Abstract: Load forecasting is a crucial task in the operation and planning of power systems. To address the challenges posed by the nonlinearity and multiple influencing factors of power load data, a load forecasting model based on bimodal decomposition, deep learning, and Attention mechanism is presented. Firstly, empirical mode decomposition(EMD) is applied to the input data, and K-means clustering is used to group similar components, resulting in three composite components. Secondly, variational mode decomposition(VMD) is employed to further decompose the composite components into distinct modes, and sparrow search algorithm(SSA) is utilized to optimize the parameters of variational mode decomposition. Subsequently, the VMD-derived components are combined with influencing factors and fed into a long short-term memory(LSTM) network. An Attention mechanism is employed to uncover internal data correlations, and SSA is used to optimize the parameters of the LSTM network. Finally, the model is validated using load data from a power station in Ningxia over the course of a year. Comparative analysis with different models demonstrates that the proposed model achieves higher forecasting accuracy.