Year: 2024
Type: book-chapter
Abstract: In recent years, many industrial and commercial users have begun to participate in demand response, and demand response measures will affect future load demand, thus affecting the accuracy of load forecasting. Prediction problems for these particular users, this paper describes the implementation of an industrial user load forecasting model considering regulation and incentive by using the time series characteristics of historical data of power load and the influence of reward mechanism on the power demand behavior of industrial users. Firstly, the influence of production cost and reward mechanism on the power utilization behavior of industrial users is investigated. Then, the feature engineering, data preprocessing, improved Adam optimization algorithm, LSTM neural network algorithm, and model evaluation index of industrial user load forecasting are described. Finally, the effect of this method in actual industrial user load forecasting is demonstrated. The power load data of a printing and dyeing enterprise in East China is used for verification. Through the analysis of actual simulation verification, with a mean absolute percentage error (MAPE) of 3.12% on the training set and 5.22% on the test set, respectively, as shown by the experimental results, it has high prediction accuracy.
DOI: 10.3233/faia231323
Authors: Qiang Ye, Yuqing Cai, Wuxiao Chen, Zhijun Jiang, Jiaxin Cai, Kai Liu
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4391753664