A Novel Hybrid Collaborative Forecasting Method for Offshore Wind Power Based on Intelligent Optimization

INTRODUCTION: With the rapid growth of offshore wind power integration and increasing wind power penetration, accurate power forecasting has become essential for maintaining grid stability and supporting economic dispatch. OBJECTIVES: This paper aims to develop a high-accuracy offshore wind power forecasting framework that can effectively handle noisy, non-stationary data and reduce the impact of outliers on prediction performance. METHODS: A two-stage data cleaning procedure is first constructed by combining density-based spatial clustering of applications with noise (DBSCAN) and polynomial regression to accurately identify and correct anomalous power data. The cleaned series is then decomposed using a sequential scheme that applies complete ensemble empirical mode decomposition followed by particle-swarm-optimized variational mode decomposition, producing multiple intrinsic mode components. Each component is fed into a hybrid temporal convolutional-gated recurrent unit (TCN-GRU) network, whose hyperparameters are globally tuned using an intelligent optimization algorithm, and the component-wise forecasts are aggregated to obtain the final power prediction. RESULTS: Simulation studies based on measured data from an offshore wind farm show that the proposed method significantly reduces forecasting errors compared with conventional forecasting models and single-stage decomposition approaches. CONCLUSION: The results demonstrate that the proposed adaptive optimization-based composite collaborative framework effectively improves both the accuracy and robustness of offshore wind power forecasting.