IEA PVPS Report T13!25!2022 OandM Guidelines

Task 13 Reliability and Performance of Photovoltaic Systems
PVPS
Guidelines for Operation

and Maintenance of
Photovoltaic
Power Plants in
Different Climates
2022
Report IEA-PVPS T13-25:2022

Task 13 Performance, Operation and Reliability of Photovoltaic Systems – Guidelines for Operation and Maintenance of PV Power Plants in Different Climates
What is IEA PVPS TCP?

The International Energy Agency (IEA), founded in 1974, is an autonomous body within the framework of the Organization for Ec onomic
Cooperation and Development (OECD). The Technology Collaboration Programme (TCP) was created with the belief that the future of energy
security and sustainability starts with global collaboration. The programme is made up of 6.000 experts across government, ac ademia, and
industry dedicated to advancing common research and the application of specific energy technologies.
The IEA Photovoltaic Power Systems Programme (IEA PVPS) is one of the TCPs within the IEA and was established in 1993. The mission
of the programme is to “enhance the international collaborative efforts which facilitate the role of photovoltaic solar energy as a cornerstone
in the transition to sustainable energy systems.” In order to achieve this goal, the programme’s participants have undertaken a variety of joint
research projects in photovoltaic (PV) power systems applications. The overall programme is headed by an Executive Committee, comprising
one delegate from each country or organizational member, which designates distinct ‘Tasks’ that may be research projects or activity areas.
The IEA PVPS participating countries are Australia, Austria, Canada, Chile, China, Denmark, Finland, France, Germany, Israel, Italy, Japan,
Korea, Malaysia, Mexico, Morocco, the Netherlands, Norway, Portugal, South Africa, Spain, Sweden, Switzerland, Thailand, Turkey, and the
United States of America. The European Commission, Solar Power Europe, Smart Electric Power Alliance (SEPA), and Solar Energy Indus-
tries Association are also members.
Visit us at: www.iea-pvps.org
What is IEA PVPS Task 13?

Within the framework of IEA PVPS, Task 13 aims to support market actors working to improve the operation, the reliability and the quality of
PV components and systems. Operational data from PV systems in different climate zones compiled within the project will help provide the
basis for estimates of the reliability and performance of the current PV systems.
Task 13 provides a common platform to summarize and report on technical aspects affecting the quality, performance reliability and lifetime
of PV systems in a wide variety of environments and applications. By working together across national boundaries, we can all take advantage
of research and experience from each member country and combine and integrate this knowledge into valuable summaries of best practices
and methods for ensuring that PV systems perform at their optimum and continue to provide competitive return on investment.
Task 13 has established a framework for calculations of various parameters that provide an indication of the quality of PV components and
systems. The framework, along with the results included in the high-quality reports, is useful to and appreciated by the solar PV industry.
The IEA PVPS countries participating in Task 13 are Australia, Austria, Belgium*, Canada, Chile, China, Denmark, Finland, France, Germany,
Israel, Italy, Japan, the Netherlands, Norway, Spain, Sweden, Switzerland, Thailand, and the United States of America.
This report addresses climate-specific guidelines for operation and maintenance of PV systems with the aim to serve different functions to
various stakeholders depending on their roles in the entire value chain of PV. Further information and results of Task 13 can be found at:
https://iea-pvps.org/research-tasks/performance-operation-and-reliability-of-photovoltaic-systems/.
* Belgium is no longer participating in IEA PVPS, effective from 01 July, 2022.
DISCLAIMER
The IEA PVPS TCP is organised under the auspices of the International Energy Agency (IEA) but is functionally and legally autonomous. Views, findings and publica-
tions of the IEA PVPS TCP do not necessarily represent the views or policies of the IEA Secretariat or its individual member countries.
COVER PICTURE
The solar PV farm of Les Mées is located in the hills of the plateau de la Colle des Mées, in the Alpes-de-Haute-Provence department in the south of France.
Covering an area of 200 hectares with a total of 112,780 PV panels, located at 800 metres above sea level, the installation is the largest in France. It generates a
total electrical power of 100 MWp and supplies nearly 12,000 homes with sustainable energy. Photo Courtesy of AvaxNews.
ISBN 978-3-907281-13-0: Guidelines for Operation and Maintenance of Photovoltaic Power Plants in Different Climates
Task 13 Performance, Operation and Reliability of PV Systems – Guidelines for Operation and Maintenance of PV Power Plants in Different Climates
INTERNATIONAL ENERGY AGENCY

PHOTOVOLTAIC POWER SYSTEMS PROGRAMME
IEA PVPS Task 13

Performance, Operation and
Reliability of Photovoltaic Systems
Guidelines for Operation and Maintenance of

Photovoltaic Power Plants in Different Climates
Report IEA-PVPS T13-25:2022

October 2022
ISBN 978-3-907281-13-0
Task 13 Performance, Operation and Reliability of Photovoltaic Systems – Guidelines for Operation and Maintenance in Different Climates
AUTHORS
Main Authors
Ulrike Jahn, VDE Renewables, Alzenau, Germany
Bert Herteleer, Katholieke Universiteit Leuven, Leuven, Belgium
Caroline Tjengdrawira, Tractebel, Brussels, Belgium
Ioannis Tsanakas, CEA INES - Institut National de l’Energie Solaire, France
Mauricio Richter, 3E, Brussels, Belgium
George Dickeson, Ekistica, Alice Springs, Northern Territory, Australia
Alexander Astigarraga, EURAC Research, Bolzano, Italy
Tadanori Tanahashi, AIST, Fukushima, Japan
Felipe Valencia, Atamostec, Santiago, Chile
Mike Green, Green Power Engineering Ltd, Ra'anana, Israel
Anne Anderson, Research Institutes of Sweden AB (RISE), Borås, Sweden
Bengt Stridh, Mälardalen University, Västerås, Sweden
Ana Rosa Lagunas Alonso, Centro Nacional de Energías Renovables (CENER),
Sarriguren, Navarra, Spain
Yaowanee Sangpongsanont, King Mongkut’s University of Technology Thonburi
(KMUTT), Bangkok, Thailand
Contributing Authors
Narendra Shiradkar, Indian Institute of Technology, Bombay, India
Edwin Cunow, LSPV Consulting, Gröbenzell, Germany
Magnus Herz, TÜV Rheinland, Cologne, Germany
Christian Schill, Fraunhofer ISE, Freiburg, Germany
Rosmarie Neukomm, Bern University of Applied Science (BFH), Bern, Switzerland
Elke Lorenz, Fraunhofer ISE, Freiburg, Germany
Karl A. Berger, Austrian Institute of Technology GmbH (AIT), Vienna, Austria
David Moser, EURAC Research, Bolzano, Italy
David Parlevliet, Murdoch University, Perth, Western Australia, Australia
Amrita Raghoebarsing, Anton de Kom Universiteit van Suriname, Suriname
Elías Urrejola, Atamostec, Santiago, Chile
Erin Whitney, University of Alaska Fairbanks, Fairbanks, Alaska, USA
Johan Paradis Ärlebäck, Paradisenergi AB, Göteborg, Sweden
Editor
Ulrike Jahn, VDE Renewables, Alzenau, Germany
4
TABLE OF CONTENTS
Acknowledgements .................................................................................................. 6
List of Abbreviations ................................................................................................. 7
Executive Summary ................................................................................................. 9
Introduction...................................................................................................... 12
Performance Indicators.................................................................................... 14
2.1 PV Power Plant Data (Key Performance Indicators) ............................... 14
2.2 O&M Contractor Key Performance Indicators ......................................... 18
2.3 Contractual Framework .......................................................................... 23
Power Plant Operation..................................................................................... 29
3.1 Plant Performance Monitoring Guidelines............................................... 29
3.2 Performance Analysis and Optimization ................................................. 38
3.3 Power Plant Controller and Requirements for Grid Compliance ............. 39
3.4 Power Generation Forecast .................................................................... 40
3.5 Power Plant Safety Considerations ........................................................ 50
Power Plant Maintenance ................................................................................ 61
4.1 Preventive Maintenance Actions ............................................................ 61
4.2 Corrective Maintenance Actions ............................................................. 69
O&M Recommendations in Different Regions .................................................. 72
5.1 O&M Guidelines for Moderate Climates (Europe) ................................... 72
5.2 O&M Guidelines for Hot and Dry Climates.............................................. 80
5.3 O&M Guidelines for Desert Climates in High Altitudes ............................ 86
5.4 O&M Guidelines for Hot and Humid Climates (South-East Asia) ............ 98
5.5 O&M Guidelines in Flood-Prone Regions ............................................. 105
5.6 O&M Guidelines for Building Requirements for Cyclonic Regions ........ 117
5.7 O&M Guidelines for Snowy Regions ..................................................... 123
Conclusions ................................................................................................... 136
REFERENCES ..................................................................................................... 139
5
ACKNOWLEDGEMENTS
This paper received valuable contributions from several IEA PVPS Task 13 members and
other international experts. Many thanks to Ms. Johanna Tillmann from TÜV Rheinland for
technical editing of this report and to Mrs. Mary Brunisholz for proofreading this report. Thanks
are due to Mr. Paul Kaaijk, PVPS ExCo for France, and Mr. Hubert Fechner, PVPS ExCo for
Austria, for their detailed review and proof reading. We also like to thank Karl A. Berger, AIT
for his excellent technical review. The editing of the executive summary is supported by
Mr. Kevin Punzalan, VDE Renewables, which is much acknowledged.
This report is supported by the German Federal Ministry for Economic Affairs and Energy
(BMWi) under grant number 0324304A and 0324304B.
This report is supported by the Austrian Federal Government, represented by the Austrian
Research Promotion Agency (FFG) under contract no. 876736.
It is supported by the New Energy and Industrial Technology Development Organization
(NEDO), Japan, under contract #15100576-0.
This report is supported by the Swiss Federal Office of Energy (SFOE) under contract no.:
SI/501788-01.
This work is supported by CORFO technological programme ATAMOSTEC, who funded all
the tasks associated with the results presented in chapter 5.3.
The editing of Chapter 5.3 was highly supported by Jose Alejandro Tapia Jelcic, Jose Galle-
guillos Alvarado, Erik Mella Cuitiño, and Sebastián Delgado. They are part of the ATAMOS-
TEC teamwork in charge of placing and keeping the solar desert platform (PSDA) operative.
They also contributed with the measurements and data processing that generated the results
presented in chapter 5.3.
The preparation of Chapter 5.4 was supported by Sorraphat Bubpharam and Dhirayut Chen-
vidhya from CES Solar Cells Testing Center (CSSC), Pilot Plant Development and Training
Institute (PDTI), King Mongkut’s University of Technology Thonburi (KMUTT), Bangkok, Thai-
land.
6
LIST OF ABBREVIATIONS
AI Artificial Intelligence
AMR Automatic Meter Reading
ANNs Artificial Neural Networks
AR Auto-Regressive
BAPV Building Applied Photovoltaics
BIPV Building Integrated Photovoltaics
BoS Balance of Systems
CAPEX Capital Expenditure
CCD Charge Coupled Device
CM Corrective Maintenance
DSOs Distribution System Operators
EL Electroluminescence
EPC Engineering, Procurement and Construction
EPI Energy Performance Index
ERP Enterprise Resource Planning
GFS Global Forecast System
GRP Guaranteed Performance Ratio
HMM Hidden Markov Models
IEA International Energy Agency
IML Inhomogeneous Mechanical Loading
IR Infrared
KPI Key Performance Indicator
LCOE Levelized Cost of Energy
LD Liquidated Damage
LOTO Lockout-Tagout
MAE Mean Absolute Error
ML Machine Learning
MPPT Maximum Power Point Tracking
MRA Maintenance Reserve Account
NAM North American Mesoscale
NIR Near Infrared
7
NWP Numerical Weather Prediction

OEM(s) Original Equipment Manufacturer(s)
OHS Operational Health and Safety
O&M Operation & Maintenance
OPEX Operational Expenditures
PF Power Factor
PID Potential Induced Degradation
PM Preventive (or proactive) Maintenance
PPC Power Plant Controller
PPE Personal Protective Equipment
PR Performance Ratio
RMSE Root Mean Square Error
SCADA Supervision control and data acquisition
SCB String Combiner Box
SL Soiling Loss
SR Soiling Ratio
SVM Support Vector Machine
TMY Typical Meteorological Year
TSOs Transmission System Operators
UAV Unmanned Aerial Vehicles
WHO World Health Organization
WRF Weather Research and Forecasting
8
EXECUTIVE SUMMARY
The increasing adoption of PV systems in different climate zones and conditions worldwide
has indicated that stress factors such as temperature, humidity, exposure to UV light, rain, and
wind could contribute to the occurrence of module failures. Knowing of this fact, operation &
maintenance (O&M) operators have looked to customize O&M services to the climate zone
where particular plants are located.
At present, comprehensive guidelines for climate-specific O&M programs have yet to be de-
veloped. With this gap in mind, this report aims to provide comprehensive guidance for cus-
tomized O&M service in seven different climate zones. The first four are for conditions which
broadly prevail in large parts of the world (moderate, hot and dry, hot and humid, desert at high
elevation), while the latter three are for extreme conditions (flood-prone regions, cyclonic
regions, snowy regions). These guidelines can assist PV plant engineers and designers,
financing parties, and investors in designing and maintaining PV plants, as well as in deter-
mining operational risk related to investment decisions.
The report presents these guidelines according to the following topics: O&M performance in-
dicators and standard O&M operator services (Chapter 2), guidelines for monitoring, forecast-
ing, and analysis of PV plant performance and safety (Chapter 3), the different types of mainte-
nance services and advanced inspections (Chapter 4), and finally the recommendations for
climate-specific O&M along with a report of field experiences that affected reliability, perfor-
mance and safety (Chapter 5).
The key highlights from each chapter are the following:

• Chapter 2: An O&M contract should clearly describe the scope of the services and
responsibilities of the operator to prevent any ambiguities in their respective responsibilities,
as well as acceptable compensation if operators fail to fulfil their obligations. Key to this is
including at least one of the O&M operator KPIs (key performance indicators) such as Guar-
anteed Performance Ratio, Guaranteed Plant Availability, and Response Time in the contract
to allow for straightforward standards in measuring contract compliance. The choice of which
of these KPIs to select is based on negotiations in the O&M contracting phase. Finally, O&M
contracts should take into account regional differences, such as national or local legislation
that affects the availability of staff on-site, as well as variations in the capabilities of key stake-
holders that affect project cost.
• Chapter 3: Performance monitoring systems should allow for a ´follow-up´ of the energy
flows within a PV system. The scale and complexity of plants determine the level of monitoring:
the larger and more complex the plant, the more intensive the monitoring. Minimum require-
ments are detailed in international standards such as IEC61724-1, as well as best practice
guidelines such as the SunSpec Alliance. To gather insights on specific failures and underper-
formances, predictive maintenance services with data collection devices on-site should be put
in place to create an “intelligent” monitoring system. O&M operators need to be ready to comply
with applicable grid codes and regulations, allowing for the re-evaluation of the scope of oper-
ations for contracts if grid codes change, especially in contexts where more RE plants replace
thermal sources of power.
9
• This could particularly apply to PV power forecasting services, which are increasingly
critical but could be offered either by O&M contractors or by external service providers. Asset
owners may find it convenient to choose a PV forecasting service s/he is already working with,
as these may impact contract agreements with other partners who may depend on the perfor-
mance of the plant, such as trading service providers. Finally, to ensure safety, O&M operators
must guarantee plant and worker safety by ensuring that staff is well-trained and qualified to
implement safety procedures, equip them with PPE, tools and consumables, and take into
account site-specific risks such as heights, presence of water, increased fire risks, or weather
conditions.
• Chapter 4: Preventive maintenance (PM) action plans that exclude redundant activities
can bring costs down. The preventive maintenance plan should seek to optimize the overall
PV plant and O&M budgeting, depending on the plant´s size, design, complexity, and environ-
ment. The most important actions here include periodic sampling of individual electrical meas-
urements at module level, soiling and snow mitigation, site and vegetation management, and
keeping balance of systems and SCADA (supervision control and data acquisition) monitoring
systems operational.
• Aerial infrared (IR) and visual imagery are powerful tools for diagnosing faults, espe-
cially for power losses. As of present, turnkey solutions for aerial imagery diagnostic solutions
for large-scale PV does not yet exist, as current wireless communication and camera control
technology limit the operational range. Typical costs for base O&M scope, including soiling
mitigation, range from 6.5 up to 16.5 €/kWp*year. Additional costs for advanced diagnos-
tics/analytics based on aerial IR scans (on bi-annual basis), range from 0.5 to 3 € per PV
module or array. In corrective maintenance/spare parts action plans, maintenance reserve ac-
counts are recommended to be set aside by the plant owner, to foresee possible replacement
costs.
• Chapter 5: The essential practical guidelines for the climate zones studied are:
o Temperate – An on-site evaluation of vegetation, wildlife and farm animals
should be conducted. Grass cutting should be combined with an inspection of
the status of solar PV modules to decide if cleaning and/or corrective mainte-
nance actions are required. In industrial environments, solar PV modules can
develop unexpected deterioration. Special attention must be paid to selecting
cleaning products. It is advised to follow expert recommendations on suitable
products.
o Hot and Dry – Assessments must be made of wildlife risks, appropriate plan-
ning for visits to typically remote sites (hydration, anti-venom procedures, PPE,
travel to and from sites). Wildlife risks cover poisonous animals and insects that
can harm humans directly, whereas nesting insects and animals can cause
short-circuits or arc flashes. The typically remote nature of PV sites in hot and
dry climates entails significant travel and preparation requirements, due to lo-
gistical risks in terms of supplying these facilities as well as access to emer-
gency medical care. Temperature extremes and salt exposure also increase
material degradation in modules, frames, junction boxes and transmission ca-
bles.
o Hot and Humid – Wildlife intrusion in ground-mounted systems, particularly
from rodents, snakes, and termites can cause failures in PV components and
electrical systems. Rapidly growing plants can also have a soiling impact (dust
10
accumulation). Cleaning schemes can decrease production losses of PV mod-

ules by as much as 6-8% during summer months. Fire risks can also be posed
by agricultural activities such as field clearing. Adequate ventilation during hot
months is crucial for good operating conditions of inverters.
o Flood-prone – It is important to note that PV systems are not typically designed
with flood endurance in mind. However, climate change and extreme weather
events, combined with limited availability of land in certain parts of the world,
mean that some plants are built in areas that flood 2-3 times a year. Actions to
prevent damage include switching off the plant in anticipation of flooding and
only switching it back on after technical inspection. Floods can damage module
mounting clips, PV modules, the lamination on panels, and even uproot foun-
dations. Soiling damage is particularly difficult to repair when water has been
standing on the PV modules for a long time. Submerged inverters can also short
circuit and cause burn/fire risks. Fast flowing water can also cause debris im-
pact because most PV mounting systems present a high resistance to flowing
water.
o Cyclonic Regions – Damage from typhoons and cyclones typically affect PV
modules and mounting racks. Glass breakage is also a particular issue in cy-
clonic regions. The breakage of fixing parts was also observed due to uplifts in
air pressure. Cell cracks caused by deflection with strong wind load can also
occur. While national standards for wind loads are present in the USA, Japan,
and the EU, non-uniform mechanical load testing has not yet been launched
due to a lack of experts and countries which are needed to form a project team.
It is thus recommended to estimate the effects of winds using wind tunnel tests,
ensuring that structural connectors are of sufficient strength and that PV mod-
ules have sufficient uplift resistance. It is also recommended to assess the com-
pliance with standards during construction, to conduct periodic maintenance of
all bolted connections, to store sufficient repair parts, and to remove loose de-
bris around the plant.
o Snowy Regions – Snow accumulation affects PV performance, as heavy snow
loads hinder the transmission of light to the cells and could damage modules.
A suggested limit for snow accumulation on panels is 0.7m. PV racking systems
can also be damaged by extremes between winter and summer temperatures.
In this case steel racking is preferred over aluminium racking. If active cleaning
measures are implemented, such as brushing, care must be taken not to
scratch the glass as well. Experiments have been conducted to heat PV mod-
ules by applying a controlled forward voltage to melt the snow. This would re-
quire careful load control and the use of weather forecasts.
In conclusion, a combination of well-designed O&M specifications, proactive monitoring sys-

tems and a flexible and tailored O&M regime that considers both climactic impact on systems
as well as possible changes to grid requirements are good practices to ensure that PV systems
reach or even exceed the expected lifetime. Reducing risks by ensuring that personnel are
trained and equipped for O&M operations, as well as using PV forecasting to reduce possible
downtimes, also helps to maintain PV plant performance to specifications.
11
INTRODUCTION
Solar photovoltaic (PV) plant equipment is composed of a variety of different materials. The
final products, such as solar PV modules, power conversion equipment (inverters, transform-
ers, combiner boxes, etc.), module mounting structure, etc., are put together (i.e. installed) at
the site of the PV installation. The PV plant, together with all the equipment, are then commis-
sioned into operation with a typical lifetime of 25 to 30 years.
The performance and durability of PV plant equipment are expected to change over the lifetime
of a PV plant. The climatic environment in which the equipment is operating will influence the
equipment’s aging defects, the occurrence of failures and the equipment’s degradation; this is
because different materials (and how they are processed and assembled) respond differently
to different climatic stress factors such as temperature, humidity, UV light, rain, wind, etc. A
combination of these climatic parameters will also create second-order stressors (mechanical
load from snow, soiling from dust) to PV plants. In a study conducted by Köntges et al. [1]
whereby a database of PV module failures in the field were analysed in different climate zones,
the authors reported that despite the lack of strong correlation between module failure occur-
rences and impacts with the Köppen and Geiger climate zones [2], certain failures (e.g. soiling
losses in hot and dry climate) tend to occur more in certain climates.
The world of PV plant operation and maintenance (O&M) is increasingly competitive, reflected
clearly by a significant drop in the O&M service fee over the last decade; e.g. Bloomberg New
Energy Finance has reported a drop of 73% of average full-scope O&M price in Europe be-
tween 2011 and 2017. Typically, PV plant operators typically offer a standard O&M scope of
services that could be replicated easily across to reach an optimal point (i.e. minimizing oper-
ation and maintenance efforts (thus expenditures) while maximizing PV plant uptime, perfor-
mance and durability). However, a shift from the one-size-fits-all approach to a customized
O&M approach could offer as advantage that the O&M activities are adapted to the needs of
the PV plant, focusing on the maintenance activities that are necessary. Such customization
could be setting an O&M service based on the climate zone in which a PV plant is located.
One of the main challenges in customization of an O&M programme for a specific climate zone
is the present lack of comprehensive guidelines to guide the users to do so. Existing guidelines
and standards do not fill the gaps or only clarify the minimum requirements of climate-specific
O&M and their implementation. In this context, this report seeks to provide a comprehensive
guidance on setting up a customized O&M practice for PV plants in seven different climate
zones, four of which are general (moderate, hot and dry, hot and humid, desert in high eleva-
tion), and three which are more specific to extreme conditions (flood-prone region, cyclonic
region, snowy region). The recommendations are built based on field experiences of the con-
tributing experts from various countries representing the climate zones addressed.
The climate-specific O&M guidelines presented in this report aim to serve different functions
to various stakeholders depending on their roles in the entire value chain of PV. The most
direct application is for PV plant operators and owners by setting up an optimal and appropriate
O&M programme for their PV plants, taking into account the specific climate conditions. PV
plant engineers and designers would also benefit from relevant O&M measures and recom-
mendations to use as inputs or design criteria during the design and engineering of the plants.
At the other end of the spectrum, financing parties and investors could use these guidelines
as a benchmark for PV plant operational risk assessment, inputs which are used in the deci-
sion-making process of project finance/investment.
12
With these in mind, this report consolidates and discusses key recommendations, guidelines,
and best practices towards optimized O&M for PV plants. Task 13 puts an emphasis on best
practices through the different chapters on the site-/climate-specific aspects of PV O&M, rang-
ing from the regulatory, risk/safety and asset management level, up to the operational level
(notably the monitoring/inspections, data analytics, maintenance, and optimization). Figure 1
gives an overview of exactly all these different, yet interrelated, “components” that assemble
the overall O&M agenda throughout the technical lifecycle of PV plants.
Figure 1: Overview of O&M aspects and services for PV power systems (source: CEA-
INES).
Chapter 2, chapter 3 and chapter 4 touch on various aspects of PV plant operation and
maintenance which are applicable to installations across different climate zones:
• Chapter 2 discusses the topic of O&M performance indicators, namely ways to assess
how the PV plant itself is performing, and O&M service provider services and obliga-
tions. Parameters to measure the performance of these two aspects are discussed in
the chapter;
• Chapter 3 elaborates on the guidelines for monitoring, analysis and forecasting of PV
plant performance and safety;
• Chapter 4 presents the different types of maintenance services and important trends
(advanced inspections);
• Chapter 5 presents the guidelines for O&M in the seven different climate zones. This
chapter is comprehensive and includes not only recommendations for the climate-spe-
cific O&M, but also shares examples of field experiences in regard to PV plant reliabil-
ity, performance and safety issues in these various climate zones.
13
PERFORMANCE INDICATORS
2.1 PV Power Plant Data (Key Performance Indicators)

2.1.1 Defining the Testing Boundary
Understanding of the intended testing boundary is critical to the construction of appropriate PV
power plant KPIs. Factors of performance that lie within the testing boundary should be cap-
tured by one or more of the KPIs, while factors of performance that lie outside of the testing
boundary should be explicitly excluded. The following examples demonstrate the importance
of a well-defined testing boundary in the determination of a) which KPIs are most appropriate
and b) the particular corrections and exclusions that should be applied to each KPI.
Example 1. Electricity network downtime can have a considerable impact on the total yield of
a power station, but these impacts are not typically within the control or responsibility of an
O&M contractor, and therefore lie outside of the testing boundary. Periods of network down-
time are typically excluded when assessing the plant performance.
Example 2. If the ambient temperature at the PV power plant is higher than anticipated during
modelling and design, the PV modules may exhibit decreased efficiency. Since the O&M con-
tractor has no control over the ambient temperature, this behaviour is typically deemed to be
outside of the testing boundary and should be corrected for when calculating KPIs.
Example 3. Both the size of the PV array and the available solar resource during the period
under consideration have a direct bearing on the total yield generated. However, these factors
are not within the testing boundary when assessing plant performance, and therefore must be
corrected. Correcting the plant yield for these considerations results in the standard definition
of the performance ratio (PR).
2.1.2 Performance Ratio

The most ubiquitous measure of PV plant performance is the performance ratio (PR). The PR
is a measure of the efficiency of the entire plant as a converter of solar irradiation to AC energy
and can be considered as a measure of yield, normalised to account for both the size of the
array and the available solar resource. The PR therefore captures the combined effect of all
losses occurring within the plant, including modules, inverters, transformers, electrical cabling,
and system or network downtime.
To correct for the available solar resource, the measured plane of array insolation (Hi, typically
written as HPOA) is expressed in terms of equivalent full-sun hours by dividing by the assumed
reference irradiance level, or standard testing conditions (Gi,ref, typically written as GSTC). This
measure is known as the reference yield (Yr) [3] and is given in Eq (1).
𝐻𝑖 𝐻𝑃𝑂𝐴 (1)
𝑌𝑟 = =
𝐺𝑖,𝑟𝑒𝑓 𝐺𝑆𝑇𝐶
To correct for the plant size, the measured AC energy production (Eout) is divided by the plant
peak DC power (P0 or PSTC), to produce the specific (final) yield (Yf) as given in Eq (2).
𝐸𝑜𝑢𝑡 𝐸𝑜𝑢𝑡 (2)
𝑌𝑓 = =
𝑃0 𝑃𝑆𝑇𝐶
14
The performance ratio can then be expressed as the ratio of the specific yield and the reference
yield.
𝑌𝑓 𝐸𝑜𝑢𝑡 /𝑃𝑆𝑇𝐶 (3)
𝑃𝑅 = =
𝑌𝑟 𝐻𝑆𝑇𝐶 /𝐺𝑆𝑇𝐶
An advantage of the performance ratio is its adaptability. Factors affecting plant performance
that are deemed to be outside of the testing boundary can be accounted for by making appro-
priate adjustments to either the numerator or denominator of Eq (3). Common corrections are
presented in the following sections.
A. Module temperature correction for the performance ratio

In the calculation of a temperature corrected performance ratio, the specific yield is adjusted
to reflect the level of yield that would be achieved had the modules been operating at an
agreed-upon reference temperature. This reference temperature is often the annual (weighted)
module temperature from a standard model, or the standard testing temperature of 25°C. The
temperature-corrected PR (PRTC,ref) gives better insights into a PV plant’s operation, as it re-
duces or removes temperature effects from consideration. To estimate this correction, the ef-
fect of temperature on module efficiency is modelled as a single-coefficient linear model, using
the module temperature coefficient of power (γ), as per Eq (4).
𝐸𝑜𝑢𝑡 (4)
/𝑃
1 + 𝛾 ∗ (𝑇𝑚𝑜𝑑 − 𝑇𝑚𝑜𝑑,𝑟𝑒𝑓 ) 𝑆𝑇𝐶 𝑃𝑅
𝑃𝑅𝑇𝐶,𝑟𝑒𝑓 = =
𝐻𝑆𝑇𝐶 /𝐺𝑆𝑇𝐶 1 + 𝛾 ∗ (𝑇𝑚𝑜𝑑 − 𝑇𝑚𝑜𝑑,𝑟𝑒𝑓 )
Note that Eq (4) is sometimes written in the form shown in Eq (5). While Eq (5) is mathemati-
cally accurate (if implemented correctly), it is still recommended to use Eq (4), as this is more
intuitive, and is the form presented in IEC 61724-1 [3].
𝑃𝑅 (5)
𝑃𝑅𝑇𝐶,𝑟𝑒𝑓,𝑎𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒 =
1 − 𝛾 ∗ (𝑇𝑚𝑜𝑑,𝑟𝑒𝑓 − 𝑇𝑚𝑜𝑑 )
The module temperature correction factor, resulting in de-rating for temperatures above the
reference temperature, gives an uprating for lower temperatures. When using aggregated val-
ues for a time period to calculate PR, the mean module temperature shall be calculated as an
irradiance-weighted average.
B. Corrections for curtailment for the performance ratio

Curtailment of a PV plant’s power output may occur due to a variety of reasons, ranging from
a lack of flexibility of other generating technologies on the grid, to (local) oversupply of PV
power, or grid stability events where the system operator directs plants to reduce output [3] [4].
In such cases, the typical approach is to exclude the periods where curtailment was applied,
and then calculate the non-curtailed PR. This approach generally provides valid results, yet
15
careful analysis is required, especially for cases where curtailment occurs for more than 50%
of the time. As per IEC 61724-1, data exclusions must be recorded.
C. Soiling corrections for the performance ratio

Soiling of PV plants lowers the yield that can be generated by PV modules, compared to un-
soiled (clean) conditions. The PV industry has transitioned from using (only) an all-encompass-
ing PR where soiling is seen as part of the complete PR, to seeing soiling as an external factor
that can be measured and, with the appropriate cost-benefit analyses, result in decisions to
clean modules or not.
The soiling ratio (SR) is defined as per Eq (6), with the soiling loss (SL) defined in Eq (7) [3]:
𝑃𝑠𝑜𝑖𝑙𝑒𝑑 (6)
𝑝𝑟𝑒𝑓𝑒𝑟𝑎𝑏𝑙𝑦
𝑃𝑐𝑙𝑒𝑎𝑛
𝑆𝑅 = 𝐼
𝑠𝑐,𝑠𝑜𝑖𝑙𝑒𝑑
𝑎𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑙𝑦
𝐼
{ 𝑠𝑐,𝑐𝑙𝑒𝑎𝑛
𝑆𝐿 = 1 − 𝑆𝑅 (7)
Note that the usually defined soiling ratio SR is a measure of “cleanliness”, because SR = 1
denotes zero soiling losses, while low SR values represent high losses due to soiling.
The ratio of the soiled module power to the clean module power is used as the preferred cal-
culation method, as this better reflects soiling impacts on power losses. However, if this method
suffers from issues (e.g. jitter in the voltage measurement which would then also show up
presents as jitter in the power data), then the short-circuit current is used as the alternative
method to determine the soiling ratio. The reader is referred to [3] for the complete methodol-
ogy to determine the soiling ratio.
The soiling-corrected PR is then the PR divided by SR, as per Eq (8).
𝑃𝑅 (8)
𝑃𝑅𝑠𝑜𝑖𝑙𝑖𝑛𝑔 𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑒𝑑 =
𝑆𝑅
D. Power factor correction for the performance ratio

With increasing renewable energy fractions and renewable power fractions on grids, changes
to the output power of PV plants have become standard: no longer are these allowed to inject
active power without also providing reactive power. The consequence of this is that less active
power is injected, compared to the situation where the power factor (PF) is equal to one [5] [6].
The power factor PF is defined as the active power divided by the apparent power. Depending
on the contractual framework, a correction to the PR is made to account for the deviation in
delivered active energy, and therefore the performance ratio, as given in Eq (9).
𝑃𝐹𝑟𝑒𝑓 (9)
𝑃𝑅𝑃𝐹 𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑒𝑑 = 𝑃𝑅 ∗ ( )
𝑃𝐹𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑
16
2.1.3 Energy Performance Index

The energy performance index is an alternative approach to evaluating the whole-of-system
performance of a PV power plant. While the performance ratio is concerned with direct assess-
ment of the efficiency of the plant as a converter of solar insolation (i.e. resource) into AC
electrical energy (i.e. yield), the energy performance index (EPI) instead compares the meas-
ured yield to an expected yield (Yexp), typically obtained from an agreed-upon contractual
model as given in Eq (10).
The EPI is expressed as the ratio of the measured yield (Ymeas) and the expected yield Yexp.
𝑌𝑚𝑒𝑎𝑠 (10)
𝐸𝑃𝐼 =
𝑌𝑒𝑥𝑝
In practice, a key difference between these two approaches (PR or EPI) lies in their respective
suitability towards different types of corrections and exclusions. For example, weather correc-
tions to the performance ratio calculation can be provided through explicit modifications to the
formula, such as those described in section 2.1.2, and these may be negotiated by the contract
parties. On the other hand, equivalent corrections to the EPI require that a pre-built simulation
is run against measured weather data, otherwise known as a digital twin. In this case, the
corrected EPI may be simpler to produce than the corrected PR, but the calculation is less
transparent to the contract parties.
In [7] a lack of field measurement data for state estimation and PF optimization was compen-
sated by a real-time simulation. The concept of a PV system digital twin, reflecting actual op-
erating and boundary conditions is helpful in automated failure detection supporting O&M pro-
cedures [8].
2.1.4 Availability
Availability KPIs measure the extent to which the plant was generating electricity throughout
the period of examination. Unlike the performance ratio and energy performance index, the
performance of the plant during times of successful generation is not considered in the calcu-
lation of these KPIs.
A. Technical Availability, or Uptime
IEC TS 63019 [9] provides a framework from which the availability metrics of a PV power
system can be derived and reported. Technical availability measures the portion of time in the
period examined that the plant was successfully operating and its definition is given in Eq (11).
𝐴𝑡𝑒𝑐ℎ𝑛𝑖𝑐𝑎𝑙 = 𝑇𝑢𝑠𝑒𝑓𝑢𝑙 − 𝑇𝑑𝑜𝑤𝑛,𝑢𝑠𝑒𝑓𝑢𝑙 (11)
B. Contractual Availability
Contractual availability draws upon the definition of technical availability, but excludes periods
of downtime, such as network downtime, that are deemed not considered to be within the
contractor’s control or responsibility, and therefore lie outside of the testing boundary. Typical
exclusion for O&M contracts are network downtime, and asset owner mandated actions as
defined in Eq (12).
17
𝑇𝑢𝑠𝑒𝑓𝑢𝑙 − 𝑇𝑑𝑜𝑤𝑛 + 𝑇𝑒𝑥𝑐𝑙𝑢𝑑𝑒𝑑 (12)

𝐴𝑐𝑜𝑛𝑡𝑟𝑎𝑐𝑡 =
𝑇𝑢𝑠𝑒𝑓𝑢𝑙
C. Energy Availability
IEC 61724-3 [10] provides a method to determine the energy availability, where it is defined
as the ratio of the expected energy during availability periods to the total expected energy. It
therefore considers that periods of high irradiance are of greater value than periods of low
irradiance: for example, missing three days of energy generation in winter has less of a finan-
cial impact than missing an equal duration of time in summer. On the other hand, the lower
energy loss in winter, when PV plants participate in market prices, can be economically more
serious than in summer during the hours when energy prices are low (on the spot market or in
advance) because PV plants supply large amounts of electricity. Eq (13) and (14) show two
alternative approaches to calculating the energy availability [10] [11].
𝑌𝑒𝑥𝑝 − 𝑌𝑢𝑛𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑙𝑒 𝑌𝑒𝑥𝑝,𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑙𝑒 (13)
𝐴𝑒𝑛𝑒𝑟𝑔𝑦 = =
𝑌𝑒𝑥𝑝 𝑌𝑒𝑥𝑝,𝑡𝑜𝑡𝑎𝑙
𝑌𝑒𝑥𝑝 (14)
𝐴𝑒𝑛𝑒𝑟𝑔𝑦 =
𝑌𝑒𝑥𝑝 + 𝑌𝑒𝑥𝑝,𝑙𝑜𝑠𝑠
More, commonly used availability metrics in contracts are discussed in [12].
2.2 O&M Contractor Key Performance Indicators

