FBiLSTM-Attention short-term load forecasting based on fuzzy logic

Abstract

Aiming at the problem of high uncertainty in power load data due to various factors, a fuzzy logic based FBiLSTM Attention short-term load forecasting model was proposed by combining the uncertainty of load data with deep learning algorithms to improve the accuracy of load forecasting. Firstly, the raw data, including filling in missing values, conducting correlation analysis and normalizing the data, was preprocessed. Secondly, K-Means clustering was used to transform the data of each feature into fuzzy rules and introduce fuzzy logic processing. In terms of model structure, a bi-directional long short-term memory (BiLSTM) and attention mechanism (Attention) were adopted. Finally, the prediction results of the proposed method with traditional LSTM and BiLSTM Attention models were compared. The results show that the model combined with fuzzy logic has significantly improved accuracy and robustness, and has better predictive performance. The proposed model can effectively improve the ability to handle uncertain data, providing reference for load forecasting study.

Keywords

• data processing; fuzzy logic; load forecasting; bi-directional long short-term memory(bilstm); attention mechanism(attention)