中国电力 ›› 2021, Vol. 54 ›› Issue (2): 78-89.DOI: 10.11930/j.issn.1004-9649.202001039

• 新能源 • 上一篇    下一篇

基于CNN的产消群需求响应滚动优化策略

张旭东1, 李飞1, 刘迪2, 孙毅2, 李彬2   

  1. 1. 国网河北省电力有限公司,河北 石家庄 050021;
    2. 华北电力大学 电气与电子工程学院,北京 102206
  • 收稿日期:2020-01-07 修回日期:2020-07-22 发布日期:2021-02-06
  • 作者简介:张旭东(1974-),男,高级工程师(教授级),从事电力系统营销研究;李飞(1982-),男,高级工程师(教授级),从事电力系统营销研究;刘迪(1990-),男,通信作者,博士研究生,从事能源互联网、需求响应研究,E-mail:kfliudi@163.com;孙毅(1972-),男,博士,教授,从事能源互联网及其信息通信技术,物联网及现代传感技术研究,E-mail:sy@ncepu.edu.cn;李彬(1983-),男,博士,副教授,从事电气信息技术、自动需求响应相关技术研究,E-mail:direfish@163.com
  • 基金资助:
    国家自然科学基金资助项目(可再生能源接入下的大规模负荷感知模型及调控策略研究,51777068);国家电网公司科技项目 (低压用户负荷感知、测量和调控关键技术研究及应用,SGHEDK00DYJS2000044)

CNN-based Rolling Optimization Strategy for Prosumer Group in Demand Response

ZHANG Xudong1, LI Fei1, LIU Di2, SUN Yi2, LI Bin2   

  1. 1. State Grid Hebei Electric Power Co., Ltd., Shijiazhuang 050021, China;
    2. School of Electrical and Electronic Engineering, North China Electric Power University, Beijing 102206, China
  • Received:2020-01-07 Revised:2020-07-22 Published:2021-02-06
  • Supported by:
    This work is supported by the National Natural Science Foundation of China (Research on Large Scale Load Sensing Model and Control Strategy with Renewable Energy Access, No.51777068) and Science & Technology Project of SGCC (Research and Application of Key Technologies of Load Sensing, Measurement and Regulation for Low Voltage Users, No.SGHEDK00DYJS2000044)

摘要: 随着分布式光伏的普及,具有负荷电源双重属性的产消者大量出现。在电力现货市场的环境下,首先,基于Stackelberg模型分析产消者以及售电商在价格型需求响应中的决策机理,提出基于卷积神经网络(convolutional neural networks, CNN)的产消群决策行为预测模型,实现对产消群负荷调整及反弹量的精准预测。然后,充分考虑负荷反弹现象对于价格型需求响应的影响,提出考虑负荷反弹的滚动优化策略,提升售电商收益的同时,降低产消群负荷不平衡量,并促进分布式光伏的就地消纳。仿真实验表明,所提产消者决策行为预测模型的预测准确度在99%以上,且考虑了负荷反弹现象的滚动优化策略能够将产消群可再生能源就地消纳率提升5%以上,同时,相较于分时电价以及不考虑负荷反弹的实时电价,售电商的收益分别提升了118.8%和15.1%。

关键词: 产消者, 需求响应, 卷积神经网络, 滚动优化, 遗传算法

Abstract: With the massive installation of distributed photovoltaics, a large number of prosumers with dual attributes of load and power supply have appeared. In the environment of electric power spot market, the decision mechanism of prosumers and electricity sellers in the price-based demand response are analyzed based on the Stackelberg model, and a convolutional neural networks(CNN)-based prosumer group decision behavior prediction model is proposed to achieve accurate prediction of load adjustment and rebound of prosumers. Furthermore, with full consideration of the impact of the load rebound phenomenon on the price-based demand response, a rolling optimization strategy considering load rebound is proposed to increase the revenue of electricity sellers, reduce the load imbalance of prosumers, and promote the local consumption of distributed photovoltaics. Simulation experiments show that the prediction accuracy of the prosumers decision behavior prediction model proposed in this paper is more than 99%, and the rolling optimization strategy considering the load rebound phenomenon can increase the prosumers’ local consumption rate of renewable energy by more than 5%. At the same time, compared to the time-of-use electricity prices and the real-time electricity prices that do not take load rebound into consideration, the revenue of electricity sellers can increase by 118.8% and 15.1%, respectively.

Key words: prosumer, demand response, CNN, rolling optimization, genetic algorithm