A hybrid chaotic-based discrete wavelet transform and Aquila optimisation tuned-artificial neural network approach for wind speed prediction

Wind speed prediction has received reasonable attention recently because of its clean and promising source of renewable energy. Recent studies have shown that developing efficient model to predict wind speed is a challenging task because of its nonlinear and stochastic characteristics. This paper aims to propose a new hybrid model to predict wind speed. For this purpose, Discrete Wavelet Transform (DWT), Phase Space Reconstruction (PSR) of chaos theory, Aquila Optimization Algorithm (AOA) and Backpropagation Neural Network (BPNN) are hybridised and a novel DWT-PSR-AOA-BPNN is proposed. To ascertain the proposed DWT-PSR-AOA-BPNN model performance, different hybrid model variants (DWT-PSR-GA-BPNN, DWT-PSR-PSO-BPNN, PSR-PSO-BPNN and PSR-AOA-BPNN) were developed for comparison. The comparison was done using statistical model evaluators of Mean Average Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and model efficiency of Loague and Green (ELG). The statistical results showed that the proposed DWT-PSR-AOA-BPNN model performed better and is therefore considered efficient wind speed prediction tool when compared with DWT-PSR-GA-BPNN, DWT-PSR-PSO-BPNN, PSR-PSO-BPNN and PSR-AOA-BPNN hybrid models. That is, the proposed DWT-PSR-AOA-BPNN had the lowest MAE, RMSE and MAPE values for the model testing (MAE = 1.1490, RMSE = 1.4190 and MAPE = 0.2743) and validation (MAE = 0.8122, RMSE = 0.9771 and MAPE = 0.1953). The DWT-PSR-AOA-BPNN also achieved the highest ELG values of 0.9904 (testing) and 0.99738 (validation) respectively. It is therefore concluded that considering the DWT-PSR-AOA-BPNN results, the indication corroborates the fact that this model can be utilized for efficient grid operations.