It is universally understood that the quality of a solar PV power plant is primarily driven by the
selected technologies and how well the PV plant itself is designed and constructed. These
factors are primarily under the responsibilities of the EPC contractor. It is, however, equally
important to operate and maintain a well-constructed PV installation to ensure that it is oper-
ating according to its safety and functional requirements and that the PV system is producing
the amount of solar energy as targeted in the project business book. The latter falls under the
responsibilities of the operation and maintenance (O&M) service provider (operator). Just as
having a good EPC contract is important to ensuring high quality in the implementation of a
PV system, it is no less important to have a proper O&M framework for the PV plant to ensure
its proper functioning throughout its designated lifetime. The last is achieved by an O&M con-
tract which typically starts on the PV plant’s commercial operation date and lasts until the end
of the O&M contractual term.
A good O&M contract should have all necessary aspects covered under properly set terms
that could legally bind the party carrying out the O&M services, namely the PV plant’s operator.
The principal objective is to clearly describe the scope of the services and the responsibilities
of the operator to prevent avoidance of responsibilities, and to set an acceptable level of com-
pensation (usually in terms of financial penalties) in the event the operator fails to fulfil its obli-
gations under the contractual agreement. It is therefore highly recommended that the terms
and conditions in the O&M contract should leave minimal room for ambiguities and misinter-
pretations. One key is to include O&M operator key performance indicators (KPIs) in the O&M
contract which are clearly defined and could be measured quantitatively.
18
2.2.1 Different Types of O&M Key Performance Indicators

There are different key performance indicators which are typically used in the O&M contracts
in the PV sector, and they are tied directly to the performance of the PV plant (commonly used
KPIs are the performance ratio (PR) and the plant availability, as discussed in section 2.1
above.), or the maintenance services performed (response time, plant outage time, etc.).
In this section, how these three KPIs should be arranged in the O&M contract is presented.
A good O&M contract should have at minimum one of these KPIs used to be respected by the
operator. Failure to meet the agreed upon KPIs will lead to the application of financial penalties.
As PV plant lifetime spans over many years, the KPI(s) should be set for each operational year
of the PV plant, and thus is/are assessed on an annual basis.
Although having comprehensive guarantees in an O&M contract is a must, there is to-date no
one standard way on how the contractual guarantees should be set; nevertheless, there are
plenty of best-practice O&M contracting guidelines publicly available for use and reference
[13]. One main reason is that O&M guarantees are directly tied to risk and liability alloca-
tion/ownership which is dictated by commercial objectives. The guarantees are therefore ar-
ranged on a project-by-project basis as part of commercial negotiation during the O&M con-
tracting phase.
The world of solar PV O&M is extremely competitive; in addition to using more effective ways
to achieve the same scope of O&M services (e.g. via intelligent plant monitoring or automated
maintenance), reducing operator guarantees is another (but not advisable) way to keep the
O&M cost to minimum. A survey done in the Solar Bankability project [14] on O&M contracts
for large-scale PV plants developed in Europe has illustrated this case. From the study, the
authors have found that the O&M prices could range between 30 to 70% of the total operational
expenditures (OPEX) and there are quite some differences among the surveyed contracts in
the types of guarantees set, with the stronger guarantees resulting in higher costs.
In the next few paragraphs, the different types of guaranteed KPIs for O&M contracts are dis-
cussed. They correspond to the PV plant performance indicators discussed in section 2.1 and
the intention here is to give certain guidelines on how to employ these parameters to contrac-
tually bind the O&M operator in fulfilling their obligations.
A. Guaranteed Performance Ratio
Guaranteed Performance Ratio (GPR) is one of the most important and commonly used KPI
in the O&M contract. It is related directly to the energy output of the PV plant. Expressed in
relative form (percentage, %), it represents the ratio between the actual and theoretical energy
output of the PV plant. The theoretical initial plant performance ratio is calculated at the con-
tracting phase of the PV project, utilizing typical meteorological year (TMY) solar resource
datasets. With the information of the theoretical PR, the guaranteed PR should then be set in
the O&M contract.
Two different approaches have been observed in the industry practice: either setting the theo-
retical PR as the guaranteed PR, or setting a minimum GPR with a slightly lower value than
the theoretical PR. For example, a PV plant’s theoretical PR is estimated to be 85%, the O&M
contract could either set the guaranteed PR to be 85%, or set a minimum GPR at, e.g. 2%
lower, to 83%. Although both approaches are deemed acceptable, the former could be con-
sidered stricter as it leaves no buffer between the GPR and what the plant could theoretically
achieve in its production.
19
Once the initial GPR is set, the annual guaranteed PRs after the initial period is calculated
taking into account the annual PV system degradation rate (also referred to as performance
loss rate).
The assessment of the actual performance of a PV plant in operation is done by comparing
the actual energy produced over a given monitoring period (i.e. annual), to the energy yield
that the plant should deliver calculated using the actual solar irradiation over the study period.
Some key factors to achieve a proper assessment of PV plant performance ratio are therefore:
• Agreeing on a correct formula to use to calculate the PR: There are different ways a
PR formula could be set and it is important that the formula accounts for all necessary
parameters such as the effect of temperature, the availability of the PV plant, etc. This
topic is discussed in section 2.1 above.
• Using correct input data: The correct actual plant energy output and the correct actual
solar irradiation and temperature over the period of monitoring should be used. This
implies the importance of (1) having a plant monitoring system that could properly/reli-
ably measure/collect the necessary plant parameters with good data availability, and
(2) proper data processing (quality check, inclusion, and exclusion procedures). This
topic is discussed in section 2.2.3 below.
B. Guaranteed Plant Availability

Guaranteed (plant) availability (GAV) of the PV plant (percentage, %) is the second most com-
mon guarantee used in the O&M contracts for solar power plants. When discussing the varia-
ble ‘availability’, it is important to distinguish the availability at the PV plant level from the overall
availability which includes the grid availability. From the O&M operator’s perspective, the plant
level availability is the KPI of interest as the grid functioning is beyond their control.
Most good practice guidelines recommend setting the guaranteed PV plant availability to at
least 99%. In fact, it is almost unheard of to have GAV less than 99% in PV O&M contracts
nowadays, except for very special cases.
The plant availability could be assessed either using a time-based approach or an energy pro-
duction-based approach:
• A time-based availability indicates the percentage of time during which the PV plant is
producing power. It is expressed as the ratio between the duration of production activity
and the recording period (both expressed in hours). To calculate the availability, in-
verter-level data should be used. For the calculation, it is important to set a time window
which defines the period at which the PV plant is to be considered in production mode.
This could be achieved by using either (1) an irradiance threshold (minimum 30 W/m²
is usually a good approach) or (2) an hour range (from hh:mm in the morning to hh:mm
in the afternoon or evening). Beyond this window, the measured data could be excluded
in the availability calculation. In principle, the time window should be set using threshold
values which are realistic, as well as considering the geographical location and sea-
sonal conditions of the PV plant. The first approach is recommended as it is not affected
by seasonality or weather variation, i.e., as long as the minimum irradiance threshold
is achieved, the plant is considered capable of producing power. While relatively easy
20
to calculate, the drawback of the time-based indicator is that it does not allow for the
calculation of the impact of unavailability on the overall system yield.
• An energy-based availability takes into account the reference yield, and therefore indi-
cates the energy lost during times of unavailability. The energy-based availability is
calculated as the ratio between the reference yield that has been converted to electric-
ity and the total reference yield.
Similar to performance ratio, the two important factors to properly assess the plant availability
are using a correct formula and input data. The various topics related to this are discussed in
section 2.1 and section 2.2.3.
After the guaranteed performance ratio, the guaranteed plant availability is most commonly
used for setting up the guarantees of an O&M contract. The time-based availability is more
often employed than the energy-based availability, mainly because the input data is straight-
forward and directly collected from the inverter monitoring system.
Best practice recommendations call for using both GPR and GAV together in the O&M con-
tracts to strongly bind the O&M operator contractually. For stand-alone systems, the guaran-
teed PR is usually preferred because the impact of unavailability can be accounted for by
including the plant availability parameter in the PR calculation, unless the energy-based avail-
ability guarantee is used.
C. Response Time
Some O&M Operators, on occasion, may opt to guarantee a minimum maintenance response
time to faults/alarm events. This is usually expressed in minimum time lapse (in time units such
as minutes or hours) to trigger an intervention and is timed from the moment the event or fault
occurs. In this approach, the response time should be defined in responses to critical, major,
and non-critical events. Faults or events with immediate impacts on the safety operation of the
PV plant are critical and require immediate intervention (e.g., fire event). Faults or events with
major impacts on the plant production should be considered major and responded within a day
or two from the alarm trigger. Any non-critical events or faults should be responded to in the
next periodic maintenance cycle.
Figure 2: Timeline from fault to resolution, showing the O&M KPIs related to response
and resolution time, and the associated yield losses for each phase. In many cases, the
detection time will be very short, although this depends on the nature of the SCADA
system and the type of fault. Depending on the contractual framework, the yield loss
21
(the areas indicated by the letters A to D) may be excluded or not from KPI short falls
and the agreed penalties. Figure adapted from [11].
The minimum response time guarantee is straight forward. Nevertheless, the O&M contract
should include clear definitions of the different response time categories and guaranteed as-
sociated time lapse, as well as the response time that is to be measured. Having said this, it is
important to note that minimum response time guarantees should not be considered as ade-
quate to contractually bind the O&M operator to its obligations. The response time guarantees
should be used as an additional guarantee to one of the more important KPIs discussed above
(guaranteed PR, availability, or yield). The overview of the different time values is shown in
Figure 2, which also shows the detection time, for which monitoring systems are crucial.
2.2.2 Key Performance Indicator Shortfalls and Penalties

The consequences of not meeting the guaranteed values of key performance indicators in
O&M contracts are typically realized in the form of monetary compensation called liquidated
damage (LD) or performance shortfall penalty. Beyond a certain maximum limit of LD or pen-
alty (so called LD or penalty cap), the owner of the PV plant could opt to terminate the O&M
contract. The definition of and under which conditions (not only inclusions but also exclusions)
the LDs will be applied should be clearly defined in the O&M contract. Of equal importance,
the associated LDs and how to size them are also a must in the O&M contract.
Two most common LDs are penalties associated with PR and availability shortfalls. The PV
plant performance should be calculated every year over the duration of the O&M contract term,
and the actual PR and availability should be compared against their corresponding contractual
guaranteed values for the relevant monitoring year.
The amount of the LDs to be paid by the O&M Operator should be sized correctly. From the
perspective of the plant owner, the amount should at minimum compensate the loss of elec-
tricity sales revenue associated with the MWh production loss. On the other hand, the O&M
operator would like to minimize the penalty amount as this will work against their account book.
A higher penalty could lead to a higher annual O&M fee as the operator is likely to consider
the extra risk factor in order to optimise the revenue/costs. The key in O&M contract negotiation
is to find a sweet spot for both sides of the negotiating parties.
There are two approaches to size/calculate the number of LDs. The first is straight-forward and
the LDs are determined through the loss of electricity sales revenue associated with the MWh
non-production. In the formula, the difference (in MWh) (delta from actual measured, say at
the energy meter, vs guaranteed) is multiplied with the unit electricity sales price (e.g., €/MWh).
The second calculation is tiered based (for each percentage point below the guaranteed val-
ues), and the penalty is also tiered with respect to the annual O&M fee. A typical set-up in
O&M contracts translate along the line of, e.g., for every 1% PR or availability shortfall, the
O&M operator shall pay X% of the annual O&M fee. This approach is acceptable, but care
must be taken to ensure that the LD amount as calculated relative to the annual O&M is capa-
ble of covering the equivalent loss in electricity sales revenue associated with 1% of PR or
availability of the PV plant.
22
Regarding the maximum LDs or penalties cap to be set, the best-practice recommendation is
to set it at 100% of the annual O&M fee. More importantly, these penalties and LDs should be
reset at the beginning of every O&M service year.
2.2.3 Data Acceptance and Classification

Advanced monitoring platforms apply a set of criteria for data acceptance and classification
following e.g., IEC 61724 [3] standard recommendations. This crucial step ensures the quality
of the data to be used for KPIs calculation and reporting as well as for advanced data analytics
for the early detection of failures and underperformances during the operational lifetime of a
PV plant. An overview of the most common data processing and quality checks is presented
below. Further details can be found on dedicated standards such as the IEC 61724 [3].
Daylight hours are considered for the calculation of KPIs and alarm functionalities. This is typ-
ically calculated based on minimum irradiance thresholds. Depending on the application this
specific value can differ but typically ≥30 W/m² is accepted as a common minimum threshold.
This ensures eliminating night-time values and some outliers that can negatively impact the
calculation of KPIs and data analytics.
The main quality checks consist often of removing invalid readings and dealing with missing
data. Correctly labelling the raw monitoring data for further processing is crucial for an ad-
vanced monitoring platform. Several checks and filters are applied to ensure data consistency,
plausibility, and availability. These checks are typically executed automatically in monitoring
platforms and can allow certain flexibility to adjust to different stakeholder needs. For example,
among the most common checks for invalid data are: minimum and maximum physically rea-
sonable limits, outlier identification, timestamp checks, parsing of error codes, duplicates, etc.
These checks are performed at the component level and can vary depending on the compo-
nent and monitoring level.
The correct identification, labelling and reporting of missing data is crucial to enable further
analysis. The specific methodology applied must always be clearly documented in all reports
and monitoring portals. Once the missing data has been correctly labelled, different ap-
proaches can be used for its treatment. For example, the invalid or missing data can be re-
placed by data from other valid sensors, or this can be modelled using other valid data sources.
The use of digital twin models to fill in the missing data is gaining popularity as a result of the
latest improvements in artificial intelligence (AI) techniques. However, it is important to always
respect and treat the missing data in accordance with contract specifications. As recom-
mended by IEC 61724, the treatment of missing or invalid data will ultimately depend on the
goal of the measurement and the analysis. Further details and recommendations for treatment
of missing or invalid data are included in IEC 61724 [3].
2.3 Contractual Framework

In the preceding section, the key performance indicators that can be used to contractually bind
the O&M service providers to their obligations and responsibilities were discussed. In addition
to these, the framework of the O&M contract is equally important as it defines all the necessary
aspects together with the terms and conditions and serves to avoid any avoidance of or ambi-
guities in the responsibilities of the contracting parties.
23
2.3.1 O&M Contract Framework

Typically, an O&M contract is drafted and prepared by a PV plant owner with the involvement
of legal expertise, and by taking inputs from technical and/or financial advisors. Different par-
ties usually prepare their own template based on contracts used in the PV market. There are
also publicly available templates for download (some free, others at a cost). Among the com-
mon templates used in the O&M contracting is the one of the International Federation of Con-
sulting Engineers (so called FIDIC template) adapted for solar PV power plant O&M. There is
also an O&M contract template prepared from the joint effort between the Terawatt Initiative
and the International Renewable Energy Agency (IRENA) [13]; this template is specific to solar
power plants.
Regardless of the types of templates used, the O&M contract should include at the minimum
elements discussed hereafter; focuses are placed on elements that are directly or somewhat
related to technical aspects of PV plant.
1. Parties Involved
The legal identification and registered location of the contracting parties should be clearly de-
fined.
2. Contract Term and Commencement
The duration of the service should be clearly stated. Whether the term could be extended (how
long and how) should also be specified.
The effective date of service commencement, including any conditions precedent, should be
defined.
3. O&M Fee / Contract Price
The O&M service fee (expressed in kWp per year, or for total installed plant capacity to be
maintained) should be presented.
Any indexation of the fee should be clearly specified.
Any specific exclusions of the O&M service fee should be outlined clearly.
4. Scope of Services
The O&M contract should clearly specify which activities should be executed as part of the
O&M service provided. To be noted, it is not enough to mention only the scope of the works,
but also to clearly specify if the activities performed are covered by the contract fee.
The following scopes are common to O&M contracts:
Plant operation around-the-clock monitoring and supervision, including but not limited to:
o Daily performance monitoring, maintenance program execution supervision,
alarm/fault detection to trigger appropriate maintenance).
o Operation documentation management.
o Equipment warranty and claim management.
o Representing the plant owner with the grid operator.
o Support in management of change.
o Support in maintaining plant operation permit and permit of evacuation when-
ever applicable.
o Security and surveillance.
24
PV plant maintenance: The O&M contract should clearly define (as well as distinguish) the
types of maintenance (preventive, corrective, predictive) to be performed. Equally important,
the maintenance programme should be adapted to the climate the PV plant is operating under.
For the latter, refer to the subsequent chapters of this report for climate specific maintenance
guidelines.
o Preventive maintenance (discussed further in Chapter 4.1):
■The frequency and procedure should be adapted specific to the equip-
ment to be maintained.
■ Site maintenance should be included.
■ Waste management (temporary storage, disposal and removal) should
be the responsibility of the O&M Operator.
o Corrective maintenance (discussed further in section 4.2):
■ Clear definition of what is included in the service scope and if the activ-
ities are also covered under the O&M fee.
■ Response time for different types of alarm, fault events. In general,
faults which will impact the safety operation of the PV plant, or 100%
plant outage should be treated as critical and responded to within four
to eight hours. Faults impacting major production loss should be treated
as major and should be addressed within 24-48 hours.
■ Notification procedure to plant owner.
o Predictive maintenance (optional). The progress in the field of data analytics
and artificial intelligence/ machine learning has led to the development of pre-
dictive maintenance of solar plants. Predictive maintenance utilizes historical
operational data of PV plant mined from the monitoring system and parameters
perceived from the environment (weather condition such as irradiance, temper-
ature, rainfall, etc.), with the goal to learn the behaviour or performance pattern
of the PV plant and using this to anticipate and plan for maintenance interven-
tions before an event or fault occurs. An example is using predictive mainte-
nance to anticipate and plan for a PV module cleaning cycle. Predictive mainte-
nance is not a standard feature yet in the O&M scope of service.
o Spare parts and consumables management, including storage, replenishment,
management of stocks.
Any other services should be specified including reporting to asset management, for example.
Like guarantees to O&M key performance indicators, the scope of works of the O&M operator
drives the price of the O&M service. The fuller the scope, the more expensive the service is
likely to be. In some cases, the plant owner would exclude certain scopes to minimize the O&M
costs; this is usually a commercial decision. In the latter approach, the plant owner is recom-
mended arranging for the excluded service scopes elsewhere, which could make it more diffi-
cult to hold one contractor responsible for underperformance.
5. Contractor’s Responsibilities
In addition to specific activities to be carried out under the scope of services, the responsibilities
of the O&M service provided should be specified. Among the commonly included items are:
25
• The O&M services should comply with requirements prescribed in the technical speci-
fication’s documentation of the PV plant and equipment suppliers’ requirements.
• The O&M services should be carried out according to good industry practice, and com-
ply with all applicable laws, consents, and permits.
• The O&M service provider is responsible for providing all other material, equipment,
tools and consumables necessary to perform the O&M services.
• The O&M service provider is responsible for ensuring the health and safety of all
maintenance activities performed at the project site and is responsible for the safety of
its and any approved subcontractor personnel in performing the services.
• The O&M service provider shall carry out the services so that the PV plant achieves
the performance requirements and guaranteed key performance indicators.
• The O&M service provider shall ensure it has all the necessary permits and insurances
for it to carry out the services, including maintaining the validity of such permits and
insurances.
• Provide to the plant owner, in a timely and expeditious manner, all reasonable assis-
tance and information to allow for the plant owner to obtain the permits for which the
plant owner is responsible to obtain.
6. Owner Responsibilities
A good O&M contract should also clearly define the responsibilities of the plant owner to facil-
itate the O&M service provider to carry out its services. Among these obligations are:
• Obtaining all required permits and consents to operate the PV plant, as well as to pro-
duce and inject electricity (whenever applicable).
• Making sure to give the O&M operator the right of access to the project site to carry out
the O&M services.
• Providing to the O&M operator, in a timely and expeditious manner, all reasonable as-
sistance and information for the operator to obtain the permits needed for by the oper-
ator to perform its services.
• Paying to the O&M operator the contracted O&M fee according to the agreed payment
schedule.
7. Subcontracting
The contract should clearly mention if the O&M service provider is allowed to subcontract. If
this is the case, it is recommended that all subcontracting should be reviewed and approved
by the plant owner prior to the start of the services. If there are major parts of the services
subcontracted, this should be mentioned during the O&M contract negotiations.
Moreover, if subcontracting is agreed upon, it should be clearly specified that the O&M service
provider is responsible for its subcontractor(s).
8. Spare Parts and Consumables
Spare parts could be categorized into spare parts of plant equipment (such as PV modules,
inverters, low and medium voltage fuses, tracker motors, structures, etc.), and general con-
sumable small items such as fasteners (screws, bolts, nuts etc.), cables etc. Having a good
26
level of available spare parts of plant equipment will help to avoid prolonged plant down-
time/outage due to equipment malfunctioning or damage.
Three main aspects of spare parts management should be clarified in the O&M contract: the
storage, the replenishment of used spare parts, and the management of stocks. For consum-
ables, best practise recommendation usually calls for the O&M service provider to be respon-
sible for all the storage, replenishment, and management of consumables.
For the plant equipment however, as discussed in section 4.2, the plant owner and the O&M
operator must agree on who is responsible for the storage, replenishment, and management
of stocks of spares. In general, keeping inventory of spare parts falls under the responsibility
of the O&M operator. However, storage and replenishment of spares could be arranged differ-
ently. However, it is important to have this clearly specified in the O&M contract. In the case
spare parts replenishment is not foreseen in the O&M service scope, the plant owner should
arrange for a separate spare part replenishment set-up, or foresee future maintenance ex-
penses on key equipment replacements (usually in the form of maintenance contingency or
reserve account, as discussed in section 4.2).
In addition to defining the responsible parties to spare parts management, it is important to
have in the O&M contract two other aspects on spares:
• Warranty periods of replaced defective plant components, and
• Transfer of title: in general, since the plant owner pays for the spare parts, the spares
should be the sole property of the owner, and they shall be returned to the owner at the
end of the O&M contract term.
9. Key Performance Indicators (KPIs) / Performance Guarantees
Please refer to section 4.2 for detail on recommended KPIs and the corresponding liquidated
damages.
10. Dispute, Arbitration
The applicable law for dispute and arbitration should be clearly defined.
Best practice recommendation also includes an allowance in the O&M contract for the involve-
ment of an independent third-party expertise in case of technical disputes.
2.3.2 Influence of Region on O&M Framework

Despite the significant efforts by all stakeholders in the PV industry to increase the standardi-
sation in design and O&M of PV systems, there are still differences that will result in the O&M
framework varying, in some cases even within a country or region. Examples of differences:
• Legislative requirements for O&M personnel to be permanently on-site, such as in Mex-
ico [11], which enables the possibility for higher minimum guaranteed availability values
to be negotiated.
• The proposed [15] (and subsequently declared invalid [16]) legislation in Queensland,
Australia, which would have mandated that only qualified electricians could work with
PV modules, inclusive of physical mounting.
• Variances in the maturity of PV market stakeholders (policymakers, developers, EPCs,
O&M providers) and the impact of the PV system location can strongly affect O&M
costs and the ability to maintain response times [12], [17], [18].
27
• As discussed in [17], it protects PV stakeholders, especially asset managers and O&M

providers with a multinational or multi-regional footprint to fully acquaint themselves
with the specific aspects of the PV system in question, as not all practices and assump-
tions can be fully replicated everywhere.
2.3.3 Other Influences

Given that PV systems have expected lifetimes of 25 years or more, it is highly likely that
regulatory or legislative changes will affect these during their lifetimes. In particular, the rede-
sign and repowering of PV systems may occur during an O&M contract.
As defined in [11], revamping involves replacing components, mainly inverters and modules,
without substantially changing the plant’s nominal power, while repowering (or upgrading) in-
volves replacing components, which may substantially change the plant’s nominal power.
Depending on the scope of the revamping or repowering activities undertaken (whether it
affects a portion or the whole PV plant), it may be useful to redefine or renegotiate O&M KPIs,
or even have a new O&M contract start once the revamping/repowering has been achieved.
28
POWER PLANT OPERATION
3.1 Plant Performance Monitoring Guidelines

In order to obtain operational excellence – next to the use of best value-for-money materials
and high-quality installation competencies – performance monitoring, analysis and reporting
tools are key for PV plant stakeholders to differentiate. This is to be compared to industrial
automation and enterprise resource planning (ERP) systems for a production facility.
However, monitoring capacities should be aligned with the type of installation to monitor; there-
fore, an initial classification of those is needed before starting monitoring strategies.
A monitoring system allows to follow the energy flows within a PV system. In principle, it reports
on the parameters that determine the energy conversion chain. These parameters, along with
the most important energy measures in terms of yields and losses, are illustrated in Figure 3
below.
Figure 3: Energy flow diagram in a grid-connected photovoltaic system (source: 3E

[19]).
Utility-scale PV plants use complex highly customized monitoring systems, comparable to

those used in industrial automation projects. On the other hand, monitoring of residential and
commercial systems is typically performed by inverter and data logger manufacturers providing
29
in most cases free or semi-free access to web-portals collecting the data from the local hard-
ware.
As stated in the IEC61724-1 [3], the required accuracy and complexity of the monitoring sys-
tem depends on the PV system size and user needs. A proposed classification for monitoring
systems and suggested applications is given in the Table 1 below.
Table 1: Monitoring system classifications and suggested applications as per IEC61724-

1 [3].
Typical applications Class A Class B Class C
High accuracy Medium accuracy Basic accuracy
Basic system performance X X X
assessment
Documentation of a perfor- X X
mance guarantee
System losses analysis X X
Electricity network interaction X

assessment
Fault localization X
PV technology assessment X
Precise PV system degrada- X

tion measurement
30
Class A Class B Class C

High accuracy Medium accuracy Basic accuracy
Recommendations on measurement’s accuracy requirements
Uncertainty on irradiance < 3% < 8% n/a

measurements
Irradiance sensor inclination 1⁰ 1.5⁰ 2⁰

angle alignment accuracy
Irradiance sensor azimuthal 2⁰ 3⁰ 4⁰

angle alignment accuracy
Inverter-level measurement ± 2% n/a n/a

of input (DC) voltage and
current
Inverter-level measurement ± 2% ± 3% n/a
of output (AC) voltage and
current
Plant-level AC electrical out- Class 0.2 S as per Class 0.5 S as per Class 2 as per
put active power and energy IEC 62053-2 [20] IEC 62053-22 [20] IEC 62053-21
[21]
Power factor Class 1 as per Class 1 as per n/a
IEC 61557-12 [22] IEC 61557-12 [22]
3.1.1 Analysis of Monitored Data

Several “standard” PV monitoring solutions, enabling data collection and data provision under
a given service level agreement, do exist in the market. Smart PV performance monitoring
going beyond these standard solutions is indispensable, especially for large utility-scale PV
plants. Today, such smart PV monitoring systems should have at minimum the following fea-
tures:
• Extended monitoring on the plant level including device alarm collection and external
equipment monitoring (AC cabinet, security, etc.)
• Complex aggregation functionality up to portfolio level for all analysis and contractually
required KPIs
• Notifications based on device alarms and on functional alarms calculated from complex
analysis functions
• Advanced aggregation and prioritization functionality of device and functional alarms
on portfolio level
• Inclusion of accurate and independent irradiation references, e.g. based on satellite
data
31
• Expected behaviour calculation based on validated simulation models

• Intervention management system integrated with alarm handling and analysis function-
ality on component level
• Business intelligence enabled for advanced reporting from intervention up to portfolio
level.
Further insights on state-of-the-art smart PV performance monitoring systems are provided in
section 3.2 “Performance Analysis and Optimization”.
The strategy for monitoring data depends usually on contractual agreements and it is directly
related to the scale of the PV plant (e.g., utility-scale PV plant in contrast with an industrial
rooftop PV system). Class B and Class C monitoring systems usually will only take care of
basic performance parameters with a simple sampling plan. However, Class A plants deserve
smart performance monitoring and the same happens with the alternatives for analysis of mon-
itored data.
A typical example for the Class A installation is the case of multi-MW PV plants built under
strict contractual conditions and LCOE targets. Both the parameters to be monitored and the
analysis strategy to follow their evolution should be clearly defined at PV plant design level
ensuring that adequate sensors will be considered for integration in the PV plant. Not only
physical parameters (e.g., irradiation, temperature, voltage, current, DC power, AC power,
etc.) but also synthetic parameters as defined depending on plant characteristics, can then be
tracked. Specific attention should also be given to solar forecasting in order to estimate the PV
energy yield to be achieved on e.g., a daily or even intra-daily basis.
For the analysis of monitoring data, statistical data control and evaluation of tendencies on the
parameters, more than individual values, are taken into consideration. This strategy allows
having as soon as possible an “a priori” identification of unwanted evolution, thus avoiding
potential faults and/or (excessive) performance degradation. Artificial intelligence and big data
techniques are usually applied for the analysis and interpretation of the monitored data in this
case. A detailed explanation of the application of these methodologies from small industrial
systems up to commercial utility-scale ones, can be found in “Report IEA PVPS T13-07:2017
Improving Efficiency of PV Systems Using Statistical Performance Monitoring” [23].
Reporting of monitoring analysis is done in line with contractual agreements and is usually
accessible via web application. Depending on the intelligence of the monitoring system, and
the interest of the customers in knowing about the performance of their installations, the evo-
lution of the parameters, and proposals for actions to be taken can be presented in a detailed
way.
3.1.2 Best Practice Guidelines

As introduced in the previous section, different PV plant monitoring solutions are available in
the market for different PV plant sizes and stakeholder needs. In all cases it is recommended
to always follow standards and best practice recommendations to ensure the correct calcula-
tion of all KPIs such as, for example, PR and availability calculated from the monitored data.
These minimum requirements are detailed in international standards such as IEC61724-1 [3],
covering different aspects ranging from data quality checks to remove invalid data points up to
the correct identification, labelling and reporting of missing data as summarized in
section 2.2.3 “Data acceptance and classification”. Furthermore, consistency checks to ensure
32
that the correct PV plant configuration is selected as input into the PV monitoring software is
key to enable further analysis of the data through the application of PV digital twin models.
Finally, best practice guidelines such as, for example, SunSpec Alliance [24] should be fol-
lowed carefully to ensure open data accessibility for optimized transition between monitoring
platforms during the operational lifetime of the PV plants.
A non-exhaustive summary of best practice guidelines for PV plant performance monitoring is

presented in the table below. These guidelines reflect the current state-of-the-art in monitoring
and PV plant performance assessment, as represented in various standards and research
results including those from the PerformancePlus project [25], among others, as well as the
SolarPower Europe Operation & Maintenance Best Practice Guidelines [11].
33
Table 2: Best practice guidelines for PV plant performance monitoring.

Measurement Recommendations Explanation
Irradiance Solar irradiance in the plane of the array (POA) is The pyranometer class (e.g., Class A, Class B, or Class C) accord-
measurements measured on-site by means of at least one irradi- ing to ISO 9060-2018 will define the final uncertainty. Furthermore,
ance measurement device according to ISO best practice is to apply at least one pyranometer in the horizontal
9060:2018. plane and two pyranometers in the plane of the PV array. In case of
different array orientations within the plant, at least one pyranometer
is required for each orientation. Use of satellite-based irradiance
data is strongly recommended as back-up in case of sensor’s una-
vailability and for regular data accuracy checks.
Module The temperature of the modules should be meas- The temperature sensor should be glued with appropriate and stable
temperature ured with a temperature sensor attached directly to thermally conductive glue to the middle of the backside of the mod-
measurements the backside of the module. The use of an appropri- ule (in the centre of the cell) in the middle of the array table.
ate device with an accuracy < ±1 °C is recom-
mended.
Local It is best practice to measure ambient temperature Wind and ambient temperature data are normally not required for
meteorological and wind speed on-site. Ambient temperature calculating PR unless this is a contractual requirement/agreement
data should be measured with a shielded thermometer (e.g., according to specific recommendations such as for calculating
and wind speed is measured with an anemometer, the weather corrected PR as proposed by NREL [26]). However,
at 10 m height above ground level. they are required when the PV plant is to be modelled in operation
or in retrospect.
String Individual string current measurements should be String level monitoring, as compared with inverter level, allows for
measurements considered when not supported by the inverters. more precise trouble-shooting procedures and advanced data ana-
lytics at higher spatial granularity. Measuring current and voltage of
every string with at least 15 minutes resolution is recommended. To
reduce costs, the current sensor can potentially measure more than
one string (depending on the sensor’s measurement range and the
number of modules per string), but it is not recommended to parallel
more than two of them.
34

Inverter It is recommended to collect data from all measured Although the precision of inverter integrated measurements may not
measurements variables at the highest possible resolution, includ- be very high and this is often not documented by the manufacturers,
ing inverter alarms being a valuable source of infor- these measurements become a valuable source of information when
mation for fault diagnosis. Best practice for measur- performing fault diagnosis and root cause identification.
ing inverter-based variables is a <1 min sampling
and a temporal granularity of up to 15 minutes.
Energy meter A high accuracy energy meter to measure energy Gathering energy meter data is required for invoicing purposes but
produced and consumed by the plant is normally re- it is also the best reference for measuring energy and calculating
quired by the utility. When this is not the case it is a plant PR and yield and is much more accurate than using inverter
best practice to install a meter with a maximum un- data. It is recommended to have a meter with two communication
certainty of ± 0.5%, especially for plants >100 kWp. bus ports as well as automatic meter reading (AMR) service from
When providing reactive energy, it is necessary to the utility or meter operator. For meters that can store historical data
have metering not only for active power P, but also it is a best practice to have a monitoring system capable of retrieving
reactive power Q, and/or power factor PF in place. the historical data to avoid any production data loss in case of mon-
itoring system outages.
Control settings It is recommended to monitor all control settings of Monitoring all control settings of the plant is important for contractual
the plant at inverter level (DC side) as well as grid reporting or performance assessment. Many plants apply control
injection level (AC side) if available. settings for local grid regulation (injection management) or optimiza-
tion of the market value of the PV generation portfolio (remote con-
trol).
Alarms As a minimum requirement, the monitoring system As best practice, the following alarms should also be sent by the
must have the following alarms: monitoring system:
• Loss of communication • String without current
• Plant shutdowns • Plant under UPS
• Inverter stops • Discretion alarm (or alarm aggregation)
• Plant with low PR • Intrusion detection
• Inverter with low PR (e.g., due to overheating) • Fire alarm detection
35

AC circuit / It is recommended to monitor the position of all AC Whenever possible, it can also be useful to read and register the
switches through digital inputs. alarms generated by the protection relay control unit via the commu-
protection relay
nication bus.
Infrared When performing IRT measurements, global POA ir- IRT imagery provides clear and concise information about the exact
radiance must be at least 600 W/m2 and should be location (i.e., detection) and type of a large range of failure modes
thermography (IRT)
continuously measured on-site. It is recommended (i.e., diagnosis, classification) [28]. As such, IRT is broadly used in
to use infrared cameras with a minimum optical res- both preventive and corrective maintenance. However, the use of IR
olution of 320 x 240 pixels and a thermal sensitivity thermography alone might be insufficient to reach a conclusive di-
(resolution) of at least 0.1 K. Measurements must be agnosis on the root cause and its impact in quantitative terms (power
taken at a distance which ensures that the resolution output losses). Therefore, it is recommended to combine this with
of the infrared image allows each solar cell to be de- data analytics and complementary field tests [29].
picted by at least 5 x 5 pixels. IEC TS 62446-3 [27]
provides a complete list of such recommendations.
IV curve tracing In prior point current and voltage measurements at The shape of the string I-V curve allows identifying the origin of the
string level have been already described. However, loss of power by giving information on the potential cause of the fail-
the I-V curve trace gives valuable information con- ure. The theoretical addition of individual I-V curves of PV modules
cerning certain degrees of degradation on the can be considered reference and differences with respect to it in
PV modules performance that not in all cases can power or field factor can be interpreted in order to find the problem
be found by just Isc and Voc values at string level. and later go to the string to identify the malfunction on individual
Measurement is quite fast so not big loss of produc- modules, unwanted shadowing or any other problem to be solved at
tion is expected and should be done in line with the O&M level.
procedures established by IEC-61289:2016.
The necessity of performing individual module I-V curve measure-
For the individual module IV curve measurements, it ments should be carefully assessed, as it results in a costly activity
is also necessary to disconnect full string and time in terms of production loss and labour time. Other inspection tools
to measure might be longer per module, with the ex- (visual inspection, IRT imagery) should be tested before deciding on
pected loss of production. using it.
36

Electrolumines- Electroluminescence (EL) imaging is another tool for EL is an image technique that allows identifying important damages
cence on-site quality control and fault diagnosis at PV mod- in the PV modules (as cracks and micro cracks) that in some cases
ule and solar cell level. It is recommended to perform do not provoke a significant impact on the performance at initial
EL measurements to a certain sample of PV mod- level, but over time could evolve with outdoor exposure into im-
ules upon their arrival to the PV plant’s site and prior portant performance degradation. Detecting the potential presence
to their installation/commission, in order to detect of defects in this first stage of degradation is the key for preventing
unwanted problems due to transport and/or han- their future propagation or extension when the electrical perfor-
dling. Moreover, additional EL inspections of PV mance may be already compromised. Concerning another type of
modules right after their installation and during plant defects, EL can also detect damage due to impacts on the module,
operation, are typically recommended for scrutinous not only in the front side but also in the back side, or other kinds of
quality control and detailed fault diagnosis. EL problems due to stress produced when they were installed in the
measurements are usually carried out in dark condi- field. Finally, EL technique is also capable of detecting possible deg-
tions (late evening, night, or early morning times). radation related to electrical problems as PID (potential induced deg-
However, new solutions allow for performing these radation) [30].
EL measurements also at day light [29].
Soiling It is recommended to measure dust or dirt accumu- Several methodologies exist for soiling monitoring, the most basic
lation (soiling) locally in order to optimize cleaning being human inspections. A widely used soiling measurement
measurements
schedules and thus revenues particularly in areas method is using ground-based soiling reference modules consisting
where soiling represents an important risk [31]. of a module that remains soiled, a cleaned reference cell, an auto-
matic cleaning station and measurement electronics. Upcoming dig-
ital solutions for soiling monitoring include the analysis of satellite
imagery with remote sensing techniques, machine intelligence algo-
rithms and statistical methods.
37
3.2 Performance Analysis and Optimization

