Long‐term degradation rate of crystalline silicon PV modules at commercial PV plants: An 82‐MWp assessment over 10 years

Due to high competitiveness in the PV sector, despite the low degradation rate of crystalline silicon PV modules (below 0.5%/year), it is still important for utilities to know its actual value due to its impact on energy yield and hence, profitability, over the lifetime of a PV plant. However, uncertainties related to both the influence of downtime periods due to problems that may appear under normal operation of a commercial PV plant and to the measurement of degradation rates at PV plant level make this a challenging task. In order to obtain a significant value, in this paper, three measuring methods with different uncertainty sources are used for 82 MWp of PV modules on different locations in Spain and Portugal over 10 years. According to the different methods used and PV plants analyzed, excluding PV plants with problems, a range of degradation rates between 0.01 and 0.47%/year has been found. The overall average value observed is 0.27%/year. The findings of this work have also revealed the great importance of good operation and maintenance practices in order to keep overall low degradation rates.


| INTRODUCTION
The long-term degradation and stability of PV modules has great impact on the economics of PV plants. Financial models usually assume a long-term degradation rate for crystalline silicon, x-Si, modules of around 0.5% per year. 1,2 This is in accordance with the results of an extensive compendium of over 200 studies from the open literature up to 2015, which has found median degradation for x-Si technologies in the 0.5%-0.6% per year range, 3 and in accordance with the guaranties offered by manufactures, most typically in the 0.5-0.7%/year range. However, other studies show degradation rates closer to 0.2%/year. [4][5][6][7][8][9] The difference between assuming a degradation rate of 0.5%/year or 0.2%/year results in a difference of 3% of the energy yield of the PV plant during 20 years, meaning that it is of great importance to know the actual degradation rate of PV modules.
Measuring degradation rates in commercial PV plants is difficult to achieve experimentally due to the environmental variabilities that arise in consecutive outdoor measurements and due to the small magnitude of the measuring itself. A possibility consists of discrete measurements 10 of I-V curves for deriving the characteristic maximum power at Standard Test Conditions, P STC MPP , at selected modules in consecutive years. Difficulties arise from the high accuracy required, well below 1% of repeatability, and from the extrapolation of results to the whole PV plant. Another possibility consists of continuously observing a performance related parameter such as the Performance Ratio, PR. Now, difficulties arise from the modifications the PV plant or the measuring devices may suffer over the years.

The PV portfolio of Acciona includes 13 PV plants in Spain and
Portugal installed between 2004 and 2008, totaling up to 82 MW p of x-Si modules from four different manufacturers that are being routinely operated and carefully evaluated from this time. As far as we know, this is one of the largest commercial PV fleets in the world with more than 10 years of operation. Degradation rate of these plants is being assessed by both discrete I-V curves measurement separated by some years, and by the analysis of the monthly PR evolution. A 2014 publication 11 of the same authors of this paper assessed the degradation rates of four PV plants adding up to 15 MW during 4 years, revealing no apparent degradation during those years. The present paper expands the work presented in previous work 11 in terms of power, timespan, and methodologies. The relevance of our contribution comes from the fact that only 3% of nowadays global PV capacity was built before 2008, and few of those PV systems installed before 2008 have been monitored; moreover, the PV plants analyzed in this work represent around 0.5% of all the PV modules installed globally by the end of 2008. Although the main objective of this paper is to report on midlife normal degradation of x-Si PV modules, also observed failure modes and low P STC MPP at the beginning of life are discussed. The paper is organized as follows: section 2 describes the relevant characteristics of the involved PV plants, Section 3 presents the degradation rate measurement methodology, Section 4 shows the results and discussion and, finally, in Section 5, the main conclusions are explained. data from nearby state-owned weather stations, whose data is accessible at Gobierno de Navarra. 13 As a representative example, Figure 1A shows the PV plant P1, constituted by 153 individual systems, and Figure 1B Table 1, grouped by cell technology.

| DEGRADATION RATES ESTIMATION
Several methods have been proposed for assessing PV degradation.
Traditionally, long-term performance is expressed as a constant rate, in percentage per year, resulting in a gradual and homogeneous decline in annual performance. Implicit is the assumption of linear performance loss. More recently, methods for considered nonlinearities 24,25 and for reducing estimation uncertainties 26 have been proposed. However, these new methods require of continuous monitoring of irradiation, temperature, and energy production data. This is not available for the Spanish "solar farms." As mentioned above, monthly energy production is the only routinely monitored variable.
That lead us to assess the degradation of the here concerned PV Year instal.

