Overview

Designing electricity markets

for a high penetration of variable
renewables
Jenny Riesz1∗ and Michael Milligan2

Renewable technologies are often characterized as being somewhat different to

‘conventional’ generating technologies in three ways, each with different impli-
cations for electricity markets. Firstly, some have highly variable and somewhat
uncertain availability, meaning that electricity markets must be designed to elicit
adequate flexibility. Secondly, many have very low short-run marginal costs (oper-
ating costs), meaning that the mechanisms for managing resource adequacy must
be carefully considered. Thirdly, some are nonsynchronous, meaning that grid
codes and regulatory requirements must be appropriately designed. Access to flex-
ibility can be enhanced by a range of market design choices, such as short dis-
patch intervals, short delays from gate closure to dispatch, large balancing areas,
high demand side participation, and exposing renewable technologies to market
price signals commensurate with other technologies. The design of markets for fre-
quency control ancillary services (FCAS) also provides opportunities to increase
access to flexibility, by creating active real-time markets for a wide range of FCAS,
allowing renewable technologies to provide FCAS, and determining FCAS reserve
requirements dynamically in real time. Mechanisms for managing resource ade-
quacy are a source of ongoing debate, with many of the key issues having been
exacerbated by the entry of renewables. Rapid market change makes investment
decisions difficult, regardless of the market model applied. Ultimately, given the
existence of arguably successful examples of both energy-only and capacity mar-
ket designs, the choice of market model may be less important than the quality of
governance with which it is implemented and maintained. © 2014 John Wiley & Sons,
Ltd.

How to cite this article:

WIREs Energy Environ 2014. doi: 10.1002/wene.137

INTRODUCTION of global generating capacity and more than 21%

of global electricity, and the capacity of renewables
E lectricity markets around the world are at the cusp
of a dramatic transformation. One of the most vis-
ible aspects of this change is the rapidly growing pen-
installed each year now exceeds that of all other fuel
sources combined.1 Thus, over the next few decades,
many electricity market operators can expect to be
etration of renewable technologies. Already, renew-
dealing with much higher amounts of renewable gen-
ables (including hydro) contribute more than 26%
eration. Owing to their relatively lower cost, much of
this new capacity will be variable renewables, such as
∗ Correspondence to: [email protected] wind and solar photovoltaics (PV).
1 Centre for Energy and Environmental Markets and School of The technical characteristics of a system unde-
Electrical Engineering and Telecommunications, University of New niably affect its ability to cost-effectively integrate
South Wales, Sydney, New South Wales, Australia
2 National Renewable Energy Laboratory, Transmission and Grid variable renewables. For example, larger power sys-
Integration, Denver, CO, USA tems with strong transmission interconnections to
Conflict of interest: The authors have declared no conflicts of neighbors are likely to find it lower cost to integrate
interest for this article. higher quantities of variable renewable generation,

© 2014 John Wiley & Sons, Ltd.

 Overview wires.wiley.com/wene

because variability can be smoothed over a larger geo-

graphic area, and managed with a larger selection of
Variable and
plant. High proportions of flexible generation such as uncertain
hydro power, aero derivative turbines, or reciprocat- Small
ing engines are also likely to alleviate many of the hydro
challenges. CST w/o
storage
However, it is becoming increasingly apparent PV
CST w/
that the technical parameters of a system are only storage
wind
Non-
one part of the story, and the design of electric-
Low SRMC synchronous
ity industry arrangements can also have an immense
Waste
influence.2–4 For many industries, these arrangements biomass
include wholesale electricity markets seeking to estab-
lish competitive operation of generation and invest- Geothermal
ment decision making. In some cases, poor market
design dramatically increases the costs and challenges FIGURE 1 | Characterizing renewable technologies. CST refers to
of integrating renewables.5,6 Thus, before investing ‘concentrating solar thermal’ technology.
in expensive physical infrastructure it makes sense to
examine the ways in which market design may be facil- sophisticated (and often somewhat experimen-
itating or inhibiting the most efficient system response. tal) power electronics they do not contribute
In some cases, comparatively simple changes to mar- to frequency stability, system inertia, and other
ket design may provide cost-effective opportunities to grid-related services in the same way as syn-
increase the ease and efficiency with which emerging chronous generation.
renewable technologies can be integrated into power
systems.
These properties of renewable technologies are
illustrated in Figure 1, with a characterization of some
THE DISTINGUISHING of the common renewable generator types. Note that
not all renewable technologies exhibit all of these char-
CHARACTERISTICS OF VARIABLE acteristics, and also that the characterization illus-
RENEWABLE TECHNOLOGIES trated here may not apply in every case. For example,
In considering the operation of electricity markets biomass availability can depend upon the supply of
with a high penetration of variable renewables, it waste products that provide the fuel, which may not
is instructive to first define the characteristics of be easy to store in large quantities, and run-of-river
renewable technologies that make them different. hydro may be somewhat variable and uncertain. Thus,
Three key characteristics are identified here:7 the intention of the framework illustrated in Figure 1 is
not to conclusively ‘pigeon-hole’ renewable technolo-
gies, but rather to illustrate their widely varying char-
• Variability and uncertainty: Some renewable
acteristics, and to emphasize that each technology and
technologies, such as wind and solar PV exhibit
system should be considered on a case-by-case basis,
variable availability, driven by the time-varying
based upon its particular properties.
availability of the wind and solar resources
The categories outlined in Figure 1 provide a
that drive them.5 This variability is partially pre-
framework for exploring the aspects of electricity mar-
dictable but also partially uncertain.
ket design that will best facilitate efficient and low
• Low short-run marginal costs: Many renewable cost integration of those technologies, as illustrated in
technologies have very low operating costs, or Figure 2. Markets that incorporate large quantities of
‘short-run marginal costs’ (SRMC). This is the variable renewable technologies will need to ensure
marginal cost of producing an extra increment adequate access to flexibility. Those that include a
(MWh) of energy. large share of generation with low SRMCs will need
• Nonsynchronous: Some renewable technologies, to carefully consider the way in which generation
such as wind and solar PV, are nonsynchronously investment signals are managed. Systems with a sig-
connected to the grid. This means that they inter- nificant proportion of nonsynchronous generation will
act with the grid in a quite different manner to need adequate grid codes and regulatory frameworks
synchronously connected technologies, such as to ensure appropriate provision of system stability
thermal generators and hydro with large syn- requirements. Each of these aspects is discussed in
chronous electrical generators. In the absence of more detail in the following sections.

