Meteorological prediction has important application value in the energy industry and related fields. Accurate prediction of meteorological factors, such as sunshine intensity, ambient temperature and wind speed, is crucial to the operation of power transmission lines and the safety of equipment. However, traditional meteorological prediction methods may suffer from missing data and outliers, as well as challenges in model accuracy and generalization capabilities. This study uses the Encoder-Decoder LSTM network to predict the original wind speed, ambient temperature, and sunshine intensity of power transmission lines, by applying the Bi-LSTM neural network as the encoder of the network and the LSTM neural network as the decoder of the network. Through processing and predictive analysis of data collected from six overhead transmission lines of Hunan Power Grid, we verified that the model has strong feature extraction capabilities for long-term series, and the model has good generalization on unknown data. Not only can it accurately predict key meteorological parameters such as temperature, sunshine intensity and wind speed, but the deviation of the prediction data is small and has good validity and accuracy. It provides reliable meteorological information for the power industry and related fields, especially the application of dynamic capacity expansion technology. method of prediction.