction to PyLESA ................................................................................ 71
4.2. Modelling Methodology ........................................................................................... 72
4.3. Motivation for Choosing Python ............................................................................ 73
4.4. Tool Capabilities ........................................................................................................ 74
4.5. Tool Framework ........................................................................................................ 78
4.6. Tool Applications: Potential Design Studies ......................................................... 80

4.6.1. Feasibility Study ............................................................................................................................. 80
4.6.2. Operation Study ............................................................................................................................. 81
4.6.3. Sizing Study .................................................................................................................................... 82
4.7. Summary and Next Steps ......................................................................................... 83
5. Detailing PyLESA: Underlying Models and Assessment Methods ................ 84
5.1. Resource and Demand Assessment and Input Methods..................................... 85

5.1.1. Local Resources ............................................................................................................................. 87
5.1.2. Electrical Demand ......................................................................................................................... 89
5.1.3. District Heating Demand ............................................................................................................. 90
5.2. Electrical Production Technologies ........................................................................ 94

5.2.1. Wind Turbine Model and Validation ......................................................................................... 95
5|
5.2.2. Photovoltaics (PV) Model and Validation ................................................................................ 97
5.3. Heat Pumps and Auxiliary Heat Units ................................................................... 98

5.3.1. Generic Regression Model .........................................................................................................102
5.3.2. Lorentz Model ..............................................................................................................................103
5.3.3. Standard Test Regression Model ..............................................................................................105
5.3.4. Auxiliary Heat Units and Schematic of Setup ........................................................................107
5.4. Electrical Storage .....................................................................................................108
5.5. Hot Water Tank .......................................................................................................110

5.5.1. Ambient Heat Loss .....................................................................................................................113
5.5.2. Charging and Discharging ..........................................................................................................117
5.5.3. Mixing Between Nodes ..............................................................................................................118
5.5.4. Final Energy Balance Equation .................................................................................................119
5.5.5. Validation ......................................................................................................................................120
5.6. Electricity Tariffs .....................................................................................................129

5.6.1. Existing ..........................................................................................................................................131
5.6.2. Future.............................................................................................................................................132
5.7. Fixed Order Control ...............................................................................................134
5.8. Model Predictive Control .......................................................................................137
5.9. Modelling Outputs (KPIs) .....................................................................................140

5.9.1. Levelized Cost of Energy ...........................................................................................................141
5.9.2. Levelized Cost of Heat ...............................................................................................................141
5.9.3. On-Site RES Used .......................................................................................................................142
5.9.4. Grid RES Used ............................................................................................................................142
5.9.5. Demand Met by RES ..................................................................................................................142
5.9.6. Heat Met by RES .........................................................................................................................142
5.9.7. Peak HP to Demand Ratio ........................................................................................................143
5.9.8. Heat Pump Utilisation ................................................................................................................143
6|
5.9.9. Days of Storage Content ............................................................................................................143
5.10. Discussion ............................................................................................................143
6. Applying PyLESA: Proposed Design and Sizing Study ................................. 149
6.1. Study Aims, KPIs, and Methodology ...................................................................150
6.2. Case Study Description...........................................................................................153
6.3. Proposed Design .....................................................................................................156
6.4. Input Data and Output Requirements .................................................................160

6.4.1. Resource and Demand Assessment and Input Methods .....................................................160
6.4.2. Electrical Production Technologies .........................................................................................160
6.4.3. Heat Pumps and Auxiliary Heat Units.....................................................................................161
6.4.4. Hot Water Tank ...........................................................................................................................161
6.4.5. Electricity Tariffs .........................................................................................................................162
6.4.6. Fixed Order Control ...................................................................................................................163
6.4.7. Model Predictive Control ...........................................................................................................163
6.4.8. Output Requirements .................................................................................................................163
6.5. Application of PyLESA Modelling Methodology ..............................................164
6.6. Outputs from PyLESA ...........................................................................................166

6.6.1. Comparison to EnergyPLAN ...................................................................................................173
6.7. Operational Analysis Results .................................................................................177

6.7.3. Discussion .....................................................................................................................................193
6.8. Sizing Results............................................................................................................195

6.8.3. Discussion .....................................................................................................................................220
6.9. Results Summary .....................................................................................................221
7|
7. Conclusions ................................................................................................. 223
7.1. Contributions ...........................................................................................................223

7.1.1. Addressing Modelling Capability Gaps ...................................................................................223
7.1.2. Exploration of Developed Control Strategies and Electricity Tariffs ................................226
7.2. Future Tool Development .....................................................................................228

7.2.1. Tool Framework ..........................................................................................................................228
7.2.2. Demand Modelling......................................................................................................................229
7.2.3. Heat Pump and Auxiliary Units ................................................................................................230
7.2.4. Thermal Storage ...........................................................................................................................230
7.2.5. Electrical Storage and Fuel Synthesis .......................................................................................231
7.2.6. Electricity Tariffs and Balancing Markets ...............................................................................231
7.3. Future Tool Applications .......................................................................................232
7.4. Final Summary .........................................................................................................233
8. References .................................................................................................. 235
8|
List of Figures
Figure 1.1: Average carbon intensity factor for electricity consumed in Scotland ........... 22
Figure 4.1: Models and energy flows of PyLESA.................................................................. 77
Figure 4.2: Framework of PyLESA showing workflow of a run ........................................ 79
Figure 5.1: Electrical and heat demand input requirements ................................................ 86
Figure 5.2: Heat demand profile generator input requirements .......................................... 87
Figure 5.3: Weather resource input requirements ................................................................. 87
Figure 5.4: Screenshot of renewables.ninja website for obtaining weather resource data
from the MEERA dataset with the example of Findhorn, Scotland .................................. 89
Figure 5.5. Flow diagram of district heating demand prediction method with user inputs
in text bold .................................................................................................................................. 93
Figure 5.6: Wind turbine input requirements ......................................................................... 94
Figure 5.7: PV input requirements ........................................................................................... 95
Figure 5.8: Monthly wind power output for each month and output for year ................. 96
Figure 5.9: Hourly wind power output for example winter and summer weeks .............. 96
Figure 5.10: Hourly PV output for year at Findhorn ............................................................ 98
Figure 5.11: Heat pump modelling input sheet with standard test regression model
activated .....................................................................................................................................101
Figure 5.12: Generic ASHP regression model COP variation for a 14kW heat pump .102
Figure 5.13: Flow diagram of generic regression model .....................................................103
Figure 5.14: Flow diagram of Lorentz model ......................................................................104
Figure 5.15: Lorentz model output from EnergyPRO .......................................................104
Figure 5.16: Standard test regression model variation of COP .........................................106
Figure 5.17: Standard test regression model variation of duty (maximum output) ........106
Figure 5.18: Flow diagram of standard test regression model ...........................................107
Figure 5.19. Heat pump, thermal store, electric heater, and heat network using a 2-port
connection .................................................................................................................................108
Figure 5.20: Electrical storage input requirements ..............................................................108
9|
Figure 5.21. Simple battery model schematic .......................................................................109
Figure 5.22: Hot water tank modelling input requirements ...............................................110
Figure 5.23: Decision tree for modelling stratification of hot water tank........................112
Figure 5.24: Energy balance for node i, including the top, bottom, above, and below nodes
.....................................................................................................................................................113
Figure 5.25. Hot water tank schematic to be modelled ......................................................114
Figure 5.26: Schematic showing horizontal conduction loss from hot water tank through
the insulation layer ...................................................................................................................115
Figure 5.27: Flow diagram for hot water tank model .........................................................120
Figure 5.28: 10-hour charging period showing modelled and monitored node
temperatures. Four 15-minute timesteps is one hour. ........................................................121
Figure 5.29: Absolute error on node temperatures for 10-hour charging period. Four 15-
minute timesteps is one hour. ................................................................................................122
Figure 5.30: Discharging period showing modelled and monitored node temperatures.
Four 15-minute timesteps is one hour. .................................................................................124
Figure 5.31: Absolute error on node temperatures for discharging period. Four 15-minute
timesteps is one hour. ..............................................................................................................125
Figure 5.32: Two-day period with charging and discharging periods showing modelled
and monitored node temperatures. Sixteen 15-minute timesteps is four hours. ............126
Figure 5.33: Absolute error on node temperatures for two-day period. Sixteen 15-minute
timesteps is four hours. ...........................................................................................................127
Figure 5.34: Comparison of output from multi-node model and energetic model. Sixteen
15-minute timesteps is four hours. ........................................................................................128
Figure 5.35: Input requirements for time-of-use tariff linked to wholesale market .......129
Figure 5.36: Input requirements for variable periods tariff ...............................................130
Figure 5.37: Input requirements for a future wind-based tariff coupled with a flat rate base
.....................................................................................................................................................130
Figure 5.38: Example of future work to be done by including balancing mechanism and
grid services ...............................................................................................................................131
Figure 5.39: Existing electricity tariffs over 72 hours .........................................................132
10 |
Figure 5.40: Top graph: wind farm modelled output including upper band and lower band
over 72 hours; Bottom graph: renewable electricity tariff with discounts and premiums
applied over the same 72 hours..............................................................................................133
Figure 5.41: Flow diagram for electricity tariff generator...................................................134
Figure 5.42: Fixed order control input requirements ..........................................................134
Figure 5.43: Flow diagram showing process i as a chunk of the flow of results and checks
when running the fixed order controller...............................................................................136
Figure 5.44: Model predictive control input requirements ................................................137
Figure 5.45: Flow diagram for model predictive control....................................................140
Figure 6.1: Panoramic view of WWHC with the 6 high rise towers and 5 low rise blocks
.....................................................................................................................................................154
Figure 6.2: WWHC hot water tank and flue chimney in foreground, and boiler house in
background ................................................................................................................................154
Figure 6.3: Schematic of existing setup of the WWHC energy centre .............................155
Figure 6.4: Schematic of proposed design ............................................................................159
Figure 6.5: Exponential function of hot water tank volume and specific price ..............162
Figure 6.6: Total PV, wind turbines and combined renewable power production in bar
charts ..........................................................................................................................................167
Figure 6.7: Electrical demand, renewable generation, and renewable generation usage
plots for summer week ............................................................................................................168
Figure 6.8: Grid interaction plot for summer week ............................................................169
Figure 6.9: Heat demand, heat pump output, and heat pump electrical usage plots for
summer week ............................................................................................................................171
Figure 6.10: Heat pump performance with COP and duty................................................171
Figure 6.11: Thermal storage charging and discharging as energy, and node temperatures
for summer week ......................................................................................................................172
Figure 6.12: Operational graphs with FOC and flat rate tariff over a summer week ....178
Figure 6.13: Operational graphs with FOC and flat rate tariff over a winter week........179
Figure 6.14: Operational graphs with FOC and variable periods tariff over a summer week
.....................................................................................................................................................180
11 |
Figure 6.15: Operational graphs with FOC and variable periods tariff over a winter week
.....................................................................................................................................................181
Figure 6.16: Operational graphs with FOC and time-of-use tariff over a summer week
.....................................................................................................................................................182
Figure 6.17: Operational graphs with FOC and time-of-use tariff over a winter week .183
Figure 6.18: Operational graphs with FOC and wind tariff over a summer week .........184
Figure 6.19: Operational graphs with FOC and wind tariff over a winter week ............185
Figure 6.20: Operational graphs with MPC and flat rate tariff over a summer week ....186
Figure 6.21: Operational graphs with MPC and flat rate tariff over a winter week .......187
Figure 6.22: Operational graphs with MPC and variable periods tariff over a summer week
.....................................................................................................................................................188
Figure 6.23: Operational graphs with MPC and variable periods tariff over a winter week
.....................................................................................................................................................189
Figure 6.24: Operational graphs with MPC and time-of-use tariff over a summer week
.....................................................................................................................................................190
Figure 6.25: Operational graphs with MPC and time-of-use tariff over a winter week.191
Figure 6.26: Operational graphs with MPC and wind tariff over a windless winter 10-day
period .........................................................................................................................................192
Figure 6.27: 3D plot of capital cost for a range of size combinations .............................197
Figure 6.28: 3D plot of LCOH (levelized cost of heat) for FOC with flat rate tariff ....198
Figure 6.29: 3D plot of ORESpv (on-site PV self-consumption) for FOC with flat rate
tariff ............................................................................................................................................199
Figure 6.30: 3D plot of HRESpv (heat demand from on-site PV) for FOC with flat rate
tariff ............................................................................................................................................200
Figure 6.31: 3D plot of LCOH (levelized cost of heat) for FOC with variable periods
tariff ............................................................................................................................................201
Figure 6.32: 3D plot of ORESpv (on-site PV self-consumption) for FOC with variable
periods tariff ..............................................................................................................................202
Figure 6.33: 3D plot of HRESpv (heat demand from on-site PV) for FOC with variable
periods tariff ..............................................................................................................................203
12 |
Figure 6.34: 3D plot of LCOH (levelized cost of heat) for FOC with time-of-use tariff
.....................................................................................................................................................204
Figure 6.35: 3D plot of ORESpv (on-site PV self-consumption) for FOC with time-of-
use tariff .....................................................................................................................................205
Figure 6.36: 3D plot of HRESpv (heat demand from on-site PV) for FOC with time-of-
use tariff .....................................................................................................................................205
Figure 6.37: 3D plot of LCOH (levelized cost of heat) for FOC with wind tariff .........206
Figure 6.38: 3D plot of ORESpv (on-site PV self-consumption) for FOC with wind tariff
.....................................................................................................................................................207
Figure 6.39: 3D plot of HRESpv+windtariff (heat demand from on-site PV and electrical
imports during high wind periods) for FOC with wind tariff ...........................................208
Figure 6.40: 3D plot of LCOH (levelized cost of heat) for MPC with flat rate tariff ....209
Figure 6.41: 3D plot of ORESpv (on-site PV self-consumption) for MPC with flat rate
tariff ............................................................................................................................................210
Figure 6.42: 3D plot of HRESpv (heat demand from on-site PV) for MPC with flat rate
tariff ............................................................................................................................................210
Figure 6.43: 3D plot of LCOH (levelized cost of heat) for MPC with variable periods
tariff ............................................................................................................................................211
Figure 6.44: 3D plot of ORESpv (on-site PV self-consumption) for MPC with variable
periods tariff ..............................................................................................................................212
Figure 6.45: 3D plot of HRESpv (heat demand from on-site PV) for MPC with variable
periods tariff ..............................................................................................................................213
Figure 6.46: 3D plot of LCOH (levelized cost of heat) for MPC with time-of-use tariff
.....................................................................................................................................................214
Figure 6.47: 3D plot of ORESpv (on-site PV self-consumption) for MPC with time-of-
use tariff .....................................................................................................................................215
Figure 6.48: 3D plot of HRESpv (heat demand from on-site PV) for MPC with time-of-
use tariff .....................................................................................................................................216
Figure 6.49: 3D plot of LCOH (levelized cost of heat) for MPC with wind tariff ........217
Figure 6.50: 3D plot of ORESpv (on-site PV self-consumption) for MPC with wind tariff
.....................................................................................................................................................218
13 |
Figure 6.51: 3D plot of HRESpv+windtariff (heat demand from on-site PV and electrical
imports during high wind periods) for MPC with wind tariff ...........................................219
14 |
List of Tables
Table 3.1: Screening process for 51 tools using a set of criteria ......................................... 39
Table 3.2: Input data support capabilities ............................................................................... 44
Table 3.3: Electrical and thermal supply technologies and district heating ....................... 46
Table 3.4: Design optimisation, outputs, controls and DSM controls capabilities .......... 48
Table 3.5: Storage modelling capabilities and underlying models ....................................... 54
Table 3.6: Electrical and thermal storage technologies and advanced models (beyond
SSM) ............................................................................................................................................. 55
Table 3.7: Practical considerations ........................................................................................... 61
Table 3.8: Output from application of tool selection process ............................................. 68
Table 4.1: Screening requirements applied to PyLESA ........................................................ 75
Table 4.2: Categorisation of PyLESA tool capabilities ......................................................... 76
Table 5.1: Thermal conductivity of insulation materials ....................................................116
Table 5.2: Empirical approximations of heat loss due to tank insulation openings .......116
Table 5.3: Statistical analysis of example charging period ..................................................123
Table 5.4: Statistical analysis of example discharging period .............................................125
Table 5.5: Statistical analysis of example 2-day period .......................................................127
Table 5.6: Existing electricity tariff descriptors and examples ..........................................131
Table 5.7: Rules for fixed order control strategy, split into condition based on an import
setpoint ......................................................................................................................................135
Table 5.8: Set of key performance indicators which are outputs of the modelling
methodology .............................................................................................................................141
Table 6.1: Set of KPIs for sizing study .................................................................................151
Table 6.2: Outputs from application of PyLESA and EnergyPLAN ...............................174
Table 6.3: Optimum LCOH results for the existing electricity tariffs and control strategies
including KPIs (brackets is the relative change from FOC to MPC) ...............................196
Table 6.4: : Optimum LCOH results for the wind electricity tariff and control strategies
including KPIs (brackets is the relative change from FOC to MPC) ...............................196
15 |
Acknowledgements
Foremost, I’d like to give a huge thank you to my supervisor, Dr Paul Tuohy, for the
support and guidance over the years and for the helpful editing of various draft papers
and manuscripts. His positive and ambitious attitude throughout this PhD has encouraged
and challenged me to constantly improve and develop my work. I’d also like to thank the
rest of the ESRU department for software help and guidance.
I’d like to thank all my friends who have put up with me and cheered me up. Thanks
to my family for their unconditional support of all kinds and a home to occasionally
escape to, and thanks to my pets Ollie and Chips for their constant distractions. Finally,
I’d like to thank my partner, Rameeza, who makes my world a happier place and kept me
going throughout.
16 |
Useful Abbreviations
ASHP – Air Source Heat Pump

CAES – Compressed Air Energy Storage
CHP – Combined Heat and Power
CIBSE – Chartered Institution of Building Services Engineers
COP – Coefficient of Performance
DSM – Demand Side Management
EMPC – Economic Model Predictive Control
ES – Electrical Storage
FOC – Fixed Order Control
GSHP – Ground Source Heat Pump
HP – Heat Pump
KPI – Key Performance Indicator
LCOE – Levelized Cost Of Energy
LCOH – Levelized Cost Of Heat
LZCT – Low or Zero Carbon Technology
MPC – Model Predictive Control
NPC – Net Present Cost
OO – Operational Optimisation
PCM – Phase Change Material
PID – Proportional-Integral-Derivative
PLC – Programmable Logic Controller
PV – Photovoltaics
PyLESA – Python for Local Energy Systems Analysis
RES – Renewable Energy Sources
SSM – Simple Storage Model
WSHP – Water Source Heat Pump
WWHC – West Whitlawburn Housing Cooperative
17 |
Source code for PyLESA
PyLESA stands for Python for Local Energy Systems Analysis and is pronounced "pai-
lee-suh". It is an open source tool capable of modelling local energy systems containing
both electrical and thermal sector technologies modelled in hourly timesteps. It was
developed by the author as part of the work of this thesis with the aim of aiding the
planning-level design of local energy systems, and the focus is on modelling systems with
heat pumps and thermal storage alongside time-of-use electricity tariffs and model
predictive control. Additionally, it is anticipated that the tool provides a framework for
future development including electrical battery studies and participation in grid balancing
mechanisms.
The source code of PyLESA can be accessed from the following GitHub repository:
https://github.com/andrewlyden/PyLESA.
18 |
1. Introduction
This chapter introduces the general context and background underpinning the motivation
for the work carried out in this thesis. It begins with a discussion of the current transition
of the energy system in Scotland and the UK. Local energy systems are clearly defined to
help contextualise the scale and scope of this work. Heat pumps, thermal storage,
electricity markets, and control strategies are then explored and discussed with a view to
motivating the vital role they can play in flexible, low-carbon, low-cost, and local energy
systems. Finally, an overview of widely used energy modelling tools, and reviews of
modelling tools, is provided to give a flavour of the current status and future challenges
of local energy system modelling.
1.1. General Context

The usage and sourcing of energy is currently undergoing a fundamental transition as
society aims to tackle climate change, improve urban environments, and progress energy
equity. Reducing energy use across all sectors including buildings, industry, and transport,
and increasing sourcing energy from sustainable sources are both required to meet these
aims. This transition mandates a paradigm shift in how energy systems at all scales are
designed. Supply of energy using renewable energy sources such as wind and solar are
low-grade and stochastic and cannot easily be matched to inflexible demand. This
contrasts with the high energy density and abundant nature of fossil fuels which when
burned in power plants provide a means of meeting demand. This problem necessitates
new methods for introducing flexibility in future energy systems to match supply and
demand.
Various storage technologies can introduce the required flexibility at different
timescales (sub-second, day, seasonal). However, there are significant future uncertainties
and differences in the efficiencies and costs of these storage technologies. Future systems
need to be designed carefully to ensure optimal solutions for different applications are
implemented.
Another consequence of a renewable energy-based system is the localisation of energy
production. While fossil fuels are used to produce electricity in large-scale and centralised
19 |
power plants, renewable energy can be produced regionally, across the country in smaller
scale installations. This has given rise to the increased importance of local energy systems
which are formed of a variety of co-located demand, supply, and storage technologies
which can be integrated within the wider national energy system.
Heat pumps can play a vital role in local energy systems as they can utilise low-carbon
power generation and sources of low-grade heat to meet space heating and hot water
demands. They can be employed in single dwellings or be scaled up to supply district
heating networks. Importantly they can be used with hot water tanks which are one of
the cheapest forms of storage technology currently available. Heat pumps and thermal
storage together have the potential capability to add flexibility to local energy systems in
a cost and carbon effective way.
A low-carbon, local energy system is complex and needs to be modelled to be
understood and properly designed. These systems need suitable control strategies to be
able to optimise the use of time-of-use electricity tariff structures and local energy
production via renewable energy sources.
1.2. Transitioning Energy System

Future energy system pathways to aid in tackling climate change in the UK have emerged
which involve interactions between a centralised power grid with large amounts of
fluctuating renewable power and decentralised local systems including generation and
storage [1–3]. Technologies already exist which can contribute to this potential pathway:
rooftop PV on domestic households with bi-directional grid flows [4], community-owned
distributed wind farms [5], micro-hydro schemes [6], smart grids [7], district heating
utilising waste heat [8] or combined heat and power (CHP) [9], hydrogen from renewable
energy sources [10], etc.
Scotland is undergoing a transition to a low-carbon energy system, underpinned by
the introduction of renewable energy sources into the electricity grid network. Renewable
sources accounted for 74.6% of gross electricity consumption in 2018 [11]. However,
electricity is only 23.5% of final energy consumption of which heat constituted 52.0% and
transport 24.4% in 2016 [11]. Efforts are necessary to decarbonise the heat sector where
non-electrical, renewable sources contributed 5.9% to heat demand in 2017 [11].
20 |
Traditionally Scotland has developed a gas grid in urban areas and used oil and
electricity to heat homes in rural areas. Little district heating has been developed with
approximately only 30,000 buildings connected to district heating as of 2018 [11]. District
heating can potentially increase the use of renewable heat. This has led to interest in
micro-district heating in rural towns and villages with low heat demand density, as well as
larger scale district heating projects such as housing schemes. Many of these rely on
expensive electrical heating due to restrictions on gas infrastructure in high-rise dwellings.
This, coupled with government incentives and grants, has made district heating an
economically favourable alternative with the added benefit of the perceived low-carbon
aspect.
There has been a significant uptake of biomass boilers in Scotland. These installations
are typically set-up to provide heat for large public buildings and are often built by local
councils to meet their sustainability targets. However, it is unclear that these systems have
been built to best practice and there is concern around sustainability and air pollution
issues related to the burning of wood for domestic heating [12]. Additionally, biomass
may have a pivotal role to play in the wider energy system in decarbonising difficult sectors
such as high-temperature industry and heavy transportation [13].
Electrification of heat could prove to be an effective method of decarbonising heat
due to the transition of the electrical grid to higher penetrations of renewable energy
sources. As this transition is occurring, electrical heat options such as heat pumps will
become increasingly low-carbon as seen from the falling grid carbon intensity factors in
Figure 1.1 [11]. From a holistic view of the wider energy system it is worthwhile exploring
design options incorporating heat pumps as the primary heat source.
21 |
Figure 1.1: Average carbon intensity factor for electricity consumed in Scotland [11]
1.3. Local Energy Systems

Local energy systems consist of energy production units and demand side management
(DSM) enabling technologies co-located with demands [14]. There are many terms used
to describe similar energy systems such as community, district, decentralised, etc. Each
have their own specific definitions and different interpretations in technical, social, and
economic contexts. The term local energy system is used here in a technical context and
defined on the co-location of technologies and the scale, ranging from a cluster of
buildings to an entire region of a country.
An example of a local energy system may contain locally owned wind turbines and/or
PV which provide electricity to local customers, and heat generation units utilising local
resources (e.g. biomass, local renewable generation) serving district heating networks.
The implementation of these has been motivated by reducing imported energy costs,
increasing self-reliance, reducing carbon footprints, and providing revenue streams via
government subsidies and exported electricity. The development of such systems in the
UK marks a shift from the historical means of delivering electricity from centralised fossil
fuel power production plants and providing heating from gas via large-scale piping
infrastructure.
Local energy systems also open opportunities for sector integration possibilities.
Energy sectors which have been historically considered in silo can be designed holistically.
This could be designing a heat pump with a thermal store controlled to turn on when
22 |
there is local excess generation of wind power. These require connected systems to
facilitate logic decisions based on external signals.
Local energy systems should also be designed to be integrated within the wider
national energy system. Only considering on-site renewable power production could lead
to very large capacities of expensive storage technologies being installed, but often local
energy systems are connected to a wider national grid network. This grid in the future
could heavily rely on wind power from across the country and have periods of large excess
production. A local energy system should be designed to take advantage of this.
Additionally, grid services which provide response to stress events to ensure grid stability
could in the future be aided by local energy systems.
1.4. Heat Pumps

Heat pumps are a decentralised technology which couple the electrical and thermal
sectors. They efficiently use electricity and low-grade heat sources to provide heat at useful
temperatures, commonly for small-scale household purposes of hot water and space
heating although district heating and industrial applications do exist [15]. Previous studies
have identified large-scale heat pumps for district heating as providing 25-30% of heat in
future roadmaps for Europe [16], and concluded the technology is mature for deployment
[17].
CHP provide competition to heat pumps as a low-carbon form of district heating
supply, however it is possible a combination of both could be employed to optimise
energy balancing in larger schemes [18]. CHP can increase overall primary energy
efficiency in comparison to traditional power generators by making use of waste heat.
They are usually located close to heat demands and are categorised as a decentralised form
of electricity and heat production. Natural gas is currently the most commonly used fuel
in CHP plants. Biomass is potentially a cleaner fuel [19], however due to limited
availability and sustainability concerns, it is likely to be more useful in other sectors [13].
Synthetic fuels, such as hydrogen, produced from renewable power could be used and
these have been identified as important for 100% renewable energy systems [20].
Viability of each of the competing technologies is dependent on local conditions and
grid dynamics such as power prices and carbon intensities. A local system may have
existing renewable generation or heat infrastructure, e.g. gas grid, available to utilise which
23 |
influences technology choices. When the grid power price and carbon intensity is high,
CHP electricity production is favourable, however, when these are low and lots of
renewable power production is on the grid (or there is low demand), electrical
consumption with heat pumps can provide low-cost and low-carbon heat. This simple
example illustrates the importance of using control strategies which account for grid
dynamics and local power generation in design. As power grids become increasingly low-
carbon it is possible heat pumps could become the dominant source of low-carbon heat
for district heating networks.
1.5. Thermal and Electrical Storage

A flexible energy system consists of generation, demand, and storage components which
can respond to stochastic weather conditions and the dynamic carbon intensity and
pricing of a renewable-dependent electrical grid. The control of shifting demand and
production is commonly known as Demand Side Management (DSM) and is enabled by
storage technologies such as electrochemical batteries, hot water tanks, hydrogen
electrolysers, etc., and non-technology methods such as encouraging behavioural change.
Flexibility technologies have been analysed and compared in previous studies in the
electrical network [21] and across sectors [22].
Thermal storage can provide flexibility to an energy system when paired with heat
pumps by decoupling the heat demand and electrical consumption. Applications of
thermal storage include hot water tanks and heat pumps in domestic buildings with smart
control [23], phase change materials (PCM) for solar water heating [24], thermal mass of
residential buildings [25], etc. Hot water tanks can be used in district heating and have
been identified as an important component of 4th generation district heating systems [26].
The economic savings due to scale for hot water tanks at the district-level provides
motivation for encouraging their usage [27].
There are a number of potential applications of electrical storage technologies and
these range from large scale generation and transmission network systems, to smaller-
scale generation and distribution network systems, to individual consumers and end-users
[28]. These applications have been categorised in literature [29]: (i) Generation – arbitrage,
power reserve, area control, frequency regulation, and black-start; (ii) Transmission and
distribution – system stability, voltage regulation, and asset deferral; (iii) Energy service –
24 |
Energy management, power quality, and power reliability; (iv) Renewables – Transmission
curtailment, load-shifting, grid frequency support, and fluctuation suppression.
1.6. Electricity Markets

Electricity markets are changing to reflect the transition from dispatchable power
generation to stochastic, renewable power generation. Traditional tariffs such as flat rate
or day/night periods are being challenged by emerging half hourly time-of-use tariffs
issued a day ahead. These incentivise users with reduced prices during periods of surplus
zero marginal cost renewable generation, and correspondingly dis-incentivise with
increased prices during periods of peak demand and low renewable generation. This
reflects pricing already being seen in wholesale markets with negative pricing in high wind
and low demand periods [30]. The time-of-use pricing is currently largely driven by
electricity demand profiles which only require simple time-based controls to gain most of
the benefits of flexibility. However, as the proportion of renewable generation increases
the tariffs will become more variable and motivate deploying improved communications
and control technologies.
Future commercial arrangements through aggregators, and others, could further
reward flexibility which can contribute to local network and wider grid electricity services
such as frequency response and other longer-term balancing requirements. Engagement
in these services will also require communication and control solutions.
1.7. Control Strategies

Communication and control are essential in enabling flexibility to deliver value in future
energy systems. Secure communications, monitoring and control software and hardware
platforms are being standardised, developed and deployed at commercial [31,32] and
community cooperative scales [33–36]. Communications services available to these
platforms include weather forecasts and next 24-hour time-of-use tariffs.
These platforms allow controls to be developed that optimise the operation of the
system to meet the customer needs while maximising financial parameters or meeting
other objectives such as maximising local or global renewable (self) consumption etc.
Predictive controls and non-predictive controls have been studied with extensive
literature on control methods with building integrated thermal energy storage [37,38]. A
25 |
rule-based controller has been compared to an MPC control finding that MPC control is
vital when sizing thermal storage [39].
Classical local-loop control is typically PID (Proportional-Integral-Derivative) and
PLC (Programmable Logic Controller). These use predetermined setpoints and
instructions instead of predictions of load and supply which allow improved optimisation.
A PID control for drying temperature has been designed for a heat pump for drying [40],
and a PLC control has been designed to provide commercial demand side management
to shed load during peak hours [41]. PID and PLC control offer the local control which
implements the control decisions from supervisory controllers.
Two classifications of supervisory control have been identified in literature [38]: hard
and soft control. Soft controls have the capability to learn complex relationships between
system input and output variables, e.g. neural network control, fuzzy logic control, and
reinforcement learning control. Neural network control has been applied to optimise
control of batteries in a household [42] and fuzzy logic control used to optimise the
performance of a PV, solar thermal and heat pump hybrid system [43]. Reinforcement
learning control uses machine learning to determine how to take actions in order to
maximize a long-term objective over a sequence of decisions. It has been applied to a
demand response aggregator of electrical water heaters using a 40-45 day learning period
[44].
Hard control uses physical models to determine control signals which optimise a
system performance parameter. Model predictive control (MPC) captures the dynamics
of an energy system, and this can be based upon building and system simulation models
or artificial intelligent techniques. An MPC controller consists of several key components:
• Objective function which an optimiser minimises/maximises.
• Prediction horizon which is the period over which the optimisation is performed.
• Decision timestep which is the interval between solving optimisation problem.
• Manipulated variables can be varied by the controller.
• Optimisation solver which is chosen based upon optimisation type and required
speed.
• Feedback signal which provides updated system variables for next optimisation
step.
26 |
Adaptive control is another hard control approach which accounts for the changing
dynamics of a system and requires less accuracy of the physical system model. It consists
of three main components: the optimal control models which contain the objective
functions and constraints; the identification of the time-varying control parameters which
are adjusted; and optimal algorithms which make decisions and determine the values for
the control variables. Downsides include overcomplication and large computational time
requirements meaning that this control has been rarely applied. However, it has been
applied to microgrid operation to evaluate flexibility benefits [45].
Combinations of control methods have also been applied with a typical approach
using two stages. Firstly, a preliminary learning stage where a simple dynamic model is
developed in order to train the controller. Secondly, a refined learning phase where the
controller is applied to an actual system and continues to learn and further improves its
performance via feedback from actual performance.
A key challenge is how to capture these controls together with appropriate system
characteristics in early stage modelling to appropriately inform design.
1.8. Modelling Local Energy Systems

Given the importance of local energy systems and wide variation in possible supply,
storage control options, and variation in contexts such as climates and user expectations,
there have been many efforts to provide modelling support for the planning process from
a range of different perspectives. A general method for community energy planning is
described in [46] and a key element identified is the use of modelling tools. This thesis
focusses on the planning-level stage of design where often there is a limited set of data
available. Many planning-level energy system tools have been developed and applied to a
range of situations, and a sample of the most popular ones are discussed below.
EnergyPLAN [47] is a national and regional planning tool which has been used to
model a 100% renewable energy future for Denmark [48] and for many other studies [49].
It is applicable at community scale, and was used to model the island of Mljet in Croatia
[50] in a comparative study with H2RES, an alternative tool designed for simulating the
integration of renewables and hydrogen storage into island systems [51]. In this study, it
was shown that both tools gave very similar results; H2RES focus is technical while
EnergyPLAN supports technical and economic analyses. Both tools are deterministic and
27 |
use an hourly energy balance over a year to calculate energy generated, stored, rejected,
consumed, exported, lost, and produced in excess, as well as percentage of energy
consumed from renewable sources.
HOMER [52] is a community scale tool, originally developed to support design of
off-grid electrical energy systems but expanded to model grid connected and thermal
systems [53]. One example is modelling a hybrid solar-biomass system for a remote area
in Pakistan [54]. This study used electricity demand, available solar and biomass resource,
and costs to analyse the techno-economic viability of such a system. HOMER was used
to optimise system size using an hourly energy balance and with minimum net present
cost (NPC) as objective function.
Merit [55] is another community scale tool which has been used to model a hybrid
wind/solar system for a care home in Scotland [56]. Merit models demand, supply and
storage using an hourly energy balance and provides results showing demand/supply
match and renewable and non-renewable supply. Multiple systems were modelled, and
those shown to satisfy demand all year round analysed. The tool provides technical
analysis only with cost calculations being done outside of the tool.
TRNSYS [57] has a user-defined timestep as small as 1 second. A comprehensive
library of components is available. Systems are described in detail and the solver is
dynamic which means that TRNSYS is usually a building-level simulation tool [58]. The
number of components and parameters required for a community scale system could be
complex requiring expert level of technical systems knowledge. These complex
calculations take considerable time. It has been used to model hybrid solar PV/thermal
systems with thermal and electrical storage [59] etc. TRNSYS and similar building-level
simulation tools can be scaled up for use at community scale.
The tools described above are a sample of those available and serve to illustrate
different approaches. There is general agreement that hourly modelling timesteps (or less)
are required to adequately model such systems [60]. Tools are often first developed from
a specific perspective e.g. hydrogen for H2RES, off-grid for HOMER, building systems
for TRNSYS, and then adapted to support broader planning of community scale systems.
How to choose between the plethora of different tools, particularly for planning of
renewable energy systems where storage and DSM are to be considered is a key challenge.
A number of reviewers have previously provided an overview of modelling tool
capabilities specific to the effective integration of renewable energy. In general it was
28 |
found that the prior work, although a useful foundation for tackling the aims of this thesis,
did not fully: (i) address all storage and DSM options, (ii) provide a sufficiently detailed
categorisation of the models used to represent storage and DSM, and (iii) provide a
structured tool selection process. The most relevant of these previous works are briefly
described below.
[61] reviewed 37 tools (narrowed down from 68) regarding their suitability for the
integration of renewable energy into energy systems; the details on the storage
technologies used in the tools are high level i.e. stating whether a tool is capable of
modelling pumped hydroelectric, battery, compressed air and hydrogen storage. Thermal
storage and DSM are not included in the provided tables; ‘thermal storage’ is mentioned
for 3 of the tools in textual descriptions. The underlying models for electrical and thermal
storages are not discussed in detail; such information can be useful to inform tool
selection as some models can be more accurate than others [62,63]. The authors provide
the review to inform tool selection and the provided information is indeed useful in this
regard, but a formal selection process is not specified.
[64] considered 72 tools to find those capable at city scale of modelling multi energy
systems considering all relevant energy carriers (electricity, heating, cooling, transport
etc.). They considered in detail 13 of the tools which were open source. Information
regarding the tools was usefully tabulated including: available RES components, storage
options, economic parameters, scale, availability, objective, modelling approach, timestep,
evaluation criteria, user friendliness and training requirement. The paper identified the
different storage technologies included in the energy tools but did not give detail on the
underlying models. While it was highlighted that grid balancing is essential in districts
utilising stochastic energy sources the DSM and grid support modelling capability of the
tools was not captured. No tool selection process was specified.
[65] reviewed 20 tools chosen based on their ability to “simulate and analyse urban
energy systems”. Storage discussion was limited to seasonal thermal storage modelling,
with building level storage capability documented within the tables but not in detail, DSM
also is not covered in detail.
Several further reviews of energy system tools have been undertaken. [66] reviewed
219 studies, examining areas of urban energy systems (technology design, building design,
urban climate, systems design, policy assessment, land use and transportation modelling)
to evaluate their potential for integrated urban design. [67] reviewed 6 bottom-up tools
29 |
which focus on optimisation of community energy systems, finding DER-CAM and
MARKAL/TIMES to be the most appropriate. [68] documented the capabilities and
inputs/outputs of 11 energy tools, a short paragraph on each was provided in terms of
their energy, economic and environmental analysis capabilities. [69] undertook a review
of 12 tools to consider the methods available for integrated energy analysis for cities and
territories.
1.9. Uncertainty in Energy System Modelling

Energy system modelling tools are being used to provide decision support, and therefore
it is important to consider the types of uncertainties and the methods which exist for
incorporating uncertainty into the design process.
Typologies of uncertainty have been developed for different purposes [70], and
Mirakyan and De Guio [71] developed an overall framework of uncertainty specific to
modelling and decision support in the context of energy planning. The sources of
uncertainty which form this particular framework are described here:
• Linguistic uncertainty: has been defined “as uncertainty that arises because our
natural language is vague, ambiguous, and because the precise meaning of words can
change over time” [72]. It has an influence at all stages of modelling because the
information used in energy planning and modelling is in linguistic terms.
• Variability uncertainty: is due to inherent variability of human and natural systems

(e.g. wind and temperature changes).
• Decision uncertainty: can occur where there is ambiguity over how to define
modelling objectives.
• Planning procedural uncertainty: can occur because of available resources or time.
• Knowledge uncertainty: occurs due to the limitation of our knowledge. It may be

reduced by additional research and empirical efforts. There are different subcategories
of this uncertainty and these are described below.
o Context and framing uncertainty: occurs when setting out of the context
and boundary conditions of the modelling and is influenced by the
modellers/stakeholders perception of the scope of the planning exercise.
30 |
o Model inputs and parameters: Input data can describe external conditions
(i.e. weather data), or model system data which defines characteristics of the
system. Uncertainty of the input data and parameters may occur due to:
measurement errors, data reading, user-error, and biases in data retrieval.
o Model structure uncertainty: occurs due to incomplete understanding of
the system processes (e.g. performance of components), when approximating
a real, complex system to a mathematical model. The complexity of the model
is dependent on the context of the modelling exercise, the available data input
to the model, and the required outputs from the model (reliant on the
requirements for validation and interpretation of the results).
o Model technical uncertainties: occur due to errors/bugs in the software
and/or hardware.
o Model output uncertainty: defined as the accumulation of all the other
described uncertainties.
Methods exist for incorporating uncertainty analysis into the decision making of
energy system planning. A number of these methods have been reviewed in literature in
relation to their application in energy system planning and feasibility [73]. A sample of
these are explored below:
• Stochastic optimisation techniques: Stochastic optimisation studies introduce

uncertainty to input parameters which exhibit stochastic behaviour using a range
of different modelling approaches such as fuzzy, dynamic, and interval
mathematical programming [74]. One study used interval parameter fuzzy linear
programming and multistage stochastic programming which were integrated with
a mixed-integer linear programming framework in order to aid planning of energy
and environmental systems management under multiple uncertainties [75].
Examples of stochastic input parameters are electricity prices, fuel prices,
production costs of power plants, CO2 emission policy, energy demand, and
technological efficiency.
• Monte Carlo simulation: Monte Carlo simulation performs uncertainty analysis

by modelling possible outputs using a range of values (in the form of a probability
function) for an input parameter which has inherent uncertainty. It repeatedly
calculates results, each time using a different set of random values from the
31 |
probability functions. For example, this method has been applied to optimal sizing
of cogeneration systems under long-term uncertainty in energy demand [76].
• Scenario analysis: Uncertainty can be incorporated by evaluating various

scenarios which incorporate the dynamics and the drivers resulting in a specific
future scenario. Scenarios utilised often represent situations that are likely to occur
or extreme cases (worst-case or best-case scenarios). For example, in a study
modelling Egyptian office buildings, to capture the inherent variability in
operation and behaviour a set of ‘best and worst case parameter sets’ were used
to investigate their impacts on performance [77].
Other methods for incorporating uncertainty analysis include mean-variance portfolio
analysis [78], real-options analysis [79], and multi-criteria decision analysis (MCDA) [80].
This section provides a discussion of the types of uncertainty related to energy system
modelling and identified that knowledge uncertainty can be reduced through research and
empirical efforts. It also explored methods for incorporating uncertainty analysis into
energy system decision making. The approach taken in this thesis is to develop a
modelling framework which can be integrated into these methods.
1.10. Related Projects

The work undertaken in this thesis builds on the experiences and outputs from related
previous and ongoing projects, as well as the literature explored in this chapter.
The ORIGIN project (Orchestration of Renewable Integrated Generation In
Neighbourhoods) aimed to develop and implement load shifting mechanisms in order to
both maximise the usage of local renewable generation and minimise the import of energy
[81]. The project developed algorithms which were capable of (i) encouraging occupant
behavioural change, and (ii) remote control of systems. An ecovillage, Findhorn, was used
to trial the developed algorithms and different types of system were tested on-site. One
of these system types consisted of a close to Passivhaus standard dwelling with space
heating and hot water delivered by an air-source heat pump, with auxiliary electric boost,
coupled with solar thermal. MPC was implemented on this system type and was shown
to reduce use of the auxiliary electric boost by 44% in summer and increase the match of
the boost heat with renewable generation by 22%. This highlights the potential benefits
of employing MPC along with heat pumps, thermal storage, and local renewable
32 |
generation (i.e. solar thermal in this example). These systems require planning-level
modelling in order to capture these benefits at the early stage of design.
The author of this thesis was involved in Annex 31 of the IEA Technology
Collaboration Programs (TCPs), Energy Conservation through Energy Storage, “Energy
storage with energy efficient buildings and districts: Optimization and automation” [82].
The aim of this annex was to address challenges in integrating thermal energy storage
systems in buildings/districts from the perspective of design, development of simplified
modelling tools, and optimization techniques. The work undertaken in this annex
identified the need for simplified design modelling tools at the district-level which capture
heat pumps, thermal storage, electrical storage, and optimal control strategies.
These projects, along with the literature reviewed in this chapter, point towards the
need for design modelling tools which can incorporate heat pumps, thermal storage, local
renewable generation, and MPC.
1.11. Problem Statement

Overall, it is clear that there is a need to transition the energy system from the current
paradigm of centralisation, and reliance on fossil fuels for flexibility to using zero-carbon
technologies across all energy sectors aided by decentralisation and measures to
incentivise and enable renewable-led flexibility.
The design of future local energy systems which incorporate heat pumps, thermal
storage, future electricity markets, and predictive control strategies is complex and
requires sufficiently accurate planning-level modelling tools to help aid in understanding
the performance of these systems. This is vital in comprehending the role that properly
designed low-carbon, integrated, local energy systems will have in the transition to a 100%
renewable energy system.
33 |
2. Research Question, Aims, and
Methodology
2.1. Research Question

The research question to be addressed in this thesis is:
Can a modelling methodology and supporting modelling tool be developed
that usefully aids the planning-level design of local energy systems incorporating
heat pumps, thermal storage, local renewable electricity production, time-of-use
electricity tariffs, and predictive controls?
2.2. Research Aims

The following aims are set out to help answer the research question:
• Identify gaps in existing planning-level modelling tools.

