Mechanical Condition Assessment Of On-load Tap-changer In Converter Transformer Based On Multi-feature Analysis

With the rapid development of HVDC technique in China,the reliability of converter transformer especially its on-load tap-changer(OLTC)to regulate the precise voltage has aroused more and more concern.Meanwhile the mechanical fault is the main fault type.Thereby,it’s necessary to find the reasonable and effective method to evaluate the mechanical condition of OLTC in converter transformer,as well as to early diagnose the hidden problem and improve the stability of OLTC and converter transformer.Considering the mechanical condition of OLTC is closely related to the vibration characteristics of switching process,the vibration theory has been analyzed first and some characteristic extraction methods have been proposed.Then the mechanical condition of OLTC will be assessed by all the characteristics.The main contents are as follows:In order to understand and obtain the characteristics of OLTC switch,the movement formula of contact hit has been formed and calculated based on the switching process of OLTC to get the average percussive force and acceleration curve.Then the 3-D finite element model of OLTC has been constructed and the whole acceleration characteristics have been obtained.The results have shown that the accelerations of X & Y directions are highest at the beginning of force.Then they are dissappared after the force ends.Meanwhile the acceleration of Z direction increases with the accumulation of force.The peak value occurs at the border of head cover and the maximum one is at the place nearest to the contact.The results demonstrate the movement characteristics of switching process and the whole vibration characteristics,which also provides the reference for the vibration test and choice of measuring points.The model machine of R type OLTC used in converter transformer is chosen as the experimental object.By setting different mechanical conditions of normal,contact looseness and insulated panel looseness,the vibration signals of each condition have been collected.Then the correspondence of contact movements and vibration peaks has been verified by the high-speed photographic technique.The results have shown that the signals under the same conditions are very similar,which verifies the repeatability of the tests.Moreover,the bursts cannot correspond to the hits of each contact with a 10～15ms deviation,which shows the influence of random factors on the vibration signal of converter transformer OLTC.Thus the characteristics of vibration signals need to be extracted from new point of view.Considering the relativity between waveform and switching process,the signal context method along with the Teager energy operator(TEO)has been proposed to describe the characteristics of OLTC vibration signal in time domain.The vibration waveforms of different OLTC conditions are compared by context histogram and the distribution of bins with maximum number of vectors.The results have shown that the TEO can exactly identify the vibration area.Meanwhile the signal context in this area can effectively reflect the variation of vibration waveforms under different conditions.Each waveform under normal condition is basically the same(c>0.94),which verifies the repeatability of tests.When the signal is different from that of normal condition(c<0.7),however,the fault may occur.In addition,the distribution of bins with maximum number of vectors is also different from the normal condition.Thus the mechanical condition of OLTC can be initially judged by signal context.Since the vibration signal of OLTC has much instantaneity and randomness,the chaotic dynamics of vibration signal are taken into consideration in this paper.The collected vibration time series have been reconstructed based on Phase Space Reconstruction theory.Then the reconstructed surface is drawn according to the high dimensional reconstructed points in high dimensional space,which is described by the proposed Lorentz Information Measure(LIM)based on the Gaussian curvatures and normal vectors of curved surface.The surface similarity index d has also been defined by LIM to compare the different reconstructed surfaces.The results have shown that the LIM can effectively reflect the variances between reconstructed surfaces under different conditions.The repeatability of normal condition is still satisfying(d>0.9),while the similarity between normal and fault is quite low(d<0.5).It can be further identified whether the switching device fault(d<0.3)or the appurtenance fault(0.3<d<0.5)is.Considering the differences and relativities of the vibration signals during the contacts switch of OLTC at multiple measuring points,the Narrowband Noise Assisted MEMD(NNA-MEMD)has been put forward to analyze the vibration signals at different measuring points in a uniform scale,as well as to effectively reduce the aliasing effect of current EMD methods.The power feature matrices of multidimensional vibration signals have been calculated and the power matrix similarity m has been defined as the comparative index.The results have shown that the power feature matrix based on each IMF can effectively reflect the mechanical characteristics of OLTC.The matrix similarity is high under normal condition(m>0.9),while it is low between normal and fault condition(m<0.7).Thus the energy viaration of fault condition can be precisely identified.In order to effectively separate the vibration signals of OLTC and transformer in real converter transformer under operation,the improved Fast-ICA method with self-adaptive convergence factor has been put forward to effectively separate two vibration signals from the collected mixed signal.Meanwhile the similarity index and relativity index of waveform have been defined to assess the similarity and independence of separated results.The results of simulation and experiment have shown that the similarity indexes between sepatated signal and original signal are all higher than 0.9,which proves that they are similar.While the relativity indexes between the separated signals are all lower than 0.1,which proves that they are irrelevant and the separation is effective.Besides,the irrelevance between vibration signals of OLTC and transformer verifies the feasibility of OLTC vibration extraction in the field.Finally,after combining the previously extracted multiple characteristics of OLTC vibration signals from perspectives of time,chaos and energy,the evaluation method of OLTC mechanical condition has been formed based on improved Analytic Hierarchy Process(IAHP)and Fuzzy Comprehensive Evaluation(FCE).In detail,the weights of all the characteristics related to the mechanical conditions are first calculated by IAHP.Then the evaluation memberships corresponding to the mechanical conditions have been determined by FCE using evaluation factor matrix.The evaluation results of offline and online tests match well with each other,meanwhile the determined evaluation level of OLTC mechanical condition is identical with the practical conditions,which leads to a more efficient,accurate,objective and comprehensive evaluation resul based on the proposed evaluation method.The findings of this paper can provide theoretical foundation and practical reference for mechanical condition assessment and online monitoring of OLTC in converter transformer,which is of great significance for the safe stable operation of OLTC and reduction of manitenence cost and labour force input.