PERFORMANCE ANALYSIS OF GRID-CONNECTED PV SYSTEMS Stefan Mau arsenal research, Business Area of Renewable Energy Technologies Giefinggasse 2, A-1210 Vienna, Austria, Phone: +43 (0) 50550 - 6652, Fax: +43 (0) 50550 – 6390 Email: Stefan.Mau@arsenal.ac.at Ulrike Jahn Bavarian Center for Applied Energy Research (ZAE Bayern) Am Weichselgarten 7, D-91058 Erlangen, Germany, Phone: ++49-9131-691-295, Fax: +49-9131-691-181 Email: email@example.com ABSTRACT: Performance, quality and reliability issues are becoming more and more important for the emerging photovoltaic markets worldwide. The purpose of this work is comparing and presenting operational performance results of grid-connected PV systems, as collected and elaborated for Task 2 of the Photovoltaic Power Systems Programme (PVPS) of the International Energy Agency (IEA). The data analysed here are derived from analytical monitoring data, hourly values and 5-minute-values of 21 selected PV systems installed in five countries in Europe and Japan. Graphical analysis methods are used to identify the energy frequency distribution and to explain the operational system behaviour, of the inverter in particular. The performance ratio is analysed at identical climatic conditions by extracting data with identical module temperatures and irradiances to detect small shifts of system performance. This contribution describes and displays different approaches and methods of PV system performance analysis to understand and improve the operational behaviour of PV systems. Keywords: Performance, Reliability, Analysis
Task 2 of the IEA PVPS Programme  is an international collaborative group focusing on the operational performance, long-term reliability and sizing of PV systems and subsystems, providing technical information to PV experts, research, PV industry, utilities, system designer & installers as well as to the education sector & end-users. Task 2 started its work in 1999 with the overall objective to improve PV system operation and sizing by analyzing and disseminating information on technical performance. Deliverables include a PV Performance Database, technical reports and workshop proceedings, which can be accessed through the website: www.iea-pvps-task2.org. Activities to date include the work on investigations of buildingintegrated PV, long-term system performance and maintenance analysis, reliability issues, monitoring techniques, normalised evaluation of PV systems and various national procedures. This work is part of a performance analysis activity within Task 2. The approach in this paper is to elaborate about the system performance values over time, based on in-field monitoring data, at identical climatic conditions. The goal is to state more precisely on small shifts of performance ratio that are usually superposed by larger seasonal time fluctuations. The paper focuses on analysis results on long-term performance and reliability issues of 21 selected PV systems in five different countries in Europe and in Japan. Long-term performance data (hourly and 5-minute values) from PV installations in Belgium, Germany, Italy, Japan, Sweden and Switzerland are used to demonstrate the performance ratio (PR) over time under constant climatic conditions such as in-plane irradiance, incidence angle of irradiance and module temperature. The PR indicates the overall effect of losses on the system output due to incomplete
utilization of the irradiance, temperature effects and system component inefficiencies or failures. Upon this research, performance shifts of PV systems over time are derived and can be used as a failure detection method. 2 DATASETS AND ANALYSED SYSTEMS
The data analysed here are derived from analytical monitoring data, hourly values and 5-minute-values, which were averaged to hourly mean values. The 21 analysed PV systems are located in Belgium, Germany, Italy, Japan, Sweden and in Switzerland with installed nominal powers between 1 and 16 kWp. They include roof top systems, freestanding systems as well as facades. An overview of the operational years and the analysed monitoring period for the 21 grid-connected PV systems is given in Table 1. Table 1: Overview of the analysed datasets of the 21 grid-connected PV systems, their operational years (in gray colour) and monitoring periods (in black colour)