• Develop a modelling methodology capable of aiding design of local energy
systems incorporating heat pumps, thermal storage, local renewable electricity
production, time-of-use electricity tariffs, and predictive controls.
• Develop a supporting modelling tool with the capabilities to address the identified
gaps.
o Open source to enable accessibility and build upon previous work.
o Heat pump model with explicit temperature-dependent performance.
o Hot water tank model with explicit thermal characteristics.
o Future electricity markets.
o Set of supervisory control strategies including model predictive control.
• Explore developed control strategies and existing and future electricity tariffs
through demonstration of application of the modelling methodology by
undertaking a sizing study.
o Demonstrate application of the developed modelling tool and position it
in the context of the state of the art.
34 |
o Compare the performance of predictive and non-predictive control
strategies.
o Compare the use of existing electricity tariffs.
o Investigate the use of a future wind-based renewable electricity tariff.
The scope of the work undertaken in this thesis is limited to the planning-level design
of local energy systems where it is desirable to use modelling tools with hourly or sub-
hourly timesteps, low-carbon technology models, and storage and DSM capabilities. This
type of tool would preferably provide useful outputs with minimal computational time or
input requirements. More detailed building and system design tools which require high
user expertise, are computationally heavy, and have detailed input demands, have been
considered outside of the scope of this paper.
2.3. Research Methodology

The research methodology sets out the work undertaken in order to tackle the stated
research question and aims. The thesis chapters follow the structure outlined below.
Chapter 3
1. Identification of gaps in existing tools through a review of existing modelling
tools, categorisation of capabilities, selection process to identify appropriate tools,
and application to a case study.
Chapter 4
2. Introduction of the novel modelling tool including modelling methodology,
capabilities, framework and applications with the ability to address the identified
gaps.
Chapter 5
3. Description and validation of new modelling tool capabilities: resources and
demands; electricity production technologies; heat pumps; hot water tanks;
electricity tariffs; fixed order controls; model predictive controls; and key
performance indicators.
Chapter 6
4. Carry out a sizing study for a proposed design of a residential district heating
system aided by an application of the developed modelling methodology to
35 |
explore and inform design decisions regarding control strategies and electricity
tariffs.
a. Carry out operational analysis.
b. Size heat pump and thermal storage capacities.
Chapter 7
5. Discussion and conclusions of the contributions and strengths in relation to
addressing both the identified modelling tool capability gaps and the exploration
of the control strategies and electricity tariffs; limitations and future developments
of the novel modelling tool; and potential future applications.
36 |
3. Review of Existing Modelling
Tools
This chapter reviews existing modelling tools with the aim of identifying gaps to be
addressed in the proposed modelling tool. The focus is on the tool capabilities regarding
storage technologies and demand side management (DSM), but the categorisation section
covers a wide range of capabilities and forms a thorough review of the overall ability of
the identified tools to perform according to user requirements. This work appears in a
peer-reviewed journal paper co-authored by the author [83].
The review consists of the following steps:
• Initial screening process to identify 13 applicable local energy system modelling

tools from an initial list of 51.
• Categorisation and tabulation of the modelling capabilities overall with focus on
heat pumps, storage, and DSM capabilities.
• Development of a tool selection process using the tables and demonstration for
a case study.
• Discussion of the tool selection process and modelling tool capabilities.
• Identification of the gaps to be addressed in this research.
The scope of the review has been limited, in line with the thesis scope, to tools
designed for hourly or sub-hourly timestep modelling of local energy systems containing
low-carbon technology, storage and DSM, for use at the planning stage. More detailed
building and system design tools have been considered outside of the scope of this thesis.
There is an increasing trend towards using modelling tools in conjunction with other
modelling tools or external software such as MATLAB [84], GEN-OPT [85],
EnergyTRADE [86] etc. particularly to support mathematical optimisations or realistic
controls. These multi-tool processes are also outside the focus of this review.
It is recognised that tools are continuously being developed and that the screening
analysis and the tool classification exercise will need to be refreshed periodically. The
work of this review, in addition to providing a current snapshot, provides a useful
framework for this refresh within the context of the proposed tool selection process.
37 |
3.1. Initial Screening Process
An initial list of 51 tools with some ability to model an energy system was derived from:
literature including review papers and papers describing the development and application
of tools; tool user manuals and websites; and communications with tool providers. Tools
captured in previous reviews but clearly not capable of modelling local energy systems
were discounted. For example, Envi-met is a microclimate and landscaping tool [87], and
Radiance is used in daylight prediction [88].
A set of criteria were applied to the 51 tools in order to determine in more detail their
potential suitability. A tool passed the criteria if it could be used at local energy system
scale (i.e. was defined as such or had a case study demonstrating this capability), was
appropriate to the planning stage, incorporated renewable and low-carbon technology
and storage and DSM, had hourly or sub-hourly timestep and could cover either thermal
or electrical energy supply. The screening process is captured in Table 3.1 along with
relevant references. Additionally, the dark shading indicates fail and light shading indicates
potential fail.
This process resulted in the identification of 15 tools suitable for modelling
community scale energy systems incorporating renewable energy sources, storage and
DSM, for use at planning design stages. Two of the 15, MODEST and Mesup/PlaNET
were discounted due to lack of accessible information required for more detailed analysis.
This left 13 tools to be carried forward into the categorisation of capabilities and tool
selection processs.
38 |
Table 3.1: Screening process for 51 tools using a set of criteria
Planning
Criteria Local energy Case Storage/
Tools -level LZCT Time step Electrical Thermal References
Met? system scale study DSM
design
National/
AEOLIUS No No Yes Yes Yes Minutes Yes No [61]
regional
Balmorel No No No Yes Yes Yes Hourly Yes Yes [89]
BCHP
Screening No No No Yes No Yes Hourly Yes Yes [61,90]
Tool
Biomass
decision Yes Yes - Yes Yes Yes Hourly No Yes [91]
support tool
CitySim No Yes - No Yes Yes Hourly Yes Yes [65,92,93]
COMPOSE Yes Yes - Yes Yes Yes Hourly Yes Yes [61,94]
DECC 2050
No No No Yes Yes Yes Yearly Yes Yes [95]
Calculator
DER-CAM Yes Yes - Yes Yes Yes 5 mins Yes Yes [96,97]
E4Cast No No No Yes Yes Yes Yearly Yes Yes [61]
EMPS No No No Yes Yes Yes Weekly Yes No [61,98]
National/ [47,50]
EnergyPlan Yes Yes Yes Yes Yes Hourly Yes Yes
regional
EnergyPRO Yes Yes - Yes Yes Yes Minutes Yes Yes [99,100]
ENPEP- National/
No No Yes Yes No Yearly Yes Yes [61,69,94]
BALANCE regional
39 |
ESP-r No Yes - No Yes Yes Seconds Yes Yes [58,101]

ETEM/Mark
No Yes - Yes Yes Yes Yearly Yes Yes [69,102,103]
al-lite
eTransport Yes Yes - Yes Yes Yes Hourly Yes Yes [104,105]
GTMax No No No Yes Yes Yes Hourly Yes Yes [61,106]
H2RES Yes Yes - Yes Yes Yes Hourly Yes Yes [50,107,108]
HOMER Yes Yes - Yes Yes Yes Minutes Yes Yes [52,109,110]
Hybrid2 Yes Yes - Yes Yes Yes Minutes Yes No [111,112]
HYDROGE
No Yes - No Yes Yes Minutes Yes No [61,113]
MS
IDA-ICE No No No No Yes Yes Minutes No Yes [65]
iHOGA Yes Yes - Yes Yes Yes Minutes Yes No [114–116]
IKARUS No No No Yes Yes Yes 5 years Yes Yes [61,69]
INFORSE No No No Yes Yes Yes Yearly Yes Yes [61]
National/
Invert No Yes Yes Yes No Yearly Yes Yes [61,117]
regional
KULeuven
OpenIDEAS No Yes - No Yes Yes Minutes Yes Yes [61,118,119]
framework
LEAP No No No Yes Yes Yes Yearly Yes Yes [69,120,121]
MARKAL/T
Yes Yes - Yes Yes Yes Hourly Yes Yes [122,123]
IMES
MERIT Yes Yes - Yes Yes Yes Minutes Yes Yes [124]
Mesap/PlaN
Yes Yes - Yes Yes Yes Minutes Yes Yes [61,69,125]
et
[61,69,105,12
MESSAGE No No No Yes Yes Yes 5 Years Yes Yes
40 |
6]
National/
MiniCAM No No Yes Yes Yes 15 years Yes Yes [61]
regional
MODEST Yes Yes Yes Yes Yes Yes Hourly Yes Yes [61,127,128]
NEMS No No No Yes No Yes Yearly Yes Yes [61]
Neplan No Yes - No Yes Yes Minutes Yes Yes [129]
NetSim No Yes - No Yes No Hourly No Yes [130,131]
ORCED No No No Yes Yes Yes Hourly Yes No [125,132]
PERSEUS No No No Yes Yes Yes 36-72/year Yes Yes [61]
Polysun No No No Yes Yes Yes 15 minutes Yes Yes [84,133]
PRIMES No No No Yes Yes Yes Yearly Yes Yes [130,134]
ProdRisk No Yes - Yes Yes Yes Hourly Yes No [61,135]
RAMSES No No No Yes Yes Yes Hourly Yes Yes [61,136]
[61,69,137,13
RETScreen No Yes - Yes Yes Yes Monthly Yes Yes
8]
SimREN Yes Yes - Yes Yes Yes Minutes Yes Yes [61,120,139]
National/
STREAM No No Yes Yes Yes Hourly Yes Yes [140]
regional
Termis No Yes - No Yes No Minutes No Yes [141,142]
[57,59,61,65,9
TRNSYS No Yes - No Yes Yes Seconds Yes Yes
4,116]
UniSyD3.0 No No No Yes Yes Yes Bi-weekly Yes Yes [61]
WASP No Yes - Yes Yes Yes 12/year Yes Yes [61]
WILMAR
Planning No No No Yes Yes Yes Hourly Yes Yes [61]
Tool
41 |
3.2. Categorisation of Modelling Tool
Capabilities
Tool capabilities tables were generated for the 13 modelling tools that document:
• Input data requirements and input support capabilities.

• Electrical and thermal supply technology modelling capabilities including district
heating.
• Design optimisation, outputs capabilities, controls and DSM modelling
capabilities.
• Storage modelling capabilities and underlying storage models.
• Practical considerations.
These tables are intended to be useful in the tool selection process (described later in
Section 3.3) by providing information on the capability of tools to be assessed against
requirements for a local energy system analysis task.
3.2.1. Input Data Requirements and Input Support Capabilities

Tools have different levels of input data requirements; some tools require the energy
demand profiles, local climate, system characteristics, or generation profiles to be
explicitly input as time series directly by the user. Other tools have embedded functions
and libraries that provide support in generating detailed datasets from simple inputs,
and/or support a mix of both directly entered and tool generated calculation inputs. This
functionality could be essential, desirable, or not applicable depending on availability of
data or expertise.
The key characteristics related to data input requirements for the various tools are
captured in Table 3.2 and described below.
Demand profile generator
Tools were deemed to contain a demand profile generator (‘Yes’ in Table 3.2) if
functionality exists to support synthesis of electrical, thermal or fuel demand profiles in
hourly or sub-hourly timesteps from simple inputs such as monthly or annual bill data or
descriptions of building numbers and types, demographics, etc. Others which take the
approach that either explicit half-hourly or hourly metered data needs to be obtained, or
42 |
potentially generated using a secondary modelling process (e.g. using building
performance simulation tools), were categorised as ‘No’ for this category.
Resource assessor
A resource assessor gives access to weather and other resources (e.g. solar radiation, wind,
water, biogas and biomass) in a suitable data input format (e.g. from national or
international datasets) based on simple inputs (e.g. location). The resources covered were
identified for each tool.
Supply profile generator
A supply profile generator provides electric, thermal or fuel-producing system outputs for
use in the modelling. ‘Modeller’ describes a tool which generates the supply profile from
the resource input (e.g. climate) and the device specifications. For example, in HOMER,
local wind speeds (the resource input) and a specific wind turbine specification (a power
curve and other details) are used to calculate the wind turbine supply profile. ‘Database
and input’ describes a tool where the hourly or sub-hourly supply profiles are input
directly requiring the user to do some outside tool calculations or source such datasets.
43 |
Table 3.2: Input data support capabilities
Demand profile Resource Supply profile

Tools
generator assessor generator
Biomass decision
Yes No Modeller
support tool
Database and
COMPOSE No No
input
DER-CAM No S, T, Wi Modeller
Database and
EnergyPLAN No No
input
EnergyPRO Yes B, H, S, T, Wi Modeller
eTransport Yes Yes* Modeller
H2RES No B, H, S, Wi Modeller
HOMER Yes B, H, S, T, Wi Modeller
Hybrid2 Yes S, Wi Modeller
iHOGA Yes H, S, Wi Modeller
MARKAL/TIMES No B, H, S, T, Wi Modeller
Merit Yes S, T, Wi Modeller
SimREN Yes Yes* Modeller
Resource Assessor Key: Biomass (B); Hydro (H); Solar radiation (S); Temperature (T); Wind
(Wi)
*indicates that a resource assessor exists but the specifics were unable to be determined.
44 |
3.2.2. Electrical and Thermal Supply Technology Modelling
Capabilities
Tools vary with respect to the range of supply technologies that can be directly modelled.
This section captures information about available supply technologies within the different
tools and more detailed description is given in Table 3.3.
A wide range of electrical supply systems can be modelled, most tools support
modelling of connection to the external electricity grid. Two categories have been
assigned for modelling of the grid connection: ‘Grid simple’ allows for limitless import
and export, with static pricing; more complex ‘Grid’ models include features such as
connection limits and charges, complex time-based import and export tariffs etc. In
general, the tools lack the ability to model the evolving electricity markets and tariffs.
The modelling of district heating systems, if available in the tools, is only as an
estimated heat loss. This is a continuous heat loss as a percentage of peak load in the
Biomass decision support tool, or a percentage of real-time load as in EnergyPRO. The
heat demand density, distribution temperature and other factors such as controls which
have a large effect on ancillary energy use and losses in district systems are not directly
considered and are required to be captured by the user in inputting thermal demand
profiles.
District heating is becoming more popular in the UK [143,144], and is ubiquitous in
Scandinavia and Eastern and Central Europe [145]. It has potential to increase energy
system overall efficiency and provide flexibility for more effective use of waste heat and
renewables using thermal storage which is much cheaper at district scale than for
individual buildings, and much cheaper than an equivalent capacity of electrical storage
[146]. It is therefore important to consider district heating while it will not necessarily be
appropriate in all circumstances.
Heat pumps are commonly modelled using simple energetic models which do not
account for temperature dependent COP. Additionally, available data for specific heat
pumps is often limited such as a COP provided under one set of conditions. This leads
to an overestimation of seasonal performance.
45 |
Table 3.3: Electrical and thermal supply technologies and district heating
District
Tools Electrical supply Thermal supply
heating
Biomass
decision support No FBo Yes
tool
CHP, EBo, FBo,
COMPOSE B, C, CHP, G, Gr, PV, Wi No
HP, ST
CHP, EBo, FBo,
DER-CAM CHP, D, G, Gr, PV, Wi No
Geo, HP, ST
B, C, CHP, D, G, Geo, Gr, GrS, CHP, EBo, FBo,
EnergyPLAN Yes
H, N, PP, PV, T, Wa, Wi Geo, HP, I, ST, Was
B, C, CHP, D, G, Gr, H, PV, CHP, EBo, FBo,
EnergyPRO Yes
Wi HP, ST
eTransport CHP, Gr, PP CHP, FBo, HP Yes
H2RES B, C, D, G, GrS, H, PV, Wa, Wi, EBo, FBo No
HOMER B, C, CHP, D, G, Gr, H, PV, Wi CHP, FBo No
Hybrid2 D, PV, Wi None No
iHOGA D, G, Gr, H, PV, Wi None No
MARKAL/TIM B, C, CHP, D, G, Geo, Gr, GrS, CHP, EBo, FBo,
No
ES H, N, PP, PV, T, Wa, Wi Geo, HP, I, ST, Was
Merit C, CHP, G, GrS, PV, Wi, CHP, HP, ST No
SimREN Geo, H, PP PV, Wi CHP No
Key:
Electrical: Biomass power plant (B); Coal power plant (C); Combined heat and power plant (CHP); Diesel
plant (D); Gas plant (G); Geothermal plant (Geo); Grid (Gr); Grid simple (GrS); Hydro (H); Nuclear (N);
Generic power plant (PP), Photovoltaic (PV); Tidal (T); Wave (Wa); Wind (Wi)
Thermal: Combined heat and power (CHP); Electric boiler (EBo); Fuel boiler (FBo); Geothermal (Geo);
Heat pump (HP); Industrial surplus (I); Solar thermal (ST); Waste incineration (Was)
46 |
3.2.3. Design Optimisation and Output Capabilities
Two attributes important in supporting design tasks are: the capability of the tool to aid
the identification of optimum design solutions, and the ability of the tool to directly
provide outputs required to support decision making. Key capabilities of the 13 tools in
these areas are captured in the first two columns of Table 3.4 and further discussed below.
Design optimisation
Optimisation tools find the minima, or maxima, for a defined objective function by
systematically searching a defined modelling space according to a mathematical algorithm.
Design optimisation involves a search for the optimal system w.r.t. combination and
sizing of components. Most of the reviewed tools, where they support optimisation, use
a full factorial deterministic approach based on user defined inputs to solve the
optimisation problem and use a simple financial and/or carbon emissions objective.
HOMER historically has executed a grid search based on user defined inputs specifying
the system options to be included but recently provided an update allowing users to only
input upper and lower limits to the grid search. iHOGA was the only identified tool with
multi-objective function capability, it includes a choice of available objective functions
and embedded genetic algorithms [147]. The Biomass decision support tool supports the
optimisation of thermal storage size. A number of reviews have covered the mathematical
optimisation methods that could potentially be employed [148,149]. Tools which do not
directly support mathematical optimisation could be used within an external mathematical
optimisation process by an iterative approach, but this can be logistically complex or
require advanced software skills to automate.
Outputs
The outputs are key in assessing system performance. Different tools focus on different
aspects of the system performance; most tools provide financial analysis such as
cost/kWh of energy produced or information on energy market interactions, some are
purely technical and focus on the energy production, system analysis, demand/supply
match, or fuel consumption, others assess emission and renewable penetration, and others
consider social factors such as job creation and the human development index. Specific
tool outputs can be used in external calculations to generate a wider range of analysis
outputs but only the in-tool capabilities are documented here.
47 |
Table 3.4: Design optimisation, outputs, controls and DSM controls capabilities
Tools Design Outputs Controls DSM

optimisation control
Biomass Decision S E, EP, FA, FC, FO, NO FO
Support Tool RP, SA
COMPOSE E, F E, EP, FA, FC, MO, OO (F) OO (F)

SA
DER-CAM E, F A, E, EP, FA, DC, EV, LS, DC, EV,
FC, SA MO, OO (F, LS, OO
E) (F, E)
EnergyPLAN No E, EP, FA, FC, FO, LS, MO, FO, LS,
SA, RP OO (F) OO (F)
EnergyPRO No E, EMI, EP, FA, EV, MO, NO, EV, OO
FC, SA OO (F), UO (F)
eTransport F E, EMI, EP, FA, MO, OO (F) OO (F)
FC, SA
H2RES No EP, FC, RP, SA FO, MO FO
HOMER F A, E, EP, FA, AC, LS, MO, LS, OO
FC, RP, SA NO, OO (F), (F)
UO
Hybrid2 No EP, FA, SA FO, LS, MO, FO, LS
NO
iHOGA F, A, E, F, A, E, EP, FA, FO, MO, NO, FO, OO
HDI, JC, NPC FC, HDI, JC, OO (F) (F)
RP, SA
MARKAL/TIMES F E, EMI, EP, FA, MO, NO, OO OO (F)
FC, RP, SA, (F)
Merit No EP, FC, M, SA FO, LS, MO FO, LS
SimREN No EMI, EP, SA - -
48 |
Key for Table 3.4:
Design Optimisation: Autonomy (A); Emissions (E); Financial (F); Human development index (HDI); Job creation
(JC); System (S)
Outputs: Autonomy (A); Emissions (E); Energy market interaction (EMI); Energy production (EP); Financial analysis
(FA); Fuel consumption (FC); Human development index (HDI); Job creation (JC); Demands/supply match (M);
Renewable penetration (RP); System analysis (SA)
Controls/DSM Controls: Advanced control (AC); Demand curtailment (DC); Electric vehicles (EV); Fixed order
(FO); Load shifting (LS); Modulating output (MO); Non-modulating output (NO); Operational optimisation (OO)
with objective function in brackets; User-defined order (UO)
3.2.4. Control Modelling Capabilities including DSM

The ability to correctly capture controls is important in assessing the performance of local
energy systems. Particularly when assessing the impacts of storage and DSM. Modelling
tools often have in-built control logic intended to mimic real or idealised controls, it is
important to comprehend and assess the control regime underpinning each of the models.
Control capabilities of the 13 tools are captured previously in Table 3.4 and are further
discussed below.
3.2.4.1. General Control Capabilities
Controls regulate how supply, storage and DSM technologies meet loads by determining
the control logic and constraints applied. A simple local energy system control strategy
can include: (i) an order of dispatch for the different resources, and (ii) a set of constraints.
Operational Optimisation
Operational Optimisation (OO) control is where the tool optimises, at each timestep, the
order of dispatch of supply, storage, and DSM technologies to satisfy an objective
function which may relate to cost, emissions, etc. There are differences in detailed logical
implementation between tools; a general description is given here.
Most tools use the OO control chronologically i.e. calculations are performed at each
individual timestep to establish an optimum based on prevailing conditions at that
timestep only, before the next timestep is then considered. Storage is generally charged
and discharged when it is deemed favourable to do so according to the specific logical
implementation and objective function. Typically charging will occur when there is excess
energy from renewable or non-modulating supply where storage is deemed to have
benefit over export or curtailment, or where grid parameters, e.g. tariff, make charging
from grid advantageous. Discharge from available storage is generally treated as a
dispatchable supply option. The value attached to storage charge and discharge takes
account of characteristics of the storage system, e.g. efficiencies and costs, plus parameters
49 |
such as tariffs and carbon contents. For example, in HOMER the discharge energy cost
includes average charge energy cost, efficiencies, and battery wear, lifetime and
replacement costs.
OO control is applied non-chronologically in some tools e.g. in EnergyPRO the whole
calculation period is scanned for energy supply costs and an optimised supply schedule
determined, with excess low-cost generation charging storage and discharge occurring to
meet demand in subsequent favourable high cost periods. These OO control
functionalities may replicate real control systems for situations where local renewable
consumption is prioritised or where a set tariff structure is established for energy import
and export; the non-chronological OO implementation may in some circumstances
provide an optimistic view of system performance as perfect foresight is implied.
Fixed Order Control
Fixed Order Control (FOC) is where there is an available set of functions with pre-defined
order of dispatch of supply, and fixed conditions for the use of storage and DSM
technologies. Dispatchable supply is dispatched in a fixed order in periods where non-
dispatchable, typically renewable, supply is below demand. EnergyPLAN, H2RES, and
Merit charge electrical storage in periods of excess renewable production and prioritise
discharge from electrical storage over generators and power plants. In Merit thermal
storage discharge is prioritised over other thermal supply options. In EnergyPLAN
thermal storage charging is prioritised to absorb excess electricity or heat production and
discharged to avoid non-renewable generation. In iHOGA batteries can charge/discharge
at fixed, user input tariff values. In the Biomass decision support tool excess heat from
the biomass boiler is stored in a thermal storage and discharged when demand exceeds
supply. EnergyPLAN includes several selectable functions for dealing with excess
electricity production. Hybrid2 contains embedded functionality for 13 pre-defined fixed
order controls relating to the practical performance of electric systems [150].
User-defined Order
User-defined Order (UO) control is where the order of dispatch, for at least some part of
the supply, is defined by the user. For example, UO in EnergyPRO requires all supply
options to be given an order of preference, which can also include separate priorities for
production to satisfy different (peak, high, low) loads; storage priority setting is not an
option and in this tool storage operation always follows the OO control strategy.
50 |
Modulating Output
Modulating Output (MO) control applied to a dispatchable supply allows modulation of
output to match load above some minimum supply output level. In all tools the grid
connection, if enabled, can modulate output to follow electrical load with a minimum
supply level of zero. HOMER can only designate grid or generator supplies to this control
while in EnergyPRO, DER-CAM, and eTransport any dispatchable supply can be
assigned.
Non-modulating Output
Non-modulating Output (NO) control sets the constraint that a designated supply must
run at a fixed output whenever it is running. In the Biomass decision support tool, the
designated supply is the biomass boiler. In EnergyPRO the user selects supplies. In
iHOGA and HOMER the designated supplies are the generators. In these two tools a set
state of charge can be specified, and the designated supply will continue operating,
regardless of availability of renewable generation, until the set point is reached. This
mimics a common feature in real systems used to maximise battery life but which reduces
the potential for renewable inputs to the store.
Advanced Control
HOMER offers the capability to use Advanced Control (AC) strategies where users can
define more complex control operating regimes than those previously outlined by
interfacing with externally written code in MATLAB [151].
3.2.4.2. DSM-related Control Capabilities
The general control modelling capabilities described in the previous section, such as OO
and FOC, can be used where there is storage in the system to capture DSM functionality
associated with storage charging and discharging. Several tools have further DSM specific
functionality to represent ‘Load Shifting’, ‘Demand Curtailment’ and ‘Electrical Vehicles’
in the system. All DSM related control capabilities are captured in the 'DSM control'
column of Table 3.4, further DSM specific functionalities are described below.
Load Shifting
Load Shifting (LS) is where a flexible load is defined which can be met or deferred to a
later timestep within a limited deferrable time period, while incurring no loss. The flexible
load can be input as a specific energy quantity over the deferrable period in EnergyPLAN
which uses 1 day, 1 week, or 4 weeks deferrable periods, and in Hybrid2 which allows
51 |
users to input the deferrable period. In DER-CAM the flexible load is sized as a
percentage of the main load over a 1-day deferrable period. The flexible loads in these
tools are actuated when lowest cost or surplus energy is available within the flexibility
period. HOMER and Hybrid2 can accommodate more detailed model parameters such
as: average deferrable load (kWh/day), capacity (kWh), peak load (kW), and minimum
load ratio, flexible load in these tools is treated as secondary to the main load but
prioritised over charging storage.
Demand Curtailment
Demand Curtailment (DC) is where demand can be curtailed under certain conditions,
and, unlike load shifting, is not shifted but reduced. DER-CAM is the only reviewed tool
capable of modelling DC and curtails demand when tariff prices exceed a user defined
curtailment cost (£/kWh) within an annual maximum number of curtailment hours.
There is also additional functionality to allow for up to 5 daily hourly profiles capturing
the proportions of the main load which can be curtailed at each timestep.
Electric vehicles
Electric vehicles are going to play a vital role in the future of energy systems [152,153],
and there has been research into the system flexibility they can provide [154,155]. Only
two of the identified tools include models for an electric vehicle to grid interaction.
EnergyPRO has a model based on the energetic capacity of the batteries in the cars, and
limits on the charging and discharging along with associated efficiencies. The demand for
the vehicles is input as a time series and there are options accounting for availability.
Charging/discharging can be set to on/off with charging allowed at zero demand, it can
be set to proportional to the driving demand time series, or it can be set its own time
series. EnergyPLAN contains a similar model. The inputs are for maximum
discharge/charge, capacity of batteries in vehicles, efficiencies, and a time series for
demand. Simpler assumptions are made on the availability, with the fraction of cars
driving at peak demand and, of cars parked used to calculate the connection of cars to
grid.
52 |
3.2.5. Storage Modelling Capabilities and Underlying Models
This section looks at relevant capabilities of the 13 screened tools and underlying models
with respect to storage functionality. Such functionality enables DSM and, in the reviewed
tools, is used with the operational optimisation and fixed order controls (see Section
3.2.4.1).
Storage capabilities are captured in two look-up tables for use in tool selection. Table
3.5 describes the range of storage modelling capabilities available in each tool, with more
detailed descriptions of these capabilities in the sub-sections below. Table 3.6 gives a
summary of the more advanced models i.e. more detailed models than the simple storage
model (SSM) for each storage technology; SSM can be used to model all storage types and
is not included in Table 3.6 for this reason. A brief summary of each capability and
underlying model is given below, further details including model equations can be found
in the relevant references.
53 |
Table 3.5: Storage modelling capabilities and underlying models
Thermal Fuel Fuel

Tools Electrical storage
storage synthesis storage
Biomass decision
No MB No B
support tool
COMPOSE KiBaM CS, SSM No No
DER-CAM FB, SSM MB No No
BF, BG,
EnergyPLAN CAES, PH, SSM SSM, STS G, O, M
EF, GtL, H
BF, BG,
EnergyPRO PH, SSM CS, MB G, O, M
EF, GtL, H
eTransport Yes Yes Yes Yes
H2RES Yes Yes No Yes
FB, KiBAM,
HOMER No H H
MkiBaM, PH, SSM
Hybrid2 EKiBaM No No No
KiBAM, MKiBaM,
iHOGA No H H
SSM
MARKAL/TIMES Yes Yes Yes Yes
Merit EKiBaM SSM No No
SimREN Yes No No No
Key:
Electrical: Compressed air energy storage model (CAES); Extended kinetic battery model (EKiBaM); Flow battery
model (FB); Kinetic battery model (KiBaM); Modified kinetic battery model (MKiBaM); Pumped hydro model (PH);
Simple storage model (SSM)
Thermal: Cold storage model (CS); Moving boundary model (MB); Seasonal thermal storage model (STS); Simple
storage model (SSM)
Fuel synthesis: Biofuel (BF); Biogas (BG); Electrofuel (EF); Gas to liquid (GtL); Hydrogen (H)
Fuel storage: Biomass (B); Gas (G); Hydrogen (H); Methanol (M); Oil (O)
*”Yes” indicates that the tool has a certain capability but specific models used were not able to be confirmed; these
tools were assumed to have SSM as minimum electrical and thermal storage models
54 |
Table 3.6: Electrical and thermal storage technologies and advanced models (beyond SSM)
Electrical storage Advanced ES Thermal storage Advanced TS

(ES) type models used (TS) type models used
EKiBaM, KiBaM,
Lead-acid battery Hot water tank MB
MKiBaM
EKiBaM, KiBaM,
Li-ion battery Cold storage CS
MKiBaM
Seasonal thermal
Flow battery FB STS
storage
Pumped hydro PH
CAES CAES
Key:
Electrical: Compressed air energy storage model (CAES); Extended kinetic battery model (EKiBaM); Flow battery
model (FB); Kinetic battery model (KiBaM); Modified kinetic battery model (MKiBaM); Pumped hydro model (PH);
Simple storage model (SSM)
Thermal: Cold storage model (CS); Moving boundary model (MB); Seasonal thermal storage model (STS); Simple
storage model (SSM)
Note: SSM can be used to model all storage types and is not included
55 |
3.2.5.1. Electrical Storage Modelling Capabilities and Underlying Models
Electrical storage is a general term used here to include electrochemical (li-ion, flow, lead-
acid batteries), electromagnetic (supercapacitors), and mechanical (CAES, hydro,
flywheels) forms. Electrical storage can be represented using a number of different
mathematical models, the different models used in the tools are categorised and described
below. The level of detail required at the planning stage depends on the specifics of the
system being modelled and the outputs to be derived from the modelling.
Simple Storage Model
A tool possessing a Simple Storage Model (SSM), which can interact with supply and load,
can model any storage technology. EnergyPLAN and EnergyPRO use the SSM to define
all types of storage, including all electrical storage types. iHOGA, DER-CAM and
HOMER support the use of the SSM, e.g. for high-performance batteries [136]. HOMER
also recommends its use for simple pumped hydro storage systems. The SSM consists of
a simple energy in/out balance via an energy store. Energy can enter the store below a
threshold maximum charging rate up to a maximum store capacity. There can be self-
discharge from the store e.g. a percentage or other function at each timestep. Energy can
leave the store below a threshold maximum discharging rate. For charging and discharging
there are associated efficiencies, which combine with self-discharge to give a round-trip
efficiency. Charge and discharge efficiencies are both generally fixed values. The SSM has
fixed maximum charge and discharge rates independent of the state of the system, this
approximation may be sufficient for some analyses, but may not be realistic in other cases,
more detailed models are available. Storage lifecycle analysis is included in some tools with
the SSM, e.g. in HOMER lifetime is modelled as both an energy throughput and time,
however performance degradation effects are only included in the MKiBaM model
described later.
Kinetic Battery Model
The Kinetic Battery Model (KiBaM) was first developed for modelling lead-acid batteries
in hybrid energy systems [156]. It is described as a two tank model [157], where one tank
holds the available energy to directly support charge and discharge and the other holds
the bound energy which transfers energy to and from the available tank according to a
defined exchange function representing the chemical process. The model supports
charge/discharge rates as functions of stored energy in the two tanks. The underpinning
56 |
electronic mechanisms are still simplified with voltage modelled only as a linear function
of energetic state etc. iHOGA and HOMER both possess this model and have libraries
of electrochemical batteries with parameters established from test data.
Extended kinetic battery model
Work was done to improve the KiBaM in terms of modelling voltage behaviour [158].
These models are denoted here as Extended Kinetic Battery Models (EKiBaM). Hybrid2
includes such an improved model [159], with voltage, charging and discharging
efficiencies and current as non-linear functions of the state of charge. Merit also contains
a different but similar model with improved voltage modelling [124].
Modified Kinetic Battery Model
A further Modified Kinetic Battery Model (MKiBaM) is used by HOMER and iHOGA
to give deeper insights. This includes a thermal model component whereby the resistive
properties of the battery produce heat which affects temperature, capacity and lifetime.
Secondly, it involves cycle-by-cycle degradation of the battery as a function of depth of
discharge; this is accounted for using the Rainflow counting algorithm [160], which
iHOGA also utilises to account for corrosion effects over time. iHOGA offers
customised models for lead-acid batteries [161,162] and Li-ion batteries [163–165].
Flow battery model
Flow batteries can also be modelled explicitly with models which account for the
independence between capacity and charge/discharge and other flow cell characteristics.
Flow battery specific models based on manufacturers data are included in DER-CAM
[166] and HOMER [157].
Pumped hydro model
Pumped hydro is often modelled using the SSM by factoring in the capacity and efficiency
of the pump and generator as well as the capacity of the reservoir. EnergyPLAN and
HOMER include pumped hydro as a technology using the SSM. Only EnergyPRO
includes an explicit pumped hydro model and includes inputs such as reservoir volume,
friction factors and head difference.
Compressed air energy storage model
A simple compressed air energy specific storage model (CAES) is included in
EnergyPLAN, with a focus on the economic trading possible [167].
57 |
3.2.5.2. Thermal Storage Modelling Capabilities and Underlying Models
Thermal storage allows for sensible or latent heat to be kept for meeting a demand later.
It can include hot water tanks, brick radiator stores, phase change materials (PCM), and
cold storages. It can also be designed for buildings or community/district scales. A
summary of different thermal storage models including underlying equations is given by
[63]. The tools that are the focus of this paper use only the least complex models, some
of the limitations associated with this are discussed later. The categorisation of thermal
storage models found in the tools is captured in Table 3.5 and Table 3.6, and described
below.
Simple storage model
The SSM model does not consider temperatures but only accounts for energy and was
described earlier. EnergyPLAN uses the SSM to model all thermal storage technologies.
Moving boundary model
The most common model for thermal storage in the examined tools is the moving
boundary model (MB), where the additional inputs over the SSM are top and bottom tank
temperatures. It assumes that there is no mixing between the upper hot zone and the
lower cold zone and the thermocline boundary layer is infinitesimally small. This is again
an energy balance model with inflows and outflows of energy moving the boundary layer
up and down the store and stored energy calculated based on the thermocline position.
The model does not explicitly capture temperature variation due to losses and
destratification. This model is incorporated in the Biomass decision support tool, DER-
CAM, EnergyPRO, and Merit. The model can be adjusted in EnergyPRO using a
utilisation factor which reduces the useful energy which can be used for supply. DER-
CAM allows for different high temperature and low temperature stores within the system
to allow for different heat generation devices [168]. EnergyPRO also uses the MB model
for cold storage (CS) and was the only tool identified to have electrical, heat, and cold
storage modelling capability.
Seasonal thermal storage model
A seasonal thermal storage model is included in EnergyPLAN. It is simplified and only
two inputs are required: capacity, and ‘days of optimising storage’ which allows for the
model to identify inter-seasonal variations in demand. [65] set out the state of art in
58 |
modelling seasonal thermal storage in building-scale simulation tools, but in general this
functionality is not supported in the tools analysed here apart from EnergyPLAN.
Other thermal storage models
Temperature variations, and therefore entropy considerations, are vital in real thermal
storage analysis [169]. There may appear to be enough energy in a tank to meet the energy
demand, but if the temperature does not meet the supply requirement it is not useful
energy. The MB model does not account for changes in the temperature zones; there are
no entropic considerations. A summary of modelling approaches for sensible thermal
storage tanks [63] includes the MB model and highlights the models which would be used
to include entropy, with increasing detail at the expense of computational and data input
complexities.
3.2.5.3. Modelling of Fuel Synthesis and Storage
Fuel synthesis is the production of fuels within a system creating a new energy vector
which can be used across a range of energy sectors, and acts as storage to be used later
[170]. EnergyPLAN, iHOGA and HOMER can model the synthesis of hydrogen. This is
produced using electricity with an electrolyser to form hydrogen, stored in a hydrogen
tank, and then converted to meet transport, heat, or electricity demands. All three
technical components can be modelled within the three tools. EnergyPRO contains a
simple model for the synthesis of any fuel. EnergyPLAN allows for synthesis of different
types of fuel: biofuel, biogas, hydrogen from electrolysis, electrofuel, and gasification to
liquid transport fuel. These fuels are used to form interactions between energy sectors
and ensure high-value energy is used for high-value processes.
These fuels must then be kept in storage. The Biomass decision support tool can size
biomass fuel storage, while iHOGA and HOMER can model hydrogen storage tanks.
EnergyPLAN can model gas, oil and methanol storages, and EnergyPRO can model any
fuel storage as a generic model.
59 |
3.2.6. Practical Considerations
Table 3.7 sets out the practical considerations associated with selecting a tool: cost, access,
support, whether it is academic or commercial, user-friendliness, and whether there is
existing available expertise.
Cost may be a vital factor in choosing an energy system tool and depends on the
resources available to a user. A student is likely to choose a free tool which there is
abundance of: Biomass decision support tool, COMPOSE, DER-CAM, EnergyPLAN,
iHOGA, Hybrid2, Merit and MODEST. Often tools are available at discounted prices
for students. A government agency or an engineering consultancy may have the resources
available to afford the cost for a tool such as 3,000+ EUR for EnergyPRO, 500-1,500
USD for HOMER, or 1,275-3,130 EUR to manipulate the code for MARKAL/TIMES.
Accessibility is defined in terms of availability, purchase requirement, and if the tool
was downloadable or browser based. Available support as indicated by tool websites and
verified by the authors is listed. These include user manual, available contact details,
videos, training, and an online forum. The tools are classed as academic or commercial
based on the development and ownership of the tools through either a
university/research group, or a private company, respectively.
User friendliness was judged on the provision of an intuitive model-building pathway
which was subjectively graded by the authors at a low, medium, or high level. This
required first-hand knowledge of the tools so where the tool was not available to the
authors, the grade by [94] was referenced.
Most modelling tools require a significant investment in time to develop expertise in
order to be used correctly and proficiently so there will be a strong practical driver to use
a modelling tool which has established available expertise if this exists. If there is no
established expertise available and the aim is to develop such an expertise, then this driver
will be less strong or zero.
60 |
Table 3.7: Practical considerations
Tools Cost Access Support Academic / User

