引用本文: | 吴晨,姚菁,薛贵元,王剑晓,吴垠,何凯.基于MMoE多任务学习和长短时记忆网络的综合能源系统负荷预测[J].电力自动化设备,2022,42(7): |
| WU Chen,YAO Jing,XUE Guiyuan,WANG Jianxiao,WU Yin,HE Kai.Load forecasting of integrated energy system based on MMoE multi-task learning and LSTM[J].Electric Power Automation Equipment,2022,42(7): |
|
摘要: |
精确的多元负荷预测是实现综合能源系统优化调度与经济运行的关键技术。在考虑多元负荷相关性的基础上,提出一种基于MMoE多任务学习和长短时记忆网络(LSTM)的多元负荷预测方法。利用皮尔逊相关系数分析冷热电负荷及气象因素存在的强相关性和弱相关性;构建MMoE多任务学习模型,利用专家子网和门控单元学习多元负荷间耦合特性的差异;使用LSTM构建子任务模型,对多元负荷进行预测。利用公开数据集进行性能验证,结果表明所提基于MMoE多任务学习和LSTM的模型能够有效提升多元负荷预测精度。 |
关键词: 多元负荷预测 综合能源系统 相关性分析 MMoE多任务学习 长短时记忆网络 专家网络 |
DOI:10.16081/j.epae.202204083 |
分类号:TM73;TK01 |
基金项目:国家电网公司科技项目(1400-202118231A-0-0-00) |
|
Load forecasting of integrated energy system based on MMoE multi-task learning and LSTM |
WU Chen1, YAO Jing2, XUE Guiyuan1, WANG Jianxiao3, WU Yin1, HE Kai2
|
1.Economic Research Institute of State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210008, China;2.Beijing Tsintergy Technology Co.,Ltd.,Beijing 100080, China;3.State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Beijing 102206, China
|
Abstract: |
Accurate multivariate load forecasting is the key to realize optimal scheduling and economic opera-tion of IES(Integrated Energy System). On the basis of considering the correlation of multivariate loads, a multivariate load forecasting method based on MMoE(Multi-gate Mixture-of-Experts) multi-task learning and LSTM(Long Short-Term Memory network) is proposed. The Pearson correlation coefficient is used to analyze the strong and weak correlation between cooling, heating, electric load and meteorological factors. Then, the MMoE multi-task learning model is constructed, and the expert subnetworks and gating units are used to learn the difference of coupling characteristics among multivariate loads. Moreover, the subtask model is constructed using LSTM to forecast multivariate loads. The performance is validated by public datasets, and results show that the proposed model based on MMoE multi-task learning and LSTM can effectively improve the accuracy of multivariate load forecasting. |
Key words: multivariate load forecasting integrated energy system correlation analysis MMoE multi-task lear-ning LSTM expert network |