© 2014 John Wiley & Sons, Ltd.

 WIREs Energy and Environment Designing electricity markets

are added, and these mechanisms may need to be

Renewable technologies Therefore markets
can be: should ensure:
expanded and enhanced (Box 1).

BOX 1
Variable & uncertain Flexibility
TERMINOLOGY FOR REFERRING TO
VARIABLE GENERATION

Low SRMC
Effective investment The term ‘variable’ is applied here to refer
signals to technologies such as wind and solar PV,
whose available generating capacity varies
over time as the primary energy resources that
Non-synchronous Suitable grid codes drive them vary. The terms ‘intermittent’ or
‘non-dispatchable’ are also commonly applied,
but it has been proposed that these are not
FIGURE 2 | The distinguishing characteristics of renewable accurate descriptions of the properties of these
technologies, and the associated features of electricity markets to technologies.5 Intermittent implies a rapidly
effectively manage these qualities. ‘flickering’ quality, suggesting a resource that is
fully available 1 second and completely unavail-
Many of the aspects discussed could be consid- able the next. This is a poor description for
ered good market design even in markets that do not the characteristics of technologies such as wind
generation, for which the available capacity will
include any renewable generation. As discussed below,
generally vary gradually over time periods of
all markets already include some degree of uncer-
hours.10 Solar PV can vary more rapidly as cloud
tainty and variability, all market designs face chal-
cover moves, but over a large power plant, or
lenges around creating accurate investment incentives,
with dispersed residential PV, this will occur
and all systems must ensure appropriate grid codes over timescales of minutes rather than seconds,
and regulatory planning frameworks. However, the and the aggregate solar power generation over
addition of renewable technologies will often exacer- a whole system will generally vary even more
bate these issues, and thus the value of good design gradually.
choices is enhanced as the proportion of renewables Similarly, technologies such as wind and PV
increases.5 are typically highly dispatchable, being able to
be rapidly ramped over their entire available
capacity.5 Arguably, wind and PV are much more
VARIABILITY AND UNCERTAINTY flexibly dispatchable than many thermal units,
which may take hours to ramp, days to cycle on
Variability and uncertainty are features of all power
and off, and are limited to operation above a
systems. Demand is inherently variable and uncertain
significant minimum loading. Some system oper-
to predict. Aggregate residential loads vary signifi-
ators even report an increased degree of com-
cantly over the duration of a day, and industrial loads fort when renewable generators are operating,
such as electric arc furnaces can exhibit extreme fluc- because they offer access to very flexible capac-
tuations of large capacities.8,5 Similarly, all generators, ity that can be rapidly turned down if required
including fossil fuel and nuclear generators, have the to maintain system security (System operator of
potential for unplanned forced outages at any time. In the South-West Interconnected System, Western
the case of large conventional generators this can very Australia 2011, personal communication).
suddenly remove hundreds to thousands of megawatts Thus, the term ‘variable generation’ is pre-
of capacity from the grid without notice. ferred here for referring to technologies with
Thus, over many decades, power system these characteristics.
engineers have already developed sophisticated
mechanisms for managing considerable amounts of
variability and uncertainty. At low levels of penetra-
tion, variable renewable technologies can often be Effective Market Design for Managing
absorbed into the system without adding to its flexi- Variability and Uncertainty
bility requirements.9 However, beyond a certain point Power systems must always maintain balance between
the aggregate degree of variability and uncertainty in supply and demand for electricity within small pre-
the system is likely to increase as variable renewables scribed tolerances, to maintain system frequency

© 2014 John Wiley & Sons, Ltd.

