| AC contactor is a control appliance for low voltage level.Its main function is to control the on-off of AC circuit and control circuit with large current level.AC contactors are widely used in power systems because of their advantages of frequent operation,strong ability to withstand large currents,and long operating distances.Its working state plays a key role in the safe operation of the entire power grid.Therefore,predicting the remaining life of the AC contactor can effectively prevent the paralysis of the power control system due to its failure.It can be predicted and replaced in advance before the end of its life,which can effectively maintain the healthy and safe operation of the power system.This thesis studies the characteristic parameters that can effectively characterize the operating state of the AC contactor,and uses the method based on deep learning to establish a DDAE-LSTM model to predict the remaining life of the AC contactor.Firstly,the structure,operation and failure characteristics of the AC contactor are analyzed,and the operating characteristic parameters that can effectively reflect the degradation trend of the AC contactor are explored.Test the CJX2-5011 contactor through the AC contactor operating state degradation test platform,the electrical signals and contact vibration signals of the three-phase main contacts and coils in operation are obtained,and the contactor operating characteristic parameters related to electrical parameters and mechanical parameters are extracted from them.Secondly,in order to eliminate redundant information in the extracted characteristic parameters,the influence degree of each characteristic parameter on the remaining life of the AC contactor and the correlation between each characteristic parameter are comprehensively considered.The grey relational analysis method and the Pearson correlation coefficient method were used to select the characteristic parameters,and the characteristics that had a great influence on the remaining life and had a low degree of correlation were retained.Considering that the remaining life research method of deep neural network,the larger the dimension of its input,the more difficult the training and the amount of computation increase exponentially,it is necessary to reflect the operating state of the AC contactor to the greatest extent with the least features,and comprehensively consider the characteristic parameters of electrical parameters and mechanical parameters.Therefore,the deep denoising autoencoder（DDAE）method is used to fuse and reconstruct the filtered features to reduce the dimension to form the optimal life prediction model dataset.Finally,the dataset is fed into a long short-term memory neural network（LSTM）for model training and AC contactor remaining life prediction.After comparing the results,in terms of prediction models,the long-short-term memory neural network model has higher prediction accuracy than the traditional recurrent neural network model and the support vector machine regression prediction model;in terms of model input,the prediction accuracy of using mechanical parameter features alone is low,and the stability of prediction results using electrical parameter features alone is not good.Considering the accuracy and stability of the prediction results,the effect of remaining life prediction of the LSTM model using the fusion features of electrical parameters and mechanical parameters as input is better. |
