基于场景分析的多风电场无功优化

发布时间：2018-08-05 11:35

【摘要】：相对于传统能源而言,风电具有随机性以及波动性。由于风电的不可控性,对于电力系统而言,如何在电力系统规划以及运行层面上考虑风电的随机性以及波动性成为当今电力系统的热点研究问题,尤其是给电力系统无功/电压控制方面带来一系列巨大的挑战。由于风电出力对电力系统运行的影响具有复杂的非线性特征,传统风电场景模型难以保证风电场景与电力系统优化运行保持一致。为此,不同于传统方法的先对风电场景聚类、再进行无功优化,而是建立系统运行特性聚类得到系统运行特性场景。由此出发,本文从如下几方面进行研究:首先,风电场景分成了静态场景模型和动态场景模型:利用非参数核密度估计的方法以及LHS的方法产生了风电静态模型;利用风电预测的预测误差分布以及风电的波动性分布,结合多元标准正态分布逆变换抽样产生了风电动态场景。其次,在风电静态场景当中考虑到K-means聚类方法难以确定聚类数的问题,通过聚类指标得到运行场景的最佳聚类数。将澳大利亚2个风电场实际数据产生的静态场景接入到IEEE30节点系统中,分别进行传统风电场景分析和所提出的运行场景分析,比较了系统网损和电压的概率特性,验证了所提出的运行场景分析方法的合理性和优越性。其次考虑到运行场景计算时间上的效率问题,并基于电力系统灵敏度的指标,提出了风电静态灵敏度场景,同样接入到系统当中验证了该模型的有效性。再者,在风电动态场景当中,不同于静态场景聚类,而是采用场景削减的方法,提出了一种动态运行场景模型。首先,风电接入系统当中进行动态无功优化,得到控制变量样本,其次对控制变量进行场景削减得到风电动态运行场景,最后,带入到系统当中进行动态无功优化,得到动态无功优化的结果。基于IEEE30节点和爱尔兰风电数据,比较了系统网损和电压的概率特性,验证了所提出的动态运行场景分析方法的合理性和优越性。
[Abstract]:Compared with traditional energy, wind power has randomness and volatility. Due to the uncontrollability of wind power, how to consider the randomness and fluctuation of wind power in power system planning and operation level has become a hot research issue in power system nowadays. In particular, it brings a series of great challenges to the reactive power / voltage control of power system. Because of the complex nonlinear characteristics of wind power generation, it is difficult for the traditional wind power scenario model to ensure that the wind power scenario is consistent with the optimal operation of the power system. Therefore, different from the traditional methods, the wind power scenarios are clustered first, then reactive power optimization is carried out. Instead, the system operation characteristic clustering is established to obtain the system operation characteristic scenarios. Firstly, the wind power scene is divided into static scene model and dynamic scenario model. The method of nonparametric kernel density estimation and LHS method are used to produce the static wind power model. Based on the prediction error distribution of wind power prediction and the fluctuation distribution of wind power, combined with the inverse sampling of multivariate standard normal distribution, the dynamic scene of wind power is generated. Secondly, considering the problem that the K-means clustering method is difficult to determine the clustering number in the static wind power scenario, the optimal cluster number of the running scenario can be obtained by clustering index. The static scene generated by the actual data of two wind farms in Australia is connected to the IEEE30 node system, and the traditional wind power scenario analysis and the proposed operation scenario analysis are carried out, respectively, and the probability characteristics of the system network loss and voltage are compared. The rationality and superiority of the proposed method are verified. Secondly, considering the efficiency of the calculation time of the running scenario, and based on the sensitivity index of the power system, the static sensitivity scenario of wind power is proposed, which is also connected to the system to verify the validity of the model. Furthermore, in the dynamic scenario of wind power, different from the static scenario clustering, a dynamic running scenario model is proposed by using the method of scene reduction. First, dynamic reactive power optimization is carried out in wind power access system, and control variable samples are obtained. Secondly, dynamic operation scenarios of wind power are obtained by scene reduction of control variables. Finally, dynamic reactive power optimization is carried out into the system. The results of dynamic reactive power optimization are obtained. Based on IEEE30 node and Irish wind power data, the probability characteristics of system loss and voltage are compared, and the rationality and superiority of the proposed dynamic operation scenario analysis method are verified.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM614

【参考文献】