基于随机变量状态时序模拟的光伏并网系统概率潮流研究  

Probabilistic Power Flow Research of Photovoltaic Grid-connected System Based on Stochastic Variable State Time Series Simulation

作　　者：苑婷 王俊[2] 朱浩祎 王刚 张林 孙佳琪 余维涛 王欣然 段佳鑫 孟妍 YUAN Ting;WANG Jun;ZHU Hao-yi;WANG Gang;ZHANG Lin;SUN Jia-qi;YU Wei-tao;WANG Xin-ran;DUAN Jia-xin;MENG Yan(State Grid Panjin Power Supply Company,Panjin Liaoning 124000,China;College of Information and Electrical Engineering,Shenyang Agricultural University,Shenyang 110161,China;State Grid Shenyang Power Supply Company,Shenyang 11000)

机构地区：[1]国网辽宁省电力有限公司盘锦供电公司,辽宁盘锦124000 [2]沈阳农业大学信息与电气工程学院,沈阳100161 [3]国网辽宁省电力有限公司沈阳供电公司,沈阳110003

出　　处：《沈阳农业大学学报》2019年第6期753-759,共7页Journal of Shenyang Agricultural University

基　　金：国家自然科学基金项目(61903264);辽宁省自然科学基金计划重点项目(20170540810);国网辽宁省电力有限公司科技项目(SGLNPJ00FZJS1900381)

摘　　要：近年来,太阳能光伏产业迅猛发展,光伏发电并网加剧了电力系统的不确定性,考虑光伏发电功率、电网元件状态等随机变量不确定性的概率潮流能够更加全面反映电力系统运行状况。在基于实际光伏电站数据分析的基础上,探究了影响光伏发电的几大主要气候条件,其中,太阳辐射强度、环境温度、湿度以及风速与光伏发电输出功率间的相关系数分别为0.9939,0.5032,-0.3861,0.5383,表明存在极显著的相关,具有较强的统计意义。以太阳辐射强度、环境温度、湿度以及风速4种环境因素作为预测模型的输入变量,忽略其他因素,根据它们之间的相关性,建立用于光伏发电功率短期预测的BP神经网络模型,并成功的预测了未来7d的光伏功率输出情况。利用序贯蒙特卡罗法时序模拟电网中线路的运行状态,综合考虑光伏发电功率和线路运行状态的不确定性对系统进行概率潮流研究。在已知线路故障以及其修复时间的前提下,模拟出72h内线路的潮流分布情况,能够有效反映电网运行风险。通过对潮流的分析,可及时发现电网中存在的薄弱环节,有针对性地制定事故预防和电网改进措施。基于IEEE30节点系统进行算例研究,算例分析结果表明:建立的四输入一输出的光伏发电功率预测方法,对预测值与实际值相比较,计算均方差达到0.000256,可用于实际预测。另外,基于对节点电压、线路功率等概率潮流的分析,验证了针对光伏并网系统进行概率潮流研究的必要性。In recent years, with the rapid development of solar photovoltaic industry, photovoltaic power generation is connected to the grid to aggravate the uncertainty of power system. Considering the uncertainty of photovoltaic power generation power,power grid components and other random variables, the probabilistic power flow can reflect the operation of power system more comprehensively. Based on the analysis of the actual photovoltaic power station data, several main climatic conditions affecting photovoltaic power generation are explored, in which the correlation coefficients between solar radiation intensity, environmental temperature, humidity and wind speed and the output power of photovoltaic power are 0.9999, 0.5032,-0.3861 and 0.5383,respectively, indicating that there is a very significant phase, with a strong statistical significance, therefore, four environmental factors, solar radiation intensity, environmental temperature, humidity and wind speed, are used as the input variables of the prediction model, and other factors are ignored. According to the correlation between them, a BP neural network model for shortterm prediction of photovoltaic power generation is established, and the photovoltaic power output in the next seven days is successfully predicted. The sequential Monte Carlo method is used to simulate the operation state of the line in the power grid,and the probabilistic power flow of the system is studied considering the uncertainty of photovoltaic power generation power and line operation state. On the premise of knowing the line fault and its repair time, 72 small pieces are simulated. The power flow distribution of the internal line can effectively reflect the power grid operation risk. Through the analysis of the power flow, the weak links in the power grid can be found in time, and the accident prevention and power grid improvement measures can be formulated. Based on the example of IEEE30 node system, the results of example analysis show that the proposed four-input-one-output photovoltai

关 键 词：光伏并网系统 BP神经网络 序贯蒙特卡罗仿真 概率潮流 

分 类 号：TM743[电气工程—电力系统及自动化]