Monthly Electricity Revenue Prediction Based on the Feature Fusion Weighted Temporal Causal Convolutional Network
Lists
Only you see this
Abstract
With the deepening of the electricity market reform, improving the prediction accuracy of electricity revenue is crucial for the fund management of power supply enterprises. In view of the problem that existing prediction methods do not consider the complex relationships among various factors and the different influence factors of different factors, resulting in large prediction deviations, this paper proposes a deeplearning model named Feature Fusion Weighted Temporal Causal Convolutional Network (FFW-TCN) model for monthly electricity revenue prediction. The model adaptively learns to dynamically capture the influence factors of various factors such as electricity sales volume, number of users, temperature, and holidays through the Temporal Feature Fusion Weighted (FFW) module, and uses the Temporal Causal Convolutional Network (TCN) to capture long-term time series dependencies. Comparative experiments based on the monthly data of a power supply company from 2022 to 2024 show that the model proposed in this paper is significantly better than the benchmark model, with an average reduction of 8.368% in the mean absolute error compared with the traditional prediction model, which can provide accurate decision support for the fund scheduling and risk prevention and control of power supply enterprises.
BibTeX
@misc{zhang2025monthly,
author = {Tianyi Zhang and Ru Liu and Rong Wang and Hua Xu and Feihu Hu},
title = {{Monthly Electricity Revenue Prediction Based on the Feature Fusion Weighted Temporal Causal Convolutional Network}},
year = {2025},
doi = {10.1109/iceeps66790.2025.11239802},
}Loading related content...