| With the rapid growth of electricity demand,there is an imbalance between electricity load and power generation in many countries and regions around the world.Ultra High Voltage Direct Current(UHVDC)transmission system has advantages in high capacity and low loss,and its development is rapid.Especially in China where the area is large and the population distribution is uneven,in addition to traditional fossil fuel power generation,new energy power generation bases such as wind power and solar energy are far away from load centers,UHVDC transmission systems has become one of the main methods to solve the problem.As the key equipment of UHVDC transmission system,the converter transformer takes an important operational responsibility,so its insulation status is particularly important for the safe operation of the system.The valve side winding of the converter transformer needs to withstand the combined AC-DC voltage.Partial discharge(PD)is a kind of discharge phenomenon that can reflect the process of insulation failure.The PD characteristics of different defect models under different voltage forms are not the same.The PD characteristics under AC voltage only and DC voltage only have been widely studied,but the research on the PD characteristics under combined AC-DC voltage is not deep enough.Therefore,the PD characteristics of the typical defect models of the oil-pressboard insulation under combined AC-DC voltage and the application methods to reflect the transformer insulation state have become important issues.To this end,an experimental platform for PD of oil-pressboard insulation under combined AC-DC voltage is established in this paper.According to the actual situation,four typical defect models are set up.The pulse current method is used to record the experimental data for each model.Then,based on these experimental results,the PD characteristics of each model are summarized,and the discharge mechanism is analysed.Finally,a PD development degree assesment method for oil-pressboard insulation under combined AC-DC voltage is proposed,and the defect type of PD source is identified.The main contents and conclusions are as follows:1.Based on the PD experimental platform,four typical defect models of oil-pressboard insulation are established.They are needle-plate model,ball-plate model,surface discharge model and air gap model.The pulse current method is used to collect PD experimental data of these four defect models under 1:1 combined AC-DC voltage.The time trends of maximal apparent charge(Qmax)and repetition rate(N)are used to show the overall development of PD.Phase Resolved Partial Discharge(PRPD)spectra are used to summarize the phase distribution of PD pulses at the selected time point.It can be concluded that in the four models,the PD development processes of the needle-plate model and the ball-plate model are longer,while the PD process of the surface discharge model and the air gap model are shorter;in the steady discharge period of the four models,the needle plate model has the lowest apparent charge level.In the PD processes of the four defect models,the needle-plate model,the ball-plate model and the surface discharge model all have a relatively long initial period of discharge,and the discharge process of the air gap model is in a steady rising state before the pre-breakdown stage.During the whole PD process of the needle-plate model,it shows that the apparent charge and repetition rate in the negative half cycle are significantly higher than the positive half cycle;during the PD process of the ball-plate model,the apparent charge in the negative half cycle is significantly higher than the positive half cycle,but the PD repetition rate in the positive half cycle after entering the steady discharge period is slightly higher than the negative half cycle;the surface discharge model also has higher apparent charge and repetition rate in the negative half cycle;during the whole PD process of air gap model,the apparent charge is slightly higher in the negative half cycle than the positive half cycle,but the repetition rate he positive half cycle is slightly higher than the negative half cycles.2.ANSYS finite element software is used to simulate the electric field distributions of the four defect models under AC and DC voltage components,and the charge injection distributions of the four models under DC voltage component are discussed.The discharge characteristics are explained by the mechanism.The needle-plate model simulates the defect type of extremely uneven electric field.The electric field strength around the needle tip is the highest and PD occurs first there.The pressboard under the needle tip is damaged more and the breakdown occurs below the needle tip.The ball-plate model simulates the defect type of slightly uneven electric field.The electric field strength in the oil gap between the ball electrode and the pressboard is the highest,which is the main position of PD.The surface discharge model simulates the type of creepage defect at the oil-pressboard interface.The locations near the electrodes are with the high electric field strength.PD and charge injections that occur near the electrodes eventually cause insulation failures along the surface and produce flashovers.The air gap model simulates the PD when there is an air gap defect in the pressboard.Because the AC voltage component has a greater influence on the internal insulation of the air gap than the DC voltage component,the PD characteristics of air gap model under 1:1 combined AC-DC voltage are closer to the the PD characteristics under AC voltage component.3.The O-AT-N pattern for PD analysis under combined AC-DC voltage is proposed.The established Hn(?T,?)spectrum can clearly reflect the PD characteristics of the four defect models.It does not require strict calibration or normalization of PD amplitudes and is highly adaptable.For each sample of the H,(?T,?)spectrum,a matrix centroid C(xc,yc)can be obtained as a feature that comprehensively expresses the PD phase,time interval and repetition rate.Variational Mode Decomposition(VMD)can smoothly extract the primary development trend Tr(xt,yt)of the features and reduce the adverse effects of randomness in the PD development.The degree of PD development 'based on the Tr(xt,yt)of each defect model can be assessed by continuous information without considering the boundary point problem caused by the traditional PD stage division.It is very suitable for on-line monitoring and the establishment of insulation status history files of the electrical equipments based on big data.By normalizing Tr(xt,yt)and converting it into dimensionless index R(X,Y),it is easier to evaluate the PD development degree of each defect model,and to make the comparison of the PD characteristics between different defect models more concise and clear.Based on the experimental basis in this paper,the respective warnings(X,Y)for the four models are recommended.4.The waveform of the PD pulse is combined with the phase information as the input feature,and the feature dimension is normalized by the zero-padding method.For the case of large sample size and high feature dimension,Random Forests(RF)algorithm is used for the recognition of the PD defect types.In the case of a mixture of PD pulses from two models,the recognition rate can reach more than 90%;and in the case of a mixture of PD pulses from four models,the recognition rate can reach more than 85%.Compare with the recognition effect of Least Squares Support Vector Machines(LSSVM)in the recognition of PD defect types in the case of large amounts of data,comprehensively considering the recognition rate,calculation time and adaptability,RF has better performance.In summary,the PD characteristics of four common oil-pressboard insulation defect models under 1:1 combined AC-DC voltage are studied through experiments and mechanism analysis.?-?T-N analysis pattern for PD was proposed and the PD development degree assessment method based on the continuous history information is set up.Random forest algorithm is used to recognize the type of PD defects with good results,and it can provide a basis for the fault diagnosis of the convertor transformer insulation status assessment. |
PDF Full Text Request