Commercial friendly
Biomass Free Download User manual, Commercial High
decision videos, online
support course
tool
COMPO Free Download Videos, forum Academic Med
SE
DER- Free Browser User manual, Academic Med
CAM videos, forum
EnergyPL Free Download User manual, Academic High
AN contact,
videos,
training,
online course
EnergyP 3,000+ EUR Purchase User manual, Commercial High
RO for all contact,
modules training
eTranspo Not available Not Not available Academic High1
rt available
H2RES Not available Not NA Academic Not
available available
HOMER Free 2-week Purchase User manual, Commercial/ High
trial, 500 - contact, Academic
1500 USD videos, forum
Hybrid2 Free Download User manual, Academic Not

contact available
61 |
iHOGA Educational Purchase User manual, Academic Med
Free, 500 EU forum,
for 1 year contact
MARKA Costs 1275- Download User manual, Academic Low1
L/TIME 3130 EUR to paid support,
S manipulate forum
source code
Merit Free Download Training Academic Med
SimREN Not available Not Not available Commercial Not
available available
1From [94], 2 User to self-assess
62 |
3.2.7. Discussion
Through the categorisation and documentation of tool capabilities it is apparent that there
are many differences between tools. Some tools, such as EnergyPLAN, combine all
energy sectors based on the view that holistic consideration across sectors leads to optimal
solutions. Other tools are primarily single domain focussed, e.g. iHOGA has strong
capabilities for electrical analysis with a wide range of storage models but no thermal
capability.
Design optimisation capabilities in the tools generally optimise for financial or
technical considerations. Only iHOGA optimises for human considerations (human
development index, job creation) and two tools optimise for environmental
considerations. Much work has been done on external optimisation used in a two-step
process. This may influence the lack of embedded optimisation options in the tools,
another factor is the preference for the simplicity and transparency available in full
factorial parametric analysis.
The review identified a lack of detailed district heating modelling capability in any of
the local energy system tools, with only a heat loss parameter as input, factors such as the
heat demand density, distribution temperatures, network layouts and controls which have
a large effect on ancillary energy use and losses in district systems are not directly
addressed.
Analysis of controls modelling capabilities in the tools showed a wide range including
operational optimisation, fixed order, and user-defined orders, for dispatch of supply and
storage. Operational optimisation control is usually used with a cost based objective
function, other possible objective functions such as maximising local use of renewable
generation, minimising grid imports or minimising emissions are not generally directly
supported, with DER-CAM a notable exception. More advanced predictive controls
based on weather forecast and demand prediction are not supported, although the non-
chronological operational optimisation in EnergyPRO and the deferrable load
functionality in HOMER etc. can represent this type of control but with significant
simplifications. The option to run tools in combination with external control algorithms
in separate software packages is one way round this limitation.
The tools, with the exception of DER-CAM, focus on load shifting and use of storage
where there is grid connection to optimise value based on cost (arbitrage) while it is widely
63 |
accepted that other grid services (such as frequency stabilisation, peak reduction,
avoidance of capital investments etc.) may also be very important.
The review of storage functionality and modelling revealed frequent use of the simple
storage model. More complex models for electrochemical storage exist particularly for
use with lead-acid, li-ion and flow batteries.
Thermal storage and heat pumps are generally limited to simple energetic models
which do not directly take account of temperature variations other than in assessing
capacity. These may be suitable for initial planning design stages but have limitations. For
thermal storage, to take account of temperatures, heat transfer rates, stratification, and
phase change in thermal stores necessitates more complex models. These will be required
in the future to support realistic modelling of the hybrid systems and advanced controls
for which these parameters have critical importance.
There were few tools found to be directly capable of analysing fuel synthesis
technologies, such technology, however, is currently unlikely to be at a local energy system
scale in the short term. For this reason, tools developed for regional scale have most
capability.
The wide range of tools available and their differing capabilities makes a capability
categorisation of value to the end user of such tools, and of use to inform those looking
to expend effort or resources in modelling of such systems. The abundance of available
tools and rapidly developing field dictated that it was impossible to include them all. The
author believes the selection is however reasonably representative of the state of the art
in tools for planning-level design at local energy system scale.
A gap in this review is the exclusion of open source modelling tools written in
programming languages such as Python, GAMS, Java, Fortran, C++, Julia, etc. An
overview of these tools is available on the opendmod website [171].
An element not considered here is the validation of the modelling tools. So far in
available literature case studies are largely based on design and do not include monitored
data on completed schemes that include DSM and storage. Experience in the buildings
industry has found that performance gaps are common [172] and identified that industry
process needs to evolve to address these gaps [173]. It is critical that similar issues are
addressed to avoid performance gaps in future local energy systems.
64 |
3.3. Tool Selection Process
A stepwise tool selection process was developed in order to compare different tools in
the context of a specific modelling problem. The methodology is based on a software
selection method from Sandia National Laboratories [174] in order to aid in the
identification of an appropriate tool for a particular analysis. This analysis was for the
planning-level design of a local energy system incorporating storage and DSM.
3.3.1. Determination of Requirements

The first process step is to establish which of the modelling tool capabilities (documented
in the previous tables) are ‘essential’, ‘desirable’ or ‘not applicable’ and to assign values of
2, 1, and 0 respectively to each of these tool capabilities. This process requires that each
of the capabilities described in the column headings and associated keys of the tables are
individually considered against the requirements for the intended analysis.
For example, if we look at Table 3.2 then the three tool capabilities captured are
'demand profile generator', ‘resource assessor’, and ‘supply profile generator’. If the user
requires the tool to provide demand profiles, weather data and renewable generation
supply profiles from simple input data, such as location and demographics, then these
capabilities would be considered essential and each of these capabilities would be assigned
a value of 2. Alternatively, if the user has available data for demand, weather and
renewable generation and supply (e.g. from monitored data) then these capabilities are
not applicable so would be assigned a value of 0 and can be eliminated from further
consideration. If the user can source information and generate the demand, weather and
renewable generation input data, but with significant effort, then this capability could be
ranked as desirable and allocated a value of 1.
Similarly, if we consider Table 3.3 it may be that it is essential that there is capability
to model electrical generation with both PV and wind so each of these capabilities would
be allocated a 2, while if there is no potential for hydro then this capability would be
allocated a 0.
When this process is complete the essential and desirable capability requirements have
been established. The first 4 rows of Table 3.8 illustrate this process for a simple case
study example which will be described in more detail in Section 3.3.3.
65 |
3.3.2. Scoring of Tools Against Requirements
Once the requirements have been established then each of the tools can be scored against
them. The first consideration is whether all the essential capabilities are available, if a given
modelling tool has all the essential capabilities it can be considered further, those which
do not pass this check can be discounted. For the tools which pass, their scores for the
essential plus desirable capabilities are summed into an overall score and ranked with the
most suitable tools having the highest scores. Again, Table 3.8 illustrates this process for
the simple case study which is described in more detail in the following section.
3.3.3. Example Application of the Modelling Tool Selection Process

Findhorn is an ecovillage in the north-east of Scotland with an ambition to transition to
a local, low-carbon energy system. It consists of around 75 buildings, with a private wire
electrical network, wind and solar generation, a grid connection, micro-district heating
from biomass, and individual household heat pumps and solar thermal systems. The
community could be said to be net zero carbon but has large electricity surpluses and
shortfalls due to stochastic demands and renewable production. They have an interest in
the use of thermal and electrical storage with advanced controls as a potential route to
achieving their aims. They had also previously been monitored as a research and
demonstration site for advanced DSM [81].
The community overall objective is to increase their energy autonomy and use of local
renewable energy resources; they have some concerns over the sustainability of biomass.
To help achieve their objective they enlisted support from a University and after an initial
scoping process identified two future illustrative scenarios to be investigated: 1) increased
electrical generation plus battery storage, and 2) increased electrical generation plus heat
pumps and large hot water tanks replacing the micro-district biomass heat source. The
modelling tool selection process was then applied in order to identify suitable software to
use for the investigation.
The first step was to review the tool capability requirements: demand profile
generator, resource assessor, and supply profile modeller capabilities (Table 3.2) were all
deemed to have zero value (i.e. not applicable) since multi-year sub-hourly data was readily
available from monitoring.
Electrical supply technologies wind, grid, and solar PV were deemed to be essential
(Table 3.3). Thermal supply modelling of fuel boiler (biomass fuel in this case) and heat
66 |
pumps were deemed essential. Capability to model solar thermal and district heating in
detail were scored desirable. These were not considered essential at this stage as the
primary focus was on the electrical supply system and the available monitoring data which
includes heat delivery from existing heat production units, solar inputs, and distribution
losses.
Design optimisation capability (Table 3.4) was deemed desirable but not essential as
the view was taken that the relatively simple range of options to be investigated could be
covered through a full factorial deterministic investigation and modelling outputs analysed
outside of the tool to establish potential optima. The output of hourly data allowing either:
autonomy, emissions, or renewable penetration to be established was deemed essential.
This level of system performance parameter output would then allow the other required
outputs to be calculated outside of the tool.
For control capabilities (Table 3.4), either FOC or OO control was deemed essential
to support the required ordering of dispatch of supply and storage, in addition to the MO
control inherent in all the tools for representing the grid. DSM-specific control
functionality was not required in this example.
Storage modelling capability was deemed essential for both electrical and thermal
storage (Table 3.5 and Table 3.6). It was deemed that the simple storage model was
sufficient but that it would be desirable for more complex models to be available. Fuel
synthesis and fuel storage are not required in this simple illustrative study.
These technical requirements are captured in the top 4 rows of Table 3.8 and then
each of the tools are assessed against these requirements. Where a tool has an essential or
desirable capability then it scores 2 or 1 respectively against that capability, otherwise it
scores 0. Once all the potentially capable tools have been assessed they are ranked: (i) the
tools which do not have all the essential capabilities are deemed ‘FAIL’ to meet the
essential requirements, and only those that ‘PASS’ this test considered further, (ii) the
remaining tools are then ranked based on their cumulative score. This process is illustrated
in Table 3.8, with the result in this case that 6 tools are capable with similar scores of
either 20 or 21.
67 |
Table 3.8: Output from application of tool selection process
Essential Overall Design Controls

Outputs Supply technologies Storage
Capabilities Score optimisation and DSM
A, E, or Fuel District Solar Heat Battery Battery TS Hot water

Pass Score Yes FO or OO WT PV Grid
RES boiler Heating Thermal Pump SSM >SSM SSM tank >SSM
D=Desirable
D E E E E E E D D E E D E D
E=Essential
Value 1 2 2 2 2 2 2 1 1 2 2 1 2 1
COMPOSE Pass 21 1 2 2 2 2 2 2 0 1 2 2 1 2 0
DER-CAM Pass 21 1 2 2 2 2 2 2 0 1 2 2 0 2 1
EnergyPRO Pass 21 0 2 2 2 2 2 2 1 1 2 2 0 2 1
EnergyPLA
Pass 20 0 2 2 2 2 2 2 1 1 2 2 0 2 0
N
MERIT Pass 20 0 2 2 2 2 2 2 0 1 2 2 1 2 0
MARKAL/
Pass 20 1 2 2 2 2 2 2 0 1 2 2 0 2 0
TIMES
eTransport F 16 1 2 2 0 0 2 2 1 0 2 2 0 2 0
H2RES F 16 0 2 2 2 2 2 2 0 0 0 2 0 2 0
HOMER F 16 1 2 2 2 2 2 2 0 0 0 2 1 0 0
iHOGA F 14 1 2 2 2 2 0 2 0 0 0 2 1 0 0
Biomass
F 11 1 2 2 0 0 2 0 1 0 0 0 0 2 1
support tool
Hybrid2 F 9 0 0 2 2 2 0 0 0 0 0 2 1 0 0
68 |
SimREN F 6 0 0 0 2 2 0 0 0 0 0 2 0 0 0
3.3.4. Discussion
The categorisation and selection process presented is not limited to the tools identified
here but is intended to provide a framework which can be used in future to refresh the
capabilities categorisation or be applied to further tools. The review of required
capabilities as the first part of the selection process can also form a guide for modellers
to ensure relevant factors are considered. More detailed scoring systems in the selection
process would be possible, the pair-wise comparison suggested by [175] remains to be
investigated.
The tool selection process does not take into account the potential for multiple tools
to be used together to analyse the system under consideration, such work is recommended
for future studies. The more detailed simulation modelling tools currently used in
buildings and systems domains have potential to be developed for community scale
energy systems in future. These would allow more physical detail to be captured in
planning level design studies. Their capabilities could also be assessed using the same tool
selection process.
One observation from carrying out the review has been that the source code is
generally hidden in the tools. There is generally an inflexible software environment,
limiting the ability to adapt or evolve the tools or incorporate functionality from elsewhere
independent of the tool supplier.
The developed selection process will be used in the sizing study, Chapter 6 where the
modelling methodology is applied, to consider the position the novel tool to be developed
in this thesis sits within the state of the art of the existing identified modelling tools.
69 |
3.4. Gaps to be Addressed
The identified tools were reviewed against their ability to model local energy systems
incorporating heat pumps, thermal storage, local renewable electricity production, time-
of-use electricity tariffs, and predictive controls. The gaps identified, and to be addressed
in this thesis are:
• Ability to adapt source code and import or exploit functionality from elsewhere.
• Temperature dependence for the heat pump models.
• Detailed model with temperature characteristics for the thermal storage models.
• Ability to model the evolving electricity markets and tariffs.
• Ability to explore predictive controls.
This chapter has achieved the first stated research aim from Chapter 2: Identify gaps
in existing planning-level modelling tools.
These gaps will be addressed by developing a modelling framework and appropriate
models addressing the required functionality. This work expands on the growing energy
system modelling capabilities previously built in the open source programming language
Python. The proposed modelling framework will be elaborated in the next chapter, and
the subsequent chapter will detail the underlying models and their validation.
70 |
4. Introducing PyLESA: Modelling
Methodology, Capabilities,
Framework, and Applications
This chapter describes: (i) a modelling methodology capable of aiding planning-level

design of local energy systems and, (ii) the capabilities and framework of the novel
modelling tool PyLESA which was developed in this thesis. The methodology sets out
the steps for applying PyLESA which can model local energy systems incorporating heat
pumps, thermal storage, local renewable electricity production, future electricity tariffs,
and predictive controls. This chapter also explores potential types of design studies which
can be aided by application of the modelling methodology. The underlying models used
in PyLESA address the gaps identified in the review of existing modelling tools.
4.1. Brief Introduction to PyLESA

PyLESA stands for Python for Local Energy Systems Analysis and is pronounced
"pai-lee-suh". It is an open source tool capable of modelling local energy systems
containing both electrical and thermal sector technologies modelled in hourly timesteps.
It was developed by the author as part of the work of this thesis with the aim of aiding
the planning-level design of local energy systems. The focus is on modelling systems with
heat pumps and thermal storage alongside time-of-use electricity tariffs and model
predictive control. Additionally, it is anticipated that the tool provides a framework for
future development including electrical battery studies and participation in grid balancing
mechanisms.
The source code of PyLESA can be accessed from the following GitHub repository:
https://github.com/andrewlyden/PyLESA.
71 |
4.2. Modelling Methodology
The modelling methodology developed in this work sets out the steps for applying the
new modelling tool PyLESA to carry out a specified design/scoping/parametric/scenario
analysis and is outlined below:
1. Define and gather data on the local energy system to be modelled including
resources, demands, supply, storage, grid connection, and control strategy.
2. Optionally run the demand and resource assessment methods to generate hourly
or sub-hourly profiles depending on available data from step 1.
3. Input gathered the relevant input data from steps 1 and 2 on the local energy
system into the input Excel workbook.
4. Input the increments and ranges to be modelled within the required parametric
design into the input Excel workbook.
5. Run PyLESA for the specified analysis.
5.1. Using a terminal (e.g. PowerShell) navigate to the relevant directory, i.e.
“…/PyLESA/PyLESA”.
5.2. Enter “python run.py”.
5.3. When prompted enter the input Excel workbook filename (excluding the file
extension “.xlsx”).
6. Analyse the outputs saved in the “outputs” folder, particularly the KPIs, to inform
the specified analysis.
These steps will be explored further in Chapter 6 where the methodology will be
applied to a sizing study for a residential district heating scheme.
Steps 1, 2, and 3 relate to the required data input to PyLESA and this highlights the
importance of the quality of the data to perform useful modelling. Step 4 defines the
specific analysis to be carried out, step 5 executes the analysis, and step 6 post processes
the analysis outputs.
Since the modelling methodology centres around the use of PyLESA, the rest of this
chapter describes the capabilities and framework of this tool. Firstly, a discussion of the
motivation behind choosing Python as the programming language is given. Then, a
summary of the capabilities of the tool is provided with the contributions which have
been made towards identified gaps in existing modelling tools, highlighted in Chapter 3.
72 |
Finally, the framework of the modelling tool and potential design studies are described.
For further modelling details see Chapter 5 which provides a detailed description of the
models and assessments used in the modelling tool.
4.3. Motivation for Choosing Python

Open source was identified as an important requirement when choosing a programming
language. This is because open source software can be easily accessed, studied, and, built
upon. The Open Energy Modelling (openmod) Initiative have highlighted that openness
“… allows the community to advance the research frontier and gain the highest benefit
from energy modelling for society” [176]. For the tool developed as part of this thesis to
make a larger impact it should be open source as this allows the energy modelling
community the ability to learn from and improve the developed modelling methodology.
A gap in the review of modelling tools from Chapter 3 is the exclusion of open source
modelling tools written in programming languages such as Python, GAMS, Java, Fortran,
C++, Julia, etc. An overview of these tools is available on the opendmod website [171].
A significant proportion of these tools use Python as either the sole language, or partly
contributing to the modelling or processing parts of the software.
Python is an interpreted, high-level, general-purpose programming language with a
design philosophy which emphasises the readability of the code. Its object-oriented
approach allows for programmers to write clear, logical code which can be understood by
others in the energy modelling community. Potential advantages of using Python include
transparency, adaptability and re-usability of utilities developed across a range of
applications including mathematics, optimisation, machine learning as well as energy
systems [177–179]. This is, in addition to the rising popularity of Python across all
scientists and engineers, where it has been dubbed “…the de facto standard for
exploratory, interactive, and computation-driven scientific research” [180].
A particular advantage of using Python for energy modelling, as well as the ability to
effectively disseminate the modelling code, is building on existing models. pvlib [181] is a
Python package for modelling photovoltaic systems and windpowerlib [182] is a Python
package for modelling wind turbines. Instead of developing the code for models of these
technologies from scratch, using Python as the programming language means that it is
easy to directly use these packages.
73 |
Python energy system models exist which could have research synergy with the topics
tackled in this thesis. PyPSA (Python for Power Systems Analysis) [183] and pandapower
[184] are tools which model and optimise electrical power systems. These have been
applied to perform analysis of power systems, e.g. a study involving the sectoral
interactions of a highly-renewable European energy system [185] which utilises PyPSA.
oemof (Open Energy Modelling Framework) uses a novel approach of providing a
toolbox for the representation, analysis and modelling of renewable energy and sector
integrated systems [186], and OSeMOSYS (Open Source Energy Modelling System) is
intended for national and regional policy development and provides a framework for the
analysis of energy systems over the medium (10–15 years) and long term (50–100 years)
using linear optimization or mixed integer programming [187].
These frameworks and models could have provided the basis for a starting point for
the modelling developed in this thesis. However, this would have required a high-level of
Python competence and instead a modelling tool was built primarily from scratch to aid
the author in gaining proficiency in Python programming. Additionally, it would be an
insightful future exercise for the review of modelling tools in Chapter 3 to include those
identified through the openmod initiative as well as the developed tool.
Based on the potential to build directly on the state of the art, and pass forward to
others through open source, Python appears an apposite choice of a programming
platform for developing new energy system modelling utilities to address the
shortcomings of the existing tools.
4.4. Tool Capabilities

The capabilities of PyLESA are tabulated using the structure of the review of modelling
tools from Chapter 3. Table 4.1 shows the screening requirements applied with comments
detailing the result; the tool passes all the defined criteria. Table 4.2 shows the modelling
and assessment capabilities following the categorisation and highlights, with red bold text,
the capabilities which address the gaps in existing modelling tools. Figure 4.1 displays the
models and energy flows of PyLESA.
PyLESA has been developed with the aim to address the gaps identified in the review
of existing modelling tools and therefore to advance the state of the art.
74 |
Table 4.1: Screening requirements applied to PyLESA
Criteria Result Comment

Scale Yes Developed specifically for local energy systems
Application of tool to residential district heating
Case study Yes
scheme, see Chapter 6
Planning-level design Yes Developed specifically for planning-level design
Low and/or zero
Yes Wind turbine, PV, heat pump
carbon tech
Storage: Electrical storage, hot water tank
Storage/DSM Yes
DSM: Fixed order control, MPC
Hourly timestep chosen for easier data collection
Timestep Hour
and lower computational run time
Electrical demand, electrical RES production,
Electrical Yes
electrical storage, and grid
Heat demand, heat pumps, auxiliary heat, and hot
Thermal Yes
water tanks
75 |
Table 4.2: Categorisation of PyLESA tool capabilities
Input data requirements and input support

Demand profile generator Yes
Resource assessor Yes
Supply profile generator Modeller
Electrical and thermal supply technology modelling capabilities
Electrical supply Grid, PV, Wind turbine
Thermal supply Auxiliary electric heat, Fuel boiler, Heat pump
District heating Yes
Design optimisation and output capabilities
Design optimisation Parametric analysis
Outputs EMI, EP, FA, FC, M, RP, SA
Controls/DSM controls AC, OO, FO, MPC
Storage modelling capabilities and underlying models
Electrical storage Simple storage model
Thermal storage Multi-node model
Fuel synthesis No
Fuel storage No
Practical considerations
Cost Free (open source)
Access Download (open source)
Support Author and ESRU
Academic/commercial Academic
User friendly Medium. Chosen because while Python is very popular, and the
code is commented and structured, many existing tools do not
require any programming proficiency
Available expertise Yes/No
Red bold text indicates the capabilities which address the gaps in existing modelling tools
Key – Outputs: Autonomy (A); Emissions (E); Energy market interaction (EMI); Energy production (EP); Financial
analysis (FA); Fuel consumption (FC); Human development index (HDI); Job creation (JC); Demands/supply match
(M); Renewable penetration (RP); System analysis (SA)
Controls/DSM Controls: Advanced control (AC); Demand curtailment (DC); Electric vehicles (EV); Fixed order
(FO); Load shifting (LS); Modulating output (MO); Model predictive control (MPC); Non-modulating output (NO);
Operational optimisation (OO) with objective function in brackets; User-defined order (UO)
76 |
Figure 4.1: Models and energy flows of PyLESA
77 |
4.5. Tool Framework
The framework of PyLESA is described in this section which explores the interaction
of the different models and assessment methods employed to model a local energy
system. In brief, an Excel workbook is used for inputting data; the modelling is performed
in hourly timesteps and models for the technologies and assessment methods are written
in Python scripts; and outputs are in the form of an array of numerical outputs and
graphical 3D plots of Key Performance Indicators (KPIs). Figure 4.2 is a graphical
representation of the framework of the modelling tool and is described and shown below.
The modelling tool follows an object orientated class structure throughout and
typically technology models and assessment methods are calculated by initialising the
appropriate class object and executing a method. Models are run using hourly timesteps
with the input data of first timestep and number of timesteps used to calculate the
modelling period.
Data is input via an Excel workbook where a cover sheet, Figure 4.1 from earlier,
displays the models and energy flows of PyLESA. Through this sheet the worksheets
relating to the inputs for each of the models and assessment methods can be accessed and
the necessary data input. Then a Python script is run which reads, processes and locally
saves the input data from the Excel workbook. Demand assessment scripts can be run to
synthesise heat and electricity demands which can inform the data input to the Excel input
workbook. Additionally, parametric analysis inputs are also input into the Excel
workbook.
The fixed_order.py or mpc.py Python scripts contain control classes and methods which
determine how the assessment and technology models are run. The first step for both is
to run the renewable generation models and obtain renewable electricity production over
the modelling period. Each timestep in the modelling period is taken sequentially and
algorithms applied to determine the order of dispatch and magnitude of supply of the
various input supply technologies in order to meet supply (further details in Chapter 5).
Class objects for each of the technologies and the assessment methods feed into the
control class to set constraints and calculate potential actions.
78 |
Using the run.py script automates the process of running all of the iterations of the
model defined in the parametric analysis step.
Results over the modelling period relating to a wide range of data are appended to a
dictionary object and manipulated using the outputs.py script. The resulting output from
running this script are a CSV file containing an array of numerical outputs and .png
graphical 3D plots of KPIs.
Figure 4.2: Framework of PyLESA showing workflow of a run
79 |
4.6. Tool Applications: Potential Design Studies
Three types of design studies which can benefit from applying PyLESA are discussed in
this section: a feasibility study, an operation study, and a sizing study. General information
regarding the type of design study is given, before an example study is described to
illustrate the benefits of applying PyLESA.
4.6.1. Feasibility Study

A feasibility study is an evaluation of the practicality of a proposed plan. For a local energy
system this could aim to investigate the practicalities of the installation of a variety of
sustainable energy options [188]. Several different steps can be taken to achieve this aim.
An example feasibility study template is outlined to illustrate the role of PyLESA:
1. Review a full range of energy supply and efficiency options and narrow down
options based on their applicability to the local energy system.
2. Technical appraisal of the potential performance of the narrowed down options.
3. Economic assessment of the costs and benefits relating to the capital and
operational costs as well as sources of funding and external financial benefits.
4. Assessment of the environmental and social impacts of options.
The first step is highly dependent on the specifics of the local energy system, including
the geographical location, the weather resources, and availability of skills. An assessment
of the local energy system, without the need for modelling, would allow the designer to
rule out a number of options. This results in a narrowed down list of options which have
practical potential.
The second and third step require sufficiently detailed modelling such that the
potential benefits of the narrowed down list of options are captured. Modelling the system
using PyLESA results in technical and economic output KPIs which can be used to
inform the technical appraisal and economic assessments required. In its current form
PyLESA is limited in the technologies it can model. Therefore, where the narrowed down
list contains technologies beyond PyLESA’s capabilities, either alternative modelling tools
can be used, or new methods can be developed to further the capabilities of PyLESA.
The fourth step involves an assessment of the environmental impacts, some of which
can be calculated from the technical outputs from modelling along with additional
environmental information of the technologies, i.e. emissions from CO2, NOx and SOx,
80 |
etc. Expert knowledge of the geographical area and population are required to perform a
full environmental impact assessment (EIA) [189]. Assessment of the social impacts are
out of the scope of this thesis.
In conclusion, application of the PyLESA to a local energy system can inform the
technical and economic aspects of the example feasibility study provided, which aims to
investigate the practicalities of the installation of a variety of sustainable energy options.
4.6.2. Operation Study

An operation study examines the operation of an existing local energy system and aims
to investigate avenues for improving performance. Often the controls on existing heating
systems with low-carbon technologies are poorly designed or installed and analysis is
required to identify opportunities for improvement.
An example operation study could be to analyse an underperforming existing heat
pump and hot water tank heating system. The study could use the following steps:
1. Monitor the major components of the existing heating system, e.g. heat output
and electrical usage of heat pump, hot water tank temperatures at different levels,
demand flow rates and temperatures, etc.
2. Model the heating system according to design to determine expected
performance.
3. Compare to identify differences between design and actual performance, and
subsequently the components which are causing the performance gap.
4. Model the heating system with alternative control strategies to identify potential
improvements to the control design.
The first step requires an energy monitoring kit capable of collecting all the relevant
data, see [36] for an example open source monitoring kit. Description of this step is
outside the scope of the work in this thesis.
The modelling required in the second and fourth steps can be performed using
PyLESA as long as the control to be modelled is the fixed order or model predictive
control strategies. However, given the flexible and open source nature of PyLESA,
bespoke control strategies can be developed and incorporated into the tool. In future a
substantial catalogue of control strategies could be contained within PyLESA which
would mean that operational studies, such as the example outlined here, would be able to
compare several control strategies easily.
81 |
4.6.3. Sizing Study
A number of methods have been used for carrying out sizing studies for energy systems
and have been classified in literature [190] into three categories: probabilistic, analytical,
and iterative.
Probabilistic methods combine models for generation and load to create a risk model
using a variety of optimisation indicators and design constraints. A number of studies
have used this type of method to size solar, wind, and/or battery storage hybrid systems
[191–193].
Analytical methods typically use computational tools to model energy systems and
assess performance (the tools reviewed in Chapter 3 are often used in analytical methods).
Different system configurations and component sizes are modelled using the
computational tools and the outputs relating to system performance compared. Often
tools have design optimisation algorithms which use a single or multiple objective
function(s) to automate this process. HOMER [194] is an example of a computational
tool which has been designed to carry this type of sizing study.
Iterative methods use a recursive process which runs until a configuration is modelled
which meets design specifications. Genetic algorithms are an optimisation method which
are used to find global optimal solutions from a population of candidate solutions which
are evolved towards better solutions [195]. A study developed a genetic algorithm to
optimize the capacity and operation strategy for a CHP and ground source heat pump
coupled heating system [196].
The data used in methods for sizing studies is also important. Methods which use
average values such as monthly weather data and worst-case scenarios such as coldest day
over 20 years to size different energy system components can lead to oversizing because
of large variability in sub-monthly weather data and the low probability of worst-case
scenarios.
In the UK buildings industry, sizing methods rely on estimating peak demands for a
design day (i.e. coldest day) and sizing the primary unit to match this peak [91]. For hybrid
systems, typically the heat pump is sized to match base load and the auxiliary units are
oversized to provide back-up and peaking capability. When including thermal storage
another approach is to size the primary unit such that continuous output matches the
demand for the design day, with the thermal storage sized to meet peaks. This approach
82 |
may enable the thermal storage benefits of plant optimisation and enhancing service and
resilience but do not enable the other benefits: grid services, shifting electricity
consumption according to tariffs, and increasing on-site generation self-consumption.
The modelling methodology developed in this thesis uses an analytical approach
coupled with a parametric analysis step to perform sizing studies. The capacities of the
heat pump and hot water tank can be input as a range and the tool will automatically run
multiple simulations. These are then graphically output as 3D plots where KPIs of the
various size combinations can be readily compared. Extension of this method could be
integration of an optimisation algorithm such as that employed in the analytical method
used by HOMER, or an iterative approach using genetic algorithms. Additionally, a
simpler future step is to extend the parametric analysis to other technologies such as
capacity of battery storage, number of PV, and number of wind turbines.
PyLESA will be showcased in Chapter 6 by carrying out a sizing study for an existing
residential district heating scheme looking to incorporate heat pumps, thermal storage,
time-of-use electricity tariffs, and predictive controls.
4.7. Summary and Next Steps

This chapter provides the context and outlines a high-level specification of the
development and potential applications of the proposed modelling tool, PyLESA.
The next steps are to delve into the details of the underlying models which underpin
PyLESA and tackle the identified gaps in existing modelling tools. Then, in order to
further validate and demonstrate the use of PyLESA, an application of the modelling
methodology will be made. This will aid design decisions for a sizing study focussing on
the optimal sizing combinations for heat pump and hot water tank capacities; comparison
of the developed control strategies; and comparison of a range of existing and future
electricity tariffs.
83 |
5. Detailing PyLESA: Underlying
Models and Assessment
Methods
This chapter covers the input requirements, detailed description and validation of the
underlying models and assessment methods used in PyLESA (Python for Local Energy
Systems Analysis). A published peer-reviewed, conference paper by the author introduces
aspects of the underlying models [197], but this chapter provides greater detail and
contains a more complete description.
Validation consists of discussion of previous applications of adopted methods,
comparison to real data, or inspection and explanation of the outputs. Validation of the
control strategies is undertaken as part of the application in Chapter 6, as a well-defined
specific local energy system is required to discuss the operation of controls in detail. The
following models and assessment methods are described:
• Resource and demand assessment

• Electricity production technologies
• Heat pump and auxiliary heat units
• Electrical storage
• Hot water tank
• Electricity tariffs
• Fixed order control
• Model predictive control
• Modelling outputs including KPIs
Contributions to the state of the art are made in the development of the following
models to address previously identified gaps:
Section 4.3
• Open source development to enable accessibility and build upon previous work
by writing the majority of the code in Python.
84 |
Section 5.3
• Heat pump model which uses standard test data to generate performance maps
using multiple variable linear regression analysis with explicit temperature
dependence.
Section 5.5
• Hot water tank modelled using a multi-node approach to represent thermal

characteristics through state of charge dependence on node temperatures.
Section 5.6
• Future wind-based electricity tariff generator.

Section 5.8
• Model predictive control strategy to optimise system operation.

All of Sections 4 and 5 describe the contribution below
• Novel tool for modelling local energy systems incorporating heat pumps, thermal
storage, future electricity tariffs, and model predictive controls.
5.1. Resource and Demand Assessment and

Input Methods
Methods have been developed that support the user to input the required resource and
demand information that allows an energy system analysis to be placed in the correct
context. The user may have direct access to the required input data at the correct timestep,
or the user may need support in assessing these inputs indirectly, e.g. from available data
sources or datasets.
Resources is the term used here to describe weather and climate related parameters
such as wind speed, direct and indirect solar insolation, and ground or water temperatures.
These parameters influence the output of the models. For example, wind turbine and PV
where wind speed and solar radiation data drives the power output, or a hot water tank
located outside where air temperature is a factor in calculating heat loss. Resource
assessment methods are included which provide access to available international and
national datasets.
A range of energy demands are considered, e.g. electrical power, heat, and transport
energy demands across the range of entities that make up the district to be analysed.
85 |
Demand assessment methods are included which provide a means for synthesising hourly
electrical and heat demand profiles from simple inputs, which are often all that is available
at the planning-level stage.
The PyLESA tool has been configured to be flexible and support multiple direct and
indirect data entry methods. Data can be input into PyLESA using an Excel workbook
template that has been created for this purpose as illustrated in Figure 5.1, Figure 5.2 and
Figure 5.3. Methods that can be used to indirectly establish the required resource and
demand inputs are described in more detail in the following subsections.
Figure 5.1: Electrical and heat demand input requirements
86 |
Figure 5.2: Heat demand profile generator input requirements
Figure 5.3: Weather resource input requirements
5.1.1. Local Resources

Accessing available data on local resources requires access to the necessary databases
containing hourly datasets. For weather resources, local stations and reanalysis climate
databases can be used to obtain datasets over several years. Typically, it is easier to access
a greater number of year datasets from reanalysis databases than from weather stations
(as these often require payment to access, e.g. Met Office in the UK [198]). Reanalysis
87 |
datasets allow access to data which enables energy system modellers to characterise the
variability of renewable energy resources across days and months. They have emerged as
a popular tool to the energy modelling community and have been applied to both
national/global scale [199] and small-scale energy systems [200].
The website renewables.ninja [201] provides free and easy access to hourly data from
the NASA MERRA reanalysis [202] dataset. The website offers a simple interface where
a point on Google Maps can be clicked and the weather datasets for that point can be
downloaded, see Figure 5.4 for a screenshot. MERRA has many advantages over other
reanalysis datasets. It provides hourly data (other datasets use 3 or 6-hourly intervals),
global coverage, has a high spatial resolution across Europe (1/2° latitude and 2/3°
longitude, 50 × 50 km), and is stable over long-timescales [203]. MERRA includes direct
and diffuse solar radiation, windspeed, air temperature, etc. for the past 37 years (from
1987). The MERRA dataset is widely used in energy system modelling and has advantages,
however studies have shown that it has systematic errors and needs to be corrected in
order to reduce error.
MERRA solar data was used as input to the Global Solar Energy Estimator (GSEE)
to model PV power output for plants in different European countries and this was
compared to measured data from individual and aggregated PV plants [199]. The use of
MERRA input data was found to cause an overestimation of output when simulating
individual sites. The study found that, for aggregated sites, that applying a simple linear
correction to the MERRA solar data to adjust for the overestimation improved the fit of
the simulated to measured power output. The authors concluded that it was likely that the
corrected data was suitable for the type of energy modelling studies, such as planning-
level design, where the resultant PV power output uncertainty will be similar to other
input data and model uncertainties. However, this study did not investigate applying this
correction method when modelling individual PV sites. Therefore, while it the
functionality to use MERRA input solar data after applying a simple linear correction
factor is included in PyLESA, this approach has not been validated and it is better to input
validated data.
MERRA wind data was used to model wind turbine power output across Europe and
it was found that it causes errors and suffers spatial bias when uncorrected, overestimating
wind output by 50% in northwest Europe and underestimating by 30% in the
Mediterranean [203]. The study describes a method of correcting the systematic error in
88 |
the MERRA dataset of the wind speed by using correction factors (both a multiplicative
factor and a linear offset). The study developed a set of correction factors for countries
across Europe, based on the ratio of observed to simulated wind capacity factor for each
country. The study did not investigate the suitability for this correction method for
modelling individual sites. Therefore, while correction factors can be used in PyLESA to
account for over-estimation of the MERRA wind speed data, this method is not validated
for individual wind turbines and it is recommended that validated data sources are used
when available.
PyLESA can accept any source of local resource input, and it is recommended that
any data input is validated in order to reduce uncertainties and errors. However, at the
planning-level stage there is often no validated and stable hourly data available and a
corrected MERRA reanalysis dataset can be used to carry out concept design studies with
the caveat that it introduces significant uncertainty into the output.
Figure 5.4: Screenshot of renewables.ninja website for obtaining weather resource data from the MEERA dataset
with the example of Findhorn, Scotland
5.1.2. Electrical Demand

PyLESA can support data input for non-heating electrical demand from a range of
sources. For the UK context of the applications in this thesis, it has been configured to
use electrical demand profiles on an hourly timestep such as those generated using
HOMER software [194]. HOMER contains a module which allows the user to generate
a profile based upon building types: residential, commercial, industrial and community.
89 |
Choice of peak demand month accounts for seasonal variability, and a random variability
parameter is used to include hourly and daily variability. For additional accuracy, the
resultant hourly profile over a year can be scaled to match the building mix to be modelled
e.g. by using CIBSE benchmarks [204] for different building types. HOMER has been
used to synthesise community (or local) electrical loads and validated in previous studies
by comparing modelled output to real data [205,206]. An agreement suitable for planning-
level design was found. The resultant profile using HOMER is not a function of local
weather conditions and therefore is fixed for any weather year. It could be adjusted to
include an ambient temperature dependence to improve prediction.
PyLESA can accept hourly electrical demand profiles from any source, and the
method described here can be used where the user has access to and knowledge of the
HOMER software.
5.1.3. District Heating Demand

Predicting heat demand is necessary to generate an hourly profile over a year and for use
in a predictive control using lookahead periods. A review of existing, similar methods can
be found in [207] which highlights the need for a simplified approach.
A method for generating hourly profiles for the district heating demand of a residential
scheme was developed in PyLESA using regression analysis of pre-simulated housing
standard profiles, scaling based on floor area, and applying diversity using a moving
average method. This method builds on work done in the development of the demand
assessment method used in the Biomass decision support tool [91] which generates a
design day demand profile and an annual energy estimation. The described method uses
simple inputs typically available at the planning-level design stage and is included as a
functionality of PyLESA. The Biomass decision support tool method that was adapted in
PyLESA also covers non-domestic building types, in future this functionality could be
easily incorporated in PyLESA.
The flow diagram, Figure 5.5, shows flow of the PyLESA hourly heat demand profile
generating method and the steps below describe the method in more detail:
1. Buildings to be modelled are split into archetypes based upon the following ages
and types. It is assumed that the age of the building indicates the building
regulations applied during construction.
1.1. Ages: pre-1983, 1983-2002, 2003-2007, and post 2007.
90 |
1.2. Building types: detached, semi-detached, mid-terrace, detached bungalow,
semi-detached bungalow, ground floor flat, mid floor flat, and top floor flat.
2. Detailed building simulation models provide standard profiles for each age and
type and are a function of average day temperature and hour of day (see [91] for
floor areas, U-values, ventilation rates, and controls used in the detailed
simulation). Regression models are applied to the standard profiles to predict
demand as a function of hour of day and outdoor temperature.
3. The standard profiles are linearly scaled according to floor area which is calculated
from the user input number of bedrooms and then each of the building types are
multiplied by the total number of each type.
4. Diversity is applied by smoothing the demand in each hour across multiple hours
by applying a moving average algorithm. This calculation is based on flattening of
the peak heat demand by a factor of 0.63 and is applicable for groups of dwellings
above 10. This is in accordance with diversity factors used by district heat pipe
manufacturers Isoplus [208].
5. Underground piping heat losses are calculated using industry standard pipe sizing
software [208] which take the types, lengths and diameters of the different piping
sections of the network, the design flow and return temperatures, and a design
ambient temperature to calculate day heat loss. Using average ambient
temperatures for each day of the design year, the tool is used to calculate the heat
loss every day of a year. This is uniformly distributed across each hour of that day
and added to the diversified heat demand.
6. Hot water demand is added by either (i) a constant baseload by specifying the
demand in kWh/person/day, or (ii) scaling standard profiles obtained from
measurements in the EST 107-household survey from 2008 [209]. Further hourly
specification of hot water demand has been investigated in literature [210].
This results in a method for predicting demand for any hour of the design year as a
function of outdoor temperature for use in the control strategies.
This method was applied to a 544 flat district heating scheme at West Whitlawburn in
order to compare to an existing method which can predict annual heating demand (this is
used as the case study for an application of PyLESA in Chapter 6). The scheme consists
of a mix of ground, mid and top floor flats, and delivers both space heating and domestic
hot water (assumed 2kWh/person/day). Each flat has a floor area of 70m2 and, due to
91 |
recent retrofit improvements to the building fabric, was modelled as post-2007. Annual
underground piping losses were calculated to be 837MWh. Local monitoring data of air
temperature from 2017 was also used. The developed method predicted an annual heat
demand of 4017MWh.
BREDEM (BRE Domestic Energy Model) [211] is a monthly calculation tool which
can be used to predict space heating and hot water demand of properties connected to
district heating networks. It predicts that post-2000 flats will have a combined space
heating and hot water demand of 6218kWh [212]. The piping losses calculated in the
previous method were then added as these were not included in this calculation. This
method predicts an annual demand of 4220MWh. This is 5.1% higher than the prediction
from the developed method and indicates that the developed district heating demand
method predicts reasonable values. Further work and data are required to validate the
hourly output of the method.
Monitoring data was collected for the heat demand at WWHC over the year 2017 and
the annual demand for this year was measured to be 3301MWh. This is lower than the
demand modelled in both the developed model and in BREDEM. It was not investigated
whether 2017 was a typical weather year, or if it was particularly mild or warm, which
would have an impact on the heat demand for this year. Other factors causing this include
the social demographics of the residents, many are in fuel poverty, and under-occupancy,
with the housing scheme not always operating at full occupancy. This highlights the
uncertainty in heat demand due to the variable behaviour of different occupants. Work
has previously captured this behaviour when predicting heat demands [213], and this
could be incorporated within PyLESA in future.
PyLESA can accept a heating demand profile from any source. The demand method
presented here is a useful means of generating an hourly profile based on simple inputs,
which may be all that is available at the planning stage.
92 |
Figure 5.5. Flow diagram of district heating demand prediction method with user inputs in text bold
93 |
5.2. Electrical Production Technologies
The electrical production technologies modelled in PyLESA are wind turbines and PV. It
is assumed that these are situated on-site and can directly meet the local electrical demand
via a local network which allows self-consumption. A grid connection is also included in
PyLESA which provides the option to configure limitless import and export electricity
tariffs, described in detail in Section 5.6.
The renewable electricity production technology models enable assessment of their
generation against local demands along with the opportunities for sector coupling and
load shifting through storage and controls to increase the utilisation of on-site renewable
generation. For example, heat pumps controlled to utilise renewable generation including
load shifting using thermal stores.
Both the wind turbine and PV models in PyLESA are existing Python modules which
have been independently validated [214,215]. Figure 5.6 and Figure 5.7 show the input
requirements for the wind turbine and PV models respectively.
Figure 5.6: Wind turbine input requirements
94 |
Figure 5.7: PV input requirements
5.2.1. Wind Turbine Model and Validation

Windpowerlib [182] is an existing Python library which contains functions and classes for
calculating the power output from wind turbines and is used in this methodology. A
choice between a user input wind curve and selecting from a database of power curves
from different manufacturers is provided to simplify the input requirements of the user.
Hub height and rotor diameter are the additional technical inputs. Local condition inputs
are wind speed (including measurement height) and roughness length as mandatory, and
pressure, air density, air temperature, and wind speeds at different heights as optional.
The power is output in hourly timesteps.
A first step validation of the Windpowerlib method is through examination of the
previous applications [186]. In this work a further validation was carried out using an
example of the modelling output for the power output from 5x Vestas V27 wind turbines
located at Whitelee, Scotland, for weather data from the year 2017. These wind turbines
individually have a rated power of 225kW, 27.0m diameters, and cut-in and cut-out wind
speeds of 3.0m/s and 25m/s respectively [216]. The modelled output is shown in Figure
5.8 and Figure 5.9. The modelled capacity factor is 20% compared to a reported actual
2017 capacity factor of 18% [217] which shows sufficient agreement for the Scottish
context, which in conjunction with the prior validations confirms it is suitable for
planning-level design.
95 |
Figure 5.8: Monthly wind power output for each month and output for year
Figure 5.9: Hourly wind power output for example winter and summer weeks
96 |
5.2.2. Photovoltaics (PV) Model and Validation
PV is modelled using the PVLIB Python library [181]. The model consists of a module
and an optional inverter, and the power output is dependent on the technical model and
the location. A database for the module and inverter is used to input measured
performance characteristics based on PVUSA test conditions [218]. The surface azimuth,
surface tilt, surrounding surface type, and a multiplier are also used to complete the
technical model. PV location is defined by latitude, longitude, and altitude and local
conditions by wind speed, air temperature and at least two of direct normal irradiance,
diffuse normal irradiance, and global horizontal irradiance. The power is output in hourly
timesteps.
Validation of the PVLIB method is through previous applications [219] and using an
example of the modelling output for the power output from 6x residential PV panels
located at Findhorn, Scotland for weather data from the year 2017 as shown in Figure
5.10. The modelling gives an 1660kWh annual output. Two existing modelling tools and
a field monitoring trial were used to independently calculate annual output:
• PVGIS [220]: 1540 ± 73.5kWh

