Research on Electricity Consumption Prediction Model Based on Long Short-Term Memory Neural Networks

With the rapid development of smart grid technology, household intelligent electricity usage has become an important aspect of smart grid systems. Currently, the method of collecting customer electricity data mainly involves measuring electricity consumption through smart meters installed at the household level, which then upload the data to cloud servers. This approach can lead to decreased server efficiency due to the increasing number of users, resulting in users being unable to obtain detailed electricity information, and power companies being unable to effectively analyze future electricity consumption trends. To address this issue, this paper proposes a deep learning-based electricity consumption prediction model, focusing on the total electricity consumption of households to forecast usage over a future period. The model is centered around Long Short-Term Memory (LSTM) neural networks and employs Convolutional Neural Networks (CNN) and Attention mechanisms for data feature analysis, resulting in a Bidirectional Long Short-Term Memory (BLSTM-Attention) model. It then derives multidimensional electricity features from multivariable consumption data, which are input into the BLSTM-Attention model to predict electricity consumption for different future time periods. Finally, through simulation experiments, the model is compared with VAR algorithm models and LSTM algorithm models. The experimental results indicate that this hybrid model algorithm outperforms current deep learning algorithms, providing a more accurate prediction of future electricity consumption trends.