Prediction of energy injected into the grid using a hybrid artificial intelligence approach: case of the Ten Merina photovoltaic solar power plants (Senegal)

This paper presents a Hybrid Particle Swarm Adaptive Neuro Fuzzy Inference System (HPSANFIS) technique for predicting the energy injected into the grid by a photovoltaic (PV) power plant. In the proposed predictive model, Particle Swarm Optimization (PSO) was selected as the optimizer for the training process of the Adaptive Neuro-Fuzzy Inference System (ANFIS). The proposed method is validated by using actual data from the Ten Merina solar power plants in Senegal. The artificial intelligence (AI) method is compared with methods based on the performance ratio (A1 method) and the method of the online simulation software Photovoltaic Geographical Information System. These methods were implemented on MATLAB/Simulink. A daily production prediction was made and analyzed according to the season (dry or rainy). The performance study showed Root Mean Square Error of 0.6823 kWh, 23.9178 kWh, and 133.0048 kWh, respectively, for the proposed model, A1, and Photovoltaic Geographical Information System models. This study also showed that the proposed model has the highest yield across all seasons.