Detection of different degradation mechanisms for photovoltaic systems using data-driven algorithms

Abstract

A main requirement for the photovoltaic (PV) uptake is the assurance of the lifetime energy yield and formulation of strict warranties to reduce investment risk. This requires reliable detection of PV underperformance issues (e.g., failures and degradation modes) and accurate evaluation of the degradation rate of fielded PV systems. Degradation is the decrease in PV system performance over time due to mechanisms not reversed in the field. It is well known that all PV systems suffer from degradation and it is observed at all levels (i.e., cell, module, array, and system) with different factors and degradation mechanisms apparent at each level. The key extrinsic variables that contribute to PV degradation are temperature, humidity and radiation. A major degradation mechanism for silicon wafer-based PV modules is potential induced degradation (PID), which results from the use of increased system voltage of up to 1000 V. The factors that can cause PID include difference in potential between earth and module, high temperatures and humidity, creating leakage current through the ground from which loss in power is caused. PID does not occur on all PV modules/cells. Other degradation mechanisms include the light induced degradation (LID) and the light elevated temperature (LeTID). LID occurs during the first year of module operation due to sun exposure. It affects the performance of the installed modules with respect to name plate data delivered by some PV module manufactures. LeTID is a form of solar cell degradation seen in the field and is accelerated by high irradiation and temperatures. The purpose of this master’s thesis is to analyse data from grid-connected PV systems and diagnose different degradation modes (e.g., PID and degradation) at early stages in attempt to safeguard the optimal performance of PV systems. This will be achieved by performing statistical analysis on the acquired data. In parallel, predictive algorithms will be implemented for forecasting PV degradation rates. The proposed methodology will be benchmarked against field data from test PV systems installed at the University of Cyprus (UCY) in Nicosia, Cyprus.