Predicting daily wind speed using coupled multi-layer perceptron model with water strider optimization algorithm based on fuzzy reasoning and Gamma test

Abstract

Wind energy is a valuable renewable resource that plays a significant role in electricity generation process. Predicting wind speed (W.S.) is critical for effectively managing wind energy and producing power. This study proposes an improved multi-layer perceptron (MLP) model for W.S. prediction that incorporates a novel optimization algorithm namely water strider algorithm (WSA). The WSA optimizes the MLP parameters to increase the model’s accuracy. The MLP-WSA model’s predictive capabilities were compared with various algorithms, such as MLP-sine cosine (SCA), MLP-salp swarm (MLP-SSA), Multi-Layer Perceptron-particle swarm optimization (MLP-PSO), and MLP models. Furthermore, we proposed an inclusive multiple model (IMM) that utilizes the outputs of the MLP-WSA to predict W.S. The study utilized fuzzy reasoning to modify MLP models to remove redundant weights and reduce computation time. Finally, to predict W.S, we considered five stations in Malaysia. By utilizing the WSA and Gamma tests, we identified that the IMM model provided the most optimal input for our model. To test the I.P. station, the RMSE of the IMM model was lower than other models. Additionally, the NSE of the IMM model was found to be higher at the B.L. station, indicating superior performance. Furthermore, the IMM model’s mean absolute error was notably lower than other models in C.H. station. Overall, the results demonstrate that the combination of the WSA and Gamma tests allowed us to achieve more accurate and efficient predictions with less computation time using fuzzy reasoning.