综合智慧能源 ›› 2023, Vol. 45 ›› Issue (6): 66-72.doi: 10.3969/j.issn.2097-0706.2023.06.009

• 电力系统规划 • 上一篇    下一篇

基于决策树的上海商业建筑虚拟电厂执行力分析

熊真真()   

  1. 上海腾天节能技术有限公司,上海 200336
  • 收稿日期:2022-11-04 修回日期:2023-04-11 接受日期:2023-05-09 出版日期:2023-06-25 发布日期:2023-06-14
  • 作者简介:熊真真(1987),男,工程师,从事需求侧管理、虚拟电厂平台设计方面的研究,xzzok@hotmail.com
  • 基金资助:
    上海市科委科研计划项目(20DZ1206200)

Analysis on execution of VPPs for commercial buildings in Shanghai based on decision tree

XIONG Zhenzhen()   

  1. Shanghai Tengtian Energy Conservation Technology Company Limited,Shanghai 200336,China
  • Received:2022-11-04 Revised:2023-04-11 Accepted:2023-05-09 Online:2023-06-25 Published:2023-06-14
  • Supported by:
    Research Project of Science and Technology Commission of Shanghai Municipality(20DZ1206200)

摘要:

位于上海市中心的黄浦区商业建筑虚拟电厂是一座零排放的“电厂”,它是以商业建筑在用电高峰时期通过自主控制降低用电负荷,以“数据”为燃料实现“发电”的电厂。由于中心城区商业建筑分布广阔,不同建筑内的用电设备种类繁多,管理人员能力也不尽相同,使虚拟电厂在执行需求响应任务时虚拟发电机执行能力存在一系列问题。将近年来黄浦区商业建筑虚拟电厂中50幢商业楼宇参与需求响应真实事件作为原始数据,使用决策树算法对虚拟电厂中虚拟发电机组的执行力进行分析,从而提高需求响应事件中虚拟电厂整体发电执行力的预测精准度。通过该算法的训练和验证,在当前场景下准确度达到72.3%,且通过随机森林、贝叶斯等方法优化后有进一步提高精准度的可能。决策树模型算法在该场景下有较好的应用价值,可用于虚拟电厂执行能力预测。

关键词: 商业建筑, 虚拟电厂, 决策树, 信息熵, 需求响应, 需求侧管理, 零排放

Abstract:

A commercial building virtual power plant(VPP)lying in Huangpu district,the centre of Shanghai, is a zero-emission power plant. The VPP can lower the power load of commercial buildings during peak periods through autonomous control, that is to say, it can generate power with "data" as fuels. The wide distribution of commercial buildings in central urban areas, the wide variety of electrical equipment in different buildings, and the diverse capabilities of management personnel lead to various problems to VPPs in performing demand-response tasks. Based on the raw data from 50 commercial buildings participating in demand responses via the VPP in Huangpu district, the execution capacity of the virtual generator set in the VPP is analysed by decision tree algorithm, to improve the prediction accuracy of the VPP’s power generation execution in demand responses. Through the training and verification of the algorithm, it shows that the accuracy of the algorithm reaches 72.3% under current scenarios, and it is possible to further improve the accuracy by taking random forest, Bayesian and other algorithms in optimization. Therefore, the decision tree model algorithm is considered to be of good application value in the scenario, and can be used to predict the execution capacity of VVPs.

Key words: commercial building, virtual power plant, decision tree, information entropy, demand response, demand-side management, zero emission

中图分类号: