基于多链MCMC方法的光伏出力序列预测研究

现有电力系统规划和运行中主要考虑单一可再生能源出力的不确定性,考虑可再生能源出力间相关性的研究较少。文中提出一种多链马尔科夫-蒙特卡洛(MCMC)方法对多个光伏电站出力序列进行组合预测,该方法建立了多条相互服从完全条件分布的马尔科夫链,以模拟光伏电站上空的随机变化的大气状态,充分保留了光伏电站之间的相关特性。对三组具有不同相关水平的光伏电站的出力序列进行了预测,证明了相较于传统MCMC方法,该方法能够更精确地继承历史序列的一般统计特性,能够更有效地体现多个光伏电站出力之间相互影响的特点,更加适用于未来电力系统规划与运行设计的要求。

A PV Power Time Series Generating Method Considering Correlation Characteristics Based on Multi Markov Chain Monte Carlo Method

The existing power system planning and operation mainly considers the uncertainty of the single renewable energy output,and there are few studies considering the correlation between renewable energy outputs.In this paper,a multi Markov Chain Monte Carlo method is proposed to predict the PV power time series.This method establishes several Markov chains that obey the complete conditional distribution to each other to simulate the random variation of the atmosphere over the PV plants.These Markov chains can fully retain the correlation between the PV plants.This paper simulates the time series of three sets of PV plants with different correlation levels using multi MCMC method.It is proved that this method can inherit the general statistical characteristics of historical sequences more accurately than the original MCMC method and can more effectively reflect the characteristics of the interaction between PV power plants.The multi MCMC method is more suitable for future power electrical system planning and operation design.