PDLC-based machine learning for photovoltaic systems fault detection and localization

Photovoltaic systems operate in various climatic conditions and are affected by shading, dust, bird droppings, and cracks, which can cause various faults. Detecting and locating faults is critical for photovoltaics maintenance and ensuring continued operation. In this work, polymer-dispersed liquid crystal (PDLC) was used to detect and locate faults. The cell voltage and current are monitored by varying irradiance intensity, thus distinguishing whether the cells are working or not. Experiments were conducted to measure the transparency of the PDLC. A photovoltaic was simulated in MATLAB/SIMULINK for several faults at different locations. Machine learning tools were used to accurately detect and locate faults. Support vector machine (SVM) and random forest (RF) were utilized to determine classification. RF performed better than SVM, achieving an accuracy of 98.89% compared to 96.30% for SVM in the binary detection stage. Furthermore, RF achieved an accuracy of 98.61% compared to 96.76% for SVM in the localization phase. The PDLC film can be intelligently used to protect cells from excessive radiation and control it in real time. The behaviour of the PDLC is simulated using an adaptive neuro-fuzzy inference system (ANFIS), while an iterative learning controller (ILC) is used to control the transparency.