APPLYING SEQ2SEQ-LSTM IN A SHORT-TERM LOAD FORECASTING MODEL FOR THE POWER GRID IN TIEN GIANG

Abstract

Short-term load forecasting is critical for energy suppliers to meet the loads of consumers connected to the grid. This study explores the performance of short-term load demand forecasting models, including CNN-LSTM, Wavenet, and Seq2Seq integrating long short-term memory (LSTM). The Seq2Seq-LSTM prediction model is established by combining the sequence-to-sequence (Seq2Seq) structure with the long-short-term neuron model to improve the prediction accuracy. The study validates the models using demand data from the Tien Giang power system from 2021 to 2022, taking into consideration historical demand, holidays and weather variables as input characteristics. The results show that both CNN-LSTM, Wavenet and Seq2Seq-LSTM models can predict future demand with root mean square error (RMSE) and mean absolute percentage error (MAPE). Therefore, the proposed models are not only a useful tool in making smart decisions and planning for future energy needs, but can also play an important role in minimizing Optimize the use of energy resources, minimize negative environmental impacts, and improve the overall performance of the energy sector.