The State Changes Analysis And Aging Predicition Of High Voltage Remote Detection Module Based On Machine Learning

The main function of the remote module of the optical measurement system in the converter station is to realize the mutual conversion of electrical and optical signals in the system.As the core equipment of the HVDC(High Voltage Direct Current)transmission system,its operational reliability is of great significance to the safe and stable operation of the power system.However,due to the complicated operating environment of the remote module,in recent years,the optical measurement system of the ultra-high voltage direct current transmission company breakdowns frequently caused by the remote module.However,the research on its operation status detection and fault diagnosis technology is still blank in the field,its failure trend can’t be predicted,its operation status is difficult to judge,operation and maintenance specifications are relatively scarce,and equipment life cycle management has no technical support.Therefore,it is of great practical significance to carry out research on the detection technology of remote module operation status.First of all,based on the multiple linear regression method,this paper establishes a remote module life prediction model to realize real-time prediction of the remaining life of the remote module.This paper builds a test platform to study the characteristics of remote module aging and the main factor that leads to remote module aging,and analyze the impact of external operating environments such as different input voltages,different operating temperatures,and different input optical powers on the operating status of remote modules.Analyse and quantificat the inevitable relationship between the operating state characteristic parameters of the remote module and external environmental factors,the test results show that the environment temperature has a greater impact on the operating state of the remote module,and other factors of the external environment have no significant impact on the operating state of the remote module.Through the test and result analysis of remote modules under different operating conditions at different operating temperatures,the method for determining the operating state threshold function of the remote module and the function of the operating state threshold varys with temperature are obtained.The operation status of the remote module of the Southern Power Grid field test is judged based on the threshold function.Chapter 4 is based on the LOF outlier detection algorithm in the unsupervised machine learning algorithm to detect the operating status of the remote module in the field test of China Southern Power Grid.The results of the algorithm are the same as the conclusions of Chapter 3,which proves the operating status diagnosis model of the remote module is of validity and reliability.The diagnostic model provide a reasonable and effective method for China Southern Power Grid and other power grid companies to judge the operating status of the remote module.In order to optimize the remote module operating state diagnosis model,this paper evaluates the importance of each feature parameter based on the entropy weight method,that is,assigns a weight to each feature parameter in the model.The results show that the model is improved compared with the previous model.At last,this article conducts high and low temperature cycle accelerated aging tests on some remote modules retrieved from the field,so that the characteristic parameters of the remote modules reach the fault threshold set in Chapter 3,and the performance of different modules is measured.The characteristic parameters and remaining service life after different time aging tests are stored in the My SQL database,the remaining life prediction model of the remote module was established through the multiple linear regression method in the supervised machine learning algorithm in order to predict the remaining life of the remote module working in the field in real time.Finally,a simple remote module management system application software was designed by the Python programming language.The application software can train the life prediction model,predict the life of the remote module and check the operating status information of the module,which is convenient for the maintenance staff without basic programming.These research results provide a theoretical basis and experimental examples for the research of remote module operation status detection technology.