Name P1 P2 P3 P4 P5 P6 P7 P8 P9 P 10 P 11 P 12 P 13 P 14 P 15 P 16 P 17 P 18 P 19 P 20 P 21 … 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 Years 2 2 3 6 5 6 5 5 5 5 3 5 5 5 5 4 9 3 9 11 3
The operating years vary from plant to plant covering minimum 7 years of operation (2000-2006) to 23 years (1984-2006). The monitoring periods range from two complete years (plant 1 and 2 in Tab. 1) to 11 years (plant 20). Accordingly, we analysed between 104 and 106 datasets per plant and a total of 2*106 datasets. Each dataset contains the parameters in-plane irradiance (Gi) and horizontal global irradiance, ambient temperature and module temperature as well as DC and AC power. For some of the PV plants, data of diffuse inplane irradiance, DC voltages and DC currents as well as AC voltages and AC currents are available. System performance analysis is not always straight forward plugging the right numbers into the standard formulas for grid-connected performance analysis results. In our investigations on 21 selected PV plants using a total of 102 years of monitoring data (Tab. 1), we experienced 5 years of missing data (4.9%) and 2 years of abnormal data (2.0%). Further obstacles were here observed for several plants, when the data format or time resolution changed during the monitoring period or the time stamp was not available. In the case of missing data, they were not provided and could not be retrieved or the data were substituted by replacement characters. In some cases abnormal data of one or more parameters occurred, while the system was supposed to operate well. For PV system analytical monitoring, both figures represent typical features of data quality and data availability, while data “losses” in the order of 5% and data “abnormality” of 2% are comparably modest values, which can be handled. 3 PERFORMANCE ANALYSIS AND RESULTS
Fig 1: Final yield Yf as a function of reference yield Yr – hourly readings of 9 years of operation Furthermore, this plant’s operation is characterized by zero energy output (Yf) for all levels of irradiance values (Yr) lining up along the x-axis of Fig.1. Which is the reason for this behaviour? Fig. 2 attempts to give the answer employing the graphical analysis method introduced by Baumgartner et. al. for inverter characterization and optimisation . The DC voltage power shows a “footprint” of the hourly energy frequency distribution for plant 19 (roof-top system of 3 kWp for 9 years). As shown in Fig. 2, the contours display sums of yearly array yields YA for interval of 2% of PDC and VDC respectively. Most energy output of the PV modules occurs in the intervals between 60% and 70% of PSTC, at DC voltages from 80% to 84% of VSTC. Higher module temperatures in summer are responsible for shifting the energy production to lower voltage values, resulting in different seasonal operational states, which are typical for roof-top-systems .
3.1 Final yield as a function of reference yield Reference yield Yr and final yield Yf are the most important parameters for PV system analysis for gridconnected PV plants. The ratio of Yf and Yr is defined as the performance ratio (PR), which is independent of the plant’s location with respect to irradiation, but the PR is effected by the module temperature and thus by the ambient temperature of the system’s site. For characterizing the performance of a system, the PR is used as the main index. In Fig. 1, hourly mean values of final yield, based on 5-minute-values, are plotted as a function of reference yield for nine monitoring years of plant 19 (1997-2005). It also displays PR values as the ratio of Yf and Yr showing that the majority of the hourly PR values range between 0.60 and 0.80. PR values smaller 0.6 occur for different values of irradiance, while most of the PR scattering is observed at low Yr and hence, for low inplane irradiance (Gi). Gi values smaller than 10 W/m2 are deleted to eliminate the operation at night times. Fig. 1 shows rare cases of hourly PR values > 1, which can be explained by timely fluctuations in irradiance and PV energy output during the monitoring period of one hour. Inconsistent PR values bigger than 1 can be avoided by choosing smaller monitoring intervals such as 5-minute-values.
