Research on Integrated Learning Electric Load Forecasting Model Based on Transformer Model

Towards the goal of carbon neutrality, the exploration and application of low-carbon energy technologies have become particularly important. The goal of power forecasting is to optimize resource allocation, reduce energy losses, and reduce operating costs while meeting power demand. In this context, the accuracy of power load forecasting has become a core issue in the operation and management of power systems. The general forecasting method is to deal with non-lines and complex time series data often face limitations. In this work, an ensemble learning framework based on Transformer model is proposed, which combines LSTM and XGBoost model to improve the accuracy of power load forecasting. The results show that the proposed model is significantly better than the transform, XGBoost and LSTM methods alone in terms of mean absolute error (MAE) and coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>).