Abstract: With the vigorous development of new clean energy sources and the ongoing reform of the electricity market, electricity prediction has become increasingly important in the production and operation of power companies. To achieve precise electricity prediction, this paper proposes a combined short-term electricity prediction model. The model integrates an improved whale optimization algorithm (IWOA) with variational mode decomposition (VMD) and extreme gradient boosting (XGBoost)-corrected autoregressive integrated moving average (ARIMA). Firstly, the whale optimization algorithm is improved by incorporating nonlinear factors, adaptive inertia weights, and perturbation control factors to enhance its solving and search capabilities for optimizing VMD parameters. Secondly, the VMD with optimized parameter selection is used to decompose the power data, reduce data volatility and facilitate the learning process of the prediction model. Finally, an ARIMA-XGBoost power prediction model is constructed for the decomposed components, and the final prediction values are obtained by reconstructing the prediction results. Experimental results show that the proposed model outperforms the comparison models in terms of prediction evaluation metrics. The symmetric mean absolute percentage error decreased by 2.46% and 1.55% compared to least squares support vector regression and random forest regression, respectively, validating the higher accuracy of the proposed model in electricity prediction.