青藏铁路接地网设计中的遗传优化神经网络

Design for Substation Grounding Grids in Qinghai-Tibet Railway Based on BP Artificial Neural Network Optimized by Genetic Algorithm

为了提高青藏铁路变电站接地网设计准确度,同时解决CDEGS软件价格高,运算、操作复杂的问题,建立了用于青藏铁路接地网设计的遗传优化神经网络。首先分析了影响接地网接地电阻的主要因素,提出了接地网电阻系数的概念,根据CDEGS软件数据建立了求解接地网接地电阻的BP神经网络,为了提高网络的计算精度,根据青藏铁路接地网的实际情况确定了接地网网格数、长宽比和面积的选择范围,采用遗传算法对BP神经网络的训练过程进行优化。通过与实际计算对比表明,该遗传优化神经网络具有较高的准确性和可信度,且简单易行,可代替CDEGS软件,为青藏铁路沿线变电站接地网的设计提供帮助。

To enhance the design precision of substation＇ grounding resistance and recuce the design cost within Qinghai-Tibet Railway grounging grid, the BP artificial neural network （ANN） optimized by genetic algorithm （GA） is built. First, the main influence factors to grounding resistance of grounding grid are analyzed, and the concept of grid resistance coefficient is established. According to input-output data of CDEGS software, the system model of BP ANN to calculate the grounding resistance is built. According to the practical design of Qinghai-Tibet Railway grounding grid, the ranges of gridding, area and shape of grounding grid are analyzed. The genetic algorithm （GA） is used to improve the computing precision of BP ANN, and the exact system model to calculate grounding resistance is obtained. Compared with the actual calculation results, this system model is reliable. This method is helpful reference to electrical design and engineering of Qinghai-Tibet Railway grounding system.