A Hybrid Method for Electricity Bill Prediction that Integrates Signal Processing and Model Recognition Under Business Scenario Structure

Electricity revenue is the core source of income for power enterprises. Accurately predicting their daily cash flow is of great significance for daily fund scheduling, stock optimization and financing decisions. In response to the prediction challenges brought about by the complex structure of electricity bill cash flow and the interweaving of multi-source signals, this paper proposes a composite prediction framework that integrates business scenario identification and time series signal processing. This framework deconstructs the capital flow process into a three-stage dynamic system of “issuance - payment - arrival”, introduces multiple external and state signals such as holidays and the arrival time of channel agreements, and builds an adaptive prediction model by identifying the signal characteristics in different business scenarios to accurately capture the cash flow pattern of electricity charges. The experimental results show that the MAE and MAPE metrics of the proposed framework on the test dataset are superior to those of the LSTM, ARIMA and Prophet models.