Hybrid forecasting system for renewable energy generation incorporating seasonal variability under sustainable development targets

Wind energy is strategically crucial for sustainable development amid global environmental and energy crises. However, nonlinear dynamics and multiscale seasonal fluctuations in wind speed generate significant prediction uncertainties, endangering grid stability. Therefore, this paper proposes a hybrid forecasting system, ICEEMDAN-CIAPO-ELM, to capture the spatiotemporal coupling characteristics within meteorological data, enabling accurate wind power generation prediction. The system employs Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) to decompose the original sequence into multiple dimensions, precisely resolving its cross-seasonal time-frequency characteristics for mitigating modal confusion and residual noise. To optimize the performance of Extreme Learning Machine (ELM), a novel Chaotic Inversion Arctic puffin Optimization (CIAPO) algorithm is proposed, inspired by chaos theory and Arctic puffins' collective behavior. CIAPO utilizes a chaotic inversion learning strategy for high-quality population initialization and integrates six heuristic mechanisms for efficient global exploration. Applied to hyperparameter tuning, CIAPO optimizes the output layer’s weight matrix and bias vector in single-hidden layer feedforward network, significantly strengthening the stability and generalization of the benchmark ELM. The prediction results based on the cross-seasonal wind-power dataset from the Jiangsu region, China, demonstrate that the proposed hybrid system outperforms other advanced forecasting approaches in handling wind speed fluctuations and multiscale characteristics. • A ICEEMDAN-CIAPO-ELM renewable energy hybrid forecasting system is proposed, integrating signal decomposition, optimization algorithms, and machine learning techniques, aimed at capturing the seasonal fluctuations of wind power generation data in dynamic environments. • A chaotic inversion Arctic Puffin Optimization (CIAPO) algorithm is developed for hyperparameter tuning, featuring a unique population initialization strategy and six heuristic search mechanisms. • The proposed ICEEMDAN-CIAPO-ELM hybrid forecasting system is rigorously tested on real-world wind power datasets with diverse seasonal and meteorological conditions, with performance evaluated using standard metrics. • The proposed CIAPO algorithm is thoroughly evaluated using the IEEE Congress on Evolutionary Computation (CEC) Benchmarking Test Suites, showcasing its rapid convergence and exceptional global optimization performance.