A Study on Al-Based Algorithms for Predicting the Output of Deteriorated Photovoltaic Module

Recently, it is important that can effectively utilize a large amount of data to analyze the operating factors for photovoltaic (PV) system with the AI technology. Predicting the use of solar modules can significantly contribute to reducing global greenhouse gas emissions by lowering carbon emissions. It is possible to diagnose deterioration and aging condition using various data of PV system for usage and operation of the PV modules. In order to assess the predictability of power of aged PV modules, the electrical characteristics of solar cell modules was measured in the dark state using the light irradiation. The output of aged PV modules was predicted through three different artificial intelligence-based algorithms. The power of aged PV modules decreased by more than 35% compared to the initial stage, and the main reasons of output degradation were determine to be fill factor (FF) reduction because of the increased resistance due to cell cracks or corrosion caused by aging. As a result of performance evaluation, the random forest (RF) model out of them showed the highest prediction accuracy with R2 = 0.885, and normalized mean bias error (NMBE) of 1.466%.