Developing a photovoltaic energy generation forecast system using neural networks

Photovoltaic (PV) system is one of the trending and alternative sources of energy. Harnessing reliable energy in these PV panels is a cumbersome task equipped with several challenges such as continuous monitoring, adaptability in varying weather conditions, solar irradiance, wind speed and many more. It requires an optimized system to forecast solar energy efficiently. Thus, the given paper introduces a subnet-based feed forward neural network (SFFNN) to forecast solar PV energy generation based on varying weather conditions. The neural network is trained using Levenberg–Marquardt (LM) algorithm to discover connections among learning variables in the neural network. LM algorithm optimizes SFFNN by avoiding training failures and performing forward as well as backward computation for all neurons in the network layer thereby leading to faster convergence rate. Lastly, the performance of the proposed SFFNN is validated and compared with existing recent studies based on metrics such as root mean square error (RMSE) and mean absolute error (MAE).