Hybrid KNN-LSTM Modeling for Short-Term Feeder Peak Load Forecasting

Abstract

Load forecasting is important in power system planning and management. Accurate forecasting is key in maintaining the balance of energy supply and demand. This research develops a hybrid KNN-LSTM method for load forecasting using historical load and voltage data. KNN is used in finding local patterns and LSTM is used in capturing long-term patterns. The result is that the KNN-LSTM method provides MSE 30289.4952, RSME 174.0387, and MAE 98.9081. These results are better than the KNN and LSTM methods alone. In addition, by adding the voltage feature, the prediction result increases by 50.5%.

Keywords: Load forecasting, KNN, LSTM, KNN-LSTM