• Renewables.ninja [201]: 1719kWh
• BRE field trial [221]: 1500kWh (results for this latitude suggest 800kWh/kWp)
• PVLIB: 1660kWh
The PVLIB shows a sufficiently good correspondence with these monitoring studies
and models to allow it to be used for the planning-level stage that PyLESA has been
developed to support.
Windpowerlib and PVLIB have been demonstrated to be useful Python modules and
are particularly useful because they are open source and can be easily accessed and the
source code inspected. They form the models utilised to model on-site renewable power
technologies within PyLESA.
97 |
Figure 5.10: Hourly PV output for year at Findhorn
5.3. Heat Pumps and Auxiliary Heat Units

The focus of this work, as established earlier, is to investigate the role of heat pumps in
conjunction with thermal storage and controls in converting renewable electricity into
heat energy that usefully meets heating demands. Heat pumps are often not implemented
as stand-alone heat generation but rather they are commonly integrated with direct electric
'boost heaters' and back-up heating such as gas boilers. In this work these integrated
systems have been modelled to be used to provide hot water for use in district heating.
However, the underlying models can be directly used for mini-districts, single residential
buildings, etc., with only minor modification of PyLESA. An example of this would be
modifying the code to model a mini-district energy centre delivering heat to a few flats
with separate hot water tank and heat pump operation for space heating and domestic
hot water provision (this has been done but is not reported in this thesis).
Heat pumps are commonly modelled at the planning stage using simple energetic
models which do not account for temperature dependent COP. Additionally, available
data for specific heat pumps is often limited and COP under one set of conditions is
provided. This leads to an overestimation of seasonal performance.
98 |
Here, a more detailed heat pump modelling approach is developed going beyond the
current state of the art in available modelling tools highlighted in Chapter 3. This approach
requires standard test data to generate performance maps using multiple variable linear
regression analysis with explicit temperature dependence.
The thermal production technologies modelled in PyLESA include large and smaller
scale heat pumps as primary units, back up and boost direct electric heaters, and fuel
boilers as auxiliary units.
There are several key characteristics which influence the performance of heat pumps.
There is variation in the steady state thermal output and power consumption across the
operating range of heat source and heating system water temperatures and flow rates.
Transient behaviour including during start-up, shut-down and setpoint variations are
largely influenced by the thermal inertia of the heat pump, i.e. the energy absorbed by the
different components (primarily the condenser). A study [222] concluded that for an
ASHP the transient induced reduction in COP is less than 2% for correctly designed
systems with continuous run times above 15 minutes. Fixed output heat pumps can run
the compressor in two modes: on or off. Their performance can be significantly reduced
by excessive cycling caused by oversizing, with a study suggesting a seasonal efficiency
reduction up to 25% under stated circumstances [223]. Variable speed heat pumps can
vary thermal output by controlling the speed of the compressor which is typically achieved
through an inverter-based variable speed drive or a stepwise compressor speed controller.
ASHP evaporators can grow frost which requires defrost mechanisms. Studies have
reviewed defrosting techniques and show COP can be reduced by around 15% in some
circumstances [224,225]. ASHPs therefore require additional modelling consideration
with respect to the defrost cycles which effect performance in temperature regions where
freezing conditions are possible. Water source heat pumps (WSHP) are assumed to have
a constant flow or limitless supply of ambient water, meaning that there is no degradation
of the source temperature. Ground source heat pump (GSHP) models have been
developed throughout literature [226] but an explicit ground model is not included in this
work.
Approaches for modelling heat pumps can be categorised into black box, grey box,
and white box. Black box models do not factor in the physical system but consist of look-
up tables or equation fit models. Grey box models can include aspects of the physical
system such as refrigeration cycle equations, fundamental system characteristics such as
99 |
defrost cycles, and empirical-based equation fit models. White box models use explicit
models to represent each of the major components of the heat pump thermodynamic
cycle.
Three separate grey box approaches for modelling ASHP and WSHP, with different
modelling detail input requirements, have been developed in PyLESA and can each be
used in the methodology depending on the available input data and desired level of detail
in the available outputs. The three approaches are:
• The Generic Regression Model
• The Lorentz Model
• The Standard Test Regression Model

The Generic and Standard Test regression models use performance maps based upon
empirical data while including an algorithm to incorporate defrosting for ASHPs. The
Lorentz model uses one empirical data point along with the idealised Carnot cycle
performance to calculate an estimation of heat pump performance. The Generic
Regression Model and The Standard Test Regression Model approaches are grey box
quasi-steady state as they include a reduction in performance to account for the dynamic
effects and defrosting without fully capturing the associated physical details of evaporator
design, the system defrosting cycle, and associated controls. The Lorentz Model is grey
box steady state as it does not account for dynamic effects and defrosting, rather being
based on a simple equation.
General inputs required for the heat pump modelling in PyLESA are heat pump type
(ASHP or WSHP/GSHP), modelling approach, rated thermal capacity of the heat pump,
the difference between the source in and out temperatures, operation mode (variable or
fixed speed), auxiliary heat requirement (monovalent or bivalent), and data input type
(peak performance if data does not include defrost cycling). Figure 5.11 shows the heat
pump modelling input Excel worksheet with the standard test regression model activated.
The three methods are set out in the following sections.
100 |
Figure 5.11: Heat pump modelling input sheet with standard test regression model activated
101 |
5.3.1. Generic Regression Model
This approach uses a generic regression performance map to calculate the COP as a
function of flow temperature and ambient temperature. The following regression
relations were obtained from [227] and are based upon surveys of manufacturer
datasheets and field trials. While they were obtained for household-scale heat pumps, the
assumption is made that they are also applicable to large-scale heat pumps. Ideally, a
dataset of performance data for large-scale heat pumps would be used, however, this was
not readily available, but this can be done as future work. Additionally, for ASHPs a 15%
reduction in COP is assumed below 5°C. For the COP of an ASHP, where ∆𝑇 is the
difference between flow temperature and ambient temperature:
𝐶𝑂𝑃𝐴𝑆𝐻𝑃 = 6.81 − 0.121∆𝑇 + 0.000630∆𝑇 2 𝑓𝑜𝑟 15 ≤ ∆𝑇 ≤ 60
The same paper includes a regression function for a generic ground source heat pump
(GSHP) and is included here as also representative of a water source heat pump (WSHP)
due to similar dynamics of the ambient sources:
𝐶𝑂𝑃𝑊𝑆𝐻𝑃 = 8.77 − 0.150∆𝑇 + 0.000734∆𝑇 2 𝑓𝑜𝑟 20 ≤ ∆𝑇 ≤ 60
This approach is useful when quickly appraising heat pumps without data for a specific
heat pump. Figure 5.12 shows the application of the generic regression model to a 14kW
ASHP by displaying the COP variation with flow and ambient temperature.
Figure 5.12: Generic ASHP regression model COP variation for a 14kW heat pump
102 |
Figure 5.13: Flow diagram of generic regression model
5.3.2. Lorentz Model

The Lorentz approach involves calculating a real COP based upon one set of operating
conditions and then calculating the maximum COP, the Lorentz efficiency, under the
same operating conditions. The real COP divided by the Lorentz efficiency gives the heat
pump efficiency. For each timestep the Lorentz efficiency for the conditions is calculated
and then multiplied by the heat pump efficiency to give the modelled COP. The maximum
thermal output in a timestep is given by the input maximum electrical capacity multiplied
by the modelled COP. This follows the same modelling approach as used in a standard
industry modelling tool EnergyPRO (Figure 5.15) and detailed equations can be found in
the user manual [228]. This approach is useful with limited data, i.e. COP under a single
operating condition, but likely leads to overestimation of COP in other operating
conditions.
103 |
Figure 5.14: Flow diagram of Lorentz model
Figure 5.15: Lorentz model output from EnergyPRO [218]
104 |
5.3.3. Standard Test Regression Model
The Standard Test Regression Model is based on multiple variable linear regression
analysis using measured COP and duty (maximum thermal output) at a range of test
conditions. Ambient temperature (Ta) and flow temperature (Tf) are used as the two
independent variables. Coefficients for the following 2nd degree polynomial functions for
COP and duty are calculated automatically based on the input data. Predictions are made
in each of the timesteps using these equations and the flow and ambient temperatures.
𝐶𝑂𝑃 = 𝛼0 + 𝛼1 𝑇𝑎 + 𝛼2 𝑇𝑓 + 𝛼11 𝑇𝑎2 + 𝛼22 𝑇𝑓2 + 𝛼12 𝑇𝑎 𝑇𝑓
𝑑𝑢𝑡𝑦 = 𝛽0 + 𝛽1 𝑇𝑎 + 𝛽2 𝑇𝑓 + 𝛽11 𝑇𝑎2 + 𝛽22 𝑇𝑓2 + 𝛽12 𝑇𝑎 𝑇𝑓

If the data input is at peak performance, defrost cycling and dynamic effects, such as
start up and shut down, are not included in the data. Thus, for ASHPs a 15% reduction
in COP to account for defrosting is assumed below 5°C but not included for WSHPs. If
the data input is at integrated performance, then cycling behaviour and dynamic effects
are included in the testing. This should be the case if measurements are taken under
standard test conditions according to EN14511 [229].
This method is the most detailed of the three approaches and should yield realistic
heat pump performance across a wider operational range. While the necessary data is not
always readily available, using standard test conditions means that manufacturers should
possess the necessary data and correspondence could be sought to obtain this data. Part
load effects on COP for variable speed heat pumps are neglected in all three modelling
approaches. Inclusion of this would be a useful future development.
Validation of the heat pump models is difficult because the models are generated
based upon the input data. Therefore, it is up to the user to ensure that the input data is
validated and accurately reflects the performance of the heat pump to be modelled. In
Chapter 6 the modelling methodology is applied to the case study using the Standard Test
Regression Model and the performance of the heat pump will be explored.
105 |
Figure 5.16: Standard test regression model variation of COP
Figure 5.17: Standard test regression model variation of duty (maximum output)
106 |
Figure 5.18: Flow diagram of standard test regression model
5.3.4. Auxiliary Heat Units and Schematic of Setup

Heat pumps are not generally used as the sole heating unit and are typically installed
alongside auxiliary heat units such as gas boilers and direct electric heaters. In these
systems the heat pump is a priority unit and the auxiliary units allow under sizing of the
heat pump to reduce capital costs. Additionally, they provide resilience to the heat supply
for periods of breakdown or planned maintenance.
Figure 5.19 is a schematic of the connections between the heat pump, electric heater,
and thermal store, all connecting to a district heating network.
107 |
Figure 5.19. Heat pump, thermal store, electric heater, and heat network using a 2-port connection
5.4. Electrical Storage

Storage models for battery storage and hot water tanks are employed to investigate DSM
strategies. Electrical storage models have been developed in detail in previous studies and
utilised in different tools [83]. The input requirements for the electrical storage model are
shown in Figure 5.20.
Figure 5.20: Electrical storage input requirements
In PyLESA a simple battery model is used which captures the essential technical
parameters: capacity, initial state, max charging/discharging, efficiency
charging/discharging, min/max state of charge, and self-discharge. A schematic of the
energy flows is displayed in Figure 5.21.
108 |
Given the generic nature of the model, any electrical storage technology, e.g. flow
batteries, lithium-ion batteries, lead-acid batteries, which operates on the appropriate
timestep can be modelled in a low-detail manner by using simplifying assumptions.
This simplistic model has been included as a starting point for future work on
incorporating a more advanced electrical storage model for technologies such as batteries,
pumped hydro, flywheels, etc. Therefore, this model has not been validated but given its
widespread usage in existing modelling tools, see Section 3.2.5, it is widely accepted within
current practice as a model capable for planning-level modelling and is included in
PyLESA.
Figure 5.21. Simple battery model schematic
𝑄(𝑡) + 𝜂𝑐 𝑄𝑐 (∆𝑇) − 𝑄𝑠 (∆𝑇) 𝑖𝑓 𝑄𝑐𝑚 ≥ 𝑄𝑐 𝑎𝑛𝑑 𝑄𝑑 = 0

𝑄(𝑡) − 𝜂𝑑 𝑄𝑑 (∆𝑇) − 𝑄𝑠 (∆𝑇) 𝑖𝑓 𝑄𝑑𝑚 ≥ 𝑄𝑑 𝑎𝑛𝑑 𝑄𝑐 = 0
𝑄(𝑡 + ∆𝑡) =
𝐶 𝑖𝑓 𝑄(𝑡) + 𝜂𝑐 𝑄𝑐 (∆𝑇) − 𝑄𝑠 (∆𝑇) ≥ 𝐶 𝑎𝑛𝑑 𝑄𝑑 = 0
{ 𝑀 𝑖𝑓 𝑄(𝑡) − 𝜂𝑑 𝑄𝑑 (∆𝑇) − 𝑄𝑠 (∆𝑇) ≤ 𝑀 𝑎𝑛𝑑 𝑄𝑐 = 0
The stored energy Q at the time 𝑡 + ∆𝑡 can be expressed in the above equation where
∆𝑇 is the timestep, 𝜂𝑐 is the charging efficiency, 𝜂𝑑 is the discharging efficiency, 𝑄𝑐 (∆𝑇)
is the charging energy, 𝑄𝑑 (∆𝑇) is the discharging energy, 𝑄𝑠 (∆𝑇) is the self-discharge,
𝑄𝑐𝑚 is the max charging rate, 𝑄𝑑𝑚 is the max discharging rate, C is the capacity, and M
is the minimum state of charge.
109 |
5.5. Hot Water Tank
Hot water tanks provide a cheap form of mid-term (from days to minutes) thermal
storage, particularly when used at scale for district heating, and in comparison, to battery
storage. A contribution is described in this section of the use of a multi-node approach in
PyLESA to represent the stratification and to incorporate thermal characteristics through
state of charge dependence on node temperatures.
The contributions of this work consist of (i) an extension of the Duffie and Beckman
model [230] by the development of an ambient heat loss method, and (ii) Python code
implementing the multi-node model which can be utilised in PyLESA and in other
planning-level tools.
The approach taken is similar to those implemented in detailed building design
simulation such as TRNSYS [231]. This addresses the gap which has been identified in
planning-level modelling tools in representing the temperature dependence of hot water
tanks. Figure 5.22 shows the PyLESA input requirements for modelling the hot water
tank.
Figure 5.22: Hot water tank modelling input requirements
The hot water tank is modelled as a cylinder which is vertically orientated with an
outside shell of insulation. The tank is configured using a 2-port connection and the use
of 5 temperature sensors, in accordance with CIBSE guidance [232] for district heating
design, see Figure 5.19 for a schematic showing this configuration.
The main characteristics [230] which require capturing in the modelling are:
• Capacity per unit volume
110 |
• Temperature range of operation
• Means and power requirements of charging and discharging
• Structural elements of tank
• Control
• Degree of stratification
Physical processes of hot water tanks include: (i) heat losses through tank due to a
difference in internal temperature and external ambient temperature, (ii) conduction heat
transfer in the water due to temperature differences at different layers, (iii) convective
flows due to cooling of water at edge of tank resulting in density differences, (iv) buoyancy
induced flows due to load temperature being lower than temperature of layer it is entering
at, (v) entering fluid mixing with lower temperature water due to high flow rate (carrying
kinetic energy), and (vi) recirculation of water from connections.
A selection process [233] was used to select a model to represent the stratification in
the tank, see Figure 5.23. The main selection decisions were to select a suitable model for
simulating without data and a balance between accuracy and computational time. The
multi-node model was chosen as it fits these criteria. In addition, due to the use of 5
temperature sensors a 5-node model was selected as a default, however, the number of
nodes can be changed by the user.
111 |
Figure 5.23: Decision tree for modelling stratification of hot water tank
In the multi-node model, the hot water tank is divided into N nodes which are disks
of fixed volumes. Each node has energy and mass balance equations which can be used
to determine the changes in temperature due to input and output flow. This results in N
differential equations which are solved simultaneously to calculate the final node
temperatures for each node. Duffie and Beckman’s description of a multi-node model
provides the basis for the model developed here [230].
Figure 5.24 shows the energy balance for node i, including the top and bottom nodes
and above and below nodes. The final energy balance equation for each node, denoted i,
contains four terms which each have an explicit node temperature dependence:
• Ambient heat loss between inside and outside of the tank.

• Charging and discharging.
• Mixing between nodes.
112 |
Figure 5.24: Energy balance for node i, including the top, bottom, above, and below nodes
5.5.1. Ambient Heat Loss

The ambient heat loss is applied to each node in the multi-node model, with additional
heat loss for the top and bottom nodes. The calculation uses the following steps:
• Calculate ideal conduction losses through the cylinder at the side and the
top/bottom, including a correction factor to account for insulation imperfections.
• Estimate losses from openings in insulation, e.g. pipe connections.
• Factor by a user input empirical sensitivity parameter.
Heat losses from a hot water storage tank occurs through the insulation material due
to the temperature difference between the water medium inside the tank and the
temperature outside the tank. If the tank is located indoors then a 15°C constant room
temperature is assumed, and if it is located outside then outdoor air temperature is
assumed. Figure 5.25 shows a schematic of the hot water tank to be modelled; a vertical
cylinder with an outside shell of insulation. The curvature of the top and bottom surfaces
is assumed to be negligible and modelled as flat.
113 |
Figure 5.25. Hot water tank schematic to be modelled
The heat loss from the tank can be estimated by assuming the insulation material
provides the only significant resistance to heat flow, i.e. neglecting the relatively small
convective resistances on the inside and outside tank surface, and the minor conductive
resistances provided by the tank wall and the tank cladding. This calculation also assumes
the tank is at steady state conditions. The heat losses through the horizontal section and
the top and bottom plane wall sections are included in the calculation. The losses at the
corners are neglected.
For the horizontal losses, Fourier’s law of conduction (the only heat transfer
considered) is used to calculate the heat transfer between inside the tank and outside the
tank (for cylindrical tank):
𝑑𝑇
𝑄𝑐𝑜𝑛𝑑,𝑐𝑦𝑙 = −𝑘𝐴
𝑑𝑟
where 𝐴 = 2𝜋Δ𝑟𝐿
integrating both sides…
114 |
𝑟2 𝑇𝑖
𝑄𝑐𝑜𝑛𝑑,𝑐𝑦𝑙
∫ 𝑑𝑟 = ∫ 𝑘𝑑𝑇
𝐴
𝑟1 𝑇0
𝑇𝑖 − 𝑇0
𝑄𝑐𝑜𝑛𝑑,𝑐𝑦𝑙 = 2𝜋𝑘𝐿
ln (𝑟2 ⁄𝑟1 )
Taylor expansion of the natural log, assuming small r2 – r1…
2Δ𝑟
ln(𝑟2 ⁄𝑟1 ) =
𝑟2 + 𝑟1
𝑇𝑖 − 𝑇𝑜
𝑄𝑐𝑜𝑛𝑑,𝑐𝑦𝑙 = 𝜋𝐿𝑘(𝑟2 + 𝑟1) [ ]
𝑟2 − 𝑟1
where Qcond,cyl is the horizontal heat transfer, k is the thermal conductivity, A is the
surface area, L is the height of the cylinder, Ti is the temperature of node i, T0 is the
ambient temperature, r2 is the total cylinder radius, and r1 is the internal cylinder radius.
Figure 5.26: Schematic showing horizontal conduction loss from hot water tank through the insulation layer
115 |
For the vertical losses from the top and bottom plane walls:
𝑑𝑇
𝑄𝑡𝑜𝑝/𝑏𝑜𝑡𝑡𝑜𝑚 = −𝑘𝐴
𝑑𝑥
𝑄𝑡𝑜𝑝/𝑏𝑜𝑡𝑡𝑜𝑚 = 𝑘𝜋𝑟1 2 [ ]
𝑟2 − 𝑟1
Combining the horizontal and vertical heat losses:
𝑄𝑐𝑜𝑛𝑑,𝑐𝑦𝑙 + 𝑄𝑡𝑜𝑝/𝑏𝑜𝑡𝑡𝑜𝑚 = 𝑘𝜋𝑟1 2 𝐿 [ ] (𝑟 + 𝑟1 )
𝑟2 − 𝑟1 2
The calculation for this first step then includes a correction factor, Fc, to account for
insulation imperfections, such as voids and thermal bridges between the inner tank shell
and outer casing (e.g. a stainless-steel ring supporting the inner tank shell on the ground).
𝑄𝑆𝑡𝑒𝑝 1 = 𝐹𝑐 (𝑄𝑐𝑜𝑛𝑑,𝑐𝑦𝑙 + 𝑄𝑡𝑜𝑝/𝑏𝑜𝑡𝑡𝑜𝑚 )
Table 5.1: Thermal conductivity of insulation materials
Insulation Material Thermal conductivity [W/m.K]

Polyurethane 0.025
Fibreglass 0.040
Polystyrene 0.035
Further corrections are made by considering the openings in the insulation such as for
pipe connections, valves, thermostats and electric elements.
Table 5.2: Empirical approximations of heat loss due to tank insulation openings
Tank insulation opening Heat loss increase [kWh/24 h

@ ΔT = 55 K]
Tank opening (e.g. thermostat pocket) 𝐴𝑜𝑝𝑒𝑛𝑖𝑛𝑔 × 27
Uninsulated pipe or fitting (e.g. PTR valve) 𝑑𝑝𝑖𝑝𝑒 × 5
Insulated pipe or fitting 𝑑𝑝𝑖𝑝𝑒 × 3.5
Losses from these are added to the previously calculated heat losses to give a final
heat loss (divide by 0.024 to convert from kWh/day to W). PTR valve is a pressure relief
valve, thermostat pocket is for thermostats (i.e. temperature sensors), and insulated pipe
or fitting is where water is taken in and out of the tank.
116 |
𝑄𝑙𝑜𝑠𝑠𝑒𝑠 = 𝑄𝑆𝑡𝑒𝑝 1 + 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛 𝑙𝑜𝑠𝑠𝑒𝑠
The final step is to add an empirical factor, Fe, which can be input by the user who
has data on heat loss from the hot water tank to be modelled. This is introduced as hot
water tanks have been monitored to have higher heat losses than theoretical models
suggest.
𝑄𝑙𝑜𝑠𝑠𝑒𝑠 𝑒𝑚𝑝𝑖𝑟𝑐𝑎𝑙 = 𝑄𝑙𝑜𝑠𝑠𝑒𝑠 ∗ 𝐹𝑒
An alternative approach would have been to determine a regression equation based
on existing hot water tank data and to determine a lumped heat loss coefficient which
encompasses all the theoretical considerations included here. However, this approach
would require monitored data. Therefore, it was decided, since the work here focusses on
design, that this data may not be available and that a theoretical approach with correction
factors was more suitable.
5.5.2. Charging and Discharging

The hot water tank can be in a state of charging, discharging, or standby. These states
dictate which of the flows and returns are active to and from the heat source and demand.
Flow from the heat source to the tank and flow from the tank to the heat demand are
only to or from the top node. Return from the heat demand to the tank and return from
the tank to the heat source are only to or from the bottom node. This means that in each
modelling timestep the maximum charge/discharge is limited to the volume of one node.
To get around this an internal hot water tank timestep is used based on the number of
nodes. For example, take a hot water tank with 5 nodes and a volume of 1000L. The
maximum charge/discharge in one timestep is a total flow of 200L but there may be
situations where the charge/discharge is larger than this. Therefore, for each model
timestep (1 hour) the number of internal hot water tank simulation timesteps is the same
as the number of nodes. In this example there are 5 nodes, so 5 hot water tank internal
timesteps are run for each model timestep in order for the tank to be able to
charge/discharge fully in each simulation timestep.
The state of the hot water tank is calculated using the following charging and
discharging functions.
117 |
If the charging function is 1 then the node is charging, and the following charging
term is non-zero.
𝑄𝑐ℎ𝑎𝑟𝑔𝑖𝑛𝑔 = 𝐹𝑐 𝑚𝑖𝑛
̇ 𝑐𝑝 𝑇𝑖𝑛
where Fc is the charging function, 𝑚𝑖𝑛 is the mass flow of charging water into the
node i, 𝑐𝑝 is the specific heat of water, and 𝑇𝑖𝑛 is the temperature of charging water.
Similar equations are used for the discharging term.
𝑄𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑖𝑛𝑔 = −𝐹𝑑 𝑚𝑜𝑢𝑡

̇ 𝑐𝑝 𝑇𝑖
where Fd is the discharging function, 𝑚𝑜𝑢𝑡 is the mass flow of discharging water out
of the node i, 𝑐𝑝 is the specific heat of water, and 𝑇𝑖 is the node temperature.
5.5.3. Mixing Between Nodes

The mixing between nodes is dependent on the state of the hot water tank, and therefore
the charging and discharging functions from above. The following equations, along with
the state of the tank, are used to determine the mixing term.
If the hot water tank is in a state of charging, the direction of mixing between node i
and the node above is downwards:
𝑄𝑚𝑖𝑥 = 𝐹𝑐 𝑚𝑑𝑜𝑤𝑛
̇ ∗ 𝑐𝑝 ∗ 𝑇𝑖−1
If the hot water tank is in a state of discharging, the direction of mixing between node
i and the node above is upwards:
𝑄𝑚𝑖𝑥 = 𝐹𝑑 ∗ −1 ∗ 𝑚̇𝑢𝑝 ∗ 𝑐𝑝 ∗ 𝑇𝑖
In these equations a negative sign is representative of flow leaving the node i.
118 |
If the hot water tank is in a state of charging, the direction of mixing between node i
and the node below is downwards:
𝑄𝑚𝑖𝑥 = 𝐹𝑐 ∗ −1 ∗ 𝑚𝑑𝑜𝑤𝑛
̇ ∗ 𝑐𝑝 ∗ 𝑇𝑖
If the hot water tank is in a state of discharging, the direction of mixing between node
i and the node below is upwards:
𝑄𝑚𝑖𝑥 = 𝐹𝑑 𝑚̇𝑢𝑝 ∗ 𝑐𝑝 ∗ 𝑇𝑖+1
5.5.4. Final Energy Balance Equation

The final energy balance equation for a node, denoted i, contains the four terms which
are described above and can be written as:
𝑑𝑇𝑖
(𝑚𝑖 𝑐𝑝 ) = 𝐹𝑐 𝑚̇𝑖𝑛 𝑐𝑝 𝑇𝑖𝑛 − 𝐹𝑑 𝑚̇𝑜𝑢𝑡 𝑐𝑝 𝑇𝑖 + 𝑚̇𝑑𝑜𝑤𝑛 𝑐𝑝 𝑇𝑖−1 − 𝑚̇𝑢𝑝 𝑐𝑝 𝑇𝑖 − 𝑚̇𝑑𝑜𝑤𝑛 𝑐𝑝 𝑇𝑖
𝑑𝑡
+ 𝑚̇𝑢𝑝 𝑐𝑝 𝑇𝑖+1 − 𝐹𝑒 [𝐹𝑐 𝑘 [ ] 𝜋[𝑟1 2 + ℎ(𝑟2 + 𝑟1 )] + 𝑐. 𝑙. ] (𝑚𝑖 𝑐𝑝 )
𝑟2 − 𝑟1
This equation can be written in a simpler format using coefficients:
𝑑𝑇𝑖
= 𝐴𝑇𝑖 + 𝐵𝑇𝑖−1 + 𝐶𝑇𝑖+1 + 𝐷
𝑑𝑡
where the coefficients are:
𝐴 ∗ (𝑚𝑖 𝑐𝑝 ) = −𝐹𝑑 𝑚̇𝑜𝑢𝑡 𝑐𝑝 − 𝑚̇𝑢𝑝 𝑐𝑝 − 𝑚̇𝑑𝑜𝑤𝑛 𝑐𝑝

𝑇𝑖
− 𝐹𝑒 𝐹𝑐 𝑘 [ ] 𝜋[𝑟1 2 + ℎ ∗ (𝑟2 + 𝑟1 )] (𝑚𝑖 𝑐𝑝 )
𝑟2 − 𝑟1
𝐵 ∗ (𝑚𝑖 𝑐𝑝 ) = 𝑚̇𝑑𝑜𝑤𝑛 𝑐𝑝
𝐶 ∗ (𝑚𝑖 𝑐𝑝 ) = 𝑚̇𝑢𝑝 𝑐𝑝
𝑇𝑜
𝐷 ∗ (𝑚𝑖 𝑐𝑝 ) = 𝐹𝑐 𝑚̇𝑖𝑛 𝑐𝑝 𝑇𝑖𝑛 + 𝐹𝑒 𝐹𝑐 𝑘 [ ] 𝜋[𝑟1 2 + ℎ(𝑟2 + 𝑟1 )](𝑚𝑖 𝑐𝑝 )
𝑟2 − 𝑟1
+ 𝐹𝑒 𝑐. 𝑙. (𝑚𝑖 𝑐𝑝 )
Each node can be written in this form which together are a set of N first-order
ordinary differential equations. These are solved simultaneously using the odeint function
from the SciPy software package which is commonly used to solve a system of ordinary
differential equations [234].
119 |
Figure 5.27: Flow diagram for hot water tank model
5.5.5. Validation
The developed multi-node model was validated by comparison between modelled output
and monitored data collected from a biomass and hot water tank district heating system
at West Whitlawburn (see description in Section 6.2). A 10-hour charging period, a 3-
hour discharging period, and a two-day period of operation are analysed by interpreting
the difference in node temperature evolutions between the model output and the data. 6
nodes were used in the model to directly compare to the 6 temperature points available
from the monitored data, and 15-minute timesteps were used in both the modelling
output and monitored data. The analysis is undertaken both graphically, with explanation
of differences, and statistically, through calculation of the absolute error and the mean
absolute percentage error. While graphically the evolution of the node temperatures looks
different for the modelled output and the monitored data, the statistical analysis shows
that there is a small absolute percentage error for each node and on average across all
nodes. This validation section ends with a comparison of the outputs from the multi-node
model and a simple energetic model.
5.5.5.1. Charging Period (10-hour)
Figure 5.28 shows the evolution of the node temperatures over a 10-hour charging period
for the modelled output and monitored data. The model output shows in the first timestep
the top node charging, along with increases in the temperatures of the nodes below the
top node, due to mixing from the nodes above. In subsequent timesteps all nodes are
120 |
Figure 5.28: 10-hour charging period showing modelled and monitored node temperatures. Four 15-minute timesteps
is one hour.
charged and, additionally, increase in temperature due to mixing from the nodes above.
The monitored data displays a node temperature evolution with short periods of each
node increasing in temperature from low to high.
Comparison of the two shows that the time taken to charge the hot water tank is
similar in both the data and model. It also indicates the thermocline is smaller for the real
tank than is modelled, meaning there is greater stratification in the real tank. This is caused
by the mixing between the nodes in the model, an effect which is increased with low flow
relative to node volume. The use of the internal timesteps for the hot water tank reduces
the flow in each computational step and further increases the mixing between the fixed
volume nodes. It is possible to capture greater stratification by using a larger number of
nodes relative to flow rates. However, this results in slower computational time.
Alternatively, a moving boundary model, which uses two variable mass nodes, could be a
more computationally efficient approach for modelling perfect stratification.
The absolute error on the node temperatures for the modelled hot water tank is shown
over the charging period in Figure 5.29. The absolute error briefly exceeds 6°C for the 3rd
and 6th nodes (purple and green on graphs), and these nodes are discussed below.
121 |
Figure 5.29: Absolute error on node temperatures for 10-hour charging period. Four 15-minute timesteps is one hour.
For the 3rd node the monitored data shows the thermocline pass at the 4th timestep,
and after this the temperature exceeds the nodes above. This could be due to uncertainty
around the exact position of the inlet, which in the real tank could be closer to the third
node than the top node. There could also be dynamic turbulent effects near the top of
the tank which cause this temperature inversion. The multi-node model assumes that the
inlet is to the top node and does not include turbulent effects.
For the 6th node the real tank shows slight increase in temperature until a sharper
gradient around the 30th timestep. This indicates low mixing from the nodes charging
above, and the movement of the thermocline through the node around the sharp
temperature gradient. The model shows a more consistent temperature gradient slope and
does not capture the sharp temperature gradient caused by the thermocline. As discussed
earlier, this is due to the low flow and number of nodes modelled causing greater mixing
between the nodes over the charging period.
The mean absolute error and mean absolute percentage error were calculated for each
node over the charging period. The average of the mean absolute percentage error across
the nodes is 2.2% and this is in agreement with a similar multi-node tank model which
122 |
was validated against experimental data [235]. Additionally, due to the reasons discussed
above relating to greater node mixing in the model output, the bottom node has the largest
margin of error. An average across the nodes of a mean absolute percentage error of 2.2%
indicates that the model is suitable to represent the charging period of a hot water tank.
Table 5.3: Statistical analysis of example charging period
Statistical Measure 1 2 3 4 5 6 Average

Mean absolute error (°C) 1.2 1.4 1.5 1.8 1.8 2.5 1.7
Mean absolute percentage error (%) 1.5 1.6 1.8 2.3 2.3 3.4 2.2
5.5.5.2. Discharging Period (3-hour)

Figure 5.30 shows the evolution of the node temperatures over the discharging period for
the modelled output and monitored data. The model output shows the node temperatures
decreasing as the hot water tank meets a heat demand. The lower nodes decrease
temperature first, while the above nodes decrease in temperature due to mixing from the
nodes below. The hot water tank stops discharging when the top node drops below the
service temperature of the district heating network.
The monitored data displays a node temperature evolution where each node
consistently reduces in temperature, and short periods where there is a more rapid
decrease in temperature from low to high as the thermocline moves within the tank. The
tank is emptied when the top node decreases to the service temperature (in this example
this is 72°C).
Comparison between the model and data shows that the time taken to charge the hot
water tank is similar in both the data and model. However, it shows the thermocline is
smaller for the real tank than is modelled, meaning there is greater stratification in the real
tank. This effect appears to be less pronounced for the discharging period compared to
the charging period analysed above.
The end of the discharging scenario has ending node temperatures in a range between
the return temperature and the service temperature. There will be scenarios where nodes
are charged above the return temperature but below the service temperature. In these
scenarios the charging is not wasted because in subsequent periods less charging is
required from the heat source to charge the hot water tank to a stage where some nodes
are above the service temperature.
123 |
Figure 5.30: Discharging period showing modelled and monitored node temperatures. Four 15-minute timesteps is
one hour.
The absolute error on the node temperatures for the modelled hot water tank is shown
over the charging period in Figure 5.31. The absolute error briefly exceeds 5°C for the 1st
and 4th nodes (red and yellow on graphs). The 1st node mixes with other nodes earlier in
the multi-node model than the real tank data, and as a result towards the end of the
discharging period is below the real tank node temperature. The 4th node discharges more
quickly in the model than in the real tank.
node over the discharging period. The average of the mean absolute error across the
nodes is 1.4°C and this is, as with the charging period, in broad agreement with a similar
multi-node tank model which was validated against experimental data [235]. An average
across the nodes of a mean absolute percentage error of 1.9% indicates that the model is
suitable to represent the discharging period of a hot water tank.
124 |
Figure 5.31: Absolute error on node temperatures for discharging period. Four 15-minute timesteps is one hour.
Table 5.4: Statistical analysis of example discharging period

5.5.5.3. Two-Day Period

An example two-day period is analysed in order to investigate the ability of the model to
respond to both charging and discharging period, and to carry out statistical analysis over
a longer time period.
Figure 5.32 shows the evolution of the node temperatures over the two-day period
for the modelled output and monitored data. The model output shows the node
temperatures increasing and decreasing over the period as the hot water tank is charged
and discharged, and the monitored data shows similar patterns.
Figure 5.33 shows the absolute error, and there are periods with an absolute error
higher than that identified in the charging and discharging periods. This can be due to
125 |
Figure 5.32: Two-day period with charging and discharging periods showing modelled and monitored node
temperatures. Sixteen 15-minute timesteps is four hours.
reliability of the data. For example, the 2nd node shows the highest absolute error over the
period, and this is due to a charging event which occurs in the real tank data, as seen by
the increase of temperature of the 2nd node. However, this does not correlate with the
data on the output from the heat source and the demand of the district heating network,
which indicates that there is missing or wrong data, or that there are dynamics between
the measured points which is not captured. Additionally, the error is also due to not fully
capturing the stratification, as was identified in analysis of the charging and discharging
periods earlier.
node over the charging period. The average of the mean absolute error across the nodes
is 2.5°C, and an average across the nodes of a mean absolute percentage error of 3.2%.
The maximum absolute values over the 2-day period exceed 12°C which indicates that
the model is not suitable for precise tracking of the node temperatures. However, the
average error over the period is still low which indicates that while the model does not
provide accurate snapshots of the tank node temperature it is capturing the consecutive
charging and discharging events over this two-day period. The statistical analysis for the
126 |
charging, discharging, and two-day period indicates that the model is suitable to represent
the consecutive charging and discharging events, over multi-day periods, characteristic of
a hot water tank in a district heating system.
Table 5.5: Statistical analysis of example 2-day period

Figure 5.33: Absolute error on node temperatures for two-day period. Sixteen 15-minute timesteps is four hours.
5.5.5.4. Comparison to Energetic Model

The developed multi-node model was also compared to a simple energetic model which
is typically used by planning-level energy system tools similar to PyLESA.
Figure 5.34 shows the calculated available discharging capacity output by both the
multi-node model and energetic model over the same two-day period examined in the
previous sub-section. The available discharging capacity is calculated using the multi-node
model based upon node temperatures in each timestep, while the energetic model either
127 |
adds or subtracts the charging or discharging energy below a maximum capacity. The
available discharging capacity calculated using the multi-node model at the beginning of
the evaluated period is used as the starting value for the energetic model.
The energetic model predicts both higher charging and discharging compared to the
multi-node model. This can be seen by an increase in the difference between the available
discharging capacity over the two-day period. At the end of this period the energetic
model predicts the available discharging capacity is 277kWh (+35%) greater than the
multi-node model.
Figure 5.34: Comparison of output from multi-node model and energetic model. Sixteen 15-minute timesteps is four
hours.
In conclusion, this work shows the strengths and weaknesses of a multi-node model
and its importance in planning-level modelling tools in order to justify its inclusion in
PyLESA. The multi-node model is an approach which attempts to capture the thermal
characteristics of a hot water tank in order to capture the available state of charge more
realistically than the commonly used energetic model. Other models of hot water tank
may be included in PyLESA and these may more accurately capture the evolution of node
temperatures. However, it is key that balance is struck between accuracy and the need for
simple inputs along with low computational times. In future work and outside the scope
128 |
of this thesis, thermal storage modelling may benefit from standardised system
performance testing across a range of thermal storage technologies and their applications
which would enable standard modelling methods to be established.
5.6. Electricity Tariffs

Existing and future tariffs can be generated and modelled in PyLESA. A contribution to
the state of the art is made by including a future wind-based electricity tariff generator in
PyLESA. This allows PyLESA to perform analysis of future energy system scenarios
which may include electricity pricing structures which are highly differential and based on
renewable power generation. This differs to existing tariffs which are priced according to
demand and inflexible baseload generation, amongst other complex factors.
The input requirements for these are in Figure 5.35, Figure 5.36, and Figure 5.37, and
the modelling details are described below. Figure 5.38 is an example of the inputs required
for future work to be done by including the balancing mechanism and grid services.
Figure 5.35: Input requirements for time-of-use tariff linked to wholesale market
129 |
Figure 5.36: Input requirements for variable periods tariff
Figure 5.37: Input requirements for a future wind-based tariff coupled with a flat rate base
130 |
Figure 5.38: Example of future work to be done by including balancing mechanism and grid services
5.6.1. Existing
Traditionally, domestic electricity tariffs available from energy suppliers in the UK have
been flat rate tariffs where a price is agreed which is static regardless of when electricity
is used or variable periods tariffs, such as economy 7 where a cheaper electricity price is
available for 7 hours during the night. A newer form of tariff is time-of-use where
electricity prices fluctuate hourly (or sub-hourly) and are linked to the wholesale market.
This encourages users to shift demand from peak periods.
Table 5.6: Existing electricity tariff descriptors and examples
Tariff Description Pricing structure example

Flat rates Fixed price £130/MWh – All times
Variable Variable hourly with a fixed £150/MWh – Day
periods structure, e.g. day/night, £75/MWh - Night
weekday/weekend
Time-of-use Variable hourly, or sub-hourly, Linked to wholesale market,
e.g. linked to wholesale market premium pricing period between
4pm and 7pm, maximum set to
£350/MWh
131 |
Figure 5.39: Existing electricity tariffs over 72 hours
5.6.2. Future
In the future electricity grids are expected to be highly dependent on stochastic renewable
energy sources. Electricity prices could be affected by this with decreases during periods
where there are abundant zero-marginal cost renewable electricity and increases during
periods of low renewable electricity and there is increased reliance on power from
dispatchable sources.
Possible future renewable tariff synthesis can be supported in PyLESA. For example,
a future tariff could be synthesised in PyLESA using the following method. Firstly, an
existing tariff is chosen as a base: (i) a continuously fixed tariff, (ii) low demand coupled
with inflexible generation (such as nuclear) causing low price periods during the night, or
(iii) a flexible tariff based on avoidance of peak late afternoon demands. Then, a wind
farm output is modelled using the same method for the on-site wind power generation
described previously, and the resultant hourly power output is separated into top and
bottom bands of production. A discount is applied to the base tariff where wind power
output is in the top band of production and a premium applied where it is in the bottom
band. The wind bands, and the discount, and premium to be applied to the base tariff are
defined by the user. Figure 5.40 shows PyLESA synthesised tariffs with the wind
132 |
generation discount and premium applied to each of the three base cases. The
functionality in PyLESA allows other future tariff scenarios to be generated and
investigated in the modelling of future scenarios.
Figure 5.40: Top graph: wind farm modelled output including upper band and lower band over 72 hours; Bottom
graph: renewable electricity tariff with discounts and premiums applied over the same 72 hours
133 |
Figure 5.41: Flow diagram for electricity tariff generator
5.7. Fixed Order Control

The fixed order control implementation in PyLESA uses a pre-defined set of rules to
order the dispatch of supply and determine the usage of storage. The user can rearrange
the set of rules at the start of the simulation but cannot change the order according to
dynamic system variables during the simulation period. This functionality is intended as a
representation of a commonly employed control when introducing load shifting
mechanisms. It will be compared to more advanced model predictive control which is
described in the next section. Figure 5.42 shows the input requirements for the fixed order
controller.
Figure 5.42: Fixed order control input requirements
134 |
This control is used to represent a classical controller which uses fixed setpoints for
components (e.g. thermal storage temperature setpoint) to provide on/off and PID
output responses. Table 5.7 splits the set of processes between above/below an import
setpoint. This split allows for a different set of rules for day and night to take load shift
from higher prices during the day to lower prices during the night.
Table 5.7: Rules for fixed order control strategy, split into condition based on an import setpoint
Above import setpoint Below import setpoint

Electricity demand Electricity demand
1 RES to demand 1 RES to demand
2 ES to demand 2 Import to demand
3 Import to demand 3 ES to demand
Heat demand Heat demand
4 HP RES to demand 4 HP RES to demand
5 E-AUX RES to demand 5 E-AUX RES to demand
6 TS to demand 6 HP import to demand
7 ES to HP to demand 7 TS to demand
8 HP import to demand 8 ES to HP to demand
9 AUX to demand 9 AUX to demand
Thermal storage Thermal storage
10 HP RES to TS 10 HP RES to TS
11 E-AUX RES to TS 11 E-AUX RES to TS
Electricity storage 12 HP import to TS
12 RES to ES Electricity storage
Export 13 RES to ES
13 RES to export 14 Import to ES
Export
15 RES to export
The processes are run sequentially with the output from each process producing a set
of results and checks. Figure 5.43 shows the flow of results and checks when running the
fixed order controller in PyLESA. The validation of this controller can be found in
Chapter 6, where the methodology is applied to a case study. This is because it is easier
to analyse the operational decisions made using a well-defined example.
135 |
Figure 5.43: Flow diagram showing process i as a chunk of the flow of results and checks when running the fixed
order controller
136 |
5.8. Model Predictive Control
Model Predictive Control (MPC) captures the dynamic influences of energy systems and
optimises the performance of the components as a supervisory control strategy. MPC can
be based upon models from building and system simulation models or artificial intelligent
techniques. Contributions to the state of the art are made by including MPC in PyLESA
as the review identified a gap in the ability of existing planning-level energy tools to model
a model predictive control strategy.
An MPC controller consists of several key components:
• Objective function which an optimiser minimises/maximises.