 Overview wires.wiley.com/wene

within the narrow bands required for safe and stable 1. Operational time frames: Generators (and loads)
operation. This must be managed over time frames in the system must be encouraged to offer their
ranging from less than a second to years. As more flexibility to the market over operational time
variable renewable generation is added to the system, frames, to economically efficient levels. This
the degree of flexibility required from the ‘balance of includes the design of the wholesale market, and
system’ (the other generators and loads) will often the design of frequency control ancillary services
increase. (FCAS), as described in more detail below.
Flexibility in this context is used to refer to 2. Investment time frames: The developers of
aspects such as: new generators entering the system should be
exposed to appropriate market signals to ensure
• Fast ramping rates (both up and down): Many that new capacity is appropriately flexible.
thermal units are not designed to ramp rapidly,
which can limit their ability to respond to vari- The latter may, but does not always, follow from
ability in the system. the former, depending upon the interaction of energy
• Short shut-down and start-up times: For many spot markets with mechanisms designed to manage
large thermal units a full start-up or shutdown investment (such as capacity mechanisms).
of the plant is a significant process, requiring
substantial notice (such as 24 h). This limits their
Wholesale Market Design to Facilitate
ability to respond flexibly to changes in the
system. Variable Renewable Integration
Balancing of supply and demand is managed over
• Low minimum loads: Minimum loads (also
operational time frames firstly through general mar-
called ‘turn down levels’) are the lowest level the
ket design, usually with a real-time wholesale energy
plant can be turned down to, without executing
market, in which generators make offers to sell elec-
a complete shutdown. For many large thermal
tricity, and retailers (also known as suppliers or load
generators (such as coal and combined cycle gas
serving entities) make offers to buy electricity. The
turbines) this is approximately 50–70% of the
market settles at the point where demand and supply
capacity of the unit; below this level the boiler
are balanced, with the price of the marginal genera-
cannot be operated in a stable state. This con-
tor (the last generator dispatched) typically setting the
strains the operation of these plants, and can be
price for all market participants in that trading inter-
particularly problematic in small systems with val. The design and operation of this wholesale market
low overnight minimum loads. can have significant implications for the degree of flex-
ibility that generators and loads willingly offer to the
Most power systems have sufficient technical system.
flexibility to integrate high proportions of variable The ‘speed’ of the market is one of the most
renewables; analysis by the International Energy important characteristics. Very fast markets that fea-
Agency suggests that existing power systems may be ture short (5 min) dispatch intervals can integrate
able to integrate 20–63% variable renewables based much higher quantities of variable renewables at lower
upon a ‘purely hardware point of view’.2,11 With cost.5,6 Shorter dispatch intervals allow more frequent
innovation, the flexibility of existing systems may re-dispatch of the whole system, providing near con-
exceed initial expectations. For example, a recent stant access to the full degree of flexibility available
case study demonstrated that coal plants can become from all plants and loads. By contrast, long dispatch
flexible resources; a coal-fired power station originally intervals artificially ‘lock-in’ generators and loads to
designed to operate as a ‘base-load’ plant in North preset levels for extended periods of time, prevent-
America now operates successfully in a highly flexi- ing adjustments that may be possible at very low
ble manner, often undertaking complete shut-down cost.5 These markets must maintain larger quantities
and start-up cycles multiple times per day. This was of ‘reserves’ (plant set aside for managing variabil-
achieved with limited hardware modifications, but ity and uncertainty within dispatch intervals), thus
extensive modifications to operational practice.11 increasing system costs.5,12
Despite the high degree of ‘technical’ flexibility A short delay from gate closure (the time when
available in many power systems, poor market design the last offers and bids are made) to dispatch is also
may limit access to that flexibility in a range of important. A short delay allows incorporation of the
ways.3,5,6 Market design to encourage efficient levels most up-to-date wind and solar PV forecasts, and
of flexibility needs to be explored in two dimensions:5 therefore more accurate scheduling.10,13,14 This means

© 2014 John Wiley & Sons, Ltd.

 WIREs Energy and Environment Designing electricity markets

that smaller quantities of reserves can be maintained, many renewable technologies have very low short-run
reducing costs.15,16 Needless to say, the quality of wind marginal costs, full participation in the wholesale mar-
and solar forecasts is also important, in addition to ket without priority dispatch is likely to still result in
integration of forecasts with the full market dispatch their full dispatch in most normal circumstances.
process. ‘Make Whole’ payments, applied in some mar-
Managing dispatch over larger areas is also kets, add complexity to the issue. These payments
of significant benefit. In many markets significant ensure that generation resources recoup costs by pro-
physical transmission interconnection exists, but sup- viding out of market payments if necessary. Because
ply and demand are balanced over an artificially the resource is providing a service on behalf of the
much smaller area. Aggregating markets (or ‘balanc- system and should be compensated to cover costs,
ing areas’) allows greater per unit smoothing of gen- these payments are ‘out of market’ and thus may
eration and load variability, access to a larger number not ensure economic optimality. In addition, they
of generators and loads for providing flexibility,17 and shield market participants (in this case typically large
allows sharing of reserves, reducing costs.5,10 In many incumbent thermal units) from important pricing sig-
cases the barriers to aggregation (and thus cost reduc- nals, potentially not providing sufficient incentives
tion) are political and regulatory in nature, rather than to provide long-term flexibility. Some markets have
technical.18 introduced new categories that allow variable renew-
Many markets have only limited demand side ables to participate more fully in the wholesale mar-
participation (DSP). Because loads are usually not ket, while recognizing their unique characteristics.
exposed to real-time market prices, they usually have For example, the Midcontinent Independent System
no incentive to respond to real-time signals limiting the Operator (MISO) introduced the ‘Dispatchable Inter-
degree of flexibility they make available to the system. mittent Resources’ category,20 and the Australian
The potential may be very large, and possible in many National Electricity Market (NEM) introduced the
cases at low cost. For example, the value of DSP ‘Semi-Scheduled’ category.21 The New York ISO has
potential in the European Union has been estimated also extended its market-based Security Constrained
at 53 billion euros.19 Access to new technology such Economic Dispatch process to optimize the schedul-
as advanced metering infrastructure may assist in ing of wind plants, based upon their offers, as it does
alleviating some of the many and varied barriers with other generating resources.22
to DSP. However, a shift to full DSP represents a Some system designers have decided that the
fundamental change in the way electricity markets price signals provided by the market are insufficient
operate, and a complete paradigm shift in the way to elicit an adequate flexibility response, and have
that customers interact with their electricity supply. added further incentives. For example, the MISO
Many of the barriers are likely to be social and and California Independent System Operator (CAISO)
regulatory, rather than technical or economic, and have both introduced supplementary mechanisms to
thus a concerted effort is likely to be required to reward generators for providing ramping services over
unlock this potential. time frames longer than a single dispatch interval.
Ideally, markets are designed to be ‘technol-
ogy neutral’, such that renewable technologies fully
participate in the wholesale market, and are man- Frequency Control Ancillary Services
aged on a level playing field with other technolo- to Facilitate Variable Renewable Integration
gies. Some markets have applied ‘priority dispatch’ for FCAS are provided by generating capacity (or
renewable technologies, such that they are dispatched demand) that is reserved for responding to varia-
ahead of other technologies regardless of system con- tions in the system balance on time periods shorter
ditions at the time. At low levels of penetration this than a dispatch interval, or that were not forecast
can effectively assist the entry of renewables. How- at the time of scheduling. In markets with long dis-
ever, at higher penetration levels the shielding from patch intervals and long delays from gate closure to
market price signals becomes problematic, because dispatch, much larger quantities of reserves must be
it inhibits an economically efficient market response. maintained. Thus, the design of the wholesale market
Many renewable generators (such as wind and PV) can significantly affect the cost of providing FCAS.
are extremely flexible, being able to ramp very rapidly Some systems have real-time markets for a range
from full capacity to complete shutdown with very lit- of FCAS that operate alongside the wholesale energy
tle notice, but they are unlikely to do so unless they market. Through these FCAS markets, suitably capa-
are appropriately incentivised by exposure to the same ble generators and loads can submit offers to provide
wholesale price signals as other generators. Because various ancillary services in each dispatch interval,