Asset managers and operators need to analyse the monitoring data at different levels, going
from large scale portfolios down to plant level and even components analysis. Smart alarming
systems based on user specific thresholds such as, for example, business plan data or real-
time deviations between on-site assets are crucial to enable fast reaction times and perfor-
mance optimization [11].
As introduced in section 3.1 above, smart PV performance monitoring systems go beyond
simple analytics and alarm systems by providing fault detection and diagnosis capabilities giv-
ing actionable recommendations and insights for the operator and asset manager. The value
and related complexity of big data analytics changes depending on different O&M objectives
as represented in Figure 4 below.
Figure 4: Big data analytics - Value and complexity for different operation and mainte-
nance objectives (source: 3E).
For example, a simple analysis based on monitoring data provides limited value, but it is not
complex to execute and therefore all monitoring platforms today offer such basic functionality.
However, this typically provides limited information and will still require the operator or asset
manager to take further actions based on the information, i.e. no actionable recommendations
are provided at this stage. Subtle trends that would otherwise go unnoticed until the next site
inspection and that indicate underperformance or upcoming component or even system fail-
ures, can in most cases, be identified remotely through regular performance data analysis of
the PV plant by using a smart solar performance monitoring system.
38
Today, however, users (especially O&M managers/engineers) require further insights on spe-
cific failures and underperformances and therefore, smart monitoring systems start providing
additional functionalities around fault detection and fault diagnosis with insights and particu-
larly, with actionable recommendations. A PV performance monitoring system today provides
not only analysis features but also fault detection and diagnosis capabilities providing insight
on the failure or underperformance root cause(s) and generating clear and actionable recom-
mendations for performance optimization.
A “predictive” smart solar monitoring would in theory provide foresight capabilities towards op-
timization of O&M activities. For example, providing insights on the optimal replacement cycle
including cost per time unit under the optimal age-based replacement policy information would
be highly valuable information for an operator towards predictive maintenance activities. How-
ever, given the high complexity of this type of analysis, and the limitations in terms of lack of
communication between devices and lack of standardization, this technology has not been
deployed yet at full industrial scale and has not been sufficiently validated and demonstrated.
A prerequisite for a good predictive maintenance service is that the devices on-site can provide
information about their state, in such a way that the O&M contractor can evaluate trends or
events that signal deterioration of the device [11]. The asset owner or operator that wants to
benefit from predictive maintenance should, as a best practice, select “intelligent” equipment
set with enough sensors, and opt for a smart monitoring service. Furthermore, a standardiza-
tion of status and error codes through inverters and data loggers within the same brand should
be followed and, in the future, this standardization should become common to all manufactur-
ers.
3.3 Power Plant Controller and Requirements for Grid Compliance

Continuous grid compliance is a requirement for all generators connected to the electrical grid.
The evolving knowledge about PV systems and improvements in their capabilities, coupled
with their increased uptake on grids (and therefore larger relative and absolute impacts on
grids) [32], have resulted in increased requirements for grid compliance. This has evolved from
“simple” anti-islanding protection functions of PV inverters [33] in the early 2000s, to additional
grid support by these by the 2010s, such as reactive power control, (power) drop control, power
factor control and the ability to remotely set values (i.e. enable curtailment) [34]. Updates to
inverter standards over time also increase grid compliance requirements [35] which in some
cases can be resolved through software (firmware) upgrades, and in other cases require hard-
ware replacements or retrofits. The specific context where the PV plant operates can also
affect grid connection or compliance requirements, for example where PV systems operate in
high Renewable Energy Fraction (REF) environments [36].
For larger, possibly less solicited grids with lower REF values, grid compliance requirements
are currently not as stringent, nor are the requirements homogeneous over similar geographic
regions (albeit in different jurisdictions), with Distribution System Operators (DSOs) and Trans-
mission System Operators (TSOs). For example, PV inverters in Belgium must have reactive
fault current support deactivated [37], whereas just across the border in Germany, PV systems
must provide reactive current support during faults [38], [39]. Homogenisation and adoption of
more stringent connection requirements for neighbouring countries is expected over time, with
a significant step taken having been in the EU, with the Commission Regulation (EU) 2016/631
39
[40], which established a network code on requirements for grid connection of generators,
which is then combined with the appropriate standard according to the PV plant size, either
EN 50549-1 (PV connected to LV networks) [41] or EN 50549-2 (MV networks) [42]. In the
USA and associated jurisdictions, IEEE 1547 [43] sets the requirements for distributed energy
resources, which includes PV systems. Some countries or geographical regions have their
location-dependent challenges or characteristics, such as found in Australia’s National Elec-
tricity Market with its extensive and sparse high-voltage grid which is seeing large amounts of
renewable power coming online while thermal generators (mainly coal-fired stations) are de-
commissioned [44], posing challenges to the traditional power system operation paradigm.
Given the different physical characteristics of inverter-based generation compared to synchro-
nous generators, PV systems typically apply for negotiated generator performance standards
[45] to connect and remain connected to the grid.
While compliance with grid requirements can be expected to last as an inherited trait from the
design, construction and commissioning phases, changing requirements for grid compliance
are very likely to occur over the plant’s lifetime, given the technical and financial lifetimes of
PV plants and the rapid deployment of renewable power and changes to the grid. The period
to implement changes to the plants (repowering, retrofits) will likely be excluded from O&M
KPIs, yet the O&M operator will have to adapt its processes to ensure that the PV plant con-
tinues to operate within the updated compliance framework.
PV plants of larger size are often required to be ready for dispatch by the grid manager, re-
quiring added capex for installing the control and communication functionalities between the
dispatcher and the plant. The dispatched plant can be curtailed, and the power factor changed
by the dispatcher as the grid requires, at the cost of lost energy generation to the PV plant.
As PV plants increase in size to many tens of megawatts, the implementation of enabling con-
trol by the grid manager becomes more challenging due to the modular distributed nature of a
PV power plant. The solution for this is a power plant controller (PPC), which centralises control
over the plant so that it behaves as a single entity, in line with the applicable regulations. De-
pending on the jurisdiction and PV plant power rating, a certified PPC is obligatory for ensuring
grid compliance, e.g., with [38], [39].
Given the above, it is the responsibility of the O&M operator to ensure compliance with all
applicable regulations and grid codes, knowing that if and when these codes change, the O&M
contractor’s scope of operation and the KPIs originally defined may have to be re-evaluated.
3.4 Power Generation Forecast

3.4.1 Importance within PV O&M
More than 80% of the PV plants installed worldwide today have been commissioned only within
the last 5 years. This reflects a constantly and significantly growing share of solar PV electricity
in the global energy mix and, of course, an impressive increase rate of PV deployment and
feed-in of electricity to the power grid. The growing PV penetration could also trigger the need
for new regulations, in order to guarantee grid stability and correct balancing of electricity sup-
ply/consumption at all times, inevitably resulting in curtailment and losses to plant owners.
40
The two main challenges towards addressing such high penetration rates of PV systems are
variability and uncertainty; in other words, the fact that PV power output exhibits variability at
all timescales – from a few seconds to daily, seasonal or (multi)annual degree – and the fact
that predicting such variability (PV power forecasting) is a task of high intrinsic uncertainty.
Within a PV asset management and O&M plan, PV plant power forecasting is an important
element of the PV operations, which refers to the adoption of forecasting tools to calculate the
expected PV power production for a certain timeframe, based on weather forecasts, satellite
data, or measured irradiance and PV power. PV power forecasts depend on country/climate
and plant/site. They can serve different stakeholders related to a PV plant, from PV asset
owners up to grid operators and energy traders.
PV power forecasting services are generally offered by PV monitoring service providers and/or
O&M contractors; however, other external service providers can also provide this function,
based on proprietary or publicly available meteorological forecasts, satellite data, statistical
methods, and site observations/measurements. Requesting such kinds of (optional) services,
particularly from the O&M contractor, implies different service level agreement(s); which, in
turn, may have an influence on the contract agreement for electricity dispatching, between the
asset owner and the trading service provider.
From PV plant financing, project development and O&M perspective, the importance of PV
power forecasting is multifold [46]:
PV power forecasting is the basis for trading PV-power on day-ahead or intra-day electricity
markets. To ensure grid stability, deviations of forecasts from the actual PV power must be
compensated with costly balancing power. Depending on the specific regulations for grid- in-
tegration of PV power in different markets, there are different models to cover these costs.
With feed-in tariff models, plant owners receive a fixed price per kWh, and grid operations are
in charge of balancing costs. For direct marketing of PV power, in some markets, PV power
marketers directly must cover balancing costs (e.g., in Germany). In other markets contract
agreements come with an associated level of reliability or performance “fidelity”, implying fi-
nancial penalties for underperformance (e.g., China). Therefore, accurate PV power forecast-
ing is indispensable for the bankability of PV plant projects not receiving a fixed feed-in tariff.
PV power forecasting allows for commitment and dispatch of conventional resources (e.g.,
diesel generator) with their given ramp-up time. This is particularly crucial in cases of remote
systems (energy islands, geographical islands) or in micro grid applications.
PV power forecasting is a basis to optimize energy management for PV plants combined with
storage in order to e.g., avoid curtailment or to compensate for forecast errors in short-term
market.
When benchmarked with PV monitoring data, PV power forecasts can facilitate cross-checking
and quantitative assessment of underperformance issues. Here especially satellite based now-
casting can give valuable information.
PV power forecasts can be used as an additional input to maintenance schedules.
41
3.4.2 Requirements, Methodologies and Recommended Practices

Essential characteristics of PV power forecasts include the forecast horizon, the spatial and
temporal resolution, and the update frequency. The requirements with respect to these char-
acteristics vary with the envisaged application, the data availability, etc. For applications rele-
vant to PV power systems or the power market in general, forecast horizons are typically up
to 48 hours, the time resolution is 15 minutes to one hour (depending on the power system or
the market). Forecasts are provided either for single PV power plants, for the overall PV power
in control areas or for grid nodes. In this sense, common commercial services provide day-
ahead and intraday power forecasts.
Typically, day-ahead forecasts are delivered in the morning hours (for the next day) and up-
dated once or twice during the same day; whereas intraday forecasts are delivered and up-
dated automatically every 15 minutes or every hour by the forecast provider. PV forecasts with
longer time horizons (typically one week or more) comprise a more relevant practice when
long-term planning, dispatching, unit commitment or maintenance schedules are to be ad-
dressed.
In terms of implementation methodology, we can distinguish three segments of solar PV power
forecasting, namely: i) physical methods, ii) statistical and artificial intelligence (AI) based
methods and iii) hybrid models. Most PV forecasting products and services combine several
of these methods, often in a case-by-case context. An overview of PV power forecasting meth-
ods and associated accuracies is given by Lorenz et al. [47], where also limits and benefits of
the different methods are discussed.
Irradiance prediction is an essential step in solar power prediction. PV power predictions are
derived from irradiance predictions with PV simulation models and/or statistical methods (Fig-
ure 5), except time-series approaches for very short-term forecast horizons that are solely
based on PV measurements. As Figure 5 depicts, irradiance prediction can be implemented
based on different forecasting models for different forecast horizons (cloud-motion sky imager
and satellite data, numerical weather prediction) and in combination with statistical learning
approaches for optimized site-specific predictions. Next, PV power prediction steps refer to the
conversion of irradiance to PV power with parametric PV simulation models and/or statistical
learning approaches. A last step refers to the upscaling requirement of the regional PV power
prediction [48].
42
Regional
PV power
forecasts
Local
PV power Upscaling
forecasts
PV Statistical learning
irradiance to power conversion
measurements and/or post-procssing
parametric modelling PV system

POA irradiance description
PV power PV simulation nominal power
tilt and orientation
prediciton
Irradiance
prediciton
Irradiance Statistical learning
measurements optimized site-specific predictions
Numerical
Cloud motion Cloud motion
weather
sky imager satellite
prediction
Forecast minutes hours days

horizon
Figure 5: Overview of basic modelling steps in PV power prediction. Adapted from

[49].
In the physical methods segment, PV power forecasting is implemented by analysing atmos-

pheric data (e.g., cloud cover, temperature, pressure, or humidity), numerical weather predic-
tion (NWP) tools or cloud observations (i.e., sky and satellite imagery).
Numerical weather prediction (NWP) models, routinely operated by meteorological services,

are employed for forecast horizons from several hours up to 15 days- ahead several days.
NWP models describe and predict the physical and dynamic processes in the atmosphere,
employing current weather state observations as a basis to forecast weather patterns. Predict-
ing dynamic changes of the atmosphere, including formation and dissipation of clouds as well
as advection, essentially relies on this physical modelling. NWP-based modelling is highly
computing-intensive. It is used for predicting the atmospheric characteristics on either a world-
wide (global) scale or a continental/regional scale (mesoscale). Broadly used and validated
NWP global models are the Integrated Forecast System (IFS) operated by the European Cen-
tre for Medium-Range Weather Forecasts (ECMWF) and the Global Forecast System (GFS)
by the US National Oceanic and Atmospheric Administration (NOAA).
43
Examples of mesoscale models are the North American Mesoscale (NAM) model and the
Weather Research and Forecasting (WRF) model. Generally, NPW models are differentiated
mainly in terms of their employed physical models, their spatial resolution and the input pa-
rameters. Typical resolutions for Global NWP models are ten kilometres to 50 kilometres and
temporal resolutions of one, three or six hours. Mesoscale models have finer grids of typically
three kilometres to 10 kilometres with hourly resolution.
For intra-day horizons of several hours ahead, forecast with higher spatial and temporal reso-
lution and higher accuracy can be inferred from satellite data. In the satellite-imaging domain,
the cloud pattern is determined through the analysis of images (in both the visible and the
infrared spectrum), generated with the use of satellite-based sensors and imaging systems.
Multiple, consecutive satellite images are combined to determine cloud structures during ear-
lier recorded time steps and to generate cloud-motion vector fields, which, in turn, can be uti-
lized to forecast future locations of clouds and their variability. As such, these methods are
used effectively in predicting irradiance (particularly GHI) with a temporal resolution of 5
minutes to one hour with a spatial resolution of one kilometer to five kilometres up to several
hours ahead. For intra-hour PV power forecasting sky-imagers offer the potential to forecast
irradiance with a temporal resolution down to minutes or even less and a spatial resolution in
the range of several meters to 100 m.
Sky-imagery can be effectively employed for sub-kilometer observations of cloud shadows,

e.g., over a utility-scale PV power plant, suited thus for PV power forecasting of high spatial
resolution. Sky imaging systems employ digital – typically, charge coupled device (CCD) type
– cameras and refer to the ground-based identification of clouds, cloud motion determination
and measurement of cloud height above ground. PV power forecasting via sky imagery anal-
ysis comprises of four steps:
1. Image acquisition near the forecast site, using the sky imager.
2. Sky image data analysis for cloud pattern recognition.
3. Cloud-motion vector estimation.
4. Short-term probabilistic and deterministic forecasts of cloud cover, irradiance, and PV
power.
Overall, the detailed cloud cover (structure, extent, and motion) analysis achieved by sky im-
agery renders this method a very suitable tool, for very short-term PV power forecasts or a so-
called nowcasting, for certain PV installations. Maximum forecast horizons of sky imager fore-
casts depend on the extent of the monitored cloud scenes and cloud velocities. They typically
range between ten minutes and 30 minutes ahead.
PV power forecasting based on statistical and artificial intelligence (AI) methods – also
mentioned as time-series and machine learning methods respectively – refers to the imple-
mentation of advanced analytics on historical data of solar irradiance, for a given PV installa-
tion site. Statistical and AI methods maybe employed as pure ‘time-series approaches” solely
based on local measurements. They also may be employed for statistical post-processing and
to infer PV power from NWP or satellite-based forecasts, which are described as hybrid models
below. The user community for statistical methods and artificial intelligence refers to the former
44
as “statistical models without exogenous input” and to the latter as “statistical models with
exogenous input”.
Common examples of statistical and AI methods applied for PV power forecasting, include the
use of artificial neural networks (ANNs), neuro-fuzzy models, the support vector machine
(SVM), hidden Markov models (HMM), regression analysis and auto-regressive (AR) models.
AI or machine learning algorithms are well suited for PV forecasting, all being based on the
same principle, i.e., of utilizing existing data for modelling parameters and creating (self) learn-
ing patterns towards classification, regression, and prediction.
Today, ANNs represent the most common machine learning technique, accounting to nearly a
quarter of all methods used for PV power forecasting, notably due to their ability in resolving
complex and non-linear forecasting models. The accuracy and robustness of machine learning
(ML) based PV power forecasting depend on both the training method and the evaluation met-
ric for the quality of the forecasts. Also, the access to quality-controlled PV power measure-
ments is crucial.
Pure time-series approaches benefit from the high accuracy of on-site measurements of irra-
diance and PV power and the high autocorrelation for short time lags in time-series of solar
irradiance. However, changes in cloud conditions, such as approaching clouds, can hardly be
predicted based on local measurements alone.
Hybrid models in PV forecasting combine two or more of the aforementioned methodologies,

thus aiming to leverage their individual benefits and/or overcome certain limitations, in terms
of accuracy, forecast horizon or temporal and spectral granularity. A common case is the fusion
of NWP models with ANNs (i.e., hybrid-physical models), where the outputs of the former are
used as inputs for the latter, for training the latter, in a self-learning loop.
In practice, integrated approaches can be implemented in nearly any form and combination,
i.e., linear-linear, nonlinear-nonlinear, or nonlinear-linear models, whereas all can be classified
into either competitive or cooperative approaches [50]. In this sense, it is also particularly com-
mon to utilize multiple predictors, e.g., by coupling statistical and learning machine techniques
(i.e., hybrid-statistical models), to obtain aggregated decisions for improved PV forecasting
accuracy or to further self-training. In their review study, Diagne et al. [50] discuss in detail how
integrated PV forecasting methods outperform individual ones in multiple aspects.
On the basis of reported literature and today’s research status in the field [50], [51], [52], [53],
Table 3 summarizes typical spatial and temporal resolution as well as forecast horizons for the
PV forecasting methodologies that have been discussed.
Summarizing, good forecasting practice for PV power plants requires numerical weather pre-
dictions as input for day-ahead forecasting and a combination with satellite data and/or online
PV power measurements for intra-day forecasting. Further, statistical post-processing is a rec-
ommended practice in all cases, which requires measured PV power data as a basis to adapt
the forecasts.
Here, near-real time data feeds from PV monitoring systems towards PV forecast providers
are beneficial, as input for shortest term forecasting and for continuously updating the training
45
of machine learning models with recent data. Finally, it is highly recommended as best practice
to warn and communicate (when applicable) all scheduled outages and the expected duration
of forced outages to the forecast provider.
Table 3: Typical forecast horizons and typical temporal and spatial resolutions for dif-
ferent PV power forecasting methods. Statistical/AI tool for post-processing can be ap-
plied to all models and forecast horizons.
Temporal
Forecast Horizon Spatial Resolution
Resolution
NWP 2 – 15 days 1 - 6 hours 1 km - 50 km
Satellite based
2 – 6 hours 5 – 30 minutes 500 m – 5 km
cloud motion
Sky Imager 1 second –

10 - 30 minutes 2.5 m – 500 m
cloud motion 1 minute
Statistical / AI tools
(no exogenous input) 10 minutes – 2 days 1 second – 1 day PV plant
3.4.3 KPIs/Accuracy Metrics and Uncertainties
Introducing and understanding key metrics and considerations in PV forecasting, allows to

better assess and compare the performance and accuracy of the different PV forecasting mod-
els, as well as to weigh the impact of different parameters on them.
In their work, Zhang et al. [54] point out that classical KPIs may not fully address the different
requirements of grid operators, O&M managers, or investors in relation to real-case PV power
plants. Indeed, for the latter, certain metrics that “penalize” large errors are indispensable,
considering that such forecasting errors also have highly negative impact, in both finiancial and
operational stability/reliability terms. From this perspective, PV forecasting KPIs, or metrics
can be classified into four categories:
1. Statistical metrics for different time and geographic scales.

2. Uncertainty quantification and propagation metrics.
3. Ramp characterization.
4. Economic metrics.
The most common statistical KPIs for PV forecasting quality are the root mean square error
(RMSE) and the mean absolute error (MAE), which are defined in Eq (15) and Eq (16).
46
𝑁 (15)
1 2
𝑅𝑀𝑆𝐸 = √ ∑ (𝑃𝑓𝑜𝑟 − 𝑃𝑎𝑐𝑡 )
𝑁 𝑖=1
1 𝑁 (16)
𝑀𝐴𝐸 = ∑ |𝑃𝑓𝑜𝑟 − 𝑃𝑎𝑐𝑡 |
𝑁 𝑖=1
where Pfor and Pact are the forecasted and actual (measured) power output and N refers to the
sampling count (size). In other words, RMSE points out and “penalizes” large errors in a square
error; while MAE shows the average difference between the actual and modelled (forecasted)
values, thus being suitable for evaluating uniform forecast errors. Typically, RMSE and MAE
of PV power forecasts are evaluated using day-time values only and normalized to installed
PV power.
Yet, both RMSE and MAE are only unbiased for cases of Gaussian distributions. Besides,
when used stand-alone, they present certain limitations such as in the case of over forecasting
tendency, as highlighted by Zhang et al. For this reason, other statistical parameters, particu-
larly skewness and kurtosis are also used. Further, with respect to uncertainty quantification
and ramp characterization in PV forecasting, other metrics such as the Rényi entropy, the
standard deviation of the power forecast errors, and the “swinging door” algorithm are pro-
posed. Recently, Yang et al. [55] proposed to apply the well-established Murphy–Winkler
framework for distribution-oriented forecast verification as a standard practice to analyse and
compare solar forecasts.
Furthermore, the comparison to simple reference models, quantified by “skill scores” is a fre-
quently applied check to assess forecast quality. A forecasting tool has skill if it is able to
outperform trivial models. In solar irradiance forecasting, the simplest and widely used refer-
ence model is persistence, i.e., the assumption that cloud conditions stay the same (“persist”)
in the future.
In addition to the computation of KPIs, at least basic visual analysis is recommended. A direct
comparison of measurements and forecasts in scatter plots or two-dimensional histograms
and time-series is very helpful to develop a better understanding of forecast performance.
From the economic metrics’ perspective, as explained by Antonanzas et al. [51], the way grid
operators address irradiance (thus solar and PV yield) variability is through reserves. Practi-
cally, the greater the penetration of solar energy, the bigger the energy reserves (and their
associated costs) must be, to mitigate potential variations. On this basis, the number of oper-
ating reserves and, consequently, the operating costs can be significantly reduced when ac-
curate PV power forecasts are achieved and leveraged. Perez et al. [56] propose the concept
of firm power forecast to evaluate the economic value of forecasts and as an operational strat-
egy to integrate increasing amounts of intermittent solar generation on power grids. The costs
incurred in transforming imperfect into firm predictions define the new metric: these include the
costs of energy storage and output curtailment necessary to make-up for any over/under pre-
diction situations.
47
Finally, specifying the expected uncertainty of PV power predictions gives valuable additional
information to forecast users and helps them to assess the risk associated to using the forecast
as a basis for decisions. Probabilistic forecasts or confidence intervals provide situation-spe-
cific uncertainty information, e.g., as a function of the cloud conditions and the time of the day.
As reported in the literature [54], using the 95th percentile of forecast errors is a simple and
frequently used option to e.g., quantify the minimum required operating reserves.
3.4.4 State of Play

With the increasing contribution of PV power to the electricity supply in many countries, a
strong need for reliable PV power predictions arose during the past decades. Reacting to this
new and rapidly evolving situation on the energy market, with great need for reliable solar
power predictions, increasing effort has been spent on developing and enhancing solar irradi-
ance and power prediction models. Today, PV power prediction systems are an essential part
of electric grid management in countries that have substantial shares of solar power genera-
tion, as e.g., Germany or Spain. Several companies offer good quality worldwide PV power
forecasting services.
PV power forecasting is offered for different temporal and spatial scales depending on the
needs of different stakeholders. Many forecast providers address different applications and
stakeholders, but most of them have a specific focus and/or specific strengths, depending on
their key customers and/or their background.
As described above, PV power forecasting integrates different models and data: Meteorologi-
cal models for irradiance forecasting for the different temporal and spatial scales, PV simula-
tion for irradiance to PV power conversion, statistical or machine learning approaches to im-
prove forecasts or infer PV power predictions, and PV power measurements as a basis to train
ML forecast algorithms. A good PV power forecasting service integrates all these elements,
but not necessarily all of them must be operated by a PV power prediction services provider.
Companies offering PV Power forecasts often have a strong background in one of these ele-
ments. Private weather forecasting companies extend their meteorological services to addi-
tionally offering PV power forecasts, providers of long-term satellite data also offer satellite
based short term predictions, experts in machine learning set up PV power prediction services
using data from other companies, and PV monitoring service providers exploit their access to
PV power measurements for PV power forecasting.
Also, several forecasting services have been developed to specifically address the require-
ments of grid-operators, partly by companies also offering wind power predictions. Grid oper-
ators typically require forecasting services for aggregated PV power in their control areas as a
basis for allowing PV power on the grid and for congestion management rather than forecasts
for single PV power plants. Mostly these companies have extended their portfolio to also pro-
vide forecasts for single PV power plants and direct marketing.
Any PV power forecasting service offering day-ahead predictions - be it for control areas, grid
nodes, or single PV plants - requires NWP irradiance forecast as essential input. Here, it is
good and common practice to employ NWP forecasts from public or private weather services.
IFS forecasts provided by ECMFW are a popular choice because of their high accuracy [57]
as well as freely available GFS forecast, or meso-scale models forecasts of national weather
services. Some forecast providers also run their own weather models, with the freely available
48
WRF model often used. Top forecast providers do not rely on one single NWP model but inte-
grate several models. Simple averaging is beneficial here, because forecast errors of different
models are not fully correlated and therefore partly cancel each other. More advanced forecast
combinations exploit strength and weaknesses of different models for different weather situa-
tions.
Intra-day PV power prediction services mostly combine NWP predictions with online-PV power
measurements as additional input for very short-term horizons of five minutes up to several
hours ahead. Here, the access to these online measurements is of critical importance. There-
fore, PV power prediction services are frequently offered by PV monitoring service providers
and/or O&M contractors. Another option is that access to online PV power measurements is
directly given to the trading service provider, e.g., a direct marketer, or a PV power prediction
service provider. Direct marketers often run their own PV power forecasting tools or cooperate
with PV power prediction service companies.
A valuable additional data source for improving PV power forecasts for several hours ahead
are satellite-based cloud motion forecasts. Only a comparatively small number of companies
worldwide operate satellite based short term predictions with good quality. Mostly, these com-
panies offer their own PV power prediction services with the satellite-based forecasts being a
special feature. Some also deliver the satellite-based irradiance forecast to other PV power
prediction providers.
For very high-resolution intra-hour forecasting, few specialised companies offer sky-imager
based PV forecasting, which is a comparatively new field, with respect to model and application
development.
PV power forecasts in principle may be offered based on irradiance predictions in combination
with PV power simulation. However, forecasts for all forecast horizons benefit from high quality
PV power measurements as a basis for improved forecasts with statistical or ML algorithms.
For single plant forecasts, they may be used to adapt the forecasts to the specifics of a PV
plant, including e.g., shading or adaptation of orientation which is often not correctly specified.
All high-quality PV power services providers adapt their forecasts to PV power measurements
in some form, which makes access to these data crucial.
Finally, for customers of PV power forecasting services, besides the used models, data, and
accuracy of a service, also the form of implementation as well as general considerations with
respect to contracting service partners are of importance. Most PV power service providers
offer to deliver forecasted PV power data to their customers, e.g., using a web-service or by
providing the data for download on a web-platform. Though, also the implementation of PV
power forecasting tools on-site is offered by some providers for customers who prefer this
option, e.g., because confidential data is involved. With respect to deciding for a service part-
ner, it may be convenient for an asset owner to choose a partner he is already working with,
e.g., a monitoring and/or O&M service provider or a trading service provider. Such kinds of
additional, optional services imply different service level agreement(s) and may also impact
contract agreements with other partners, e.g., using a forecasting service from the O&M con-
tractor, may have an influence on the contract agreement for electricity dispatching, between
the asset owner and the trading service provider.
49
3.5 Power Plant Safety Considerations

3.5.1 Overview
Grid connected PV power plants are expected to have a technical lifetime of decades, with
maintenance, repairs or modifications required to ensure continued power production. Several
PV plants have already demonstrated their ability to operate over time spans of decades [58].
Maintenance activities of PV plants expose personnel to safety risks which must be known,
understood, and mitigated appropriately. Legislation and standards such as ISO 45001 [59],
ANSI Z10 [60] have been put in place to ensure general workplace safety, often in conjunction
with ISO 9001:2015 [61] which aims to ensure that risk-based thinking is applied into the man-
agement system; in this case, the operation and maintenance of PV plants. Since 2017, the
IEC 63049 [62] standard is available for the PV industry which provides guidelines for effective
quality assurance in PV systems operation and maintenance. Here, quality is taken as an en-
compassing term that also covers risks and risk-based thinking.
Asset owners have the legal responsibility to warrant the health and safety1 of the people inside
and around a PV power plant. By having an O&M company under contract the responsibility
of safety is internally transferred to the service contractor.
The commissioning of a PV power plant is usually executed by the engineering, procurement,
and construction (EPC) company engaged by the owner. It must confirm a regular commis-
sioning and that the plant can safely be operated, in line with local laws and regulations, and
any other project-specific requirements. After the proof of such procedure having been suc-
cessfully passed, is delivered, the plant (and associated risks) is handed over to the principal,
who in turn may engage an O&M contractor.
The safe and reliable operation of PV power plants is essential to ensuring that the asset can
deliver power as expected. Among the best practices applied is the use of specialised and
qualified staff to deal with special components such as power conditioning units/inverters, who
are sourced either from the manufacturer itself or from a qualified and authorized service part-
ner of the manufacturer, where the use of quality management and risk management systems
such as IEC 63049 and ISO 9001:2015 are essential.
A key tool to minimise or eliminate risks is the hierarchy of controls methodology, as shown in
Figure 6. At the O&M stage the ability to eliminate risks is limited compared to the design and
construction phases (where safety can be designed into (or omitted in the design of the PV
plant) unless additional investments are performed.
1
Safety as the state of being safe as person; freedom from the occurrence or risk of injury, danger, or
loss.
50
Figure 6: Hierarchy of controls, adapted from [63].

Consequently, the focus of this section is more on best practices around known and typical
safety issues that can be encountered during the operation and maintenance of PV plants.
3.5.2 Qualification of PV Maintenance Staff

Assuring a safety power plant operation presumes skilled and experienced service staff. More-
over, the attitudes of staff and systems in place can mitigate or exacerbate risks that are pre-
sent. To keep the plant in electrical safe conditions the maintenance staff should have exper-
tise and knowledge related to the construction and operation of the electrical equipment and
installations of the PV power plant. It also should have been given special safety instructions
for high power and/or high voltage DC systems, with central inverter PV plants operating with
1000 Vdc or 1500 Vdc, whereas internal AC ring mains for larger utility-scale power plants or
distributed over large areas can operate anywhere between 12 kV and 36 kV. Typically, gen-
eral work on electrical components is restricted to qualified or registered electricians, while
work on inverters or transformers require additional training and/or certifications.
Skilled maintenance personnel should have the following qualifications to maintain PV power
plants safely and successfully:
• Qualification as certified electricians or equivalent trained persons
• Current first-aid training certification with qualification for emergency first aid for electric
shock
• Training in how to deal with hazards and risks associated with operation and maintain-
ing electrical devices and plants
• Familiarity with the special high power DC application of the PV plant
51
• Knowledge of how the PV power plant works and is operated

• Skilled or instructed according to valid electrical safety standards and the use and care
of a referring safety equipment
• Selecting and using personal protection equipment
• Knowledge and understanding of technical plant documents and component manuals
with all described safety precautions
• Authorisation to connect and disconnect medium-voltage sources if maintenance ac-
tivities are required for medium-voltage components or other parts of the power plant.
• Beside electro-technical knowledge, knowledge of relevant PV standards and direc-
tives is desirable
The execution of different maintenance works typically require different qualifications and
safety equipment.
To ensure the health and safety of workers at a PV power plant, a tailored system of main
theoretical PV basics, safety training and procedures must be established, where
IEC 63049:2017 [62] and ISO 9001:2015 [61] can inform the course design and the subse-
quent implementation and monitoring on an organisational basis. The training then also serves
as proof of the special qualification required to confront any incidents occurring.
A. Personal Protective Equipment

Authorized service staff / personnel must be equipped with appropriate personal protective
equipment (PPE) including the following:
• Safety glasses
• Ear protection
• Steel-toed safety boots
• Safety hard hats
• Padlocks and barriers
• Double-insulated tools
• Appropriate qualified measuring instruments, regularly checked and re-calibrated.
For instance: appropriate LV meter to verify that the circuits are de-energized (1 000 V
AC and 1 000 V rated, or 1 500 V DC rated, dependent on max. DC system voltage)
B. Special maintenance safety instructions

• Cleaning of PV modules
• Replacement of defective PV modules
• Safely changing a DC fuse link in a field connection / combiner box and in the DC inlet
of a PV inverter
52
• A strict smoking ban inside the PV power plant: fire risk due to increased/rapid vegeta-
tion growth in and around PV arrays
For a change of DC fuse links during plant operation, a verification with a meter is a must to
confirm that the DC circuit is de-energized. Opening a module string is only allowed after meas-
uring and confirming that there is no current on the circuit. Some DC switches used in the field
of a PV plant (e.g., field combiner boxes) may not be designed for load-break operation (dis-
connect switch). Disconnect switches shall be labelled as non-load-disconnecting. These must
never be opened while the system is operating. Before opening such a DC switch that is not
rated for load break, the referring sub-system should in any case be switched off respectively
by disconnection of the inverter in question.
3.5.3 Electrical Safety Considerations

In this section we describe relevant safety considerations regarding electrical safety in general,
with requirements regarding qualification of maintenance staff, training of the staff, personal
safety equipment, and PV special instructions. While the overall hierarchy of controls method-
ology remains valid and applicable, it is advisable to consult more specific sources such as
those discussed by [64].
PV power plant operation should take place under safe, continuous, and trouble-free condi-
tions. Planned service routines should keep the plant in an up-to-standard condition, be well
scheduled and executed and result in minimum energy losses. Extraordinary plant or sub-plant
shut offs are often considered in the yield expectations or considered by an availability factor.
In practice, when such availability factor of a PV power plant is considered in the initial financial
calculations, a clear definition is indispensable. The same applies for O&M contracts which
include an availability clause.
The electrical safety of a PV power plant in general presumes a consequent electrical safety
concept in design, by selection of high-quality products, for PV applications certified compo-
nents and the strictly conducted concept in practice [65]. Among others, health, and safety
rules, referring regulations, relevant national and valid international photovoltaic standards /
codes are considered. Knowledge of manuals of main components ensures that professional
maintenance directives will be executed, and special safety precautions applied.
A. General electrical safety of PV power plants

The electrical design of PV arrays of large PV power plants is typically classified as ground-
free, Safety Class II and designated as IT grid (insulated earth). This classification refers to all
electrical components such as PV modules, DC cabling, field connection boxes must have
double isolation. Safety Class II or double insulated electrical appliances is designed in such
a way that it does not require (and must not have) a safety connection to electrical
earth/ground.
At the international level, according to the EU standard [66], the operator of the electric facility
(the PV plant) is responsible to indicate which testing period is suitable for the facility. This
procedure leads to a self-responsibility and self-certification situation, which is mitigated by the
incentive for the principal: the operator must ensure the reliable (and therefore safe) operation
of the plant, partly also to be able to claim insurance in the event of an incident.
53
However, there are national norms and regulations or even contractual agreements with grid
operators that reverse the above argument, as these require all (PV) power plants connected
to the (medium or high voltage) grid to be always operated at a safe operational level.
B. Electrical safety from the viewpoint of maintenance

As in any facility generating electric power, electrical safety in a PV power plant is of the highest
priority. One major difference with other types of power generation plants is that the primary
on-site electrical system is associated with the production of DC power. In cases of DC short-
circuits in the field, hazardous arcs may occur. Such a case may affect persons and cause
damages of material to a certain extent by fire, as well as follow-up repair efforts and costs,
depletion of spare parts and finally, energy production losses.
In large PV power plants nearly all types of maintenance work bear safety risks. A well-orga-
nized maintenance strategy applies a special maintenance risk management to provide an as
safe as possible working environment. It covers and presents all relevant hazards and presents
adequate solutions for mitigation. Since each PV power plant has unique characteristics, the
risk management must be adjusted individually. However, most PV power plants share many
features, and organizations that perform O&M on multiple plants typically have procedural doc-
uments which have common sections, combined with specific sections for the individual plants.
For servicing and maintaining PV power equipment, a team of minimum two people is required.
In case of any accident on site or weakness of one person the other can immediately initiate
help actions. An emergency list with addresses and phone numbers to be contacted and in-
formed must be available for personnel at short notice, which typically must be carried with the
person or in the vehicle.
C. General Safety Precautions

