Extracting Trends Ensembles in Solar Irradiance for Green Energy Generation Using Neuro-evolution

Abstract

Globally, there are variations in climate, there is fossil fuel depletion, rising fossil fuel prices, increasing concern regarding energy security, and awareness about the environmental impacts of burning fossil fuels. These factors lead to a growing interest around the world in green and renewable energy resources, solar energy being a common one. A Neuro-evolutionary approach is explored to extract the trend ensembles in the solar irradiance patterns for renewable electric power generation, using the data taken from stations in Al-Ahsa, Kingdom of Saudi Arabia. The algorithm, based on Cartesian Genetic Programming Evolved Artificial Neural Network (CGPANN) was developed and trained for hourly and 24-hourly prediction, using the solar irradiance value as the input parameter. It was tested to predict solar irradiance on hourly, daily, and weekly basis. The proposed technique is 95.48% accurate in solar irradiance prediction.

Recurrent Cartesian Genetic Programming of Artificial Neural Networks

“…Series forecasting is an important application of machine learning and statistical modelling techniques, including GP [16,31,39,57] and ANNs [21,83,84], finding application in many disciplines including: economics, politics and planning. Series forecasting is also a common application of NE [13,32,61] including CGPANNs [32,52].…”