基于BP神经网络的低压配电台区电压估算

Voltage Estimation of Low-voltage Distribution Network Based on BP Neural Network

针对低压配电台区运行参数采集受限,无法实现电压全覆盖监测或电压精确计算的问题,在能够召测少数节点电压的情况下,提出基于少数据的BP神经网络电压估算算法。在总结低压配网可用参数的基础上,通过电压降落的近似计算,分析了低压配电网节点电压的影响因素,提出节点负荷矩的新概念,建立低压配网节点负荷矩——电压估算模型,并结合神经网络基于数据建模对输入输出特性的自学习能力,以少数节点数据为样本对配网全部节点电压进行估算。为防止神经网络在训练过程中陷入平坦区,在算法中添加加权因子,动态调整神经网络的学习率和动量因子,提高学习效率。采集实际配电网算例数据对算法进行验证,其误差满足估算精度要求,证明本文提出的方法具有可行性及适用性。

The Limitation of monitoring system to collect operating parameters in low-voltage distribution grid makes it difficult to monitor all users＇ voltage or calculate voltage accurately. Considering that a few users＇ voltages are available,a new method to estimate node voltage of low voltage distribution grid is proposed on the basis of the BP neural network. This paper analyzes the factors that are relevant to the voltage level of rural low-voltage distribution network in accordance with the approximate calculation of the voltage drop and the available parameters in the low voltage distribution. Then this paper defines a new variable called node-load-moment and establishes a model of voltage estimation to low-voltage distribution grid. Given the self-learning ability of BP neural network to the in-out characteristics based on the data modeling,a new method with a few available parameters is proposed to estimate all users＇ voltage of low-voltage distribution grid according to BP neural network. In addition,to avoid the error function falling into flat region and improve the learning efficiency of BP algorithm,this paper adds a weighting factor to the neural network. The weighting factor can adjust the learning rate and momentum factor in a dynamic manner. In the end,the data of pragmatic distribution grid examples are adopted to check the method and the simulation results show that the proposed method can calculate accurately which is feasible to the engineering application.