Incremental neuroconductance to analyze performance losses due to soiling in photovoltaic generators

Abstract

Many large photovoltaic plants are being installed in desert regions with very high solar irradiation but are subject to significant losses due to soiling. Quantifying soiling losses in energy production is crucial for optimizing cleaning schedules and the plants’ bankability but can be challenging, particularly when knowledge about soiling losses is desired before plant implementation. This work introduces an innovative, self-contained, and easily deployable system that quantifies energy losses from dust accumulation on photovoltaic arrays. Our approach, termed ‘Incremental Neuroconductance,’ combines a photovoltaic module, irradiance, and module temperature sensors with a calibrated electrical model and a trained artificial neural network. The models estimate the clean module’s maximum power output based on the measured irradiance and module temperature. The system simultaneously measures the module’s maximum power by using maximum power point tracking based on the incremental conductance method. Under clean module conditions, the models and measurements yield identical power values. However, once the module starts accumulating dust, the measurements yield lower power values than predicted by the electrical model and the neural network, quantifying the power loss due to soiling. We assess and contrast the precision of the electrical and the artificial neural network models in estimating the clean module’s power, emphasizing the importance of their calibration and training using experimental data. Our findings reveal comparable performances between the two models, with the trained neural network offering lower computational costs. Model recalibration and retraining can compensate for long-term module degradation. We report on the results from one year of system operation, demonstrating its capability to predict performance losses due to the soiling of the photovoltaic module without requiring the implementation of a complete photovoltaic system or plant. Over the year, maintaining a monthly cleaning schedule, the system estimates a 7 % energy loss due to soiling for a photovoltaic array located in the desertic region of Arequipa, Peru.