基于CNN-MD的军用电站健康状态评估方法

A Health Status Evaluation Method for Military PowerStation Based on CNN-MD

基于健康状态评估的军用电站预先维修对提高其可靠性、安全性有着重要意义,采用基于CNN-MD的评估方法,对军用电站健康状态进行评估。建立卷积神经网络(convolutional netural network,CNN)军用电站健康状态识别模型,通过采用不同的状态样本训练模型,可使模型识别和输出健康状态类别;再引入马氏距离(Mahlanobis distance,MD)算法,计算不同状态下样本与健康样本的MD距离,并将其归一化为健康指数,即可进一步量化军用电站健康评价结果。以某型军用电站人工模拟不同健康状态的工况获取实验数据,通过模型能对正常、退化和注意三种状态进行有效区分,并得到量化评价健康指数0.6为正常状态的临界阈值,0.6以下为退化以下状态的具体量化评估值,验证了所方法的有效性。该方法通过CNN和MD结合,可实现军用电站健康状态定性与定量评估目的,为开展预先维修提供依据。

Pre maintenance of military power stations based on health status assessment is of great significance in improving their reliability and safety.Therefore,a CNN-MD based evaluation method is proposed to evaluate the health status of military power stations.It establishes a CNN military power station health status recognition model,which can recognize and output health status categories by training the model with different state samples.Introducing the MD algorithm again,calculating the MD distance between samples and healthy samples in different states,and normalizing it into a health index can further quantify the health evaluation results of power stations.Using a certain military power stations to artificially simulate different health conditions,experimental data was obtained.The model was able to effectively distinguish between normal,degraded,and attention states,and a quantitative evaluation health index of 0.6 was obtained as the critical threshold for normal states,and below 0.6 was the specific quantitative evaluation value for degraded states.The effectiveness of the proposed method was verified.This method combines CNN and MD to achieve the qualitative and quantitative evaluation of the health status of military power stations,providing a basis for pre maintenance.