基于像素卷积生成网络的可再生能源场景生成  

Renewable Energy Scenario Generation Based on Pixel Convolutional Generative Networks

作　　者：廖文龙 杨德昌[2] 葛磊蛟[3] 杨哲 刘匡普 宋如楠 LIAO Wenlong;YANG Dechang;GE Leijiao;YANG Zhe;LIU Kuangpu;SONG Runan(AAU Energy,Aalborg University,Aalborg 9220,Denmark;College of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China;Key Laboratory of Smart Grid of Ministry of Education,Tianjin University,Tianjin 300072,China;China Electric Power Research Institute,Beijing 100192,China)

机构地区：[1]奥尔堡大学能源系,奥尔堡9220 [2]中国农业大学信息与电气工程学院,北京100083 [3]天津大学智能电网教育部重点实验室,天津300072 [4]中国电力科学研究院有限公司,北京100192

出　　处：《高电压技术》2022年第4期1320-1331,共12页High Voltage Engineering

基　　金：国家自然科学基金(51977212);国家自然科学基金智能电网联合基金(U2166208);中国农业大学中央高校基本科研业务费(2021TC069)。

摘　　要：以光伏和风电为代表的可再生能源具有很强的间歇性和波动性,会给配电网的运行和优化带来巨大的挑战。为表征光伏和风电输出功率的不确定性,提出了基于像素卷积生成网络(pixel convolutional generative network,PCGN)的可再生能源随机场景生成方法。该方法通过链式法则将可光伏和风电出力曲线的联合分布因子化成多个1维分布的乘积,利用先前生成的输出功率来预测下一个功率值,从而实现可再生能源出力曲线的逐点生成。根据光伏和风电出力曲线的高维数据特征,设计了适用于可再生能源随机场景生成的网络结构,并通过实际运行数据验证了所提方法的有效性和适应性。仿真结果表明,所提的PCGN不仅可以准确地捕获光伏和风电出力曲线的形状特征、波动性、概率分布以及时空相关性,还具有较强的适应性,通过微调生成网络的结构和参数,就能用于不同可再生能源的随机场景生成任务。Renewable energy represented by photovoltaic and wind power has characteristics of strong intermittence and fluctuation,which brings great challenges to the operation and optimization of distribution networks.To investigate the uncertainties of photovoltaic and wind power output,a stochastic scenario generation method for renewable energy based on the pixel convolutional generative network(PCGN)is proposed.The joint distribution of photovoltaic and wind power generation curves is transformed into the product of multiple one-dimensional distributions by the chain rule,and the previously generated output power is adopted to predict the next power value,thus achieving point-by-point generation of each power point in power generation curves of renewable energies.According to the high-dimensional data characteristics of photovoltaic and wind power generation curves,network structures applicable to the stochastic scenario generation of the renewable energy is designed,and the effectiveness and adaptability of the proposed method are verified by the real operation data.Simulation results show that the proposed PCGN can not only accurately capture the shape characteristics,volatility,probability distribution,and spatio-temporal correlation of photovoltaic and wind power curves,but also has strong adaptability,which can be used for stochastic scenario generation tasks for different renewable energies by fine-tuning the structure and parameters of generative networks.

关 键 词：生成网络 可再生能源 场景生成 深度学习 数据驱动 

分 类 号：TM73[电气工程—电力系统及自动化] TP183[自动化与计算机技术—控制理论与控制工程]