Portal de Periódicos da CAPES

Evidence theory‐based identification of aging for capacitive voltage transformers

2016; Institution of Engineering and Technology; Volume: 10; Issue: 14 Linguagem: Inglês

10.1049/iet-gtd.2016.0540

1751-8695

Narges Daryani, Heresh Seyedi,

High voltage insulation and dielectric phenomena

IET Generation, Transmission & DistributionVolume 10, Issue 14 p. 3646-3653 Research ArticleFree Access Evidence theory-based identification of aging for capacitive voltage transformers Narges Daryani, Narges Daryani Faculty of Electrical and Computer Engineering, University of Tabriz, 29 Bahman Avenue, Tabriz, IranSearch for more papers by this authorHeresh Seyedi, Corresponding Author Heresh Seyedi hseyedi@tabrizu.ac.ir Faculty of Electrical and Computer Engineering, University of Tabriz, 29 Bahman Avenue, Tabriz, IranSearch for more papers by this author Narges Daryani, Narges Daryani Faculty of Electrical and Computer Engineering, University of Tabriz, 29 Bahman Avenue, Tabriz, IranSearch for more papers by this authorHeresh Seyedi, Corresponding Author Heresh Seyedi hseyedi@tabrizu.ac.ir Faculty of Electrical and Computer Engineering, University of Tabriz, 29 Bahman Avenue, Tabriz, IranSearch for more papers by this author First published: 01 November 2016 https://doi.org/10.1049/iet-gtd.2016.0540Citations: 5AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract In this study, an aging identification approach is proposed for capacitive voltage transformers (CVTs). Aging is an unavoidable phenomenon which emerges and spreads in equipment structure and mainly affects the insulation parts. The aging may affect transformer output in its primary stages. If it grows, it may cause serious damages to the equipment, substation and the staff. Accordingly, it is necessary to diagnose the aging of CVTs in its primary steps in order to implement the required maintenance and repair. The proposed aging identification method is an efficient method which collects data from different sources in digital substations to diagnose aging in CVT structures. Moreover, the possible external disturbances are also taken into account. A modified method is proposed based on the distance between the evidences. In addition, the algorithm is capable to identify the CVTs providing bad data arising from their aging. 1 Introduction Electric power systems face different steady state and transient faults due to their large scale, complex structure and variety of installed equipment. Aging is known as a long term phenomenon which unavoidably emerges and spreads within the network instruments. Exhaustion, sudden over-voltages, partial discharge (PD) and etc. may cause the mentioned incident which leads to system malfunction, financial penalties for utilities, damage to the environment and in the worst cases may pose a serious hazard to employees or the public. Accordingly, it is necessary to diagnose aging in its primary steps. Then, applying the preventive and restorative actions would be impressive in order to reduce the harmful consequences. In order to detect aging of the network equipment, an efficient condition monitoring (CM) process is required. CM is a technique which continuously monitors the operating characteristic of a physical system. Consequently, any changes of the monitored characteristic can be specified and used to schedule maintenance before serious deterioration or breakdown occurs [1]. CM includes the knowledge of failure mechanisms of individual parts or the integrated system, data acquisition and analysis, and the ability to evaluate the healthy condition of the system [2]. In recent years, based on economic concerns, engineers and system operators are interested in achieving a secure electrical system with minimum number of faults and deteriorations. Aging, as a preventable phenomenon can be monitored and diagnosed in primary stages in order to minimise the aftermath disorders. Consequently, researchers work on different techniques to implement an appropriate CM process with proper performance and high resolution. Therefore, lots of CM methods are introduced in literature with the aim of monitoring the aging of various system equipment [3–8]. As mentioned, PD is one of the aggravating factors of aging. This phenomenon occurs within the network equipment when the electric field exceeds the dielectric strength of the insulation within a localised volume. This is usually associated with the impurities in the dielectric material. Although the PD may be quite small in its early stages of development, the damaged area may grow if left unchecked. Various technologies have been developed to address the problem of PD monitoring in electrical substations. Monitoring the high frequency components is applied in [3, 4] in order to detect PD in power transformer insulators. As noted, the main effect of aging will be in the insulation part of equipment. In this regard, Cabanas et al. [5] presents online analysis of power transformer leakage flux as