© 2014 John Wiley & Sons, Ltd.

 Overview wires.wiley.com/wene

along with their energy offers and bids. This allows Type of event Response Direction
generators and loads to flexibly provide the maximum time
amount of FCAS they have available in any interval,
without having to make long-term commitments. Raise
Primary
Dividing FCAS into distinct categories allows a Lower
wider range of generators and loads to provide only
the particular services that they find cost-effective. Raise
Contingency Secondary
Common categories of FCAS are illustrated in
Lower
Figure 3, and include:

Frequency reserves
Raise
Tertiary
• Type of event: It is common to make a distinction Lower
between services that are provided continuously
to match supply-demand discrepancies within a Raise
dispatch interval (often termed ‘regulation’), and Primary
Lower
services that involve standing ready to respond if
a generator or load suddenly experiences a forced Raise
outage (often termed ‘contingency’). Regulation Secondary
Lower
• Response time: Services can be further divided by
the response time in which they are required to be Raise
available. Very fast services (responding within Tertiary
Lower
seconds) are often called ‘primary’ reserves, while
slower services (responding in minutes or hours)
are termed secondary or tertiary reserves respec- FIGURE 3 | Typical categories of frequency control ancillary
tively. Some generators can respond over short services.
time frames, but cannot sustain the response for
a long period. Other generators can only respond
more slowly, but can sustain the response for as forced outage). These generators are technically capa-
long as required. Dividing FCAS into different ble of providing this service23 with minimal opportu-
categories allows these generators to seamlessly nity cost, meaning that in future markets this service
integrate, cost-effectively providing the required might be provided essentially for free in any period
system response in aggregate. where wind and PV are operating at sufficient levels.
Thus, if wind and PV are to participate cost-effectively
• Raise/lower: Every service can be further divided in the FCAS market, these services must be divided
into ‘raise’ and ‘lower’ components, where these into different categories.24,25
involve adding or subtracting energy from the Similarly, many types of loads may find it
system. cost-effective to provide contingency raise services.
Contingency events are relatively rare (called upon
Good market design principles suggest an perhaps several times per year) and typically short
approach that is tailored to the specific system, (less than an hour), such that providing a contingency
selecting categories of FCAS that allow the maxi- raise service would simply require demand response
mum number of generators and loads to provide the units to be available to rapidly reduce consumption
components that they find economically and tech- in the rare event of a generator suddenly experienc-
nically efficient, without introducing excessive and ing a forced outage. However, demand response may
unnecessary complexity. This will change as the mix not be equally able to provide a contingency lower ser-
of generators and loads evolves over time, making vice, because this would require being able to rapidly
regular review of the FCAS market design worthwhile. increase consumption at short notice. Electric vehicle
Wind and PV generators would face a signifi- (EV) charging, for example, may be well suited to pro-
cant opportunity cost from providing a contingency viding contingency raise services. Owing to the rarity
raise service, because this would necessitate their cur- of reserves being called upon, the impact on EV bat-
tailment over extended periods of time, and therefore tery cycling could be negligible, and customers mini-
mean lost revenue in the energy market. However, mally affected.
they may find it cost-effective to provide contingency Of course, to achieve these benefits it is a
lower services (being ready to rapidly reduce genera- prerequisite that renewable generators and loads are
tion in the rare event of a load suddenly experiencing a freely allowed to participate in providing all types

© 2014 John Wiley & Sons, Ltd.

