基于多目标优化组合模型的电力负荷预测  

Power load forecasting based on the multi-objective optimal combination model

作　　者：刘刚[1] 李嘉翔 魏文浩 海轩 LIU Gang;LI Jia-xiang;WEI Wen-hao;HAI Xuan(School of Information Science and Engineering,Lanzhou University,Lanzhou 730000,China)

机构地区：[1]兰州大学信息科学与工程学院,兰州730000

出　　处：《兰州大学学报（自然科学版）》2023年第3期380-386,共7页Journal of Lanzhou University(Natural Sciences)

基　　金：国家社会科学基金重大项目(21&ZD163)。

摘　　要：提出一种基于多目标优化的组合模型(CM),对短期电力负荷需求进行预测.将电力负荷数据预处理后,结合神经网络算法和改进后的多目标蜻蜓算法优化模型的权重.将CM分别通过2018、2019年的电力负荷数据集进行分析验证.在2019年电力负荷时间序列数据集的预测中,CM预测的平均绝对误差百分比MApe平均值为0.63%,与MApe最高的H-Elman算法相比低3.28%.结果表明,CM在短期电力负荷预测中具有较高的准确率.对CM与实验分析中涉及的基准模型进行迪堡马里亚诺检验,CM和绝大多数的基准模型得到的统计量值均大于99%的置信概率临界值.A combination model(CM)based on multi-target optimization was proposed to predict the demand for a short-term power load.After pre-processing the power load data,this study combined neural network algorithms and improved multi-target dragonfly algorithms to optimize the weight of the mod-el,which was used to analyze and verify the models on the 2018 and 2019 power load data sets.Among them,in the prediction of the power load time sequence data set in 2019,the average mean absolute percentage error(MApE)of CM prediction was 0.63%.Compared with H-Elman algorithm,which has the highest MApe,the MApe predicted by CM was reduced by 3.28%.The results showed that the model had a high accuracy in short-term power load forecast.The Diebold Mariano(DM)test was carried out on CM and the benchmark model involved in the experimental analysis.The DM statistic values obtained by CM and most of the benchmark models were greater than 99%confidence probability critical value.

关 键 词：多目标优化 组合模型 电力负荷 迪堡马里亚诺检验 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]