• Prediction horizon which is the period over which the optimisation is performed.
• Decision timestep which is the interval between solving optimisation problem.
• Manipulated variables can be varied by the controller.
• Optimisation solver which is chosen based upon optimisation type and required
speed.
• Feedback signal which provides updated system variables for next optimisation
step.
Figure 5.44: Model predictive control input requirements
PyLESA uses Economic Model Predictive Control (EMPC) which aims to maximise
the economic performance of a system by varying control variables to minimise costs
over a receding prediction horizon. It is useful for complex local energy systems which
consist of multiple supply options, stochastic renewable power generation, storage, and
fluctuating electricity prices. Traditional controllers are not suited to optimise the
operation of these types of systems. PyLESA allows the range of optimisation algorithms
available in Python to be accessed.
137 |
State equations are used to predict changes in state variables and are shown here for
the heat balance, thermal storage state of charge, storage charging, heat pump thermal
output, electric auxiliary thermal output, and use of surplus on-site renewable generation.
𝐻𝐷 = 𝐻𝑃𝑡𝑟𝑑 + 𝐻𝑃𝑡𝑖𝑑 + 𝐴𝑈𝑋𝑑 + 𝐴𝑈𝑋𝑟𝑑 + 𝑇𝑆𝑑

𝑑 SOC
= 𝑇𝑆𝑐 − 𝑇𝑆𝑑 − 𝑙𝑜𝑠𝑠𝑒𝑠
𝑑𝑡
𝐻𝑃𝑜𝑛/𝑜𝑓𝑓 . 𝐻𝑃𝑡_𝑣𝑎𝑟 = 𝐻𝑃𝑡𝑟𝑠 + 𝐻𝑃𝑡𝑟𝑑 + 𝐻𝑃𝑡𝑖𝑑 + 𝐻𝑃𝑡𝑖𝑠
𝐴𝑈𝑋 = 𝐴𝑈𝑋𝑑 + 𝐴𝑈𝑋𝑠 + 𝐴𝑈𝑋𝑟𝑑 + 𝐴𝑈𝑋𝑟𝑠

𝑇𝑆𝑐 = 𝐻𝑃𝑡𝑟𝑠 + 𝐻𝑃𝑡𝑖𝑠 + 𝐴𝑈𝑋𝑟𝑠 + 𝐴𝑈𝑋𝑠
𝑅𝐸𝑆𝑠urplus = (𝐻𝑃𝑡𝑟𝑠 + 𝐻𝑃𝑡𝑟𝑑 )/𝐶𝑂𝑃 + 𝐴𝑈𝑋𝑟𝑑 + 𝐴𝑈𝑋𝑟𝑠 + 𝑒𝑥𝑝𝑜𝑟𝑡
where HD is the heat demand, HPtrd is the heat pump thermal output from renewables
to demand, HPtid is the heat pump thermal output from imports to demand, AUXd is the
auxiliary thermal output to demand, AUXrd is the auxiliary thermal output from
renewables to demand, TSd is the thermal storage discharging, SOC is the state of charge
of the thermal storage, TSc is the thermal storage charging, losses is the losses from the
thermal storage, HPon/off is the binary on/off state of the heat pump, HPt_var is the total
thermal output of the heat pump, HPtrs is the heat pump thermal output from renewables
to storage, HPtis is the heat pump thermal output from imports to storage, AUX is the
total auxiliary thermal output, AUXs is the auxiliary thermal output from imports to
storage, AUXrs is the auxiliary thermal output from renewables to storage, RESsurplus is the
surplus electricity after electrical demand electrical demand has been met, COP is the
coefficient of performance of the heat pump in that timestep, and export is the surplus
electricity exported from the local energy system.
A mixed integer linear programming problem can then be formulated which
minimises electricity costs by controlling the heat pump and thermal storage. The
formulation contains: the objective function, state equations lumped into a generic state
equation, inequality constraints, and allowed values for the heat pump status (integer
on/off operation is allowed).
𝑚𝑖𝑛 𝜙 = ∑k∈ℳ[Ic,k (HPi,k + EDi,k ) + Ac,k AUX i,k − Ec,k EXe,k ]
𝑥,𝑢,𝑦
s.t.
138 |
𝑥𝑘+1 = 𝐴𝑑 𝑥𝑘 + 𝐵𝑑 𝑢𝑘 + 𝐸𝑑 𝑑𝑘
𝐻𝑃𝑑𝑢𝑡𝑦,𝑘 ≥ 𝐻𝑃t var,k ≥ 𝐻𝑃𝑚𝑖𝑛,𝑘
𝑆𝑂𝐶𝑘 ≤ 𝑇𝑆𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦
𝑇𝑆𝑐,𝑘 ≤ 𝑇𝑆max 𝑐ℎ𝑎𝑟𝑔𝑒
𝑇𝑆𝑑,𝑘 ≤ 𝑆𝑂𝐶𝑘
𝐴𝑈𝑋𝑘 ≤ 𝐴𝑈𝑋𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦
𝐻𝑃𝑠𝑡𝑎𝑡𝑢𝑠 ∈ {0, 1}
ℳ ∈ {0, 1, …, N} and N is the prediction horizon and a sampling time of 1 hour is
used. At each timestep the optimisation problem is solved, and a set of control variables
is obtained. The first control variable is implemented and new state variables and forecast
variables are updated in the next iteration of the optimisation problem.
It is assumed in this formulation that forecast variables and cost coefficients are
known with perfect foresight. However, an MPC running in real-time is dependent on
the accuracy of the predictions of the forecast variables. Therefore, the perfect foresight
MPC approach results in an idealised operational schedule; the potential benefits from
MPC will be overestimated. Stochastic MPC approaches have been developed which
incorporate the uncertainty in forecast variables [236]. Alternative approaches for
incorporating uncertainty of prediction variables have used the future value of the
reference signal [237], and historical value of the control signal [238]. In future PyLESA
can be developed to incorporate methods for representing prediction error.
The presented mixed integer linear programming problem is solved using GEKKO,
a Python package for machine learning and optimisation [239]. It uses large-scale solvers
for linear, quadratic, nonlinear, and mixed integer programming and in the MPC
developed for PyLESA the APOPT solver is used [240]. GEKKO has previously been
used in energy system analysis to optimise the performance of thermal storage to minimise
cost operation of a district energy system in a time-of-use electricity market [241] and
optimization of a hybrid solar thermal and fossil fuel system [242].
The developed MPC strategy uses a simplified energetic model for the thermal storage
in the optimisation problem. This may lead to overestimation of the ability of the thermal
store to meet demand in a later period, and an increase in the electrical import costs due
to sub-optimal deployment of the heat sources.
139 |
The validation of this controller is undertaken in Chapter 6, where PyLESA is applied
to a case study. This is because it is easier to analyse the operational decisions made using
a well-defined example.
Figure 5.45: Flow diagram for model predictive control
5.9. Modelling Outputs (KPIs)

A set of Key Performance Indicators (KPIs) are included as outputs from PyLESA in
order to compare different system configurations and component sizes.
Economic KPIs are useful for design studies such as sizing studies, as larger heat
pumps and thermal storage will generally decrease operating cost, but this needs to be
balanced against the increase in capital cost. Levelized cost of heat and energy can be used
as a metric to explore this relationship. Larger thermal storage can mean that a smaller
heat pump is required which is economically advantageous because thermal storage has a
smaller capital cost compared to heat pumps.
The economic KPIs are simplified because they are designed to allow for comparisons
between system configurations and component sizes, as opposed to being used to make
final financial decisions. For example, the levelized cost of heat does not include a
140 |
discount factor or maintenance costs. However, raw outputs from the tool, along with
additional financial parameters, can be used to calculate more advanced economic KPIs,
and/or the tool developed to incorporate these directly.
Technical KPIs are useful as economics are not always the sole driver of a project.
Additionally, future electricity tariff prices are uncertain [243,244] particularly when
calculating KPIs over 20-year system lifetimes.
The full set of KPIs output from PyLESA are listed in Table 5.8 and the equations
used are described in the following sub-sections.
Table 5.8: Set of key performance indicators which are outputs of the modelling methodology
Economic KPIs Technical KPIs

Capital cost (£) On-site RES used (%)
Operating cost (£) On-site RES used (kWh)
Cost of heat (£) Grid RES used (kWh)
Cost of electricity (£) Total RES used (kWh)
Lifetime total cost (£) Demand met by RES (%)
Levelized cost of heat (£/kWh) Heat met by RES (%)
Levelized cost of energy Peak HP to demand ratio
(£/kWh)
Heat pump utilisation (%)
Days of storage content
5.9.1. Levelized Cost of Energy

Levelized Cost of Energy (LCOE) is defined here as a measure of the average cost of
meeting electrical and heat demands over a 20-year project lifetime and has units of
£/kWh.
𝐶𝐴𝑃𝐸𝑋 + 20 ∗ 𝑂𝑃𝐸𝑋 − 𝐼𝑁𝐶𝐸𝑁𝑇𝐼𝑉𝐸
𝐿𝐶𝑂𝐸 =
(𝐻𝐸𝐴𝑇 𝐷𝐸𝑀 + 𝐸𝐿𝐸𝐶 𝐷𝐸𝑀) ∗ 20
5.9.2. Levelized Cost of Heat

Levelized Cost of Heat (LCOH) is defined as a measure of the average cost of meeting
the heat demand over a 20-year project lifetime and has units of £/kWh.
141 |
𝐶𝐴𝑃𝐸𝑋 + 20 ∗ 𝐻𝐸𝐴𝑇𝑂𝑃𝐸𝑋 − 𝐼𝑁𝐶𝐸𝑁𝑇𝐼𝑉𝐸
𝐿𝐶𝑂𝐻 =
(𝐻𝐸𝐴𝑇 𝐷𝐸𝑀) ∗ 20
5.9.3. On-Site RES Used

On-site RES used (ORES) is a metric which quantifies the percentage of self-
consumption of on-site renewable power generation. It can be adapted to specify the
source of on-site renewable power generation, i.e. ORESpv for on-site PV in the sizing
study in Chapter 6.
𝑆𝑈𝑀 𝐸𝑋𝑃𝑂𝑅𝑇
𝑂𝑅𝐸𝑆 = 1 −
𝑆𝑈𝑀 𝑅𝐸𝑆 𝐺𝐸𝑁𝐸𝑅𝐴𝑇𝐼𝑂𝑁
5.9.4. Grid RES Used

Grid RES used (GRES) is a metric which quantifies the percentage of the grid imports
which are classed as generated from RES. In this work it is only applicable for the wind
tariff. During periods where a discount is applied there is a high penetration of wind on
the grid and imports during these periods are classed as from renewable energy sources.
𝑆𝑈𝑀 𝑅𝐸𝑆 𝐼𝑀𝑃𝑂𝑅𝑇𝑆
𝐺𝑅𝐸𝑆 =
𝑆𝑈𝑀 𝐼𝑀𝑃𝑂𝑅𝑇𝑆
5.9.5. Demand Met by RES

Demand met by RES (DRES) is the percentage of the total electrical and heat demand
which is met from both Grid RES and On-site RES. As a reminder, the electrical demand
here refers to non-heat electricity. It can be adapted to specify the source of RES, i.e. on-
site PV and electrical imports during high wind periods in the sizing study in Chapter 6.
𝑆𝑈𝑀 𝐺𝑅𝐼𝐷 𝑅𝐸𝑆 + 𝑆𝑈𝑀 𝑂𝑁𝑆𝐼𝑇𝐸 𝑅𝐸𝑆
𝐷𝑅𝐸𝑆 =
𝑆𝑈𝑀 𝐻𝑃 + 𝑆𝑈𝑀 𝐴𝑈𝑋 + 𝑆𝑈𝑀 𝐸𝐿𝐸𝐶𝑇𝑅𝐼𝐶 𝐷𝐸𝑀𝐴𝑁𝐷
5.9.6. Heat Met by RES

Heat met by RES (HRES) is the percentage of the total heat demand which is met by
Grid RES and On-site RES. It can be adapted to specify the source of RES, i.e. HRESpv
and/or HRESpv+windtariff as used for on-site PV and/or electrical imports during high
wind periods in the sizing study in Chapter 6.
𝑆𝑈𝑀 𝐺𝑅𝐼𝐷 𝑅𝐸𝑆 𝐻𝐸𝐴𝑇 + 𝑆𝑈𝑀 𝑂𝑁𝑆𝐼𝑇𝐸 𝑅𝐸𝑆 𝐻𝐸𝐴𝑇
𝐻𝑅𝐸𝑆 =
𝑆𝑈𝑀 𝐻𝑃 + 𝑆𝑈𝑀 𝐴𝑈𝑋
142 |
5.9.7. Peak HP to Demand Ratio
Peak Heat Pump to Demand ratio (PHPD) is the peak output of the heat pump and peak
demand over the simulation period. This is a measure of the utilisation of the heat pump
capacity. If a heat pump is oversized this ratio will be low.
𝑃𝐸𝐴𝐾 𝐻𝐸𝐴𝑇 𝑃𝑈𝑀𝑃 𝑂𝑈𝑇𝑃𝑈𝑇
𝑃𝐻𝑃𝐷 =
𝑃𝐸𝐴𝐾 𝐻𝐸𝐴𝑇 𝐷𝐸𝑀𝐴𝑁𝐷
5.9.8. Heat Pump Utilisation

Heat Pump Utilisation (HPU) is defined as the percentage of heat demand which is met
by the heat pump. It does not incorporate losses from the heat pump, so may result in a
higher than 100% percentage.
𝑆𝑈𝑀 𝐻𝐸𝐴𝑇 𝑃𝑈𝑀𝑃 𝑂𝑈𝑇𝑃𝑈𝑇
𝐻𝑃𝑈 =
𝑆𝑈𝑀 𝐻𝐸𝐴𝑇 𝐷𝐸𝑀𝐴𝑁𝐷
5.9.9. Days of Storage Content

Days of Storage Content (DOSC) is a measure for quickly comparing the energetic
capacity of a hot water tank in the context of the heat demand to be met and has units of
days. The average day demand takes the annual demand and divides it by the number of
days in a year, 365. The equation below contains the following parameters with units:
capacity in kg, Cp in kJ/(kg °C), T in °C, and average day demand in kWh.
𝐶𝐴𝑃𝐴𝐶𝐼𝑇𝑌 ∗ 𝐶𝑝 ∗ Δ𝑇
𝐷𝑂𝑆𝐶 =
3600 ∗ 𝐴𝑉𝐸𝑅𝐴𝐺𝐸 𝐷𝐴𝑌 𝐷𝐸𝑀𝐴𝑁𝐷
5.10. Discussion
The underlying models of PyLESA detailed in this chapter have been developed to
address the identified gaps from the review of existing modelling tools, and the
contributions to the state of the art are discussed below:
• Open source: PyLESA is written primarily in Python and incorporates previous

tools for modelling wind turbines and PV. The inputs are via an Excel workbook
which is not open source software. However, building a GUI in Python for
inputting data and viewing modelling outputs would complete PyLESA as a fully
open source modelling tool. A distinct advantage of using open source software
is the ability to build upon the work done in the developed modelling tool. Several
143 |
avenues of further work are identified in this discussion (and in Chapter 7) and
PyLESA provides a useful framework for undertaking it.
• Heat pump: A few heat pump modelling approaches have been detailed, and the
choice is dependent on the available data. The standard test regression approach
includes explicit temperature dependence and is an advance on the existing
models utilised by planning-level modelling tools. Dynamic effects and part load
conditions are largely neglected, and for timesteps ≤10 minutes a dynamic model
is needed to capture start-up characteristics. To fully capture these a future
advanced model could be developed which uses a white-box model where each
of the major components of the heat pump is modelled [245]. However, this
approach could increase the input data requirements beyond that available at the
planning-level.
• Hot water tank: The multi-node modelling approach advances on the simple
energetic models utilised by other tools, as it means that thermal characteristics
can be represented while maintaining low input requirements. This modelling
approach means that the state of charge and heat loss of the hot water tank is
dependent on the node temperatures. The developed model increases
computational time significantly and the code could be optimised to decrease this.
The model does not allow for multiple heat sources to charge the tank, nor does
it allow the heat source to inject at any other node but the top node. This would
form a useful future development as technology such as solar thermal could be
incorporated to provide complimentary heat to a heat pump and hot water tank
system. The developed model exaggerates the mixing between nodes which is
likely to occur for a hot water tank designed to be highly stratified, as is desirable
for the load shifting mechanisms investigated. Therefore, while the energetic
models used in existing models likely result in underestimations of the capacity
required to perform load shifting, the model developed within PyLESA is likely
to overestimate the capacity required.
• Future electricity markets: A wind-based electricity tariff generator has been
presented as a way of including a tariff which could exist in the UK in the future.
This is because of the high resource of onshore and offshore wind power that
exists, and a future 100% renewable energy system would likely have wind power
as one of the main producers of energy. Electricity markets which incentivise
144 |
flexibility will become more prevalent as the grid decarbonises. This will likely
involve higher rewards from participating in ancillary markets such as those
existing, e.g. balancing mechanism, frequency response, and new markets, e.g.
European wide balancing energy market TERRE [246].
• Predictive control: PyLESA incorporating a model predictive control strategy
allows analysis including optimal system operation. Particular advantages from
this are where dynamic time-of-use tariffs are connected in which case simple
rule-based controls would need to be constantly updated by an expert to minimise
import costs. Additionally, where on-site renewable production competes with
the electricity tariff, it is non-trivial to identify the low-cost periods. MPC enables
an optimal operation which ensures that current and future on-site generation and
electricity prices are included in the control decisions. An important aspect of
future work on the MPC strategy presented here is the inclusion of uncertainties
in the future predictions as currently these are modelled with perfect foresight
[247,248]. This overestimates the benefits of the MPC as in its present form it
always makes the right decision.
Further methods and assessments, and future developments are also discussed:
• Resource assessment methods: Resource data can be obtained via the MERRA
reanalysis dataset which provides a free, hourly dataset for a number of years. The
method provides a means of obtaining data easily, but it does not consider analysis
such as developing a typical meteorological year [249], or investigating resilience
to extreme [250], or worst case, weather conditions. This type of analysis would
typically be performed at the detailed design level but inclusion at the planning
stage would improve robustness.
• Electrical demand: This method relies on the use of a commercial software
which is not open source. While this ensures that the method is user-friendly and
reliable, it means that there is a cost and the underlying code is not accessible.
Implementing a similar solution with open source software would contribute
towards a fully open source software solution. However, demand profile
generation is not within the scope of the aims of this thesis, and it was decided to
use an existing method.
145 |
• Heat demand: The method used to simulate heat demand is built upon the
method used in the Biomass decision support tool [91] which produced a design
day demand profile. The new method used the same underlying pre-simulated
building models to produce an hourly heat demand over the year. Importantly,
this only requires simple inputs so is suitable at the planning stage.
The focus was on residential district heating schemes, therefore similar
methods could be developed for other building types, e.g. industry, commercial.
When applying diversity the moving average algorithm used could be modified to
reflect real data from monitored UK district heating schemes. Additionally, more
detailed modelling of the district heating network losses would allow better
comparison of the effect on system performance of changes in flow and return
temperature.
Only one heat demand can be input into the current model. Allowing input of
separate space heating and hot water demand profiles would allow for the
modelling of systems where there are separate hot water tanks and heat pump
modes (high/low temperature outputs). This is commonly the case for single
buildings.
Stochastic demand profile generators exist which can be used to account for
unpredictable influences on demand such as occupancy behaviour and could be
incorporated into this tool [213,251].
• Energy production technologies: Adding an electrical generator which

produces power from a fuel would allow for off-grid energy systems to be
simulated [252–254]. A number of additional renewable energy technologies
could also be added, e.g. tidal, wave, pumped hydro, solar thermal.
• Electrical storage: This method used a simple energetic model which is the most
often employed by other modelling tools. This can give an idea of how a generic
electrical storage technology may perform. However, more detailed models which
are technology specific could be included as future work.
• Thermal storage: The developed hot water tank model is a contribution to the
state of the art and is discussed above. Additional thermal storage technologies
could be included in future work such as Phase Change Materials (PCM)
146 |
[255,256], next generation smart hot water tanks [257], building thermal mass
[258], etc.
• Control strategies: The model predictive control strategy is a contribution to the
state of the art and is discussed above. The fixed order control strategy developed
is similar to the control algorithms used in existing modelling tools and is a useful
basis as a comparator to explore the performance benefits of the MPC. This will
be explored further in Chapter 6 where the two will be compared for a case study.
PyLESA has been written in an object-orientated manner which should allow for
future work to build upon it as a framework. This means that further control
strategies can be developed within the tool. These could be simple, rule-based
controllers which have special conditions, e.g. forced opportunistic charging from
on-site renewable power production, and it could be heuristic controls which are
data driven and adapt operation according to analysis of performance data [259].
• Modelling outputs: The set of KPIs developed allow for analysis of the
economic and technical performance of the modelled energy system. When the
tool is run an output file is saved which contains all the hourly output data of a
full range of parameters. This allows users to inspect the data and develop their
own KPIs which can be specific to their analysis.
• Uncertainty: PyLESA does not explicitly account for sources of uncertainty,
however it can be integrated within existing design methodologies which include
uncertainty analysis (see Section 1.9.). Integrating PyLESA with an uncertainty
analysis method such as Monte Carlo simulation would account for the inherent
uncertainty of input parameters, such as demand, and output design solutions
which are robust to these uncertainties. Scenario analysis is another approach
which, for example, could use worst-case (coldest) weather years, to ensure robust
solutions for uncertainties in weather. PyLESA can also be used in conjunction
with design optimisation methods which are used to search for an optimum
design solution for a given objective function and often incorporate uncertainty
[260].
The modelling methodology which describes the application of the developed
modelling tool PyLESA will be applied and validated further in the next chapter using a
case study of a housing cooperative. The case study currently consists of a biomass district
147 |
heating network and are looking to investigate the potential benefits from installing heat
pumps and thermal storage with predictive controls and alternative electricity tariffs.
148 |
6. Applying PyLESA: Proposed
Design and Sizing Study
This chapter explores the application of the developed modelling tool, PyLESA, for a
sizing study undertaken for a proposed design of a specific local energy system. The
primary aim from the perspective of the case study was to identify a low-cost and highly
renewable size combination of heat pump and hot water tank for three existing electricity
tariffs, a future wind-based tariff, and the developed control strategies. Additional aims
of the application were to showcase PyLESA as a useful tool to aid planning-level design;
and provide validation of the developed control strategies and workflow of the steps of
the tool.
In a peer-reviewed, conference paper the author performed a simple application of
PyLESA to inform a sizing study for a proposed design of a district heating scheme with
the fixed order controller and a time-of-use electricity tariff [261]. This chapter expands
on this work to compare the fixed order control and model predictive control as well as
four electricity tariffs.
The chapter is structured as follows:
• The aims, KPIs, and methodology of the sizing study are outlined.
• The existing system and proposed design are described.
• The steps of applying the modelling methodology are set out including the
required input data requirements.
• The modelling input requirements for the sizing study are provided.
• The results from PyLESA are presented for the performance of the model
predictive control and the fixed order control for all the modelled electricity
tariffs, along with a comparison of outputs to EnergyPLAN.
• Sizing results for the existing electricity tariffs and a future electricity tariff with
both controllers are discussed and graphically represented using 3D KPI plots.
149 |
The chapter concludes with a discussion which considers the results and the
performance differences for the range of electricity tariffs and control strategies, as well
as the overall applicability and usefulness of PyLESA in aiding the sizing study.
6.1. Study Aims, KPIs, and Methodology

The sizing study aims are split between two perspectives: informing design decisions for
the case study, and assessing the contribution of the developed modelling capabilities of
PyLESA to the state of the art.
From the case study perspective, the aim of the application of the developed
methodology was to aid the design and sizing of a low-cost and highly renewable local
energy system. A design consisting of an air-source heat pump and hot water tank system,
plus back-up electric heat, with a connection to on-site PV generation, participation in a
variable electricity tariff, and operation by a predictive control strategy, was proposed as
a solution to meet this aim.
The heat pump and hot water tank components of the proposed design require sizing
in order to enable the following load shifting mechanisms: increase on-site PV self-
consumption; take advantage of varying electricity costs under existing electricity tariffs;
and utilise low cost wind power under a future wind-based electricity tariff.
A set of KPIs (Table 6.1) are used in this sizing study to quantify the ability of the
proposed design to enable these load shifting mechanisms and allow for comparisons of
the technical and economic performance under the different control strategies and
electricity tariffs. The KPIs were chosen from Section 5.9. and the renewable-related KPIs
were adapted to suit this sizing study and provide clarity on the specific source of RES.
The LCOH was used as the KPI for choosing the optimal heat pump and hot water
tank size combination. LCOH acts as a cost metric and as a proxy for quantifying the
ability of the proposed design to enable the various load shifting mechanisms. Technical
renewable-related KPIs were also used to further explore the performance of the
proposed design with the different control strategies and electricity tariffs.
ORES has been adapted to ORESpv to clarify that the on-site PV generation is the
only form of on-site RES generation included in this study. HRES has been divided into
two adapted KPIs. HRESpv is defined to clarify that the on-site PV generation is the only
contributing RES generation source to this KPI and is used for the existing electricity
150 |
tariffs. HRESpv+windtariff is defined to clarify that both the on-site PV generation and
the electrical imports during high wind periods (as defined in the wind tariff, Section
6.4.5.) contribute to the RES in this KPI and is used solely for the wind electricity tariff.
These KPIs were chosen to illustrate the framework application and PyLESA tool
capabilities, other choices could be made in applications that best suit the situation.
Table 6.1: Set of KPIs for sizing study
KPI Comment Section

Levelized cost of heat
Same as in previous chapter 5.9.2.
(LCOH)
On-site RES used – PV Adapted ORES to represent the on-site PV
5.9.3.
(ORESpv) generation
Heat met from RES – PV Adapted HRES for the existing electricity
5.9.6.
(HRESpv) tariffs where on-site PV is only source of RES
Heat met from RES – Adapted HRES for the wind electricity tariff
PV+Windtariff where both on-site PV and electrical imports 5.9.6.
(HRESpv+windtariff) during high wind periods are classed as RES
A set of specific aims were developed in order to illustrate the use of the tool to
investigate various load shifting mechanisms and aid sizing decisions for the proposed
heat pump and thermal store design. These aims use the above KPIs as metrics to allow
comparisons and sizing decisions to be made. Specific aims are to:
• Investigate the performance of the developed control strategies, fixed order

control and model predictive control, with respect to their ability to enable the
various load shifting mechanisms.
• Investigate the use of existing electricity tariffs (flat rate, variable periods, and
time-of-use), particularly in relation to the proposed systems ability to take
advantage of variable electricity import costs.
• Explore the ability of the proposed system to utilise low cost wind power with
the use of a future wind-based electricity tariff.
• Identify an optimal LCOH heat pump and hot water tank size combination for
the different control strategies with both the existing tariffs and the wind tariff.
151 |
From the perspective of assessing the contributions of PyLESA to the state of the art
the following aims were set out:
• Review the position of the developed modelling tool in the context of the state of
the art of existing modelling tools.
• Validation of the developed control strategies.
• Demonstrate the function of the developed tool.
The methodology of the sizing study is framed to ensure that the above aims are
achieved. It is based around the concept of an existing local energy system investigating
the possibility of converting to a new proposed design. This methodology can be easily
adapted for designing from scratch.
The sizing study methodology reflects the structure of the rest of this chapter and
consists of the following steps:
1) Describe the case study and set out the existing local energy system.
2) Outline the design of the proposed local energy system.
3) Identify input data availability and output requirements.
4) Model the proposed local energy system with PyLESA using the modelling
methodology, including multiple runs for different size combinations of heat
pump and hot water tank. Rerun for all combinations of control strategy and
electricity tariffs.
5) Carry out a qualitative inspection of the operation to verify modelling and control
strategies, including a comparison to EnergyPLAN, and to compare and explore
the control strategies and electricity tariffs.
6) Tabulate the KPIs of the optimal heat pump and hot water tank sizing results for
each control strategy and electricity tariff combination.
7) Evaluate the output 3D plots of the KPIs for each control strategy and electricity
tariff combination.
The scope of this application is to illustrate the deterministic modelling capabilities of
PyLESA for heat pump and hot water tank sizing. It does not consider other system
component sizing and uncertainties in a wider optimisation and robustness analysis,
although PyLESA could be adapted to support such analysis in future.
152 |
6.2. Case Study Description
The sizing study is applied to a residential district heating scheme operated by West
Whitlawburn Housing Co-operative (WWHC) [262]. The scheme connects to 544 flats
and is supplied by a biomass boiler and backup gas boilers. WWHC are interested in
investigating the potential of transforming their existing assets into a low-cost and highly
renewable local energy system incorporating on-site PV, a heat pump, a hot water tank,
time-of-use electricity tariffs representing potential future power purchase arrangements
with local large scale wind generation (WWHC is adjacent to Whitelee wind farm, one of
the biggest in Europe), and predictive controls.
WWHC is a fully mutual, tenant owned and controlled Housing Co-operative with
charitable status, located in the south of Glasgow in Cambuslang, South Lanarkshire. It
is in an area of multiple deprivations and they hold an overarching aim to provide
affordable, sustainable, and community energy to the households. Previously, heat was
supplied via electric storage and panel heaters in the individual flats as gas heating could
not be installed due to regulations which apply to the multi-storey and low-rise tenement
buildings. A demand reduction initiative has seen these properties have fabric upgrades
to the buildings, windows, and roofs, in addition to the substantial external cladding
installed onto the multi-storey flats.
To tackle the problem of fuel poverty in the community alternatives to the inefficient
electric heaters were sought. Therefore, a biomass district heating system was installed
with a centralised energy centre supplying domestic instantaneous heat and hot water to
544 of the flats via a district heating network. A 740kW Viessman Pyrotec biomass boiler
operates as the primary heat source and is connected in parallel to a 50m3 hot water tank.
3x 1.2MW gas boilers are included to contribute to large peaks in demand during winter
and provide back up in the event of a breakdown or maintenance of the biomass boiler.
See Figure 6.3: Schematic of existing setup of the WWHC energy centre for a schematic
of the existing setup of the WWHC energy centre.
There are concerns around sustainability and air pollution issues related to the burning
of wood for domestic heating [12]. Additionally, biomass may have a pivotal role to play
in the wider energy system in decarbonising difficult sectors such as high-temperature
industry and heavy transportation [13]. From this holistic view of the wider energy system
it is worthwhile exploring alternative design options.
153 |
Figure 6.1: Panoramic view of WWHC with the 6 high rise towers and 5 low rise blocks - This image has been
removed by the author of this thesis for copyright reasons
Figure 6.2: WWHC hot water tank and flue chimney in foreground, and boiler house in background
154 |
Figure 6.3: Schematic of existing setup of the WWHC energy centre
155 |
6.3. Proposed Design
As an alternative to the current design at WWHC it is proposed that an air-source heat
pump and hot water tank system, plus back-up electric heat, with a connection to on-site
PV generation, participation in a time-of-use electricity tariff, and operation by a
predictive control strategy, can offer a solution for low-carbon and low-cost provision of
heat. This section discusses the motivation behind the choice of the various components
and ends with a schematic illustrating the proposed setup of the design.
The choice of heat source for a heat pump influences performance and installation
cost. Generally, ground source heat pumps have a high initial capital cost but offer lower
running costs because of improved COP due to the on average higher and more stable
temperature of the ground. Air source heat pumps (ASHP) have lower capital costs but
typically perform with a lower seasonal COP due to on average lower air temperatures.
A 400kW ASHP, using R134a as the refrigerant, is capable of providing hot water at 62°C
and has been installed along with a district heating network at Hillpark, Glasgow. This
connects to 351 homes which are similar to the residential scheme of WWHC [263].
Performance data for the COP and the maximum thermal output (duty) was readily
available thanks to links to the heat pump manufacturer, Star Refrigeration [264]. For
these reasons it seems apt to investigate the potential of installing a similar heat pump at
WWHC and modelling it using the available performance data. A back-up electric heater
is included as it fits with an all-electric design and simplifies the local energy system by
removing gas as an energy vector.
Thermal storage generally either stores sensible or latent heat. The primary technology
utilising latent heat are Phase Change Materials (PCM), and these can be expensive, are
primarily installed at a small-scale in homes, and they have the complication of being
associated with a fixed phase change temperature [256]. Hot water tanks are a widely used,
inexpensive example of a sensible storage technology [265]. Hot water tanks were chosen
as the investigated thermal storage for inclusion in the proposed design because it is
cheap, technically simple, allows temperature flexibility, and can be built in a wide range
of capacities. Additionally, WWHC already has a hot water tank to aid in the operation of
156 |
the biomass boiler so it is assumed that there is community acceptance and knowledge
regarding this technology.
On-site renewable generation possibilities are dependent on the resources available in
the local energy system. PV is a renewable power generation technology which is cheap
and widely installed in the UK. PV can connect to local electrical loads, such as the
residential electrical demand from the flats, to meet the electrical consumption needs of
the heat pump and auxiliary heater, and to the national grid to be exported. Therefore,
PV is chosen to be included on-site as part of the proposed system because it is cheap,
widely used, can contribute to both local electrical and heat demands, and can be exported
when not locally needed. PV has the limitation that while it supplies electricity in the
summer, its winter performance is limited. Hence, PV on its own is not an obvious fit
with winter space heating demand.
Renewable electrical power can also be sourced from at a range of scales through the
renewables connected to the local distribution network and through transmission via the
national grid, not solely from the on-site renewable power generation discussed above.
Wind is the predominant renewable electrical generation source on the Scottish networks
and transmission grid with excess capacity available being exported or curtailed [266]
(wind farm constraint payments in the UK have risen from £174,00 in 2010 to £139
million in 2019, with Scottish onshore wind farms receiving 94% of the 2019 payment
[267]). With further offshore and onshore capacity expansion planned there are expected
to be greater incentives to increase demand on high wind days. The use of this renewable
grid power will be explored in the proposed design indirectly through a representative
local wind influenced tariff rather than directly through on-site wind generation. In this
sizing study the primary KPI used to measure the ability of the proposed design to utilise
low cost wind power is HRESpv+windtariff. This KPI is the percentage of the heat
demand met by both the on-site PV generation and the electrical imports during high
wind periods.
Several electricity tariffs currently exist which incentivise shifting demand. Flat rate
tariffs are the most common domestic electricity tariff and do not change electricity prices
according to time of use. Variable periods tariffs such as day/night and weekday/weekend
tariffs incentivise users to shift to low-cost periods. Time-of-use pricing structures vary
prices in hourly, or sub-hourly, timesteps and can include facets such as dependence on
the wholesale market, premium price periods, etc. Future tariffs may emerge which are
157 |
based on inflexible, stochastic renewable generation and incentivise shifting electricity
consumption to periods of high renewable generation. An example is the wind tariff
discussed above. It is proposed that all these types of electricity tariff are included as
options in the proposed design so that modelling can differentiate which is the most
appropriate in terms of cost and integration with renewable generation.
Both a traditional fixed order controller and a model predictive controller would be
useful to include in the proposed design to compare the performance of each. Benefits
from simple tariffs may be maximised using only simple controls, while more complex
designs may require the optimisation element from model predictive control. These
controls are modelled with a view to maximising the use of on-site renewable generation
and minimising overall electricity import costs.
Figure 6.4 is a schematic which illustrates the proposed setup, combining all the
components discussed above. The system provides both space heating and hot water
through a district heating network. Note that the design study carried out here does not
size the buffer section of the hot water tank which is required for safe operation of the
heat source but instead focuses solely on sizing the thermal store section which enables
load shifting. On the diagram, red lines indicate communication between component and
controller, and not shown is the grid connection which allows import and export priced
by the electricity tariffs.
158 |
Figure 6.4: Schematic of proposed design
159 |
6.4. Input Data and Output Requirements
This section details each component of the proposed setup outlining the required and
available input data, and ending with the modelling output requirements. The structure
follows the description of the underlying models in Chapter 5.
The year 2017 is used as the reference year and subsequently the collected data is
applicable to this year.
6.4.1. Resource and Demand Assessment and Input Methods

• Local resources: The only available local resource data was air temperature,
which was collected on-site using local sensors with 2017 data available. Wind
speed data was collected from the MERRA reanalysis hourly dataset for 2017 for
wind speed (at 10m height) due to the lack of available on-site data. A simple
multiplication correction factor of 0.67 is applied to this data, in accordance with
a study which shows a 50% overestimation of the MERRA wind speed for
northwest European countries [203]. The uncertainty introduced by using the
MERRA could be further reduced by using the correction factors as calculated
specifically for the UK [203] or by using more reliable data, such as from a weather
station.
• Electrical Demand: Due to a lack of monitoring data a generic community
electrical demand profile was synthesised in HOMER.
• District Heating Demand: Hourly monitored data was available for the district
heating demand for WWHC including for the year 2017.
6.4.2. Electrical Production Technologies

• PV: 1.74MW rated capacity, 6000 x 290W LG LG290N1C-G3 [2013] panels,
south-facing, with 40° surface tilt and LG295A1C-B3 [240V] 240V [CEC 2018]
inverters. This capacity was chosen to ensure that there is a large opportunity for
utilising excess PV generation with the heat pump and hot water tank. There will
be periods where the 1.74MW PV generation exceeds electrical demand, which
has a peak of 650kW. Additionally, there will be periods where the excess PV
generation can be used to meet heat demand, which has a peak of 1.36MW. Due
160 |
to uncertainties in the future of government incentives for PV, these are not to
be included in this modelling exercise.
6.4.3. Heat Pumps and Auxiliary Heat Units

• Heat pump: Star Refrigeration ASHP Neatpump [264] with variable speed
compressor and 65/55 flow/return temperatures feeding a district heating
network at 60/40 flow/return temperatures. The heat pump and district heating
temperatures are the same as used for an existing ASHP and district heating
scheme in Scotland, with similar building characteristics as WWHC. It is assumed
that the heat exchangers in the flats at WWHC can be adapted to work at these
temperatures. This type of ASHP was chosen because it is state of the art and the
performance curves were available. Capital costs were assumed linear at £600/kW
as given by the Danish Energy Agency [268]. Due to uncertainties in the future
of government incentives for heat pumps, these are not to be included in this
modelling exercise.
• Electric heater: Backup electric heater with 100% efficiency sized to peak heat
demand. This capacity was chosen to ensure that the electric heater can act as a
back up to the heat pump and always meet heat demand.
6.4.4. Hot Water Tank

• Hot water tank: Modelled using the following inputs: 5 nodes, polyurethane
insulation, located outside, 5 thermostat tank openings with diameter 35mm, and
2 insulated connections for the flow and return with diameter 50.8mm. Capital
cost is presumed to follow an exponential decay function for £/m3 [268], Figure
6.5. This configuration is typical of hot water tanks used in current district heating
schemes. 5 nodes were chosen to balance the accuracy in modelling tank
stratification and computational time.
161 |
Figure 6.5: Exponential function of hot water tank volume and specific price
6.4.5. Electricity Tariffs

4 different electricity tariffs are modelled in the sizing study using the following inputs.
The prices used for the different tariffs are representative of typical prices available from
domestic energy suppliers in 2017. Exports have been set to zero value in order to increase
the incentive for the system to self-consume. Additionally, while currently export prices
can be agreed and sold to an energy supplier, there is uncertainty as to the future value of
exports from on-site generation such as PV.
• Flat rate: £130/MWh.
• Variable periods: 12am to 7am - £75/MWh, 7am to 12am - £150/MWh.
• Time-of-use: Tracks wholesale market (2017 prices) with £120/MWh premium
from 4pm to 7pm and £350/MWh maximum. This mimics the currently available
Octopus Agile tariff [30].
• Wind: Combination of variable periods and wind pricing structures. 12am to 7am
- £75/MWh, 7am to 12am - £150/MWh and a £50/MWh discount during top
20% of wind output and a £50/MWh premium during bottom 20% of wind
output. Wind output is based upon Whitelee Wind Farm, which consists of 215x
Siemens SWT-2.3MW.
162 |
6.4.6. Fixed Order Control
The fixed order controller requires an import setpoint. For the flat rate electricity tariff
the import setpoint was set below the import cost to avoid unnecessary charging and
discharging of the hot water tank using high-cost grid imports. For the variable periods
electricity tariff the import setpoint was set between the day and night import costs to
enable load shifting from day to night. For the time-of-use and wind electricity tariff the
import setpoint was set to £100/MWh.
6.4.7. Model Predictive Control

MPC only requires the prediction horizon as an input. For the existing electricity tariffs a
24-hour period was used and for the wind tariff a 168-hour (1 week) prediction horizon
was used. The existing tariffs used follow a 24-hour pattern because they generally follow
demand, therefore a prediction horizon of 24 hours captures general price fluctuations.
The wind tariff is variable over a longer timescale as it is dependent on periods of high or
low wind, therefore a 1 week prediction horizon to take advantage of longer term
variations.
6.4.8. Output Requirements