an appropriate alternative in order to assess machine integrity and detect the presence of insulation failures in the earliest steps. A vast majority of aging monitoring studies are dedicated to the monitoring of other system components such as rotor and stator windings in generators. Destruction of insulation integrity within a coil may cause internal short circuit. Therefore, health of the insulation in the structure of winding, mostly guarantees the accuracy of its performance. Several factors such as aging and corrosion result in damaging of the insulation of a coil. Variety of offline and online methods for monitoring the internal coils of motors and generators, have been proposed in the literature. Various parameters that affect the insulation dielectric properties can be monitored. Among the parameters that have been used in the literature, magnetic flux components of the PD, temperature and etc. can be monitored [6]. Moreover, in [7], CM of an induction motor's stator insulation is analysed. Furthermore, in [8], condition of capacitors installed in a converter structure is monitored. Therefore, aging as a destructive factor for insulation property, can be diagnosed efficiently. According to the literature, the necessity and importance of system monitoring is clear. In other words, disorders caused by time elapse which is known as aging, makes the online and continuous CM necessary. One common problem in electricity supply is the aging population of capacitive voltage transformers (CVTs). CVTs are widely used in transmission and distribution substations to provide proportional, secondary single or three phase voltages for protection, metering and control functions. The CVT comprises three basic components: a capacitor divider made of a group of high voltage capacitors and a lower voltage grounding capacitor(s), a voltage transformer and filter element which provides the single or three phase secondary voltage. Over years, the CVT components will degrade or experience over-voltages. This may result in capacitor element failure and, hence, the secondary voltage progressively loses its accuracy. More notably, the CVT may explode if sufficient number of capacitor elements fails. The explosion can break the porcelain shell and spread porcelain fragments and hot oil within the local area. These particles are serious threats to the staff safety and surrounding components. In addition, the CVT is commonly located close to the substation bus and, therefore, bus protection will clear the emerging fault. This can result in loss of supply to a large number of customers and possibly incur a penalty from the energy regulator [9]. According to Kasztenny and Stevens [9], the magnitude and also the negative sequence components of two groups of CVTs can be compared with each other. Inequality of these parameters illustrates the disorder of CVTs. However, the mentioned approach suffers from inappropriate accuracy. Despite the importance of efficient diagnosis of CVT aging, this issue has been rarely discussed in the literature. This paper proposes a monitoring approach in order to diagnose CVT aging efficiently and with high resolution. The proposed method is based on the evidence theory. Evidence theory was first introduced by Dempster in the 1960s, and was developed by Shafer [10]. Nowadays, evidence theory is widely used in many applications, such as multisensory fusion, pattern recognition, fault diagnosis and target tracking [11, 12]. According to the proposed method, measured data of each group of CVTs are considered as a piece of evidence, and the combination of such evidences is used as a tool to make the final decision. The rest of this paper is organised as follows. Section 2 introduces evidence theory and discusses Dempster–Shafer rule's evolution. In Section 3, the proposed approach is described. In Section 4, simulation results are presented and Section 5 includes conclusions. 2 Overview of evidence theory In this section, conventional evidence theory and its modifications are briefly discussed. 2.1 Basic definitions Definition 1.Let θ be the set containing N mutually exclusive and exhaustive hypotheses. The power set of θ is denoted by 2Θ = {A |A ⊆ Θ}, which is the set of all subsets of θ. Definition 2.A mass function, called basic probability assignment (BPA) function, is a mapping from 2Θ to the interval [0, 1] while satisfying the following conditions. (1) where is the empty set. For any A ∈ 2θ, m (A) represents a piece of evidence indicating the degree of belief corresponding to A. 2.2 Evidence combination without conflicts Considering n different BPAs, m1, m2, …, mn, the combination can be obtained by Dempster–Shafer's combination theory as follows. (2) where parameter K is calculated as follows. (3) Obviously, K may be interpreted as a measure of conflict degree among the distinct pieces of evidence. According to (2) and (3), if there is no conflict among pieces of evidence, the combination result will enhance the supporting degree towards the same hypotheses. On the other hand, considering conflict between pieces of evidences may yield conclusions different from what we expect. However, the conflict often exists among the collected information and increases with the number of sources. Therefore, conflict management is a major problem especially during the fusion of numerous information sources. 