Electricity Consumption Prediction for Enterprises Based on Neural Network Algorithms

This paper investigates electricity consumption forecasting using monthly data from 37 enterprises between January and September 2024. Initially, cubic spline interpolation is employed to smooth the data into a continuous functional form. Functional Principal Component Analysis (FPCA) is then applied for dimensionality reduction, facilitating the extraction of significant features. Given the similarities in characteristics across certain enterprises, the study adopts the k -means clustering algorithm to categorize the enterprises into four groups. The electricity consumption time series for each group are subsequently predicted using the Long Short-Term Memory (LSTM) model, thus avoiding the redundancy of modeling each enterprise individually and improving forecasting efficiency.