| Metal oxide surge arresters are important overvoltage protection devices in power sys-tems,which are of great importance to the safe and efficient operation of power systems and their equipment.However,due to the long-term work of metal oxide arrester(MOA),they can be affected by temperature,humidity,over-voltages and other influence factors which could finally result in destroyed of the MOA.The destroyed MOA will explode in some special cir-cumstances which would bring great harm to the safety of equipment and staffs.Therefore,how to accurately and reliably monitor the MOA situations during their on-line work is the research emphasis in the field of intelligent lightning protection.In this paper,the aspects of noise suppression and utilization in the signals of on-line monitoring MOA are studied.The article proposed a de-noise method is proposed based on Complete Ensemble Empirical Mode Decomposition(CEEMDAN).What's more,the influ-ences of the harmonic voltage to the traditional capacitive current compensation method are analyzed,and the harmonic correction algorithm for MOA on-line monitoring are proposed on the base of the traditional method.Further,the particle swarm optimization(PSO)method is introduced to extract the resistive leakage current.The Gradient Descent is used to optimize the local optimization ability of PSO,and a hybrid particle swarm optimization(HPSO)in the monitoring of MOA is proposed.The concrete research contents are as follows:This paper puts forward a de-noise method which is based on Complete Ensemble Em-pirical Mode Decomposition.The detection signals with noise of MOA can be decomposed into a series of intrinsic mode functions(IMFs),and then the IMFs are analyzed by autocor-relation.In order to get useful components and noise components,the Savitzky-Golay filter is used to de-noise.After filtering,the remaining components are restricted into a new signal.A typical surge arrester has been simulated in MATLAB,and the results show that by de-noising of the continuous current of MOA,the root mean square error can be reduced by one order of magnitude,reaching 3.2098×10-5for un-aged and 0.0025 for ageing while the model before de-noising can reach 2.4509×10-4 for un-aged and 0.0173 for ageing,which means the pro-posed method can effectively reduce the noise to improve the data quality of on-line monitor-ing.Traditional capacitive current compensation method ignores the influence of the har-monic voltage,causing a huge error on the monitoring results.To solve this problem,a Har-monic correction and study for MOA on-line monitoring algorithm based on the relationship between the resistive current and capacitive current at voltage's zero-point is proposed.Ca-pacitance C of MOA and compensation coefficients of capacitive G was deduced.This meth-od eliminated the influence of the harmonic voltage and improved the extraction accuracy of resistive current.MATLAB simulations shown that the error of capacitance C,fundamental resistive current component iRl and the 3rd harmonic resistive current component iR3 were ?10%by this method,which could reduce the influence of harmonic voltage to the monitoring results effectively,reducing the extraction error of resistive current and improve the accuracy of the arrester on-line monitoring.According to the component characteristics of the metal oxide arrester leakage current,the optimization and fitting ability of PSO is utilized.Combining with the advantages of Gra-dient Descent method in local optimization,a hybrid particle swarm optimization algorithm is proposed to exact the resistive leakage current of MOA.It is found that the HPSO and the PSO are significantly different on the extraction of resistive leakage current.Simulated results show that the root mean square error(RMSE)of leakage current with HPSO(without har-monics:2.6291×10-6;with harmonics:2.4917×10-5)can be reduced by two order of magni-tude than those in PSO(without harmonics:1.0385×10-4;with harmonics:1.2×10-3);The er-rors of parameter a of HPSO(without harmonics:0.086%;with harmonics:1.594%)are much more smaller than those of PSO(without harmonics:9.895%;with harmonics:17.963%).This paper studies the extraction method of the MOA on-line monitoring signals.The de-noise method of the on-line monitoring signals and the resistive leakage current extraction algorithm are proposed based on the theories of Complete Ensemble Empirical Mode De-composition,autocorrelation analysis,harmonic correction,hybrid particle swarm optimiza-tion,which can better suppress the interface of the noise to the on-line monitoring signals,improve the reliability and accuracy of the monitoring results and has a certain theoretical significance and practical application value for the MOA on-line monitoring. |
PDF Full Text Request