Optimizing photovoltaic power prediction using computational methods and artificial neural networks

This paper focuses on utilizing an Artificial Neural Network (ANN) to predict photovoltaic (PV) panel output power. Since solar power output is fluctuating and depends on climatic, geographical and temporal factors, precise prediction requires the implementation of computational approaches. The aim of this research is to develop ANN algorithms that anticipate solar power output and enhance the structure of them by incorporating the derating factor due to dirt (kdirt) into account. The effectiveness and dependability of the ANN are determined using MATLAB software. By comparing the Mean Squared Error (MSE) of four different values of derating factor due to dirt which are ۰.۸, ۰.۸۸, ۰.۹ and ۰.۹۸ in ANN predictions comprehend with ۴ input layers and ۱۰ hidden layers. Direct data input is obtained through a photovoltaic solar panel at University Tun Hussein Onn Malaysia (UTHM). Comparative analysis also has been carried out after theresults has been obtained from the mathematical equations. The daily solar power output predictions are effectively achieved by the deployed ANN. As the result, the optimal kdirt has been selected which is ۰.۸ based on its ability to produce the most accurate ANN predictions than the other values of kdirt.

کلیدواژه ها:

Power prediction ، solar output ، ANN ، MSE ، Derating factor