2.3 Evidence combination with conflicts Considering conflict between evidences, some modifications should be applied in order to enhance the algorithm efficiency. Accordingly, extensive studies have been performed and various methods have been proposed in the literature including modification of combination rules [13–16] and modification of combination models. The idea of dealing with conflicting pieces of evidence, first, has received much attention among the modification categories. Afterwards, the available evidences were combined using the conventional Dempster–Shafer rule. In [17], according to Murphy's approach, each piece of evidence gets an equal weighting coefficient. However, according to the mentioned approach, all pieces of evidence seem equally important. In order to deal with this problem, a new method is introduced on the basis of Murphy's work [18, 19]. Therefore, a weighted average approach is proposed in order to combine conflicting evidences. The weighting coefficient is a function of distance between different pieces of evidence. The proposed alternative is described as follows. Definition 3.Let m i and m j be two BPAs on the same frame of discernment, containing N mutually exclusive and exhaustive hypotheses. If the evidences are expressed as distinct values, the distance between these evidences is obtained by their difference. On the other hand, for the vector forms of evidences, the distance between two pieces of evidence is calculated as follows. (4) where m i and m j are the vector forms of m i and m j, respectively. Moreover, D is a 2N × 2N matrix whose elements are achieved as follows. (5) where the cardinalities of and are considered. It is worth noting to say that the factor 1/2 is needed in (4) in order to normalise d (m i, m j) and also guarantee that . According to (5), dij, presents a criterion of the similarity of two pieces of evidence in the interval [0, 1]. In this paper, the correlation coefficient is proposed as a parameter which provides the considered criteria. The absolute value of the correlation coefficient related to two signals, is a value in the range of [0, 1] which illustrates the similarity of the considered signals. The similarity degree of two pieces of evidence is introduced as follows. (6) The support degree of evidence is defined as follows. (7) Finally, the credibility degree of evidence m i is calculated as follows. (8) As mentioned, the credibility degree is used as a weighed coefficient of each piece of evidence. Evidences are combined as follows. (9) where the parameter K is calculated as the following. (10) According to the proposed approach, if one piece of evidence conflicts with others, the majority sources will affect the conclusion. 3 Proposed approach for aging identification in CVTs 3.1 Proposed approach description Nowadays in digital substations, it is possible to collect data from various groups of measuring instruments. In this paper, measured data of each group of CVTs is considered as a piece of evidence. Afterwards, the available pieces of evidence are combined using a data combination algorithm based on Dempster–Shafer's evidence theory. The proposed method is mainly based on comparing the evidences with each other. The incidence of aging in CVTs will affect the output. The output distortion is expected to be small at first, i.e. when only few capacitor elements fail, and gradually increases. Implementing an appropriate algorithm to detect and highlight the difference between pieces of evidence would be effective. In this research, the required evidences are obtained from received data of two separate CVTs (groups of CVTs). Various parameters can be extracted from the output signal and considered as the inputs of the algorithm. Since the target is to detect the aging of CVTs in its primary stages, choosing the appropriate parameters and features of the CVT signal will improve the algorithm performance. In this paper, various parameters such as the root mean square (RMS), negative sequence and etc. are studied. The achieved results will be compared and the best one will be determined. 