As in any other electrical installation, basic safety precautions must be strictly adhered to in
PV power plants’ operation. Plant owners are legally liable for the asset, and, among others,
they are responsible for safety issues. Risks are covered by an indemnity insurance. In case
of an accident inside the PV plant the asset owners and operators are obliged to give evidence
to have taken all preventive measures for protection of people against accidents. Even visitors
of a PV plant, before entering the site, must undergo safety instructions and act in accordance
with precaution guidelines. Similarly, a list of emergency contact numbers should be made
available for easy access by staff and visitors, which can reduce the time between an issue
occurring and a call for help being made.
To ensure safety at the place of work whilst e.g., interventions, maintenance or repairs are
taking place, it is necessary to respect the following internationally accepted five rules. These
ensure proper lockout-tagout (LOTO), so that work is performed on de-energised systems as
shown in Table 4.
54
Table 4: Safety Rules for Works in Electrical Facilities [11], [12], [67].
Rule Action Special conditions / Remarks spe-
cific for PV plants
Rule 1 Isolate / separate completely (iso-
late the installation from all possi-
ble sources of electrical power)
Rule 2 Protect against reconnection, fix
(protect against reconnection) in
the open position all the breaker
components or switching device in
the on position, or adopt preventa-
tive measures when that is not fea-
sible
Rule 3 Check that voltage is not pre-
sent, verify there is no electrical Caution, PV special 1: at daylight
power, after having previous identi- PV modules / module strings are volt-
fied the place of work and the in- age-carrying. For getting real voltage-
stallation which has been placed free conditions covering modules /
without electrical power module strings with opaque material is
required.
Caution, PV special 2: works at an

opened PV inverter cabin requires veri-
fication of absolute de-energized invert-
ers: internal capacitors must be de-en-
ergized! The time to be considered
must be provided by the inverter manu-
facturer. It may e.g., last 5-15 minutes!
Rule 4 Ground and connect in a short
circuit
Rule 5 Protect against nearby power
sources and cordoned off the work-
ing zone
The above-mentioned rules are especially valid for MV/HV switching equipment and discon-
necting devices in transformer stations or in medium stations of large PV power plants. When
switching operations are carried out at such points, only authorised specialists are allowed to
perform such operations, wearing protective equipment if necessary.
3.5.4 Weather Conditions Increasing Risk during and Post Events

A. Thunderstorms
Due to massive wide-spread metallic structures and cabling PV power plants are more likely
to be affected by thunderstorms than other structures in the area. Even though a lightning
55
protection system is typically installed following best practices and according to local regula-
tions – which should protect the investment - it does not guarantee a protection of people inside
the plant. Thus, during a thunderstorm the complete plant site must be vacated by maintenance
personnel for safety reasons. Best practices in this area include proactive weather monitoring
(e.g., “push” notifications on smartphones to warn of approaching storms) by O&M personnel
to prevent any personnel being on-site if a thunderstorm is likely to appear.
B. Heavy rain
PV power plants often cover a large area. Heavy rain falls may turn the terrain to mud, and
areas may be inundated. Walking around the site by maintenance crews in such conditions
will see increased risks to personnel such as sprains or dislocations, the potential for vehicles
to become bogged, as well as increased risk of electrocution due to reduced electrical insula-
tion compared to dry conditions. In some areas, heavy rains can even result in flash flooding
occurring. Given the increased risk of electrocution due to the lowered resistivity of the area,
proactive de-energisation (ideally, remotely operated) of inverters or strings prior to movement
among arrays is recommended.
C. Damaging winds
In the event of high winds (above 50 km/h sustained, or wind gusts above 80 km/h) being
forecast for the area of the PV plant, it is recommended to reschedule maintenance activities
to another time. This reduces the risk of elements (in the worst case, modules) becoming air-
borne while the maintenance crew is onsite. After such an event has occurred, staff should
exert special care, as modules may have come loose, or the presence of debris within the
power plant may increase risks, such as the likely presence of broken glass and potentially
exposed or frayed electrical wiring.
D. Natural hazards
The presence of insects or venomous animals in PV plants may be a risk throughout the year,
or peak in certain periods. In the case of poisonous snakes or arachnids, rapid response pro-
cedures must be known and adhered to; in some cases, this requires having anti-venom on
site if bites occur. In other cases, having a wildlife catcher on speed dial can help in safely
catching the insect or animal. Note also that killing such animals is often illegal according to
local legislation, and significantly increases the likelihood of bites occurring. Works in mosquito
areas require special protection such as mosquito nets and repellent, and in specific cases
additional prophylactic actions, such as taking anti-malaria pills.
E. Fire safety ground-mounted
Fire constitutes a major risk and hazard to the safe operation of a PV plant and its surround-
ings. To reduce the risk of fire spreading over large areas, ground-mounted PV plants are
typically designed with fire breaks, such as roads or areas where vegetation is kept to a mini-
mum. Moreover, local rules and regulations often require site visits and consultations with local
fire brigades for development approvals, aiming at ensuring that fire risks are eliminated or
minimised at the design stage. The visit should include the information on how the fire brigade
is embedded in the PV plant’s emergency plan and the communication strategy. An appropri-
ate firefighting strategy must be developed [68]. Guidelines for fire-fighters to operate in or
near PV plants have already been published in multiple countries such as Japan, the United
States, Germany, and Australia. On the other hand, the British BRE National Solar Centre
56
released some recommendations for photovoltaic industry with the input of British fire brigades
[69] and a detailed literature and standard review [70].
Vegetation management such as clearing of weeds and cutting the grass on site is a key com-
ponent of fire risk reduction. Additional fire risks stemming from PV components (modules,
inverters, transformers) require on-site inspections combined with monitoring software and/or
the use of SCADA to rapidly shut off parts of the plant in the event of fire, or the detection of
indicative parameters [23], [71].
3.5.5 Building-Mounted PV Safety Considerations

Building-mounted PV systems have specific safety risks and considerations in addition to
ground-mounted ones. These risks stem from the fact that the PV installation is mounted on a
building, and the PV installation is at a height where there is a risk of falls.
Building-mounted PV can be either applied (BAPV) on or physically integrated (BIPV) in the
building skin. BAPV systems include all the PV installations added to flat or inclined roofs.
BIPV elements can be part of the roof, façade or part of a balustrade or another integral part
of the building (envelope). Such elements have both structural and electricity production func-
tions. Apart from “pure” building applications, PV can be applied or integrated in individual
infrastructure, such as noise barriers, railway station shelter roofs, and carports, which typically
have different or additional security requirements, compared to BIPV or BAPV.
Typically, building mounted PV installations are electrically connected via the power connec-
tion of the building to the grid. As such, the electric low-voltage installation standard IEC 60364
is applied, particularly Part 4 ( [72], [73], [74], [75]) and Part 7-712 [76] for safety and PV in-
stallation aspects respectively.
In practice, constructors, and O&M managers of BAPV or BIPV installations must ensure com-
pliance with certain additional norms, technical specifications and directives that regulate:
• Requirements for non-flammability and fire load for modules and underlying construc-
tion elements, particularly thermal insulation, or roof membranes, but also cables and
installation tubes
• The distance of the array from firewalls or roof edges
• Fall protection related to working at heights
• Fire protection depending on the buildings height and use, which can have conse-
quences for the PV system installation
• Dimensioning and routing of cables in the building, and their allowed load with regards
to fire protection
• Glass safety and construction of modules for overhead glazing or balconies
• Periodic controls of the electrical installations, in particular of building mounted PV in-
stallations
Furthermore, a PV installation must be integrated into existing protection elements of the build-
ing, such as lightning and surge protection or the equipotential bonding, as well as preventive
fire protection measures. These elements should be verified with all involved parties and the
insurer.
Many of such regulations depend on the use of the building, as well as on national and local
regulations. Also, in general, BAPV has less stringent requirements because in its case the
PV array is clearly separated from the building envelope, whereas in the BIPV case the array
57
consists of multi-functional BIPV elements or are even construction products, with dedicated
standards [77], [78]. An overview of the existing regulations for building-mounted PV, including
guidelines for prevention of losses can be found in existing literature [79], [80].
Some of the additional requirements for building mounted PV stem from the fact that in and
around the building persons are present and not only in the case of maintenance of the PV
installation, as is often the case for ground-mounted systems.
For example, in case of an inclined roof in snowy regions, persons passing by may be in danger
because of a sudden roof snow avalanche. Depending on the situation, temporary measures
that eliminate the risk – cordoning off the affected area, clearing off the roof from ice and snow
- or constructive measures such as a snow retention system must be previewed.
In building-mounted PV installations, which are connected via the power connection of the
building, self-consumption of the PV electricity in the building and net metering are also con-
sidered, depending on the local legislation. In addition, batteries or an uninterruptible power
supply can be integrated to enhance auto-consumption or to ensure electric power supply for
critical usage in case of grid failure, but it must also then be an integral part of the security
concept.
A. Maintenance
Maintenance of BAPV or BIPV systems can be done according to a maintenance plan that has
been already developed at the planning phase of the PV installation. With this, it is ensured
that visual inspection and maintenance can be done safely and efficiently, at predefined time
intervals. On flat roofs that normally include fall protection measures such as lifelines at anchor
points, sufficient spacing should be considered between the PV rows, so that the modules can
be cleaned, and any unwanted vegetation or other shading obstacles can be safely removed.
It is necessary to facilitate maintenance interventions to building-mounted PV installations
while avoiding damage to building components such as tiles, skylights, or roof membranes and
that their function is not compromised. Similarly, safe access to other technical equipment
installed on flat-roofs, e.g., thermal PV collectors or ventilation mono-blocks, must be ensured,
for trained technical personnel. Typically, the O&M plan of the building-mounted PV system
should be an integral part of the O&M plan of the building.
B. Working at height: risk of falls
This risk stems from working at heights and on the (roof) structure onto which a PV system is
installed, where there is a risk of falls. For these reasons, staff that perform maintenance on
building-mounted PV systems often require additional training, which may or may not be man-
datory per jurisdiction. Examples of such requirements range from “working at heights” certifi-
cations, to the need to use climbing harnesses (and sometimes to have climbing certifications)
to enable safe access to PV equipment in exposed areas (façades, close to roof edges). Ex-
amples of poor workmanship or system design have resulted in loss of life e.g., as the roof
could carry the weight of the PV equipment, but not the additional weight of the O&M person-
nel, resulting in the roof collapsing.
Potential risks can also arise from roof windows and smoke/heat extraction installations, which
must either be breakthrough-proof or have a sufficiently indicated location and large safety
distance from the PV elements, to prevent accidental falls.
58
C. Fire safety for building-mounted PV systems

For building-mounted PV systems, the installation must not only be “fireproof”, but also "fire
brigade friendly". Hence, it is recommended that fire brigade experts inspect such PV systems
in advance, particularly for systems of installed capacity >30 kWp (corresponding to approxi-
mately 200 m² solar module area). Ideally, the installer or owner should carry out an inspection
with the fire brigade upon commissioning, which also assumes that the fire brigade has been
involved in pre-installation or design stakeholder meetings.
For this, according to existing example cases [81], the technical documentation with all plans
and electrical diagrams, as well as the components of the PV system must be available. A
copy of the documentation must be located with the 230 V/400 V electrical panel (AC) or with
the inverters. This document must contain the following information that is important for the
local fire brigade:
1. Location of the AC switch-disconnector
2. Location of the 230 V/400 V electrical panel (AC) with FI circuit breaker
3. Location of the DC load break switches and inverters
4. Location of PV fireman's switch, if available
5. Location of the generator connection boxes, if available.
6. Information on whether the array boxes have DC load break switches or not; and if
these can be operated manually or by means of PV fireman's switches
7. A guarantee that the operable safety components are accessible
8. Location of the solar modules
9. Information on roof access/paths without stepping onto solar modules
10. Information on where the roof skin can be opened without damaging the solar modules
11. Warning and information signs on the PV system and its components. The operable
safety components can be quickly identifiable during operation
12. Telephone numbers of the operating personnel, safety managers and/or contact per-
sons
For systems with a minimum size of 30 kWp, a fire brigade deployment plan with the most
important mission-relevant information must be drawn up. Yet, it must be noted that firefighters'
practices and requirements can differ [68]. Analyses of the causes of fires in PV installations
and on the firefighters’ interventions have been made and recommendations are given in the
BRE recommendations [70].
D. Special Safety Aspects of BIPV
In addition to the above considerations, special safety, and reliability aspects for BIPV systems
can also include:
• Compared to free-standing modules, the rear sides of modules in BIPV and BAPV in-
stallations are often not easily accessible for visual control. Therefore, adapted inspec-
tion methods may be required
• Higher module operation temperatures are to be anticipated, also for the junction boxes
and bypass diodes, because of limited ventilation conditions, in cases of confined BIPV
modules
59
• Higher requirements of non-flammability of the underlying construction elements, par-

ticularly regarding thermal insulation or roof membrane materials
• In the built environment, shading of a part of the installations during several days of the
year by neighbouring buildings, trees, or building details such as balconies (for BIPV
façades) or technical infrastructure elements such as ventilation outlets and/or HVAC
systems may cause hot spot effects. As a mitigating measure today, DC optimizers are
often used for modules in BIPV systems where shading is inevitable, eliminating hot
spots (and with-it further hazards, such as melting or fire), while maximizing the PV
energy yield. Yet, particularly in the case of BIPV systems, with their higher operational
temperatures, there is no sufficiently long-time experience with respect to the durability
of the DC-DC-transformation components
3.5.6 Conclusion
One of the factors for PV’s success has been its reduced O&M requirements compared to
other types of electricity generation. However, low maintenance does not mean no mainte-
nance. Whereas the PV industry has made great strides in detecting faults and even permitting
some faults to be corrected remotely, the safe and reliable operation of PV plants still entails
scheduled and unscheduled maintenance work, for which personnel will have to be onsite.
While the bulk of safety issues are designed out (or in) during the design and construction
phases of PV plants, there are elements which need to be considered, most of which can be
planned or mitigated:
• The safety systems and procedures in place
• The training of staff with the appropriate qualifications for the tasks to be undertaken
• The presence of appropriate equipment to perform maintenance tasks: PPE, consum-
able as well as durable maintenance tools
• Site-specific risks to be considered, such as heights (PV on buildings), presence of
water (PV on water), or increased fire risks
• Weather and site conditions for onsite visits
While many of PV systems and plants are becoming more standardised, each PV power plant
has its individual characteristics, and therefore the safety briefing and procedures should ad-
dress those specific aspects together with general safety considerations.
The various methods and systems described in this chapter address risks and their mitigation
for the safe operation of PV power plants. The human and organisational aspects, such as
implementing the hierarchy of controls (Figure 6) and developing PV systems in keeping with
ISO 45001, IEC 63049, ISO 9001, or ANSI Z10 are key issues to ensure that the PV power
plant and the staff who maintain it can operate safely.
60
POWER PLANT MAINTENANCE
4.1 Preventive Maintenance Actions

Preventive (or proactive) Maintenance (PM) represents the foundation of an O&M plan ap-
plied to a PV power plant, comprising a wide range of "per-schedule" activities and services of
maintenance in PV plant components. Typically, PM actions refer to all routine screening (e.g.,
physical, and visual inspections), onsite or remote testing and servicing/intervention at (pre)de-
termined frequencies and time intervals. The latter can be determined on a case-by-case ba-
sis, considering the equipment type and original equipment manufacturers (OEMs) require-
ments, system size and complexity, environmental/site conditions, historical data and, of
course, certain PV O&M agreement terms and conditions (warranty, financing, insurance).
Ultimately, PM is aiming to maximize the long-term energy yield and operational lifetime of a
PV plant, by preventing the occurrence of costly or catastrophic failures and breakdowns at
both PV system level and at individual “key” components, in compliance with the OEMs oper-
ating manuals and recommendations. As such, PM actions also ensure that PV equipment and
component warranties are in place, while reducing the risk of failures or (performance and
physical) degradation. On this basis, PM approaches are well-established and standard prac-
tices in PV O&M, particularly for their perceived effect in lowering the risk of unplanned PV
system downtime.
Yet, a PM action plan – due to its inherent “per-schedule” nature – can have a considerable
financial cost within the overall PV O&M agenda, often inducing excessive or redundant activ-
ities, thus labour (and costs), especially if not optimally planned and/or implemented. There-
fore, PM must be well balanced against the overall PV plant and O&M budgeting. Such balance
is pursued considering mainly direct costs of scheduled PM interventions and services, as well
as yield and cash flow through the life of the system. Balancing and optimizing PM protocols
and “best practices” will also depend on PV system size, design, complexity, and environment.
In addition, for PV systems and O&M plans where support by predictive maintenance (based
on highly granular PV data) is technically and economically applicable, PM scheduling can be
greatly improved both at the manufacturing level (better lifetime predictions for the PV compo-
nents) and at the O&M level (inspection of PV components at more realistic time intervals).
Lastly, technical experience and relevant track records can also be leveraged to further opti-
mize PM action plans.
In practice, though, it must be noted that PM action should generally comply with prescribed
OEM and O&M manuals as well as respective legislative and regulations, e.g. (inter)national
standards for periodic inspections of certain electrical components in PV installations. Besides,
the O&M contract should include this scope of PM services and the frequency of each task. It
is then under the responsibility of the O&M contractor to draft and implement the PM task plan
and report all activities to the PV asset owner/manager. Reporting of PM activity is important
to record tracking and follow up the plan according to the conditions in the contractual PV O&M
agreement, as well as to justify any deviations, if required. Typical exemplary PM “manuals”,
annual O&M plans and best practices have been recently compiled by NREL and SolarPower
Europe task forces [11].
61
Independent of the PV plant’s size, the most common PM actions in PV O&M include:
• Periodic “sampling” of individual electrical (I-V) measurements at module level
• Periodic inspections of PV modules or strings: mainly IR and/or (auxiliary) EL imaging
• Cleaning of PV modules (including soiling and/or snow removal)
• Site maintenance including vegetation management, removal of loose objects
• Upkeep of BOS (inspection, inverter servicing, tracker maintenance)
• Upkeep of SCADA/monitoring systems (including weather stations, data acquisition
units, sensors, etc.)
• Other actions related to site management (water drainage, isolation from wildlife,
fence/road repairs, environmental compliance, and security)
During PV operational years, it is important to consider vegetation growing on adjacent PV
assets (e.g., growth of grass/plants in the short term, growth of trees in the long term), espe-
cially if these will cause shading on PV arrays, which can then result in hot spots, which are
shown in the small inserted IR images (Figure 7). Site and vegetation management, depend-
ing on the site/climatic conditions and the size of the PV plant, can be both challenging and
labour/cost intensive. In arid areas, mowing may be unnecessary with proper soil stabilization.
On the other hand, for areas where mowing is required, there is the added risk of projectiles
damaging modules. In the latter case, vegetation management “strategies” that have gained
increasing popularity in the last years include the collaboration with local farmers in using cer-
tain species of grazing animals.
Figure 7: Examples of shading caused by non-managed vegetation at a utility-scale PV

plant © CEA-INES.
Similarly, regular cleaning of modules from dust (soiling) or snow is an imperative PM action
in PV plants installed in certain geographical locations and climates. As well known, soiling
can result in considerable losses in the PV energy yield on both seasonal and annual scale.
Also, as in the case of unwanted shading from vegetation, cases of uneven, localized soiling
(dirt e.g. from bird droppings) can lead to hot-spot effects on PV modules and risks of follow-
up failures (glass/cell fractures, melting of the backsheet) or even fire hazards (Figure 8). Upon
cleaning of PV arrays, care should be taken, in compliance with the OEM’s recommendations,
to avoid damaging the different PV components. For instance, cleaning solutions shall be
based on plain demineralized water and mild detergents.
62
Figure 8: Aerial view of a rooftop commercial PV plant installed at a coastal site, pre-
senting severe issues of localized soiling due to dirt (bird droppings) © CEA-INES.
At PV inspection-level, PM practices concentrate on primarily two categories of onsite in-
spections: 1) manual electrical (I-V) tests at module or string level and 2) visual and IR/EL
imagery. Today, such inspections are indispensable not only in PM but in the overall PV O&M
context, since data from PV SCADA/monitoring systems present certain inherent limitations in
terms of both accuracy and spatial resolution, especially considering the rapidly increasing
sizes of utility-scale PV plants.
Manual electrical testing such as open-circuit voltage, operating current, or field I-V curve
tracing can be used to detect faults on the DC side of PV systems, at submodule, module, or
string level. Such faults, which often remain undetected or are misdiagnosed by monitoring
systems, can result in follow-up failures, physical degradation of PV modules, constantly un-
derperforming PV plants and, ultimately, in significant energy yield (thus financial) losses.
Since such electrical measurements are carried out in real field conditions, the accuracy of the
testing equipment is limited by the combined accuracy of the necessary irradiance, tempera-
ture, and electrical sensor inputs; and in the case of a standard I-V tracing set-up, it is limited
to around 5%. In practice, electrical testing can only reveal existing defects and failures at
string, module, or submodule level that at MPPT and field conditions result in measurable
power output losses. However, when correlated with the IR/EL image patterns, electrical sig-
natures can also yield valuable insights into underlying module-quality issues, thus often ena-
bling the timely prevention of more severe, follow-up failures, e.g., fire hazards.
4.1.1 Thermal Imagery PV Plant Inspection

Thermal imagery-based inspections refer to the collection and processing of IR images at PV
system, string, module or even submodule/cell level, to diagnose and classify certain thermal
patterns of PV failures (small inserted IR images in Figure 7 and Figure 9). As a widespread
best practice today, IR imagery of PV plants is performed in the form of aerial inspection. By
63
analysing abnormal thermal variations and thermal patterns across the inspected PV arrays,
nearly any critical defect or failure that is causing even a minor performance loss can be lo-
cated and diagnosed, in addition to the proactive detection of hot spots and potential fire risks,
as aforementioned. IR imagery scans can be performed in addition to or instead of manual
electrical testing, as an integral part of the annual PM plan in PV O&M, while they can also be
deployed in cases such as PV system commissioning, site-management related checks (e.g.,
vegetation, soiling mitigation), insurance claims and end-of-warranty inspections, as well as
for IR inspections of AC substations. (Figure 10).
Figure 9: Aerial IR imaging by aircraft (left) and drone (right), revealing different failure
modes at cell, submodule, module, and string level © Heliolytics Inc. (left image), © CEA-
INES (right image).
Figure 10: Aerial visual imagery for general inspection in general site management of a
PV plant © CEA-INES.
64
Today, aerial thermal imaging of PV plants can be performed using manned survey-type air-
craft or unmanned aerial vehicles (UAVs, drones). The diagnostic and quality of the assess-
ment depends largely on the aerial imagery sensors, the measurement conditions (and their
compliance to best practices and technical specifications, e.g., the IEC TS 62446-3:2017), as
well as the employed post-processing systems and algorithms that are used.
The resulting imagery should be processed by a validated processing routine to correctly iden-
tify thermal “signatures” of PV failures/defects and their exact physical location in the field.
Such PV failure data, in the form of thermal patterns, are typically of very high spatial granu-
larity from system up to cell level; and their labelling allows O&M engineers and field techni-
cians to quickly diagnose, classify and remediate PV faults and their root cause. When properly
applied, aerial IR/visual imagery is a powerful tool for diagnosing:
• PV module faults and solar cell defects:
o Electrical mismatches and hot spots
o Bypass diode and junction box failures
o Glass fracture and/or cell cracks
o Snail trails
o EVA delamination and discoloration
o Localized soiling (bird droppings, debris, vegetation)
o Broken interconnections
• PV string and system faults:
o Burnt fuses
o Inverter failures
o Cabling/connector failures
o Open- or short-circuited PV modules or strings
o MPPT faults
• Racking and BOS faults:
o Major racking shifts
o Systemic shading
o Major erosion
o Tracking systems’ mismatch
The use of proper post-processing and image data analytics is critical to accurately localize,
diagnose, classify all these different PV failure modes, or even quantify them in terms of asso-
ciated power loss. Further cross-validation against ground data and auxiliary measurements
(I-V tracing, EL imaging) is often recommended, to allow proper root cause analysis, limit false-
positives and optimize decision-making on corrective interventions.
At present days, there are quite a number of suppliers active in aerial IR/visual imaging ser-
vices. Many of these players provide not only the inspection of PV plant but also solutions for
65
PV diagnostics. This full turnkey aerial-IR inspection services typically include artificial intelli-
gence (AI)-based data analytics, fault diagnostics and reporting as well as consulting, i.e., rec-
ommendations for corrective maintenance actions to PV asset owners and O&M engineers. In
the general PV O&M context, such services typically include scans for both regular inspection
as part of preventative maintenance, and for reactive troubleshooting during corrective mainte-
nance, as well as for commissioning or asset transfer.
Furthermore, market leaders in aerial PV inspections can leverage large and consistently in-
creasing portfolios (already beyond the GW-scale) of inspected PV plants, to offer compari-
sons of different PV underperformance issues and anomalies against extensive fault libraries,
utilizing proprietary imagery analysis and IR/RGB mapping software. As such, the offered so-
lutions commonly identify a large range of PV plant anomalies, from extensive string, combiner
or array-level outages to module-, submodule- and cell- level failures and subtle defects, e.g.
PID, electrical mismatches and concomitant hot spots. In addition, fast return of investments
(ROIs) for aerial inspections is justified by most service providers, based on both a decrease
of preventative O&M costs by over 10% (compared to those employing manual and ground-
based inspections) and recovery of an average 1-2% PV power capacity losses.
In principle, the data acquisition equipment and approaches employed for aerial PV
inspections can vary among the different downstream service providers. For instance, in con-
trast to UAV/drone-based aerial IR imaging scans, some suppliers opt to employ IR imagers
mounted on aircrafts, offering flyovers at much higher altitude and speed; thus, allowing for
very high inspection rates (e.g., 100 MW per hour), multiple times faster compared to compet-
itors’ drone-based inspections. However, such approach can considerably limit the spatial
granularity, being rather impractical to fully comply with the minimum recommended diagnostic
resolution, for correct identification of certain failure modes at PV submodule and cell level.
Yet, all services face certain operational limitations. In particular, their current high reliance on
(semi-)manual image data processing, represents a major drawback as human error drives
down the diagnostic accuracy (and its consistence) of IR imaging for PV. In addition, such
solutions are typically "restricted" to a qualitative diagnosis of PV failures, being inefficient (until
today) in providing reliable, quantitative, and real-time assessments on the PV yield losses
associated to the detection of failures. It should also be noted that shifting from today’s “per
schedule” aerial IR inspections towards data-driven ones, will have a significant impact on the
technical bankability, financial feasibility, and consequent long-term competitiveness of these
commercial services.
4.1.2 Electroluminescence Imagery PV Plant Inspection

Electroluminescence (EL) imagery inspection is becoming quite popular to inspect defects or
flaws in PV cells and modules. EL imagery has many advantages as it shows the details of the
defects, so it is helpful to identify most of cell and module defects [1].
Electroluminescence screening in the beginning of a PV plant’s lifetime is a concept that en-
sures the quality of cell material during constructions and gives the security to all stakeholders
that, at the handover, the plant’s power generating surface is not only operational but of flaw-
less quality. Workers and constructor must operate more carefully knowing that the result is
screened. The investor/operator has a warranty that there are no construction errors such as
cracked and damaged modules. Insurance companies have an initial footprint of the plant to
which they can reference, and damage can be more easily tracked to single events such as
66
hail and wind loads, maltreatment during cleaning or maintenance work during operation. In
fact, legal problems are likely to arise due to a lack of initial reference when screening is per-
formed only after damage becomes apparent. Insurers are reluctant to accept the screening
results as proof that the damage was done during operation or by an insurable event. Thus,
knowing the initial quality ensures that legal problems will not add up to financial problems
caused by underperformance.
A. Electroluminescence Test Procedure

For electroluminescence imaging, the PV cells and the modules need to be fed with reverse
current to generate near infrared (NIR) light, which is captured using a special camera. More-
over, the environment needs to be dark, and thus EL inspection is best done at night because
the daylight contains a lot of NIR which is stronger than the NIR irradiated from the PV mod-
ules. If the inspection is carried out during daytime, a dark environment (such as a dark enclo-
sure or a darkroom) needs to be set up around the PV modules which are to be inspected.
Like IR inspection, unmanned aerial vehicles (UAV) or drone inspection is favoured to quickly
cover a large amount of inspection surface. A recording of the EL imagery would be the most
comfortable solution considering the number of frames per second and hardware limitations.
There are several difficulties which are still hindering a wide adoption of drone EL inspection.
First, as previously explained, PV plant inspection is best done at night, but flying at night for
EL field inspection increases the difficulty and risk of loss of UAV control. A technique for day-
light EL drone inspection is explained in [29]; in this case, the drone stability plays a very
important role (wind conditions should be optimal).
The second challenges in EL inspection, the PV module strings to be inspected must be ener-
gized, i.e., the module strings need to be connected to a DC power supply to feed in current
to the PV cells. It requires more work than the IR inspection. If the module strings are con-
nected to DC combiner boxes which are located at a distance to each other, the DC power
supply needs to be moved from box to box connecting the strings at night, an added complex-
ity. The requirements for an EL inspection make the UAV option less efficient.
The third difficulty is the limit of the camera capability. Most of the silicon sensor commercial
cameras are not sensitive in the NIR wavelength range, so the exposure time needed is usually
long (from a few seconds up to minutes). However, the long exposure time might result in
blurry images taken by UAV.
With much lower resolution sensors, some of the InGaAs cameras have very good sensitivity
for NIR getting the luminescence signal of the PV modules in some few milliseconds. The
capability of video recording is the biggest added value in these kinds of cameras, making
things easier to the O&M operator. Depending on the type of analysis, a compromise between
UAV panel distance and lenses should be found in order to obtain the desired image quality
for the proposed analysis. From micro-cracks to submodule or string functioning analysis, the
solution combinations can be several. The UAV-panel distances for IR and EL are not compa-
rable due to the resolution constraints of the latter.
There are still no turnkey solutions in the market to do EL from UAVs. Wireless communication
and camera control is now the biggest challenge. Fortunately, the camera technology and the
inspection devices are improving, and a complete inspection device shall become commer-
67
cially available in the near future. However, inspection service companies and research cen-
tres are developing their own solutions with increasing success. In the Report IEA-PVPS T13-
24 on mobile devices for PV inspections by Herrmann et al. [29], details on the detectable
failure types for PV modules and arrays are given as well as measurement uncertainties and
cost considerations for EL inspections.
B. Electroluminescence Test Sampling

For EL test sampling, the screening can be done in different ways and 100% screening is not
always necessary. Experience has shown that a 1% scan is sufficient in most cases to monitor
the main performance related issues. Inspectors and designers have identified a number of
critical points, such as:
• Racks near the transformer/inverter stations, combiner boxes, meteorology stations
• Points with high traffic during construction near the fence, and close to roads
Screening these points makes detection of common construction problems feasible. These
points in the plant are exposed to additional stresses during the construction and the mainte-
nance operation. From experience in the field, these points are prone to more damage than
other parts of the plant. Further screening of several modules in zones with high wind loads
(corners and edges of the plant) is important to be able to proof that damage was caused by
an extreme weather event and not during construction. The bottom rows of freestanding instal-
lations need to be checked before commissioning. In practice, damages due to leaning on the
lower module rows during installation, is often reported.
When a purely statistical screening is done, it is important to be able to track the screened
modules over time for later inspection. As the modules should not be obviously marked (con-
structors might take extra care with only these) the utilization of hidden transponders in the
frame is useful to mark the modules that are selected for statistical screening. This makes
especially sense if these modules were tested at earlier points in the supply chain. In this
fashion, the gap between incoming goods inspection and commissioning is bridged. If the first
test is performed in the field, recording of GPS position or relative position in the park of the
respective modules is sufficient. Both ways, the sample modules are easily findable and iden-
tifiable in the plant.
Ideally, the selected modules should be screened before and after installation and periodically
during O&M measures. By establishing a routine screening, prior to commissioning and after
certain O&M measures, the cause of the fault is more easily identified.
4.1.3 Preventive Maintenance Scope vs O&M Costs

Regarding cost aspects, obviously, costs associated with individual PM tasks (and the overall
PV O&M budgeting) cannot be determined in a simple “one-size-fits-all” approach. Multiple
and interrelated parameters, such as the PV plant’s size/use (residential or small-scale; com-
mercial/industrial; utility-scale), its characteristics in both design/layout and PV/BOS technol-
ogy terms, the site (location or land characteristics, climatic stress profiles, etc.), its contractual
arrangements, as well as the adopted O&M plan (rigorousness, approach, stipulated labour,
etc.) all contribute to a wide range of costs in relation to the aforementioned PM tasks.
68
In general, O&M service providers tend to overestimate PV O&M budgets to offset both their
own margins and uncertainties or technical assumptions. On the other hand, PV plant devel-
opers typically follow a more moderate or underestimating budgeting approach for PV O&M,
to allow higher PV plant valuations and more “motivating” revenue prospects. As aptly noted
by Enbar et al. [82], such contrasting O&M budgeting viewpoints can ultimately undermine a
PV plant’s lifecycle performance economics. In the following Table 5, we provide a non-ex-
haustive list of representative costs and remarks associated with the main PM tasks, as pub-
licly reported in the years 2021/2022, specific for utility-scale PV.
Table 5: Representative costs of main preventive maintenance tasks.
PM Task Costs Remarks
Base O&M scope 6 - 14 €/ kWp/ year Includes: full preventive maintenance

scope, regular module cleanings, security
(remote or on-site); excl. IR and EL
scans.
Varying highly with the site characteris-
tics, labor, and frequency of activity.
Cleaning/washing of 0.5 - 2.5 €/ kWp/ year Varying with the module technology,
PV modules [83] labor, cleaning solution and method,
climatic conditions (affecting the
frequency), etc.
IR scans 0.5 – 3.0 € / module Includes drone inspections, analysis,

and reporting
EL scans 3.0 – 10.0 € / module
4.2 Corrective Maintenance Actions

Corrective maintenance (CM) forms the second pillar of the PV plant maintenance during
operational years. In contrast to preventive maintenance discussed previously, corrective
maintenance actions are triggered only by the occurrence of events such as failures, break-
downs, malfunctions, anomalies, or damages. The trigger sources could be alarms in the mon-
itoring system, or from findings during regular preventive maintenance inspections.
CM’s primary goal is to restore the PV plant to its proper functioning state. Thus, the interven-
tion time (i.e., from detection, to response, to rectifying the defect) from the moment the events
occur is an important aspect in CM actions. The response time should be categorized for dif-
ferent types of fault events. In general, faults which will impact the safety operation of the PV
plant, or 100% plant outage should be treated as critical and responded within four to eight
hours. Faults resulting in major production loss should be treated as major and should be
addressed within 24-48 hours. Ideally, the intervention time should be chosen to minimize the
69
plant downtime. However, a more responsive CM programme could entail a more expensive
O&M service price.
In this light, there has been an emerging maintenance to predict a fault event ahead of time.
The progress in the field of data analytics and artificial intelligence/ machine learning has led
to the development of predictive maintenance of PV plants. Predictive maintenance utilizes
historical operational data of a PV plant obtained from the monitoring system and data on
environmental parameters (weather conditions such as irradiance, temperature, rainfall, etc.),
with the goal to learn the behaviour or performance pattern of the PV plant and using this to
anticipate and plan for maintenance interventions before an event or fault occurs. An example
is using predictive maintenance to anticipate and plan for the PV module cleaning cycle. Pre-
dictive maintenance is not a standard feature yet in the O&M scope of service but is gaining
traction, especially with the advance of powerful and scalable tools for big data analytics.
4.2.1 Spare Parts

One important aspect of a good corrective maintenance plan for PV power plants is ensuring
that spare parts are available and accessible when needed to avoid prolonged plant down-
time/outage due to equipment malfunctioning or damage. An O&M programme of a PV plant
thus should foresee a certain number of spare part stocks of the plant equipment, especially
the key ones such as PV modules, inverters, tracker components (if applicable) and transform-
ers. Best-practice recommendations regarding the operation of PV plants include well man-
agement of spare parts and consumables needed to maintain the proper functioning of the
plants. This commences with the EPC contractor preparing a minimum spare parts list and
procuring the initial stock of the spare parts. The spare parts (and the list) are then handed
over to the plant owner or the O&M Operator during takeover.
During the operational years of PV plants, typically the management (managing stock levels,
storage, and replenishment) of spare parts is typically included in the scope of O&M services.
However, purchasing replacements of expensive key equipment leads to higher costs, and
consequently, the O&M annual fee would be increased to cover the activities related to the
management of spare parts stock. It is not uncommon that the owner of PV plant opts to ex-
clude spare part management (sometimes only for certain types of key plant components) to
be excluded from the O&M Operator’s responsibilities. Under this context, the plant owner
could then address the foreseeable future maintenance of the excluded equipment either
through an arrangement for extended service or warranty of the equipment from the equipment
supplier, or through reserving the necessary budget. For the latter, a reserve account (MRA -
maintenance reserve account) is thus recommended to be set aside by the plant owner to
anticipate replacement costs of the excluded key equipment.
4.2.2 Future Inverter Replacements

The failure rate of inverters is driven by various factors; it is inverter specific (design, topology,
supplier manufacturing quality) as well as affected by external factors from the quality of instal-
lation and maintenance of the PV plant the inverter is installed in, to the environmental condi-
tions the inverter is operating in. A good source of inverter failure rate information will therefore
include all these factors. Past studies of inverter failure rates in PV plants across the globe in
different climates and over many years have indicated a mean time between failures of around
11-12 years for solar inverters.
70
At present, PV plants are expected to have a lifetime of more than 25 years. Considering the
above average lifetime of solar inverters, inverter replacements should thus be anticipated
roughly halfway through the PV plant life. The associated future inverter replacement costs
thus need to be foreseen as operational costs of PV plants. The future replacement expenses
are in fact commonly set aside as a maintenance reserve account. The amount of money to
be set aside should be estimated based on the future costs of purchasing inverter replace-
ments and calculated using reliable and relevant information regarding the failure rate of the
inverters over the PV plant operational years.
71
O&M RECOMMENDATIONS IN DIFFERENT REGIONS
Our planet has many different climates, each one affecting PV systems in particular ways. In
this chapter O&M recommendations are presented for different climate scenarios. The moder-
ate climate O&M guideline will cover the basic or most common aspects and conditions that
are relevant for all climates and regions. The recommendations will become more specific in
the next subchapters where extreme climates present diverse particularities and may require
greater attention of O&M operators.
Seven different types of climates are presented while addressing a wide spectrum of issues
with the goal to raise the awareness of the O&M complexity to the PV community.
As a basis of any O&M activity in any region, we can find the following international standards
in Table 6, which could offer support in topics such as test procedures, interpretations, and
safety.
Table 6: O&M related international standards and technical specifications.
Standard / Technical Title Highlights
Specification
IEC TS 63049 Terrestrial photovoltaic (PV) systems - Good practices for
Guidelines for effective quality assur- O&M management
ance in PV systems installation, oper- and development
ation, and maintenance
IEC 62446-1:2016 Photovoltaic (PV) systems - Require- Verifications, test
ments for testing, documentation, and procedures (I-V
maintenance - Part 1: Grid connected curve, IR inspection,
systems - Documentation, commis- etc.), results inter-
sioning tests and inspection pretation
IEC 62446-2:2020 Requirements for testing, documenta- Maintenance proto-
tion, and maintenance - Part 2: Grid cols, verifications
connected systems - tasks, safety proce-
Maintenance of PV systems dures
IEC TS 62446-3:2017 Photovoltaic (PV) systems - Equipment require-
Requirements for testing, documenta- ments, inspection
tion, and maintenance - Part 3: Photo- procedure, evalua-
voltaic modules and plants - Outdoor tion
infrared thermography
5.1 O&M Guidelines for Moderate Climates (Europe)

5.1.1 Description of Climatic Conditions
Moderate climates are very suitable for vegetation and wildlife development throughout the
year, making the seasonal changes smooth enough for the survival of many species. Although
this may seem contradictory in these times, the development of nature close to the PV modules
is not always welcomed. Vegetation and wildlife, among others, may become persona non
72
grata during certain periods of the year for the benefit of energy yields. There are many ways
to avoid an undesirable situation and develop solutions where life can be respected. Soiling
instead is potentially more harmful than vegetation and wildlife impacts, especially when it is
composed of corrosive chemicals coming from an industrial environment.
5.1.2 Field Experiences

Vegetation around the PV modules may create undesired shadows or situations where
branches wrap cables or cables trays. It may also occur that branches exert pressure on the
backside of the modules as seen in Figure 11.
Figure 11: Growing bush in front of a tracker (left) and underneath an array (right) [Eurac
Research].
Rodents pose a threat to electrical cabling integrity. The more accessible and thinner the ca-
bles are, the more damage can be done. Figure 12 shows the damage in string and sensor
cabling in different PV installations.
Figure 12: Cable bitten by a marten (left) (pveurope.eu © Leitl). Temperature sensor ca-
ble cut in half by a rodent (right) [Eurac Research].
The impact of soiling in a moderate climate will vary depending on environmental factors, dust
type, location, and installation factors [84]. Cleaning thin-film modules in sunlight can lead to a
73
permanent power loss, as the shaded cell parts switch to high resistance mode and therefore
the remaining low resistance cell parts overheat due to the string current. Therefore, the rec-
ommendations of the module manufacturers must be strictly followed to avoid loss of warranty.
A general introduction of this topic has been previously presented in another Task 13 report
[1] and will be further analysed in later chapters of this report. The following examples in Figure
13, collected from a recent publication [83], present a general visualization of this problem.
Figure 13: Examples of different types of soiling of PV arrays [83].

