遗传神经网络在载人飞船环控决策系统中的应用研究

Application of Genetic Neural Network to Decision Support System for Environmental Control and Life Support System

针对载人飞船环控生保系统的状态监控由于参数数量及不确定性因素多,导致学习模型训练周期长,不能满足快速、实时、准确参数预测的现实,运用遗传算法对神经网络进行优化,提出了基于遗传神经网络的环控生保参数预测模型,设计并实现了相应的仿真软件.以轨道舱总压预测为例,通过飞船的真实飞行数据测试,证实在达到同样误差的情况下,遗传神经网络的训练周期数比BP神经网络的训练周期数减少了30%,而且遗传神经网络的平均误差小于BP神经网络的平均误差,说明基于遗传神经网络的参数预测算法和模型能为载人飞船环控决策支持系统提供更准确和实时的关键参数预测.

The environmental control and life support system （ECLSS） of manned spacecraft consumes long training period for learning model due to huge number of ECLSS parameters and uncertain factors,so that it is hard to predict those parameters quickly and accurately in real time. This paper presents a new ECLSS parameters prediction model based on a genetic neural network by optimizing BP neural network with a genetic algorithm,and develops a simulation software. The model is verified by real spacecraft flying data in the case of predicting obit cabin pressure. It is proved that the training epochs of the genetic neural network are 30% less than that of BP network within the same error limitation,and the former has smaller mean error. So it is believed that the genetic neural network based model can predict the key parameters more accurately and fast for ECLSS decision support system.