基于生成对抗网络的综合能源负荷场景生成方法

doi:10.12204/j.issn.1000-7229.2021.12.001

电力建设 ›› 2021, Vol. 42 ›› Issue (12): 1-8.doi: 10.12204/j.issn.1000-7229.2021.12.001

• 综合多元能源及信息技术的能源互联网规划与运行·栏目主持刘洋副教授、韩富佳博士· • 上一篇下一篇

基于生成对抗网络的综合能源负荷场景生成方法

朱庆¹(), 郑红娟¹(), 唐子逸²(), 韦思雅¹(), 邹子骁³(), 吴熙³()

1.国电南瑞科技股份有限公司,南京市 211106
2.国网浙江杭州市余杭区供电有限公司,杭州市 311100
3.东南大学电气工程学院,南京市 210096

收稿日期:2021-06-08 出版日期:2021-12-01 发布日期:2021-11-26
通讯作者: 吴熙 E-mail:dreamathstat@163.com;451173096@qq.com;25804126@qq.com;weisiya_bess@163.com;961498170@qq.com;wuxi@seu.edu.cn
作者简介:朱庆(1981),男,博士,高级工程师,主要从事虚拟电厂、人工智能、5G通信、电力信息通信方面的研究工作,E-mail: dreamathstat@163.com;
郑红娟(1989),女,工程师,长期从事综合能源、需求响应业务方向研究工作,E-mail: 451173096@qq.com;
唐子逸(1994),男,硕士研究生,主要从事综合能源系统等方面的研究工作,E-mail: 25804126@qq.com;
韦思雅(1997),女,硕士研究生,主要从事虚拟电厂等方面的研究工作,E-mail: weisiya_bess@163.com;
邹子骁(1998),男,硕士研究生,主要从事人工智能在电力系统中应用方面的研究工作,E-mail: 961498170@qq.com
基金资助:
国家电网有限公司总部科技项目“大规模综合能源计量系统数字与真型混合仿真关键技术研究”(5600-201955167A-0-0-00)

Load Scenario Generation of Integrated Energy System Using Generative Adversarial Networks

ZHU Qing¹(), ZHENG Hongjuan¹(), TANG Ziyi²(), WEI Siya¹(), ZOU Zixiao³(), WU Xi³()

1. NARI Technology Co., Ltd., Nanjing 211106, China
2. Hangzhou Yuhang District Power Supply Company of State Grid Zhejiang Electric Power Co., Ltd., Hangzhou 311100, China
3. School of Electrical Engineering, Southeast University, Nanjing 210096, China

Received:2021-06-08 Online:2021-12-01 Published:2021-11-26
Contact: WU Xi E-mail:dreamathstat@163.com;451173096@qq.com;25804126@qq.com;weisiya_bess@163.com;961498170@qq.com;wuxi@seu.edu.cn
Supported by:
State Grid Corporation of China Research Program(5600-201955167A-0-0-00)

摘要/Abstract

摘要：

综合能源负荷场景生成是研究能源计量、规划运行等领域问题的基础,具有重要意义。但由于数据采集困难、综合能源负荷多能耦合等因素的限制,综合能源负荷场景的多样化生成仍是一大难题。提出了一种基于生成对抗网络(generative adversarial networks, GAN)的综合能源负荷场景生成方法。首先建立梯度惩罚优化的Wasserstein生成对抗网络模型,解决综合能源负荷的高随机性可能带来的不收敛或模式崩溃问题。其次,基于深度长短期记忆(long short-term memory, LSTM)的循环神经网络构建生成对抗网络的生成器和判别器,使模型更适用于复杂综合能源负荷数据生成。算例结果表明,所提模型的生成负荷场景在概率分布、曲线标志性特征和冷热电负荷之间相关性等方面相较于蒙特卡洛法和原始生成对抗网络均获得了较好结果,可以在不同模式下生成具有多样性且逼真的负荷场景。