 WIREs Energy and Environment Designing electricity markets

of FCAS, as long as they can demonstrate that they when the system is likely to experience a shortfall in
are technically capable of doing so (they are in many flexibility, and therefore where there are opportunities
cases23 ). This is not yet the case in many markets. to install new generation that can profitably provide
Another important aspect of FCAS market that service. However, this may not eventuate in prac-
design is the manner in which the quantity required tice. Firstly, during a period of rapid market transfor-
of each type of reserve is determined. Many systems mation developers may struggle to anticipate future
apply a ‘static’ approach, based upon examination market needs, and may underestimate the value of
of typical system variability at different times of day, increased flexibility. Secondly, many markets include
or by season, peak versus off-peak, and so on.26 explicit capacity remuneration mechanisms (CRMs)
However, the operation of variable renewable gener- that supplement generator revenue. Where capacity
ation at any particular point in time has a significant payments constitute a large proportion of generator
influence on the degree of system variability, allowing revenue, the design of these CRMs can have significant
a more dynamic approach that takes into account implications for generator incentives around new tech-
the system state in a particular dispatch interval.27–30 nology decisions. It is also possible that some FCAS
Some systems set the reserve requirement in a com- markets are thin—i.e., even in the face of high price
pletely dynamic fashion based upon system properties spikes, little additional FCAS are needed to ensure a
such as the time error (closely related to the system sufficient supply and mitigate price spikes. If this is the
frequency).31,32 This allows the reserve requirement to case, a new entry (or entrants) into the market could
evolve dynamically in real time as the system changes, result in a price collapse that could result in revenue
such that the scheduled reserve in each interval is the insufficiency for the entrant.
minimum required (reducing costs). In a properly operating energy-only market,
The interaction of markets for ancillary services generators are exposed to extremely high prices at
and energy requires careful attention, particularly as times of scarcity, and very low (negative) prices during
the number and sophistication of ancillary services times of oversupply. In some circumstances these
markets increases over time. In some cases, poor con- prices can occur suddenly, creating incentives for
sideration of these interactions has led to detrimental generators to rapidly adjust their output. Thus, in this
perverse incentives. For example, in some Eastern US market there is an incentive to install more flexible
systems, the energy market rules financially penalize plant that can respond to these price signals and reap
generators that deviate from their energy schedules, the benefits.5 By contrast, in markets where a CRM is
aiming to provide incentives for generators to operate applied, the market price cap is typically much lower.
at their instructed levels. However, many generators This reduces the benefit of flexible operation.
are designed to automatically assist in maintaining sys- Some markets provide added incentives for
tem frequency during contingency events. If a genera- flexibility by distinguishing between different kinds
tor’s governor is tuned appropriately, a drop in system of capacity in the capacity payment process. For
frequency will cause it to increase output, assisting in example, in Spain only generators of certain types are
arresting system frequency decline. Despite this being eligible for capacity payments.34 In the German mar-
behavior that is helpful to the system, the energy mar- ket a ‘focused’ capacity market has been proposed,
ket rules in this example would penalize this gener- involving multiple tiers of capacity credits that result
ator for deviating from its energy schedule. This has in differentiated capacity payments depending upon
likely contributed to many generators in these markets the flexibility of the plant.35
detuning their governor responses, and exacerbating
frequency response issues.33 These perverse incentives
can be removed by only penalizing deviations from the LOW SHORT-RUN MARGINAL COSTS
energy schedule when those deviations are exacerbat- For most renewable technologies the majority of the
ing frequency problems, such as is implemented in the expense is in the capital cost of constructing the
sophisticated ‘causer pays’ methodology in the Aus- generator, potentially with some fixed maintenance
tralian NEM.31 costs that do not vary significantly with the amount
of energy generated. Given their very low SRMC,
in a competitive market it is rational for renewable
Investment Signals for Flexibility generators to offer their energy at a price close to zero.
In theory, short-term price signals for generators to Generation at any level above their SRMC increases
provide flexibility to a system also translate into operational profitability (although profitability in the
long-term investment signals. With sufficient fore- long-run requires sufficient revenue to recuperate fixed
sight, developers of new generation can anticipate costs).

© 2014 John Wiley & Sons, Ltd.