3.2 Presentation of evidence and decision rules As there is no definite way to define the value of evidence, it is generally dependent on the application. In this section, an efficient method is used in order to establish the required evidence. The mentioned method is based on the distance between typical swatches and testing swatches. Let the frame of discernment be {Anormal, Aaged}, where Anormal and Aaged represent normal condition and aged condition of the considered group of CVTs, respectively. Suppose Xj as a testing swatch which is, in fact, a certain parameter obtained from the CVTj output signal. On the other hand, Nj and Fj are considered as the set of normal and aged typical swatches of the same CVT, respectively. Where the set {Nj, 1, Nj, 2, …, Nj, n 1} contains n1 typical swatches for Anormal and set {Fj, 1, Fj, 2, …, Fj, n 2} contains n2 typical swatches for Aaged. Accordingly, the distance between Xj and both sets of typical swatches are introduced as P and Q and measured as follows. (11) (12) Afterwards, the evidences of CVTj are calculated as follows. (13) (14) After combination of different pieces of evidence, the decision can be made under the following rules. If , the fusion result will illustrate the normal condition. If , the fusion result will show the aged condition. 3.3 Flowchart of the proposed approach In this section, the process of the proposed approach for aging detection is described. In addition, a flowchart is prepared in order to facilitate understanding of the procedure. The mentioned flowchart is illustrated in Fig. 1. Step 1 : The process starts with some initialisation operations such as selecting appropriate typical swatches. Step 2 : To collect data, a moving window loads sampled data of one cycle from two (two groups of) redundant CVTs. Step 3 : Afterwards, an appropriate parameter of the sampled data for each (groups of) CVTs is calculated. For example, RMS, negative sequence etc. Step 4 : In the next step, the evidence for every group of CVTs is established using (11)–(14). Step 5 : Then, the similarity degree Sim(m i, m j), the support degree Sup(m i) and the credibility degree Crd(m i) are calculated using (6)–(8). Step 6 : The credibility degree is assumed as the weighted coefficient of each piece of evidence. Afterwards, all pieces of evidence are combined according to (9). Step 7 : Finally, the existence of aging phenomenon is checked, the normal or aged situation is reported and the aged CVT is identified. Fig. 1Open in figure viewerPowerPoint Flowchart for the proposed aging detection method 4 Simulations and analysis 4.1 Description of simulation principles To demonstrate the effectiveness of the proposed method, simulation studies are carried out. The simulations and analysis are carried out in PSCAD and also MATLAB 9 using PC with Intel Core i5 2.5 GHz and the RAM of 6GB. In order to reduce the complexity, all the CVTs are assumed to have the same structure. The main parameters of the considered CVT are available in Table 1. Table 1. Main parameters of the considered CVT Parameter Value C1 8.35 (nF) C2 190.67 (nF) compensating reactor 80 (H) transformer ratio 87 Within the considered CVT, C1 consists of 137 similar capacitors connected in series while C2 contains six capacitors. As mentioned, the capacitive voltage transformers are installed in substations and are mainly responsible for supplying low voltage instruments. A CVT is fed by the line-to-ground voltage and provides a nominal voltage (generally (V)) as the output. In order to simulate such a situation, a typical feeder is provided in PSCAD environment which is schematically illustrated in Fig. 2. Fig. 2Open in figure viewerPowerPoint Schematic of a simulated typical feeder Creation of typical swatches. Typical swatches include two main categories: normal operation data and aging situation data. Since various parameters of the CVT output signal are observed, the corresponding typical swatches will be needed. In order to achieve the typical swatches, simulations should be performed through the ideal conditions which means the absence of any kind of disorder such as noise, measuring error and etc. Aging mainly affects the insulation parts of the instruments. In a capacitive voltage transformer, aging mostly leads to dielectric failure in capacitors of the capacitive divider. Therefore, in this paper, the CVT aging is simulated by capacitor outage and by the changes of reactance value in the divider section. Therefore, the aging swatches include measured data related to various degrees of aging, modelled by capacitor outage from the capacitor divider. Finally, RMS, negative sequence, zero sequence and the D parameter of park transformation of the data logged per cycle are taken as typical swatches. Creation of testing swatches. Testing swatches are produced based on typical swatches by adding some conditions which lead to more realistic simulations. 4.2 Case studies and analysis In this section, first, the simulations and analysis are carried out considering four different parameters of CVT output as the input of the proposed algorithm. After choosing the appropriate parameter as the input of the algorithm, simulations are performed for the situation considering external faults besides aging. In this case, the authors aim to investigate the algorithm accuracy. Finally, some modifications are proposed in order to improve the algorithm performance and to enhance its accuracy. The corresponding simulations are performed and the obtained results are reported. 