Concerning bird droppings, Figure 14’s left image shows a thermal image from a bird dropping
that may lead to hotspot conditions, possibly damaging the PV module.
Figure 14: Impact of bird droppings and locally increased temperature (left), Impact of
freshly cut green and unexpected hotspots (right) [Fraunhofer ISE].
74
What might be less known is the fact that O&M actions might also be a source of soiling if
conducted in an erroneous way. Figure 14 right image shows a thermal image of a rack after
fresh grass cuttings. It does make a difference if the lawnmower goes from left to right or vice
versa if it throws out debris on one side only.
This heterogeneous soiling can also be seen in I-V curves from single modules. The difference
to uniform soiling is the fact that heterogeneous soiling has a stronger influence on MPP rather
than on Isc, as can be seen in Figure 15.
Figure 15: IV and P-V curves of clean, uniformly, and heterogeneously soiled PV mod-
ules [ATONOMETRICS soiling measurement system manual].
A special anecdote concerns the effect of the fungus Baudoinia compniacensis which prefers
airborne alcohol that occurs close to whiskey distilleries and forms a black patina on PV mod-
ules, as can be seen in Figure 16 [85].
75
Figure 16: Effect of Baudoinia fungus in Scotland [85].

76
5.1.3 O&M Recommendations and Guidelines

In moderate climates, frequently, soiled PV modules bird droppings, agricultural emissions,
pollen, lichen, and traffic emissions, such as engine exhausts, and this occurs with a strong
seasonal variation. With frequent precipitation, most of the soiling will be washed away but
there are conditions, especially during a dryer period or heavy agricultural activity, where the
soiling rate - the comparison of “clean vs. dirty” - can vary by several percent.
Several methods to mitigate soiling - preventive and restorative - exist (see Report IEA-PVPS
T13-21 on soiling [31] for a more in-depth discussion), such as manual, semi or fully automatic
cleaning solutions. Fully automatic pre-installed solutions with very low water consumption or
even dry brushing are attractive for utility-scale systems in regions with a very high risk of
soiling, e.g., in the Middle East region. On the other hand, semi-automatic or manual cleaning
systems using demineralized water or some biodegradable chemicals to wash off organic sub-
stances might be more cost-effective in moderate climates. There is no one-size-fits-all solu-
tion for the mitigation of soiling effects. It is a local economic decision to clean based on labour
costs, water availability and costs, feed-in tariffs, and medium-term weather forecast. Further
details are given in Report IEA-PVPS T13-21 on soiling for generic “Best-time-to-clean” models
[31].
Bird dropping is the cause for soiling of PV modules in moderate climates. It is important to
develop and execute a cleaning plan based on the frequency of droppings which will be very
dependent on the kind of birds living in the area and those crossing it when migrating. This
planning should avoid droppings sticking to the module glass for a long time for the following
main reasons:
• Energy performance loss
• Hot spots, provoking material degradation and electrical risk increase
• Etching effect on the glass leaving a non-removable trace on the surface.
Many are the solutions to clean the PV panels [86], [87] and there are many solutions to deter
birds in order to keep them away from PV systems [88]. The budget for this O&M measure will
determine the type of solution. To conclude, it is important to be aware of the risk of different
chemical products that can be harmful to the glass, glass coating and frames of the PV mod-
ules. They can put both the module and the product warranty in danger. In case of any doubt,
it is recommended to be advised by professionals.
Proceeding with more issues, roof mounted systems and floating PV systems are preferred
places for animal nesting such as birds or squirrels in rooftops. Both places offer shelter and
minimal disturbance from humans for nesting and resting. In the case of squirrels or similar
rodents, electrical wires are in danger of being chewed and damaged. Installing a mesh to
close gaps can be a successful solution for rooftops, while for floating PV bird deterrent sys-
tems can be employed [89].
In large PV installations where the reflective surface may imitate sky or water, like the illusory
effect in windows on a building, birds and waterfowls may impact or land on the panels re-
sulting in broken wings or death due to high temperatures exposure [90].
From the O&M point of view, such situations are difficult to prevent. Good practices are to
identify bird populations passing near the PV plant and to call the closest wildlife rehabilitation
centre if an animal is injured. The areas where these incidents are most likely to occur seem
to be deserts [91] and lakes for floating PV systems [92]. It goes without saying that after any
77
such event, a visual inspection should be carried out in the impact area to detect any mechan-
ical damage.
For rodents, a first recommendation is to fix cable loops and hanging cables as far as possible
from the ground or from structures that are easy to climb. If more protection is needed, anti-
rodent solutions are available on the market, such as cable shields.
Most animals living next to ground-mounted PV modules are farm animals such as sheep and
small livestock (Figure 17). They can control vegetation in PV plants and usually do not climb
on or damage the PV modules. The first row of modules might be exposed to them which could
be pressed upon by their bodies when reaching for grass underneath the PV panels.
Figure 17: Farm animals in British solar farms [93].

Larger farm animals such as horses and cattle are considered unsuitable since they have the
weight and strength to dislodge standard mounting systems, while pigs or goats may cause
damage to cables [93].
It is worth mentioning that both PV modules and structures are an opportunity for animals to
rub their bodies against. A follow-up of the mechanical status of the structure and first row of
panels is recommended.
Before finishing with the wildlife chapter, we should not forget insect pests. The cool shade
beneath a PV panel and the inner empty spaces of metal structures are ideal places for wasp
or bee nests. In the case of the metal structures, blocking the holes may prevent any future
nest to be built inside. In the case of the PV modules backside, nests may find structural sup-
port in frames and cables, making maintenance more difficult. In case of trouble, the safest
thing is to hire professional service for pest control.
Cutting or herbicide spraying is the common way to control the vegetation. It should be applied
in a proper way without spraying the PV modules. If it occurs, the modules can be washed off
with water. After some days, once the vegetation is dead (Figure 18), it should be pulled up by
the roots and the waste should be treated accordingly.
78
Figure 18: Vegetation before (left) and after spraying the herbicide (right) in a white
stone ground [EURAC Research].
5.1.4 OHS Recommendations and Guidelines

Concerning occupational health and safety, section 5.1 does not address special issues re-
lated to extraordinary climate conditions. International standards as mentioned in a previous
chapter and national codes are the references to follow. Public guidelines developed by the
industry, research or national entities are as well a good source of summarized and detailed
information as listed in Table 7.
Table 7: Available guidelines in different parts of the world.
Country / Continent Title Author/Editor
Europe Operation & Maintenance Best SolarPower Europe
Practice Guidelines / Version 4.0 [11]
India Best Practices in Operation and Gujarat Energy Re-
Maintenance of Rooftop Solar PV search & Manage-
Systems in India ment Institute
Japan Report on Guidelines for Periodic Japan Electrical
Inspection and Failure Examination Safety & Environ-
of PV Power Systems ment Technology
Laboratories
United States Best Practices for Operation and National Renewable
Maintenance of Photovoltaic and Energy Laboratory
Energy Storage Systems; 3rd Edition and others
79
5.2 O&M Guidelines for Hot and Dry Climates

Even though multiple PV-specific climate classification schemes have been proposed [94],
[95], the best known is the Köppen-Geiger system [96]. In the context of this section, a hot and
dry climate is the desert climate, classified as either BWh or BWk. These criteria are shown in
Table 8. In a desert climate there is very little precipitation (by approximation: a total precipita-
tion expressed in mm of less than 10 times the mean absolute annual temperature, e.g., less
than 200 mm of precipitation per year for 20°C), hot summer daytime temperatures that can
exceed 40°C (104°F), an excess of evaporation over precipitation, and low humidity throughout
the year. The typical desert landscape in such locations has low vegetation and receives very
little moisture most of the year, though heavy precipitation for short periods of time in the year
can occur. This heavy precipitation may cause land erosion that can affect PV plant founda-
tions.
Table 8: Köppen-Geiger requirements for hot and dry climates, from [96].
Desert Arid climate: hot steppe/ Desert Arid climate: cold steppe/
desert desert
Köppen- BWh BWk

Geiger
classification
Mean annual Tann ≥ +18°C Tann < +18°C
temperature
Tann
Average an- Pann ≤ 5 ∗ Pth
nual precipi- 2
tation Pann 2 ∗ |Tann | if at least
of annual precipitation in winter
3
Pth = 2
2 ∗ |Tann | + 28 if at least of annual precipitation in summer
3
{2 ∗ |Tann | + 14 otherwise
The challenging thermal range of hot and dry climates typically means that PV plants are lo-
cated in areas with low population density and are at much more remote locations than similar
systems in milder climates. The combination of physical remoteness and challenging climatic
conditions results in a strong preference for remote monitoring of plants, with few plants having
O&M personnel on-site or nearby. This is reflected in O&M KPIs, where availability values
lower than 99% are occasionally seen.
5.2.2 Soiling Impact

There is a great amount of possible variation for the type of ecosystem defined by these two
terms. A region defined as BWh can be comprised of loose sand dunes of coarse silicate sand
with very little vegetation or a rocky terrain with scrub vegetation. The climatic variation may
be the same, however the type of soiling experienced will be quite different.
80
Industrial development in these regions exists, though arguably less so than in the BWk re-
gions. Industrial regions add further complexity to the soiling issue, due to the addition of soiling
agents to the soil build-up on a solar module.
BWk regions may experience higher levels of moisture compared to BWh regions, particularly
dew production at night, due to the temperature drop reducing the atmosphere’s water reten-
tion ability. The level of moisture in the atmosphere will influence the choice of automatic mod-
ule cleaning systems. Dry cleaning systems will work well where moisture in the atmosphere
is very low, more typical for the BWh climate. Where dew can be expected, a dry-cleaning
system will not be effective, exacerbating module soiling. An extreme example would be the
choice of a dry-cleaning solution where moisture in the atmosphere is a possibility in the prox-
imity of a cement factory.
It is not inconceivable for a small geographical area with varying altitudes to include areas
requiring different automatic module cleaning characteristics. The O&M design should carefully
examine the available meteorological data available to ascertain the level of moisture during
defined cleaning times.
Figure 19 gives an example of automated cleaning of soiled PV modules.
Figure 19: Automated cleaning of soiled PV modules in the Negev Desert, Israel.
81
Experience in the Negev desert in Israel reflects manual washing of modules five times a year
with additional washing sessions added as required during the transitional seasons when heat
waves accompanied by dust clouds are broken by low volume rainfall. By contrast, PV systems
in Central Australia which are close to population centres have typically low soiling rates and
consequently have one to two manual cleaning sessions with water, with as-needed additional
cleaning mandated by weather events, such as occasional dust storms.
5.2.3 Operational Health and Safety Issues

Aside from the usual OHS issues associated with a PV plant, roof top or ground mounted,
there do exist several additional risks when operating in the desert. The typically highest risk
in hot and dry climates as calculated through a risk matrix (likelihood or frequency multiplied
by the consequence) are dehydration and sun stroke to O&M personnel performing the mainte-
nance when onsite. An often overlooked yet important risk is travel to the site(s) in relation to
the weather: to avoid the hottest hours of the day, early morning or night-time travel to remote
sites may necessitate driving in the dark in areas where wildlife can be encountered. Ensuring
that staff have the appropriate PPE as well as specialty communication tools (satellite phones
or emergency locator beacons) is important.
As the cost of PV plants drop over time due to learning effects and commercial pressures,
there is a tendency to maximize profits through economy of scale, including packing more
installed DC capacity on a smaller footprint, thanks to increasing module efficiencies. In this
context, plant designers and investors must be reminded that the final product must be main-
tainable, and O&M providers must evaluate the risk profile of plants ahead of taking on the
O&M role. In a PV plant built in the desert or in remote locations, efficient operations by the
maintenance personnel are of high importance. Foregoing access paths for ease of access to
critical maintenance points in the field such as string boxes may cause maintenance staff to
walk longer periods of time in the sun, reducing their efficiency and increasing their OHS risks.
Similarly, removing control rooms or reducing the size of shaded LV stations to reduce CAPEX
will increase OPEX costs due to lack of efficiency as the human element involved is weakened
by the adverse desert climate.
OHS concerns in these areas include the danger to personnel of endemic reptiles and other
threats. Electrical distribution panels are best opened with care in regions where such threats
are realistic. Anecdotal evidence points to snakes, birds, or termites nesting in or near these
distribution panels, with the risk of either short-circuits, arcing, or bites from snakes, scorpions
or lizards occurring upon opening cabinets.
Vegetation control addresses several risk factors, by reducing usable habitat for reptiles, avoid-
ing energy losses by keeping vegetation growth below the lowest module level, and reducing
combustible material that may lead to fires. Examples of vegetation management and potential
consequences are shown in Figure 20 and Figure 21.
82
Figure 20: Vegetation management of PV plant in Central Australia.
Figure 21: PV module damaged during vegetation management due to pebbles. Array
in Central Australia.
83
Hot and dry climates present a challenge for PV equipment: The high temperature, solar irra-
diation (often with high UV content) and dry atmosphere accelerate decomposition of many
PV elements, which include rubber seals, fiberglass cabinets, cable insulation, and cable ties.
Significant cable insulation degradation has been recorded after only a few years for cables
subjected to direct sunlight, despite the cables being certified as having been tested for such
use.
It is recommended that all cables be shaded from direct sunlight, fiberglass and polymer cab-
inets installed in shaded areas and that rubber seals (e.g., around cabinet doors) be slated for
replacement in realistic time spans. Similarly, cable ties should be shaded from direct sunlight,
be made from metal instead of plastic polymers, and be installed to accommodate the thermal
expansion seen over the day and the year.
An additional danger exists in the increased arc flash risk at low humidity due to static build-
up. The use of protective equipment when approaching and operating electrical switch gears
of any kind is of great importance.
The high temperature and high voltages introduced by the 1500 VDC standard equipment
tends to increase the probability for module degradation due to PID or other degrading mech-
anisms. Though PID is typically considered a high risk in areas with both high heat and humid-
ity together, many cases of PID have been registered in desert climates with little humidity.
When the sky is clear, radiation cools the modules, causing the module surface temperature
to be several °C lower than the air temperature, which leads to wetting of the rear (and front)
sheet surface, especially if the temperature falls below the dew point. At sunrise, low module
temperatures lead to high open-circuit and string operating voltages and, together with the
damp module surfaces, cause high leakage currents before the modules dry out in the sun.
Whether modules are susceptible to PID caused by (high) leakage currents can be checked
according to the parts of TS 62804 (-1 and -1-1 for crystalline silicon, -2 for thin-film modules),
see also PID chapters 6.2.5, 7.5 in [31].
Table 9 compares two desert regions, both defined as BWh, yet with different operational
parameters.
Table 9: O&M considerations in hot and dry climates.

Hot and Dry: No Vegetation Hot and Dry: With Vegetation
Example region/ Israel Central Australia

country
Köppen-Geiger BWh (Hot desert climates) BWh (Hot desert climates)
classification
Wildlife risks Humans: Poisonous snakes Humans: Poisonous snakes, spi-
ders, scorpions.
Infrastructure: various species of
wildlife attack cabling, nesting Infrastructure: snakes, termites
birds attempt to enter equipment, and birds may attempt to enter
84
large birds drop objects on mod- electrical equipment, causing

ules short-circuits.
Hot and Dry: No Vegetation Hot and Dry: With Vegetation
External fire risks Low – there is seldom little vege- Seasonal; requires fire breaks to
tation to feed a fire. be maintained (dedicated space
around PV farms to be kept devoid
of vegetation), need to monitor &
manage rapid vegetation growth in
wetter months.
OHS requirements Procedures to deal with extreme Procedures to deal with extreme
heat: avoiding full sun activities, heat: avoiding full sun activities,
appropriate PPE (protection appropriate PPE (protection
against sunburn), increased hy- against sunburn), increased hy-
dration needs. dration needs.
Extreme weather: Flash flooding,
intense rain events require moni-
toring of weather. Similarly, bush-
fire risks need to be gauged before
O&M activities are undertaken.
Module cleaning Transition seasons bring muddy Soiling typically of low impact on
requirements rainfall. Occasionally occurring in energy yield. Difficulty or cost of
summer as well. sourcing water for cleaning.
Electrical risks High temperatures reduce cur- Static build-up due to low humidity
rent carrying capacity and can increase arc flash risks. Ther-
switchboard design calling for mal expansion of components
forced ventilation requires extra such as cables can cause the
care during summer months to cables to loosen and arc at point of
avoid breakdown and dangerous connection.
overheating.
Special causes of Thermal cycling of modules (up to 40-50°C module temperature
module damage swings in one day), violent hailstorms, large bird droppings, rocks or
animals on modules.
Practical consider- Careful planning of activities and Relative remoteness of sites (dis-
ations properly maintained signage and tance to local and major popula-
maps to enable O&M staff to effi- tion centres) needs to be consid-
ciently undertake the tasks ered for O&M activity planning &
quickly during summer months. risk mitigation.
High irradiance and temperature High irradiance and temperature
can stress components (fuses, can stress components (fuses, ca-
cables, inverters) and shorten bles, inverters) and shorten their
their lifetimes. lifetimes.
85
5.2.4 Recommended Practices

World-wide standard O&M best practices as discussed in section 5.1, apply to hot and dry
climates, with a few aspects that need to be reiterated, such as evaluating wildlife risks, ap-
propriate planning for visits to typically remote sites (hydration, anti-venom procedures, PPE,
travel to and from sites).
Other aspects that can be re-evaluated include the installation of an automatic module cleaning
system due to the remote nature of the typical large desert site, the objective difficulty of human
activity in such climates and the unknown frequency of climatic events that require immediate
module washing, such as when a transitional season heatwave is broken with a short period
of low volume rain, creating a layer of dried soiling.
The weather extremes that can be encountered require equipment to be robust and resilient,
which may counter-intuitively point to manual cleaning over automated solutions, or a hybrid
combination. Given the remoteness and cost pressures encountered in both CAPEX and
OPEX, some plant owners will accept higher O&M inventory costs to ensure higher uptime
(e.g., storing high-cost items such as inverters or transformers at or near the PV farm), whereas
other plants operate as lean as possible, and thus run the risk of prolonged periods of under-
performance if critical parts cannot be sourced in time.
Depending on the local weather patterns, a contract for one to five washing sessions per year
timed around the transition seasons, at the discretion of the owner’s engineer can be assumed
to be an average for plants in this type of climate. Aside from the real time monitoring and
alarm protocol typical for daily operations, due to the remote nature typical to a desert plant,
visual inspections of the plant are typically undertaken four times a year, although fewer in-
spections are accepted for (very) remote systems. During these inspections, all electrical pan-
els are opened and visually examined. Land erosion is evaluated and compared to previous
visits. Metal elements are examined for corrosion and the security system is examined.
During annual maintenance visits, special attention must be paid to the risk of termites, birds
or snakes attempting to enter or nest in inverters and electrical switchboards, with cables or
tubing being often inspected for entry points. Similarly, a verification that modules and cables
remain fixed is required, as the thermal cycling throughout the year and the high irradiance
and UV content can loosen modules, cables, or cable ties. Ideally, once a year, I-V curves are
measured on a rotating sampling basis. All electrical panels and PV modules are photographed
with an IR camera and analysed for defects.
5.3 O&M Guidelines for Desert Climates in High Altitudes

Desert regions have received much attention in recent years due to their solar potential for the
implementation of photovoltaic power plants. They can take advantage of the outstanding con-
ditions: clear skies, large number of sun hours per day throughout the year, higher (UV) radi-
ation levels than most other places in the world, etc. Typically, deserts also face water scarcity
and are barren, treeless, and sandy. This does not necessarily imply that they are regions that
have high temperatures. Given the different characteristics that determine desert regions, they
are classified according to the Köppen classification standard [2] in accordance with the local
86
meteorological conditions and with the altitude compared to the sea level as: deserts with
cloudy days, deserts, high altitude deserts, and high steppe deserts. This classification pro-
vides insights for the deployment of large-scale solar projects in each region.
Figure 22 presents a cross-section view of the Atacama Desert and its classification according
to the Köppen classification standard. As it can be noticed, deserts offer a wide variety of
possibilities for the implementation of photovoltaic power plants, and thus the effects of the
weather conditions at these regions are being studied to generate operation and maintenance
recommendations so that their solar potential can be effectively exploited. As a result of these
studies, for instance, the Atacama Desert appears as one of the most interesting landscapes
for installing PV worldwide.
Figure 22: Cross-section view of the Atacama Desert according to the Köppen classifi-
cation standard. BWn stands for deserts with cloudy days, BWk for deserts, BWk’ for
high altitude deserts, and ETh for high steppe deserts [97].
The Atacama Desert is exposed to a very high irradiation, many sun hours per year, one of the
clearest skies in the world, the highest annual expected energy yield, and relatively low tem-
peratures. The particular features that offer the Atacama Desert for the deployment of photo-
voltaic power plants are 1) irradiation levels of 2 500 kWh/m2 (GHI), 3 500 kWh/m2 (DNI), on
average on a yearly basis; 2) 4 000 h/year, average number of sunlight hours per year; 3) 65%
87
UV-B and about 25% UV-A irradiation (over the European average); 4) temperature levels
below 30ºC, on average, in the summer season; and 5) 2 mm/year average rainfall in some
areas of the desert. Additionally, there are systematic east-west winds that serve as a natural
cooling system for photovoltaic modules, and a growing energy demand (due to the mining
activity) that could take advantage of the potential of the desert to (partially) fulfil that growing
demand with an environmentally friendly energy source.
In the following sections the main issues in the operation of PV power plants located at high
radiation and high-altitude desert conditions will be addressed, and it will be shown from the
experience gathered in the Atacama Desert how to deal with these conditions so that the per-
formance and life span of the plant are not highly affected.
5.3.2 Field Experiences of Reliability in Desert Climate

The Atacama Desert offers unique conditions for high production of solar energy, but at the
same time important challenges to guarantee the reliability and durability of PV systems. Own-
ers of photovoltaic plants must carefully select the technology they are planning to install (e.g.,
modules, cables, mounting structures, and inverters), since international standards for accel-
erated ageing applied in the photovoltaic industry largely underestimate the ultraviolet (UV)
conditions of the Atacama Desert. For instance, the standard IEC 61215 “Terrestrial photovol-
taic (PV) modules - Design qualification and type approval - Part 1-1: Special requirements for
testing of crystalline silicon photovoltaic (PV) modules” [98], used to pre-condition the modules
for ageing tests, does not represent the UV conditions in the Atacama Desert.
Current technologies for the implementation of photovoltaic power plants are sensitive to ex-
tremely high irradiation levels and specific solar spectrum (in the UV-B range). Furthermore,
some deserts have 1) corrosive environments due to the presence of salts and water conden-
sation during night hours; and 2) sticking fine powder that adheres to the components of the
power plant and causes operational problems. Therefore, the performance and life span of the
plant may be severely affected by the desert conditions. One way to diminish the effect of the
environmental conditions of desert regions on the performance of the plant is to make an over-
view of possible incidents in the plant and then to elaborate an operational and maintenance
plan focused mainly on those incidents that have a high probability of occurrence and a high
impact on the plant performance. For instance, in Chile, the Comite Solar developed a study
focused on the failures in photovoltaic power plants [99]. From this study, 27% of the failures
declared by the owners were in the PV modules and 49% of the failures were in the inverters.
The remaining 24% of the failures were in the medium and low voltage installations, in the
tracking system, in the communications and in the SCADA. Figure 23 presents the results
reported by Comite Solar.
88
1%
5% 2%
7%
9%
49%
27%
Inverter PV modules MV installation EV installation

Tracker Communications SCADA
Figure 23: Typical failures in PV power plants in Chile reported by Comite Solar [99].
5.3.3 Performance and Safety Issues in Desert Climate

Although the typical failures have been identified for several desert regions around the world,
there is not a compendium of recommendations focused on those desert regions with high
radiation, low temperature, and high altitude, such as the Atacama Desert. The present section
addresses the degradation of the main components of PV power plants located in high radia-
tion and high-altitude desert regions. The components analysed are the following: modules,
cables and cable ties, junction boxes, and mounting structures. To illustrate the degradation
of these components, the experience gathered from the operation of power plants located in
the Atacama Desert were considered, since the solar spectrum of the Atacama Desert does
not completely correspond to the ASTM G173 spectrum [100], which is often used to assess
the effect of the radiation on modules and other components of the power plants. Figure 24
compares both the ASTM G173 and the Atacama Desert spectrum [101].
The main differences between both spectrums lie in the range of the UVA and UVB, being
more notorious in the UVA range. The key factors responsible for these spectral differences
are 1) the high mean elevation above sea level; 2) the many days with clear skies due to the
absence of cloud cover and, hence, a low diffuse radiation fraction; 3) small aerosol optical
depth; 4) small total ozone column (TOC); and 5) small water vapour column. This finding
indicates that the technologies to be installed in high radiation and high-altitude desert regions
should be designed/selected to support considerable energy in the UVA range.
89
Figure 24: Comparison with the ASTM G173-03 reference spectra (purple) [100] and the
Atacama Desert solar spectrum (blue) [101].
In literature, several factors have been reported that may be responsible for the degradation
of PV modules [101]. In high irradiation and high-altitude desert regions, the main mechanisms
are 1) the high UV radiation present in the solar spectrum, and 2) the temperature oscillations
that could cover in a day from -10ºC to 30ºC. The former mechanism produces changes in the
colour and/or brightness of the cells, indicating that the entire cell or the encapsulant, respec-
tively, are being affected by the UV radiation, and consequently started an accelerated degra-
dation process. The later mechanism could produce the condensation of the water present in
bubbles inside the modules, which might drive towards 1) corrosion inside the cell; 2) corrosion
in the frame of the module and in the supporting structures; and 3) the accelerated degradation
of the encapsulant because of the interaction with the condensed water.
Both degradation mechanisms: UV radiation and temperature oscillation may lead to hotspots
and isolation issues of the modules (often in the form of arc flashes). The severity of the
hotspots and their number, and of the isolation issues as well, are directly correlated with the
degree of degradation of the module (whatever the degradation mechanism) and are inversely
correlated with the measures taken to prevent the progress of the degradation. Figure 25 pre-
sents some cases of degraded modules of PV power plants installed in the Atacama Desert.
For explanations of which types of degradation can be seen in Figure 25, see technical reports
[1], [28], and the PV module failure sheets in [102].
90
Figure 25: Degradation mechanisms evidenced in PV power plants installed in the Ata-
cama Desert.
In addition to module degradation, the UV content of the solar spectrum and the temperature
oscillations affect the cables and the cable ties. The UV content of the solar spectrum dimin-
ishes the flexibility of cables and cable ties increasing the possibility of cracking and cutting,
whereas the temperature oscillations pose additional mechanical efforts to them. These fac-
tors, reduction in the flexibility and additional mechanical efforts, may lead to energy losses,
short circuits, and arc flashes. Moreover, the presence of animals (mainly rodents) must be
prevented. Wild animals could cause damages to cables and cable ties since they tend to bite
these components, reducing their lifespan. Furthermore, wild animals can cause short circuits
in cables by damaging the isolation.
Figure 26 shows examples of a cable affected by the UV radiation and bitten by a rodent. In
the former (picture a), the isolation of the cable is completely cracked even though it was cer-
tified for operation under UV radiation. In the picture b, a rodent damaged the isolation of the
cable. In both cases, a replacement is necessary to prevent any damage to the facilities.
91
Figure 26: Examples of damages in cables in power plants located in the Atacama
Desert.
Moreover, junction boxes are also highly sensitive to the environmental conditions of high ra-
diation and high-altitude deserts. They are made and fixed to the modules with polymeric ma-
terials that are sensitive to heat and temperature oscillations (in general junction boxes are not
exposed to the radiation, just the radiation reflected from the ground). Heat and temperature
oscillations make the boxes themselves deteriorate and the polymer used for bonding them to
the modules can become damaged, losing their fixing capacity.
The resultant failures are module detaching, case opening, and connection losses. Module
detaching appears when the polymer used to fix the boxes to the modules expands and shrinks
because of the heat and temperature oscillations. Increases in heat and temperature followed
by a decrease at night can make the polymer become crystallized and lose its properties,
resulting in the detachment of the box. The issue of case opening follows a similar mechanism,
where the box is expanded and then shrunk from the heat and temperature oscillations. Then,
the screws become loose, and the box gradually starts to open.
Connection losses arise due to the factors that affect the flexibility/mechanical properties of
the cable, or because of an inappropriate installation. Degradation problems in junction boxes
may lead towards energy losses, short circuits, and internal arc flashing that endanger the
module and the operation of the facilities. Figure 27 presents these two cases: a detaching of
the junction box and a junction box opening.
92
Figure 27: Failures in a junction box in operation during about a year due to degradation
issues.
Other components that are highly degraded by the environmental conditions of high radiance
and high-altitude desert regions are the mounting structures that support the modules. These
structures degrade through temperature oscillations and the presence of salts. Temperature
oscillations might cause water condensation, and the interaction between the condensed water
and the structure’s exposed parts could make them corrode. Indeed, a wrong selection of the
materials for the structures and/or a lack of adequate surface finish may cause the structures
to be damaged and/or corroded. This is because, for example, differences in the electrochem-
ical potentials (corresponding to the electrochemical series) of the connected materials as well
as moisture can cause electrolytic corrosion.
The presence of salts makes for a corrosive environment: the higher the concentration of salts,
the higher the environmental corrosion level. Thus, the material selection for the structures as
well as appropriate surface finish is mandatory. In the case of the Atacama Desert, for instance,
both conditions are present: water condensation during night and dawn hours and salt pres-
ence in the environment since almost all the desert is a salt surface. Figure 28 shows typical
corroded structures in photovoltaic power plants.
As can be noticed, the corrosion of the structure affects not only the surface of the structure
but also its inner parts. This diminishes the mechanical properties of the structure as a whole
and becomes a source of risks for the operators of the plant. Furthermore, a corroded structure
could affect other nearby structures, hence reducing the life span of the plant itself.
93
Figure 28: Rusted mounting structure in a PV power plant located in the Atacama
Desert. This mounting structure has been in operation for about one year.
5.3.4 Recommendations of O&M and OHS in Desert Climate

The present section puts forward several recommendations/guidelines for both:
• operation and maintenance of photovoltaic power plants in high radiation and high-
altitude desert conditions
• and operation, health, and safety in these power plants
As in section 5.3.3, the operation and maintenance recommendations are focused on the fol-
lowing elements: modules, cables and cable ties, junction boxes, and mounting structures.
Next, recommendations to prevent the risks associated with the exposure of the operators to
high radiation and low oxygen environments are given.
A. Operation and maintenance recommendations/guidelines

Since the intensity of UV radiation and temperature oscillations are factors that cannot be con-
trolled, it is recommended to perform periodic visual inspections of modules to detect if any of
the phenomena illustrated in Figure 25 are taking place. These inspections should be accom-
panied with thermographic analyses and measurements of the quality of the modules’ isolation
94
to check for the formation of hotspots and the probability of losing electrical isolation. To de-
termine when complementary analyses must be conducted, it is suggested to compare the
energy production of the plant with the results of the visual inspections.
If a continuous decrease in the energy production is noticed, even after the plant has been
cleaned, the maintenance plan must be executed to reduce the energy losses and avoid pos-
sible permanent damages to the facilities. It is important to remark that any change in the
colour of the cells must be reported since such change indicates that something is happening
with the encapsulant and/or the cells themselves. These changes must be carefully followed
and the plant’s zones with presence of these changes must be identified. This way, it is possi-
ble to focus the maintenance efforts on these zones that are most affected by problems in the
modules, reducing the time taken to evaluate the plant, to determine the failures, and to isolate
the failed zones.
To prevent that damage will happen from issues in the cables and cable ties, it is recom-
mended to perform a visual inspection of the facilities with a certain regularity. These visual
inspections can be done at the same time as the other inspections, in order to verify the deg-
radation level of the modules. Visual inspections allow to verify that all the cable ties are work-
ing as expected. The focus is on the cable ties, as they could fail when exposed to UV radiation
and wild animals. However, if the cable ties are intact, it is expected that the cables have not
suffered any damage. Only if they were installed inappropriately, while being well tied, it may
be expected that the cables have flexibility/mechanical problems. Furthermore, it is recom-
mended that all cable ties are replaced periodically. Notably when they are made of a polymer
that gradually, with the heat, temperature oscillations, and UV radiation, loses its properties.
Indeed, the heat at midday combined with the weight of the cables make the ties become
gradually warped. Thus, despite the efforts made to make them work properly, within a short
period the cable ties may completely lose their capabilities to keep cables fixed and will fail,
leaving the cables exposed to the hazards. Note that if there are cables exposed to radiation.
They must be examined to determine if they must be replaced or tied adequately to avoid the
exposure. If this is not carried out, any damage in the cable is expected to progress leading to
undesired consequences for the operation and performance of the plant.
Like the recommendation for modules, cables, and cable ties, the main recommendation to
avoid junction box failures due to degradation is to perform periodical visual inspections to
assess the operating conditions of the junction boxes. All these inspections could be done
simultaneously following a checklist that allows the operators to determine if the equipment is
operating in adequate conditions or not. Furthermore, systematizing these checklists might
help in determining the plant’s zones with a larger number of degradation issues, and those
zones with degradation problems (e.g., junction box failures mainly). In the specific case of
detecting junction box problems, a report must be made so that corrective measurements can
be immediately taken. Any junction box failure added with water condensation, for instance,
could lead to short circuits causing fires impacting the facilities’ operation. Hence, as in the
case of the cable ties, it is highly important that the junction boxes prone to fail are detected in
advance. Decisions on how to solve the issues detected and/or about their replacement should
then be taken.
Finally, due to the potential risk associated with the loss of mechanical properties of the mount-
ing structures, their “degree of health” must be evaluated constantly. The degree of health is
assessed as the evolution of the corrosion process, starting from the day that it was first de-
tected. This evolution not only comprises the corrosion process of the structure evaluated but
95
also if this process is impacting other nearby structures or elements (e.g., cap screws and
module frames). Moreover, the structure’s mechanical properties must firstly be visually eval-
uated. This can be done by detecting if an abnormal buckling is taking place or not. The ab-
normal buckling of a structure indicates that it is becoming warped. If any of the aforementioned
factors are present: evolution of the rusting process in the structure, evolution of the rusting
towards closer structures, or abnormal structure warp, the recommendation is to schedule a
replacement of the affected structure so that the associated risks are reduced.
B. Operation, health, and safety recommendations/guidelines

