Prediction Of MH-Ni Battery Capacity And Intilligent Charge Based On Neural Networks

At present, there were increasing concerns about environmental and energyproblems, electric vehicles have been developed for its benefit of zero emissionand low noise. Many countries make electric vehicles as the developmentdirection of motor industry, and nickel metal hydride (MH-Ni battery) becomesthe first elect of the storage battery in motor industry, because of the excellentsynthetic performance of MH-Ni battery, such as high specific energy, longcircle life, adapting to big currents discharge and no pollution. Recently, thebattery management system (BMS) of MH-Ni battery, in which the prediction ofMH-Ni battery residual capacity and the realization of intelligent quick andharmless charge are the two important contents, is the hotspot question. Basedon literature in existence, this paper induces fuzzy technology and neuralnetwork to study these two questions.Firstly, the question of prediction of MH-Ni battery residual capacity isstudied. This paper adopts RBF neural network to predict the state of charge(SOC) of MH-Ni battery, considering the slow rate of convergence and localminimum of BP neural network, the dynamic nearest neighbor-clusteringalgorithm is proposed based on The RBF design difficulty of appointing thenumber of the hidden layer nodes and center dynamically, the stimulation resultsshow that this method is effective and the prediction results can meet the requireof industrial precision.Secondly, the charge character of Ni-MH battery is studied. This paper citesan approximate charge model of MH-Ni battery, and an effective neural networkoptimization algorithm is proposed based on The RBF design difficulty ofappointing the number of the hidden layer nodes and center dynamically: Firstadjusted the number of the hidden layer nodes and initialized the center usingnearest neighborhood clustering learning algorithm, then optimized the centernumbers of the hidden layer and resolve the output weights and threshold basedon the generalization inverting matrix. The stimulation results indicate that thismethod can realize precise prediction result.Finally, the neural network and fuzzy control are combined to designed fuzzyneural network controller which can be used for charging the MH-Ni batteryintelligently, the stimulation results show that this method can short the chargetime of MH-Ni battery and realize intelligent charge of MH-Ni battery.