Early Warning Method for Overload of Massive Transformer Based on TL-LSTM Model

To address the issues of low prediction accuracy and limited capability in handling large datasets in power load forecasting methods, this paper proposes a novel load forecasting model based on transfer learning (TL) and long short-term memory (LSTM) neural networks. Firstly, using the maximum mutual information coefficient method (MIC) to select high relevance load features; Secondly, a load prediction model is constructed by employing transfer learning and long short-term memory networks to effectively predict the loads of a large number of substations; Finally, using the predicted load rate and the heavy overload assessment criteria, the final heavy overload warning results for distribution substations are determined. Compared with the traditional method, this method can greatly improve the model's ability to handle big data and retain the timeliness of load forecasting while maintaining a certain level of forecasting accuracy.