关键词: 综合能源系统, 场景生成, 深度学习, 生成对抗网络(GAN), 长短期记忆网络(LSTM)

Abstract:

Load scenario generation is the basis of studying energy measurement, operation scheduling and other fields, which is of great significance. Due to difficulty of data collection and multi-energy coupling of integrated energy system, it is still a big challenge to generate load data with diversity. A novel multi-load scenario generation method based on generative adversarial network (GAN) is proposed in this paper. Firstly, the Wasserstein generative adversarial network model with gradient penalty optimization is established to overcome the misconvergence and mode collapse caused by high randomness of load. Secondly, on the basis of the recurrent neural network with deep long-term and short-term memory, the generator and discriminator in the GAN are constructed to be more suitable for load data generation of complex integrated energy system. The result shows that the scenarios generated by proposed model achieves better results in probability distribution, curve signature features and correlation in cooling, heating and power load than original GAN and Monte Carlo method. The model can generate realistic load scenarios with diversity in different modes.

Key words: integrated energy system, scenario generation, deep learning, generative adversarial network (GAN), long short-term memory(LSTM)

中图分类号:

TM712

朱庆, 郑红娟, 唐子逸, 韦思雅, 邹子骁, 吴熙. 基于生成对抗网络的综合能源负荷场景生成方法[J]. 电力建设, 2021, 42(12): 1-8.

ZHU Qing, ZHENG Hongjuan, TANG Ziyi, WEI Siya, ZOU Zixiao, WU Xi. Load Scenario Generation of Integrated Energy System Using Generative Adversarial Networks[J]. ELECTRIC POWER CONSTRUCTION, 2021, 42(12): 1-8.

图/表 11

图1 生成器和判别器结构

Fig.1 Structure of generator and discriminator

表1 负荷生成模型的超参数阈值

Table 1 Hyper-parameter threshold of load generation model

图2 GAN训练流程

Fig.2 Training process of generative adversarial networks

图3 Wasserstein距离随训练变化情况

Fig.3 Wasserstein distance changes with training

图4 生成场景与真实场景样本数据比较

Fig.4 Comparison of sample data between generated and real scenarios

图5 本文方法生成场景与真实场景自相关系数比较

Fig.5 Comparison of autocorrelation between generated and real scenarios

图6 生成场景与真实场景数据概率分布特性比较

Fig.6 Comparison of probability distribution characteristics between generated and real scenarios