With respect to health issues in the operation of photovoltaic power plants located at high
radiation and high-altitude desert regions, there are two main factors that must be considered:
the first one is the exposure to high radiation of the operators, and the second one is the
exposure of the operators to environments with low atmospheric pressure.
Both factors can significantly affect the capabilities (and even threaten the life) of the operators
of the PV power plants. The former might produce burns to the skin from solar exposure and
heatstroke, whereas the later would produce a decrease in the activities that the operators can
safely do during a day and altitude sickness, which may lead to headaches, dizziness, and
generate the feeling of sickness that may even require urgent medical care. Therefore, to pre-
vent the effects of high radiation and high-altitude, the tasks must be carefully scheduled so
that the risks for the operators are minimized. Furthermore, additional medical care equipment
is recommended to locally address any medical emergency.
For example, to prevent sun burn, it is required that the operators, in addition to the personal
protection equipment, wear long-sleeve shirts (desirable with some UV filter), trousers with UV
protection, neck cover, glasses with UV filters, and “sunblock cream” to protect the parts ex-
posed to the radiation (the face mainly). To prevent heatstroke, the activities of the operators
must be done during the morning and the afternoon mainly. This allows avoiding the day hours
with the highest radiation and therefore the hours with the highest risk for the personnel. Only
critical activities must be done during these hours. This fact is crucial during the implementation
phase and maintenance of photovoltaic power plants since activities with a high demand of
physical work could be required. If these recommendations are not considered, several delays
might be expected as well as health problems in the personnel. All the recommendations
should be accompanied by a clearly understood health-care plan and self-care policy. It must
be mandatory that every operator informs the health and security manager whether he/she is
getting sick due to solar radiation exposure. In turn, the health and security manager must
have internalized the procedure to be carried out. The World Health Organization (WHO) has
published a practical guide to the "Global Solar UV Index", which should be taken into account
and may be of practical benefit [103].
With regards to the exposure of operators to low oxygen environments, the measures to re-
duce the risk are more focused on health monitoring and on the scheduling of the activities
than on the acquisition/wearing of additional equipment itself. In this context, the recommen-
dations start with requiring a medical test that certifies that every operator can work at high
altitude locations. Then, every operator must be trained in providing first aid and in the use of
first aid equipment. The training must be accompanied by 1) a procedure for the attendance
of medical emergencies that may occur when the team is travelling and when it performs ac-
tivities at the photovoltaic power plant, and 2) the acquisition and placement in the plant of
96
proper equipment to address medical emergencies. In addition, despite the first aid training of
the operators, it is desirable that doctors/paramedics are present in the photovoltaic power
plant in case of emergency situations and to periodically monitor the health of the operators.
Moreover, the displacements from/to the plant must be carefully scheduled. For this displace-
ment it is recommended that 1) at least two people have a driving license with high-altitude
driving certificate, 2) there exists programmed stops that allow the team to get accustomed to
the changes in altitude (the number of stops and their duration will depend upon the altitude at
which the activities will be done), and 3) the maximum driving hours per day are previously
defined (it is recommended no more than eight hours per day per driver, is should be forbidden
to drive at night hours). As a complement, oximetry measurements must be periodically taken
(both during the travel and when performing activities at the plant) to anticipate the presence
of undetected symptoms associated with altitude sickness.
In this regard, oximetry measures below 80% indicate that the operator must stop his/her ac-
tivities and go for first aid (in this case, breathe oxygen from an oxygen bottle or similar). Fi-
nally, the activities to be done in the photovoltaic plant must be scheduled considering that, at
least during the first days, the personnel must become accustomed to the new environmental
conditions, and that it is therefore not safe to carry out activities that demand high physical
effort.
5.3.5 Comparison with Issues and O&M in Moderate Climate

This section presents additional operation recommendations for photovoltaic power plants in-
stalled in high radiation and high-altitude desert regions. Besides the broadly used practices
in this area, the recommendations presented here are focused on 1) the use of water and 2)
soiling effects on the inverters. These specific topics were selected since water scarcity and
therefore water costs increase the operation costs of photovoltaic power plants in these desert
regions.
For instance, the water costs in high radiation and high-altitude desert regions could be ten
times higher than in urban areas, and high radiation exposure could cause skin cancer to the
operators. These issues make the operation of photovoltaic power plants located at high radi-
ation and high-altitude desert regions a challenging task from a resource management point
of view. Furthermore, in practice, it has been found that water scarcity, high temperatures in
the inverter area, and the accumulation of dust particles in the refrigeration systems increase
the operation costs and produce the largest number of failures.
Regarding the use of water, the main options explored/reported are the following: 1) the recy-
cling of the water used for cleaning purposes, and 2) the use of dry systems for cleaning the
modules. These options have been considered despite the cleaning task being accompanied
either with diagnostics tools such as those based on drones and image processing or with
other technological tools, such as cleaning robots. The idea behind this practice is to use of
the water as intensely as possible and hence reduce the amount of water to be purchased for
cleaning purposes (or reduce the frequency of water purchasing). To accomplish this objective,
an alternative option is to install water capture systems at the bottom of the modules, so that
the water is conducted towards a storage tank, instead of falling to the soil. After filtering the
dirt from the water, it would be possible to use the water again.
97
It is important to note that, in the design of the water capture system, the evaporation rate must
be considered to reduce the water losses. An alternative for reducing water losses is to sched-
ule the cleaning task during the less warm hours of the day. The main advantage of this alter-
native is that the scale formation on the surface of the modules is prevented (and the exposure
of the operators to high radiation). For photovoltaic power plants with tracking systems located
in desert regions with water condensation (such as the Atacama Desert), an alternative is to
leave the modules in an orientation that fosters self-cleaning during the night hours instead of
the safer flat position. With this practice, the amount of water required for the cleaning task is
highly reduced (or in some cases even avoided). Finally, the use of dry-cleaning systems is
gaining attention, despite the need for a structure (fixed or with a solar tracking system). De-
pending on the type of dust particles (size, structure, chemical composition, and interaction
with the glass surface of the modules), these systems can partially or totally clean the modules.
Then, as in the case of the tracking systems, the use of water can be avoided. In the case that
dry-cleaning systems are not totally able to remove the dust particles accumulated on the PV
modules, the combination with a water capture system is the best option to reduce the water
use in the plants.
With respect to high temperatures and the accumulation of dust particles in the refrigeration
systems, the use of hermetic cases for the inverters and liquid cooling systems with heat ex-
changers have been proposed. However, the investment and maintenance costs associated
with these alternatives make them unappealing. Thus, the use of air conditioners and air-based
cooling systems for the inverters is likely to remain common practice. To prevent damage to
this equipment, it is recommended that the air filters are cleaned and replaced periodically.
Furthermore, the temperature inside the inverter areas must be continuously monitored to de-
tect temperature increases that could cause either a loss of performance of the devices or
even fires. In addition, it is recommended to periodically check that the heat dissipation sys-
tems installed in the inverters are clean. The accumulation of dust particles or any other foreign
element may cause an increase in the internal temperature of the inverter, which makes it
unable to work under its optimal range of operating conditions. This may lead to energy losses
as well as the increased probability of major damage to the inverter’s electronics. Indeed, in
the Atacama Desert, failures in the inverters’ electronics are the most common type of failure
for these devices.
5.4 O&M Guidelines for Hot and Humid Climates (South-East Asia)
The increasing number of PV deployments in Asia, Africa and South America has led to sig-
nificantly more PV modules being deployed in hot and humid climates. Moreover, many of the
big cities in these regions are located in coastal areas, thus the PV deployments in and around
these cities are exposed to heat and humidity. High temperature is generally known to accel-
erate several degradation modes in PV modules, however, when high temperature is com-
bined with high humidity environments, different kinds of degradation modes can manifest
themselves.
Southeast Asia lies in the tropical climatic zone and includes eleven countries. Thailand has
three distinct seasons, namely hot, rainy, and dry or relatively cool season and an average
98
temperature around 30°C Celsius throughout the year. The highest temperature typically oc-
curs in April, regularly above 40°C. Most of the year, the region is covered by warm and moist
air particularly in coastal areas. The relative humidity may be significantly reduced in the winter
and summer months. Irradiation in Thailand is highest in April and lowest in December [104].
5.4.2 Field Experience of Reliability and O&M Issues Seen in Hot and Humid
Climates
A. Potential Induced Degradation
The most common type of Potential Induced Degradation (PID) is caused by the shunting of
P-N junctions in solar cells due to accumulation of Na+ ions in the stacking faults present in
the wafers. PID can rapidly cause significant power loss in PV modules thus forcing large
amounts of modules in a plant out of operation very early in their service life. It is known that
for a given type of PV module, PID is accelerated by an increased negative voltage experi-
enced by the cells with respect to ground, temperature, humidity, and the surface conductivity
of the glass (all leading to higher leakage current).
As the maximum system voltage of PV systems has been increasing from 600 V a few years
ago to 1500 V currently, newer PV modules must endure higher voltages during their service
life. Due to power conditioners without transformers, it is often not possible to physically ground
any end of the string of PV modules, and thus even for an electrically floating PV system, some
modules in the string end up experiencing considerable amounts of negative voltage with re-
spect to ground - thus attracting Na+ ions from glass, potentially leading to PID.
Figure 29 DEGRADATION RATE Histogram of linear degradation rates of Pmax: All c-Si
PV modules in which modules from PID affected sites differentiated along with their
position in the string.
99
Figure 29 shows the histogram of module degradation rates for c-Si PV modules after dis-
counting for 2% of initial power loss due to LID as observed in the All-India Survey of PV
Module Reliability 2018 performed by the National Centre for Photovoltaic Research and Ed-
ucation (NCPRE), IIT Bombay. The modules coming from PID affected sites (highlighted in
colours other than blue) tend to belong to the higher degradation rate side tail of the histogram.
Moreover, the modules belonging to the negative end of the string tend to experience higher
negative voltage with respect to ground. It is seen in Figure 29 that such modules (highlighted
in red) tend to have higher degradation rates than the rest of the modules. Also, the severity
of PID is shown by the fact that these PID affected modules tend to have significantly higher
annual degradation rates than those warranted by the module manufacturers (~0.7%/year).
B. Wildlife
Most of the operation and maintenance risks for PV power plants, especially in ground-
mounted systems in Thailand, originate from wildlife, especially rats, termites, and snakes. As
most of PV power plants are installed in rural areas and surrounded by agricultural fields, wild-
life management becomes an important issue to prevent failures in PV components and elec-
trical breakdown of the PV system. The highest failure occurrence was recorded for effects
due to termites, followed by rats and snakes, as shown in Figure 30. Solar cables and PV
modules are the critical components that most frequently cause failures due to the impact by
wildlife over a long period of time.
Figure 30: Impact by wildlife on PV components, such as a rat in the main distribution
box (left) and termites around the DC cable under the PV module (right).
100
C. Soiling issues
Due to the dry climate and PV installations in agricultural areas, plants may be the main reason
of the soiling impact on PV power plant, particularly for components such as PV modules and
inverters. Different locations show specific types of dust which may be more or less difficult to
remove. Dust accumulation must be removed from the top of the soiled inverters to avoid over-
heating and them consequently switching off. A field study on the impact of dust on the perfor-
mance of PV systems and suitable cleaning schemes for Thailand leads to the conclusion that
appropriate cleaning can decrease the production loss due to soiling on the PV modules by 6-
8% during the summer months [105].
D. Fire risks
Fire can easily occur in areas with a dry climate for various reasons. There are two kind of fire
risks: internally and externally caused fires. Within the first group of risks, fire accidents may
be caused by smoking or by fuel igniting inside a PV plant. External fire risks may typically be
present at the end of the agricultural season, when agricultural waste is typically disposed of
by burning it. If the wind blows the fire towards the PV installation, it can bring the flames to
the PV power plant, which is the main cause of fire accidents in most PV power plants in
Thailand.
E. Degradation rates
In hot and humid climates, weather conditions can very often result in partly cloudy skies caus-
ing rapid changes in module temperatures due to full solar irradiance (possibly enhanced by
reflections from clouds up to 1200 W/m²) and a sudden change to diffuse irradiance conditions
at much lower irradiance (e.g., below 300 W/m²) when the sun is hidden behind a cloud. The
thermomechanical stress on the module components caused by such very frequent tempera-
ture changes was found to be more stressful (solder joint fatigue) than the thermal cycles in
desert areas (change between cold night and hot daytime), although the amplitude is much
lower. In IEC 62892 [106], Annex B provides a calculation based on temperature monitoring
data to estimate this thermomechanical stress to decide whether to test modules with extended
thermal cycling.
5.4.3 Operational Health and Safety Issues

The occupational health and safety (OHS) guidelines, described in section 5.2.3, are also valid
for PV power plants in hot and humid climates such as that of Thailand. OHS guidelines should
assist operating personnel and staff members to help with compliance and protect them from
sudden situations, personal risks, and damage to their health.
In combination with the climatic conditions, e.g., hot temperatures and the agricultural season,
certain risks of fire accidents may occur. The main reason for this is agricultural waste disposal
at facilities to clear land and prepare for the next growing season. However, to prevent the risk
of fire accidents, regular fire training (Figure 31) and the localisation of all potential fuel sources
are the most important measures to prevent a fire accident. Finally, first aid training of PV
power plant personnel is also essential. Receiving a basic background of first aid will encour-
age staff to prevent any calamity from worsening as well as enable them to properly take care
of injured personnel (Figure 31).
101
Figure 31: Regular fire accident training (left) and first aid training (right) in Thailand.
Figure 32: Regular checking and performing preventive maintenance in PV power plants
in Thailand.
102
5.4.4 Recommended Practices

Standard O&M best practices as described in section 5.1 also apply to PV power plants in-
stalled in hot and humid climates. Additional key recommendations for O&M to prevent typical
risks in PV power plants operating in hot & humid climates are given in the
Table 10.
Table 10: Recommendations for O&M of power plants in hot and humid climates.
Risk Cause Impact Recommendation
Wildlife such as rats Animal bites will threaten First aid training for PV power plant
and snakes the health of operators staff (Figure 31).
working on site.
Animals inside the PV Preventive maintenance actions and

system destroy compo- routines (Figure 32) and system moni-
nents and cause failures. toring to detect failures and to avoid
performance losses.
Fire accidents Health of operators and Regular fire accident trainings and
staff affected by fire first aid trainings (Figure 31).
risks.
Fire destroys PV compo- Fire protection measures to be imple-

nents and the PV sys- mented. Increase communal respon-
tem. sibility to prevent agricultural waste
disposal being burned.
Climate conditions When working under ex- Install proper OHS guidelines for per-
(summer season) treme weather conditions sonnel and operators of PV power
(hot temperature), the plants.
operating personnel may
experience dehydration
and heart stroke.
High temperatures in the Adequate design of air ventilation is

inverter cabinet (60°C in crucial for remaining good operation
Thailand) may lead to conditions for inverters as well as re-
power losses and switch ducing temperature inside the inverter
off. cabinet/ room.
Imported inverters may Specific value settings for inverters

have different initial set- considering the operating climate are
tings given by the manu- highly important for the PV power
facturer. This may lead plant commissioning process.
103
to the mismatch due to

hot temperatures.
Climate conditions Flooding of the PV sys- Using Personal Protective Equipment

(rainy season) tem might impact opera- (PPE) and performing the operation
tors during O&M service under the safety rules (Table 4) would
(electric shock). Further- help to prevent risks of electric
more, dangerous ani- shocks and dangerous animal attacks
mals may come to the and bites.
PV power plant and
threaten the operators
during their work.
As most DC cables in Leakage issues might be detected by

Thailand are installed using effective data monitoring sys-
underground, floods may tem.
cause electrical leakage
and system failures.
Soiling effects Plant performance may Regularly monitoring of PR output for

be impacted by soiling cleaning decisions and frequency. In-
on PV modules. clude PV module cleaning in O&M
contact as basis practice for this cli-
mate.
Inverter filter may be Regularly cleaning inverter filter

covered by dust, which would help the resistance of airflow
might cause resistance by dust accumulation.
of airflow, leading the
performance loss.
Glass breakage Stone impact on the Regularly monitor the PR output of

module’s surface can strings and array and check the PV
cause glass breakage modules by visual inspections.
and cause performance
decline.
104
5.5 O&M Guidelines in Flood-Prone Regions

5.5.1 Background and Motivation for Studying Reliability and Safety Issues in
Flood Affected PV Plants
Ground mounted PV systems in flood-prone areas are susceptible to damages resulting in
safety hazards and performance loss. Flooding can also increase the risk of erosion of the
support structure and foundations, depending on geotechnical conditions. The findings and
recommendations regarding flood-affected PV power plants in section 5.5 are largely based
on the published paper by one of the contributing authors of this report in Shiradkar et al. [107].
Climate change has increased the frequency and severity of extreme weather events through-
out the world [108]. Although PV plants are expected to have a service life of greater than 25
years under normal conditions, extreme weather events such as floods can cause significant
and irreparable damage to them. Understanding the vulnerability of existing PV plants is an
important step in the development of guidelines for flood resilient PV plants.
In recent years, flooding in the northern parts of India has caused significant damage to prop-
erty and livelihood [109]. Flooding and waterlogging have also been observed in semi-arid
areas that were considered less prone to it [110]. Moreover, floods in the southern state of
Kerala were recorded to be the worst in a century [111]. PV plants in flooded areas experienced
significant damage and/or disruption in the operations in many cases. The National Centre for
Photovoltaic Research and Education (NCPRE) at the Indian Institute of Technology (IIT) Bom-
bay had conducted a field study to identify reliability and safety issues seen in flood-affected
plants to come up with guidelines for better operation and maintenance (O&M) and plant de-
sign in flood-prone areas. In this section, the technical, financial, and strategic motivation for
studying flood-affected PV plants is presented.
While designing a PV plant, several decades of history of flooding are considered and typically
channels are dug in the ground to direct the water during events of flash flooding. Since floods
are considered as rare events, no other aspect of the PV plant is typically designed with flood
endurance in mind. However, several studies have predicted that the frequency and severity
of flooding are going to increase globally in the years to come and flooding would become
common in areas where it is currently a rare event [112].
Therefore, the effects of climate change need to be considered while designing the PV plant
that is built to last for 25 years. Merely looking at the history of flooding would result in a sig-
nificant underestimation of the risk. For some PV plants, risk due to flooding is not a potential
future problem caused by climate change but is a clear and present danger. For example, due
to the scarcity of land in India and large land requirements of multi-MW PV plants, it is recom-
mended that the land acquired for large PV parks should be unsuitable for agriculture, industry,
forest, inhabitation, etc. This has forced the development of future large solar parks in some
of the harshest environments observed in the country. For example, one location for a planned
solar park is known to be affected by flooding and waterlogging two to three months in a year.
Thus, a significant number of PV systems will be deployed in the harsh and flood-prone regions
in the coming years. This requires innovative designs of modules and a suitable balance of
system components (especially structures) that can survive in flood-prone locations.
It was also observed that insurance policies can vary significantly in terms of their coverage
and conditions for extreme weather events. Some policies that are specially made for flooding
events, require that the PV plant should be switched off in the anticipation of flooding and it
105
should be only switched on, after technical experts have given permission. This is difficult to
accomplish because most utility-scale PV power plants with central or string inverters do not
have a ‘rapid shutdown’ feature. In other words, even when the inverters are turned off in
anticipation of flooding, the strings of PV modules continue to generate high voltages and it is
often impractical to disconnect all PV modules from each other during an emergency event.
Finally, even though the PV modules are not designed for continuous immersion in water, it is
important to understand which types of modules and balance of system components show
resilience to immersion and floods.
5.5.2 Field Experience of Investigations of Reliability and Safety Issues Ob-

served in Flood Affected PV Plants
A. Demographics of Surveyed Sites and Classification of Damage
The demographic details of PV plants selected for the detailed study are given in Table 11 The
damage caused by flooding can be classified into two types: 1) caused by fast-flowing water
and/or impact with debris 2) caused by waterlogging and hence continuous submergence un-
der water. The magnitude of type 1 damage depends on the resistance posed by the structures
and modules to the fast-flowing water. It was found that no structure in a typical ground-
mounted PV plant is strong enough to endure fast-flowing waters if it poses significant re-
sistance to the water flow. On the other hand, the amount of type 2 damage depends on the
duration and depth of submergence and the specific equipment such as modules, string com-
biner boxes (SCBs) or inverters that were submerged. Figure 33 shows an image of a partially
submerged plant P1.
Figure 33: Partially submerged PV plant showing strings fully submerged in water (in
blue circle) and those that are above water (in black circle).
106
Table 11: Details of the selected PV plants for investigation.

Plant Number Plant Details Flooding Status
P1 15 MW, Ground-Mounted, South facing, Waterlogging, partially
latitude tilt submerged
P2 6 MW, Canal Top, South Facing, 6° tilt Submerged under
fast-flowing water
P3 2 MW, Ground-Mounted, South Facing, Submerged under
latitude tilt fast-flowing water
P4 10 MW, Ground-Mounted, South Facing, Waterlogging, partially
latitude tilt submerged
B. Structural Issues
Significant structural issues were seen in PV plant P3. As seen in Figure 34 (a), the structures,
module mounting clips, as well as modules were broken. In some cases, the laminates were
seen to be separated from frames, wires detached from junction boxes, and broken glass was
observed due to the impact of debris. In some cases, (Figure 34 (b)) it was seen that the
modules remained intact on the structures, but the pressure from water and debris stuck be-
hind the module uprooted the concrete foundations of the structures. Further, erosion caused
by soil around the foundations was also observed. It should be noted that even though this PV
plant was south facing at latitude tilt, and the water flow direction was East/West, the pertur-
bations in flowing water were sufficient for the modules and structures in the PV plant P3 to
get severely damaged by the fast-flowing water as they offered too much resistance to the
flow. Thus, typical steel structures with concrete foundations, modules and mounting fixtures
used in PV plants are seen to be extremely vulnerable to damage when they pose resistance
to fast-flowing flood water.
107
Figure 34: (a) Structural damage to PV plant P3 showing broken structures and mod-
ules. (b) An example where modules remained intact, but the foundations of structures
were seen to be uprooted.
Figure 35 shows the top and bottom views of a PV plant (P2) that has been constructed above
a water channel. This PV system has concrete structures with steel bars for module mounting
which have been proved to be significantly resilient against floods. This PV plant has south-
facing modules, mounted at an inclination of 6°, while the direction of the floodwaters was
East/West. It is conjectured that due to the south-facing nature and low inclination angle, the
modules posed minimal resistance to fast-flowing water. Moreover, the concrete beams proved
to be considerably sturdy despite being in the way of flowing water. However, such concrete
structures are rarely seen at ground-mounted utility-scale power plants.
108
Figure 35: (a) Top view of Canal top plant P2 showing that hardly any damage to mod-
ules or structures was seen despite this plant being submerged under fast-flowing wa-
ter. (b) Bottom view of the all-concrete structure of P2 showing concrete beams.
C. Soiling / Staining Issues

As shown in Figure 36 (a), silt was deposited on many modules of PV plant P2, that were
mounted at a low inclination angle of 6°. After cleaning the silt by the O&M team of PV plant
P2, uneven stains were observed on the glass of the modules as shown in Figure 36 (b).
109
Figure 36: (a) Silt was seen to be deposited post-flooding on most of the low inclination
angle mounted modules in P2. (b) Uneven stains were observed near the bottom edge
of most of the modules which could not be removed even after cleaning with soap water
and scrubbing.
Hotspots with a typical temperature difference of 10°C were observed in modules with stains
on glass (Figure 37). These stains (and therefore hotspots) were permanent and could not be
removed even after vigorous cleaning by soap water followed by scrubbing with a cloth. Since
the stains were only observed near the bottom edge of the modules, it was conjectured that
the dissolved minerals in the stagnant flood water that remained near the lower edge of the
modules with low inclination angle could be the cause of staining (and not merely deposition
of silt – which was distributed throughout the module). This shows that low inclination angle
mounting may provide benefits of low resistance against floodwaters, but it can be vulnerable
to soiling/staining issues if the water stays on PV modules for a long time.
110
Figure 37: Modules with uneven stains on glass were seen to have developed hotspots
with ΔT ~ 10°C.
5.5.3 Performance / Safety Issues in this Region

A. Effect of Continuous Submergence Underwater
In some parts of the PV plant P1, modules remained under about 0.3- to 1-metre-deep water
for over two days. Since most commercial PV modules are not designed to withstand continu-
ous underwater immersion, this could cause several kinds of reliability issues for the PV mod-
ules, depending on the design. For example – short-circuited/damaged bypass diodes, accel-
erated corrosion in the presence of water (in the junction box, connectors and even fingers on
cells), accelerated potential induced degradation (PID), cell/backsheet cracking, etc. Note that
these are problems merely due to submergence and not due to resistance to fast flowing water
and debris hitting the modules.
Modules of PV plant P1 had IP67 rated junction boxes filled with pottant. Note that the test for
IP67 rating involves immersion of junction boxes at 1 metre depth in water for 30 min. A sig-
nificant number of bypass diode failures that have been found in PV plant P1 were limited to
only one to two strings. Some of these were accompanied by visible burn marks on the junction
boxes (Figure 38). Furthermore, diode failures were often accompanied by burn marks at the
junction box. Any penetration of water inside the junction box would have caused an immediate
short circuit and catastrophic failure. However, many submerged modules in PV plant P1 did
not experience diode failures. This indicates that the IP67 rated junction box with pottant pro-
vided adequate protection for continuous submergence under 0.3- to 1-metre-deep water for
two days.
111
Figure 38: Burn mark and deformation is seen at the centre of the junction box with
diode failure. The manufacturer's name is masked in black.
To assess the effect of submergence on power degradation, seven modules each from two
submerged strings and seven modules each from two non-submerged strings of PV plant P1
were selected for I-V analysis. The I-V curves of the modules were measured outdoors at
irradiance > 700 W/m2 and the power was translated to STC using modified IEC 60891 Proce-
dure 1 [113]. From the results, it could be concluded that there were no statistically significant
differences in power degradation between the submerged and non-submerged modules as
shown in Figure 39. Furthermore, there were no signatures of accelerated PID or corrosion.
Also, no significant cell cracking was observed in the submerged modules. However, on some
modules backsheet scratches/cracks have been observed. It appears that the actual damage
caused on the modules of P1 due to the submergence was much less than what could have
been with a different (poorer) module design.
112
Figure 39: No statistically significant difference was seen in the power degradation of
submerged and non-submerged modules.
B. Safety Issues
As the water level rose, the water entered the SCBs / Inverters of PV plant P1 and P2 first, as
they were at a lower elevation than the modules. Since the modules were still producing volt-
age when this happened, every submerged SCB / inverter failed catastrophically. For example,
Figure 40 (a) shows that fire was observed on an inverter, several hours after the floodwater
had receded. Burn marks like those shown in Figure 40 (b) were observed in large numbers
of submerged SCBs / inverters that had failed. Since flooded SCBs / inverters may be wet or
internally damaged, they pose safety hazards even several hours after the floodwaters recede.
It is difficult to further increase the height of SCBs / string inverters in PV plants (like P1) to
avoid submergence as they can cause partial shading on the nearby modules.
113
Figure 40: (a) Fire was seen at an inverter several hours after the floodwaters had re-
ceded (b) A similar fire probably caused burn marks and failure of several submerged
inverters. The manufacturer’s name is masked in black.
In the case of PV plant P4, the SCBs were submerged and were damaged with burn marks as
shown in Figure 41 (a). Even though the water level around this PV plant did not rise enough
to submerge the modules, peculiar burn marks were observed near the frames of the modules
as seen in Figure 41 (b). The burn marks were so severe that the module glass was melted.
The SCBs in this plant did not have reverse current protection. The root cause of this phenom-
enon was not clear. Reverse currents flowing in the strings due to shorting as the SCB goes
underwater could be one possible explanation. This case study shows that it may be danger-
ous to approach PV modules during flooding even though they are not submerged in water.
In addition to the electric safety issues discussed above, flood-affected PV plants also pose
other kinds of safety risks such as displaced and distressed wildlife after flooding. For example,
there were sightings of poisonous snakes and crocodiles in urban areas after the Kerala floods
of 2018 [111]. Therefore, the personnel visiting the PV power plant immediately after floods
should exercise precautions and expect to find wildlife at unusual places.
114
Figure 41: (a) Damaged SCB due to submergence. Prominent burn marks are seen.
(b) Even though the modules were not submerged, severe burn marks that cause the
melting of module glass were seen near module edges.
5.5.4 Recommendations
Since the frequency and severity of floods are expected to increase worldwide due to climate
change, increasing numbers of PV plants would become vulnerable to flooding in the coming
years. Various types of failure modes and mechanisms have been observed from flood-af-
fected PV plants and have been discussed in this section. Floods expose the modules and
balance of system (BoS) components to two types of stressors: 1) Fast flowing water and/or
impact with debris 2) Continuous submergence.
Typical steel structures with concrete foundations used in ground-mounted PV plants have
been seen to be extremely vulnerable for severe damage in the case of a type 1 stressor,
wherever they pose resistance to fast-flowing water. Further, it was found that none of the
structures of the typical ground-mounted PV plants were strong enough to endure fast-flowing
waters if they pose significant resistance to the water flow. South facing and fixed tilt mounting
at low inclination minimizes the resistance to water flow (East/West) and mitigates the damage
due to the type 1 stressor. Also, an all-concrete structure with bars for module mounting in a
canal-top plant was shown to be especially resilient against fast-flowing water. However, flood-
water may contain minerals that could permanently stain the glass leading to hotspots. This
happens when the floodwater stagnates on the low inclination mounted module surface and
evaporates over time.
115
Despite being 0.3 to 1 metre under water for two days, the PV modules with IP67 rated junction
boxes did not show any signs of accelerated degradation. Therefore, IP67 or better rated junc-
tion boxes with pottant are seen to provide enhanced protection even in the case of continuous
submergence (for which the junction boxes are not certified). Therefore, such junction boxes
are strongly recommended in flood-prone areas. Lack of significant degradation in submerged
modules also shows that the IP67 rating could be used for all the PV modules and some mod-
ules (as in the ones encountered by NCPRE) may pass the test without needing any changes
in the bills of materials (BoM). Module manufacturers could use this as a differentiating factor
in favour of their products for example for the applications in flood-prone areas. This could be
a significant added value for PV plant owners who have modules submerged in water but do
not seem to have any apparent damage as seen in I-V / IR / EL measurements. Also, this
would require the development of procedures to assess cases in which the manufacturer’s
warranty could be continued post submergence due to flooding.
IP67 or better rated boxes for inverters and SCBs could mitigate some of the problems due to
submergence for a few days during flood events. However, it would significantly reduce their
maintainability as the electrical access would not be possible in the case of pottant filled boxes.
Increasing the height of SCBs / string inverters may not be an option as they may cause partial
shading on the nearby modules. Using modules with micro inverters can be explored in such
areas as the micro inverters are often potted and rated IP67 or better. They would not only
provide advantages under continuous submergence scenarios, but they are also often
equipped with a module-level rapid shutdown feature. This can be useful to quickly disconnect
individual modules when floods are expected.
Before implementing PV plants in flood-prone areas, it is recommended to conduct studies
regarding the floodplain area and the water surface elevation. The history and the future pre-
dictions of the water surface elevation for approximately 100 years for flood-prone areas con-
sidering the effects of climate change are recommended. If the maximum water level caused
by the floods is known, weather bars or upstands can be used to prevent flooding of electrical
equipment buildings. Further, elevated pads can be provided to prevent flooding of ground-
mounted equipment. Existing and new drainage should also be considered to ensure that the
run-off is controlled to minimise erosion [114].
Finally, extreme precautions should be taken by the O&M team when returning to the PV power
plants after the floodwaters have receded. This because various electrical parts can still be
wet, and live fires are often seen days after the floodwaters have receded. Moreover, danger
due to the presence of wildlife at unusual places should also be considered while performing
the O&M operations immediately after flooding. Overall, the PV system can behave unpredict-
ably once its components have been submerged in water and all physical contact should be
avoided until it is deemed safe to enter the PV plant. Post-flooding, drone-based inspections
of PV plants can be performed to assess the status.
When developing procedures and standards to quantify the vulnerability of PV plants in case
of flooding, the following should be considered: the module design, IP rating, structural design,
SCB and inverter elevations, reverse current protection, etc. This may also be useful for the
financial risk assessment of PV plants deployed in flood-prone areas.
.
116
5.6 O&M Guidelines for Building Requirements for Cyclonic Regions

Strong winds attributed to a tropical cyclone (also known as hurricane, typhoon, or cyclone,
depending on the region in which it developed) may cause severe damage to the PV systems
located in its path. Damages that have occurred around the world are described in this section.
5.6.2 Field Experiences on Damages Caused by Strong Winds with Hurricane /

Typhoon / Cyclone
In the last few years, several typhoons with strong winds (over 54 m/s) have hit the coasts of
northeast Asian countries / regions. It has been recently revealed that the PV systems installed
in these areas suffered from severe damage to PV modules, mounting structure, and their
relevant equipment / architectures, which resulted in heavy financial losses (including the long-
term reduction of revenue). As shown in Figure 42, two strong typhoons (JEBI and TRAMI)
damaged respectively 20 and 12 PV facilities (in over 50 kWp facilities) in 2018, according to
official reports published by the Japanese government [115], [116].
The damages in these PV facilities were primarily observed in PV modules and mounting
racks. Nearly half of the installed PV modules were blown out or broken by ca. 60 m/s wind-
gusts in a PV facility (6.5 MWp installed with ca. 28 000 modules) located at the coastal area
in Japan, and the mounting rack on the rooftop was also broken. Even in a PV facility (10 MWp)
installed in an inland area, the glass breakage was confirmed in about 13 400 out of 36 500 PV
modules, although these PV modules were not blown out from the rack on the ground. Also he
floating PV systems were impacted [115], as many PV modules were turned over (after the
bolts connecting the anchors and floats fractured), and fire damage (maybe caused by electric
arcing) was also observed [117], [118].
117
Figure 42: Damages of PV facilities by typhoons in Japan in 2018 and 2019.
Figure 43: Damages of PV facilities by hurricanes in U.S. Virgin Islands in 2017.

