- Scientists from Spain’s University of Granada have presented and compared four regression methods to estimate potential-induced degradation (PID) in PV strings.
- Their methods use the strings’ open-circuit voltage (Voc) measurement for the predictions.
“The proposed method solves a heat stress health safety problem, prevents water supply failures, and estimates the size of repowering purchases,” said the academics. “Days with very high temperatures are the best times to detect PID in the field, but they represent a safety problem for the maintenance workforce in very hot areas. The method is fast, and it is not necessary to reach the roof.”
The developed model was based on data collected from two 75 kW PV power plants, installed on the roof of agricultural water pumping facilities that they power. The p-type monocrystalline silicon plants were installed in 2009 in Spain’s province of Córdoba, one of the hottest areas in the country’s summer. The manufacturer’s specifications guaranteed a maximum power loss of 10% after 10 years and 20% after 25 years. However, measurements show that after years their production dropped by 67.58% and 47.22%, respectively.
“In 2019, a repowering project was launched that included an assessment of the degradation of the three plants, and the replacement of the most degraded panels: 246 modules out of 656 (37.57%) in plant 1 and 186/608 (30.59%) in plant 2,” said the academics. “The repowering partially recovered production in 2020 to 73.04% of the mean of the first four years in plant 1 and 72.88% in plant 2 by 2021.”
Using this chance, the team did some field measurements with which they could develop their models. Data taken from those plants resulted in pairs – the open-circuit voltage of strings and the number of degraded panels. Those pairs accounted for 41 strings in plant 1 and 33 in plant 2. In addition, some panels were taken to a lab test to confirm PIDs.
They initially used Linear regression to model this relationship so that the number of degraded panels of a string could be estimated from its open-circuit voltage. They then added a quadratic regression. “The methods are initially fitted with data from plant 1, giving regressions P1L (linear), P1Q (quadratic), and without outliers P1LO, P1QO. Equivalently, fitting with data from both plants, the models are P12L, P12Q, P12LO, and P12QO,” they said.
This analysis showed that the coefficient of determination (R2) between the Voc and the number of estimated degraded strings was best in P12QO, with 0.8204. P12LO had an R2 of 0.8189, P12Q of 0.7020, and P12L had an R2 of 0.6974. Using the best model, the team calculated how many new panels they needed to order for the repowering of plant 3, which was of a different size and had issues from the other plants. Per their calculation, a purchase order of 96 panels has been placed.
“The main contribution is that it is born out of a real practical need; a quick estimation of the panels that are in a condition of failure, with no need to go up to the roof of the plant, avoiding long measurements of voltages module by module at times of danger of thermal shock for maintenance workers. In this way, the readings of the Voc of the strings in the protection cabinet facilitate estimating the number of panels in the described failure conditions,” concluded the researchers. “The method does not locate the failing modules in the string, and it is starting to be used by the authors for the repowering actions of the plants.”
Their findings were presented in “A simple method to estimate the degraded photovoltaic modules from the string open-circuit voltage in solar pumping systems,” published in Solar Energy.
Author: Lior Kahana
This article was originally published in pv magazine and is republished with permission.
