首页|基于融合集成算法的配电网负荷预测研究

基于融合集成算法的配电网负荷预测研究

扫码查看
原文链接
  • 万方数据
  • 维普
配电网的负荷预测在电力运行状态监测中尤为重要.负荷预测的精度提升为电网的安全、稳定运行提供了保障.通过对融合集成算法的研究,提出了一种基于关联特征选择的融合集成算法.在数据集的选择上,使用相关系数和灰色关联算法综合对样本中负荷影响较小的特征进行剔除,使得样本数据集的相关性更高;同时,对传统Stacking集成学习的输入和输出特征进行优化,提高了模型的预测效果.试验结果表明,基于融合集成算法的配电网负荷预测模型与传统的Stacking集成算法、XGBoost、灰狼优化-反向传播算法相比,负荷预测的精度提升了 3.07％.该模型总体性能表现较好.该研究结果有效地支撑了配电网的负荷监测和规划,也为电力系统故障诊断提供了参考.
Research on Load Forecasting of Distribution Networks Based on Fusion Integration Algorithm
Load forecasting of distribution networks is particularly important in power operation condition monitoring.The improvement of the accuracy of load forecasting provides a guarantee for the safe and stable operation of power grids.Through the study of fusion integration algorithm,a fusion integration algorithm based on correlation feature selection is proposed.In the selection of the data set,the features with less load influence in the samples are eliminated using the correlation coefficient and the gray correlation algorithm in a comprehensive way,which makes the sample data set more relevant;at the same time,the input and output features of the traditional Stacking integration learning are optimized,which improves the prediction effect of the model.The experimental results show that the load prediction model for distribution networks based on the fusion integration algorithm improves the accuracy of load prediction by 3.07％compared with the traditional Stacking integration algorithm,XGBoost,gray wolf optimization-back propagation algorithm.The overall performance of the model is better.The results of the research effectively support the load monitoring and planning of distribution networks,and also provide a reference for power system fault diagnosis.

Intelligent algorithm fusionDistribution networksIntegrated learningLoad forecastingStacking integrated learningXGBoostFeature extractionTarget prediction

李强、赵峰、吴金淦、谭守标

展开 >

国网信息通信产业集团有限公司,北京 102200

安徽继远软件有限公司,安徽合肥 230088

安徽大学集成电路学院,安徽合肥 230601

智能算法融合 配电网 集成学习 负荷预测 Stacking集成学习 XGBoost 特征提取 目标预测

国网信通产业集团两级协同研发基金资助项目

T1821011650

2024

10.16086/j.cnki.issn1000-0380.2022110039

自动化仪表
中国仪器仪表学会 上海工业自动化仪表研究院

自动化仪表

CSTPCD
影响因子：0.655
ISSN：1000-0380
年,卷(期)：2024.45(1)
  • 12