并网光伏电站运行特性分析及发电预测

发布时间：2018-11-16 19:28

【摘要】：光伏并网发电作为太阳能的主要使用方式,得到世界各国的大力扶持,发展迅猛,虽然国内外在光伏并网发电方面的研究已经取得很多成果,但是在一些关键性的技术方面,依然有很多问题有待解决。电网里如果含有并网光伏发电单元,必须要充分考虑其发电不确定性、间断性对电网的影响。本文调研了江苏地区并网光伏电站,分析了典型光伏电站的运行特性,深入研究江苏地区光伏电站出力规律。首先研究单一光伏电站出力特性,包括不同天气下日有功出力特性,出力波动特性,出力概率分布特性,最大出力时间分布特性,出力波动的随机分布特性,同一光伏电站在不同时间尺度下的出力相关性分析和单一光伏电站接入电网对配电网电压的影响。接着研究多光伏电站的出力特性,出力相关性及对电网电压的影响,以及光伏单元并网后电网电压越限的问题。接着提出准确的光伏预测可以解决光伏并网给电网安全稳定运行带来的不利影响。本文在发电预测部分,首先介绍了几种常见的预测方法,然后提出组合预测方法。本文的组合预测方法是基于方差-协方差的组合预测方法,先分别介绍本文组合预测方法中需要用到的三个单一预测法：人工神经网络预测方法、基于马尔可夫链的预测方法、基于相似日和最小二乘支持向量机的光伏发电预测。最后介绍基于方差一协方差的组合预测方法。本文的组合预测方法需要建立在三种单一预测方法的结果上来确定权重系数。因此在针对徐州协鑫光伏电站的工程预测实例部分,本文先采用三种单一预测法进行预测,最后结合单一预测的结果确立组合预测权重系数,采用基于方差一协方差的组合预测法得到组合预测值。通过对比三种单一算法和组合预测算法的平均误差,可以看出组合预测法更加准确。
[Abstract]:Photovoltaic grid-connected power generation as the main use of solar energy, has been vigorously supported by countries around the world, rapid development, although domestic and foreign research in photovoltaic grid-connected power generation has made a lot of achievements, but in some key technology, There are still many problems to be solved. If there are grid-connected photovoltaic units in the power grid, it is necessary to take full account of the uncertainty of its generation and the intermittent impact on the grid. This paper investigates the grid-connected photovoltaic power stations in Jiangsu Province, analyzes the operation characteristics of typical photovoltaic power stations, and deeply studies the law of photovoltaic power generation in Jiangsu area. Firstly, the output characteristics of a single photovoltaic power plant are studied, including the characteristics of daily active power output, force fluctuation, force probability distribution, maximum force time distribution, and random distribution characteristics of output force fluctuation under different weather conditions. The correlation analysis of the output force of the same photovoltaic power plant at different time scales and the influence of a single photovoltaic power station on the distribution network voltage. Then, the characteristics of power generation of multi-photovoltaic power stations, the dependence of output force and its influence on grid voltage, as well as the problem of grid voltage exceeding limit after grid connection of photovoltaic units are studied. Then it is proposed that accurate photovoltaic prediction can solve the adverse effects of grid-connected photovoltaic network on the safe and stable operation of the grid. In the part of generation prediction, several common forecasting methods are introduced, and then combined forecasting methods are proposed. The combined forecasting method is based on variance-covariance. First, three single forecasting methods are introduced: artificial neural network forecasting method, Markov chain based forecasting method. Photovoltaic generation prediction based on similarity day and least squares support vector machine. Finally, the combined prediction method based on variance-covariance is introduced. The combined forecasting method in this paper needs to be based on the results of three single forecasting methods to determine the weight coefficient. Therefore, in view of Xuzhou Xiexin photovoltaic power station engineering forecast example part, this article first uses three kinds of single forecast method to carry on the forecast, finally unifies the single forecast result to establish the combination forecast weight coefficient, The combination prediction method based on variance-covariance is used to obtain the combined prediction value. By comparing the average error between the three single algorithms and the combined prediction algorithm, we can see that the combined prediction method is more accurate.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM615

【相似文献】