 Overview wires.wiley.com/wene

In markets where renewable generators supply of energy-only markets, where market participants
large quantities of energy the so called merit order receive revenue only through the wholesale energy and
effect has led to falling electricity prices, reducing the ancillary services markets.41,42 Others remain con-
profitability of incumbent generators.36–39 The entry vinced that energy-only markets are fundamentally
of low SRMC renewables pushes other generators up flawed for a variety of subtle reasons, and support the
the dispatch merit order, causing lower prices and introduction of explicit CRMs.43,44 These introduce
reduced dispatch for incumbents. a more certain revenue stream for investors in firm
This effect applies not only to renewable genera- capacity by providing a payment for capacity avail-
tors. Many conventional generating technologies also ability (MW) in addition to energy supplied to the
have low operating costs. Nuclear power is similarly market (MWh). This can create stronger incentives
capital intensive, and some Australian mine-mouth to invest in new capacity, and may allow investors to
coal-fired power stations can have an SRMC at a acquire capital at a lower cost by reducing risk and
similar level to the estimated variable operations and uncertainty. They can also support higher generation
maintenance costs of wind generators.40 In present adequacy and security of supply for the whole system,
power systems, these low SRMC technologies are typ- and allow lower price caps to be introduced in the
ically combined with a proportion of much higher wholesale market, reducing price volatility and there-
SRMC technologies, such as gas-fired generation. fore reducing risks for customers and other market
When these intermediate and peaking generators oper- participants.24
ate, all generators are paid the higher marginal price However, given that CRMs represent a regula-
(in a marginal pricing market) and thus recoup fixed tory intervention, they also have some drawbacks.24
and capital costs. The choice on the amount of reserve margin main-
Not all renewable technologies have a very low tained in the system must be made via a regulatory
SRMC. For example, some biomass technologies are decision, which can incline the system toward a con-
likely to have a cost associated with collection, pro- servative oversupply of capacity. Some have argued
cessing and storage of biomass fuel. Hydro generation that this may be of benefit, because it may be better
is also complex to value in short-run terms, because it to err on the side of oversupply and limit the risk
is energy limited and therefore often carries a substan- of extended periods of high prices for customers.45
tial ‘opportunity cost’ which will strongly influence the Design and implementation of CRMs can also be
way that hydro capacity is offered to the market. Thus, very complex,24 and many have required substantial
power systems composed of a mixture of technologies adjustments over time as the market evolves. Care
involving a proportion of these higher SRMC renew- must also be taken regarding coordination with
ables may find that they exhibit pricing profiles not neighboring markets; implementing different market
dissimilar to present power systems, even at very high models without care for cross-border effects can cause
renewable penetrations. distortions and hinder functioning.24
However, as illustrated in Figure 1, many renew- There are a wide range of different types of
able technologies do exhibit a very low SRMC, and a CRMs, and some of the most innovative, such as ‘reli-
future power system could conceivably be composed ability options’, may be remarkably close to properly
entirely of these technologies. In this situation, it is log- functioning energy-only markets in practice. Ulti-
ical to foresee market prices falling to close to zero for mately, it is likely that most types of market designs
the majority of the time. The questions then arise: how can work well if they are implemented well; perceived
do generators make money in such a market? Why failures are often related to flawed implementation or
would anyone invest? These questions relate to what excessive government intervention.46
is termed ‘resource adequacy’, or ‘system adequacy’ Renewables do not undermine the theoretical
mechanisms. operation of energy-only markets and do not fun-
damentally change the debate, but they are bringing
many of the long-discussed issues to prominence.7
Effective Market Design for Investment With lower wholesale prices occurring more often,
Signals with Low SRMC Generation an increasing proportion of generator revenues in
The ideal mechanisms for managing resource ade- energy-only markets will need to be concentrated in
quacy have been debated for decades, without rare scarcity periods, exacerbating uncertainty and
resolution. Creating accurate long-term investment risks for market participants.7 In theory, market par-
incentives via an electricity market is challenging, ticipants can manage these risks with increased levels
even in the absence of variable renewable generation. of forward contracting, but there may be reasons why
Some proponents favor the economically ‘pure’ theory sufficient contracting does not occur.

© 2014 John Wiley & Sons, Ltd.

 WIREs Energy and Environment Designing electricity markets

In many ways, the most challenging aspect of the control software. This is an emerging issue that is
resource adequacy is likely to be related to the rate being examined so that the performance of synthetic
of change that markets need to manage. All models inertia can be better understood and metrics can be
are likely to struggle during periods of high uncer- developed to help determine the need for inertia.
tainty, and with rapidly changing political and reg- This characteristic of renewable technologies
ulatory drivers. These characteristics fundamentally is highly technical in nature, and is more related to
make investment decisions very challenging, especially grid codes, network regulation and the engineering
when planning long-lived and lumpy infrastructure.24 requirements placed upon new entrant generators. It
Proponents of energy-only markets suggest that is generally a step removed from the operation of the
they could operate successfully, if all the barriers wholesale electricity market, although it will interact
preventing proper market operation were removed, with the provision of FCAS. One of the most impor-
including low price caps, regulated retail prices, and a tant messages from current research is that grid codes
lack of DSP.24 However, the proportion of renewable must be forward-looking to anticipate the require-
technologies is rapidly increasing in many markets, ments of the future grid, but not overly onerous or
and addressing all these issues may take too long, discriminatory against new entrant technologies.49
given political orientations and the lengthy processes It can be challenging to strike this balance
likely to be involved in increasing DSP. A CRM correctly.
may assist in maintaining investment incentives in the
interim, implemented as a temporary measure with
eventual return to an energy-only market model being CONCLUSION
the ultimate goal.24 However, once introduced, CRM One of the most important findings from recent
policies may be hard to reverse.47 research on the integration of renewable technolo-
gies into electricity markets is that market design
really matters. The costs and challenges of integrat-
NONSYNCHRONOUS GENERATION ing renewables can vary significantly depending upon
Some renewable technologies, such as wind and PV, apparently minor differences in market design. This
are nonsynchronous. This means that they are techni- makes cross-comparison of integration studies and
cally quite different to most conventional technologies experiences challenging, and highlights the impor-
(such as coal, gas, hydro, and nuclear) which are syn- tance of context specific analysis for the particu-
chronously connected to the grid. A synchronous lar system under consideration. It also points to an
connection contributes to frequency stability, and immensely important opportunity: by adjusting mar-
also assists with various technical aspects of grid ket design alone there is the potential to significantly
operation. Increasing displacement of synchronous reduce the costs of integrating renewable technologies
generation may eventually lead to the requirement for into power systems.
additional market dispatch constraints that require a There is no denying that system operators and
minimum proportion of synchronous generation to market designers have a plethora of interesting chal-
remain in operation at all times (leading to spilling of lenges ahead. Questions about the degree of system
nonsynchronous renewable generation). For example, flexibility required and how best to encourage it are
in the Australian National Electricity Market it has only beginning to emerge. Similarly, it is unlikely that
been suggested that it may be necessary to maintain debates about the best way to manage resource ade-
a minimum of 15% synchronous generation at all quacy will be resolved any time soon. Regardless of
times,48 although far more detailed modeling and the frameworks implemented in each system, the best
analysis are required. Modern wind turbines can outcomes are likely to be achieved by rigorous analy-
provide synthetic inertia, which is implemented in sis, facilitated by ongoing international collaboration.

