A CNN-LSTM-based Method for Power Consumption Prediction and Anomaly Detection
The development of power system makes importance to accurately predict power consumption and detect anomalous power usage behaviors in a timely manner.A large number of studies are devoted to improving the accuracy of electricity con-sumption prediction,but few works focus on anomaly detection.Based on this fact,this paper proposes a CNN-LSTM deep learning method by combing a 1-dimensional CNN with an LSTM model to model time-series electricity data with a sliding win-dow approach.The network is trained to use household electricity data collected by ESP32 integrated power sensors.The re-sults of the study show that the mean square error of the CNN-LSTM model is reduced by 29％and the prediction accuracy is improved compared to the LSTM model alone.The study provides a new deep learning approach that can be used to improve the performance of power consumption prediction and grid anomaly detection.
万方数据
维普