4.2.1 Case 1: RMS of CVT output signals In this case, per cycle RMSs collected from two separated CVTs, are considered as the inputs of the detection algorithm. In order to simulate real conditions, measuring error and also noise are added to the available data and then considered as testing swatches. The fusion results of testing swatches during five seconds time duration are given in Fig. 3. The signal to noise ratio (SNR) is considered equal to 30 dB. Fig. 3Open in figure viewerPowerPoint Fusion results of testing swatches of Case 1. (SNR = 30) According to Fig. 3, when both CVTs work without aging, the value of [m (Anormal) – m (Aaged)] is close to 1. Thus, the fusion result indicates both CVTs are in normal condition. On the other hand, for seven and more capacitor outages, the decision parameter would be close to 0. Consequently, the algorithm distinguishes that one of the CVTs experiences aging. Now to detect the defective CVT, the mean value of m (Anormal) is calculated for each CVT. The CVT with the mean value of m (Anormal) close to 1, will be introduced as the hale and the other one will be the aged. At this stage of simulation studies, noises with different SNRs are added to the signals. The obtained results are reported in Table 2. Table 2. Accuracy of the algorithm for Case 1 considering different noise intensities Noise intensity Accuracy of the algorithm (Total number of capacitors = 137) Threshold, % SNR = 40 three and more capacitor outage 3% outage SNR = 35 five and more capacitor outage 4% outage SNR = 30 seven and more capacitor outage 6% outage SNR = 25 13 and more capacitor outage 10% outage Clearly, the stronger the noise is, the extracted values from the signal will differ from ideal values. Consequently, correct diagnosis and efficient performance of the algorithm will be affected. In power systems, the approach of symmetrical components is used to facilitate analysis of unbalanced three-phase power systems. The idea is based on the fact that an asymmetrical set of N phasors can be expressed as a linear combination of N symmetrical sets of phasors. In the most common case of three-phase system, the symmetrical components are named as positive, negative and zero. Physically, in a three phase winding a positive sequence set of currents produces a normal rotating field, a negative sequence set produces a field with the opposite rotation, and the zero sequence set produces a field that oscillates but does not rotate between phase windings. Consequently, by utilising the symmetrical components, the asymmetric fault analysis would be more tractable. In the following sections, the negative and zero sequence components of groups of CVT are used as the inputs. 4.2.2 Case 2: negative sequence components of groups of CVTs In this section, negative sequence components of two groups of CVTs are calculated. Each group contains three CVTs installed on three phases of a feeder. The same analyses as previous section are performed. Simulations have been accomplished for normal state and different levels of CVT aging. The algorithm outputs during five seconds for the mentioned states are illustrated in Fig. 4. Fig. 4Open in figure viewerPowerPoint Fusion results of testing swatches of Case 2. (SNR = 30) If one of the involved CVTs faces aging, {m (Anormal) − m (Aaged)} is between 0.5 and 1 and close to 1, and the fusion result correctly indicates the non-aged status of CVTs. On the other hand, if the above-mentioned difference is close to zero, one of the CVTs with the lowest amount of m (Anormal) is detected as the aged one. The algorithm with the considered input (negative sequence) is able to detect outage of 10 capacitors and more which is equal to 7% measurement accuracy. Moreover, Table 3 illustrates the accuracy of algorithm for this case, considering different noise intensities. Table 3. Accuracy of the algorithm for Case 2 considering different noise intensity Noise intensity Accuracy of the algorithm (total number of capacitors = 137) Threshold, % SNR = 40 five and more capacitor outage 4% outage SNR = 30 10 and more capacitor outage 8% outage SNR = 25 15 and more capacitor outage 11% outage 4.2.3 Case 3: zero sequence components of groups of CVTs In this case, the zero sequence components of two groups of CVTs are considered as inputs. All simulations and calculations are similar to the previous section. One of the CVTs is assumed to confront aging. Fig. 5 illustrates the value of {m (Anormal) − m (Aaged)} during five seconds for normal state and also considering different levels of aging in one CVT. Fig. 5Open in figure viewerPowerPoint Fusion results of testing swatches of Case 3. (SNR = 30) According to Fig. 5, the proposed algorithm with the considered inputs is able to diagnose outage of 10 capacitors and more which is equal to 8% threshold. 