Wind power prediction plays a crucial role in ensuring the reliable integration of wind energy into the grid. This study proposes a novel hybrid model combining random forest (RF) and convolutional neural network (CNN), referred to as the RF-CNN model, specifically designed for short-term wind power prediction. The model integrates the advantages of RF integration technology, random selection of attributes, and CNN capturing the spatiotemporal characteristics of wind power, to enhance prediction accuracy and robustness. Firstly, by analyzing the analog equivalence between decision trees and CNNs, the theoretical basis for combining RF and CNN is established. Next, an evaluation system for wind power prediction models that includes root mean square error (RMSE), determination coefficient, and Spearman correlation coefficient is introduced. Finally, validatinos are conducted using three open-source wind power datasets from European wind farms. The results demonstrate that, compared to other five models, the RF-CNN model outperforms in all three datasets, thus confirming the model’s effectiveness and accuracy for wind power prediction.