Also, in strong hurricane prone areas (especially in Caribbean islands), severe damages of PV
systems have been reported [119], [120], [121], [122], as shown in Figure 43. Hurricanes Har-
vey, Irma, and Maria inflicted some hard blows on the PV systems installed in US Virgin Is-
lands, in 2017. Two major damages were the blown-out of PV modules and the breakage of
mounting racks. These damages occurred not only in the large PV plants but also in the rooftop
118
PV systems on individual residences. The breakage of fixing parts (bolts and clamping) was
often observed in such damaged PV systems, which could be attributed to the uplift wind-
pressure.
Europe is not exempt of harsh environments where wind is the main factor of natural catastro-
phes. The next case is an example of a windstorm destroying 30% of a roof mounted 115 kWp
PV system in Northern Italy (Figure 44). The storm hit several Northern regions, where these
unusual events have been increasing over the past few years and left 14 million trees razed at
the end of October 2018 [123], [124].
Figure 44: PV System status after a windstorm in Baselga di Piné, Italy. Source: Dino
Loriatti.
As shown in Figure 45, the maximum wind speed measured around Baselga di Piné that fatal
day was 127 km/h (35.3 m/s). Although this wind speed was not evaluated as an extreme wind
gust, the designs and constructions of the fixing system were arguably not solid enough, even
for these quasi-extreme wind gusts and their consequences.
119
Figure 45: Gust speed during the windstorm day in Baselga di Piné, Italy. Source: Lo-
renzo Mercuri [125].
5.6.3 Performance and Safety Issues in this Climate

Cell cracks induced by deflection caused by strong wind loads is a major performance issue
in this climate. As indicated in our previous Task 13 Report [28], the electrical isolation of a
portion of PV cells is a crucial reason for power loss. Major electrical and structural safety
issues are caused by the breakage of fixing parts / mounting structure, including PV modules
being lifted of their structure or their roof and glass breakage of PV modules, as described in
section 5.6.2.
5.6.4 Strong Wind Measures Described in Building Requirements and O&M

Guidelines
Basically, the design of PV facilities is subjected to the building and electrical codes enacted
in the individual countries or regions. In this subsection, the established guidelines and the
proposal for the strong wind measures have been summarized.
In the USA, some codes and standards provided the design guidance for rooftop PV systems
[122]. The wind load criteria for rooftop PV systems was added in the 2016 edition of ASCE 7
[American Society of Civil Engineers “Minimum Design Loads and Associated Criteria for
Buildings and Other Structures'' (ASCE 7-16, 2017)] [126], as well as the updating of SEAOC
PV2 [Structural Engineers Association of California “Wind Design for Solar Arrays”
(SEAOC PV2-17)] [127]. ASCE 7-16 provides the criteria for determining wind loads on rooftop
PV systems, although those on ground-mounted PV systems have not been included. If
needed, the recommendations and guidelines on ground-mounted PV systems could be ac-
quired from SEAOC PV2-17 and “FM global loss prevention data sheet 7-106” [128]. In the
guidance for ground-mounted PV systems, the importance of wind tunnel testing on PV arrays
and the dynamic effect of winds on ground-mounted PV arrays are emphasized to prevent the
blow-off of PV modules, structural failures, and other types of damage.
120
In Japan, where some strong typhoons hit every year, “Load design guide on structures for
photovoltaic array” has been established as a Japanese Industrial Standard (JIS C 8955:
2017) [129]. This standard provides a guidance for allowable stress in the design of structures
that constitute a PV array to be installed on the ground or on building structures. However, the
following are not covered by this standard:
• PV arrays exceeding nine m in maximum height.
• Building integrated arrays to be replaced with the building materials such as roofing,
walling, or windows.
• PV arrays to be installed at a ground height exceeding 60 m.
In this standard, the design wind loads (unit: N) for PV array and structural components shall
be calculated by the multiple of the wind factor, design wind pressure, and wind-receiving area.
Depending on the mounting mode (on the ground, the sloped roof, or the flat roof), the tilt angle
of PV array, the wind direction (same or opposite direction), and the position of interest in the
PV array, the wind factor for the PV array is estimated through the defined formula in [129].
Although the wind factor for structural components should be determined by a wind tunnel test,
it could also be estimated by the cross-sectional shape which is applied to the structure. The
design wind loads are computed from the reference wind speed (which is assigned to the
location), the environmental factor (which is specified by the ground surface roughness class,
the gust factor, and the average height of PV array plane), and the quantified importance fac-
tor. The practical design principle and cases (including the case that the PV array would be
installed in a sloped area) also have been published as an interpretive manual [130].
In Europe, the Eurocodes, a series of ten European Technical Standards that provide a com-
mon approach to the structural design of buildings and other civil engineering works applies
[131] and [132]. Part 1-4 applies for wind actions on structures and the wind load design. Basic
wind velocity values, as a 10-min-mean, at a height of 10 m above ground above an open
terrain (category II) with a once in 50-year probability can be calculated based on country maps
with wind zones. These values are then further modified by factors considering surface rough-
ness, height, and shape of the structure, etc. [133].
Recently, in the IEC Photovoltaic Standardization Committee TC82, a new Working Group,
WG9 was launched, because of a need to address standardization of the design and fabrica-
tion of structures used to support PV arrays and their associated system components [134].
First projects will focus on the update of tracker structures’ design qualification (IEC 62817)
[135] and safety requirements (IEC 63104) [136]. Recently, a new work item proposal was
made to write a technical specification for the interface between modules, clamping and the
mounting structure(s) [137].
During the last years, non-uniform dynamic mechanical load testing of PV modules and sub-
structures was also investigated [138]. However, although a new work item within IEC TC82
was voted positively, it could not be launched because of a lack of experts/countries to form a
project team.
5.6.5 Recommendations
In the “Operation & Maintenance Best Practice Guidelines” [11], the maintenance to prevent
the damages with strong wind loads is positioned in the category of extraordinary maintenance,
but not in the preventive-, corrective-, or predictive-maintenance, because the maintenance
121
activities for strong wind loads are not generally covered by the O&M contract. That is, the
damages with strong wind loads have been recognized as “Force Majeure” events. However,
as described in section 5.6.1, the catastrophic failures of PV systems are easily induced trop-
ical cyclones once they make landfall. Therefore, various recommendations for the specific
damages with strong wind loads should be required, beyond ordinary maintenance activities
(in the current O&M guidelines [11], [139], [140], [141], according to IEC 62446-1/-2 [142],
[143], IEC TS 63049 [144], and other standards [145], [146], visual and physical inspections
to detect the failures of PV modules/arrays and structural assemblies/components are speci-
fied as the regular maintenance procedures). In accordance with the valuable experiences,
recommendations that will be useful for the prevention of damages due to strong wind loads
are summarized in this subsection.
A. Planning and Design Stage
To estimate the effects of turbulent wind gust, wind tunnel tests (with high-speed wind) on the
scale model of individual PV systems should be carried out in the first stage, in addition to the
wind load calculations for that system. The obtained results should be carefully reviewed on
wind load criteria of all assemblies and attachments.
A site-specific wind dynamic load should be considered for the reviewing of the design (includ-
ing the combined effects of static and dynamic wind loads), not just the static wind load esti-
mated by the calculation or wind tunnel test.
PV modules with sufficient uplift resistance should be specified, to meet the estimated wind
loads.
In the structural design, sufficient size of structural members and connectors (including the
fixing parts for PV modules) shall be specified, to meet the anticipated high wind magnitude
and cyclic wind loads.
The closed ovular- or rectangular-section framing members should be specified in the mount-
ing structure design, because the torsional resistance of these shapes is better than those in
the open-section members (e.g., C-shape).
To avoid the loosening of bolted connections, stainless-steel locking (double) nuts with a nylon
insert should be specified for the PV module clamping bolt-nut systems, and the appropriate
torque levels of all bolted connections should be specified.
If possible, micro-inverters should be used for the PV systems, to allow the electrical-power
production by the undamaged PV modules even when one PV module is blown out or dam-
aged by wind-borne debris.
B. Construction Stage
To ensure the construction will be built according to the design, adequate quality-control/-as-
surance systems should be realized. All bolted connections should be made with a calibrated
torque wrench, and the specified torque levels should be applied to respective connections.
C. Maintenance Stage
Periodical maintenance: To shorten the restoration time after the damages by strong wind
loads, the appropriate types and number of repair parts should be stored, and the procedures
for repairs should be made available.
122
Periodical maintenance: The torque levels of all bolted connections should be checked (pref-
erably, an annual check is recommended).
Prior to the approach of a tropical cyclone: Tightness should be checked on all bolted connec-
tions with a torque wrench. At least, the connections at the clamps in PV modules should be
checked.
Prior to the approach of a tropical cyclone: Additional anchors to reinforce the structural as-
sembly should be applied, if needed.
Prior to the approach of a tropical cyclone: Debris and loose objects around PV systems should
be removed, to avoid damage caused by windborne debris.
After a severe windstorm: loose PV modules should be removed, and the torque levels of all
bolted connections in the remaining PV modules/arrays should be checked (if possible, within
a few days/weeks after the storm).
D. Others
In accordance with the most recent codes and standards, the practical activities for design,
maintenance, and others should be implemented. For example, because the cyclic (dynamic)
mechanical load test on PV modules would be (has been) included in the new international
qualification and type approval standards for PV modules with various technologies (IEC
61215 series) [98], [147], designers should specify an approved model(s) of PV modules in the
planning of PV installations.
5.7 O&M Guidelines for Snowy Regions

5.7.1 Climates in Snowy Regions
As PV deployment grows, the need to quantify the effects of snow accumulation on array per-
formance becomes increasingly important for accurate system sizing, yield forecasting, and
service life prediction.
High plateau and mountain areas all over the world have snowy climate zones. 90% of the
world population lives in the northern hemisphere (where most of the landmass is to be found),
in regions where snow is common during the winter season.
Snow may have a positive effect on power generation from solar PV as it increases the albedo
of the ground significantly. The positive effect increases when module tilt angle increases. In
Figure 46 there is an example of calculated ground reflected irradiance Gg as a function of
module tilt angle with albedo of 0.2 and 0.8 using Eq. (17), where  is the tilt angle,  is the
albedo and Gh is the horizontal irradiance.
Gg = αGh [1 − cosβ]/2 (17)
123
Figure 46: Ground reflected irradiance calculated as function of module tilt angle and
different albedo at horizontal irradiance of 800 W/m2, using Eq. (17).
A cold climate can be favourable for the efficiency of PV and frequent precipitation helps to
avoid soiling of the modules. However, snow can also have adverse effects on photovoltaic
electricity generation as heavy snow loads may cause loss of electricity generation due to
hindered transmission of light to the cells or damage of the modules due to its heigh weight or
because of icing [148].
Maps of mean seasonal snow cover extent for land in the northern hemisphere for the period
1981 to 2010 are shown in Figure 47. Snow cover extent reaches its maximum in January and
minimum in August, ramping up quickly in the fall and melting at a slower pace in spring [149].
124
Figure 47: Mean spatial distribution of seasonal snow cover extent over northern
hemisphere lands from January 1981 to December 2010 [149].
Scientists separate the Arctic into two major climate types. Near the ocean, a maritime climate
prevails. In Alaska, Iceland, northern Russia and Scandinavia, the winters are stormy and wet,
with snow. Summers in the coastal regions tend to be cool and cloudy; average temperatures
hover around 10°C.
The interior regions of the Arctic have a continental climate. The weather is dryer, with less
snow in the winter and sunny summer days. Winter weather can be severe, with frigid temper-
atures well below freezing. In some regions of Siberia and Alaska average January tempera-
tures are lower than -40°C. In the summer, the long days of sunshine thaw the top layer on the
permafrost and bring average temperatures above 10°C. At some weather stations in the in-
terior, summer temperatures reach 30°C and higher.
Although native people traditionally have been living in the Arctic, most of the roughly four
million living in the Arctic can be found living in modern towns and cities.
125
The rather mild climate in Western Europe south of the Arctic is mainly due to the warm Gulf
Stream, which makes the climate warmer than their eastern counterparts. In Sweden, snow
coverage effect is dominant for regions north of 60N, i.e., north of the Swedish capital Stock-
holm. The mountains in Norway lead to more and stronger snow fall than in Sweden. Finland
has a similar climate as Sweden north of 60N. South Alaska is on the same latitude as Stock-
holm but has much colder winters with more snow. Canada and parts of the USA also have
cold winters with snow issues. High mountain areas, more than 1 500 m over sea level, all
over the world experience snow, while the properties of the snow differ from site to site. Solar
irradiation is generally low in winter in areas where snow tends to cover the PV modules.
Pawluk et.al. [148] state that annual losses due to snow coverage are less than 10% in most
climates. This is also confirmed in studies made in Sweden by RISE Energy Technology Cen-
ter [150].
How to calculate snow load on a building as in EN1991-1-3 Eurocode 1 [151] is based on
historical weather data, and therefore might be subject to future changes due to climate
change. Dimova et. al. [152] summarized results of a European workshop on climate change
adaption of structural design, stating that - besides less frequent snow events - warmer and
more stable weather conditions may also cause high snow density and extreme snowfalls in
some regions.
5.7.2 Quality of Snow and Impact on Modules

A. Weight of snow and impact on modules
The impact of snow on modules is dependent on the character of snowfall, such as density of
snow, snowdrift, and icing. The character of the snow is dependent on ambient temperature,
wind and solar irradiance [148]. New snow generally has a lower density than wet or packed
snow. Ambient temperature also influences the density and volume of snow. Wind packed
snow or firn (dense granular snow) can have a density of more than ten times that of new
snow, see [153]. The numerical values of densities in kg/m³ are equal to the pressure in N/m²
for a 10.2 cm thick snow cover (last column). For a 1.4 m wind packed snow cover, the pres-
sure would result in 4.8-5.5 kN/m². Therefore, it is of importance that snow is not accumulated
on top of PV modules as these may be damaged by the snow mass.
Table 12: Density of snow/ice and corresponding pressure at 10.2 cm snow load [153].
Type of Snow or Density Pressure
Ice (kg/m3) (N/m2)
New snow 50-70 50-70
Damp new snow 100-200 100-200
Settled snow 200-300 200-300
Depth hoar 100-300 100-300
Wind packed snow 350-400 350-400
Firn (granular) 400-830 400-830
Very wet snow 700-800 700-800
Glacier ice 830-902 830-902
126
The threshold above which snow must be removed from the PV modules cannot be answered
unequivocally. Paterson [153] shows that about 1.4 m of heavy wind packed snow is needed
to reach load values higher than 5.4 kN/m2, the limiting value in mechanical load tests given in
IEC 61215-2 [147]. But only half that height, 0.7 m of heavy wet snow gives the same load
value. The study leads to a suggestion that 0.7 m of heavy snow can be a limit above which
snow is recommended to be removed, especially if the snow sticks to the panel. A risk of
damage due to repeated freezing-melting cycles is difficult to avoid in some climates.
Even though PV installations may be placed in windy areas so that snow coverage due to
snowfall is not a problem, snowdrift often occurs. Snowdrift may accumulate large amounts of
dense snow on PV panels.
B. Temperature of the snow
The temperature of snow cover may affect the shedding of snow from the PV panels. Research
activities in Alaska indicate increased propensity for snow shedding when temperatures are
around or slightly above freezing [154].
Existing mechanical durability test sequences typically perform mechanical loading prior to
environmental exposures such as thermal cycling or humidity freeze. Recent work has shown
that the fracture strength of silicon solar cells can be reduced after exposure to temperatures
below -20°C [155]. To better evaluate modules with respect to cell crack durability, the use of
a single thermal cycle prior to mechanical loading was explored. The results show a significant
increase in the number of cell cracks that are generated at a given load after a single cold
exposure. These results can be used to further optimize the qualification test sequence for
mechanical durability [156].
Electroluminescence (EL) measurements of snow loaded PV modules with cracked cells have
shown some open cracks which can close arbitrarily from one day to another. Researchers
found that variations in current and temperature generated from resistive heating during EL
measurements strongly influence crack closure. Because crack closure can lead to some gain
in maximum power, there may be ramifications for IEC standards, namely that performing EL
measurements before I-V measurements may lead to inflated results [157].
5.7.3 Racking and Spacer Elements

There are several issues to be aware of when it comes to racking in colder climates. Some
installers who have installed solar PV in interior Alaska, where swings in temperatures between
the winter lows and summer highs can reach 85°C, have reported that they prefer steel racking
as opposed to the more common aluminum racking because of concerns with expansion and
contraction. Aluminum expands and contracts up to 30% more than steel. However, steel be-
comes more brittle in the cold, therefore special considerations need to be made before choos-
ing racking systems. In the Alaska Solar Manual [158], racking systems made of steel are
advised.
Spacer elements have been explored as a novel solar panel mounting scheme. Such spacers
can be applied to either the rear of the module or to the rails on the mounting structure and
could be introduced for both new installations and as protective retrofits to existing systems.
These spacers significantly reduce the panel deflection under load, and researchers have
demonstrated a dramatic reduction in cell cracking at heavy load levels and in crack opening
after cyclic loading [159].
127
5.7.4 Field Experiences of Reliability in Snowy Climate

Figure 48 and Figure 49 show examples of what can happen with heavy snow loads and
freeze-thaw cycles in Northern Sweden, including roof collapse and destruction of PV modules.
Especially if snow freezes and melts in cycles, as is common in spring, the snow load on
modules can become heavy enough to burst and deform frames.
Figure 48: Picture taken on heavy snow and ice formation in spring in BJURÅS,
Northern Sweden. Later, the frame showed a damage as in Figure 49 [Courtesy of Klaus
Lorenz Högskolan Dalarna].
Figure 49: After winter, loose frames were observed on PV modules in Piteå in Northern
Sweden [Picture from Mats Axelsson].
128
Performance examples are also provided from flat roofs and modules mounted at low tilt an-
gles. A Swedish pre-study with results from north of 60° latitude on roof-mounted standard
framed modules indicates an improvement of the Performance Ratio (PR) when modules were
mounted with tilt angles greater than 30 degrees [160]. Most often the losses caused by snow
for standard framed modules at latitudes greater than 60°N in Sweden are less than 10% on
a yearly basis [160].
Powers et al. [161] assessed experimentally the power output of cleaned and snow-covered
modules installed side-by-side in the Sierra Mountains in the United States and found annual
production losses of 18%, 15%, and 12% for PV modules tilted at 0°, 24°, and 39° respectively.
Townsend and Powers [162] showed that the effect of snowfall on PV energy production could
be modelled as a function of tilt angle with acceptable result, using the same location as Pow-
ers in [161]. Estimated annual loss values were, using the model developed, determined as
26%, 17% and 13% for PV modules with 0°, 24° and 39° tilt angles, respectively.
Snow sliding has been reported several times as the most common cause of snow removal
[150], [163]. Figure 50 shows an installation called “Snow-wave”, consisting of frameless bifa-
cial modules in Piteå, northern Sweden, at 65N. The picture shows that with a larger tilt the
snow slides off the modules easier.
In some cases, snow naturally slides off tilted panels. First results from a study of unframed
bifacial modules in northern Sweden (65°N), see Figure 50. Granlund et al. showed that if the
panel tilt angle was greater than 15 degrees, snow slid off with small effects on the perfor-
mance ratio (PR) [150].
Figure 50: Snow wave, frameless bifacial modules, Piteå, N65, Sweden early spring
2019.
A study from Bavaria, in southern Germany, of frameless PV modules at 28° tilt found that
annual losses from snow cover for a grid connected system varied between 0.3% and 2.7%
[164].
129
A study of production losses associated with snow coverage in Konya, Turkey, [165] in January
2017 showed that daily losses due to snow cover could be almost 100%, and the monthly loss
was 23%. Konya is located 38°N and 32°E at an elevation 1030 meters above sea level.
Cleaned modules were compared with snow covered modules of the same type and situated
at the same place, as shown in Figure 51.
Figure 51: The system in Konya, Turkey, where one string was cleaned from snow every
day at 09:00 AM [165].
5.7.5 Corrective Mitigation Strategies to Reduce the Impact of Snow Accumula-

tions
This section considers corrective mitigation only, in terms of measures that are applied once
the plant is built. Preventive mitigation measures (i.e., coatings, system design, elevation, etc.)
have been presented in the Task 13 Report on soiling [31].
A. Cleaning or no cleaning?
In most places in Sweden the installer recommendation is to not actively remove snow from
modules since often the guarantee is not valid after mechanical actions on the modules. In
addition, most installations are residential and are to be found in the southern half of Sweden
where production losses due to snow most often are small. Also risk for human injury must be
accounted for, since private persons are not used to work on roofs. However, at locations in
the northern half of Sweden there is in March strong sun and snow simultaneously that can
lead to up to 10% annual losses due to snow.
Exceptions where cleaning is advised often is for flat roofs, see following sections. When plan-
ning for PV system on flat roofs it is recommended to leave room between rows for persons to
be able to move. The roof shall be dimensioned to withstand the load of both snow and panels.
The calculation for acceptable roof load shall be made by the supplier before installation.
B. Active cleaning methods such as brushing
Removing snow mechanically via brushing or other types of active cleaning can sometimes be
necessary if the weight threatens to damage roof structures. However, there is also a risk of
130
damaging the modules or scratching the glass while cleaning from snow, especially if there is
ice build-up on the surface (Figure 52). There is also the potential issue of needing to find an
appropriate place to dispose of the snow, especially in urban areas [166].
For roof mounts, care must also be taken to minimize risk of human injury. If really needed,
snow can be removed with a soft brush as shown in Figure 53. Only loose snow should be
brushed off. It is not advised to try scraping off frozen snow or ice, which can damage the
surface of modules and reduce their efficiency.
Figure 52: Careless removal of snow may destroy the surface of PV modules [166].
Figure 53: Brushing to remove snow from solar roofs [167].
C. Cleaning a buffer zone

Another option is to clean snow from the roof area immediately below the modules to enable
snow sliding off the modules. This approach prevents snow accumulation below the panels
and on the lower parts of the modules, which can lead to frame damage.
131
D. Reverse current melting

A Norwegian study [168] investigated the potential and technical limitations of applying a con-
trolled forward bias (10-19 A) on PV modules to melt snow. In this study, the forward biased
PV modules’ abilities to melt snow was assessed both in a climate chamber as well as in an
operating rooftop PV system covered with snow. The main purpose of the PV system investi-
gated [168] is to mitigate snow from under-dimensioned roofs before, during, or after heavy
snowfalls, and to keep the roof from collapsing from snow loads.
Calculations of the energy and economic payback times were made using assumptions of
Norwegian power tariffs and taxes. The authors claim it is possible to recoup both energy and
economic investment in snow mitigation measures, and that it was possible to earn money by
removing snow to increase PV production. Actively melting snow does, however, require deli-
cate peak load control and utilization of weather forecasts, to prevent that power used to miti-
gate snow does not increase the monthly peak load of the facility or would require more energy
than the avoided energy losses.
E. Value of cleaning
Simple cost savings estimates for cleaning snow from a standard system, assuming a 10%
annual loss from snow coverage in Sweden, are done as an example. Assuming an annual
energy yield of 1000 kWh/kW, a 10% annual loss corresponds to a yearly reduction of
100 kWh/kW due to unremoved snow. Assuming an electricity value of 0.10-0.15 EUR/kWh,
this lost production has a value of 10-15 EUR/kW. This figure can be compared with the cost
of snow removal during the winter season to find out if there is any profitability for snow re-
moval.
5.7.6 Standardisation and Test Methods

According to test standard IEC 61215 [147], modules are tested for mechanical loads ML up
to 5.4 kPa (kPa = kN/m2). The load is evenly distributed and must last for one hour. Snow
loads, on the other hand, can last for weeks or months and can be unevenly distributed over
the module [166].
Despite the stringent requirement of modules passing the above 5.4 kPa ML tests, frame
breakages due to snow on pitched roofs happens in markets known for long winters, such as
Europe, the USA and Japan. In order to better simulate the mechanical stress of snow settling
on pitched residential installations, TÜV Rheinland has designed the inhomogeneous mechan-
ical load (IML) test, which is also referred to as non-uniform snow load test [169]. The test
begins with 240 hours of humidity freeze test [147] on the module, to simulate the freezing
conditions. Then a carefully designed set of weights is placed on top of the test module in-
stalled at a 37° angle. Most of the weight is located near the bottom frame, with lesser weights
toward the top of the module. The result is an uneven distribution of weight spread across the
bottom two-thirds of the test module. This installation simulates the increased load exerted by
settled snow around the eaves, as shown in Figure 54.
132
A standard module has successfully passed the 6.000 kPa IML test with less than 5% power
degradation. A closer inspection of the force diagram shows the 6.000 kPa IML force is the
vector sum of a 4.792 kPa perpendicular force and a 3.611 kPa horizontal force. The horizontal
force directly pushes against the weakest points of the frame, as a pile of snow pulled down
by gravitational force would. By adding a 1.5 safety factor, the module is then certified to with-
stand 4.000 kPa of non-uniform snow load, a pressure that simulates around 50 cm of settled
snow on a pitched roof.
Figure 54: Left) IML test setup [169] using even weights on a pitched module installation.
Right) A simplified force diagram that simulates the effects of settled snow.
In EN 1991-1-13 Eurocode 1 [151], the European Ground Snow Load Map is defined in EN
1991-1-3:2003, 2005 for mapping the European ground snow loads as shown in Figure 55.
Different distributions are adopted for the statistical analysis of extreme snow loads. The most
widely used distribution, the Gumbel distribution, is implemented for the determination of basic
ground snow loads in the development of EN 1991-1-3 (EN 1991-1-3:2003, 2005), Canadian
Building Code (NBCC, 2010), Chinese Building Code (GB-50009, 2012), Architectural Institute
of Japan (AIJ) and the recommendations (Architectural Institute of Japan, 2015), while the Log-
Normal distribution was adopted in ASCE7–10 (American Society of Civil Engineers, 2013).
There are ground snow load maps, “how to” do in places with exceptional snow loads and
conversion factors from ground to roof loads. The snow load on the roof is derived from the
snow load on the ground, multiplying by conversion factors (shape, thermal and exposure co-
efficients). The shape of the snow has importance for the load on a roof, as the snow load is
often caused by snow drift.
The characteristic ground snow loads (Sk) in Eurocode 1 [151] are given by the national Annex
for each country in CEN, the European standardization body for the non-electrical issues.
Regional maps for snow load are published in Annex C of EN 1991-1-3. In addition, snow loads
for various altitudes are given.
For locations where exceptional loads may occur the ground snow load may be treated as
accidental action with the value AAd according to Eq. (18), where Cesl = 2 is often recom-
mended.
AAd = Cesl ∗ Sk (18)
133
Figure 55: Eurocode 1, showing the European Ground Snow Load Map defined in EN
1991-1-3:2003, 2005 [151].
The Swedish National Board of Housing, Building and Planning (Boverket) construction rules
uses a national recommendation for constructing buildings from the Swedish Eurocode 1: Part
1-3 [151]. In Figure 56, a map shows the snow loads on the ground in Sweden with a probability
of 0.98, i.e., once in 50 years. The snow loads north of 60N in Sweden reach 2.5-5.5 kN/m2.
The highest values are reached in the mountain boundaries to Norway in the west. The values
in Eurocode 1 come from a very thorough investigation using 148 measurement stations in
Europe and 40 measurement stations in Sweden [170].
134
Figure 56: Snow load zones north of 60°N in Sweden, where it is most important to con-
sider snow loads when constructing a building. The numbers given in this map range
between 2.5 kN/m2 and 5.5 kN/m2. They can be used to dimension for extra roof loads
due to snow when constructing a building in the regions described in the Swedish ap-
plication of Eurocode 1 [171].
Monitoring the solar PV system can help operators to become more aware of the PV
system's performance. This can offer information about energy generation and about possible
damage to the PV solar system. Monitoring and follow-up also help to avoid downtime and
may give an increase in production yield.
In IEC 61724-1 Photovoltaic system performance – Part 1: Monitoring [3], the soiling ratio SR
is suggested to be estimated by comparing performance ratio PR with soil and PR without soil
using Eq. (19) and definitions of PR as explained in section 2.1.2. In [3], it is suggested that
snow is a type of soil that can be measured the same way as other types of soil. If one
measures instantaneous values, it is suggested to integrate to get daily averages or monthly
averages. After calibration and cleaning and after any significant rainfall SR is close to one
and subsequently soiling loss SL=1-SR is close to zero.
PR(as it is) (19)
SR =
PR(after cleaning)
135
CONCLUSIONS
Comprehensive and standardized guidelines for O&M programmes have been difficult to de-
velop. Increasing adoption of PV systems in different climate zones and conditions worldwide
has indicated that varied stress factors such as temperature, humidity, exposure to UV light,
rain, and wind contribute to the occurrence of PV module and system failures. In view of this,
O&M operators have been developing tailored O&M services to the climate zone in which the
respective plants are located. With this gap in mind, this report aims to provide comprehensive
guidance for customized O&M services in seven different climate zones.
The strategy for selecting a monitoring solution depends mainly on contractual agreements
and it is directly related to the scale of the PV plant. Different international standards and pub-
lications provide guidelines on minimum requirements for different monitoring needs as sum-
marized in this report. Smart PV performance monitoring solutions are indispensable, espe-
cially for large utility-scale PV plants today. A monitoring solution providing smart alarms with
aggregation functionalities and possessing a strong simulation model or digital twin of the plant
as a basis of comparison are key elements in providing business intelligence for performance
optimization.
Large utility-scale PV plants require advanced functionalities in asset management and moni-
toring solutions allowing for fast reaction times and performance optimization. Fault detection
and diagnosis capabilities, providing actionable recommendations and insights for the opera-
tors and asset managers are key to staying competitive.
The solar monitoring industry is moving towards “predictive” monitoring solutions. However,
due to the high complexity of communication between devices and due to limited standardiza-
tion, these capabilities are not yet fully deployed on an industrial scale and require further
development, validation, and demonstration. A “predictive” smart solar monitoring system will
provide insights on the optimal replacement cycle, including cost per time unit under the opti-
mal age-based replacement policy information, which provides highly valuable information for
an operator or asset manager.
While it can be assumed that grid compliance has defined the plants’ characteristics during the
design, construction, and commissioning phase, it is very likely that grid compliance require-
ments will change over the lifetime of the installation, especially given the expected rapid ex-
pansion of renewable energy and foreseen changes to the grid. It is up to the O&M provider to
plan and adapt accordingly.
The forecast of PV power is essential for trading PV electricity on the day-ahead or intraday
electricity markets. It is particularly important for ensuring grid stability and for the bankability
of PV plant projects without fixed feed-in tariffs. Furthermore, PV power forecasting enables
the deployment and dispatching of conventional resources with their given ramp-up time.
Therefore, PV power forecasting is particularly beneficial when combined with energy man-
agement/storage, and for microgrid applications in general.
136
The essential characteristics of PV power forecasts include the forecast horizon, the spatial
and temporal resolution, and the update frequency, whereby the requirements vary depending
on the intended application, data availability, etc. Basically, the forecast horizons are typically
up to 48 hours, the temporal resolution is 15 minutes to one hour. Satellite-based cloud motion
forecasts can improve PV power forecasts for several hours in advance, but so far there are
only a comparatively small number of companies worldwide offering satellite-based short-term
forecasts with good quality.
One of the factors behind the success of photovoltaics is its lower maintenance requirements
compared to other types of electricity generation. However, low maintenance does not mean
that no maintenance is required. While the PV industry has made great strides in detecting
faults and even allows some faults to be fixed remotely, the safe and reliable operation of PV
systems still requires scheduled and unscheduled maintenance for which personnel must be
on site. Although most of the safety aspects are considered during the design and construction
phases of PV plants, there are some elements that need to be considered during operations,
most of which can be planned or mitigated:
• The safety systems and procedures in place.

• Well-trained personnel with the appropriate qualifications for the tasks to be per-
formed.
• The availability of appropriate equipment to carry out maintenance tasks: PPE, con-
sumables as well as durable maintenance tools.
• Site-specific risks to be considered, such as height (PV on buildings), presence of
water (PV on water bodies) or increased fire risks; and,
• Weather and site conditions for site visits.
Both the human as well as the organisational aspects, such as the implementation of the con-
trol hierarchy and the development of PV systems in accordance with ISO 45001, IEC 63049,
ISO 9001 or ANSI Z10, are central to ensuring that the PV power plant and the personnel who
maintain it can operate safely.
Various field experiences have guided the formulation of recommendations and guidelines for
O&M services. These have raised awareness of the fact that nature may have a significant
impact on the performance and health of PV modules and systems. Recommendations for
O&M work in all climates include:
• On-site assessment of vegetation, wildlife, and livestock.

• Mowing grass means checking the condition of the solar PV modules for the pos-
sible need for cleaning or possible damage.
• The industrial environment may lead to unexpected deterioration of the solar mod-
ules.
• Special attention must be paid when selecting cleaning products for PV modules.
It is advisable to follow the recommendations of experts.
137
The global standard O&M procedures discussed in this report also apply to hot and dry cli-
mates, although some aspects such as wildlife risk assessment and appropriate planning for
visits to typically remote sites (hydration, poisoning control procedures, PPE, travel to and from
sites) deserve special attention. Wildlife risks include poisonous animals and insects that can
directly harm humans, while nesting insects and animals can cause short circuits or arcing.
The typical remoteness of PV sites in hot and dry climates comes with significant travel and
preparation requirements, as not all equipment and spare parts can be readily sourced, while
personnel may be at risk of any injuries being exacerbated at non-remote sites due to distance
and the longer time needed for getting medical attention.
To prevent various damages from occurring to PV modules/systems due to extreme weather

events, including strong tropical cyclones, international, regional, and national stand-
ards/codes/recommendations have been discussed and endorsed. Therefore, stakeholders
should comply with these guidelines for proper operation and maintenance of their PV plants,
regardless of the size of the system.
In summary, a combination of well-designed O&M specifications, proactive monitoring systems

and a flexible and tailored O&M regime that considers both possible weather impacts on sys-
tems as well as possible changes to grid requirements are good practices to ensure that PV
systems perform according to or even beyond expected lifespans. Reducing risks by ensuring
that personnel are trained and equipped for O&M operations, as well as employing PV fore-
casting to reduce possible downtimes, also contribute to maintaining PV plant performance
according to specifications.
138
REFERENCES
[1] M. Köntges, G. Oreski, U. Jahn, M. Herz, P. Hacke, K. A. Weiss, G. Razongles, M. Paggi, D. Parlevliet, T. Tanahashi and
R. H. French, Assessment of Photovoltaic Module Failures in the Field, Report IEA-PVPS T13-09: ISBN 978-3-906042-
54-1, 2017.
[2] W. Köppen. Translated by E. Volken.; S. Brönnimann, “Die Wärmezonen der Erde, nach der Dauer der heissen,
gemässigten und kalten Zeit und nach der Wirkung der Wärme auf die organische Welt betrachtet [The thermal zones of
the earth according to the duration of hot, moderate and cold periods],” Meteorologische Zeitschrift, vol. 20, no. 3, pp.
351-360, 2011.
[3] International Electrotechnical Commission, “IEC 61724-1:2017 Photovoltaic system performance - Part 1: Monitoring,”
Geneva, Switzerland, 2017.
[4] B. Herteleer, G. Dickeson, L. McLeod, B. van Ree, C. Paynter, D. Airen, P. Maker, S. Latz, A. Dobb and L. Frearson,
“Visions from the Future: The Interaction between Curtailment, Spinning Reserve Settings and Generator Limits on
Australian Projects with Medium to High Renewable Energy Fractions,” in 35th European Photovoltaic Solar Energy
Conference and Exhibition, Brussels, Belgium, 2018.
[5] S. Gonzalez, J. Neely and M. Ropp, “Effect of non-unity power factor operation in photovoltaic inverters employing grid
support functions,” in IEEE 40th Photovoltaic Specialists Conference, Denver, CO, USA, 2014.
[6] J. Johnson, R. Bründlinger, C. Urrego and R. Alonso, “Collaborative Development of Automated Advanced
Interoperability Certification Test Protocols for PV Smart Grid Integration,” in 29th European Solar Energy Conference
and Exhibition, Amsterdam, Netherlands, 2014.
[7] J. T. Johnson, A. Summers, R. Darbali-Zamora, C. Hansen, M. J. Reno, A. Castillo, S. Gonzalez, J. Hernandez-Alvidrez,

N. S. Gurule, B. Xie, C. Zhong, P. Meliopooulos and C. R. Showalter, “Optimal Distribution System Voltage Regulation
using State Estimation and DER Grid-Support Functions,” 2020.
[8] E. Garoudja, F. Harrou, Y. Sun, K. Kara, A. Chouder and S. Silvestre, “Statistical fault detection in photovoltaic systems,”
Solar Energy, vol. 150, pp. 485-499, 2017.
[9] International Electrotechnical Committee, “IEC TS 63019: Photovoltaic power systems (PVPS) - Information model for
availability,” Geneva, Switzerland, 2019.
[10] International Electrotechnical Commission, “IEC TS 61724-3:2016 Photovoltaic system performance - Part 3: Energy
evaluation method,” Geneva, Switzerland, 2016.
[11] SolarPowerEurope, “Operation & Maintenance Best Practice Guidelines Version 4.0,” 2019.
[12] G. T. Klise and J. R. Balfour, “A Best Practice for Developing Availability Guarantee Language in Photovoltaic (PV) O&M
Agreements,” 2015.
[13] International Renewable Energy Agency and Terrawatt Initiative, “ IRENA PV contract templates,” 2019. [Online].
Available: https://opensolarcontracts.org/. [Accessed 15 April 2020].
[14] C. Tjengdrawira, M. Richter and I. T. Theologitis, “Analyses of Technical Assumptions in PV Electricity Cost,” 27 07 2016.
[Online]. Available:
139
http://www.solarbankability.org/fileadmin/sites/www/files/documents/649997_SolarBankability_D3.1_v1.0p_20160727.pdf.
[Accessed 14 July 2020].
[15] Queensland Government, “Electrical Safety (Solar Farms) Amendment Regulation 2019,” 2019. [Online]. Available:
https://www.legislation.qld.gov.au/view/html/asmade/sl-2019-0046. [Accessed 14 July 2020].
[16] Supreme Court of Queensland, “Maryrorough Solar Pty Ltd v The State of Queensland,” 29 May 2019. [Online].
Available: https://archive.sclqld.org.au/qjudgment/2019/QSC19-135.pdf. [Accessed 14 July 2020].
[17] B. Herteleer, A. Dobb, O. Boyd, S. Rodgers and L. Frearson, “Identifying risks, costs, and lessons from ARENA‐funded
off‐grid renewable energy projects in regional Australia,” Progress in Photovoltaics, vol. 26, pp. 642-650, 2018.
[18] L. McLeod, G. Dickeson, C. Paynter, B. Herteleer, L. Frearson, M. Tuckwell, M. Miller and D. Scheltus, “Lessons from
Large-Scale Solar in Australia,” in 36th European Photovoltaic Solar Energy Conference and Exhibition, Marseille,
France, 2019.
[19] M. Richter, C. Tjengdrawira, J. Vedde, L. Frearson, B. Herteleer, M. Green, B. Stridh, U. Jahn, M. Herz and M. Köntges,
Technical Assumptions Used in PV Financial Models - Review of Current Practices and Recommendations, Report IEA-
PVPS T13-08: ISBN 978-3-906042-46-6, 2017.
[20] International Electrotechnical Commission, “IEC 62053-22 Electricity metering equipment - Particular requirements - Part
22: Static meters for AC active energy (classes 0,1S, 0,2S and 0,5S),” IEC, Geneva, Switzerland, 2020.
[21] International Electrotechnical Commission, “IEC 62053-21:2020 Electricity metering equipment - Particular requirements -
Part 21: Static meters for AC active energy (classes 0,5, 1 and 2),” IEC, Geneva, Switzerland, 2020.
[22] International Electrotechnical Commission, “IEC 61557-12:2018 Electrical safety in low voltage distribution systems up to
1 000 V AC and 1 500 V DC - Equipment for testing, measuring or monitoring of protective measures - Part 12: Power
metering and monitoring devices (PMD),” IEC, Geneva, , Switzerland, 2018.
[23] M. Green, E. Brill, B. Jones and J. Dore, Improving Efficiency of PV Systems Using Statistical Performance Monitoring,
Report IEA-PVPS T13-07: ISBN 978-3-906042-48-0, 2017.
[24] S. Alliance, “Best Practices in Solar Performance Monitoring Version 1.0,” SunsPec Alliance, 2014.
[25] M. Richter, K. D. Brabandere, J. Kalisch, T. Schmidt and E. Lorenz, “Best Practice Guide on Uncertainty in PV Modelling,”
Performance Plus, 2015.
[26] T. Dierauf, A. Growitz, S. Kurtz, J. L. B. Cruz, E. Riley and C. Hansen, “Weather-Corrected Performance Ratio,” National
Renewable Energy Laboratory, Golden, Colorado, 2013.
[27] International Electrotechnical Commission, “IEC TS 62446-3:2017 Photovoltaic (PV) systems - Requirements for testing,
documentation and maintenance - Part 3: Photovoltaic modules and plants - Outdoor infrared thermography,” IEC,
[28] M. Köntges, S. Kurtz, C. Packard, U. Jahn, K. A. Berger, K. Kato, T. Friesen, H. Liu, M. Van Iseghem, J. Wohlgemuth, D.
Miller, M. Kempe, P. Hacke, F. Reil, N. Bogdanski, W. Herrmann, C. Buerhop Lutz and G. Friesen, Review of Failures of
Photovoltaic Modules, Report IEA-PVPS T13-01: ISBN 978-3-906042-16-9, 2014.
[29] W. Herrmann, B. Kubicek, G. Friesen, B. Farnung, M. Köntges, A. Morlier, G. Eder, J. Vedde, D. Parlevliet, I. Tsanakas,
M. Aghaei, L. Haitao, A. Astigarraga, L. S. Bruckman, R. French, A. M. Karimi, R. Wieser, J. Lin and E. Fleiß, Good
140
Practice Recommendations to Qualify PV Power Plants using Mobile Devices, Report IEA-PVPS T13-24: ISBN 978-3-
907281-12-3, 2021.
[30] U. Jahn, M. Herz, M. Köntges, D. Parlevliet, M. Paggi, I. Tsanakas, J. Stein, K. Berger, S. Ranta, R. French, M. Richter
and T. Tanahashi, Review on Infrared and Electroluminescence Imaging for PV Field Applications, Report IEA-PVPS
T13-10: ISBN 978-3-906042-53-4, 2018.
[31] C. Schill, D. Parlevliet, A. Anderson, B. Stridh, L. Burnham, C. Baldus-Jeursen, L. Micheli, E. Urrejola, E. Whitney and G.
Mathiak, Soiling Losses – Impact on the Performance of Photovoltaic Power Plants, Report IEA-PVPS T13-22: ISBN 978-
3-907281-09-3, 2022.
[32] A. Woyte, V. Van Thong, R. Belmans and J. Nijs, “Voltage fluctuations on distribution level introduced by photovoltaic
systems,” IEEE Transactions on Energy Conversion, vol. 21, no. 1, pp. 202-209, 2006.
[33] A. Woyte, R. Belmans and J. Nijs, “Testing the islanding protection function of photovoltaic inverters,” IEEE Transactions
on Energy Conversion, vol. 18, no. 1, pp. 157-162, 2003.
[34] “Is the distribution grid ready to accept large-scale photovoltaic deployment? State of the art, progress, and future
prospects,” Progress in Photovoltaics: Research and Applications, vol. 20, no. 6, pp. 681-697, 2012.
[35] N. Stringer, A. Bruce and I. MacGill, “Review of voltage management approaches and network regulation: who pays and
how?,” in Asia-Pacific Solar Research Conference, Sydney, Dec 2018.
[36] B. Herteleer, “Visions from the Future: The Interaction between Curtailment, Spinning Reserve Settings and Generator
Limits on Australian Projects with Medium to High Renewable Energy Fractions,” in 35th European Photovoltaic Solar
Energy Conference and Exhibition, Brussels, Belgium, 2018.
[37] Synergrid, “Specific Technical Prescriptions Regarding Power-Generating Plants Operating in Parallel to the Distribution
Network (English translation),” Synergrid, Brussels, 2019.
[38] VDE, “Technical requirements for the connection and operation of customer installations to the medium voltage network
(VDE-AR-N-4110),” VDE, Berlin, 2018.
[39] VDE, “Technical requirements for the connection and operation of customer installations to the high voltage network
(VDE-AR-N 4120),” 19 October 2018. [Online]. Available: https://www.vde.com/en/fnn/topics/technical-connection-
rules/tar-for-high-voltage. [Accessed 26 August 2020].
[40] European Commission, “Commission Regulation (EU) 2016/331,” Official Journal of the European Union, p. 68, 2016.
[41] CENELEC, “Requirements for generating plants to be connected in parallel with distribution networks - Part 1: Connection
to a LV distribution network - Generating plants up to and including Type B,” 01 August 2019. [Online]. Available:
https://www.cenelec.eu/dyn/www/f?p=104:110:1399530632340401::::FSP_PROJECT,FSP_LANG_ID:63319,25.
[Accessed 27 August 2020].
[42] CENELEC, “Requirements for generating plants to be connected in parallel with distribution networks - Part 2: Connection
to a MV distribution network - Generating plants up to and including Type B,” 01 August 2019. [Online]. Available:
https://www.cenelec.eu/dyn/www/f?p=104:110:1399530632340401::::FSP_PROJECT,FSP_LANG_ID:63321,25.
141
[43] IEEE, “1547-2018 - IEEE Standard for Interconnection and Interoperability of Distributed Energy Resources with
Associated Electric Power Systems Interfaces,” 06 April 2018. [Online]. Available:
https://standards.ieee.org/standard/1547-2018.html. [Accessed 27 August 2020].
[44] AEMO, “2020 Integrated System Plan,” 30 July 2020. [Online]. Available: https://aemo.com.au/energy-systems/major-
publications/integrated-system-plan-isp/2020-integrated-system-plan-isp. [Accessed 26 August 2020].
[45] Australian Energy Market Commission, “National Electricity Rules - Current,” 27 August 2020. [Online]. Available:
https://www.aemc.gov.au/regulation/energy-rules/national-electricity-rules/current, pp. 467-469. [Accessed 27 August
2020].
[46] M. Sengupta, A. Habte, S. Wilbert, C. Gueymard and J. Remund, Solar Resource for High Penetration and Large Scale
Applications, Report IEA-PVPS 16-04:2021: ISBN 978-3-907281-19-2, 2021.
[47] E. Lorenz, J. Ruiz-Arias and W. S., Forecasting Solar Radiation. Chapter in: "Best Practices Handbook for the Collection
and Use of Solar Resource Data for Solar Energy Applications, (Second Edition): Edited by M. Sengupta, A. Habte, C.
Gueymard, S. Wilbert, D. Renné, T. Stoffel. Technical Report NREL/TP-5D00-68886., 2017.
[48] A. Betti, M. Pierro, C. Cornaro, D. Moser, M. Moschella, E. Collino, D. Ronzio, D. van der Meer, L. Visser, J. Widen, T.
AlSkaif and W. G. J. H. M. van Sark, Regional Solar Power Forecasting 2020, Report IEA-PVPS T16-01: 2020: ISBN
978-3-906042-88-6, 2020.
[49] E. Lorenz, Solar Irradiance Forecasting for System Integration of Solar Energy, Habilitation Thesis: Oldenburg University,
Germany , 2018.
[50] M. Diagne, M. David, P. Lauret, J. Boland and N. Schmutz, “Review of solar irradiance forecasting methods and a
propositionfor small-scale insular grids,” Renewable and Sustainable Energy Reviews, vol. 27, pp. 65-76, 2013.
[51] J. Antonanzas, N. Osorio, R. Escobar, R. Urraca, F. J. Martinez-de-Pison and F. Antonanzas-Torres, “Review of