REFERENCES
1. REN21. Renewables 2013 – Global Status Report. publications/freepublications/publication/Harnessing_
2013. Available at: http://www.ren21.net/REN21Activit Variable_Renewables2011.pdf. (Accessed May 30,
ies/GlobalStatusReport.aspx. (Accessed May 30, 2014). 2014).
2. International Energy Agency. Harnessing Variable 3. Van Hulle. Large scale integration of wind energy
Renewables. 2011. Available at: https://www.iea.org/ in the European power supply: analysis, issues and

© 2014 John Wiley & Sons, Ltd.

 Overview wires.wiley.com/wene

recommendations. European wind energy association. 17. Green RJ. Electricity wholesale markets: designs now
2005. Available at: http://www.ewea.org/fileadmin/ewea and in a low-carbon future. Q J IAEE’s Energy Econ
_documents/documents/publications/grid/051215_Grid_ Educ Found 2008, 29:95–124.
report.pdf. (Accessed May 30, 2014). 18. Vandezande L, Meeus L, Belmans R, Saguan M, Glan-
4. Milligan M, Holtinnen H, Soder L, Clark C, Pineda I. chant JM. Well-functioning balancing markets: a pre-
Markets to facilitate wind and solar energy integration requsite for wind power integration. Energy Policy
in the bulk power supply: an IEA Task 25 collaboration. 2010, 38:3146–3154.
National Renewable Energy Laboratory. 2012. Avail- 19. Faruqui A, Harris D, Hledik R. Unlocking the €53
able at: http://www.nrel.gov/docs/fy12osti/56212.pdf. billion savings from smart meters in the EU: how
(Accessed May 30, 2014). increasing the adoption of dynamic tariffs could make
5. Milligan M, Kirby B. Market characteristics for effi- or break the EU’s smart grid investment. Energy Policy
cient integration of variable generation in the Western 2010, 38:6222–6231.
interconnection. National Renewable Energy Lab- 20. Federal Energy Regulatory Commission. Order con-
oratory. 2010. Available at: http://www.nrel.gov/ ditionally accepting in part and rejecting in part
docs/fy10osti/48192.pdf. (Accessed May 30, 2014). tariff filing and requiring compliance filings. Docket
6. ISO/RTO Council. Increasing renewable resources: No. ER11-1991-000. 2011. Available at: http://www.
how ISOs and RTOs are helping meet this public policy troutmansandersenergyreport.com/wp-content/uploads/
objective. 2007. Available at: http://www.consultkirby. 2011/03/Midwest-ISO-DIR.pdf. (Accessed May 30,
com/files/IRC_Renewables_Report_101607_final.pdf. 2014).
(Accessed May 30, 2014). 21. MacGill I. Electricity market design for facilitating the
7. Riesz J, MacGill I. 100% renewables in Australia – will integration of wind energy: experience and prospects
a capacity market be required? In: Proceedings of the with the Australian National Electricity Market. Energy
3rd International Workshop on the Integration of Solar Policy 2010, 38:3180–3191.
Power into Power Systems, London, 2013. 22. New York Independent System Operator. Integration
of wind into system dispatch – A New York ISO White
8. Kirby B, Hirst E. Customer-Specific Metrics for the
Paper. 2008.
Regulation and Load-Following Services. Oak Ridge
National Laboratory, Oakridge, TN, USA. 2000. 23. Ela E, Gevorgian V, Fleming P, Zhang YC, Singh
Available at: http://energy.lbl.gov/ea/certs/pdf/ornl-con- M, Muljadi E, Scholbrook A, Aho J, Buckspan A,
474.pdf. (Accessed May 30, 2014). Pao L, et al. Active power controls from wind power:
bridging the gaps. 2014. Available at: http://www.nrel.
9. Holttinen H, Kiviluoma J, Estanqueiro A,
gov/docs/fy14osti/60574.pdf. (Accessed May 30, 2014).
Gomez-Lazaro E, Rawn B, Dobschinski J, Meibom
P, Lannoye E, Aigner T, Wan YH, et al. Variability 24. Eurelectric. RES Integration and Market Design: Are
of load and net load in case of large scale distributed Capacity Remuneration Mechanisms needed to ensure
wind power. In: Proceedings of the 10th International generation adequacy? 2011.
Workshop on Large-scale Integration of Wind Power 25. Eurelectric. Eurelectric Position Paper Towards Mar-
into Power Systems, Aarhus, Denmark, 2011. ket Integration of Reserves & Balancing Markets.
10. Holttinen H, Meibom P, Orths A, Hulle Fv, Lange 2008.
B. Design and operation of power systems with large 26. Ela E, Kirby B, Navid N, Smith JC. Effective ancillary
amounts of wind power. Final Report, Phase one 2006- services market designs on high wind power pene-
08. 2009. tration systems. National Renewable Energy Labo-
ratory. 2012. Available at: http://www.nrel.gov/docs/
11. Cochran J, Lew D, Kumar N. Flexible Coal—Evolution
fy12osti/53514.pdf. (Accessed May 30, 2014).
from Baseload to Peaking Plant. Golden, Colorado:
National Renewable Energy Laboratory (NREL); 2013. 27. Ela E, Milligan M, Kirby B. Operating Reserves and
Variable Generation. Golden, Colorado: National
12. Milligan M, Kirby B. Capacity Requirements to Support
Renewable Energy Laboratory (NREL); 2011.
Inter-balancing Area Wind Delivery. Golden, Colorado:
National Renewable Energy Laboratory (NREL); 2009. 28. Ela E, Milligan M, Kirby B, Lannoye E, Flynn D,
O’Malley M, Zavadil B. Evolution of Operating
13. DENA. DENA Grid Study II – Study for the Integration
Reserve Determination in Wind Power Integration
of a Share of 30% Renewable Energy in the German
Studies. Golden, Colorado: National Renewable
Electricity Market. 2010.
Energy Laboratory (NREL); 2010.
14. DENA. Planning of the Grid Integration of Wind Energy 29. National Renewable Energy Laboratory. Western
in Germany Onshore and Offshore up to the Year 2020 Wind and Solar Integration Study. Golden, Colorado:
(dena Grid Study). 2005. National Renewable Energy Laboratory (NREL); 2010.
15. Borggrefe F, Neuhoff K. Balancing and intra-day market 30. National Renewable Energy Laboratory. Eastern Wind
design: options for wind integration. 2011. Integration and Transmission Study. Golden, Colorado:
16. Tradewind. Integrating wind. 2009. National Renewable Energy Laboratory (NREL); 2011.