表2 真实场景下的负荷相关性

Table 2 Correlation of load in real scenarios

表3 本文方法生成场景下的负荷相关性

Table 3 Correlation of load scenarios generated by proposed method

表4 原始GAN生成场景下的负荷相关性

Table 4 Correlation of load scenarios generated by original GAN

表5 蒙特卡洛生成场景下的负荷相关性

Table 5 Correlation of load scenarios generated by Monte Carlo

参考文献 27

[1]	杨海柱, 李梦龙, 江昭阳, 等. 考虑需求侧电热气负荷响应的区域综合能源系统优化运行[J]. 电力系统保护与控制, 2020, 48(10):30-37.
	YANG Haizhu, LI Menglong, JIANG Zhaoyang, et al. Optimal operation of regional integrated energy system considering demand side electricity heat and natural-gas loads response[J]. Power System Protection and Control, 2020, 48(10):30-37.
[2]	FEKRI M N, GHOSH A M, GROLINGER K. Generating energy data for machine learning with recurrent generative adversarial networks[J]. Energies, 2019, 13(1): 130. doi: 10.3390/en13010130 URL
[3]	GUO Z J, WAN Y M, YE H. A data imputation method for multivariate time series based on generative adversarial network[J]. Neurocomputing, 2019, 360: 185-197. doi: 10.1016/j.neucom.2019.06.007 URL
[4]	熊文, 刘育权, 苏万煌, 等. 考虑多能互补的区域综合能源系统多种储能优化配置[J]. 电力自动化设备, 2019, 39(1): 118-126.
	XIONG Wen, LIU Yuquan, SU Wanhuang, et al. Optimal configuration of multi-energy storage in regional integrated energy system considering multi-energy complementation[J]. Electric Power Automation Equipment, 2019, 39(1): 118-126.
[5]	PAPAEFTHYMIOU G, KLOCKL B. MCMC for wind power simulation[J]. IEEE Transactions on Energy Conversion, 2008, 23(1): 234-240. doi: 10.1109/TEC.2007.914174 URL
[6]	刘纯, 吕振华, 黄越辉, 等. 长时间尺度风电出力时间序列建模新方法研究[J]. 电力系统保护与控制, 2013, 41(1): 7-13.
	LIU Chun, LÜ Zhenhua, HUANG Yuehui, et al. A new method to simulate wind power time series of large time scale[J]. Power System Protection and Control, 2013, 41(1): 7-13.
[7]	方华亮, 李大虎, 彭辉, 等. 基于“互联网+”的分散式太阳能规划方法[J]. 中国电机工程学报, 2017, 37(5): 1316-1324.
	FANG Hualiang, LI Dahu, PENG Hui, et al. Distributed solar energy planning method based on Internet plus[J]. Proceedings of the CSEE, 2017, 37(5): 1316-1324.
[8]	董雷, 孟天骄, 陈乃仕, 等. 采用马尔可夫链—多场景技术的交直流主动配电网优化调度[J]. 电力系统自动化, 2018, 42(5): 147-153.
	DONG Lei, MENG Tianjiao, CHEN Naishi, et al. Optimized scheduling of AC/DC hybrid active distribution network using Markov chains and multiple scenarios technique[J]. Automation of Electric Power Systems, 2018, 42(5): 147-153.
[9]	常小强, 宋政湘, 王建华. 基于蒙特卡罗算法的电动汽车充电负荷预测及系统开发[J]. 高压电器, 2020, 56(8): 1-5.
	CHANG Xiaoqiang, SONG Zhengxiang, WANG Jianhua. Electric vehicle charging load prediction and system development based on Monte Carlo algorithm[J]. High Voltage Apparatus, 2020, 56(8): 1-5.
[10]	张辰毓, 许刚. 分布式多元随机动态场景生成及快速精准场景降维算法[J/OL]. 电网技术, 2021 (2021-05-27)[2021-06-01]. https://doi.org/10.13335/j.1000-3673.pst.2021.0373.
	ZHANG Chenyu, XU Gang. Distributed multivariate random dynamic scenario generation and fast & accurate scenario simplified algorithm[J/OL]. Power System Technology, 2021 (2021-05-27)[2021-06-01]. https://doi.org/10.13335/j.1000-3673.pst.2021.0373.
[11]	吴钉捷, 李晓露, 陆一鸣. 基于生成对抗网络和联邦学习的非侵入式负荷分解方法[J/OL]. 电测与仪表, 2021 (2020-04-21) [2021-06-01]. http://kns.cnki.net/kcms/detail/23.1202.TH.20200421.1515.032.html.
	WU Dingjie, LI Xiaolu, LU Yiming. Non-intrusive load disaggregation method based on GAN and federated learning[J/OL]. Electrical Measurement & Instrumentation, 2021 (2020-04-21) [2021-06-01]. http://kns.cnki.net/kcms/detail/23.1202.TH.20200421.1515.032.html.
[12]	林珊, 王红, 齐林海, 等. 基于条件生成对抗网络的短期负荷预测[J]. 电力系统自动化, 2021, 45(11): 52-60.
	LIN Shan, WANG Hong, QI Linhai, et al. Short-term load forecasting based on conditional generative adversarial network[J]. Automation of Electric Power Systems, 2021, 45(11): 52-60.