Fig 2: DC voltage power plot for plant 19 using hourly values of 9 years. A maximum of produced energy YA occurs at DC voltages of 80% to 84% of VSTC Fig. 2 shows a straight line on the left side indicating that the inverter is limiting the voltage at 80% of the STC voltage. For higher module temperatures and hence, operating states of lower voltages <80%, the energy production YA as well as Yf is zero (Fig. 1). These nonproductive operating states can be overcome by better inverter sizing of this plant. Increasing the string voltage
by connecting an additional PV module in series will avoid the inverter limitation at the lower voltage end and result in higher system efficiency and PR. 3.2 Module temperature effects on the system performance The direct impact of the PV module temperature on the performance ratio can be displayed as shown in Fig. 3. Hourly values of PR are shown for any day time (y-axis) and for each month (x-axis) using PR data of plant 16 for four monitoring years. For each month and each day time, data were extracted and then averaged under the condition that at least 3 values were existing. The colour defines the level of PR ranging from 0.5 (black) to 0.9 (orange). Irradiance data lower than 10% of the maximum Gi value were extracted, which results in different morning starts and winter stops in summer (6 or 7 a.m. respectively 5 p.m.) and in winter (8 a.m. respectively 3 or 4 p.m.). During morning hours in winter, the PR scattering is much higher leading to less confident averaged PR values. Monitoring failure or missing data are easily detected for two months (April, May) in 2001. For each month, the PR changes over the course of the day, showing very high PR values in the morning (orange) and lower PR values at noon time, which is clearly correlated to increased module temperatures at noon. Particularly, in the summer months at noon time, high PV module temperatures up to 60°C cause the PR to decrease to 0.75, which is expressed in form of dark green islands. This pattern is repeating from year to year. In the second half of 2004, we observe a significant reduction of PR to 0.55 due to system failures. This kind of PR mapping is useful to explain different effects such as module temperature, shading and system failures on the PV system performance .
total monitoring period. One example is shown in Fig. 4a. Hourly mean values of PR are displayed as a function of module temperature for the irradiance interval 200 – 400 W/m?. The colours define the temporal sequence of the data as indicated by the colour bar. Extensive scattering of PR values for this plant as well as for all other plants was observed for low irradiance levels over the whole range of module temperatures. Investigations regarding the angle of incidence of solar irradiation on the module’s surface and on the inverter efficiency did not lead to a significant reduction of PR scattering. But it was observed that scattering of PR in winter was generally higher than during summer periods.
Start Monitoring Period End
Fig 3: Monthly values of Performance Ratio as a function of the month and of the day time 3.3 Performance Ratio at identical climatic conditions For PV system analysis the PR is often analysed and visualised as a function of time. But as seasonal fluctuations of climatic conditions show a large impact on the PR, even for well-operating systems, small shifts of system performance due to partly malfunction are difficult to determine. To overcome this problem the PR was analysed at identical climatic conditions by extracting data with identical module temperature and irradiance from the
Fig 4: Hourly mean values of Performance Ratio as a function of module temperature for Plant 17 (1997-2005) and Plant 21 (2002-2004). The colours define the sequence of the monitoring data: blue = begin and red = end of monitoring data.
The timely fluctuations of irradiance are one of the reasons for the large scattering of PR. This phenomenon is demonstrated in Fig. 5. The standard deviation of irradiance during full hours is calculated on the basis of 5minute-values and plotted as a function of the mean hourly value of the irradiance for the years 2004 and 2005. It can be seen that for low irradiance levels, the time fluctuations of irradiance are higher than for high irradiance levels (Fig. 5). Hence a significant reduction of scattering of PR can be observed at higher irradiance levels as shown in Fig. 4b. For the same plant 17, hourly data in the irradiance interval 800 – 1000 W/m? were extracted. The reduction of PR with increasing module temperatures by about 0.4%/°C can clearly be observed and is attributed to the temperature behaviour of crystalline silicon modules. As indicated by the colour, the PR is stable over the nine years of operation and monitoring. But more important is the reduction of scattering of PR, e.g. at 40 °C module temperature σ=0.03 (k=2). Accordingly, 95% of all PR values during nine years of operation are located within the small range of 0.74 ±0.03. Due to little scattering of PR at high irradiance levels, even small shifts of system PR can be detected. One example for such failure detection is given in Fig. 4c. The PV array (tilt angle = 80°) consists of one string of 48 crystalline silicon modules connected in series. Due to penetration of moisture into the connection box, two PV modules were short-circuited and replaced some months afterwards. The reduction of PR due to this failure can clearly be seen in Fig. 4c. As indicated by the colours at the beginning of 2003 (light blue) one module and one year later (orange), the second module is short-circuited. By mid of 2004 (red) the modules were already replaced and the PR values jump back to the initial value as of 2002 (dark blue).
