基于特征影响因子和改进BP算法的直驱风机风电场建模方法

Modeling Method of Direct-driven Wind Generators Wind Farm Based on Feature Influence Factors and Improved BP Algorithm

风电场/场群规模化接入电网背景下,电网的故障暂态特性发生了根本性改变。然而,现有单机等值型无法精确表征风电场/场群的故障暂态特性。该文提出一种基于特征影响因子和改进人工神经网络反向传播(backpropagation neuronnetworks,BP)算法的直驱风机风电场建模方法。首先,建立直驱风机暂态模型,通过理论分析构建风机与公共连接点(point of common coupling,PCC)距离、直流侧限流措施投入情况、风速、出口处无功功率等故障特征影响因子。然后,对特征影响因子集计算欧式距离,基于改进最大最小距离法提取风机的分类初始中心。通过改进BP算法,以特征影响因子和分类初始中心为训练集,实现神经网络的快速收敛。最后,通过仿真算例,对所提方法进行验证。仿真结果表明,所述方法在收敛速度、建模精度方面,与传统BP算法和单机等值建模方法相比均有较大提升。

As the large-scale wind farms/fields accessed to the grid,the fault transient characteristics of the grid had undergone fundamental changes.However,the existing single-machine equivalent model cannot accurately characterize the fault transient characteristics of wind farms/field groups.This paper proposed a modeling method of direct-driven wind generators wind farm based on feature influence factors and improved back-propagation neural networks(BP)algorithm.Firstly,the transient model of the single direct-drive wind power generators was established.Through the theoretical analysis,the influence factors of the fault characteristics such as the distance between the generator and the point of common coupling(PCC),the situation of the DC-side current limiting measure,the wind speed and the reactive power of wind generator were constructed.Secondly,the Euclidean distance was calculated for the factor set,and the classification initial center of the wind generators were extracted based on the improved maximum and minimum distance method.By improving the BP algorithm,the feature influence factors and the initial center of the classification were used as training sets to achieve fast convergence of the neural network.Finally,the proposed method was verified by simulation examples.The simulation results show that the proposed method had a greatly improvement in convergence speed and modeling accuracy compared with traditional BP algorithm and single-machine equivalent modeling method.