A hybrid model for short-term offshore wind power prediction combining Kepler optimization algorithm with variational mode decomposition and stochastic configuration networks

• A novel hybrid artificial intelligence model is proposed for wind power prediction. • Data features are effectively selected through the optimized decomposition model. • Parameter optimization of stochastic configuration networks is proposed. • Multi-season data analysis is conducted to verify the reliability of the model. With the burgeoning development of the wind power industry, the significance of wind power forecasting in enhancing electricity generation efficiency, minimizing energy waste, and improving electrical grid management is increasingly highlighted. To enhance the stability and accuracy of wind power forecasting, a hybrid model integrating Kepler optimization algorithm (KOA), variational mode decomposition (VMD), and stochastic configuration network (SCN) is proposed. Firstly, the series of wind power data is decomposed using the VMD method optimized by the KOA, aiming to smooth the wind power series while preserving its inherent characteristics. Subsequently, permutation entropy (PE) is employed to order and reconstruct the decomposed wind power subsequences, with the selection of input features by the maximal information coefficient (MIC) and autocorrelation function (ACF). Following this, KOA is utilized to optimize the parameters of the SCN model, further enhancing the predictive performance of the SCN. Finally, a multi-seasonal and multi-scenario wind power forecasting analysis is conducted by using an actual data set from an offshore wind farm in China. Compared with the basic VMD model, the data decomposition efficiency of the optimized VMD model has been improved by 28.86%. Meanwhile, the prediction average error of the proposed model has decreased by 0.1385 compared with the basic prediction model. The results demonstrate that the proposed hybrid model exhibits superior stability and accuracy in short-term wind power prediction.