photovoltaic power forecasting,” Solar Energy, vol. 136, pp. 78-111, 2016.
[52] S. Sobri, S. Koohi-Kamali and N. Abd. Rahim, “Solar photovoltaic generation forecasting methods: A review,” Energy
Conversion and Management, vol. 156, pp. 459-497, 2018.
[53] U. Das, K. Tey, M. Seyedmahmoudian, S. Mekhilef, M. Idris, W. Van Deventer, V. Horan and A. Stojcevski, “Forecasting
of photovoltaic power generation and model optimization: A review,” Renewable and Sustainable Energy Reviews, vol.
81, pp. 912-928, 2018.
[54] J. Zhang, A. Florita, B.-M. Hodge, S. Lu, H. Hamann, V. Banunarayanan and A. Brockway, “A suite of metrics for
assessing the performance of solar power forecasting,” Solar Energy, vol. 111, pp. 157-175, 2015.
[55] D. Yang, S. Alessandrini, J. Antonanzas, F. Antonanzas-Torres, V. Badescu, H. G. Beyer, R. Blaga, J. Boland, J. M.

Bright, C. Coimbra, M. David, Â. Friman, C. A. Gueymard, T. Hong, M. Kay, S. Killinger, J. Kleissl, P. Lauret, E. Lorenz
and D. van der Meer, “M. Paulescu, R. Perez, O. Perpiñán Lamigueiro, I. Peters, G. Reikard, D. Renné, Y. Saint-Drenan,
Y. Shuai, R. Urraca, H. Verbois, F. Vignola, C. Voyant and J. Zhang, “Verification of deterministic solar forecasts,” Solar
Energy (in press), 2020.
[56] P. M. Perez R., M. Pierro, J. Schlemmer, S. Kivalov, J. Dise, P. Keelin, M. Grammatico, A. Swierc, J. Ferreira, A. Foster,
M. Putnam and T. Hoff, “Operationally Perfect Solar Power Forecasts: A Scalable Strategy to Lowest-Cost Firm Solar
Power Generation,” in Proc. 46th IEEE PV Specialists Conference, Chicago, Il, USA, 2019.
142
[57] E. Lorenz, J. Kühnert, D. Heinemann, K. P. Nielsen, J. Remund and S. Müller, “Comparison of Global Horizontal
Irradiance Forecasts Based on Numerical Weather Prediction Models with Different Spatio-Temporal Resolutions,”
Progress in Photovoltaics: Research and Applications, vol. 24, no. 12, pp. 1,626–1,640, 2016.
[58] G. Becker, F. Flade, B. Giesler, W. Rehm, B.Schiebelsberger and W. Weber, “20 Jahre erfolgreicher Betrieb der 1-MW
PV-Anlage Messe München,” in Nationales Photovoltaik Symposium Staffelstein 2018, Staffelstein, Germany, 2018.
[59] “ISO 45001:2018 Occupational health and safety management systems — Requirements with guidance for use,”
International Organization for Standardization, Geneva, Switzerland, 2018.
[60] “ANSI/ASSP Z10.0-2019 Occupational Health & Safety Management Systems,” American National Standards Institute,
Washington, DC, USA, 2019.
[61] “ISO 9001:2015 Quality management systems — Requirements,” International Organization for Standardization, Geneva,
Switzerland, 2015.
[62] “IEC TS 63049:2017 Terrestrial photovoltaic (PV) systems - Guidelines for effective quality assurance in PV systems
installation, operation and maintenance,” International Electrotechnical Commission, Geneva, Switzerland, 2017.
[63] “Hierarchy of Controls,” 13 January 2015. [Online]. Available: https://www.cdc.gov/niosh/topics/hierarchy/default.html.

[64] H. L. Floyd, “A Practical Guide for Applying the Hierarchy of Controls to Electrical Hazards,” IEEE Transactions on
Industry Applications, vol. 51, no. 5, pp. 4263-4266, Sep-Oct 2015.
[65] International Electrotechnical Commission, “IEC 61730: Photovoltaic (PV) module safety qualification,” Geneva,
Switzerland, 2016.
[66] CENELEC, “EN 50110-1:2015 Operation of electrical installations Part 1 – General requirements,” European Committee
for Electrotechnical Standardization, Brussels, Belgium, 2013.
[67] “DIN VDE 0105-100 – Betrieb von elektrischen Anlagen,” VDE Verlag, Berlin, Germany, 2015.
[68] S. Namikawa, G. Kinsey, G. Heath, A. Wade, P. Sinha and K.Komoto, Photovoltaics and Firefighters’ Operations: Best
Practices in Selected Countries, Report IEA-PVPS T12-09: ISBN 978-3-906042-60-2, 2017.
[69] S. Pester, C. Holland and C. Coonick, “Fire and Solar PV Systems. Recommendations for fire and rescue services,” BRE
National Solar Centre, Watford, UK, 2017.
[70] S. Pester and S. Woodman, “Fire and Solar PV Systems. Literature Review, Standards and training,” BRE National Solar
Centre, Watford, UK, 2015.
[71] T. Nordmann, L. Clavadetscher, W. G. van Sark and M. Green, Analysis of Long-Term Performance of PV Systems,
Report IEA-PVPS T13-05 : ISBN 978-3-906042-21-3, 2014.
[72] International Electrotechnical Commission, “IEC 60364-4-41:2005: Low voltage electrical installations – Part 4-41:
Protection for safety – Protection against electric shock,” Geneva, Switzerland, 2005.
[73] International Electrotechnical Commission, “IEC 60364-4-42:2010: Low-voltage electrical installations - Part 4-42:
Protection for safety – Protection against thermal effects,” Geneva, Switzerland, 2010.
143
[74] International Electrotechnical Commission, “IEC 60364-4-43:2017, Low voltage electrical installations – Part 4-43:
Protection for safety – Protection against over current,” Geneva, Switzerland, 2017.
[75] International Electrotechnical Commission, “IEC 60364-4-44:2007 Low voltage electrical installations – Part 4-44:
Protection for safety – Protection against voltage disturbances and electromagnetic disturbances,” Geneva, Switzerland,
2007.
[76] International Electrotechnical Commission, “IEC 60364-7-712 Low voltage electrical installations – Part 7-712:
Requirements for special installations - Solar photovoltaic (PV) power supply systems,” Geneva, Switzerland, 2017.
[77] International Electrotechnical Commission, “IEC 63092-1:2019 Photovoltaics in buildings - Part 1: Building integrated
photovoltaic modules,” Geneva, Switzerland, 2019.
[78] International Electrotechnical Commission, “IEC 63092-2:2019 Photovoltaics in buildings - Part 2: Building integrated
photovoltaic systems,” Geneva, Switzerland, 2019.
[79] K. Berger, S. Boddaert, M. Del Buono, A. Fedorova, F. Frontini, S. Inoue, H. Ishii, K. Kapsis, J.-T. Kim, P. Kovacs, M.
Machado and N. Martín Chivelet, Analysis of requirements, specifications and regulation of BIPV, Report IEA-PVPS T15-
08: ISBN 978-3-906042-82-4, 2019.
[80] “CFPA-E Guideline No 37:2018 F, Photovoltaic systems: Recommendations on loss prevention,” CFPA Europe,
Hvidovre, 2018.
[81] Building Insurance Berne (GVB), “Photovoltaic systems in fire brigade use,” Bern, Switzerland, 2016.
[82] N. Enbar, D. Weng and G. Klisse, Budgeting for Solar PV Plant Operations & Maintenance: Practices and Pricing, Electric
Power Research Institute (EPRI) and Sandia National Laboratories: Report ID: SAND2016-0649R, 2016.
[83] K. Ilse, L. MIcheli, B. Figgis, K. Lange, D. Daßler, H. Hanifi, F. Wolfertstetter, V. Naumann, C. Hagendorf, R. Gottschalg
and J. Bagdahn, “Techno-Economic Assessment of Soiling Losses and Mitigation Strategies for Solar Power Generation,”
Joule, vol. 3, pp. 2303-2321, 2019.
[84] M. Maghamia, H. Hizama, C. Gomes, M. Radzi, M. Rezadad and S. Hajighorbani, “Power loss due to soiling on solar
panel: A review,” Renewable and Sustainable Energy Reviews, vol. 59, pp. 1307-1316, 2016.
[85] A. Alcaro, “PV Operations Europe 2019,” 2 March 2019. [Online]. Available: https://www.linkedin.com/pulse/why-person-
working-photovoltaic-should-against-scotch-antonio-alcaro/. [Accessed 28 October 2020].
[86] N. Khadka, B. Adhikari and A. S. A. Bista, “Solar Panel Cleaner Technology: A Review,” in 5th International Conference
on Developments in Renewable Energy Technology (ICDRET'18), Kathmandu, Nepal, 2018.
[87] P. Patil, J. Bagi and M. Wagh, “A review on cleaning mechanism of solar photovoltaic panel,” International Conference on
Energy, Communication, Data Analytics and Soft Computing, pp. 250-256, 2017.
[88] L. J. Walston Jr., K. E. Rollins, K. P. Smith, K. E. LaGory, K. Sinclair, C. Turchi, T. Wendelin and H. Souder, “A Review of
Avian Monitoring and Mitigation Information at Existing UtilityScale Solar Facilities,” SunShot Initiative and Office of
Energy Efficiency & Renewable Energy, 2015.
[89] World Bank Group, ESMAP, SERIS, Where Sun Meets Water: Floating Solar Handbook for Practitioners, Washington,
DC, USA: World Bank, 2019.
144
[90] K. Terio, D. McAloose and J. S. Leger., Pathology of Wildlife and Zoo Animals, Elsevier: https://doi.org/10.1016/C2015-0-
01586-6, 2018.
[91] S. Roth, “The Desert Sun,” 17 12 2018. [Online]. Available:

https://eu.desertsun.com/story/tech/science/energy/2016/08/17/how-many-birds-killed-solar-farms/88868372/. [Accessed
19 08 2020].
[92] M. Grippo, J. W. Hayse and B. L. O’Connor, “Solar Energy Development and Aquatic Ecosystems in the Southwestern
United States: Potential Impacts, Mitigation, and Research Needs,” Environ Assess., vol. 55, no. 10.1007/s00267-014-
0384-x, pp. 244-256, 2015.
[93] J. Scurlock, “Agricultural Good Practice Guidance for Solar Farms,” BRE National Solar Centre, Watford, UK, 2014.
[94] T. Karin, C. B. Jones and A. Jain, “Photovoltaic Climate Zones: The Global Distribution of Climate Stressors Affecting
Photovoltaic Degradation,” in 36th European Photovoltaic Solar Energy Conference and Exhibition, Marseille, 2019.
[95] J. Ascencio-Vásquez, K. Brecl and M. Topic, “Methodology of Köppen-Geiger-Photovoltaic climate classification and
implications to worldwide mapping of PV system performance,” Solar Energy, vol. 191, pp. 672-685, 2019.
[96] M. Kottek, J. Grieser, C. Beck, B. Rudolf and F. Rubel, “World Map of the Köppen-Geiger climate classification updated,”
Meteorologische Zeitschrift, vol. 15, pp. 259-263, 2006.
[97] C. Silva Cruz and J. Calderón Suenzen, Guía Climática Práctica, Santiago de Chile, Chile: Dirección Meteorológica de
Chile, https://es.scribd.com/document/114425344/Guia-Climatica-de-Chile, 2008.
[98] International Electrotechnical Commission, " IEC 61215-1: 2020 Terrestrial Photovoltaic (PV) Modules - Design,
Qualification and Type Approval - Part 1: Test Requirements," Geneva, Switzerland, 2020.
[99] Programa Energía Solar, “Estudio Benchmarking de Plantas Solares Fotovoltaicas en Chile [Benchmarking Study of
Solar Photovoltaic Power Plants in Chile],” Report of the Solar Committee of Chile , 2017.
[100] ASTM G173-03 12, “Standard Tables for Reference Solar Spectral Irradiances: Direct Normal and Hemispherical on 37°
Tilted Surface,” ASTM International, West Conshohocken, PA, USA, www.astm.org, 2012.
[101] A. Marzo, P. Ferrada, F. Beiza, P. Besson, J. Alonso-Montesinos, J. Ballestrín, R. Roman, C. Portillo, R. Escobar and E.
Fuentealba, “Standard or local solar spectrum? Implications for solar technologies studies in the Atacama desert,”
Renewable Energy, vol. 127, no. https://doi.org/10.1016/j.renene2018.05.039, pp. 871-882, 2016.
[102] M. Herz, U. Jahn, D. Moser, S. Lindig, K. Berger, M. Richter, M. Köntges, G. Friesen, R. French and J. Vedde,
Quantification of Technical Risks during Operation and Maintenance, Report IEA-PVPS T13-23: ISBN 978-3-907281-11-
6, 2021.
[103] World Health Organization, “Global Solar UV Index: A Practical Guide,” 2002. [Online]. Available:
https://www.unep.org/resources/report/global-solar-uv-index-practical-guide. [Accessed 30 October 2022].
[104] Department of Alternative Energy Development and Efficiency, “Ministry of Energy, Thailand, Areas with solar power
potential,” 2020. [Online]. Available: http://weben.dede.go.th/webmax/content/areas-solar-power-potential. [Accessed 07
December 2020].
[105] N. Sakarapunthip, D. Chenvidhya, S. Chuangchote, K. Kirtikara, T. Chenvidhya and W. Onreabroy, “Effects of dust
accumulation and module cleaning on performance ratio of solar rooftop system and solar power plants,” Japanese
Journal of Applied Physics, vol. 56, no. 08ME02, 2017.
145
[106] International Electrotechnical Commission, “IEC 62892:2019 Extended thermal cycling of PV modules - Test procedure,”
[107] N. Shiradkar, “Reliability and Safety Issues Observed in Flood Affected PV Power Plants and Strategies to Mitigate the
Damage in Future,” in 46th IEEE Photovoltaic Specialists Conference, Chicago, IL, USA, 2019.
[108] Intergovernmental Panel on Climate Change, “Global Warming of 1.5°C,” A Special Report by IPCC, 2018.
[109] G. Simpkins, “Extreme Rain in India,” Nature Climate Change, vol. 7, no. 11, p. 760, 2017.
[110] Thomson Reuters Foundation, “How Climate Change Has Affected Gujarat – More Floods,” 8 September 2017. [Online].
Available: https://everylifecounts.ndtv.com/15-flood-prone-gujarat-villages-relocate-new-schools-homes-16468.
[Accessed 1 June 2019].
[111] H. Upadhyaya, “Kerala floods: What to expect when none of the 61 dams have any emergency plan?,” 18 September
2018. [Online]. Available: https://www.downtoearth.org.in/news/natural-disasters/kerala-floods-what-to-expect-when-
none-of-the-61-dams-have-any-emergency-plan--61416. [Accessed 1 June 1 2019].
[112] Climate Central, “Climate Change Increasing Frequency of Coastal Flooding,” 4 October 2018. [Online]. Available:
https://www.climatecentral.org/gallery/graphics/climate-change-increasing-frequency-of-coastal-flooding. [Accessed 1
June 2019].
[113] S. Chattopadhyay, R. Dubey, V. Kuthanazhi, S. Zachariah, S. Bhaduri, C. Mahapatra, S. Rambabu, F. Ansari, A.

Chindarkar, A. Sinha, H. K. Singh, N. Shiradkar, B. M. Arora, A. Kottantharayil, K. Narasimhan, S. Sabnis and J. Vasi, “All
India Survey of Photovoltaic Module Reliability 2016,” http://www.ncpre.iitb.ac.in/research/reports.html, Mumbai, India,
2016.
[114] International Finance Corporation World Bank Group, “Utility-Scale Solar Photovoltaic Power Plants,” International
Finance Corporation IFS), 2015.
[115] Ministry of Economy, Trade and Industry (Japan), Features of accidental damages at PV facilities in 2018 summer
season (in Japanese), Tokyo, Japan, 2018.
[116] Ministry of Economy, Trade and Industry (Japan), Current situation and countermeasures on the penetration of PV
facilities in Japan (in Japanese), Tokyo, Japan, 2019.
[117] E. Bellini, “Japan’s largest floating PV plant catches fire after typhoon Faxai impact,” PV Magazine, 09 09 2019.
[118] Ministry of Economy, Trade and Industry (Japan), Report on the disruptive accident in Yamakura Floating PV Plant (in
Japanese), Tokyo, Japan, 2020.
[119] E. Hotchkiss, “Observations of PV systems post-hurricane,” in Proc. 2018 PV Reliab. Work, pp. 747-769, 2018.
[120] C. Burgess and J. Goodman, Solar under storm: Select best practices for resilient ground-mount PV systems with
hurricane exposure, Basalt, CO, USA: Rocky Mountain Institute, 2018.
[121] Federal Emergency Management Agency, Hurricanes Irma and Maria in the U.S. “Virgin islands: Building performance
observations, recommendations, and technical guidance, P-2021, Washington DC, USA: FEMA, 2018.
[122] Federal Emergency Management Agency, Rooftop solar panel attachment: Design, installation, and maintenance, USVI-
RA5, Washington DC, USA: FEMA, 2018.
146
[123] S. Donati, “Storms in Italy’s Dolomites rase centuries–old forests,” 05 November 2018. [Online]. Available:
https://www.italymagazine.com/news/storms-italys-dolomites-raze-centuries-old-forests. [Accessed 18 August 2020].
[124] Wikipedia, “Wikipedia, Storm Adrian (“Tempesta Vaia” in Italian language),” 2020. [Online]. Available:
https://en.wikipedia.org/wiki/Storm_Adrian. [Accessed 28 October 2020].
[125] L. M. Fedel, “Meteo Altopiano di Piné, Miola, Italy, The Piné plateau (L'ALTOPIANO DI PINÉ),” 2018. [Online]. Available:
http://meteopine.altervista.org/pagina-altopianopine.php. [Accessed 18 08 2020].
[126] American Society of Civil Engineers, Minimum design loads and associated criteria for buildings and other structures,
ASCE 7-16, 2017.
[127] Structural Engineers Association of California (SEAOC), “SEAOC PV2-2017 Wind design for solar arrays,” Sacramento,
CA, USA, 2017.
[128] FM global property loss prevention data sheet: ground-mounted solar photovoltaic power, Johnston, RI, Data Sheet 7-
106, FM Global, 2014.
[129] Japanese Industrial Standard, “ JIS C 8955 Load design guide on structures for photovoltaic array,” 2017.
[130] Japan Photovoltaic Energy Association (JPEA) and Okuji, Reviewing Committee on the Safety Design of PV Facilities,
“Design guidelines for terrestrial PV facilities 2019” (in Japanese), New Energy and Industrial Technology Development
Organization (NEDO), 2019.
[131] European Commission, “Eurocodes, Brussels, Belgium,” [Online]. Available:

https://ec.europa.eu/growth/sectors/construction/eurocodes_en. [Accessed 19 08 2020].
[132] European Commission Joint Research Centre, “Eurocodes - Background and Applications: “Dissemination of Information
for Training” workshop, Brussels, Belgium,” [Online]. Available: https://eurocodes.jrc.ec.europa.eu/showpage.php?id=332.
[Accessed 19 08 2020].
[133] S. O. Hansen, “EN 1991-1-4: 2005 Wind actions, Background and Applications: “Dissemination of Information for
Training” workshop, [Online],” in Proc. Eurocodes, 2008.
[134] International Electrotechnical Commission, Technical Committee 82, “ 82/1728/Q Solar Photovoltaic Energy Systems
Proposal to create new working group for PV support structures,” Geneva, Switzerland, 2020.
[135] International Electrotechnical Commission, “IEC 62817:2014+AMD1:2017 Photovoltaic systems - Design qualification of
solar trackers,” Geneva, Switzerland, 2017.
[136] International Electrotechnical Commission , Results of 82/1728/Q, Proposal to create new working group for PV support
structures, Technical Committee 82: Solar Photovoltaic Energy Systems, 82/1752/RQ,, Geneva, Switzerland, 2020.
[137] International Electrotechnical Commission, Technical Committee 82: Solar Photovoltaic Energy Systems Evaluation of
photovoltaic (PV) module to mounting structure interface, Geneva, Switzerland: New Work Item Proposal 82/1740/NP,
2020.
[138] S.-T. Hsu, W.-Y. Lin, and S.-J. Wu, Environmental factors for non-uniform dynamic mechanical load test due to wind
actions on photovoltaic modules, vol. 150, Energy Procedia, Sep. 2018, pp. 50-57.
147
[139] Japan Electrical Safety & Environmental Technology Laboratories, Report on guidelines for periodic inspection and failure
examination of PV power systems, Tokyo, Japan: Study Committee for Monitoring and Inspection Technologies for PV
Power Systems, 2016.
[140] Japan Photovoltaic Energy Association & Japan Electrical Manufacturers’ Association, Guideline on maintenance of PV
systems (in Japanese), JM 19Z001, Japan, 2019.
[141] Photovoltaic Power generation Technology Research Association, Report on the items to maintain the safety in the
modest scale PV facilities (in Japanese), Tokyo, Japan, 2018.
[142] International Electrotechnical Commission, "IEC 62446-1:2016+AMD1 Photovoltaic (PV) systems - Requirements for
testing, documentation and maintenance - Part 1: Grid connected systems - Documentation, commissioning tests and
inspection," Geneva, Switzerland, 2018.
[143] International Electrotechnical Commission, "IEC 62446-2 Photovoltaic (PV) systems – Requirements for testing,
documentation and maintenance – Part 2: Grid connected systems – Maintenance of PV systems," Geneva, Switzerland,
2020.
[144] International Electrotechnical Commission, "IEC TS 63049 Terrestrial photovoltaic (PV) systems - Guideline for increased
confidence in PV system installation," Geneva, Switzerland, 2017.
[145] International Electrotechnical Commission, "IEC TS 62738 Ground-mounted photovoltaic power plants - Design
guidelines and recommendations," Geneva, Switzerland, 2018.
[146] International Electrotechnical Commission, "IEC TS 63157 Guidelines for effective quality assurance of power conversion
equipment for photovoltaic systems," Geneva, Switzerland, 2019.
[147] International Electrotechnical Commission, "IEC 61215-2: 2020 Terrestrial Photovoltaic (PV) Modules - Design
Qualification and Type Approval - Part 2: Test Procedures," Geneva, Switzerland, 2020.
[148] R. E. Pawluk, Y. Chen and Y. She, “Photovoltaic electricity generation loss due to snow – A literature review on influence
factors, estimation, and mitigation,” Renewable and Sustainable Energy Reviews, vol. 107, pp. 171-182, 2019.
[149] T. W. Estilow, A. H. Young and D. A. Robinson, “A long-term Northern Hemisphere snow cover extent data record for
climate studies and monitoring,” Earth Syst. Sci. Data, https://doi.org/10.5194/essd-7-137-2015, vol. 7, pp. 137-142,
2015.
[150] A. Granlund, J. Narvesjö and A. M. Petersson, “The Influence of Module Tilt on Snow Shadowing of Frameless Bifacial
Modules,” in 36th European Photovoltaic Solar Energy Conference and Exhibition, pp.1646-1650, ISBN 3-936338-60-4,
Marseille, France, 2019.
[151] European Committee for Standardization, “EN 1991-1-3. Eurocode 1 - Actions on structures - Part 1-3: General actions -
Snow loads,” CEN, Brussels, Belgium, 2003.
[152] S. Dimova, M. Fuchs, A. P. Vieira, B. N. Kamenarova, M. L. Raposo and S. Iannaccone, “State of implementation of the
Eurocodes in the European Union. Support to the implementation, harmonization and further development of the
Eurocodes,” EUR 27511 EN, ISBN 978-92-79-52706-7, 2015.
[153] W. S. B. Paterson, “The Physics of Glaciers,” Pergamon, ISBN 978-0-08-037944-9, 1994.
[154] C. Pike, E. Whitney, M. Wilber and J. Stein, “Field Performance of South-Facing and East-West Facing Bifacial Modules
in the Arctic,” MDPI Energies, https://doi.org/10.3390/en14041210, vol. 14, p. 1210, 2021.
148
[155] M. W. Rowell, S. G. Daroczi, D. W. J. Harwood and A. Gabor, “The Effect of Laminate Construction and Temperature
Cycling on the Fracture Strength and Performance of Encapsulated Solar Cells,” in 7th World Conference on Photovoltaic
Energy Conversion, Hawaii, USA, 2018.
[156] E. J. Schneller., H. Seigneur, J. Lincoln. and A. M. Gabor, “The Impact of Cold Temperature Exposure in Mechanical
Durability Testing of PV Modules,” in 46th IEEE Photovoltaic Specialists Conference, Chicago, IL, USA, ISBN 978-1-
7281-0494-2, 2019.
[157] H. Seigneur, A. M. Gabor, E. J. Schneller and J. Lincoln, “Electroluminescence Testing Induced Crack Closure in PV
Modules,” in 46th IEEE Photovoltaic Specialists Conference, Chicago, IL, USA, ISBN 978-1-7281-0494-2, 2019.
[158] A. Nash; C. Pike; R. A. Seifert, “A Solar Design Manual for Alaska, Fifth edition,” Alaska Center for Energy and Power
and the Cooperative Extension Service, University of Alaska, Fairbanks, Alaska, USA, 2018.
[159] A. M. Gabor, R. Janoch, A. Anselmo, J. Lincoln, E. J. Schneller, H. Seigneur, D. J. Harwood and M. W. Rowell, “Mounting
Rail Spacers for Improved Solar Panel Durability,” in 46th IEEE Photovoltaic Specialists Conference, Chicago, IL, USA,
ISBN 978-1-7281-0494-2, 2019.
[160] M. v. Noord, T. Berglund and M. Murphy, “Snöpåverkan på solelproduktion - Om snöförluster på takanläggningar i Norra
Sverige,” Energiforsk, Rapport 2017:382, 2017.
[161] L. Powers, J. Newmiller and T. Townsend, “Measuring and modelling the effect of snow on photovoltaic system
performance,” in 35th IEEE Photovoltaic Specialist Conference, Hawaii, USA, ISBN 978-1-4244-5892-9, 2010.
[162] T. Townsend and L. Powers, “Photovoltaics and Snow: An Update from Two Winters of Measurements in the SIERRA,” in
37th IEEE Photovoltaic Specialists Conference, Seattle, WA, USA, ISBN 978-1-4244-9965-6, 2011.
[163] R. W. Andrews, A. Pollard and J. M. Pearce, “The effects of snowfall on solar photovoltaic performance,” Solar Energy,
vol. 92, pp. 84-97, 2013.
[164] G. Becker, B. Schiebelsberger, W. Weber, C. Vodermayer, M. Zehner and G. Kummerle, An approach to the impact of
snow on the yield of grid connected PV systems, Bavarian Association for the Promotion of Solar Energy: Munich,
Germany, 2007.
[165] M. A. Anadol and E. Erhan, “Determination of Power Losses Occurring due to Snowfall based on Grid-Connected Inverter
Data,” in International Advanced Researches and Engineering Congress, Osmaniye, Turkey, 2017.
[166] E. Andenæs, B. P. Jelle, K. Ramlo, T. Kolås, J. Selj and S. E. Foss, “The influence of snow and ice coverage on the
energy generation from photovoltaic solar cells,” Solar Energy, vol. 159, pp. 318-328, 2018.
[167] B. J. Jelle, T. Gao, S. A. Mofid, T. Kolås, P. M. Stenstad and S. Ng, “Avoiding Snow and Ice Formation on Exterior Solar
Cell Surfaces - A Review of Research Pathways and Opportunities,” Procedia Engineering, vol. 145, pp. 699-706, 2016.
[168] B. B. Aarseth, M. B. Øgaard, J. Zhu, T. Strömberg, I. A. Tsanakas, J. H. Selj. and E. S. Marstein, “Mitigating Snow on
Rooftop PV Systems for Higher Energy Yield and Safer Roofs,” in 35th European Photovoltaic Solar Energy Conference
and Exhibition, Brussels, Belgium, 2018.
[169] International Electrotechnical Commission, “IEC 62938:2020, Photovoltaic (PV) modules - Non-uniform snow load
testing,” ISBN 978-2-8322-8074-4, Geneva, Switzerland, 2020.
[170] E. M. Eriksson and R. Taesler, “Analysis and mapping of the basic values of snow loads on ground (In Swedish),”
Boverket, 1995.
149
[171] Boverket, “Boverkets konstruktionsregler, EKS11. BFS 2011:10,” Boverket: The Swedish National Board of Housing,
Building and Planning (English translation), Sweden, 2019.
[172] S. Donati, “Storms in Italy’s Dolomites rase centuries–old forests,” 05 11 2018. [Online]. Available:
https://www.italymagazine.com/news/storms-italys-dolomites-raze-centuries-old-forests. [Accessed 18 08 2020].
[173] “Wikipedia, Storm Adrian (“Tempesta Vaia” in Italian language),” 2020. [Online]. Available:
https://it.wikipedia.org/wiki/Tempesta_Vaia and https://en.wikipedia.org/wiki/Storm_Adrian. [Accessed 18 08 2020].
150
151
152

IEA PVPS Report T13!25!2022 OandM Guidelines

Uploaded by

Copyright:

Available Formats

IEA PVPS Report T13!25!2022 OandM Guidelines

Uploaded by

Document Information

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

IEA PVPS Report T13!25!2022 OandM Guidelines

Uploaded by

Copyright:

Available Formats

Task 13 Reliability and Performance of Photovoltaic Systems

Guidelines for Operation

Report IEA-PVPS T13-25:2022

What is IEA PVPS TCP?

Visit us at: www.iea-pvps.org

What is IEA PVPS Task 13?

* Belgium is no longer participating in IEA PVPS, effective from 01 July, 2022.

INTERNATIONAL ENERGY AGENCY

IEA PVPS Task 13

Guidelines for Operation and Maintenance of

Report IEA-PVPS T13-25:2022

NWP Numerical Weather Prediction

The key highlights from each chapter are the following:

accumulation). Cleaning schemes can decrease production losses of PV mod-

In conclusion, a combination of well-designed O&M specifications, proactive monitoring sys-

2.1 PV Power Plant Data (Key Performance Indicators)

2.1.2 Performance Ratio

A. Module temperature correction for the performance ratio

B. Corrections for curtailment for the performance ratio

C. Soiling corrections for the performance ratio

D. Power factor correction for the performance ratio

2.1.3 Energy Performance Index

𝑇𝑢𝑠𝑒𝑓𝑢𝑙 − 𝑇𝑑𝑜𝑤𝑛 + 𝑇𝑒𝑥𝑐𝑙𝑢𝑑𝑒𝑑 (12)

More, commonly used availability metrics in contracts are discussed in [12].

2.2 O&M Contractor Key Performance Indicators

2.2.1 Different Types of O&M Key Performance Indicators

B. Guaranteed Plant Availability

2.2.2 Key Performance Indicator Shortfalls and Penalties

2.2.3 Data Acceptance and Classification

2.3 Contractual Framework

2.3.1 O&M Contract Framework

2.3.2 Influence of Region on O&M Framework

• As discussed in [17], it protects PV stakeholders, especially asset managers and O&M

2.3.3 Other Influences

POWER PLANT OPERATION

3.1 Plant Performance Monitoring Guidelines

Figure 3: Energy flow diagram in a grid-connected photovoltaic system (source: 3E

Utility-scale PV plants use complex highly customized monitoring systems, comparable to

Table 1: Monitoring system classifications and suggested applications as per IEC61724-

System losses analysis X X

Electricity network interaction X

Precise PV system degrada- X

Class A Class B Class C

Uncertainty on irradiance < 3% < 8% n/a

Irradiance sensor inclination 1⁰ 1.5⁰ 2⁰

Irradiance sensor azimuthal 2⁰ 3⁰ 4⁰

Inverter-level measurement ± 2% n/a n/a

3.1.1 Analysis of Monitored Data

• Expected behaviour calculation based on validated simulation models

3.1.2 Best Practice Guidelines

A non-exhaustive summary of best practice guidelines for PV plant performance monitoring is

Table 2: Best practice guidelines for PV plant performance monitoring.

Measurement Recommendations Explanation

Measurement Recommendations Explanation

Measurement Recommendations Explanation

3.2 Performance Analysis and Optimization

3.3 Power Plant Controller and Requirements for Grid Compliance

3.4 Power Generation Forecast

3.4.2 Requirements, Methodologies and Recommended Practices

parametric modelling PV system