[13]	张承圣, 邵振国, 陈飞雄, 等. 基于条件深度卷积生成对抗网络的新能源发电场景数据迁移方法[J/OL]. 电网技术, 2021 (2021-07-23) [2021-08-28]. https://doi.org/10.13335/j.1000-3673.pst.2021.1008.
	ZHANG Chengsheng, SHAO Zhenguo, CHEN Feixiong, et al. A renewable power generation data transferring method based on conditional deep convolutions generative adversarial network[J/OL]. Power System Technology, 2021 (2021-07-23) [2021-08-28]. https://doi.org/10.13335/j.1000-3673.pst.2021.1008.
[14]	董骁翀, 孙英云, 蒲天骄. 基于条件生成对抗网络的可再生能源日前场景生成方法[J]. 中国电机工程学报, 2020, 40(17): 5527-5536.
	DONG Xiaochong, SUN Yingyun, PU Tianjiao. Day-ahead scenario generation of renewable energy based on conditional GAN[J]. Proceedings of the CSEE, 2020, 40(17): 5527-5536.
[15]	YANG F, YIN S P, ZHOU S Y, et al. Electric vehicle charging current scenario generation based on generative adversarial network combined with clustering algorithm[J]. International Transactions on Electrical Energy Systems, 2021, 31(8): e12971.
[16]	JIANG C M, MAO Y F, CHAI Y, et al. Scenario generation for wind power using improved generative adversarial networks[J]. IEEE Access, 2018, 6: 62193-62203. doi: 10.1109/ACCESS.2018.2875936 URL
[17]	WANG Y, HUG G, LIU Z J, et al. Modeling load forecast uncertainty using generative adversarial networks[J]. Electric Power Systems Research, 2020, 189: 106732. doi: 10.1016/j.epsr.2020.106732 URL
[18]	GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]// Proceedings of the 27th International Conference on Neural Information Processing Systems. Montreal, Canada: MIT Press, 2014: 2672-2680.
[19]	ZHANG C, KUPPANNAGARI S R, KANNAN R, et al. Generative adversarial network for synthetic time series data generation in smart grids[C]// 2018 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm). Aalborg, Denmark: IEEE, 2018: 1-6.
[20]	张宇帆, 艾芊, 林琳, 等. 基于深度长短时记忆网络的区域级超短期负荷预测方法[J]. 电网技术, 2019, 43(6): 1884-1892.
	ZHANG Yufan, AI Qian, LIN Lin, et al. A very short-term load forecasting method based on deep LSTM RNN at zone level[J]. Power System Technology, 2019, 43(6): 1884-1892.
[21]	ZANG H X, XU R Q, CHENG L L, et al. Residential load forecasting based on LSTM fusing self-attention mechanism with pooling[J]. Energy, 2021, 229: 120682. doi: 10.1016/j.energy.2021.120682 URL
[22]	MUZAFFAR S, AFSHARI A. Residential load forecasting based on LSTM fusing self-attention mechanism with pooling[J]. Energy Procedia, 2019, 158: 2922-2927. doi: 10.1016/j.egypro.2019.01.952 URL
[23]	BENDAOUD N M M, FARAH N, BEN AHMED S. Comparing generative adversarial networks architectures for electricity demand forecasting[J]. Energy and Buildings, 2021, 247: 111152. doi: 10.1016/j.enbuild.2021.111152 URL
[24]	MASCIALINO B, PFEIFFER A, PIA M G, et al. Evaluation of the power of Goodness-of-Fit tests for the comparison of data distributions[C]// 2006 IEEE Nuclear Science Symposium Conference Record. San Diego, CA, USA: IEEE, 2006: 101-103.
[25]	李洪均, 李超波, 张士兵. 噪声稳健性的卡方生成对抗网络[J]. 通信学报, 2020, 41(3): 33-44.
	LI Hongjun, LI Chaobo, ZHANG Shibing. Noise robust Chi-square generative adversarial network[J]. Journal on Communications, 2020, 41(3): 33-44.
[26]	王德文, 李业东. 基于WGAN图片去模糊的绝缘子目标检测[J]. 电力自动化设备, 2020, 40(5): 188-198.
	WANG Dewen, LI Yedong. Insulator object detection based on image deblurring by WGAN[J]. Electric Power Automation Equipment, 2020, 40(5): 188-198.
[27]	CHEN Y Z, WANG Y S, KIRSCHEN D, et al. Model-free renewable scenario generation using generative adversarial networks[J]. IEEE Transactions on Power Systems, 2018, 33(3): 3265-3275. doi: 10.1109/TPWRS.2018.2794541 URL

