引用本文:胡实,唐昊,吕凯,杨晨芳.考虑广义需求侧资源的深度置信网络短期负荷预测方法[J].控制理论与应用,2023,40(3):493~501.[点击复制]
HU Shi,TANG Hao,LV Kai,YANG Chen-fang.Short-term load forecasting method of deep belief network by considering generalized demand-side resources[J].Control Theory and Technology,2023,40(3):493~501.[点击复制]
考虑广义需求侧资源的深度置信网络短期负荷预测方法
Short-term load forecasting method of deep belief network by considering generalized demand-side resources
摘要点击 940  全文点击 320  投稿时间：2021-03-13  修订日期：2022-08-10
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2021.10209
  2023,40(3):493-501
中文关键词  短期负荷预测  广义需求侧资源  深度置信网络  负荷聚合商
英文关键词  short-term load forecasting  generalized demand side resources  deep belief network  load aggregator
基金项目  国家电网有限公司总部科技项目“弹性环境下基于深度学习的智能调度技术”(SGTYHT/19–JS–215)资助.
作者单位E-mail
胡实 合肥工业大学 hushi@mail.hfut.edu.cn 
唐昊* 合肥工业大学 htang@hfut.edu.cn 
吕凯 合肥工业大学  
杨晨芳 合肥工业大学  
中文摘要
      随着智能电网信息化水平的不断提高以及可控负荷、分布式电源和储能等广义需求侧资源的大量接入, 将 产生海量负荷数据集并改变负荷特性. 为了提高负荷预测精度, 提出了一种考虑广义需求侧资源的深度置信网络 (DBN)负荷预测方法. 首先, 借助负荷聚合商确定了广义需求侧资源参与电力市场的机制, 构建了基于合同的广义 需求侧资源调度模型, 并利用该模型求解广义需求侧资源参与电力市场的最优调度计划. 其次, 引入DBN结构, 并 将广义需求侧资源的最优调度计划作为其输入量, 建立了短期负荷预测模型. 最后, 以实际数据进行仿真测试, 结果 表明, 本文所提方法具有更高的预测精度.
英文摘要
      With the continuous improvement of the informatization level of smart grid and the massive access of generalized demand-side resources such as controllable load, distributed power sources and energy storage, massive load data sets will be generated and load characteristics will be changed. In order to improve the load forecasting accuracy, a deep belief network (DBN) load forecasting method considering generalized demand-side resources is proposed. Firstly, the mechanism of the participation of generalized demand-side resources in the power market is determined with the help of load aggregators, and a contract-based generalized demand-side resource scheduling model is constructed, which determines the optimal scheduling plan of generalized demand-side resources participating in power market. Then, a short-term load forecasting model is established by introducing the DBN structure and taking the optimal scheduling plan as its input. Finally, a simulation test is conducted with the actual data, and the results show that the proposed method has higher prediction accuracy.