Publication

Abstract : The steady integration of wind energy into modern power systems highly relies on accurate short-term wind speed forecasting for consistent grid operations. In this study, a unique hybrid forecasting methodology is proposed that integrates the strengths of variational mode decomposition (VMD) with a novel machine learning model—firefly intensified revamped grey wolf optimized regularized extreme learning machine (FF-RGWO-RELM). The proposed framework decomposes raw wind speed data into sub signals (modes) using VMD to reduce non-stationarity and isolate meaningful frequency components. The decomposed modes are then fed into a RELM whose parameters are optimally tuned using the Firefly-Intensified Revamped Grey Wolf Optimizer. This enhanced Grey Wolf Optimizer incorporates Firefly-intensified local search, adaptive dimension learning hunting, weight-based position updating, parametric chaotic control, and a nonlinear convergence factor. With regularized and optimally tuned parameters, the FF-RGWO-RELM model effectively captures complex temporal wind speed patterns while maintaining strong generalization capability. Experimental validation on four benchmark datasets demonstrates that the proposed framework achieves RMSE reductions of 33.44%, 25.36%, 40.69%, and 24.56%, respectively, compared with existing models. Overall, the proposed approach provides a robust and reliable forecasting tool for handling non-stationary wind speed data, thereby supporting improved operational planning, grid stability, and sustainable wind energy integration into modern power systems.