On Line Temperature Monitoring And Fault Diagnosis Of Large Generator Rotor Based On Multi-source Data Fusion

With the vigorous support of new energy industry in China,the hydro-power industry is also developing rapidly in recent years,the installed capacity of hydro generators is also growing,and the proportion in the whole power system is also growing.As the core component of power system,the safe operation of hydro generator is particularly important.As the core part of hydro generator,the rotor is of great research value.Firstly,this paper introduces the development of temperature measurement of Hydro-generator rotor at present,analyzes and compares the temperature measurement methods of Hydro-generator rotor,and draws the conclusion that the error brought by direct temperature measurement is greater than indirect temperature measurement,and the cost of detection software and hardware is high,so this paper chooses indirect temperature measurement.Then,the indirect temperature measurement error is studied: the indirect temperature measurement interference can be divided into signal noise and generator parameter gradual change.In MATLAB software,the signal interference and generator parameter gradual change are simulated respectively.The simulation results show that: the inertia filter and the average filter method of de extremum can eliminate the signal interference noise better;the contact resistance in the gradual change parameters of generator can affect the rotor temperature The influence of degree measurement can be ignored,but the change of rotor reactance and trigger angle can not be ignored,which will cause great error to the measurement of rotor temperature.On this basis,the filtering algorithm and reactance voltage drop calculation algorithm are reasonably selected in the practical application of Huya power station to ensure that the rotor temperature measurement is superior to the accuracy requirements of the national standard.Then take the rotor turn to turn short circuit fault as an example,establish the model,simulation,record and analyze the fault waveform in MATLAB,extract the change characteristics of each electrical quantity and non electrical quantity during the rotor turn to turn short circuit.The fault diagnosis model based on BP neural network is established,and the sample space of fault data ? fault feature is obtained.BP neural network is trained to realize the whole process of fault data input ? fault feature recognition ? fault diagnosis.Finally,the integration of the fault diagnosis system and SCADA system isintroduced.The excitation voltage and excitation current are collected in the local layer.They are transmitted to the database server and the monitoring host through the communication manager.The MATLAB fault diagnosis platform accesses the database to complete the diagnosis of rotor turn to turn fault and display the fault diagnosis results.