Research Based On Deep Learning Of Condition Monitoring And Fault Diagnosis Of Combinated OLTC

Power transformer is important equipment in power network,it is equipped with OLTC.When voltage fluctuation occurs in the network,does not interrupting the load current,the OLTC can be adjusted transformer output a stable voltage,so as to ensure the stability of the voltage quality of the grid to fulfil the power supply requirements.As an important component of transformer,OLTC is also the only part with mechanical movement inside of transformer.Its operating condition is directly related to voltage quality stability,safety and reliability in the network.Condition evaluation and fault diagnosis of OLTC are of great significance to ensure the safe and stable operation of power system.This thesis uses combinated OLTC as the research object,through studying its mechanical structure principle,electrical mechanism and switching signal,to find out the relationship between the specific parts closely related to the motion state phenomenon of its internal mechanical parts and analyze the relationship between waveform and time changes during COS switching,clockwise and counterclockwise switching,determine the switching time limit of each component in integration with design specification of OLTC and routine test database to acquire its non-intrusive internal signal.Aiming at the relation between the CAM rotation Angle of the combinated OLTC driving mechanism and the position of the internal mechanical parts of the tap changer,the mathematical model and mathematical operation are established.Determine the rotation Angle and position set of mechanical components inside the tap switch.A torque sensor suitable for installation in the combinated OLTC system is selected to measure the driving torque,and the current or voltage signal information which can reflect the change of the mechanical running state of the tap changer is collected.Through sample analysis and calculation of the switching torque limit of each component,and the key indicators suitable for reflecting the condition monitoring of the tap changer are sorted out.The D-S evidential theory is applied to evaluate the condition state of the mechanical parts of the combinated OLTC.According to the data provided by the tap changer condition monitoring system,the combinated OLTC condition evaluation model based on the evidence theory is built.Integrating the historical data and the current data,the deep learning theory is applied to the fault diagnosis of the combinated OLTC.The sparse algorithm is introduced into the DBN model formed by the RBM.Through the test comparison between the pre-training set of the unlabeled sample dataset and the labeled sample dataset,the method still gets a high accuracy under the condition of fewer training samples.Finally,the fault diagnosis process and fault diagnosis mode of combinated OLTC based on Logsum DBN are constructed.Through the test of unlabeled samples,the test results are consistent with the research requirements of fault diagnosis of the OLTC.Using VVS III 400Y-76-10193 W product as the experimental sample,the application of GX WORKS2,LABVIEW,MATLAB,PC and other computer programming technology and hardware system equipment,to obtain its switching operation experiment data,integarted with historical data,through evaluation its 36 times operation sequence test data at different condition of compoents,the condition evaluation and fault diagnosis of the mechanical parts of the combinated OLTC are realized,which provides a reliable decision basis for the condition evaluation and fault diagnosis of the combinated OLTC.