AB5型储氢合金初始放电容量的预测

Predication on Initial Discharge Capacity of AB5-Based Hydrogen Storage Alloy

摘要:

将模拟退火算法用于神经网络，对AB5型储氢合金的初始放电容量性能进行预测。通过实验确定了冷却进度表中各项参数并讨论了其对网络预测性能的影响，提出了实用的冷却制度；将模拟退火算法与传统的梯度下降法结合，用于人工神经网络的优化与预测，得到了预测效果等与迭代时间性能更好的神经网络。

Abstract:

Predication on initial discharge capacity of AB5-based hydrogen storage was done by adapting the simulated annealing method to artificial neural network (ANN). The parameters in cooling schedule were determined and their influences on network predication performance were discussed in experiments. A practical cooling regime was established then, and a network with better performances was obtained by combining simulated annealing algorithm with gradient descent method.