The digital twin technology of the distribution network is an important product resulting from the integration and development of the power system and information technology. The technology simulates the physical behavior and operational status of the distribution network in a digital space by constructing a virtual model of the physical distribution network, enabling comprehensive simulation and analysis. Due to the diverse systems and complex states involved, the existing digital twin simulation platform technology for distribution networks still requires improvement. A wavelet-LSTM fusion model for power state and weather factors was constructed based on the existing wavelet transform and long short-term memory (LSTM) neural network. The high-dimensional input data were converted into detail and contour coefficients using discrete wavelet transform. Subsequently, LSTM neural networks were constructed to process the data and fuse the results, thereby forming accurate prediction outcomes. This method was validated on real datasets, showing that the wavelet-LSTM fusion model significantly improves the mean absolute percentage error (MAPE) compared to the existing LSTM network. Additionally, the method was tested on datasets from different industries. Compared to wavelet-Lasso, LSTM, and STL-LSTM, it exhibits better performance in terms of MAPE, demonstrating that the wavelet LSTM prediction method can be applied to state data from various sectors, thereby providing robust support for future state prediction of digital twins.