基于神经网络的钒基储氢合金充放电循环性能预测分析

Prediction and Analysis on Charge-discharge Circulation Property of Vanadium-based Hydrogen Storage Alloy Based on Neural Network

以16种合金元素含量为输入层节点参数,以充放电循环性能为输出层节点参数,构建了16×48×1三层神经网络预测模型,并对预测能力进行了试验验证,同时对模型选出的合金进行了化学成分、显微组织、物相组成和充放电循环性能的测试与分析。结果表明,该神经网络模型的预测精度较高,V3TiNi0.56-0.1Sc合金具有最佳的充放电循环性能;该合金由V基固溶体相、TiNi相和Ti2Ni相组成,经过15次充放电循环后放电容量保持率高达82%,较V3TiNi0.56合金提高了80%。

The neural network prediction model with three layers of 16×48×1 was built with the contents of16 kinds of alloying elements as input parameters,and with the charge-discharge circulation property as output parameter. The verification test for the prediction model was carried out. Furthermore,the chemical composition,microstructure,phase composition and charge-discharge circulation property of the optimized alloy picked out by the model were tested and analyzed. The results show that the neural network prediction model has high precision,and the V3TiNi0.56-0.1Sc alloy,which is composed of V-based solid solution phase,TiNi and Ti2 Ni phases,has optimum charge-discharge circulation property. Moreover,the maintenance rate of discharge capacity for the V3TiNi0.56-0.1Sc alloy keeps at 82% after 15 times of charge-discharge circulation,which increases by 80% in comparison to that of V3TiNi0.56 alloy.