Economic and technical outputs are required to inform comparisons of the control
strategy and electricity tariff combinations. Additionally, sizing the heat pump and hot
water tank components require outputs which can quantify the ability of the proposed
design to perform the following load shifting mechanisms: increase on-site PV self-
consumption; take advantage of varying electricity costs under various existing electricity
tariffs; and utilise low cost wind power under a future wind-based electricity tariff.
A set of KPIs to be used in this sizing study was described in Section 6.1. and displayed
in Table 6.1. LCOH was the economic metric chosen as it can act as a proxy for
quantifying the ability of the proposed design to enable the various load shifting
mechanisms. Technical renewable-related KPIs (ORESpv, HRESpv, and
HRESpv+windtariff) were also used to further explore the performance of the proposed
design with the different control strategies and electricity tariffs. These KPIs can be
calculated from economic and renewable-related outputs, and are directly calculated with
PyLESA.
163 |
6.5. Application of PyLESA Modelling
Methodology
This section describes applying the developed modelling methodology, from Chapter 4,
which sets out the steps for running PyLESA, for the proposed design of WWHC. The
aim is to size a heat pump and a hot water tank as part of a low-cost and highly renewable
local energy system. The following are the methodology steps including the details relating
to this specific application.
1. Define the local energy system to be modelled:
The proposed design for WWHC was defined in Section 6.3, and the input data
available outlined in Section 6.4.
2. Optionally run the demand and resource assessment methods:
Resource method and electrical demand methods were run as monitored data was
not available. However, not needed for the heat demand as monitored data was
available.
3. Input gathered data to PyLESA:
Take data from first two steps and input to PyLESA using the Excel workbook.
4. Input ranges to be modelled for parametric analysis:
• Fixed Order Control and Model Predictive Control with existing tariffs:
o Hot water tank capacity range: 0 -> 800m3 in 100m3 steps.
o Heat pump thermal output capacity: 0 -> 2000kW in 250kW steps.
• Fixed Order Control with wind tariff:
• Model Predictive Control with wind tariff:
5. Run PyLESA:
• Use the run.py script inputting the appropriate input Excel workbook name.
• Rerun the tool for all combinations of control strategy (fixed order and model
predictive control) and electrical tariffs (flat rate, variable periods, time-of-use, and
wind).
164 |
6. Analyse the outputs:
• In order to illustrate (and further validate) the system operation in detail, an
operation analysis snapshot is provided allowing inspection of four example week
operational graphs consisting of (i) heat pump thermal output, auxiliary electric
heat output, and heat demand, (ii) hot water tank node temperatures, (iii) import
costs, and (iv) surplus and exports.
• To showcase how PyLESA can support component sizing and compare control
strategies and electricity tariffs, KPIs for the optimal LCOH heat pump and hot
water tank size combinations with the range of control and tariffs are tabulated.
These KPIs are LCOH, ORESpv, HRESpv, and HRESpv+windtariff. See
Section 6.1. for further details on these.
• To explore the ability of the proposed design in enabling the various load shifting
mechanisms, the same set of KPIs are presented for the modelled ranges of heat
pump and hot water tank sizes using 3D graphs.
165 |
6.6. Outputs from PyLESA
Running PyLESA produces several graphical and numerical outputs which can inform
planning-level design decisions such as those encountered in this sizing study. This section
gives a brief overview of example of outputs from PyLESA to inform the reader the type
of analysis which is readily available after running the tool. The next two sections will then
present the operational analysis results and sizing results for the sizing study.
To provide an illustrative example of PyLESA outputs, the results for a possible
WWHC design option is presented here, with a 1000kW heat pump and 500m3 hot water
tank size combination, a variable periods tariff and MPC.
Raw data containing the hourly, sub-data of the underlying outputs is made available
in the outputs.pkl file. This is saved at the end of running the tool and can be used to
perform external analysis. This is the file which the outputs.py script uses to perform the
output analysis included in PyLESA to produce numerical and graphical outputs.
The output graphical plots are a mixture of bar charts, 2D plots, and 3D plots. The
2D plots are generated for a summer week, winter week, and the year in hourly timesteps.
If the user has defined a modelling period shorter than a year then the plots are generated
for the user-defined period.
Key performance indicators (KPIs) have been developed within PyLESA and are
described in Section 5.9. LCOH was the economic metric chosen as it can act as a proxy
for quantifying the ability of the proposed design to enable the various load shifting
mechanisms. Technical renewable-related KPIs (ORESpv, HRESpv, and
HRESpv+windtariff) were also used to further explore the performance of the proposed
design with the different control strategies and electricity tariffs. These KPIs are used in
Section 6.8. where 3D plots illustrate the sizing decisions and performance of the
electricity tariffs and control strategies investigated.
The total PV, wind turbines and combined renewable power production (Figure 6.6)
are output as bar charts which display the gross power production for each month, along
with a heading which contains the annual power production for the modelled year. This
example does not include on-site wind power production and hence the wind power
production is zero, this is included here because the outputs described in this section
follow a generic template and are designed to be useful for a wide variety of modelling
exercises.
166 |
Figure 6.6: Total PV, wind turbines and combined renewable power production in bar charts
Electrical demand, renewable generation, and renewable generation usage plots are
included together in Figure 6.7 for a summer week as this season has the highest PV
production. The electrical demand plot allows the modeller to quickly assess which supply
options are meeting electrical demand from imports, renewable generation (RES), and/or
electrical storage (ES). The renewable generation plot displays, on an hourly scale, the
wind and PV power generation and is included to compliment the above bar charts. The
renewable generation usage plot shows how the PV and wind power generation is being
used within the local energy system. In the example figure the electrical demand uses a
baseload proportion of the generation, while the heat pump and auxiliary electric heater
consume most of the remainder. Little of the power is exported showing that, for this
week, the system has enough flexibility built in to self-consume almost all of the on-site
renewable power generation.
167 |
Figure 6.7: Electrical demand, renewable generation, and renewable generation usage plots for summer week
The grid interaction plots, Figure 6.8, shows the flow of imports and exports along
with the financials of the import cost and cashflow for a summer week. As was seen from
the renewable power usage plot, little power is exported from the local energy system for
the example summer week. The variable periods pricing structure can be readily viewed,
along with the cashflow which is useful as a proxy metric of viewing the combined import
and export interaction. In this example exports have no value which explains why the
short period of export does not result in a positive cashflow.
Two figures are output to explore the modelling results of the heat pump. Figure 6.9
shows the heat demand, heat pump output, and heat pump electrical usage. Additionally,
Figure 6.10 displays the variation in heat pump performance in terms of the COP and
maximum thermal output. The heat demand plot shows when the heat demand is met by
the heat pump, the hot water tank discharging, and the auxiliary heat. It can be seen by
viewing the heat demand and heat pump thermal output plots together that, in this
example, load shifting is occurring as there are distinct periods where the heat pump
thermal output is charging the hot water tank, the heat pump turns off, and then the hot
168 |
Figure 6.8: Grid interaction plot for summer week
water tank meets the heat demand. The heat pump is utilising excess power generation
from the PV to charge the hot water tank to meet the heat demand in a later period, where
the PV is not producing.
In this analysis the usage of excess PV generation is assigned a zero-marginal cost and
exports zero value. In this case the MPC makes no distinction in the electrical cost
between running the heat pump or auxiliary electric heater. The MPC does include the
advantage that using the heat pump is a more efficient conversion of the surplus electricity
into heat than using the auxiliary electric heat. However, there are periods where there is
no advantage from the more efficient use of the surplus, such as where the hot water tank
can be filled, or the hot water tank charged sufficiently to cover the heat demand until the
next PV surplus, by either the heat pump or auxiliary electric heat. This is seen in the
periods where the heat pump turns off and the electric heater meets demand and charges
storage using surplus PV generation, and discharging the hot water tank is then capable
of meeting the heat demand until the next PV surplus. See Section 6.8.2.2. for further
discussion of the operation of this example.
169 |
The hot water tank plots combine the calculated energy flow of charging and
discharging along with the modelled node temperature variations, Figure 6.11. The top
and middle plots show the charging and discharging over a summer week, and these
coincide with the heat pump and auxiliary electric heater utilising the hot water tank to
shift load in order to maximise use of on-site PV power generation. The bottom plot
shows the fluctuation of the temperature of the nodes, which matches the periods of
charging and discharging. In the summer the heat demand is substantially lower than in
the winter, and this leads to an underutilisation of the hot water tank in the summer. This
is seen in this plot as the temperatures of the lower nodes are consistently at lower
temperatures than the maximum, which means that effectively only a fraction of the hot
water tank capacity is being used.
170 |
Figure 6.9: Heat demand, heat pump output, and heat pump electrical usage plots for summer week
Figure 6.10: Heat pump performance with COP and duty
171 |
Figure 6.11: Thermal storage charging and discharging as energy, and node temperatures for summer week
Thus far, the graphical outputs described have generally focussed on individual
components of the local energy system. However, to understand the complex interactions
between the supply, demand, and storage components it is useful to view these plots side
by side in a figure. Operational graphs are output to help understand the relationships
between the system components, and consist of four plots: heat pump output, auxiliary,
and heat demand; hot water tank node temperatures; import cost; and, surplus and export.
These are explored fully for the different control strategies and electricity tariffs in the
Section 6.7.
172 |
6.6.1. Comparison to EnergyPLAN
The aim of this section is to compare the outputs from a widely used planning-level
modelling tool, EnergyPLAN, to PyLESA, to help build confidence in the modelling
capabilities of PyLESA. This was done by comparing a sample of outputs from both tools
and explaining differences in the output due to the different hot water tank modelling and
control approaches. EnergyPLAN was chosen because it is widely used, can model all the
required components, is free to download and use, and the author has user expertise. The
scope of this section is not a detailed analysis of EnergyPLAN, which is one of the tools
reviewed in detail in Chapter 3, or PyLESA, where the outputs are explored in detail
elsewhere in this chapter. The purpose of this section is to check if the outputs from both
tools are similar, and that any differences can be explained by the different functionality.
Given the widespread usage of EnergyPLAN, it is assumed that if PyLESA can produce
similar outputs then it is producing reasonable outputs. This can then help build
confidence in the modelling capabilities of PyLESA.
Both tools were used to model the proposed design of WWHC (described in Section
6.3.), with a 1000kW heat pump and 500m3 hot water tank size combination, a variable
periods tariff and MPC. This size combination was chosen as it has large enough
capacities to use the heat pump and hot water tank to load shift to both utilise excess PV
generation and to take advantage of the day/night electrical import cost timings of the
variable periods tariff.
The inputs to EnergyPLAN are described here:
• The electricity and heat demands are input using the annual demand and distributions
of the hourly variation. For the heat demand monitored data was input, and for the
electricity demand the synthesised profile from HOMER was input.
• PV generation is input as a capacity and a distribution profile. The solar radiation

global horizontal incident from the MERRA reanalysis dataset was input as the
distribution profile, while the capacity was adjusted to match the annual power
production as modelled using PVLIB.
• The heat pump is input as a fixed electrical capacity and COP. This was input as
417kW to obtain a 1000kW heat output capacity using a COP of 2.4. This is the
seasonal COP over the year as calculated by PyLESA, and is used because
173 |
EnergyPLAN can only accept a fixed COP. A 1000kW auxiliary electric boiler is also
added for backup and peak demand periods.
• For the hot water tank, the energetic capacity is calculated using the specific heat
formula, where the mass is 500,000kg and the delta T is 25°C (from 65°C heat pump
flow and 40°C district heating return temperatures). The energetic capacity is
calculated and input as 14.5MWh.
• The variable periods tariff is input as a timeseries distribution, with day prices of
£150/MWh and night prices of £75/MWh.
• The technical control strategy selected balances heat and electrical demands, while
maximising self-consumption of excess electrical generation. A sample of the available
control strategies were tested, and this one was found to minimise operational costs.
Table 6.2 shows the imports, percentage of PV exported, import costs, and LCOE
are similar for both tools. This suggests that the outputs from PyLESA are in broad
agreement with the outputs from the widely used planning-level tool, EnergyPLAN, and
therefore, producing outputs which are reasonable. However, the outputs are not exactly
the same, and this is due to the differences in the way PyLESA and EnergyPLAN models
and controls the hot water tank. These differences are discussed in the rest of this section.
Table 6.2: Outputs from application of PyLESA and EnergyPLAN
Output (over a year) PyLESA EnergyPLAN

Imports (MWh) 2011 2000
Percentage of PV generation
4.5 9.3
exported (%)
Import costs (£k) 247 273
LCOE (p/kWh) 5.8 6.3
The tools use different approaches for modelling the hot water tank. PyLESA uses
the multi-node model which accounts for the evolution of temperature throughout the
tank, and for losses to both the environment and mixing. EnergyPLAN uses an energetic
model which does not account for any losses and does not incorporate temperatures of
the tank. This contributes to the higher imports and lower percentage of PV generation
exported (increased PV self-consumption) as output by PyLESA compared to
174 |
EnergyPLAN, as there is a greater energy requirement for meeting demands plus the
losses.
There are also differences in the approaches of the tools to the control of the hot
water tank. In this example, PyLESA has been applied with MPC which optimises the
operation of the hot water tank over a 24-hour period in order to minimise electricity
costs. This means that the hot water tank is used to flexibly operate the heat pump and
auxiliary electric heater in order to meet demand by using excess PV generation and
avoiding the high-cost periods of the variable tariff. EnergyPLAN controls the hot water
tank using a technical strategy which stores excess PV generation whenever possible and
is independent of fluctuations of prices of the variable tariff.
The control logic used by EnergyPLAN results in periods where the storage reaches
capacity and forces excess PV generation to be exported. PyLESA operates to maximise
the use of the excess PV generation, as it is considered zero marginal cost electricity.
Additionally, it also occasionally uses the auxiliary electric heater to use or store excess
PV generation. This is because operation is minimised for cost and does not distinguish
between the greater energy utilisation possible by using the heat pump over the auxiliary
electric heater. The optimisation strategy, use of the auxiliary electric heater, and losses
from the hot water tank all contribute to a lower percentage of the PV generation which
is exported in the outputs from PyLESA compared to EnergyPLAN.
EnergyPLAN does not attempt to shift electricity imports to low-cost periods, in
contrast to PyLESA which accounts for both PV generation and the variability of the
electricity tariff when operating the heat pump and hot water tank. This contributes to
lower import costs and LCOE from PyLESA compared to EnergyPLAN. This is despite
the higher energy demand for PyLESA due to incorporating hot water tank losses, which
will contribute to increasing import costs and LCOE for PyLESA.
This analysis does not explore the extent to which the different approaches change
the outputs, and this can be done as future work to complete a more detailed comparison
of the two tools. However, the above discussion of the effects of the different approaches
to control and modelling of the hot water tank explains the reasons for the differences
between the outputs.
PyLESA could be configured to imitate the control logic of EnergyPLAN by setting
losses to zero and utilising a significantly longer prediction horizon. However, this
approach is not realistic as keeping the hot water tank hot over periods of days, as
175 |
EnergyPLAN does in periods of high PV generation, will result in higher heat losses. Not
modelling these heat losses results in a lower overall energy demand than would be
expected with a system containing a hot water tank. This highlights the advantage of using
PyLESA with its more realistic hot water tank modelling and control. Additionally,
EnergyPLAN cannot be configured to imitate the more realistic control used by PyLESA.
This highlights the greater user flexibility of PyLESA over EnergyPLAN.
In conclusion, the similar outputs show that PyLESA produces reasonable outputs,
while the differences in the outputs can be explained by the more realistic approach taken
to the modelling and control of hot water tanks in PyLESA than in EnergyPLAN.
176 |
6.7. Operational Analysis Results
This section contains the results and discussion of the operational analysis for all the
modelled combinations of control strategies and electricity tariffs. The algorithms and
decision making behind the controllers will be discussed by examination of the behaviour
of the supply and storage components as they meet demand. This contributes to two aims
of the sizing study, (i) to compare the performance of the control strategies and, (ii) to
compare the performance of the electricity tariffs.
This section also contributes to another aim of the sizing study which was to validate
the control strategies by inspection. This is crucial as the development of these constitute
a contribution made to the state of the art. Additionally, the analysis in this section helps
validate the workflow of PyLESA; ensuring that the underlying models work together and
can produce outputs which are logical and realistic.
The operational graphs presented here consist of four plots: heat pump output,
auxiliary, and heat demand; hot water tank node temperatures; import cost; and, surplus
and export. A summer week and winter week are shown for most tariff and control
combinations, with a windless winter week shown for the wind tariff with MPC.
The following key applies to all the operational graphs: HPt – heat pump heat output,
aux – auxiliary heat output, and HD – heat demand.
The case study for the control strategy and electricity tariff combinations use a
1000kW heat pump and 500m3 hot water tank size combination as the illustrative example
to carry out the analysis, except those with the wind tariff where a 3000kW heat pump
and 3000m3 hot water tank combination was used. A different size combination is chosen
for the wind tariff to explore the ability of a larger sized system to take advantage of the
larger price differentials on a longer timescale.
177 |
6.7.1.1. Flat Rate Tariff
The operation of the fixed order control with the flat rate tariff is displayed for a summer
week in Figure 6.12 and a winter week in Figure 6.13. The main load shifting mechanism
on display is shifting heat pump electricity consumption to match excess PV generation
which charges the hot water tank to meet heat demand during periods of little or no PV
generation. Since the flat rate tariff does not vary with time (3rd plot) there is no cost
benefit for shifting heat pump operation according to import cost.
Figure 6.12: Operational graphs with FOC and flat rate tariff over a summer week
The heat pump capacity, stated as 1000kW, is the rated capacity under rated operating
conditions. The heat pump output exceeds 1000kW in the 1st plot because it is operating
in summer conditions which are warmer than the rated conditions which results in a high
heat pump maximum output.
In the first few hours of the plotted period, the heat pump is meeting demand (1st
plot) from surplus PV generation (4th plot) and charging the storage (2nd plot). When there
is no longer surplus PV generation the heat pump turns off and the hot water tank
discharges to meet demand. Then, the hot water tank becomes depleted, the top node
178 |
drops below the required flow temperature of the district heating network, and the heat
pump modulates output to meet heat demand. Surplus PV generation is available again
and the heat pump operates at maximum output to meet demand and charge storage. The
controller also chooses to turn on the auxiliary electric heater to aid charging of the hot
water tank in the periods where relying on the heat pump using surplus PV is insufficient
to cover the heat demand. After the PV is no longer producing a surplus, the hot water
tank discharges to meet the heat demand. Over the remainder of the plotted week a similar
pattern as described is repeated. Excess PV generation is utilised by the heat pump and
electrical heater to meet demand and charge the hot water tank. This then discharges to
meet demand where there is no excess PV generation.
During the winter there is little surplus PV generation and therefore, few opportunities
to shift load using the hot water tank. Figure 6.13 shows the heat pump modulating output
to meet heat demand (1st plot), negligible hot water tank charging/discharging (2nd plot),
and small peaks of surplus generation (4th plot) which will be used by the heat pump to
directly meet the heat demand.
Figure 6.13: Operational graphs with FOC and flat rate tariff over a winter week
179 |
6.7.1.2. Variable Periods Tariff
The operation of the fixed order control with the variable periods tariff is displayed for a
summer week in Figure 6.14 and a winter week in Figure 6.15. Load shifting occurs in this
example for both utilising excess PV generation and avoiding importing during high-cost
electricity tariff periods during the day (3rd plot).
Figure 6.14: Operational graphs with FOC and variable periods tariff over a summer week
At the start of the displayed summer week, the heat pump is meeting demand from
surplus PV generation (4th plot). When there is no longer surplus PV generation the heat
pump turns off and the hot water tank discharges to meet demand (2nd plot). When the
import cost drops from £150/MWh to £75/MWh (3rd plot) the heat pump output is
turned up to the maximum output and simultaneously meets the heat demand and charges
the hot water tank. During the low-cost period, when the hot water tank is full, the heat
pump modulates its output to match demand. During the high-cost period the hot water
tank discharges and the heat pump turns off unless there is surplus PV generation, in
which case the heat pump meets demand and charges the hot water tank. However,
180 |
because the hot water tank has only briefly been discharging after fully charging from the
low-cost overnight period, there is little spare capacity to utilise the surplus PV generation.
During the winter, there is little surplus PV generation and the primary load shifting
mechanism is moving demand from day to night. The higher demand in the winter means
that the heat pump operates at maximum output during the low-cost period to meet
demand and charge the hot water tank. In the displayed example, at the start of the high-
cost period, the hot water tank discharges to meet demand but there is not enough stored
heat to meet the entire demand during the high-cost period. When the hot water tank is
empty the heat pump modulates its output to meet heat demand.
Figure 6.15: Operational graphs with FOC and variable periods tariff over a winter week
181 |
6.7.1.3. Time-of-use Tariff
The operation of the fixed order control with the time-of-use tariff is displayed for a
summer week in Figure 6.16 and a winter week in Figure 6.17. Load shifting occurs mainly
to avoid importing during the premium period which occurs each day between 4pm and
7pm (3rd plot), and due to the lack of sophistication of the fixed order control the surplus
PV generation is largely exported.
Figure 6.16: Operational graphs with FOC and time-of-use tariff over a summer week
The heat pump operation (1st plot) generally follows the heat demand but spikes to
charge the hot water tank immediately after the premium price period. The hot water (2nd
plot) stays fully charged for long periods and discharges to meet heat demand during the
premium price period. While there are periods of surplus PV generation, this is largely
exported because of the lengthy periods where the hot water tank is fully charged.
The size of the hot water tank is also sufficient in the winter week to cover the larger
heat demand during the premium period.
182 |
Figure 6.17: Operational graphs with FOC and time-of-use tariff over a winter week
183 |
6.7.1.4. Wind Tariff
The operation of the fixed order control with the wind tariff is displayed for a summer
week in Figure 6.18 and for a winter week in Figure 6.19.
The larger size of heat pump allows the demand to be met during the low-cost periods
where there is a large amount of wind on the grid. The large hot water tank meets the low
summer demand for long periods using surplus PV generation and cheap wind tariff
imports.
Figure 6.18: Operational graphs with FOC and wind tariff over a summer week
Load shifting occurs mainly to avoid importing during the premium period which
occurs each day between 4pm and 7pm (3rd plot), and due to the lack of sophistication
of the fixed order control the surplus PV generation is largely exported.
Again, in the winter week the large heat pump and hot water tank are operated to take
advantage of the low-cost periods. This shows the potential of coupling large hot water
tanks with this type of future electricity tariff which has large price differentials over long
periods.
184 |
Figure 6.19: Operational graphs with FOC and wind tariff over a winter week
185 |
The operation of the MPC with the flat rate tariff is displayed for a summer week in Figure
6.20 and a winter week in Figure 6.21. The main load shifting mechanism is moving heat
pump electricity consumption to match excess PV generation which charges the hot water
tank to meet heat demand during periods of little or no PV generation. Since the flat rate
tariff is constant (3rd plot), shifting heat pump operation according to import cost does
not happen.
Figure 6.20: Operational graphs with MPC and flat rate tariff over a summer week
In the first hours of the plotted period, the heat pump is running on surplus PV
generation, charging the hot water tank and meeting the heat demand. Then the surplus
PV generation goes to zero and the hot water tank is discharged to meet demand, until it
is fully discharged, and the heat pump turns back on to meet demand. When surplus PV
generation later returns, it is used by both the heat pump and electric heater to meet
demand and charge storage.
The heat pump only operates in the plotted period when there is surplus PV
generation. The MPC optimises the operation based on minimising operational costs.
186 |
Since the usage of excess PV generation is assigned a zero-marginal cost, the algorithm
makes no distinction in cost between running the heat pump or auxiliary electric heater.
This is seen in periods where the heat pump turns off and the electric heater meets
demand.
In the winter period, where there excess PV generation is negligible, the heat pump is
primarily controlled to meet demand and charge storage in periods where the COP is
highest. This results in a high number of heat pump cycling as the controller optimises
the performance.
Figure 6.21: Operational graphs with MPC and flat rate tariff over a winter week
187 |
The operation of the MPC with the variable periods tariff is displayed for a summer week
in Figure 6.22 and a winter week in Figure 6.23.
In the summer week, the system shifts heat pump and electrical heater usage to match
the surplus PV generation. The operation manages to cover most of the heat demand
using the surplus. This means that in the summer week shown, there is little interaction
with the variable periods tariff. Between the 25th and 50th hour of the summer week the
heat pump is turned on at the end of the low-cost period and charges the hot water tank
to avoid a proportion of the consumption during the high-cost period, despite the excess
of PV generation and spare hot water tank capacity in the 26th hour. The implemented
MPC uses a simplified energetic model in the optimisation problem, which leads to an
overestimation of the use of the hot water tank. In the illustrated period, the hot water
tanks ability to meet demand from the 26th hour is overestimated, and this shortfall met
at the end of the low-cost period. This shows a limitation of the implemented MPC in
PyLESA.
Figure 6.22: Operational graphs with MPC and variable periods tariff over a summer week
188 |
In the winter period the surplus PV generation is negligible and the MPC instead
maximises the heat pump output during the night to shift as much load as possible from
the high-cost period to the low-cost period. There is insufficient heat pump capacity to
charge the hot water tank to cover the entire high-cost period. During the day, the heat
pump charges the hot water tank during higher COP periods to improve overall
performance, and ultimately minimise the operational cost.
Figure 6.23: Operational graphs with MPC and variable periods tariff over a winter week
189 |
The operation of the MPC with the time-of-use tariff is displayed for a summer week in
Figure 6.24 and a winter week in Figure 6.25. The time-of-use tariff is variable hourly
throughout the day and the MPC should be adept at ensuring that electrical consumption
coincides with the lowest cost periods.
Figure 6.24: Operational graphs with MPC and time-of-use tariff over a summer week
In the summer week, operation is similar to the MPC and variable periods
combination. The MPC successfully shifts most of the electrical consumption to match
the surplus PV generation, while utilising the lowest cost periods of the time-of-use tariff
to meet the remaining heat demand.
In the winter period, the surplus PV generation is negligible and the MPC instead
optimises the heat pump output to match the lowest cost periods of the time-of-use
tariffs, as well as incorporating the effect of outdoor temperature on the COP. This results
in a complex operation of the heat pump cycling on and off throughout the day,
responding to fluctuations in electricity prices and ambient temperature.
190 |
Figure 6.25: Operational graphs with MPC and time-of-use tariff over a winter week
191 |
The operation of the MPC with the wind tariff is displayed for a windless, winter 10-day
period. The MPC is modified to a 168-hour prediction horizon when modelling using the
wind-based tariff with a 3000kW heat pump and 3000m3 hot water tank. The increase in
prediction horizon increases computational time but allows for operation to be optimised
during windless periods.
Figure 6.26: Operational graphs with MPC and wind tariff over a windless winter 10-day period
.
In the first 50 hours there are periods of high wind resulting in low cost, and it is
during these periods the heat pump operates at maximum output to fill storage and meet
demand. Additionally, the auxiliary electric heat turns on because the direct electric heat
is cheaper in these periods than operating the heat pump in the high-cost periods. The
hot water tank is then used to cover a large proportion of the high-cost period, as can be
seen by the trend of reducing node temperatures. However, there is not enough capacity
to cover this entire period and the heat pump occasionally operates to charge the hot
water tank. This will occur during the highest heat pump performance periods which are
when the air temperature is highest.
192 |
6.7.3. Discussion
The operational analysis results allow discussions on the performance of the developed
control strategies and the existing electricity tariffs, as well as exploring the use of a future
wind-based electricity tariff. Additionally, this close examination of the operation ensures
that the control strategies are operating as expected and that the workflow connecting the
underlying models of PyLESA can produce logical and useful outputs.
Flat rate tariff:
• The fixed order control enables load shifting to increase the self-consumption
from on-site PV generation. This is because load shifting heat pump and electric
heater power consumption for surplus PV generation is the only avenue of
reducing costs. The flat rate tariff offers no incentive for shifting load.
• Consequently, qualitative analysis of the operation does not reveal an advantage
of using MPC as it performs the same function – increasing utilisation of surplus
PV generation.
Variable periods tariff:
• The fixed order control attempts to load shift for both increasing use of surplus
PV generation and avoiding high-cost tariff periods. However, the hot water tank
is often fully charged at the end of the low-cost period and just before there are
surpluses of PV generation. This results in a significant portion of PV generation
being exported, despite this having zero value to the local energy system in this
case study. Additionally, keeping the hot water tank at a high temperature for long
periods will incur higher tank losses.
• MPC allows the operation to optimise for both avoiding high-cost periods and
utilising PV with the variable periods tariff. This results in a much higher
utilisation of surplus PV generation, and lower import costs as a result of only
charging the hot water tank at low-cost periods in order to cover the remaining
heat demand.
Time-of-use tariff:
• This is a highly variable tariff outside the premium period meaning that using the
fixed order control is limited to avoiding imports during the premium period. The
fixed order control is not suited to avoid the premium period and load shift based
on price variations outside of the premium.
193 |
• MPC makes it possible to avoid premium prices and take advantage of the other
variable prices. Additionally, similar to the variable periods tariff, the PV
utilisation can be maximised and imports limited to the lowest cost periods and
restricted to only meeting the remainder of the demand in the calculation period.
Wind tariff:
• Using the fixed order control with the wind tariff shows great potential for the
use of large hot water tanks and heat pumps with this type of tariff. The heat
pump only runs in the low-cost and renewable periods which should lead to low
operating costs and a high percentage of heat met by renewables.
• MPC requires a longer prediction horizon than that used for the existing electricity
tariffs to account for the large windows of fluctuations, over periods closer to a
week. It can be seen that the electric heater is still needed to see the system
through long periods of high cost. However, the sizing results should show
employing the MPC resulting in a higher proportion of renewables meeting
demand.
194 |
6.8. Sizing Results
The heat pump and hot water tank are sized for the proposed design using three existing
electricity tariffs: flat rate, variable periods, and time-of-use; and a future wind-based
electricity tariff. Price data for the existing tariffs is from 2017 and taken to be
representative of the types of price differentials available and allow this sizing study to
investigate and compare system performance using the different tariff structures.
The KPI levelized cost of heat (LCOH) was used to select the optimum size
combination in this study. The LCOH metric does not reflect the total cost of the system,
as only the heat pump and hot water tank capital expenditures are included, but it does
provide a helpful insight into trade-offs between the larger capital cost and the smaller
operating cost associated with increasing capacities of heat pump and hot water tank. The
capital cost plot is shown in Figure 6.27 and the shape of this plot shows the relative
expensive cost of increasing heat pump capacity as compared to hot water tank capacity.
Cost may not be the sole driver of a project, as environmental, e.g. CO2 emissions, or
wider societal benefits may be more important. Therefore, 3D plots are included for the
existing electricity tariffs of the following KPIs: LCOH, ORESpv, and HRESpv. For the
wind tariff 3D plots of the following KPIs are displayed: LCOH, ORESpv, and
HRESpv+windtariff. ORESpv is the percentage of on-site PV generation which is self-
consumed. HRESpv is the percentage of the heat demand which is met through usage of
the on-site PV generation. HRESpv+wind is the percentage of the heat demand which is
met through both the usage of the on-site PV generation and the usage of electrical
imports during high wind periods. See Section 6.1. for further details on these.
Results for the optimum LCOH size combinations of heat pump and hot water tank
are in Table 6.3 for the existing electricity tariffs and developed control strategies, and in
Table 6.4 for the wind electricity tariff and developed control strategies.
195 |
Table 6.3: Optimum LCOH results for the existing electricity tariffs and control strategies including KPIs (brackets is
the relative change from FOC to MPC)
HP TS HRESpv ORESpv LCOH

Tariff Control
(kW) (m3) (%) (%) (p/kWh)
FOC 750 400 33.8 92.7 4.75
Fixed Rate 4.62
MPC 750 500 32.7 96.7
(-2.7%)
FOC 1000 400 18.7 70.2 4.49
Variable
4.13
Periods MPC 1000 500 33.6 95.5
(-8.0%)
FOC 750 300 15.8 67.0 3.86
Time-of-use 3.11
MPC 750 500 32.1 95.9
(-19.4%)
Table 6.4: : Optimum LCOH results for the wind electricity tariff and control strategies including KPIs (brackets is
the relative change from FOC to MPC)
HP TS HRESpv+windtariff ORESpv LCOH

Tariff Control 3
(kW) (m ) (%) (%) (p/kWh)
FOC 1000 1500 52.8 73.8 5.81
Wind 3.25
MPC 1000 2000 70.2 98.1
(-44.1%)
196 |
Figure 6.27: 3D plot of capital cost for a range of size combinations
197 |
The main load shifting mechanism for fixed order control with the flat rate tariff is shifting
heat pump electricity consumption to match excess PV generation. The LCOH optimum
size combination is a 750kW heat pump and a 400m3 hot water tank.
The LCOH is reduced by increasing hot water tank capacity before levelling off above
capacities of 400m3 (Figure 6.28), and by increasing heat pump capacity to 750kW after
which the LCOH increases. The main driver for the reduction in LCOH is increasing the
heat pump capacity, as the higher usage of the heat pump over the auxiliary electric heat
meets the heat demand more efficiently. Another driver is a larger hot water tank which
allows higher on-site PV self-consumption (Figure 6.29). Both graphs show the same
levelling off point with increasing hot water tank capacity.
Figure 6.28: 3D plot of LCOH (levelized cost of heat) for FOC with flat rate tariff
198 |
The on-site PV self-consumption increases with a larger hot water tank, and then
levels off above hot water tank capacities of 400m3. The decrease in on-site PV self-
consumption as heat pump size increases is due to the more efficient conversion of
electricity to heat using the heat pump as opposed to the auxiliary electric heater.
This is clarified in Figure 6.30 which shows that increasing the heat pump and hot
water tank capacities also increases the percentage of the heat demand met by on-site PV.
Therefore, while the on-site PV self-consumption decreases with greater heat pump
usage, the increase in efficient use of the produced electricity results in a higher percentage
of heat demand met by on-site PV.
Figure 6.29: 3D plot of ORESpv (on-site PV self-consumption) for FOC with flat rate tariff
199 |
Figure 6.30: 3D plot of HRESpv (heat demand from on-site PV) for FOC with flat rate tariff
200 |
The fixed order control with the variable periods tariff shifts load to move electricity
consumption from the high-cost period during the day to the low-cost period at night,
and to utilise excess PV generation. The LCOH optimum size combination is a 1000kW
heat pump and a 400m3 hot water tank. This is a larger heat pump capacity, but the same
hot water tank capacity as compared to the optimum found for the flat rate tariff (750kW
heat pump and 400m3 hot water tank).
The size of the hot water tank is limited to the heat pump running at full output for
the duration of the low-cost period, as seen in the levelling of LCOH (Figure 6.31) for
additional hot water tank capacity larger than 500m3.
Figure 6.31: 3D plot of LCOH (levelized cost of heat) for FOC with variable periods tariff
201 |
An interesting phenomenon occurs when increasing the size of heat pump beyond
250kW. The percentage of heat demand from on-site PV decreases, and this is because
the increased charging of the hot water tank during the night, decreases the potential for
charging using the surplus PV generation further (Figure 6.33). While increasing heat
pump capacity decreases the LCOH by shifting additional load from day to night, the
renewable fraction lowers because of the hot water tank having less available charging
during the day where there is surplus of generation. This disadvantage of the controller
can be seen as additional storage has little effect on increasing the usage of on-site PV
above 100m3 (Figure 6.32).
Figure 6.32: 3D plot of ORESpv (on-site PV self-consumption) for FOC with variable periods tariff
202 |
Figure 6.33: 3D plot of HRESpv (heat demand from on-site PV) for FOC with variable periods tariff
203 |
The fixed order control with the time-of-use tariff shifts load mainly to avoid importing
during the premium period which occurs each day between 4pm and 7pm, and due to the
lack of sophistication of this control strategy the surplus PV generation is largely exported.
The LCOH optimum size combination is a 750kW heat pump and a 300m 3 hot water
tank.
The tariff varies through all periods of the day making it difficult to choose a setpoint
for the fixed order controller. The setpoint in this example is below the premium period
cost to ensure that the controller is attempting to use the hot water tank and avoid heat
pump usage in this period. A sensitivity analysis would be required to fine tune the
setpoint. With this primary objective of avoiding the premium period, the optimum hot
water tank for minimising LCOH is smaller than for the other tariffs as there is only a
three-hour period for which the hot water tank is used to shift load.
Figure 6.34: 3D plot of LCOH (levelized cost of heat) for FOC with time-of-use tariff
Similar to the variable periods, the on-site PV self-consumption and percentage of

heat demand met by on-site PV is limited due to insufficient hot water tank capacity
during PV generation which is due to limitations with the fixed order control strategy.
204 |
Figure 6.35: 3D plot of ORESpv (on-site PV self-consumption) for FOC with time-of-use tariff
Figure 6.36: 3D plot of HRESpv (heat demand from on-site PV) for FOC with time-of-use tariff
205 |
The wind tariff incentivises load shifting by offering high price differentials between
windy and non-windy periods, in addition to the day/night differential. As with the time-
of-use tariff the fixed order control is difficult to setup due to the importance of defining
the setpoint. The LCOH optimum size combination is a 1000kW heat pump and a
1500m3 hot water tank.
The fixed order control does not successfully take advantage of the price differentials
available with the wind tariff. The modelling results show a higher LCOH than the
existing tariffs. This is because the control is not shifting enough load from the high cost
periods, which are substantially higher than the existing tariffs, to the low-cost periods,
which are substantially lower than the existing tariffs.
Figure 6.37: 3D plot of LCOH (levelized cost of heat) for FOC with wind tariff
206 |
Similar to the variable periods and time-of-use tariffs, the on-site PV self-consumption
and percentage of heat demand met by on-site PV and high wind grid import is limited
due to insufficient hot water tank capacity during PV generation which is due to charging
the storage according to the import price and only charging from the surplus PV
generation opportunistically. However, unlike the existing tariffs, importing using the
wind tariff during periods of high-wind can be classed as from RES. Therefore, the
percentage of heat demand met from on-site PV and high wind grid import increases with
additional heat pump capacity up to 1500kW and for any additional hot water tank
capacity.
Figure 6.38: 3D plot of ORESpv (on-site PV self-consumption) for FOC with wind tariff
207 |
Figure 6.39: 3D plot of HRESpv+windtariff (heat demand from on-site PV and electrical imports during
high wind periods) for FOC with wind tariff
208 |
The only load shifting opportunity for MPC with the flat rate tariff is shifting heat pump
electricity consumption to match excess PV generation. The LCOH optimum size
combination is a 750kW heat pump and a 500m3 hot water tank.
Compared to the fixed order control, use of the MPC reduces the LCOH by 2.7%.
This is small because the fixed order control is only storing heat generated via excess PV
generation and discharging the hot water tank when no excess is available. As this is the
only form of load shifting available, the MPC performs a similar process.
Figure 6.40: 3D plot of LCOH (levelized cost of heat) for MPC with flat rate tariff
The MPC control strategy is slightly better at utilising the surplus PV generation. This
is because the MPC optimisation is cost-driven and PV generated electricity is modelled
with zero marginal cost which means that use of the heat pump and auxiliary electric heat
are equal in priority of dispatch. Meanwhile, the fixed order control always prioritises the
heat pump. This means that when employing MPC there will be more periods where the
auxiliary electric heat uses the surplus generation.
209 |
As with the fixed order control, while the on-site PV self-consumption decreases with
greater heat pump usage, the increase in efficient use of the produced electricity results in
a higher percentage of heat demand met by on-site PV.
Figure 6.41: 3D plot of ORESpv (on-site PV self-consumption) for MPC with flat rate tariff
Figure 6.42: 3D plot of HRESpv (heat demand from on-site PV) for MPC with flat rate tariff
210 |
The MPC with the variable periods tariff aims to optimise operation in order to shift load
to move electricity consumption from the high-cost period during the day to the low-cost
period at night, and to utilise excess PV generation. The LCOH optimum size
combination is a 1000kW heat pump and a 500m3 hot water tank.
Using MPC with the variable periods tariff decreases the LCOH by 8.0%. This saving
is made because the MPC allows greater usage of the on-site excess PV generation. This
is achieved by limiting the charging of the hot water tank during the cheaper night-time
periods such that there is sufficient capacity during the day to increase the storing of heat
from excess PV generation.
Figure 6.43: 3D plot of LCOH (levelized cost of heat) for MPC with variable periods tariff
211 |
The MPC enables almost all of the surplus PV generation to be self-consumed above
a hot water tank capacity of 300m3 (Figure 6.44). Increasing the heat pump capacity
reduces the self-consumption of the PV, and this is because of a greater use of the heat
pump which is more efficient than the auxiliary electric heat. This is again, similar to
previous tariffs, seen in the increase in percentage of heat demand met from on-site PV
by increasing either heat pump or hot water tank capacities.
Figure 6.44: 3D plot of ORESpv (on-site PV self-consumption) for MPC with variable periods tariff
212 |
Figure 6.45: 3D plot of HRESpv (heat demand from on-site PV) for MPC with variable periods tariff
213 |
The time-of-use tariff is variable hourly throughout the day and the MPC should be adept
at ensuring that electrical consumption coincides with the lowest cost periods, including
utilising surplus PV generation. The LCOH optimum size combination is a 750kW heat
pump and a 500m3 hot water tank.
Using MPC over the fixed order control decreases LCOH by 19.4%, making it the
lowest LCOH tariff for this control strategy. The savings come about because the fixed
order controller is limited to avoiding the premium period and does not use the storage
to shift load in the other price-varying periods. The MPC has the advantage of not
requiring a setpoint and can therefore utilise all the storage to shift load outside the
premium period to minimise operating cost across all periods. Additionally, as was seen
with the variable periods tariff, the MPC optimises the usage of the excess PV generation.
Figure 6.46: 3D plot of LCOH (levelized cost of heat) for MPC with time-of-use tariff
214 |
As with the variable periods tariff, the MPC enables almost all of the surplus PV
generation to be self-consumed above a hot water tank capacity of 300m3 (Figure 6.47).
A drop in the self-consumption for large heat pump and hot water tank combinations is
due to the greater proportion of heat demand being met by the heat pump which is more
efficient than the auxiliary electric heat. Again, similarly to previous tariffs, an increase in
percentage of heat demand met from on-site PV is achieved by increasing either heat
pump or hot water tank capacities.
Figure 6.47: 3D plot of ORESpv (on-site PV self-consumption) for MPC with time-of-use tariff
215 |
Figure 6.48: 3D plot of HRESpv (heat demand from on-site PV) for MPC with time-of-use tariff
216 |
The wind tariff incentivises load shifting by offering high price differentials between
windy and non-windy periods, in addition to the day/night differential. The MPC with
the wind tariff uses a 168-hour prediction horizon which allows the operation to account
for long periods of lots of wind or no wind. The LCOH optimum size combination is a
1000kW heat pump and a 2000m3 hot water tank, marking a significant increase in optimal
hot water tank size and similar optimal heat pump size compared to the existing tariffs.
Due to the larger parametric steps used for the hot water tank sizes, two additional
simulations were undertaken for a 1000kW heat pump with both 1500m3 and 2500m3 hot
water tank capacities. These both result in an increase in LCOH, therefore a 2000m3 hot
water tank remains the optimal size.
Figure 6.49: 3D plot of LCOH (levelized cost of heat) for MPC with wind tariff
Using MPC over the fixed order control decreases LCOH by 44.1%, which clearly
shows that using MPC is beneficial. These substantial savings are possible due to the
ability of the MPC, with the week-long prediction horizon, to optimally shift the heat
pump electrical consumption to the periods of low-cost. The wind tariff is highly variable
with a large differential between low-wind and high-wind periods, and this heavily
217 |
incentivises the load shifting mechanism which is enabled by a large hot water tank and
use of MPC.
Figure 6.50: 3D plot of ORESpv (on-site PV self-consumption) for MPC with wind tariff
As with the existing electricity tariffs, the MPC enables almost all of the surplus PV
generation to be self-consumed. A drop in the self-consumption for larger heat pump and
hot water tank combinations is due to the greater proportion of heat demand being met
by the heat pump which is more efficient than the auxiliary electric heat.
Unlike the existing tariffs, importing using the wind tariff during periods of high-wind
can be classed as from RES. In Figure 6.51 it can be seen that the percentage of heat
demand met from on-site PV and high wind grid import increases with additional heat
pump capacity and hot water tank capacity. With the wind tariff and MPC, along with a
large hot water tank, the percentage of heat demand met from on-site PV and high wind
grid import is greater than any of the other tariff and control combinations. This illustrates
the importance of combining MPC with a large hot water tank in future highly renewable
energy systems in order to maximise the local energy system renewable usage.
218 |
Figure 6.51: 3D plot of HRESpv+windtariff (heat demand from on-site PV and electrical imports during
high wind periods) for MPC with wind tariff
219 |
6.8.3. Discussion
Tariffs with greater variability provide more incentive to use the hot water tank to load
shift demand, as can be seen with lower LCOH for variable periods and time-of-use tariffs
over the flat rate tariff (using the given cost assumptions).
When using the fixed order controller, the flat rate and time-of-use tariffs both have
an optimal heat pump capacity of 750kW. This similarity is linked to the comparable,
short duration of the surplus PV generation and the premium period. The flat rate tariff
utilising surplus PV generation is the main load shifting mechanism, while with the time-
of-use tariff avoiding heat pump usage during the three-hour premium period is the
primary objective. Additionally, the variable periods tariff has a higher optimal heat pump
capacity of 1000kW which reflects the longer duration of the low-cost period. For the
variable periods tariff, shifting heat pump electrical consumption from day to night is the
main utility of the hot water tank.
When using the fixed order controller, the flat rate and variable periods tariffs both
have an optimal hot water tank capacity of 400m3. Increasing hot water tank capacity for
the variable periods tariff does not reduce operating cost with the fixed order controller
because the controller is not predictive and cannot increase the usage of surplus PV
generation and is limited in shifting load from day to night. The time-of-use tariff with
the fixed order controller has the lowest optimum hot water tank capacity because it is
only avoiding heat pump usage in a three-hour window.
Using MPC does not affect the optimal size of heat pump but increases the optimal
hot water tank size. With MPC, the hot water tank can load shift for multiple objectives
such as day to night shifting, optimising heat pump COP by shifting operating times, and
utilising excess PV generation.
Comparisons between the control strategies found that MPC offers savings over the
fixed order control for all the existing electricity tariffs, but particularly for a time-of-use
tariff where savings of 19.4% were obtained. Comparisons between the existing electricity
tariffs found that the time-of-use tariff delivered the lowest levelized cost of heat. This
conveys the advantage of combining tariffs which promote flexibility through load
shifting with optimally sized hot water tank.
MPC has the largest reduction in LCOH when paired with the wind tariff. Using MPC
over the fixed order control decreases LCOH by 44.1%. This shows how when using a
220 |
tariff which heavily incentivises the load shifting mechanism, a larger hot water tank and
use of MPC is advantageous. Additionally, the percentage of heat demand met by on-site
PV and high wind grid import increases from 52.8% to 70.2% for the fixed order control
and MPC.
The wind tariff optimal heat pump and hot water tank combination, 1000kW heat
pump and 2000m3 hot water tank, has the same heat pump capacity as the existing tariff
optimal sizes but has a larger hot water tank capacity. This means that systems which are
sized for the existing electricity tariffs can be modified later by adding hot water tank
capacity to take advantage of future tariffs, such as the wind tariff used in this sizing study.
6.9. Results Summary

