A comparison between BNN and regression polynomial methods for the evaluation of the effect of soiling in large scale photovoltaic plants

• Published in: Applied energy Vol. 108; pp. 392 - 401
• Main Authors: Massi Pavan, A., Mellit, A., De Pieri, D., Kalogirou, S.A.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.08.2013; Elsevier
• Subjects: Applied sciences; Bayesian NN; Cleaning; Energy; Equipments, installations and applications; Exact sciences and technology; human resources; Large scale photovoltaic plant; losses; Maintenance; Mathematical models; Natural energy; Neural networks; Photovoltaic cells; Photovoltaic conversion; Pollution; Polynomial regression; Position (location); Power plants; Regression; regression analysis; Soiling; Solar cells; Solar energy; Bayesian NN; Pollution; Soiling; Maintenance; Polynomial regression; Large scale photovoltaic plant; Evaluation; Solar energy; Photovoltaic system; Electric power production; Photovoltaic cell
• ISSN: 0306-2619; 1872-9118
• DOI: 10.1016/j.apenergy.2013.03.023

•The soiling effect can have a significant impact on the PV plant performance.•Bayesian neural network performs better than polynomial regression model.•Grid operators can benefit from the proposed technique. This paper presents a comparison between two different techniques for the determination of the effect of soiling on large scale photovoltaic plants. Four Bayesian Neural Network (BNN) models have been developed in order to calculate the performance at Standard Test Conditions (STCs) of two plants installed in Southern Italy before and after a complete clean-up of their modules. The differences between the STC power before and after the clean-up represent the losses due to the soiling effect. The results obtained with the BNN models are compared with the ones calculated with a well known regression model. Although the soiling effect can have a significant impact on the PV system performance and specific models developed are applicable only to the specific location in which the testing was conducted, this study is of great importance because it suggests a procedure to be used in order to give the necessary confidence to operation and maintenance personnel in applying the right schedule of clean-ups by making the right compromise between washing cost and losses in energy production.