Forecasting Short-term Power Grid Load Based on Recurrent Neural Network

Abstract

Electric power is an important energy source for national development. Power resources are not only closely related to people's lives, but also related to a country's economy. In addition, the three main processes of power generation, power transmission, and power consumption must be highly coupled together. In recent years, domestic power consumption has gradually increased, causing huge power consumption, which may lead to unexpected power outages and other major accidents. The relevant part must strike a balance between power supply quality and economic benefits. The predicted quality directly affects the economic feasibility and reliability of each power company. In this paper, combining the real power load data of a power plant in Slovakia and the real power load data of a certain region of my country, the two-layer LSTM model, the LSTM model combined with the Attention mechanism, and the two-branch LSTM model are proposed and verified. Based on the experimental results of the two datasets, this paper finds that because the two-branch LSTM model has the highest model complexity, it has the best fit to the load data, and the prediction results are the most accurate.