A Sensor-less Estimation Strategy Of Fuel Concentration Based On Improved BP Neural Networks

With more and more environment problems exposed, such as hazy weather, new kinds of power are taken into consideration by our society. Direct Methanol Fuel Cells(DMFC) takes methanol solutions as anode fuel, have environment-friendly products, and methanol as a common chemistry product is rather economic. DMFC systems have small scales; it can be applied in distributed energy stations, portable power suppliers and automobile engines, etc. Methanol concentration has an important effect on the performance of DMFC, so how to control methanol concentration within the right scope and increase no complexity of systems is very important to improve the economic benefit and marketization of DMFC.This article mainly focuses on portable DMFCs, have designed a sensor-less control strategy of methanol concentration, my study work are arranged as follows:(1) After looking into many literatures on methanol concentration measurements of DMFC, voltage mode is bulit on the basis of MEA decay and influence by methanol concentration.(2) According to existing methods about improved BP neural networks used in other fields, a new improved BP neural network was designed to matching the voltage model with a large number of repeated experiments and empirical formulas.(3) Repeated tests have been done on a fuel cell and a stack to optimize parameters of improves BP neural network and to compare the calculations and measured values of methanol concentration. The results of test proved itself useful to DMFCs.(4) A methanol concentration control system was construct based on previous method and it has been proved by experiments with good results.