4.2.4 Case 4: Park transformation parameter of groups of CVTs In electrical engineering, direct–quadrature–zero is a mathematical transformation to simplify the analysis of three-phase circuits. In fact, the DQ0 transform is often referred to as Park transformation. In the case of balanced three-phase circuits, application of the DQ0 transform reduces the three AC quantities to two DC quantities. Further details on the calculation of Park quantities are available in [20]. This section utilises the direct (D) parameter of Park transformation as the input of the proposed algorithm and its performance is surveyed. Fig. 6 illustrates the results of simulations and calculations of this case during five seconds of time interval. Fig. 6Open in figure viewerPowerPoint Fusion results of testing swatches of Case 4. (SNR = 30) Clearly, outage of 15 capacitors and more is diagnosable considering D parameters of two groups of CVTs as the inputs of the proposed algorithm. According to the prepared four cases and obtained results, the first case considering the RMS values of two separated CVTs as the algorithm inputs, demonstrates the best performance. The reason can be explained as follows. The considered parameters such as negative sequence, zero sequence and direct parameter are sensitive to distortions. On the other hand, from one point of view, RMS value is a representative of the average of available data. Hence, using RMS value, the influence of noise would be faded and the algorithm would perform more efficiently. 4.2.5 Simulation studies considering external disorders Up to now, various parameters of CVT outputs have been utilised as the inputs of the algorithm with the aim of achieving more accuracy in aging detection. In this section, the accuracy of proposed algorithm while considering probable external faults is investigated. Malfunction of a measuring instrument (a CVT) occurs for two reasons. First, the disruption of the internal structure of the equipment, results in its wrong performance. Second, if the input signal has disorder for any reason, it will affect the appropriate operation of the device. Accordingly, the efficient performance of a measuring device, such as a CVT, may be affected by either internal or external disorders. Aging of a CVT is known as an internal distortion which expands during time and leads to capacitor elements breakdown. The voltage drop of a feeder is a probable phenomenon which may exist for a long period of time and known as an external fault for a CVT. In this section, the involved CVTs are fed from a line which has the value of voltage equal to 0.95 p.u. Efficiency of the proposed algorithm in detecting the aged CVTs is studied. The corresponding simulations are executed similar to case 1 while considering input under-voltage. Fig. 7 illustrates the results of simulations and calculations of this case during five seconds of time interval. Fig. 7Open in figure viewerPowerPoint Fusion results of testing swatches considering both internal and external faults. (SNR = 30) According to Fig. 7, the proposed algorithm is capable to diagnose the normal and aging situations individually. However, the considered external under-voltage affects the accuracy of the algorithm. According to obtained results in case 1, i.e. normal operation, the outage of seven capacitors (and more) is detectable using the proposed algorithm. However, in the case of external under-voltage, the algorithm will be able to detect aging after breakdown of fourteen capacitors. Hence, at least 11% of capacitor elements should fail so that the algorithm could detect aging. Comparing the obtained result with the corresponding one in Table 2, demonstrates that the detection accuracy is almost reduced by half. 4.2.6 Modifications of the proposed algorithm The previously discussed results demonstrate the impact of disturbances, such as noise and external disorders, on the efficiency of the proposed algorithm for aging detection. In this section, some modifications have been proposed and implemented in order to enhance the accuracy of the algorithm. In this regard, each sampling period, increases from a cycle to a second. Based on the fact that aging is a long term phenomenon, per second sampling would be an acceptable period. Increasing the sampling period leads to reducing the impact of temporary disturbances such as noise. Considering a 50 Hz system, each sampling period contains 50 cycles of samples and consequently 50 values are obtained for RMSs. Accordingly, the algorithm receives vectors of evidences as inputs. Considering the mentioned assumptions, simulations are repeated for case 1. Fig. 8 illustrates the obtained results. Fig. 8Open in figure viewerPowerPoint Fusion results of testing swatches for Case 1 (modified approach). (SNR = 30) According to Fig. 8, performance of the algorithm is significantly improved. The algorithm with the proposed modifications is capable to detect three capacitor outages and more, which is equal to 2.2% of total capacitor elements. Comparing with the corresponding situation in case 1, the detection threshold is decreased by half which means the accuracy is enhanced. Consequently, it can be claimed that the proposed modifications, efficiently improve the performance of algorithm. 