Abstract A wind power generation forecast model based on WOA-SVM is presented to exploit effective information in different data sets completely. This model addresses the difficulties associated with parameter selection, low prediction accuracy, and susceptibility to local optima in short-term wind energy prediction. The model systematically investigates the relationship between various wind parameters (mean wind speed, maximum wind speed, minimum wind speed, mean wind direction, and mean hull position) and wind power. It then evaluates the model’s performance using mean absolute error and coefficient of determination. The predictive outcomes of the WOA-SVM model are contrasted with those of the SVM model, PSO-SVM model, and extreme learning machine (ELM) model, utilising authentic data from a wind farm located in northern China in the year 2021. The overall evaluation indicates that the WOA-SVM model outperforms the other models in short-term wind energy prediction, demonstrating superior prediction accuracy.