are supposed to run without any failure during the nine years of monitoring period. As the latter shows significant increase of scattering, system failures during the 5 and 11 years of operation could not be excluded. To detect possible failures, the data of these plants were analysed by ENECOLO using their failure detection routine. This sophisticated routine calculates the probability and aggravating of different kinds of failures (e.g. soiling, snow coverage) for each day. After deleting the days indicated by the routine in both plants, the analysis was repeated. The results show a large reduction of standard variation for both plants as well as for the two methods. The comparison of the two methods shows that PR vs. Tmod shows significant lower standard deviations for all plants. Table 2: Comparison of scattering of data for different analysing methods Standard deviation/ mean value of data (800 - 1000 W/m?) PR vs. Tmod Yf vs. Yr
Original data + ENECOLO Original data + ENECOLO
P 17 P 19 P 20 P8
± 3.2% ± 4.2% ±11.2% ±11.6% ± 7.8% ± 9.2%
± 7.4% ± 8.2% ± 14.8% ± 15.6% ± 9.2% ± 11.6%
Detailed PV system analysis was performed for 21 PV plants operating at different climatic conditions over a period of 7 to 23 years. All results from the analysis and observations made in this paper indicate that one can expect a stable PV system performance over a period of many years. On the other hand, 5 of 21 PV systems showed long-term variations of performance over time that might be related to soiling of modules and decreasing efficiency of components. Five plants showed temporary failures that could be attributed to problems such as short-circuiting of modules, under-sizing of inverter and wrong irradiance measurements. The “footprint method” and PR at identical climatic conditions provide useful methods to detect temporary as well as long-term effects such as module temperature, snow coverage and shifts in PR. Furthermore, it was shown that PR at identical climatic conditions shows little scattering of data and thus, it is a most useful tool for detecting even very small variations of PR. 5 OUTLOOK
Fig 5: Standard deviation of in-plane irradiance as a function of irradiance calculated from 5-minute-monitoring values for Plant 17 in 2004-2005 3.4 Comparison of methods The different analysing methods used in this work show different scattering behaviour of data. An overview of the standard deviation of data at irradiance interval 800 - 1000 W/m? for the methods PR vs. Tmod and Yf vs. Yr (see figures 4 and 1) is given in Table 2. As also the systems show different scattering behaviour, two systems with low (P 17 + P 19) and two systems with high (P 8 + P 20) scattering were exemplarily chosen. The first two
The performance analysis at identical climatic conditions applied to detect small shifts of PR is demonstrated to give reliable information about system failures. On-going analysis will prove whether further parameter, like clear sky index, are applicable to gain more detailed information. The graphical analysis
method using the normalized DC-voltage and DC power as plotting base will be extended to inverter efficiency, final yield and performance ratio. The characterization of components allows to improve system sizing and behaviour. The surface plots of PR over time axis will easily be extended to array yield and “array performance ratio” to determine the effect of module temperature on the array performance. 6 ACKNOWLEDGEMENTS
The authors would like to thank all colleagues of IEA PVPS Task 2 who contributed their information and PV system data for the performance analysis of this publication. Special thanks go to Christopher Caesar and Roney Thomas for their valuable contribution to the analysis. The Austrian contribution of this work is supported by the programme “Energiesysteme der Zukunft” – a cooperation of the Austrian Ministry for Transport, Innovation and Technology and the Austrian Research Promotion Agency. This German contribution of this work is supported by the German Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (BMU) under contract No. 0329640D. 7 REFERENCES
 S. Nowak: ?The IEA PVPS Programme: Towards Sustainable Global Deployment of PV“. In: Proceedings 20th European Photovoltaic Solar Energy Conference, June 6-10, 2005, Barcelona, Spain. Volume III of 3, pp. 2827  F. Baumgartner, H. Scholz, A. Breu, S. Roth: ?MPP voltage monitoring to optimise grid-connected system design rules“. In: Proceedings 19th European Photovoltaic Solar Energy Conference, June 7-11, 2004, Paris, France. Volume II of 3, 2005-2010  H. Scholz: ?Better clues to losses and performance of PV systems through comparative surface plots over time axis”. In: Proceedings 20th European Photovoltaic Solar Energy Conference, June 6-10, 2005, Barcelona, Spain, pp. 2167