基于BP神经网络的电网电压暂降源定位方法

A novel location method of power grid voltage sag source with BP neural network

针对目前主配网电压暂降源定位不准确、适用范围有限、定位特征参数利用不足等问题,设计一种基于BP神经网络对多种电气特征量进行整合利用的电压暂降源综合判据定位方法。首先,对主流定位算法所利用的电气特征量进行分析,选择利用电压、电流、相角和等效阻抗这4种电气特征量作为综合指标体系;其次,利用BP神经网络定位算法的泛化特性和非线性拟合功能,拟合上述指标体系和电压暂降源位置的内在非线性关系,设计出综合定位判据,利用该综合判据能较全面地描述故障发生过程电力网络的状态,进而可以通过该综合判据判断电压暂降源的位置。综合判据定位法通过网络训练可以将网络架构信息和不同故障类型信息存储在神经网络的权值和阈值中,可应用于不同网络架构和不同短路故障类型。最后,以广东电网某大型城市220 k V及以上电网系统作为仿真算例,验证了所设计方法的有效性和可行性。

Aiming at the problems such as low accuracy, limited application scope, and inadequate use of location feature parameters in the voltage sag source location method for active distribution networks, a new location method with comprehensive criterion was proposed in this paper, which used the BP neural network to integrate various location parameters. Firstly, the electrical characteristics of the mainstream sag source location algorithms were analyzed, and then voltage, current, phase angle and equivalent impedance were chosen to the index system. Secondly, the generalization feature and nonlinear fitting of BP neural network were utilized to fit the intrinsic nonlinear relationships between the index system and voltage sag source location, and the location comprehensive criterion was designed. The criterion described the faulted situation of power grid and determined the voltage sag source location. The location method can store the network architecture information and fault type information in the weights and threshold values of neural network through network training, which enables the method application in different network architectures and different fault types. Finally, a 220kV and above power grid of a large city in Guangdong was simulated to verify the effectiveness and feasibility of the proposed method.