基于生成对抗网络的综合能源负荷场景生成方法

Load Scenario Generation of Integrated Energy System Using Generative Adversarial Networks

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 27

相关文章 15

编辑推荐

Metrics

本文评价

[1]	刘至纯, 李华强, 王俊翔, 陆杨, 游祥, 何永祥. 基于多时间尺度综合需求响应策略的综合能源系统优化运行[J]. 电力建设, 2022, 43(9): 54-65.
[2]	杨志豪, 刘沆, 文明, 赵海彭, 廖菁, 苗世洪. 考虑综合需求响应混合不确定性的综合能源系统优化调度策略[J]. 电力建设, 2022, 43(9): 66-76.
[3]	李国堂, 宋福浩, 陈杰, 刁利, 王永利, 郭远征, 刘桢梁. 考虑碳排放的综合能源系统“结构-型号-容量”优化模型[J]. 电力建设, 2022, 43(7): 37-47.
[4]	钟劲松, 王少林, 冉懿, 冉新涛, 于金平, 俞海猛. 基于互信息和LSTM的用户负荷短期预测[J]. 电力建设, 2022, 43(7): 96-102.
[5]	刘雪飞, 庞凝, 王云佳, 姚伯岩, 魏子强, 温鹏, 高立艾. 农村综合能源系统多层协同优化运行方法[J]. 电力建设, 2022, 43(5): 63-71.
[6]	吴涵, 刘洋, 杨祺铭, 许立雄, 钟磊. 计及需求响应的分布鲁棒博弈区域综合能源系统运行优化策略[J]. 电力建设, 2022, 43(4): 108-118.
[7]	肖发龙, 吴岳忠, 沈雪豪, 何震凯, 秦烨. 基于深度学习和知识图谱的变电站设备故障智能诊断[J]. 电力建设, 2022, 43(3): 66-74.
[8]	丁曦, 姜威, 郭创新, 奚增辉, 高洁. 考虑需求响应的电/热/气云储能优化配置策略[J]. 电力建设, 2022, 43(3): 83-99.
[9]	边辉, 陈丽娜, 马凡琳, 张新, 姜金朋. 基于中枢解耦与演化博弈的多农业园区综合能源系统优化运行[J]. 电力建设, 2022, 43(2): 26-36.
[10]	孙黎霞, 彭嘉杰, 田屹昀, 陈欣凌, 袁欢. 基于混合门控循环单元子层的多任务暂态稳定评估[J]. 电力建设, 2022, 43(2): 63-69.
[11]	魏震波, 马新如, 郭毅, 魏平桉, 卢炳文, 张海涛. 碳交易机制下考虑需求响应的综合能源系统优化运行[J]. 电力建设, 2022, 43(1): 1-9.
[12]	蔡杰, 赵海彭, 关钦月, 郭婷, 张迪, 苗世洪. 基于参数线性规划的电-气综合能源系统最优能量流算法[J]. 电力建设, 2022, 43(1): 29-37.
[13]	高建伟, 梁鹏程, 高芳杰, 门慧娟, 郎宇彤. 基于多方势博弈的综合能源系统运行优化[J]. 电力建设, 2021, 42(9): 32-40.
[14]	彭雨筝, 李晓露, 李聪利, 丁一. 基于残差卷积自编码的风光荷场景生成方法[J]. 电力建设, 2021, 42(8): 10-17.
[15]	王若谷, 陈果, 王秀丽, 钱涛, 高欣. 计及风电与电动汽车随机性的两阶段机组组合研究[J]. 电力建设, 2021, 42(8): 63-70.