Scholarlink

<p>Load forecasting for diverse customers in different industries plays a crucial role in the electricity market, helping power system operators optimize scheduling strategies, reduce operating costs, and improve the operational efficiency of the power system. Due to the fact that user loads in different industries are affected by weather factors to different degrees and have different characteristics of electricity consumption behavior, traditional forecasting methods face significant challenges in dealing with the complex and variable forecasting environment. For this reason, this paper proposes a multi-user time-sharing load forecasting method based on the advantageous fusion of multivariate heterogeneous data to improve the accuracy and stability of the forecast results. First, by analyzing the power consumption characteristics of users in different industries, identifying typical power consumption behavior patterns and their discriminative basis, and realizing the classification of industry users based on the multivariate feature clustering model. Then, we quantitatively evaluate the influence of each meteorological feature on the load of users in different industries by using the maximal information coefficient, and screen and reconstruct the key meteorological features with significant influence. Finally, the Transformer model is constructed to integrate multiple heterogeneous features, and the hidden layer state extracted by temporal neural network is used as the input of encoder, and the reconstructed meteorological features are used as the input of decoder, so as to fully integrate the intrinsic temporal cycle features and extrinsic meteorological dependence features of the time-sharing loads to realize the time-sharing loads of users in different industries. It realizes accurate time-sharing load prediction for users in different industries. The experimental results show that the reconstructed prediction model shows significant advantages in different industries, and its prediction accuracy is improved compared with the traditional model.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-User Time-Sharing Load Prediction Based on Advantageous Fusion of Multiple Heterogeneous Data

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Wensheng Li,
  • Rui Lu,
  • Lidanlan Tao,
  • Min Jiao,
  • Yankun Feng,
  • Qiang Li

摘要

Load forecasting for diverse customers in different industries plays a crucial role in the electricity market, helping power system operators optimize scheduling strategies, reduce operating costs, and improve the operational efficiency of the power system. Due to the fact that user loads in different industries are affected by weather factors to different degrees and have different characteristics of electricity consumption behavior, traditional forecasting methods face significant challenges in dealing with the complex and variable forecasting environment. For this reason, this paper proposes a multi-user time-sharing load forecasting method based on the advantageous fusion of multivariate heterogeneous data to improve the accuracy and stability of the forecast results. First, by analyzing the power consumption characteristics of users in different industries, identifying typical power consumption behavior patterns and their discriminative basis, and realizing the classification of industry users based on the multivariate feature clustering model. Then, we quantitatively evaluate the influence of each meteorological feature on the load of users in different industries by using the maximal information coefficient, and screen and reconstruct the key meteorological features with significant influence. Finally, the Transformer model is constructed to integrate multiple heterogeneous features, and the hidden layer state extracted by temporal neural network is used as the input of encoder, and the reconstructed meteorological features are used as the input of decoder, so as to fully integrate the intrinsic temporal cycle features and extrinsic meteorological dependence features of the time-sharing loads to realize the time-sharing loads of users in different industries. It realizes accurate time-sharing load prediction for users in different industries. The experimental results show that the reconstructed prediction model shows significant advantages in different industries, and its prediction accuracy is improved compared with the traditional model.