Application of this methodology to a sizing study for a residential district heating scheme
has showcased the ability of PyLESA as a useful aid to investigate the benefits of model
predictive control (MPC) and different electricity tariffs including a novel future wind
tariff. PyLESA has proved to be a useful addition to the current set of existing modelling
tools to model the proposed design for WWHC including heat pumps, hot water tank,
variable electricity tariffs, and MPC. Close examination of the operational analysis
provided a validation of the underlying algorithms of the control strategies and showed
that running PyLESA can produce logical and useful outputs.
The operational analysis results provided a qualitative discussion of the comparisons
of the developed control strategies and the electricity tariffs. It identified limitations with
the fixed order controller when attempting to shift load beyond simple situations, i.e.
when attempting to utilise on-site renewable generation as well as avoiding high-cost
periods of a tariff. The operation of the MPC was explored and this illustrated the
advantages of the MPC which can optimise the operation of the heat pump and hot water
tank, for a number of objectives, in order to lower operating costs. The operational
analysis also revealed the potential for the use of large hot water tanks and heat pumps
with the future wind tariff.
A 750kW heat pump and 500m3 hot water tank using MPC and a time-of-use
electricity tariff were found to deliver the lowest LCOH in comparison with the existing
electricity tariff structures and control strategies. This marks a significant 10x expansion
of the existing hot water tank at WWHC which is used to aid peaks in demand for the
221 |
existing biomass boiler. This signifies a shift in the methods which are used to size hot
water tank; sizing to enable load shifting and not only for flattening peak demands during
design periods.
An optimal size combination of a 1000kW heat pump and a 2000m3 hot water tank
was found with MPC and the wind tariff. For the wind tariff performance improvements
were found by using MPC over the fixed order control: LCOH reducing from 5.81p/kWh
to 3.25p/kWh (44.1% reduction); and heat demand met by on-site PV and high wind grid
import increasing from 52.8% to 70.2%. The optimal heat pump size for the wind tariff
was found to be similar, or the same, as for the existing tariffs. The optimal hot water
tank capacity is significantly larger. Therefore, the proposed design could be sized for an
existing electricity tariff and later additional hot water tank capacity can be added to take
advantage of future tariffs.
222 |
7. Conclusions
The previous chapters have described the development of the local energy system
modelling tool, PyLESA, and the application of it to inform a sizing study of a proposed
design for a district heating network.
This chapter discusses the strengths and limitations of the research undertaken. This
will consist of discussion on the contributions to the state of the art in terms of the
modelling capabilities of PyLESA; exploration of the performance of the control
strategies and electricity tariffs; the limitations of PyLESA and potential future tool
developments; and potential applications of PyLESA. The chapter ends with a final
summary.
7.1. Contributions
Several contributions have been made in this work. They were achieved in order to answer
the central research question, which was introduced in Chapter 2 and is restated below:
Can a modelling methodology and supporting modelling tool be developed
that usefully aids the planning-level design of local energy systems incorporating
heat pumps, thermal storage, local renewable electricity production, time-of-use
electricity tariffs, and predictive controls?
Specific aims were set in order to answer this research question and a research
methodology was developed to organise the work required to achieve these aims. These
aims are discussed below and are split between those which address the gaps in existing
modelling tools, and those which explore the developed control strategies and electricity
tariffs in the context of the proposed district heating system design. The contributions
made are highlighted in this discussion.
7.1.1. Addressing Modelling Capability Gaps

The first research aim which was to identify gaps in existing planning-level modelling
tools. This was achieved by categorising the capabilities of the identified tools and
developing a modelling tool selection process. By analysis of these capabilities against the
223 |
local energy system specified in the research question, a set of modelling gaps were
identified.
The second and third research aims were to develop a modelling methodology and
supporting modelling tool which would address the identified gaps and aid design
decisions at a planning level by modelling local energy systems incorporating heat pumps,
thermal storage, local renewable electricity production, time-of-use electricity tariffs, and
predictive controls. This was achieved by the development of a modelling methodology
which consists of the steps for applying the novel modelling tool, PyLESA. The modelling
capabilities of PyLESA was then described, including details of the models which
contributed to identified gaps.
The discussion below concerns how the specific gaps were addressed, and the
strengths of the contributions made. The limitations and future developments of PyLESA
and its underlying models will be discussed in Section 7.2.
Ability to adapt source code and import or exploit functionality from elsewhere.
PyLESA was written in the programming language Python which was chosen because it
is open source and widely used across science and engineering fields. This meant that
PyLESA could build on the state of the art, utilising energy system models previously
built, as well as providing a platform from which the developed models can be shared
with other researchers in the energy system modelling community. As compared to the
reviewed tools which were primarily written in programming languages which require
expert software development skills, or are completely closed source, PyLESA offers the
explicit ability for others to adapt the source code or to easily couple PyLESA with other
energy system models developed in Python.
Temperature dependence for the heat pump models.
It was identified that in the reviewed tools that heat pumps were being modelled using
simple energetic models and are based upon a limited set of performance data. This leads
to an overestimation of performance across the operating conditions. In PyLESA a more
detailed heat pump modelling approach is developed which uses standard test data to
generate performance maps using multiple variable linear regression analysis with explicit
temperature dependence. This extends the current state of the art of approaches
employed by energy system planning tools.
224 |
Detailed stratification model for the thermal storage models.
Similar to the approach of the reviewed tools for modelling heat pumps, when modelling
thermal storage simple energetic models are utilised which do not capture the thermal
characteristics associated with stratified hot water tanks. The multi-node modelling
approach advances on these simple energetic models with thermal characteristics
modelled in more detail. The state of charge and heat loss of the hot water tank is
dependent on the node temperatures. This approach means that a hot water tank can be
represented with more detail while maintaining low input requirements which is helpful
at the planning stage.
Ability to model the evolving electricity markets and tariffs.
Existing modelling tools either have in-built simple tariffs which used fixed prices and
unlimited import and export or allow the user to input unique tariffs. As an advancement
on this, PyLESA can generate a range of electricity tariffs. Some of which reflect existing
and emerging tariffs such as the flat rates, variable periods, and time-of-use tariffs.
Additionally, a wind-based electricity tariff generator has been presented as an example
of a tariff which could exist in the UK in the future. A future UK 100% renewable energy
system would likely have wind power as one of the main producers of energy, and
therefore, a tariff pricing structure which is linked to wind power production is likely to
be similar to a future electricity tariff.
Ability to explore predictive controls.
A distinct disadvantage of the existing modelling tools was the lack of options for control
strategies with a notable gap for utilising predictive controls. These typically use simple
rule-based controls which need to be adjusted by an expert in order to minimise import
costs. There are also cases where on-site renewable production competes with a variable
electricity tariff which makes it difficult to identify a low-cost operation schedule.
Modelling local energy systems including a model predictive control strategy allows the
performance to be optimised. MPC enables the operation to minimise cost, even when
multiple competing dynamics influences are active, such as heat pump performance,
variable tariffs, and on-site renewable power generation. In PyLESA an economic MPC
strategy was developed and incorporated. These aspects were further explored in the
sizing study where the performance of the MPC was compared to the fixed order control,
with both using a range of electricity tariffs.
225 |
7.1.2. Exploration of Developed Control Strategies and Electricity
Tariffs
The fourth research aim was to demonstrate the application of PyLESA by undertaking
a sizing study and explore the control strategies and electricity tariffs. This aim was
achieved by carrying out a sizing study for a proposed design of a residential district
heating system which incorporates a heat pump, a hot water tank, on-site PV electricity
production, variable electricity tariffs, and MPC. The sizing study allowed for the
developed control strategies and the use of existing electricity tariffs to be compared.
These were compared by operational analysis which was used to qualitatively describe the
control strategies and the influence of the different electricity tariffs, and also compared
by sizing of the heat pump and hot water tank capacities for the different control strategies
and electricity tariffs. Additionally, these included an exploration of a future wind-based
renewable electricity tariff.
The discussion below concerns the findings on the comparisons of the developed
control strategies and existing electricity tariffs. It also discusses the exploration of the use
of a future wind-based renewable electricity tariff which is useful for studies of future
energy systems with large renewable penetrations. Finally, a discussion of the results from
the sizing study of the proposed design for WWHC for the optimal heat pump and hot
water tank size combination with control strategy and electricity tariff conclude this
section.
Compare the performance of the developed control strategies for existing tariffs.
The fixed order control and MPC perform differently for each of the electricity tariffs.
There is little advantage of using MPC over the fixed order control when the only load
shifting mechanism available is utilisation of surplus PV generation. The fixed order
control enables as much load shifting as the MPC to increase the self-consumption from
on-site PV generation. The advantages of the MPC are illuminated when there are
multiple load shifting mechanisms available, e.g. surplus generation, low-cost tariff
periods, heat pump performance. The MPC can optimise the operation of the heat pump
and hot water tank for all of these objectives simultaneously, resulting in lower operating
costs.
226 |
Comparisons between the control strategies found that MPC offers savings over the
fixed order control for all the existing electricity tariffs, but particularly for a time-of-use
tariff where savings of 19.4% were obtained.
Compare the use of existing electricity tariffs.
Load shifting with the flat rates tariff is only incentivised by attempting to increase the
self-consumption from on-site PV generation. This is because load shifting heat pump
and electric heater power consumption for surplus PV generation is the only avenue of
reducing costs.
With the variable periods tariff load shifting is useful for both increasing use of surplus
PV generation and avoiding high-cost tariff periods. However, the hot water tank is often
fully charged at the end of the low-cost period and just before there are surpluses of PV
generation. This results in a significant portion of PV generation being exported, despite
this having zero value to the local energy system in this case study. Employing MPC
results in a much higher utilisation of surplus PV generation, and lower import costs as a
result of only charging the hot water tank at low-cost periods in order to cover the
remaining heat demand.
The time-of-use tariff is highly variable, while the fixed order control does not take
advantage of this variability. The MPC makes it possible to avoid premium prices and
take advantage of all the periods of variability. Additionally, similar to the variable periods
tariff, the PV utilisation can be maximised, and imports limited to the lowest cost periods
and restricted to only meeting the remainder of the demand in the calculation period.
Comparisons between the existing electricity tariffs found that the time-of-use tariff
delivered the lowest levelized cost of heat. This conveys the advantage of combining
tariffs which promote flexibility through load shifting with optimally sized hot water tank
capacity.
Explore the use of a future wind-based renewable electricity tariff.
From the operational analysis, it is apparent that using the fixed order control with the
wind tariff shows great potential for the use of large hot water tanks and heat pumps for
this type of tariff. The heat pump generally runs in the low-cost and renewable periods
which should lead to low operating costs and a high percentage of heat met by renewables.
MPC requires a longer prediction horizon than that used for the existing electricity tariffs
to account for the large windows of fluctuations, over periods closer to a week. It can be
227 |
seen in the operational graphs that the electric heater is still needed to see systems through
long periods of high cost.
The sizing results show that using MPC over the fixed order control decreases LCOH
by 44.1%. This shows that using the wind tariff, which heavily incentivises the load
shifting mechanism, a larger hot water tank and use of MPC, is advantageous.
Additionally, the percentage of heat demand met by on-site PV and high wind grid import
increases from 52.8% to 70.2% for the fixed order control and MPC.
Identify optimal heat pump and hot water tank size combination.
A 750kW heat pump and 500m3 hot water tank using MPC and a time-of-use electricity
tariff were found to deliver the lowest LCOH in comparison with the existing electricity
tariff structures and control strategies. This marks a significant 10x expansion of the
existing hot water tank at WWHC, which is used to aid peaks in demand for the existing
biomass boiler. This signifies a shift in the methods which are used to size hot water tanks;
sizing to enable load shifting and not only for flattening peak demands during design
periods.
The wind tariff optimal heat pump and hot water tank combination is a 1000kW heat
pump and 2000m3 hot water tank, this has the same heat pump capacity as the existing
tariff optimal sizes but has a larger hot water tank capacity. This means that systems which
are sized for the existing electricity tariffs can be modified later by adding hot water tank
capacity to take advantage of future tariffs, such as the wind tariff used in this sizing study.
7.2. Future Tool Development

In this section the limitations of the framework and underlying models of PyLESA are
explored along with potential future tool developments. PyLESA has the potential to
incorporate a significant number of models for an assortment of technologies, and the
discussion presented is limited to the particularly pertinent models.
7.2.1. Tool Framework

Several existing frameworks could have provided the basis as a starting point for
developing the models which addressed the identified gaps. Instead, PyLESA was built
primarily from scratch to aid the author in gaining proficiency in Python programming.
Given the nature of Python and the accessibility of the code of PyLESA, it is possible for
others to take individual models/class objects and implement them within
228 |
existing/developing frameworks. Ultimately, PyLESA itself now offers a useful
framework for modelling local energy systems which can be expanded by incorporating
new models or developing the existing models, but alternative research directions were
available.
The modelling framework uses an Excel workbook as the method for the user to input
the necessary data, and outputs several images and .csv files. A more pythonic method
would be to develop an open source GUI written in Python where data can be input, the
tool can be run, and outputs explored.
For certain types of analysis, it would be useful to be able to model systems on smaller
timescales. An hourly timestep is used in PyLESA due to the typical data availability at
the planning stage and the modelling assumptions used. For example, the heat pump
regression model does not explicitly account for dynamic effects associated with actions
such as startup/shutdown. Therefore, using a timestep below 15 minutes would result in
inaccurate outputs for these short periods.
The current framework limits the number of technologies and demands which can be
modelled. For example, it may be useful to be able to model a system with a heat pump
connected to two hot water tanks for providing hot water and space heating separately.
PyLESA could be modified to include the option to model two hot water tanks and two
heat demands.
The capital costs which are used to calculate financial KPIs such as levelized cost of
energy/heat, currently only include the heat pump and hot water tank capital costs. In
order to perform sizing and feasibility studies looking at other components of the local
energy system, it would be useful to incorporate the full range of capital cost expenditures,
e.g. wind turbine costs and district heat network piping installation costs.
7.2.2. Demand Modelling

Simplified, deterministic assessment methods have been presented as a means of
generating electrical and heat demands at the required timestep and detail for inputting
into PyLESA. These methods only produce demands for a single year and do not account
for year-to-year variability and extreme demand events. Weather has a significant impact
on peak demand and energy use, and varies across years [269,270]. Therefore, an
improvement to the methods developed as part of PyLESA would be to generate demand
profiles based on weather data over multiple years. Additionally, sub-hourly, high-
229 |
resolution models which use stochastic techniques could be incorporated to account for
a range of occupancy behaviours, and other complex factors [213]. District heating
network piping models could also be incorporated to facilitate aiding design decisions,
and computationally inexpensive methods for modelling these have been reviewed in
literature [271].
7.2.3. Heat Pump and Auxiliary Units

The most advanced model for heat pumps in PyLESA is the standard test regression
model. This approach uses a relatively substantial amount of data input regarding the
performance of the heat pump under the standard test range of operating conditions, and
this is not always readily available at the planning-stage. This is the reason simpler models,
similar to those utilised in existing modelling tools, have been included as alternative
approaches. Despite this, a more detailed approach, requiring greater input requirements,
could fully capture the major components such as the physical details of evaporator
design, the system defrosting cycle, and associated controls [39,272]. This more detailed
approach could include important factors such as COP as a function of the variation of
the compressor speed. It would also allow for simulations at a high time resolution as the
dynamic effects below 15-minutes could be fully captured.
PyLESA does not incorporate solar thermal and this technology has the potential to
be an auxiliary, renewable heat generation unit which could assist a heat pump in meeting
heat demands. Models have been developed previously for solar thermal hot-water
heating systems [273], and photovoltaic thermal flat plate collectors [274].
7.2.4. Thermal Storage

PyLESA uses a multi-node modelling approach for hot water tanks and this was justified
as a suitable balance between accuracy and computational time. The developed model
exaggerates the mixing between nodes which is likely to occur for a hot water tank
designed to be highly stratified, as is desirable for the load shifting mechanisms
investigated. Therefore, while the energetic models used in existing models likely result in
underestimations of the capacity required to perform load shifting, the model developed
within PyLESA is likely to overestimate the capacity required.
To increase the user flexibility of PyLESA, hot water tank models with both increased
and decreased accuracy could be included. Fully mixed or moving boundary models,
similar to those used by existing modelling tools, provides a computationally efficient
230 |
approach, but fails to capture stratification. More complex approaches include 2D or 3D
CFD simulations [275], zonal approach [276], etc. PyLESA could also be developed to
include more thermal storage technologies such as phase change materials [255,256], next
generation smart hot water tanks [257], and building thermal mass [258].
A potential development of this approach would be to allow for multiple heat sources
to charge the tanks at different points of the tank. This would enable the integration of
solar thermal as a complimentary renewable source of heat along with a heat pump.
7.2.5. Electrical Storage and Fuel Synthesis

PyLESA uses a simple storage model for electrical storage, and this only captures generic
energetic technical characteristics of different types of storage. More detailed models
which are technology specific could be included, e.g. HOMER and iHOGA models for
battery types reviewed earlier [277–279], pumped hydro [280], CAES [281].
PyLESA is not capable of analysing fuel synthesis technologies as they are currently
unlikely to be applicable at a local energy system scale in the short term. However, this
capability could be included in the future, using models such as those developed for
hydrogen [282–284].
7.2.6. Electricity Tariffs and Balancing Markets

In PyLESA existing and future wind electricity tariffs can be modelled. Balancing markets
could provide a greater incentive for flexibility and may become more prevalent as the
grid decarbonises e.g. balancing mechanism, frequency response, and new markets such
as the European wide balancing energy market TERRE [48]. Including these would be a
useful future tool development. This would require a smaller simulation timestep to be
used, which requires developing the underlying models and assessment methods, in
particular the demand assessment and heat pump model.
The future tariff developed is based upon wind power production which may be
suitable in the UK context, but for other countries it may be apt to include a future tariff
which is dependent on PV. A future tariff could also be developed based upon an energy
systems model of the UK using a national-scale energy modelling tool such as
EnergyPLAN [285]. Additionally, PyLESA currently uses a flat rate for the export tariff,
however variable export tariffs are also emerging [286].
231 |
PyLESA includes model predictive control and this formed an important contribution to
the state of the art of planning-level local energy system modelling. An important aspect
of future work on the developed MPC is the inclusion of uncertainties in the future
predictions as currently these are modelled with perfect foresight [49,50]. This
overestimates the benefits of the MPC which will always makes the perfect decisions,
even when using a long prediction horizon. Further research is required to include
uncertainties with weather forecasts and electricity prices at the design stage, although a
large number of studies have developed controllers suitable for real-time control of
various energy systems, e.g. thermal storage [287,288], HVAC applications [289], building
cooling [290], community battery storage [291], community micro-grid [292], stochastic
MPC [248,293].
7.3. Future Tool Applications

