Search

• Two hybrid models for wind power forecasting MWSO-RBFNN and MWSO-GRNN are proposed. • Six cases of Turkey and Malaysia are analyzed by RE, MAE, RMSE, NSC and R 2 correlation. • Comparisons with SOTA models confirm the performance superiority of the proposed models. In order to combat the global warming, much stress is put on integrating non-conventional energy resources, such as wind power plants and solar energy systems, into standard electricity generation systems. The stochastic nature of wind energy at a large scale requires precise forecasting in short and medium terms to manage electricity demands for regional grids because the safety and control of electricity systems are highly sensitive to voltage drop. Unfortunately, the existing forecasting models achieve less desirable performance in short term power forecasting. To resolve this issue, based on hybrid stochastic algorithm based intelligent models, which are proven to be a highly effective predictive tool, this paper proposes two new hybrid models using modified white shark optimization (MWSO) algorithm based GRNN and RBFN. To enhance the forecasting performance, the parameter mv is projected as a function of time to enforce the convergence by accelerating computation as well for effective exploration and exploitation to reinforce features of data. The supervised control and data acquisition (SCADA) data provide the input metrics to the system. A hybrid decomposition model is used to prep data for feature enhancement. Six case studies are formulated from Turkey and Malaysia to validate the forecasting performance of our technically advanced models. Recommended modifications allow MWSO for enhanced training. The results are compared with four hybrid ANN topologies in combinations with PSO and BMO stochastic optimization algorithms. Seasonal results are compared graphically and statistically through 15 min ahead forecasting. The results show that the proposed MWSO-RBFNN model outperforms the classic models in all cases for point forecasting and interval forecasting with higher convergence rate and lower stochastic error. The point and interval forecasting accuracy is extremely important to ensure power systems' stability, efficiency, and safety. Our model achieves Nash-Sutcliffe constant score of 0.99 and exhibits superior performance with the least RMSE, RE, R2, and MAE. [ABSTRACT FROM AUTHOR]