Module manufacturer and cell technology Trackers
Power plants in terms of the conventional linear degradation rate. This is still a rather good representation of the degradation of x-Si modules. 27 We have proceeded with two different methodologies.

| Discrete peak power measurements
The P STC MPP of about 10 PV arrays of every plant is determined by recording their I-V curves and translating the maximum power point to STC assuming linear dependence against irradiance and constant power temperature coefficient. 28 To minimize uncertainty, the inplane irradiance, G is measured by means of the short-circuit current of a reference module, and solar cell operation temperature, T C , is measured using an infrared camera, always on clear days with very low wind speed. These few arrays are selected from among those with no operation anomalies in the year before the measurement, and the results are used to stablish a relation between P STC MPP and yearly energy production, which is then extended to calculate the P STC MPP of the rest of the arrays from the respective yearly energy productions, after discounting the differences due to other than degradation causes: on the one hand, the position of the corresponding tracker, which affects to shading and, on the other hand, possible operation anomalies (inverter shut-off, module change, etc.), as listed in maintenance records. More details are given in Llaria et al. 29 Finally, the degradation rate is determined from the results corresponding to several years by the conventional standard least square regression, SLS, approach. As a representative example, Figure 3A shows the P STC MPP measured in PV plant P5 and the derived degradation rate. The first value, in 2006, was measured 3 months after the plant's installation, which assures LID stabilization. We carried out two P STC MPP measurements campaigns. Because of that, we lack of initial P STC MPP values for the plants installed in 2008.

| PR evolution
The yearly PR of the plant is, first, determined. For that, the yearly inplane irradiation is calculated from the corresponding monthly values of horizontal irradiation recorded at close meteorological stations and free available in the web 13 ; and the energy production recorded at all the energy meters of the plant is corrected for discounting losses due to possible operation anomalies. Figure 3B shows the evolution of the yearly PR in PV plant P5 from 2006 to 2016.
It is worth commenting that the PR depends not only on P STC MPP of the modules but also on temperature and on the performance of the other PV system components: inverters, wiring, etc., that can also suffer degradation. That suggests that the degradation rate obtained with this method may represent, in general, an upper bound for the degradation rate of silicon cells. However, we have observed that yearly temperature is practically constant, and that energy loss due to equipment failure is very low and constant in well maintained PV plants.
This is the general case for these PV plants, and, moreover, data from periods with performance issues (e.g., underproduction due to damaged modules, inverters, or structures) is not used for the degradation rate estimation in this study. Hence, the annual trend of PR may well be assumed to be only caused by P STC MPP derating during this period, and the difference between degradation rates resulting from both methods can be understood as a general uncertainty indication.
The P STC MPP of the 48-MW PV plant, P13, has been directly determined from operational data provided by the SCADA (DC power, inplane irradiance, and solar cell temperature) in selected clear days.
Note that with this method, there is no need of measuring I-V curves in order to estimate the degradation rates. However, whenever the SCADA finds a performance issue, I-V curves, thermal images, visual inspection and other common diagnosis tasks are carried out in order to find the cause of the problem. Analogously, the PR evolution of this plant has been directly determined using the energy production and in-plane irradiation records registered by the SCADA. More    the substituted PV modules and of the peak power and PR observed in 2016 are also given. In order to visualize the results in Table 2, in Figure 7, it is depicted the hypothetical normalized energy yield of the PV plants, assuming constant yearly irradiation, during the first 10 years of operation using the obtained degradation rates (calculated as the average of the degradation rate obtained with the evolution of PR and the evolution of peak power). As it can be seen, PV plants in Groups   Non defective modules represent about 80% of the total modules population and degrade at about 0.27%/year in average. That is well below the limit guaranteed by the manufacturers. That means the energy production along 20 years is between 1.2% and 2.3% larger than estimated on the base of the guaranteed degradation. Two plants, representing 6.4% of the total involved power, have shown degradation rates above 0.5%/year due to cracked cells and hot spots.
Last but not least, degradations of up to 4.3%/year have been observed in short periods of time due to PV modules' failure or destructive weather events for example. In these particular cases, thanks to appropriate monitoring tools and supervisory protocols, the maintenance team was fast enough in detecting and correcting these problems in time which, otherwise, would have caused significant and permanent degradation of the PV plant and, hence, great economic losses to the owners. These events have shown that good operation and maintenance practices are essential for the correct performance of PV plants.