| Summary: | 碩士 === 輔仁大學 === 電子工程學系 === 92 === The battery management system (BMS) needs accurate estimate SOC in the intelligent electronic vehicle, various kinds of battery electric consumption had began propose , but fail accurate prediction get battery state of capacity of battery string among method these already to estimate. This thesis presents the different kinds of fuzzy neural network to estimate the state of capacity of the battery, and train the network on the principle that the single battery has four input parameter (the battery voltage, current, temperature and internal resistance value), a kind of tradition back-propagation network, adaptive-network-based fuzzy inference system, fuzzy neural network using the gradient descent and fuzzy neural network using the genetic algorithm. Through the lithium ion battery experiment and simulation network result that the fuzzy neural network using the genetic algorithm out is best, adaptive-network-based fuzzy inference system and fuzzy neural network using the gradient descent take second place, and it is worst to use tradition back-propagation network out. As the foundation with this result, will expect to be able to combined with battery equalizer technique to improve the drawbacks on the present battery management system .
|
|---|
