Ultra-Short-Term Wind Power Forecasting Based on the CEEMDAN-RBDBO-Bi-LSTM Model

Accurate wind power prediction is critical for maintaining grid stability and optimizing energy dispatch. Addressing the limitations of conventional models in handling stochastic, intermittent, and nonlinear wind power characteristics, this study proposes a CEEMDAN-RBDBO-Bi-LSTM hybrid framework. The methodology integrates three core components: improved Complete Ensemble Empirical Mode Decomposition (CEEMDAN) for adaptive signal stabilization, a bio-inspired Randomized Blackhole Dung Beetle Optimizer (RBDBO) for parameter optimization, and Bidirectional Long Short-Term Memory (Bi-LSTM) networks for multi-scale temporal dependency modeling. Validated against operational data from a Chinese wind farm, the model demonstrates enhanced noise robustness and prediction accuracy compared to traditional approaches, effectively mitigating volatility-induced errors while preserving computational efficiency.