Methodology for the Identification of Dust Accumulation Levels in Photovoltaic Panels Based in Heuristic-Statistical Techniques

The use of renewable energies is increasing around the world in order to deal with environmental and economic problems related conventional generation. In this sense, photovoltaic generation one most promising technologies because high availability sunlight, easiness maintenance, reduction costs installation production. However, panels are elements that must be located outside receive sun radiation transform it into electricity. Therefore, they exposed weather conditions many factors can negatively affect output delivered by system. One common issues location dust accumulation surface panels. particles obstruct passage reducing efficiency process making system prone experimental long-term faults. Thus, necessary develop techniques allow us assess level panel schedule a proper maintenance avoid losses associated power unexpected work, we propose methodology uses machine learning approach estimate different levels developed method takes voltage, current, temperature, radiance as inputs perform statistical feature extraction describes behavior under conditions. retain only relevant information, genetic algorithm works along principal component analysis technique an optimal selection. Next, linear discrimination carried out using optimized dataset reduce problem dimensionality, multi-layer perceptron neural network implemented classifier for discriminating among three conditions: clean surface, slight accumulation, severe accumulation. proposed real signals from installation, proving effective not determine if condition present but also when actions performed. Moreover, results demonstrate accuracy always above 94%.