Abstract: The power system requires an immediate balance between the generated power and the electricity load, which is characterized by non-linearity, time variability, and uncertainty.To address this issue, this paper proposes a combined forecasting model that integrates wavelet transform（WT）and long short-term memory（LSTM）neural networks, considering the impact of weather and date types for short-term power load forecasting.Initially, the wavelet transform is employed for feature extraction signal denoising to reduce data volatility.Then, the preprocessed data is trained using an LSTM network, and the output results undergo sequence reconstruction for the final load forecast.Finally, the data of WT-LSTM combined forcasting model is seperately compared with that of the BP neural network and LSTM model.The results show that the WT-LSTM neural network combined prediction model has the superior predictive performance, significantly enhancing forecasting precision.