PyLESA has been applied to a sizing study for a proposed design of an existing district
heating network as is described in Chapter 6. Potential applications of using PyLESA to
inform feasibility studies and operation studies have already been discussed in Section 4.5.
The applicability of PyLESA to different scales of energy system is another potential
tool development. For example, developing the tool to enable separate thermal stores and
multiple heat pump operation modes in order to model configurations commonly used
at the building or mini-district scale. It would also be useful to add to the set of control
strategies in order to be able to easily model the assortment of controls which are used at
the different scales, which would in turn increase the number of situations PyLESA is
applicable to.
PyLESA could also be utilised as the first step towards the development of the
software and hardware necessary to build a real-time model predictive controller. This
could be possible through the open-source communications and control software and
hardware available from the OpenEnergyMonitor project [36].
232 |
7.4. Final Summary
In this thesis a modelling methodology and a supporting modelling tool, PyLESA, have
been developed which can usefully aid the planning-level design of local energy systems
incorporating heat pumps, thermal storage, local renewable electricity production, time-
of-use electricity tariffs, and predictive controls.
Several gaps were identified in a review of the modelling capabilities of existing energy
system tools: (i) ability to adapt source code and import or exploit functionality from
elsewhere; (ii) temperature dependence for the heat pump models; (iii) detailed
stratification model for the thermal storage models; (iv) ability to model the evolving
electricity markets and tariffs; and (v) ability to explore predictive controls.
These gaps motivated the development of the novel modelling tool PyLESA which
can aid design studies at the planning stage and includes appropriate technology and
control models to tackle the identified gaps. The following tool capabilities of PyLESA
were described and validated: resources and demands; electricity production technologies;
heat pumps; hot water tanks; electricity tariffs; fixed order controls; model predictive
controls; and key performance indicators.
A sizing study was then devised to showcase the application of PyLESA and to
explore the developed control strategies, and existing and future electricity tariffs.
Comparisons between the control strategies found that MPC offers savings over the fixed
order control for all the existing electricity tariffs, but particularly for a time-of-use tariff
where savings of 19.4% were obtained. Comparisons between the existing electricity
tariffs found that the time-of-use tariff delivered the lowest levelized cost of heat. This
conveys the advantage of combining tariffs which promote flexibility through load
shifting with optimally sized hot water tank.
A 750kW heat pump and 500m3 hot water tank using MPC and a time-of-use
electricity tariff were found to deliver the lowest LCOH in comparison with the existing
electricity tariff structures and control strategies. This signifies a shift in the methods
which are used to size hot water tank; sizing to enable load shifting and not only for
flattening peak demands during design periods.
An optimal size combination of a 1000kW heat pump and a 2000m3 hot water tank
was found for MPC and the wind tariff. With the wind tariff, performance improvements
were obtained by using MPC over the fixed order control: LCOH reducing from
233 |
5.81p/kWh to 3.25p/kWh (44.1% reduction); and heat demand met by on-site PV and
high wind grid import increasing from 52.8% to 70.2%. The optimal heat pump size was
similar for the existing tariffs and future tariff. Therefore, the proposed design could be
sized for an existing electricity tariff and later, additional hot water tank capacity added to
take advantage of future tariffs.
The research has highlighted the advantage of combining flexible tariffs with optimally
sized thermal storage and showcased PyLESA as capable of usefully aiding the design of
local energy systems.
234 |
8. References
[1] Chmutina K, Goodier CI. Alternative future energy pathways: Assessment of the
potential of innovative decentralised energy systems in the UK. Energy Policy
2014;66:62–72. https://doi.org/10.1016/j.enpol.2013.10.080.
[2] Hammond GP, Howard HR, Jones CI. The energy and environmental implications
of UK more electric transition pathways: A whole systems perspective. Energy
Policy 2013;52:103–16. https://doi.org/10.1016/j.enpol.2012.08.071.
[3] Williams J. The deployment of decentralised energy systems as part of the housing
growth programme in the UK. Energy Policy 2010;38:7604–13.
https://doi.org/10.1016/j.enpol.2009.08.039.
[4] Bank J, Mather B, Keller J, Coddington M. High Penetration Photovoltaic Case
Study Report High Penetration Photovoltaic Case Study Report 2013:27.
https://doi.org/10.2172/1062441.
[5] Toke D. Community Wind Power in Europe and in the UK. Wind Eng
2005;29:301–8. https://doi.org/10.1260/030952405774354886.
[6] Bracken LJ, Bulkeley HA, Maynard CM. Micro-hydro power in the UK: The role
of communities in an emerging energy resource. Energy Policy 2014;68:92–101.
[7] Siano P. Demand response and smart grids - A survey. Renew Sustain Energy Rev
2014;30:461–78. https://doi.org/10.1016/j.rser.2013.10.022.
[8] Holmgren K. Role of a district-heating network as a user of waste-heat supply from
various sources - the case of Göteborg. Appl Energy 2006;83:1351–67.
https://doi.org/10.1016/j.apenergy.2006.02.001.
[9] What is Combined Heat and Power? | Resources | The Association for
Decentralised Energy n.d. https://www.theade.co.uk/resources/what-is-
combined-heat-and-power (accessed September 12, 2018).
[10] Gahleitner G. Hydrogen from renewable electricity: An international review of
power-to-gas pilot plants for stationary applications. Int J Hydrogen Energy
2013;38:2039–61. https://doi.org/10.1016/j.ijhydene.2012.12.010.
235 |
[11] Scottish Government. Annual Compendium of Scottish Energy Statistics 2019.
[12] Saidur R, Abdelaziz EA, Demirbas A, Hossain MS, Mekhilef S. A review on
biomass as a fuel for boilers. Renew Sustain Energy Rev 2011;15:2262–89.
https://doi.org/10.1016/j.rser.2011.02.015.
[13] Mathiesen BV, Lund H, Connolly D. Limiting biomass consumption for heating
in 100% renewable energy systems. Energy 2012;48:160–8.
https://doi.org/10.1016/j.energy.2012.07.063.
[14] Bačeković I, Østergaard PA. Local smart energy systems and cross-system
integration. Energy 2018;151:812–25.
[15] EHPA. Large scale heat pumps in Europe. Eur Heat Pump Assoc 2017.
[16] Connolly D, Lund H, Mathiesen BV, Werner S, Möller B, Persson U, et al. Heat
Roadmap Europe: Combining district heating with heat savings to decarbonise the
EU energy system. Energy Policy 2014;65:475–89.
[17] David A, Mathiesen BV, Averfalk H, Werner S, Lund H. Heat Roadmap Europe:
Large-scale electric heat pumps in district heating systems. Energies 2017;10:578.
https://doi.org/10.3390/en10040578.
[18] Levihn F. CHP and heat pumps to balance renewable power production: Lessons
from the district heating network in Stockholm. Energy 2017;137:670–8.
[19] Breeze P. Combined heat and power. n.d.
[20] Ridjan I, Mathiesen BV, Connolly D, Duić N. The feasibility of synthetic fuels in
renewable energy systems. Energy 2013;57:76–84.
[21] Denholm P, Hand M. Grid flexibility and storage required to achieve very high
penetration of variable renewable electricity. Energy Policy 2011;39:1817–30.
[22] Blarke MB. Towards an intermittency-friendly energy system: Comparing electric
boilers and heat pumps in distributed cogeneration. Appl Energy 2012;91:349–65.
[23] Alimohammadisagvand B, Jokisalo J, Kilpeläinen S, Ali M, Sirén K. Cost-optimal
236 |
thermal energy storage system for a residential building with heat pump heating
and demand response control. Appl Energy 2016;174:275–87.
[24] Mahfuz MH, Anisur MR, Kibria MA, Saidur R, Metselaar IHSC. Performance
investigation of thermal energy storage system with Phase Change Material (PCM)
for solar water heating application. Int Commun Heat Mass Transf 2014;57:132–
9. https://doi.org/10.1016/j.icheatmasstransfer.2014.07.022.
[25] Le Dréau J, Heiselberg P. Energy flexibility of residential buildings using short
term heat storage in the thermal mass. Energy 2016;111:991–1002.
[26] Lund H, Werner S, Wiltshire R, Svendsen S, Thorsen JE, Hvelplund F, et al. 4th
Generation District Heating (4GDH). Energy 2014;68:1–11.
[27] Lund H, Østergaard PA, Connolly D, Ridjan I, Mathiesen BV, Hvelplund F, et al.
Energy Storage and Smart Energy Systems. Int J Sustain Energy Plan Manag
2016;11:3–14. https://doi.org/10.5278/ijsepm.2016.11.2.
[28] Kousksou T, Bruel P, Jamil A, El Rhafiki T, Zeraouli Y. Energy storage:
Applications and challenges. Sol Energy Mater Sol Cells 2014;120:59–80.
https://doi.org/10.1016/j.solmat.2013.08.015.
[29] Chen H, Cong TN, Yang W, Tan C, Li Y, Ding Y. Progress in electrical energy
storage system: A critical review. Prog Nat Sci 2009;19:291–312.
https://doi.org/10.1016/j.pnsc.2008.07.014.
[30] Steele P. Agile pricing explained | Octopus Energy 2019.
https://octopus.energy/blog/agile-pricing-explained/ (accessed April 10, 2020).
[31] Usef Energy – Universal Smart Energy Framework n.d. https://www.usef.energy/
(accessed April 10, 2020).
[32] VHPready Services GmbH. Home - VHPready Services GmbH 2018.
https://www.vhpready.de/en/home/ (accessed April 10, 2020).
[33] REScoop.eu project proposals - News n.d.
https://www.rescoop.eu/blog/rescoop-eu-project-proposals (accessed April 10,
2020).
[34] FLEXcoop Project n.d. http://www.flexcoop.eu/ (accessed April 10, 2020).
237 |
[35] WiseGRID · WiseGRID n.d. https://www.wisegrid.eu/ (accessed April 10, 2020).
[36] OpenEnergyMonitor. Home | OpenEnergyMonitor 2016.
https://openenergymonitor.org/ (accessed April 10, 2020).
[37] Yu Z, Huang G, Haghighat F, Li H, Zhang G. Control strategies for integration
of thermal energy storage into buildings: State-of-the-art review. Energy Build
2015;106:203–15. https://doi.org/10.1016/j.enbuild.2015.05.038.
[38] Thieblemont H, Haghighat F, Ooka R, Moreau A. Predictive control strategies
based on weather forecast in buildings with energy storage system: A review of the
state-of-the art. Energy Build 2017;153:485–500.
https://doi.org/10.1016/j.enbuild.2017.08.010.
[39] Fischer D, Bernhardt J, Madani H, Wittwer C. Comparison of control approaches
for variable speed air source heat pumps considering time variable electricity prices
and PV. Appl Energy 2017;204:93–105.
[40] Gürel AE, Ceylan I. Thermodynamic analysis of PID temperature controlled heat
pump system. Case Stud Therm Eng 2014;2:42–9.
https://doi.org/10.1016/j.csite.2013.11.002.
[41] Roman R, Wilson R. Commercial Demand Side Management Using a
Programmable Logic Controller. IEEE Trans Power Syst 1995;10:376–9.
https://doi.org/10.1109/59.373959.
[42] Huang T, Liu D. A self-learning scheme for residential energy system control and
management. Neural Comput Appl 2013;22:259–69.
https://doi.org/10.1007/s00521-011-0711-6.
[43] Andrew Putrayudha S, Kang EC, Evgueniy E, Libing Y, Lee EJ. A study of
photovoltaic/thermal (PVT)-ground source heat pump hybrid system by using
fuzzy logic control. Appl Therm Eng 2015;89:578–86.
https://doi.org/10.1016/j.applthermaleng.2015.06.019.
[44] Ruelens F, Claessens BJ, Vandael S, Iacovella S, Vingerhoets P, Belmans R.
Demand response of a heterogeneous cluster of electric water heaters using batch
reinforcement learning. Proc. - 2014 Power Syst. Comput. Conf. PSCC 2014,
Institute of Electrical and Electronics Engineers Inc.; 2014.
https://doi.org/10.1109/PSCC.2014.7038106.
238 |
[45] Holjevac N, Capuder T, Kuzle I. Adaptive control for evaluation of flexibility
benefits in microgrid systems. Energy 2015;92:487–504.
[46] Huang Z, Yu H, Peng Z, Zhao M. Methods and tools for community energy
planning: A review. Renew Sustain Energy Rev 2015;42:1335–48.
[47] University of Aalborg. EnergyPLAN 2017. http://www.energyplan.eu/.
[48] Lund H, Mathiesen B V. Energy system analysis of 100% renewable energy
systems-The case of Denmark in years 2030 and 2050. Energy 2009;34:524–31.
[49] University of Aalborg. Literature with EnergyPLAN 2017.
http://www.energyplan.eu/category/scientific-literature-with-energyplan/
(accessed December 1, 2017).
[50] Lund H, Duić N, Krajacić G, Graça Carvalho M da. Two energy system analysis
models: A comparison of methodologies and results. Energy 2007;32:948–54.
[51] University of Zagreb. H2RES 2009. http://h2res.fsb.hr/ (accessed December 1,
2017).
[52] HOMER Energy. HOMER Pro 2017.
http://www.homerenergy.com/HOMER_pro.html.
[53] HOMER Energy. HOMER Energy Literature 2017.
http://microgridnews.com/homer-energy-bibliography/ (accessed December 1,
2017).
[54] Shahzad MK, Zahid A, ur Rashid T, Rehan MA, Ali M, Ahmad M. Techno-
economic feasibility analysis of a solar-biomass off grid system for the
electrification of remote rural areas in Pakistan using HOMER software. Renew
Energy 2017;106:264–73. https://doi.org/10.1016/j.renene.2017.01.033.
[55] University of Strathclyde. Merit 2015.
http://www.esru.strath.ac.uk/Programs/Merit.htm (accessed September 26,
2015).
[56] Morton C, Grant A, Kim J. An Investigation into the Specification of an Off-Grid
Hybrid Wind / Solar Renewable Energy System to Power an Ecobarn Building .
239 |
2017. http://www.esru.strath.ac.uk/Programs/Merit/case_study_Ecobarn.pdf
[57] TRNSYS. TRNSYS Transient System Simulation Tool 2017.
http://www.trnsys.com/ (accessed May 4, 2017).
[58] Beausoleil-Morrison I, Kummert M, MacDonald F, Jost R, McDowell T, Ferguson
A. Demonstration of the new ESP-r and TRNSYS co-simulator for modelling
solar buildings. Energy Procedia 2012;30:505–14.
https://doi.org/10.1016/j.egypro.2012.11.060.
[59] Kalogirou SA. Use a TRNSYS for modelling and simulation of a hybrid pv-thermal
solar system for Cyprus. Renew Energy 2001;23:247–60.
https://doi.org/10.1016/S0960-1481(00)00176-2.
[60] Lambert T, Gilman P, Lilienthal P. Micropower System Modeling with Homer. In:
Farret FA, Simões MG, editors. Integr. Altern. Sources Energy, John Wiley &
Sons, Inc.; 2006, p. 379–418. https://doi.org/10.1002/0471755621.ch15.
[61] Connolly D, Lund H, Mathiesen BV, Leahy M. A review of computer tools for
analysing the integration of renewable energy into various energy systems. Appl
Energy 2010;87:1059–82. https://doi.org/10.1016/j.apenergy.2009.09.026.
[62] Copetti JB, Lorenzo E, Chenlo F. A general battery model for PV system
simulation. Prog Photovoltaics Res Appl 1993;1:283–92.
https://doi.org/10.1002/pip.4670010405.
[63] Dumont O, Carmo C, Dickes R, Emelines G, Quoilin S, Lemort V. Hot water
tanks : How to select the optimal modelling approach? CLIMA 2016, AAlborg,
2016.
[64] van Beuzekom I, Gibescu M, Slootweg JG. A review of multi-energy system
planning and optimization tools for sustainable urban development. 2015 IEEE
Eindhoven PowerTech, 2015, p. 1–7.
https://doi.org/10.1109/PTC.2015.7232360.
[65] Allegrini J, Orehounig K, Mavromatidis G, Ruesch F, Dorer V, Evins R. A review
of modelling approaches and tools for the simulation of district-scale energy
systems. Renew Sustain Energy Rev 2015;52:1391–404.
[66] Keirstead J, Jennings M, Sivakumar A. A review of urban energy system models:
240 |
Approaches, challenges and opportunities. Renew Sustain Energy Rev
[67] Mendes G, Ioakimidis C, Ferrão P. On the planning and analysis of Integrated
Community Energy Systems: A review and survey of available tools. Renew Sustain
Energy Rev 2011;15:4836–54. https://doi.org/10.1016/j.rser.2011.07.067.
[68] Markovic D, Cvetkovic D, Masic B. Survey of software tools for energy efficiency
in a community. Renew Sustain Energy Rev 2011;15:4897–903.
[69] Mirakyan A, De Guio R. Integrated energy planning in cities and territories: A
review of methods and tools. Renew Sustain Energy Rev 2013;22:289–97.
[70] Walker WE, Harremoës P, Rotmans J, van der Sluijs JP, van Asselt MBA, Janssen
P, et al. Defining Uncertainty: A Conceptual Basis for Uncertainty Management in
Model-Based Decision Support. Integr Assess 2003;4:5–17.
https://doi.org/10.1076/iaij.4.1.5.16466.
[71] Mirakyan A, De Guio R. Modelling and uncertainties in integrated energy planning.
Renew Sustain Energy Rev 2015;46:62–9.
[72] Ascough JC, Maier HR, Ravalico JK, Strudley MW. Future research challenges for
incorporation of uncertainty in environmental and ecological decision-making.
Ecol Modell 2008;219:383–99. https://doi.org/10.1016/j.ecolmodel.2008.07.015.
[73] Ioannou A, Angus A, Brennan F. Risk-based methods for sustainable energy
system planning: A review. Renew Sustain Energy Rev 2017;74:602–15.
[74] Zeng Y, Cai Y, Huang G, Dai J. A Review on Optimization Modeling of Energy
Systems Planning and GHG Emission Mitigation under Uncertainty. Energies
2011;4:1624–56. https://doi.org/10.3390/en4101624.
[75] Li YF, Li YP, Huang GH, Chen X. Energy and environmental systems planning
under uncertainty-An inexact fuzzy-stochastic programming approach. Appl
[76] Urbanucci L, Testi D. Optimal integrated sizing and operation of a CHP system
with Monte Carlo risk analysis for long-term uncertainty in energy demands.
241 |
Energy Convers Manag 2018;157:307–16.
https://doi.org/10.1016/j.enconman.2017.12.008.
[77] Elharidi AM, Tuohy PG, Teamah MA. The energy and indoor environmental
performance of Egyptian offices: Parameter analysis and future policy. Energy
Build 2018. https://doi.org/10.1016/j.enbuild.2017.10.035.
[78] Awerbuch S. Portfolio-based electricity generation planning: Policy implications
for renewables and energy security. Mitig Adapt Strateg Glob Chang 2006;11:693–
710. https://doi.org/10.1007/s11027-006-4754-4.
[79] Boomsma TK, Meade N, Fleten SE. Renewable energy investments under
different support schemes: A real options approach. Eur J Oper Res
2012;220:225–37. https://doi.org/10.1016/j.ejor.2012.01.017.
[80] Troldborg M, Heslop S, Hough RL. Assessing the sustainability of renewable
energy technologies using multi-criteria analysis: Suitability of approach for
national-scale assessments and associated uncertainties. Renew Sustain Energy Rev
[81] Tuohy P, Kim JM, Samuel A, Peacock A, Owens E, Dissanayake M, et al.
Orchestration of renewable generation in low energy buildings and districts using
energy storage and load shaping. Energy Procedia, vol. 78, 2015, p. 2172–7.
[82] Haghigat F, Tuohy P, Fraisse G, Pero C Del. IEA ECES Annex 31 Final Report -
Energy Storage with Energy Efficient Buildings and Districts: Optimization and
Automation 2019.
[83] Lyden A, Pepper R, Tuohy PG. A modelling tool selection process for planning of
community scale energy systems including storage and demand side management.
Sustain Cities Soc 2018;39:674–88. https://doi.org/10.1016/j.scs.2018.02.003.
[84] Bava F, Furbo S. Development and validation of a detailed TRNSYS-Matlab
model for large solar collector fields for district heating applications. Energy
2017;135:698–708. https://doi.org/10.1016/j.energy.2017.06.146.
[85] Wetter M. Design Optimization with GenOpt. Build Energy Simul User News
2000;21:200.
[86] EMD. EMD International A/S energyTRADE - EMD International A/S 2017.
https://www.emd.dk/energytrade/ (accessed December 13, 2017).
242 |
[87] ENVI_MET. ENVI_MET tool 2017. http://www.envi-met.com/introduction/
[88] Lawrence Berkeley National Lab. Radiance tool 2017. https://www.radiance-
online.org/ (accessed December 13, 2017).
[89] The Balmorel Open Source Project. Balmorel Energy System Model 2017.
http://www.balmorel.com/ (accessed February 1, 2017).
[90] U.S Department of Energy. BCHP Screening Tool USer Manual 2016.
https://www1.eere.energy.gov/manufacturing/tech_assistance/pdfs/bchp_user
_manual.pdf, http://eber.ed.ornl.gov/bchpsc/ (accessed December 22, 2016).
[91] The Carbon Trust. Biomass Decision Support Tool 2017.
https://www.carbontrust.com/resources/tools/biomass-decision-support-tool/
[92] Robinson D, Haldi F, Kämpf JH, Leroux P, Perez D, Rasheed A, et al. CITYSIM:
Comprehensive Micro-Simulation Of Resource Flows For Sustainable Urban
Planning. Int IBPSA Conf 2009:1083–90.
[93] Ruan Y, Cao J, Feng F, Li Z. The role of occupant behavior in low carbon oriented
residential community planning: A case study in Qingdao. Energy Build
[94] van Beuzekom I, Gibescu M, Slootweg JG. A review of multi-energy system
planning and optimization tools for sustainable urban development. PowerTech,
2015 IEEE Eindhoven, 2015, p. 1–7.
[95] Department of Energy & Climate Change (DECC). 2050 Energy Calculator 2017.
http://2050-calculator-tool.decc.gov.uk/#/home.
[96] Mendes G, Ioakimidis C, Ferrao P. On the planning and analysis of Integrated
Community Energy Systems: A review and survey of available tools. Renew Sustain
[97] Marnay C, Stadler M, Siddiqui A, Deforest N, Donadee J, Bhattacharya P, et al.
Applications of Optimal Building Energy System Selection and Operation. Spec
Issue Inst Mech Eng J Power Energy 2013.
[98] Vogstad K. Utilising the complementary characteristics of wind power and
hydropower through coordinated ... Nord. Wind Power Conf., 2000.
[99] Kiss VM. Modelling the energy system of Pécs - The first step towards a
243 |
sustainable city. Energy 2015;80:373–87.
[100] EMD. EnergyPRO 2017. http://www.emd.dk/energypro/ (accessed December
25, 2015).
[101] University of Strathclyde. ESP-r 2017.
http://www.esru.strath.ac.uk/Programs/ESP-r.htm (accessed December 15,
2015).
[102] Drouet L, Thénié J. ETEM - An Energy/Technology/Environment Model to
Assess Urban Sustainable Development Policies - Reference Manual - Version 2.1.
Chêne-Bougeries: 2009.
[103] ORDECSYS. ETEM: Energy Technology Environment Model 2017.
http://apps.ordecsys.com/etem.
[104] Bakken BH, Skjelbred HI, Wolfgang O. eTransport: Investment planning in energy
supply systems with multiple energy carriers. Energy 2007;32:1676–89.
[105] Bakken BH, Skjelbred HI. Planning of distributed energy supply to suburb. 2007
IEEE Power Eng Soc Gen Meet PES 2007:6–13.
https://doi.org/10.1109/PES.2007.385786.
[106] U.S Department of Energy Office of Science, Argonne National Laboratory.
Generation and Transmission Maximization (GTMax) Model 2017.
http://ceeesa.es.anl.gov/projects/Gtmax.html%0D%0A.
[107] Duić N, da Graca Carvalho M. Increasing renewable energy sources in island
energy supply : case study Porto Santo. Renew Sustain Energy Rev 2004;8:383–99.
[108] Neves D, Silva CA, Connors S. Design and implementation of hybrid renewable
energy systems on micro-communities: A review on case studies. Renew Sustain
[109] Chmiel Z, Bhattacharyya SC. Analysis of off-grid electricity system at Isle of Eigg
(Scotland): Lessons for developing countries. Renew Energy 2015;81:578–88.
https://doi.org/10.1016/j.renene.2015.03.061.
[110] Sinha S, Chandel SS. Review of recent trends in optimization techniques for solar
photovoltaic–wind based hybrid energy systems. Renew Sustain Energy Rev
244 |
[111] Baring-gould I. Hybrid2: The Hybrid System Simulation Model. 1996.
[112] Mills A, Al-Hallaj S. Simulation of hydrogen-based hybrid systems using Hybrid2.
Int J Hydrogen Energy 2004;29:991–9.
https://doi.org/10.1016/j.ijhydene.2004.01.004.
[113] Ulleberg O, Moerkved A. Renewable Energy and Hydrogen System Concepts for
Remote communities in the West Nordic Region - The Nolsoy Case Study 2008.
http://www.nordicenergy.org/wp-content/uploads/2006/01/vest-
norden_fas_i-iii-rapport_2008-06.pdf (accessed December 4, 2017).
[114] Phrakonkham S, Le Chenadec J-Y, Diallo D, Marchand C. Optimization software
tool review and the need of alternative means for handling the problems of excess
energy and mini-grid configuration: a case study from Laos. 2009 ASEAN Symp.
Power Energy Syst., 2009.
[115] University of Zaragoza. iHOGA software 2017. https://ihoga-software.com/en/
(accessed July 7, 2017).
[116] Sinha S, Chandel SS. Review of software tools for hybrid renewable energy
systems. Renew Sustain Energy Rev 2014;32:192–205.
[117] Ragwitz M, Brakhage A, Kranzl L, Stadler M, Huber C, Eeg RH, et al. Final Report
of Work Phase 6 of the Project Invert. 2005.
[118] Baetens R, De Coninck R, Jorissen F, Picard D, Helsen L, Saelens D. OpenIDEAS
– an Open Framework for Integrated District Energy Simulations. Proc. BS2015,
14th Conf. Int. Build. Perform. Simul. Assoc., 2015, p. 347–54.
[119] Baetens R, De Coninck R, Van Roy J, Verbruggen B, Driesen J, Helsen L, et al.
Assessing electrical bottlenecks at feeder level for residential net zero-energy
buildings by integrated system simulation. Appl Energy 2012;96:74–83.
[120] Wirth S, Wang H, Yin W, Abdollahi E, Lahdelma R, Jiao W, et al. Heat Pumps in
District Heating: Final report. Renew Sustain Energy Rev 2015;52:399–405.
[121] Swan LG, Ugursal VI. Modeling of end-use energy consumption in the residential
sector: A review of modeling techniques. Renew Sustain Energy Rev
245 |
[122] Comodi G, Cioccolanti L, Gargiulo M. Municipal scale scenario: Analysis of an
Italian seaside town with MarkAL-TIMES. Energy Policy 2012;41:303–15.
[123] Faraji-Zonooz MR, Nopiah ZM, Yusof AM, Sopian M. A review of MARKAL
energy modeling A Review of MARKAL Energy Modeling. Eur J Sci Res
2009;26:352–61.
[124] Born FJ. Aiding Renewable Energy Integration through Complimentary Demand-
Supply Matching. University of Strathclyde, 2001.
[125] Prasad RD, Bansal RC, Raturi A. Multi-faceted energy planning: A review. Renew
Sustain Energy Rev 2014;38:686–99. https://doi.org/10.1016/j.rser.2014.07.021.
[126] Cai YP, Huang GH, Lin QG, Nie XH, Tan Q. Expert Systems with Applications
An optimization-model-based interactive decision support system for regional
energy management systems planning under uncertainty. Expert Syst Appl
2009;36:3470–82. https://doi.org/10.1016/j.eswa.2008.02.036.
[127] Henning D. Cost minimization for a local utility through CHP, heat storage and
load management. Int J Energy Res 1998;22:691–713.
https://doi.org/10.1002/(SICI)1099-114X(19980625)22:8<691::AID-
ER395>3.0.CO;2-E.
[128] Henning D. MODEST - An energy-system optimisation model applicable to local
utilities and countries. Energy 1997;22:1135–50.
[129] NEPLAN. NEPLAN 2017. http://www.neplan.ch/ (accessed December 4,
2017).
[130] Olsthoorn D, Haghighat F, Mirzaei PA. Integration of storage and renewable
energy into district heating systems: A review of modelling and optimization. Sol
Energy 2016;136:49–64. https://doi.org/10.1016/j.solener.2016.06.054.
[131] Carpaneto E, Lazzeroni P, Repetto M. Optimal integration of solar energy in a
district heating network. Renew Energy 2015;75:714–21.
[132] Hadley S, Hirst E. The oak Ridge Competitive Electricity Dispatch (ORCED)
Model. 2008. https://doi.org/ONRL/CON-464.
[133] Vela Solaris AG. Polysun simulaton software 2017. http://www.velasolaris.com/
246 |
[134] Blok K, Jager D de, Hendriks C. Economic Evaluation of Sectoral Emission
Reduction Objectives for Climate Change Summary Report for Policy Makers
2001.
http://ec.europa.eu/environment/enveco/climate_change/pdf/summary_repor
t_policy_makers.pdf.
[135] SINTEF. ProdRisk 2017. https://www.sintef.no/en/software/prodrisk/
[136] Lambert T, Gilman P, Lilienthal P. Micropower System Modeling with Homer. In:
Farret FA, Simoes MG, editors. Integr. Altern. Sources Energy, 2006 John Wiley
& Sons, Inc; 2006, p. 379–418. https://doi.org/10.1002/0471755621.ch15.
[137] Choi J, Yun R. Operation Strategy and Parametric Analysis of a CHP and a Tri-
Generation System for Integrated Community. Int J Air-Conditioning Refrig
2015;23:1550001. https://doi.org/10.1142/S2010132515500017.
[138] Mancarella P. MES (multi-energy systems): An overview of concepts and
evaluation models. Energy 2014;65:1–17.
[139] Herbergs S, Lehmann H, Peter S. THE COMPUTER-MODELLED
SIMULATION OF RENEWABLE ELECTRICITY NETWORKS. 2017.
[140] Technical University of Denmark. STREAM - an energy scenario modelling tool
2017. http://www.streammodel.org/ (accessed December 4, 2017).
[141] Ancona MA, Bianchi M, Branchini L, Melino F. District heating network design
and analysis. Energy Procedia 2014;45:1225–34.
[142] Schneider Electric Software LLC. Termis District Energy Optimization Software
2017. http://software.schneider-electric.com/products/termis/ (accessed
December 4, 2017).
[143] Energy and Utilities Alliance. The Heating and Hot water Industry Council trade
article - The rise of District Heating 2016. https://www.eua.org.uk/the-heating-
and-hotwater-industry-council-trade-article-the-rise-of-district-heating/ (accessed
July 19, 2017).
[144] Burohappold Engineering. Summary Report: UK Spatial District Heating Analysis.
247 |
2016.
[145] Euroheat. District Heating and Cooling Statistics 2015. 2015.
[146] Lund H, Østergaard PA, Connolly D, Ridjan I, Mathiesen BV, Hvelplund F, et al.
Energy Storage and Smart Energy Systems. Int J Sustain Energy Plan Manag
2016;11:3–14. https://doi.org/10.5278/ijsepm.2016.11.2.
[147] Dufo-Lopez R, Cristobal-Monreal IR, Yusta JM. Optimisation of PV-wind-diesel-
battery stand-alone systems to minimise cost and maximise human development
index and job creation. Renew Energy 2016;94:280–93.
[148] Baños R, Manzano-Agugliaro F, Montoya FG, Gil C, Alcayde A, Gómez J.
Optimization methods applied to renewable and sustainable energy: A review.
Renew Sustain Energy Rev 2011;15:1753–66.
https://doi.org/10.1016/J.RSER.2010.12.008.
[149] Iqbal M, Azam M, Naeem M, Khwaja AS, Anpalagan A. Optimization
classification, algorithms and tools for renewable energy: A review. Renew Sustain
Energy Rev 2014;39:640–54. https://doi.org/10.1016/J.RSER.2014.07.120.
[150] Manwell JF, Rogers A, Hayman G, Avelar CT, Mcgowan JG, Abdulwahid U, et al.
Hybrid2 - a hybrid system simulation model: theory manual. National Renewable
Energy Laboratory; 2006.
[151] HOMER Energy. Create Your Own Microgrid Control Strategies with HOMER
Pro APIs - HOMER Microgrid News and Insight : HOMER Microgrid News and
Insight 2017. http://microgridnews.com/create-microgrid-control-strategies/
[152] International Energy Agency. Global EV Outlook 2016 Electric Vehicles
Initiative. Iea 2016:51. https://doi.org/EIA-0383(2016).
[153] Urban Foresight for Transport Scotland. Energy Systems and Electric Vehicles.
2016.
[154] Navigant Research. Vehicle Grid Integration: VGI Applications for Demand
Response, Frequency Regulation, Microgrids, Virtual Power Plants, and
Renewable Energy Integration 2017.
https://www.navigantresearch.com/research/vehicle-grid-integration (accessed
July 12, 2017).
248 |
[155] García-Villalobos J, Zamora I, San Martín JI, Junquera I, Eguía P. Delivering
Energy from PEV batteries: V2G, V2B and V2H approaches. Int Conf Renew
Energies Power Qual 2015.
[156] Manwell JF, McGowan JG. Lead acid battery storage model for hybrid energy
systems. Sol Energy 1993;50:399–405.
[157] HOMER Energy. HOMER Pro Version 3.7 User Manual. HOMER Energy
2016:416.
[158] Manwell JF, McGowan JG. Extension of the kinetic battery model for
wind/hybrid power systems. Proc 5th Eur Wind Energy Assoc Conf (EWEC ’94)
1994:284–289.
[159] Manwell JF, McGowan JG, Abdulwahid U, Wu K. Improvements to the Hybrid2
Battery Model. Am Wind Energy Assoc Wind 2005 Conf 2005:1–22.
[160] Downing SD, Socie DF. Simple rainflow counting algorithms. Int J Fatigue
1982;4:31–40. https://doi.org/10.1016/0142-1123(82)90018-4.
[161] Copetti JB, Chenlo F. Lead/acid batteries for photovoltaic applications. Test
results and modeling. J Power Sources 1994;47:109–18.
https://doi.org/10.1016/0378-7753(94)80054-5.
[162] Schiffer J, Sauer DU, Bindner H, Cronin T, Lundsager P, Kaiser R. Model
prediction for ranking lead-acid batteries according to expected lifetime in
renewable energy systems and autonomous power-supply systems. J Power
Sources 2007;168:66–78. https://doi.org/10.1016/j.jpowsour.2006.11.092.
[163] Groot J, Swierczynski M, Irina A, Knudsen S. On the complex ageing
characteristics of high-power LiFePO 4 / graphite battery cells cycled with high
charge and discharge currents. J Power Sources 2015;286:475–87.
https://doi.org/10.1016/j.jpowsour.2015.04.001.
[164] Saxena S, Hendricks C, Pecht M. Cycle life testing and modeling of
graphite/LiCoO2 cells under different state of charge ranges. J Power Sources
2016;327:394–400. https://doi.org/10.1016/j.jpowsour.2016.07.057.
[165] Wang J, Liu P, Hicks-Garner J, Sherman E, Soukiazian S, Verbrugge M, et al.
Cycle-life model for graphite-LiFePO4 cells. J Power Sources 2011;196:3942–8.
https://doi.org/10.1016/j.jpowsour.2010.11.134.
[166] Stadler M, Marnay C, Siddiqui AS, Lai J, Aki H. Integrated Building Energy
249 |
Systems Design Considering Storage Technologies. 2009.
[167] Lund H, Salgi G. The role of compressed air energy storage (CAES) in future
sustainable energy systems. Energy Convers Manag 2009;50:1172–9.
[168] Steen D, Stadler M, Cardoso G, Groissböck M, DeForest N, Marnay C. Modeling
of thermal storage systems in MILP distributed energy resource models. Appl
[169] Bejan A. Two Thermodynamic Optima in the Design of Sensible Heat Units for
Energy Storage. J Heat Transfer 1978;100:708.
https://doi.org/10.1115/1.3450882.
[170] Ridjan I, Mathiesen BV, Connolly D, Duić N. The feasibility of synthetic fuels in
renewable energy systems. Energy 2013;57:76–84.
https://doi.org/10.1016/J.ENERGY.2013.01.046.
[171] Openmod. Open Models - wiki.openmod-initiative.org n.d.
https://wiki.openmod-initiative.org/wiki/Open_Models (accessed April 14,
2020).
[172] Tuohy P, Murphy GB. Are current design processes and policies delivering
comfortable low carbon buildings? Archit Sci Rev 2015;58:39–46.
https://doi.org/10.1080/00038628.2014.975779.
[173] Tuohy P, Murphy GB. Closing the gap in building performance: learning from
BIM benchmark industries. Archit Sci Rev 2015;58:47–56.
[174] Lin H, Lai A, Ullrich R, Kuca M, McClelland K, Shaffer-Gant J, et al. COTS
software selection process. Proc - ICCBSS 2007 Sixth Int IEEE Conf Commer
(COTS)-Based Softw Syst 2007:114–20.
https://doi.org/10.1109/ICCBSS.2007.11.
[175] Jong K De, Hernandez ME, Post DS, Taylor JL. Process for Selecting Engineering
Tools – Applied to Selecting a SysML Tool 2011.
[176] OPENMOD. openmod - Open Energy Modelling Initiative. 2014-10-03 2014.
[177] Pfenninger S, Hirth L, Schlecht I, Schmid E, Wiese F, Brown T, et al. Opening the
black box of energy modelling: Strategies and lessons learned. Energy Strateg Rev
2018;19:63–71. https://doi.org/10.1016/j.esr.2017.12.002.
250 |
[178] Pfenninger S, DeCarolis J, Hirth L, Quoilin S, Staffell I. The importance of open
data and software: Is energy research lagging behind? Energy Policy 2017;101:211–
5. https://doi.org/10.1016/j.enpol.2016.11.046.
[179] Pfenninger S. Energy scientists must show their workings. Nature 2017;542:393.
https://doi.org/10.1038/542393a.
[180] Millman KJ, Aivazis M. Python for scientists and engineers. Comput Sci Eng
2011;13:9–12. https://doi.org/10.1109/MCSE.2011.36.
[181] F. Holmgren W, W. Hansen C, A. Mikofski M. pvlib python: a python package for
modeling solar energy systems. J Open Source Softw 2018;3:884.
https://doi.org/10.21105/joss.00884.
[182] SabineHaas, Schachler B, Krien U, Bosch S. wind-python/windpowerlib: The new
restructuring release 2017. https://doi.org/10.5281/ZENODO.824268.
[183] Brown T, Hörsch J, Schlachtberger D. PyPSA: Python for power system analysis.
J Open Res Softw 2018;6. https://doi.org/10.5334/jors.188.
[184] Thurner L, Scheidler A, Schafer F, Menke JH, Dollichon J, Meier F, et al.
Pandapower - An Open-Source Python Tool for Convenient Modeling, Analysis,
and Optimization of Electric Power Systems. IEEE Trans Power Syst
2018;33:6510–21. https://doi.org/10.1109/TPWRS.2018.2829021.
[185] Brown T, Schäfer M, Greiner M. Sectoral interactions as carbon dioxide emissions
approach zero in a highly-renewable european energy system. Energies 2019;12.
https://doi.org/10.3390/en12061032.
[186] Hilpert S, Kaldemeyer C, Krien U, Günther S, Wingenbach C, Plessmann G. The
Open Energy Modelling Framework (oemof) - A new approach to facilitate open
science in energy system modelling. Energy Strateg Rev 2018;22:16–25.
https://doi.org/10.1016/j.esr.2018.07.001.
[187] Howells M, Rogner H, Strachan N, Heaps C, Huntington H, Kypreos S, et al.
OSeMOSYS: The Open Source Energy Modeling System. An introduction to its
ethos, structure and development. Energy Policy 2011;39:5850–70.
[188] National Energy Foundation. Renewable Energy Feasibility Studies - National
Energy Foundation n.d. http://www.nef.org.uk/service/existing-
buildings/energy-management/renewable-energy-feasibility-studies (accessed
251 |
April 27, 2020).
[189] Morris P, Therivel R. Methods of environmental impact assessment: Third edition.
vol. 9780203892. 2009. https://doi.org/10.4324/9780203892909.
[190] Luna-Rubio R, Trejo-Perea M, Vargas-Vázquez D, Ríos-Moreno GJ. Optimal
sizing of renewable hybrids energy systems: A review of methodologies. Sol
Energy 2012;86:1077–88. https://doi.org/10.1016/j.solener.2011.10.016.
[191] Yang HX, Lu L, Burnett J. Weather data and probability analysis of hybrid
photovoltaic-wind power generation systems in Hong Kong. Renew Energy
2003;28:1813–24. https://doi.org/10.1016/S0960-1481(03)00015-6.
[192] Celik AN. Techno-economic analysis of autonomous PV-wind hybrid energy
systems using different sizing methods. Energy Convers Manag 2003;44:1951–68.
[193] Tina G, Gagliano S, Raiti S. Hybrid solar/wind power system probabilistic
modelling for long-term performance assessment. Sol Energy 2006;80:578–88.
https://doi.org/10.1016/j.solener.2005.03.013.
[194] HOMER Energy. HOMER Pro 2014:2017.
[195] Whitley D. A genetic algorithm tutorial. Stat Comput 1994;4:65–85.
https://doi.org/10.1007/BF00175354.
[196] Zeng R, Li H, Liu L, Zhang X, Zhang G. A novel method based on multi-
population genetic algorithm for CCHP-GSHP coupling system optimization.
Energy Convers Manag 2015;105:1138–48.
[197] Lyden A, Tuohy P. A methodology for designing decentralised energy systems with
predictive control for heat pumps and thermal storage. E3S Web Conf
2019;111:06014. https://doi.org/10.1051/e3sconf/201911106014.
[198] Met Office. Weather and climate data - Met Office. Website 2019.
https://www.metoffice.gov.uk/services/data (accessed April 16, 2020).
[199] Pfenninger S, Staffell I. Long-term patterns of European PV output using 30 years
of validated hourly reanalysis and satellite data. Energy 2016;114:1251–65.
[200] Ramirez Camargo L, Gruber K, Nitsch F. Assessing variables of regional reanalysis
data sets relevant for modelling small-scale renewable energy systems. Renew
252 |
Energy 2019;133:1468–78. https://doi.org/10.1016/j.renene.2018.09.015.
[201] Renewables.ninja n.d. https://www.renewables.ninja/ (accessed November 22,
2018).
[202] Rienecker MM, Suarez MJ, Gelaro R, Todling R, Bacmeister J, Liu E, et al.
MERRA: NASA’s modern-era retrospective analysis for research and applications.
J Clim 2011;24:3624–48. https://doi.org/10.1175/JCLI-D-11-00015.1.
[203] Staffell I, Pfenninger S. Using bias-corrected reanalysis to simulate current and
future wind power output. Energy 2016;114:1224–39.
[204] CIBSE. Energy benchmarks. TM46. 2008. https://doi.org/1122.
[205] Moretti L, Polimeni S, Meraldi L, Raboni P, Leva S, Manzolini G. Assessing the
impact of a two-layer predictive dispatch algorithm on design and operation of off-
grid hybrid microgrids. Renew Energy 2019;143:1439–53.
[206] Bahramara S, Moghaddam MP, Haghifam MR. Optimal planning of hybrid
renewable energy systems using HOMER: A review. Renew Sustain Energy Rev
[207] Talebi B, Mirzaei PA, Bastani A, Haghighat F. A Review of District Heating
Systems: Modeling and Optimization. Front Built Environ 2016;2:22.
https://doi.org/10.3389/fbuil.2016.00022.
[208] Home / isoplus Piping Systems n.d. http://en.isoplus.dk/ (accessed November
26, 2018).
[209] Energy Savings Trust. Measurement of domestic hot water consumption in
dwellings. 2008.
[210] Ahmed K, Pylsy P, Kurnitski J. Hourly consumption profiles of domestic hot
water for different occupant groups in dwellings. Sol Energy 2016.
[211] Henderson J, Hart J. BREDEM 2012-A technical description of the BRE
Domestic Energy Model. 2013.
[212] BRE. District Heating - Heat Metering Cost Benefit Analysis. 2012.
[213] Flett G, Kelly N. A disaggregated, probabilistic, high resolution method for
assessment of domestic occupancy and electrical demand. Energy Build
253 |
[214] Sabine Haas. Implementation and validation of an open source model for
generating wind feed-in time series. Reiner-Lemoine-InstitutDe 2019.
[215] Holmgren WF, Andrews RW, Lorenzo AT, Stein JS. PVLIB Python 2015. 2015
IEEE 42nd Photovolt. Spec. Conf. PVSC 2015, Institute of Electrical and
Electronics Engineers Inc.; 2015. https://doi.org/10.1109/PVSC.2015.7356005.
[216] Petersen M. Wind turbine test Vestas V27-225 kW. 1990.
[217] Case Studies - WIND ENERGY STORAGE n.d.
http://www.esru.strath.ac.uk/EandE/Web_sites/17-18/windies/case-
studies.html (accessed April 21, 2020).
[218] Newmiller J, Hutchinson P, Townsend T, Whitaker C. PVUSA instrumentation
and data analysis techniques for photovoltaic systems. Albuquerque, NM: 1995.
[219] Stein JS, Holmgren WF, Forbess J, Hansen CW. PVLIB: Open source
photovoltaic performance modeling functions for Matlab and Python. Conf. Rec.
IEEE Photovolt. Spec. Conf., vol. 2016- Novem, Institute of Electrical and
Electronics Engineers Inc.; 2016, p. 3425–30.
https://doi.org/10.1109/PVSC.2016.7750303.
[220] Šúri M, Huld T, Cebecauer T, Dunlop ED. Geographic aspects of photovoltaics
in Europe: Contribution of the PVGIS website. IEEE J Sel Top Appl Earth Obs
Remote Sens 2008;1:34–41. https://doi.org/10.1109/JSTARS.2008.2001431.
[221] BRE Group: Report: Photovoltaics Field Trial n.d.
https://www.bre.co.uk/page.jsp?id=859 (accessed April 23, 2020).
[222] Uhlmann M, Bertsch SS. Theoretical and experimental investigation of startup and
shutdown behavior of residential heat pumps Etude théorique et expérimentale
sur le comportement des pompes à chaleur résidentielles lors du démarrage et de
l’arrêt. Int. J. Refrig., vol. 35, Elsevier; 2012, p. 2138–49.
https://doi.org/10.1016/j.ijrefrig.2012.08.008.
[223] Madonna F, Bazzocchi F. Annual performances of reversible air-to-water heat
pumps in small residential buildings. Energy Build 2013;65:299–309.
[224] Amer M, Wang CC. Review of defrosting methods. Renew Sustain Energy Rev
254 |
[225] Song M, Dong J, Wu C, Jiang Y, Qu M. Improving the frosting and defrosting
performance of air source heat pump units: review and outlook. HKIE Trans
Hong Kong Inst Eng 2017;24:88–98.
https://doi.org/10.1080/1023697X.2017.1313134.
[226] Sarbu I, Sebarchievici C. General review of ground-source heat pump systems for
heating and cooling of buildings. Energy Build 2014;70:441–54.
[227] Staffell I, Brett D, Brandon N, Hawkes A. A review of domestic heat pumps.
Energy Environ Sci 2012;5:9291–306. https://doi.org/10.1039/c2ee22653g.
[228] EMD International. energyPRO - User Guide 2001.
[229] BS EN 14511 2018.
https://shop.bsigroup.com/ProductDetail/?pid=000000000030328748 (accessed
November 27, 2018).
[230] Duffie JA, Beckman WA. Solar Engineering of Thermal Processes: Fourth
Edition. 2013. https://doi.org/10.1002/9781118671603.
[231] TESS Component Library Package - TESS Libraries | TRNSYS : Transient
System Simulation Tool n.d. http://www.trnsys.com/tess-libraries/ (accessed
April 23, 2020).
[232] Woods P. Heat Networks: Code of Practice for the UK - Raising Standards for
Heat Supply. CIBSE, Chartered Institute of Building Services Engineers; 2015.
[233] Dumont O, Carmo C, Dickes R, Emelines G, Quoilin S, Lemort V. Hot water
tanks : How to select the optimal modelling approach? CLIMA 2016, AAlborg
2016.
[234] scipy.integrate.odeint — SciPy v1.4.1 Reference Guide n.d.
https://docs.scipy.org/doc/scipy/reference/generated/scipy.integrate.odeint.ht
ml (accessed April 21, 2020).
[235] Cadau N, De Lorenzi A, Gambarotta A, Morini M, Rossi M. Development and
Analysis of a Multi-Node Dynamic Model for the Simulation of Stratified Thermal
Energy Storage. Energies 2019;12:4275. https://doi.org/10.3390/en12224275.
[236] Oldewurtel F, Parisio A, Jones CN, Gyalistras D, Gwerder M, Stauch V, et al. Use
255 |
of model predictive control and weather forecasts for energy efficient building
climate control. Energy Build 2012;45:15–27.
[237] Rehrl J, Horn M. Temperature control for HVAC systems based on exact
linearization and model predictive control. Proc. IEEE Int. Conf. Control Appl.,
2011, p. 1119–24. https://doi.org/10.1109/CCA.2011.6044437.
[238] Karlsson H, Hagentoft CE. Application of model based predictive control for
water-based floor heating in low energy residential buildings. Build Environ
2011;46:556–69. https://doi.org/10.1016/j.buildenv.2010.08.014.
[239] Beal LDR, Hill DC, Abraham Martin R, Hedengren JD. GEKKO optimization
suite. Processes 2018;6:106. https://doi.org/10.3390/pr6080106.
[240] APOPT Solutions n.d. https://apopt.com/index.php (accessed April 22, 2020).
[241] Powell KM, Edgar TF. An adaptive-grid model for dynamic simulation of
thermocline thermal energy storage systems. Energy Convers Manag 2013;76:865–
73. https://doi.org/10.1016/j.enconman.2013.08.043.
[242] Powell KM, Hedengren JD, Edgar TF. Dynamic optimization of a hybrid solar
thermal and fossil fuel system. Sol Energy 2014;108:210–8.
[243] Ventosa M, Baíllo Á, Ramos A, Rivier M. Electricity market modeling trends.
Energy Policy 2005;33:897–913. https://doi.org/10.1016/j.enpol.2003.10.013.
[244] Schwartz E, Smith JE. Short-term variations and long-term dynamics in
commodity prices. Manage Sci 2000;46:893–911.
https://doi.org/10.1287/mnsc.46.7.893.12034.
[245] Fischer D, Lindberg KB, Madani H, Wittwer C. Impact of PV and variable prices
on optimal system sizing for heat pumps and thermal storage. Energy Build
[246] TERRE n.d. https://www.entsoe.eu/network_codes/eb/terre/ (accessed April
23, 2020).
[247] Maasoumy M, Razmara M, Shahbakhti M, Vincentelli AS. Handling model
uncertainty in model predictive control for energy efficient buildings. Energy Build
[248] Oldewurtel F, Parisio A, Jones CN, Morari M, Gyalistras D, Gwerder M, et al.
256 |
Energy efficient building climate control using Stochastic Model Predictive
Control and weather predictions. Proc. 2010 Am. Control Conf. ACC 2010, 2010,
p. 5100–5. https://doi.org/10.1109/acc.2010.5530680.
[249] Polo J, Alonso-Abella M, Martín-Chivelet N, Alonso-Montesinos J, López G,
Marzo A, et al. Typical Meteorological Year methodologies applied to solar spectral
irradiance for PV applications. Energy 2020;190.
https://doi.org/10.1016/j.energy.2019.116453.
[250] Bie Z, Lin Y, Li G, Li F. Battling the Extreme: A Study on the Power System
Resilience. Proc IEEE 2017;105:1253–66.
https://doi.org/10.1109/JPROC.2017.2679040.
[251] Yilmaz S, Chambers J, Patel MK. Comparison of clustering approaches for
domestic electricity load profile characterisation - Implications for demand side
management. Energy 2019;180:665–77.
[252] Kaldellis JK, Kondili E, Filios A. Sizing a hybrid wind-diesel stand-alone system
on the basis of minimum long-term electricity production cost. Appl Energy
2006;83:1384–403. https://doi.org/10.1016/j.apenergy.2006.01.006.
[253] Gupta A, Saini RP, Sharma MP. Steady-state modelling of hybrid energy system
for off grid electrification of cluster of villages. Renew Energy 2010;35:520–35.
[254] Dufo-López R, Cristóbal-Monreal IR, Yusta JM. Optimisation of PV-wind-diesel-
battery stand-alone systems to minimise cost and maximise human development
index and job creation. Renew Energy 2016;94:280–93.
[255] Zhou D, Zhao CY, Tian Y. Review on thermal energy storage with phase change
materials (PCMs) in building applications. Appl Energy 2012;92:593–605.
[256] Baetens R, Jelle BP, Gustavsen A. Phase change materials for building applications:
A state-of-the-art review. Energy Build 2010;42:1361–8.
[257] Home - Mixergy Hot Water Tanks n.d. https://www.mixergy.co.uk/ (accessed
April 23, 2020).
257 |
[258] Chen Y, Xu P, Gu J, Schmidt F, Li W. Measures to improve energy demand
flexibility in buildings for demand response (DR): A review. Energy Build
[259] Vögelin P, Koch B, Georges G, Boulouchos K. Heuristic approach for the
economic optimisation of combined heat and power (CHP) plants: Operating
strategy, heat storage and power. Energy 2017;121:66–77.
[260] Yue X, Pye S, DeCarolis J, Li FGN, Rogan F, Gallachóir B. A review of approaches
to uncertainty assessment in energy system optimization models. Energy Strateg
Rev 2018;21:204–17. https://doi.org/10.1016/j.esr.2018.06.003.
[261] Lyden A, Tuohy P. Heat pump and thermal storage sizing with time-of-use
electricity pricing, Atlantis Press; 2019, p. 29–40. https://doi.org/10.2991/ires-
19.2019.4.
[262] District Heating | West Whitlawburn Housing Co-operative Ltd n.d.
http://www.wwhc.org.uk/our-properties/district-heating/ (accessed April 28,
2020).
[263] In the hot seat: Interview with Dave Pearson – CIBSE Journal n.d.
https://www.cibsejournal.com/technical/in-the-hot-seat-interview-with-dave-
pearson/ (accessed April 28, 2020).
[264] Neatpump | Star Refrigeration Products n.d. https://www.star-ref.co.uk/our-
products/neatpump.aspx (accessed January 25, 2020).
[265] Sarbu I, Sebarchievici C. A comprehensive review of thermal energy storage.
Sustain 2018;10. https://doi.org/10.3390/su10010191.
[266] Joos M, Staffell I. Short-term integration costs of variable renewable energy: Wind
curtailment and balancing in Britain and Germany. Renew Sustain Energy Rev
[267] Renewable Energy Foundation. A Decade of Constraint Payments 2019.
https://www.ref.org.uk/ref-blog/354-a-decade-of-constraint-payments (accessed
June 9, 2020).
[268] Danish Energy Agency. Technology Data 2018. https://ens.dk/en/our-
services/projections-and-models/technology-data (accessed March 12, 2019).
[269] Bhandari M, Shrestha S, New J. Evaluation of weather datasets for building energy
258 |
simulation. Energy Build 2012;49:109–18.
[270] Hong T, Chang WK, Lin HW. A fresh look at weather impact on peak electricity
demand and energy use of buildings using 30-year actual weather data. Appl
[271] del Hoyo Arce I, Herrero López S, López Perez S, Rämä M, Klobut K, Febres JA.
Models for fast modelling of district heating and cooling networks. Renew Sustain
[272] Graber M, Kosowski K, Richter C, Tegethoff W. Modelling of heat pumps with
an object-oriented model library for thermodynamic systems. Math Comput Model
Dyn Syst 2010;16:195–209. https://doi.org/10.1080/13873954.2010.506799.
[273] Griffith BT, Ellis PG. Photovoltaic and Solar Thermal Modeling with the
EnergyPlus Calculation Engine. World Renew Energy Congr VIII Expo, 2004:1–
5.
[274] Dubey S, Tiwari GN. Thermal modeling of a combined system of photovoltaic
thermal (PV/T) solar water heater. Sol Energy 2008;82:602–12.
[275] De Césaro Oliveski R, Krenzinger A, Vielmo HA. Comparison between models
for the simulation of hot water storage tanks. Sol Energy 2003;75:121–34.
[276] Blandin D, Caccavelli D, Krauss G, Bouia H. A zonal approach for modeling
stratified solar tanks. IBPSA 2007 - Int. Build. Perform. Simul. Assoc. 2007, 2007,
p. 678–83.
[277] Manwell JF, McGowan JG. Lead acid battery storage model for hybrid energy
systems. Sol Energy 1993;50:399–405. https://doi.org/10.1016/0038-
092X(93)90060-2.
[278] HOMER Energy. HOMER - Hybrid Renewable and Distributed Generation
System Design Software 2015.
[279] Lopez RD. iHOGA User’s manual. 2017.
[280] Schill WP, Kemfert C. Modeling strategic electricity storage: The case of pumped
hydro storage in Germany. Energy J 2011;32:59–87.
https://doi.org/10.5547/ISSN0195-6574-EJ-Vol32-No3-3.
259 |
[281] Hartmann N, Vöhringer O, Kruck C, Eltrop L. Simulation and analysis of different
adiabatic Compressed Air Energy Storage plant configurations. Appl Energy
2012;93:541–8. https://doi.org/10.1016/j.apenergy.2011.12.007.
[282] Lagorse J, Paire D, Miraoui A. Sizing optimization of a stand-alone street lighting
system powered by a hybrid system using fuel cell, PV and battery. Renew Energy
2009;34:683–91. https://doi.org/10.1016/j.renene.2008.05.030.
[283] Jensen SH, Larsen PH, Mogensen M. Hydrogen and synthetic fuel production
from renewable energy sources. Int J Hydrogen Energy 2007;32:3253–7.
https://doi.org/10.1016/j.ijhydene.2007.04.042.
[284] Mathiesen BV, Lund H. Comparative analyses of seven technologies to facilitate
the integration of fluctuating renewable energy sources. IET Renew Power Gener
2009;3:190. https://doi.org/10.1049/iet-rpg:20080049.
[285] Lund H. EnergyPLAN Advanced Energy Systems Analysis Computer Model
2015.
[286] Octopus energy. Octopus Go Frequently Asked Questions | Octopus Energy n.d.
https://octopus.energy/blog/outgoing/ (accessed May 13, 2020).
[287] Kircher KJ, Zhang KM. Model predictive control of thermal storage for demand
response. Proc. Am. Control Conf., vol. 2015- July, Institute of Electrical and
Electronics Engineers Inc.; 2015, p. 956–61.
https://doi.org/10.1109/ACC.2015.7170857.
[288] Henze GP, Dodier RH, Krarti M. Development of a predictive optimal controller
for thermal energy storage systems. HVAC R Res 1997;3:233–64.
[289] Afram A, Janabi-Sharifi F. Theory and applications of HVAC control systems - A
review of model predictive control (MPC). Build Environ 2014;72.
https://doi.org/10.1016/j.buildenv.2013.11.016.
[290] Dehkordi VR, Candanedo JA. A model-based predictive control approach for a
building cooling system with ice storage. ASME 2013 Dyn. Syst. Control Conf.
DSCC 2013, vol. 1, American Society of Mechanical Engineers (ASME); 2013.
https://doi.org/10.1115/DSCC2013-3897.
[291] Pezeshki H, Wolfs P, Ledwich G. A model predictive approach for community
battery energy storage system optimization. IEEE Power Energy Soc. Gen. Meet.,
260 |
vol. 2014- Octob, IEEE Computer Society; 2014.
https://doi.org/10.1109/PESGM.2014.6938788.
[292] Pereira M, Muñoz de la Peña D, Limon D. Robust economic model predictive
control of a community micro-grid. Renew Energy 2017;100:3–17.
[293] Parisio A, Glielmo L. Stochastic Model Predictive Control for
economic/environmental operation management of microgrids. 2013 Eur.
Control Conf. ECC 2013, 2013, p. 2014–9.
https://doi.org/10.23919/ecc.2013.6669807.
261 |

Modelling and Design of Local Energy Systems Incorporating Heat Pumps, Thermal Storage, Future Tariffs, and Model Predictive Control

Uploaded by

Copyright:

Available Formats

Modelling and Design of Local Energy Systems Incorporating Heat Pumps, Thermal Storage, Future Tariffs, and Model Predictive Control

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Modelling and Design of Local Energy Systems Incorporating Heat Pumps, Thermal Storage, Future Tariffs, and Model Predictive Control

Uploaded by

Copyright:

Available Formats

Modelling and Design of Local Energy

Systems Incorporating Heat Pumps,

Andrew Lyden B.Sc. M.Sc.

A thesis submitted for the degree of Doctor of Philosophy

Energy Systems Research Unit

1.1. General Context......................................................................................................... 19

1.2. Transitioning Energy System ................................................................................... 20

1.3. Local Energy Systems ............................................................................................... 22

1.4. Heat Pumps ................................................................................................................ 23

1.5. Thermal and Electrical Storage................................................................................ 24

1.6. Electricity Markets ..................................................................................................... 25

1.7. Control Strategies ...................................................................................................... 25

1.8. Modelling Local Energy Systems ............................................................................ 27

1.9. Uncertainty in Energy System Modelling............................................................... 30

1.10. Related Projects ..................................................................................................... 32

1.11. Problem Statement ................................................................................................ 33

2. Research Question, Aims, and Methodology ................................................ 34

2.1. Research Question..................................................................................................... 34

2.2. Research Aims............................................................................................................ 34

2.3. Research Methodology ............................................................................................. 35

3. Review of Existing Modelling Tools ............................................................... 37

3.1. Initial Screening Process ........................................................................................... 38

3.2. Categorisation of Modelling Tool Capabilities ...................................................... 42

3.2.2. Electrical and Thermal Supply Technology Modelling Capabilities ..................................... 45

3.2.3. Design Optimisation and Output Capabilities ......................................................................... 47

3.2.4. Control Modelling Capabilities including DSM ....................................................................... 49

3.2.5. Storage Modelling Capabilities and Underlying Models ......................................................... 53

3.2.7. Discussion ....................................................................................................................................... 63

3.3. Tool Selection Process .............................................................................................. 65

3.3.2. Scoring of Tools Against Requirements .................................................................................... 66

3.3.3. Example Application of the Modelling Tool Selection Process ........................................... 66

3.3.4. Discussion ....................................................................................................................................... 69

3.4. Gaps to be Addressed ............................................................................................... 70

4. Introducing PyLESA: Modelling Methodology, Capabilities, Framework, and

4.1. Brief Introduction to PyLESA ................................................................................ 71

4.2. Modelling Methodology ........................................................................................... 72

4.3. Motivation for Choosing Python ............................................................................ 73

4.4. Tool Capabilities ........................................................................................................ 74

4.5. Tool Framework ........................................................................................................ 78

4.6. Tool Applications: Potential Design Studies ......................................................... 80

4.6.2. Operation Study ............................................................................................................................. 81

4.6.3. Sizing Study .................................................................................................................................... 82

4.7. Summary and Next Steps ......................................................................................... 83

5. Detailing PyLESA: Underlying Models and Assessment Methods ................ 84

5.1. Resource and Demand Assessment and Input Methods..................................... 85

5.1.2. Electrical Demand ......................................................................................................................... 89

5.1.3. District Heating Demand ............................................................................................................. 90

5.2. Electrical Production Technologies ........................................................................ 94

5.3. Heat Pumps and Auxiliary Heat Units ................................................................... 98

5.3.2. Lorentz Model ..............................................................................................................................103

5.3.3. Standard Test Regression Model ..............................................................................................105

5.3.4. Auxiliary Heat Units and Schematic of Setup ........................................................................107

5.4. Electrical Storage .....................................................................................................108

5.5. Hot Water Tank .......................................................................................................110

5.5.2. Charging and Discharging ..........................................................................................................117

5.5.3. Mixing Between Nodes ..............................................................................................................118

5.5.4. Final Energy Balance Equation .................................................................................................119

5.5.5. Validation ......................................................................................................................................120

5.6. Electricity Tariffs .....................................................................................................129