Abstract: Accurate short-term power load forecasting is a core task for the stable operation and economic dispatch of power systems. To improve the accuracy of power load forecasting, this paper proposes a hybrid forecasting model that integrates multi-scale decomposition, improved crayfish optimization algorithm, and improved convolutional long short-term memory (ConvLSTM). Firstly, the advantages of the complete ensemble empirical mode decomposition with adaptive noise and the variational mode decomposition algorithm are combined to decompose the load data at multiple scales, thereby reducing the data complexity. Then, the crayfish optimization algorithm is improved through two improvement strategies to prevent the algorithm from falling into local optima. Finally, the Kolmogorov-Arnold networks (KAN) are introduced to enhance the ConvLSTM, improving the model's feature extraction capability and nonlinear modeling performance. Experimental results demonstrate that the proposed model outperforms current mainstream models in terms of forecasting accuracy, providing reliable decision support for power system dispatch.