基于神经网络的车载电源开关电容变换器的智能潜电路分析  

Intelligent Sneak Circuit of Switched Capacitor Converters of Vehicle Power Based on Neural Network

作　　者：何惠英[1] 李良洪[1] 付兰芳[1] 赵玲[1] 

机构地区：[1]军事交通学院基础部,天津300161

出　　处：《军事交通学院学报》2017年第6期87-91,共5页Journal of Military Transportation University

摘　　要：在基于学习机制的智能潜电路分析过程中,如何有效地从大量电路信息中抽象出神经网络所需的样本数据,是保证系统预测结果可靠性的重要前提。为提高神经网络样本数据的有效性,基于图的理论和无效路径剔除方法,提出电路信息转换成神经网络样本数据的新方法,并将此方法应用于车载电源中的基本降压式谐振开关电容变换器潜电路分析过程中的样本生成环节,再借助Matlab工具箱,用遗传算法优化的BP神经网络,对经无效路径剔除方法处理后的样本数据和原始样本数据分别进行训练。训练结果验证了此方法的准确性和实用性。通过误差分析表明,网络训练前对样本数据进行有效处理,不仅可以避免传统潜电路分析中前期大量的数据输入工作及线索表难以获取等问题,还可提高系统预测的准确性。In the process of analyzing intelligent sneak circuit based on learning mechanism,abstracting sample data for neural network from a large number of circuit information is the important premise of the reliability of prediction. In order to improve the effectiveness of the sample data of neural network,the paper firstly proposes a new method of transforming circuit information into sample data of neural network based on graph theory and excluding method of invalid paths,and applies this method in the link of sample generation in sneak circuit analysis process of switched capacitor converters of vehicle power. Then,it trains the sample data dealt with excluding method of invalid paths and the original sample data respectively with BP neural network optimized by genetic algorithm in MATLAB toolbox. The training result verified the accuracy and practicability of this method,and the error analysis showed that the effective treatment of sample data before training can reduce a lot of data entry work and solve the problem of obtaining clue lists,and it can also improve the accuracy of system prediction.

关 键 词：车载电源 潜电路分析 神经网络 有效样本 开关电容变换器 

分 类 号：TM46[电气工程—电器]