5 Conclusions Considering the development of electrical network, power system operation is dependent on the accuracy of information received from the measuring instruments. Various factors including transient and steady state phenomena may disturb the accurate performance of measuring devices. Aging, as a long term and unavoidable incident, occurs and spreads in network equipment which mainly affects the insulation parts. The mentioned incident may lead to misoperation of the aged instrument. To address this issue, an efficient approach based on evidence theory is proposed for CVT aging identification. The proposed method operates based on comparing the collected measurements from several CVTs. In this case, CVTs providing bad data can be identified. The effectiveness and performance of the proposed approach are validated through computer simulations. Moreover, some modifications are proposed in order to increase the accuracy of the algorithm. Simulation results and analysis demonstrate that the mentioned modifications improve the performance of the algorithm. 6 References 1Han, Y., Song, Y.H.: ‘Condition monitoring techniques for electrical equipment – a literature survey’, IEEE Trans. Power Deliv., 2003, 18, (1), pp. 4– 13 (doi: 10.1109/TPWRD.2002.801425) 2Steed, J.C.: ‘Condition monitoring applied to power transformers an REC experience’. Proc. IEE Colloquium Condition Monitoring Large Machine Power Transformers, 1997, pp. 41– 47 3Judd, M., McArthur, S., McDonald, J. et al.: ‘Intelligent condition monitoring and asset management. Partial discharge monitoring for power transformers’, Power Eng. J., 2002, 16, pp. 297– 304 (doi: 10.1049/pe:20020603) 4Singh, J., Sood, Y.R., Jarial, R.K.: ‘Condition monitoring of power transformers-bibliography survey’, IEEE Electr. Insul. Mag., 2008, 24, pp. 11– 25 (doi: 10.1109/MEI.2008.4591431) 5Cabanas, M.F., Melero, M.G., Pedrayes, F. et al.: ‘A new online method based on leakage flux analysis for the early detection and location of insulating failures in power transformers: Application to remote condition monitoring’, IEEE Trans. Power Deliv., 2007, 22, pp. 1591– 1602 (doi: 10.1109/TPWRD.2006.881620) 6Stone, G.: ‘Condition monitoring and diagnostics of motor and stator windings–A review’, IEEE Trans. Dielectr. Electr. Insul., 2013, 20, pp. 2073– 2080 (doi: 10.1109/TDEI.2013.6678855) 7Siddique, A., Yadava, G., Singh, B.: ‘A review of stator fault monitoring techniques of induction motors’, IEEE Trans. Energy Convers., 2005, 20, pp. 106– 114 (doi: 10.1109/TEC.2004.837304) 8Buiatti, G.M., Martín-Ramos, J.A., Garcia, C.R. et al.: ‘An online and noninvasive technique for the condition monitoring of capacitors in boost converters’, IEEE Trans. Instrum. Meas., 2010, 59, pp. 2134– 2143 (doi: 10.1109/TIM.2009.2032960) 9Kasztenny, B., Stevens, I.: ‘Monitoring ageing CCVTs practical solutions with modern relays to avoid catastrophic failures’. Power Systems Conf. Advanced Metering, Protection, Control, Communication, and Distributed Resources, 2007. PSC 2007, 2007, pp. 187– 208 10Shafer, G.: ‘ A mathematical theory of evidence’ ( Princeton University Press, Princeton, 1976) 11Hall, D.L., Llinas, J.: ‘ Handbook of multi-sensor data fusion’ ( CRC Press, Boca Raton, 2001) 12Xu, Z., Liu, M., Yang, G. et al.: ‘Application of interval analysis and evidence theory to fault location’, IET Electr. Power Appl., 2009, 3, (1), pp. 77– 84 (doi: 10.1049/iet-epa:20070439) 13Yager, R.R.: ‘On the Dempster–Shafer framework and new combination rules’, Inf. Sci., 1987, 41, (2), pp. 93– 138 (doi: 10.1016/0020-0255(87)90007-7) 14Smets, P.: ‘The combination of evidence in the transferable belief model’, IEEE Trans. Pattern Anal. Mach. Intell., 1990, 12, (5), pp. 447– 458 (doi: 10.1109/34.55104) 15Dubois, D., Prade, H.: ‘Representation and combination of uncertainty with belief functions and possibility measures’, Comput. Intell., 1998, 4, (3), pp. 244– 264 (doi: 10.1111/j.1467-8640.1988.tb00279.x) 16Lefevre, E., Colot, O., Vannoorenberghe, P.: ‘Belief functions combination and conflict management’, Inf. Fusion, 2002, 3, (2), pp. 149– 162 (doi: 10.1016/S1566-2535(02)00053-2) 17Murphy, C.K.: ‘Combining belief functions when evidence conflicts’, Decis. Support Syst., 2000, 29, (1), pp. 1– 9 (doi: 10.1016/S0167-9236(99)00084-6) 18Jousselme, A.L., Grenier, D., Bosse, E.: ‘A new distance between two pieces of evidence’, Inf. Fusion, 2001, 2, (2), pp. 91– 101 (doi: 10.1016/S1566-2535(01)00026-4) 19Yong, D., WenKang, S., ZhenFu, Z. et al.: ‘Combining belief functions based on distance of evidence’, Decis. Support Syst., 2004, 38, (4), pp. 489– 493 (doi: 10.1016/j.dss.2004.04.015) 20Ferraz, R.G., Iurinic, L.U., Filomena, A.D. et al.: ‘Park's transformation analytical approach of transient signal analysis for power systems’. IEEE North American Power Symp. (NAPS)., 2012, pp. 1– 6 Citing Literature Volume10, Issue14November 2016Pages 3646-3653 FiguresReferencesRelatedInformation