Research on Prediction Models and Optimization Methods for Electrical Current Consumption of Users

During online monitoring of users’ electrical current consumption data, data missing phenomena occasionally occur in the frequently collected current data of users’ electricity usage. To effectively estimate the sporadically missing electrical current data from users and ensure the completeness of the collected data, this paper introduces a k-estimation model for missing current data. This model takes the recent k observed current data as input and outputs the predicted value for the missing data. The model is constructed based on current data using a machine learning approach and is evaluated using two metrics: mean squared error (MSE) and effective ratio (ER). Experiments demonstrate that, compared to the k-estimation models for missing current data implemented using BP neural networks and RBF neural networks, the k-estimation model for missing current data implemented using an Adam-optimized LSTM neural network can be more effectively applied for estimating missing current data during frequent collection of users’ electrical current consumption data.