Operation Anomaly Detection And Degradation Trend Prediction Of Power Swap Station Unlocking Tools For New Energy Vehicles

As the car industry advances,the harmful emissions from cars are getting worse,leading to growing concern about cleaner energy usage from the government.The use of new energy vehicles undoubtedly represents a future trend,so the country is quite concerned about the new energy vehicle industry in recent years,and the charging equipment of new energy vehicles has also been attached to importance by the country.As a new type of technology,power swapping brings the experience of a quick recharge for the users of new energy vehicles and creates huge profits for enterprises.However,the internal structure of the changing power station is complex.Once a problem occurs in a certain part,the whole power swap station may be shut down,which is undesirable for the enterprise.Therefore,the intelligent operation and maintenance of components in the power swap station is particularly important.The unlocking tool is an important tool for disassembling and installing battery packs in the power swap station.Its running status plays an important role in the overall health status of the power swap station.Therefore,the health status management of the unlocking tool is of great significance to the enterprise.However,there are few researches on the unlocking tool of power swap station,and there are some problems such as unknown fault characteristic frequency and complex working conditions.Therefore,this study aimed at the significance and difficulties of the research of the tool in industry,carried out the research of anomaly detection and degradation trend prediction for the tool,and developed the health status management system of the tool for the power swap station.In this paper,firstly,the basic structure and failure mode of the unlocking tool of the power swap station are analyzed,considering that the torque signal can directly reflect the gun head action of the unlocking tool,but the installation of the torque sensor will cause dynamic imbalance,we choose to collect the current signal and convert it into the torque signal,and introduce the process of collecting the current signal of the unlocking tool in detail.The working conditions of the unlocking tool are divided based on the expert experience of the enterprise and the operating principle of the unlocking tool in the power swap station.The results of the working conditions division and the actual power swapping process are used to select the subsequent research data.Secondly,aiming at the anomaly detection problem of the gun head system of the power swap station unlocking tool,this study established the anomaly detection model of the power swap station unlocking tool based on criterion fusion.By using the isolated forest method to detect the anomaly in the day dimension of the unlocking tool,the model can identify the abnormal gun head system in the timing well,but the interpretation of the anomaly detection results is insufficient.In order to find the anomalies of the additional tool in the work order dimension,an anomaly detection model of the work order dimension is established by Kmeans and control chart.This model can complement the isolated forest method well,but it has misjudgment.The anomaly detection method based on criterion fusion combines the results of two methods,which makes the model have higher accuracy and better interpretability at the same time.Then,aiming at the degradation trend prediction problem of the power swap station unlocking tool,this study proposed a degradation trend prediction model based on Long Short-Term Memory(LSTM)network.By designing the fitness weighted by monotonicity index and trend index as the index of feature selection,and using the three-dimensional feature points after dimensionality reduction to represent the running state of the gun head system of the unlocking tool in the three-dimensional coordinate system,the Euclidean distance between the three-dimensional feature points after the on-line time and the threedimensional feature points at the on-line time was calculated.This is used as a health indicator to evaluate the degradation of the unlocking tool.Wiener filtering and data segmentation were used to preprocess the Euclidean distance data to facilitate the subsequent model training and learning.In addition,an adaptive size feature window is designed,and the training dataset is obtained by sliding the window.Finally,the forecast time point is adjusted from the 161 st day of signal acquisition to the 181 st day of signal acquisition,and a set of prediction results were obtained by using the LSTM model,and the results were weighted by using the weight division on the time series.This method can accurately predict the degradation trend of the unlocking tools in advance and provide a basis for enterprises to maintain equipment in advance.Finally,this paper develops a health state management system of the power swap station unlocking tool and inserts the anomaly detection method and degradation trend prediction method into the system.The system realized torque data display and visualization,feature data display and visualization,anomaly detection,degradation trend prediction,and health report generation of the power swap station unlocking tool.The system passed the system function and performance test,providing application prospects for the health management of the power swap station unlocking tool.