基于遗传算法优化Seq2Seq模型的建筑空调电能预测研究  

Research on energy prediction of building air conditioning based on genetic algorithm optimization Seq2Seq model

作　　者：李鹏 肖辉[2] 李秀花[3] 赵恒 邵頲 LI Peng;XIAO Hui;LI Xiuhua;ZHAO Heng;SHAO Ting(East China Electric Power Design Institute Co.,Ltd.,of CPECC,Shanghai 200063,China;Department of Electronic and Information Engineering,Tongji University,Shanghai 201804,China;China Haisum Engineering Co.,Ltd.,200031,China;Shanghai Shuowu Tiancheng Information Technology Co.,Ltd.,Shanghai 200052,China;Shanghai Huajian Architecture,Engineering&Consulting Co.,Ltd.,Shanghai 200041,China)

机构地区：[1]中国电力工程顾问集团华东电力设计院有限公司,上海200063 [2]同济大学电子与信息工程学院,上海201804 [3]中国海诚工程科技股份有限公司,上海200031 [4]上海硕物天成信息科技有限公司,上海200052 [5]上海华建工程建设咨询有限公司,上海200041

出　　处：《电气应用》2025年第2期30-35,共6页Electrotechnical Application

摘　　要：介绍了一种建筑空调电能预测方法。空调电能占建筑总能耗的50%以上,空调电能的预测有利于可再生能源的安装容量预测和电力系统的负荷预测。采用遗传算法自适应搜索Seq2Seq模型的最优超参数,由此生成输出序列。通过与已有常见预测方法进行对比,证实了GA-Seq2Seq模型在MAE、RMSE 2和R等指标中表现优异。此外,还研究了GA-Seq2Seq模型的超参数和分类标签设置效果。This paper introduces a method of electric energy prediction for building air conditioning.The air-conditioning electric energy accounts for more than 50%of the total energy consumption of buildings,and the prediction of air-conditioning electric energy is beneficial to the installed capacity prediction of renewable energy and the load prediction of power system.In this paper,genetic algorithm is used to search the optimal hyperparameters of Seq2Seq model and generate the output sequence.In this research,it is confirmed that GA-Seq2Seq model has 2 excellent performance in MAE,RMSE,R and other indicators by comparing with the existing common forecasting methods.The hyperparameter and classification label setting effects of GA-Seq2Seq model are also studied.

关 键 词：遗传算法 Seq2Seq模型 空调电能 能耗预测 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程] TP18[自动化与计算机技术—控制理论与控制工程] TM933.4[自动化与计算机技术—控制科学与工程]