基于多因素组合分析的电力系统长期负荷预测研究  

Research on Long-Term Load Forecasting of the Power System Based on Multi-Factor Combination Analysis

作　　者：厉瑜 益西措姆 杜宁刚 达娃央宗 郭彦君 王进仕 LI Yu;YIXI Cuomu;DU Ninggang;DAWA Yangzong;GUO Yanjun;WANG Jinshi(State Grid Xizang Electric Power Co.,Ltd.,Lasa 850000,Xizang,China;School of Energy and Power Engineering,Xi’an Jiaotong University,Xi’an 710054,Shaanxi,China)

机构地区：[1]国网西藏电力有限公司,西藏拉萨850000 [2]西安交通大学能源与动力学院,陕西西安710054

出　　处：《电网与清洁能源》2024年第7期81-87,94,共8页Power System and Clean Energy

基　　金：国网西藏电力有限公司科技项目(SGXZJY00JHJS2000007)。

摘　　要：电力系统长期负荷预测影响因素较多,仅利用单一因素进行负荷预测的精度较低,因此提出基于多因素组合分析的电力系统长期负荷预测方法。通过可辨识矩阵采集电力负荷数据后,利用ACO-PAM综合算法对电力数据进行聚类分析,获取有价值的负荷数据;将聚类获取电力负荷数据经数据类因素量化和非数据类因素量化处理后,分析多种因素与负荷的相关性,将获取的多因素作为遗传算法改进神经网络的输入,输出电力系统长期负荷预测结果。实验结果表明:在多因素的影响下,该方法的电力系统长期负荷预测结果逼近实际值;与2种对比方法相比,其平均绝对误差分别小130.98、41.65万吨标准煤,平均相对误差分别小3.77%、1.19%,说明所提方法预测效果好。Many factors influence long-term load forecasting of the power system,but the accuracy of the load forecasting using only a single factor is often low.Therefore,a power system long-term load forecasting method based on multi factor combination analysis is proposed in this paper.After the power load data is collected through the discernibility matrix,the ACO-PAM comprehensive algorithm is used to cluster the power data to obtain valuable load data.After the power load data obtained by clustering is quantified by data factors and non data factors,the correlation between multiple factors and load is analyzed,and the obtained multiple factors are used as the input of genetic algorithm to improve neural network to output the long-term load forecasting results of the power system.The experimental results show that under the influence of many factors,the long-term load forecasting result of the power system based on this method is close to the actual value.Compared with the two comparison methods,the average absolute error of this method is 1309800 and 416500 tons of standard coal respectively,and the average relative error is 3.77%and 1.19%less than the two comparison methods,respectively.

关 键 词：多因素 组合分析 电力系统 长期负荷预测 数据聚类 神经网络 

分 类 号：TM714[电气工程—电力系统及自动化]