© 2014 John Wiley & Sons, Ltd.

 WIREs Energy and Environment Designing electricity markets

31. Australian Energy Market Operator. Guide to Ancil- 41. Hogan WW. On an “Energy Only” Electricity Mar-
lary Services in the National Electricity Market. Mel- ket Design for Resource Adequacy. Cambridge, Mas-
bourne, Australia: Australian Energy Market Operator sachusetts: Center for Business and Government, John
(AEMO); 2010. F. Kennedy School of Government, Harvard University;
32. Riesz J, MacGill I. Frequency control ancillary ser- 2005.
vices – Is Australia a model market for renewable inte- 42. Hogan WW. Electricity scarcity pricing through oper-
gration? In: Proceedings of the 12th International ating reserves. Econ Energy Environ Policy 2013,
Workshop on Large-scale Integration of Wind Power 2:65–86.
into Power Systems, London, 2013. 43. Cramton P, Ockenfels A, Stoft S. Capacity market fun-
33. Ela E, Milligan M, Kirby B, Tuohy A, Brooks D. damentals. Econ Energy Environ Policy 2013, 2:27–46.
Alternative approaches for incentivizing the frequency
44. Cramton P, Stoft S. The convergence of market designs
responsive reserve ancillary service. 2012. Avail-
for adequate generating capacity with special atten-
able at: http://www.nrel.gov/docs/fy12osti/54393.pdf.
tion to the CAISO’s resource adequacy problem. A
(Accessed May 30, 2014).
White Paper for the Electricity Oversight Board;
34. Spees K, Newell SA, Pfeifenberger JP. Capacity mar- 2006. Available at: http://www.cramton.umd.edu/
kets – lessons learned from the first decade. Econ papers2005-2009/cramton-stoft-market-design-for-
Energy Environ Policy 2013, 2. resource-adequacy.pdf. (Accessed May 30, 2014).
35. Hogan M, MGottstein M. What lies "beyond capacity 45. De Vries LJ. Generation adequacy: helping the market
markets"? Delivering least-cost reliability under the do its job. Utilities Policy 2007, 15:20–35.
new resource paradigm. A "straw man" proposal for
46. Hesmondhalgh S, Pfeifenberger J, Robinson D.
discussion. The Regulatory Assistance Project. 2012.
Resource Adequacy and Renewable Energy in Com-
36. Cutler NJ, Boerema ND, MacGill IF, Outhred HR. High petitive Wholesale Electricity Markets. Published by
penetration wind generation impacts on spot prices in The Brattle Group; 2010. Available at: http://www.
the Australian national electricity market. Energy Policy thebrattlegroup.com/_documents/UploadLibrary/
2011, 39:5939–5949. Upload878.pdf. (Accessed May 30, 2014).
37. Forrest S, MacGill I. Assessing the impact of wind 47. Caldecott B, McDaniels J. Stranded generation assets:
generation on wholesale prices and generator dispatch implications for European capacity mechanisms, energy
in the Australian National Electricity Market. Energy markets and climate policy, Working Paper, 2014.
Policy 2013, 59:120–132.
48. Australian Energy Market Operator. 100 per cent
38. Sensfuß F, Ragwitz M, Genoese M. The merit-order
Renewables Study—Modelling Outcomes. Mel-
effect: a detailed analysis of the price effect of renewable
bourne, Australia; 2013. Available at: http://www.
electricity generation on spot market prices in Germany.
climatechange.gov.au/sites/climatechange/files/
Energy Policy 2008, 36:3086–3094.
documents/08_2013/100-percent-renewables-study-
39. Tveten ÅG, Bolkesjø TF, Martinsen T, Hvarnes H. modelling-outcomes-report.pdf. (Accessed May 30,
Solar feed-in tariffs and the merit order effect: a study 2014).
of the German electricity market. Energy Policy 2013,
49. Quitmann E, Erdmann E. The power system will need
61:761–770.
more! How grid codes should look ahead. In: Proceed-
40. Bureau of Resources and Energy Economics. Australian ings of the 12th International Workshop on Large-Scale
Energy Technology Assessment 2013 Model Update. Integration of Wind Power into Power Systems, Lon-
Canberra, Australia: Bureau of Resources and Energy don, 2013.
Economics (BREE); 2013.

FURTHER READING
Riesz J, Gilmore J, Hindsberger M. Market design for the integration of variable generation. In: Sioshansi FP, ed. Evolution
of Global Electricity Markets: New paradigms, New Challenges, New Approaches. San Francisco: Elsevier; 2013.

© 2014 John Wiley & Sons, Ltd.