Causes of failure of Distribution Transformers

in the East Zone of Caldas

O.J. Soto Marín, E.A. Cano Plata, A.J. Ustariz Farfan, L. F. Díaz Cadavid Electrical, Electronic and Computer Engineering Program

Universidad Nacional de Colombia, Branch Manizales

email: ojsotom@unal.edu.co, eacanopl@unal.edu.co, ajustarizf@unal.edu.co, lfdiazc@unal.edu.co

Summary – One of the main problems that the Colombian’s electrical utility faces, is the high failures rate of distribution

transformers, this situation leads to long power supply interruptions, affecting the customer’s service. These interruptions affect the economics aspects of the electrical

utility, like materials losses and repair costs. This paper undertakes the study of failures in distribution transformers;

the objective is to find a way to reduce the high failure rate that this type of equipment currently presents in the eastern rural zone of the department de Caldas, especially in the areas of

greatest keraunic level. The methodology implemented enables root cause analysis, the proposal of remedial actions and the validation of the proposed solutions. Key words: Fault trees, Inductive loops, Partial discharges, SPD, Transformer failure.

I. INTRODUCTION

Home electricity has become a basic necessity in the everyday life of families in developed and developing countries. The cost of home energy supplied by the electrical network is the lowest when compared to the energy provided by different types of conventional and unconventional processes. Some of the non-conventional systems that use renewable power, such as solar photovoltaic, are now competing with electrical networks in rural locations. The expansion of the electrical network in the rural area must be justified because: • Rural areas have only a few consumers. • Remote locations are mostly inhabited by

consumers with few prospects for the industrial and commercial sector.

• Customers from the rural area are primarily consumers whose demand increases very little and do not contribute to the improvement of poor load factors.

In the case of Colombia, most places in rural distribution companies are populated by very low-income groups with little potential to support the development of any type of industry in the future. For this reason, the decision to electrify these places should be made after a thorough technical and financial evaluation, with a subjective socio-economic contribution and without forgetting the overall benefit to society. Despite this, electrical companies in Colombia have made great efforts to expand electrical networks to all rural areas of the country. That is the case of the electrical utility Central Hidroeléctrica of Caldas (CHEC), which has improved its coverage level and has reached about 97% of total coverage. Table 1

shows the distribution of users covered in the eastern area CHEC. It is noteworthy that despite repeated efforts to improve the power supply in this area, geographical and atmospheric conditions often deteriorate and/or damage the network equipment used for the distribution of energy. Thus, the work becomes more complex and the outcome is higher downtimes affecting users, and higher costs affecting electrical utilities.

Table 1 - Electrification in the east zone chec

Location Installed meters

Installed transformers

Rural 21.205 2.120 Urban 328.88 656

For the electrical utility CHEC, and particularly eastern rural zone of Caldas, there are around 253 failed distribution transformers on average per year, with a mean lifetime of 2.73 years for transformer. A case study in the eastern rural zone of CHEC is presented and the implementation of the method used to classify cases of failure and failure rates through technical fault trees is presented.

II. DESCRIPTION OF THE PROBLEM

Most studies focus their attention on overvoltage caused by lightning as the main cause of transformer failures in Colombia. This country is located in one of the areas with the highest rate of ground flashes density (GFD) on the planet [1], [2]. Samaná, a small town in the department eastern of Caldas, registers a GFD level of 10-8 flashes/km2–year [3]. Figure 1 shows the main research related to solving the problem of transformer failure in the zone.

Figure 1. Fault studies of transformers.

The latest research on this topic was carried out by the Research Program on Acquisition and Signal Analysis (PAAS) of the Universidad Nacional de Colombia, Bogotá branch. A transformer prototype with several special features that perform better in areas of high atmospheric electrical activity was designed during this project [7]. This prototype showed increased supportability of the applied electrical stresses. During this project, 14 prototype transformers of special design with SPT mesh were installed in the rural zone of Samaná–Caldas. It was found that to this date 12 transformers have failed with an average lifetime of 2.8 years. These results show that the transformer of special design implemented does not provide an optimal solution; therefore a new study of the topic is required in order to find the main cause of transformer failures and hence propose appropriate solutions.

III. METHODOLOGY DESCRIPTION

The methodological framework below was implemented for the study of the causes of failure in distribution transformers in the eastern part of the electrical utility Central Hidroelectrica de Caldas (CHEC).

A. Processing of information The method used in the processing of information includes the identification and analysis of useful information from the databases of the electrical utility CHEC.

B. Analysis of transformer failures The methodology used to analize transformer failures is developed on the basis of the following three stages and their respective activities. 1) Forensic analysis:

The technical characteristics of electromagnetic compatibility and failure causes in dead transformers are analyzed in the laboratory. The method of fault trees allows the identification and analysis of the conditions and factors that cause or have the potential to cause or contribute to the occurrence of an event or top event [8]. The method also allows qualitative and quantitative analyses of the reliability of the system. The construction of the fault tree starts with the basic events (root cause represented by ovals in the tree) and their probability of occurrence. The probability of the top event is obtained applying the Boolean algebra [8]. 2) Field Technical visit:

After identifying the critical area of transformer failure, a field technical visit is performed to verify the condition of the installations and their compliance with standards [9], [10]. The methodology implemented allows the identification of the causes of failure or problems.

Future occurrence will be prevented once these are corrected.

C. Remedial action plans. The remedial action plan begins with the inquiry into possible solutions that comply with the standards and are likely to be implemented. The validation of the remedial actions is performed through modeling and simulation.

IV. RESULTS

A. Processing of the information From the analysis of data available in different information systems of CHEC, it was determined that: • There are 2.776 transformers installed in the

eastern area CHEC. • There are 969 transformers that were installed

but failed in the past 6 years. • There are 341 nodes in which transformers

failed. • There are 76 transformers that failed and were

diagnosed in the CHEC laboratory. A temporal analysis of failed transformers and the months of increased atmospheric electrical activity was performed to establish the relationship between electrical storms and transformer failures. Figure 2 shows a direct relationship between the amount of failed transformers per month and temporal variation of rainfall in 2012.

Figure 2 - Failed transformers/Rainfall year 2012 The above graph shows that in the months of july and august the failure of transformers did not have the same trend as the precipitation of rain, so it can be inferred that there is a different cause of failure not associated with lightning. There is a relationship between periods of rain and failure of transformers in the remainder months which indicate that lightning might be the cause of transformer failure. However, when analyzing the location of repeat nodes (see Figure 3), hereinafter called critical zone, it was determined that the largest amount of failed transformers is located on the periphery of the central mountain range, Figure 4; which is consistent with the area of higher GFD.

Figure 3 - Repeat nodes in transformer failure

Figure 4 - Topographic map of Caldas´s eastern zone

B. Analysis of transformer failures

Forensic analysis is the first stage of failure analysis and the results are shown below. 1) Forensic analysis:

Forensic analysis was done to 76 failed Transformers in the critical zone. The methodology implemented for the processing of information of the failed transformers diagnosed in the laboratory of CHEC is divided into the following stages: • Functional blocks of the installation:

The constructive parts of the transformer are categorized in functional blocks. The following elements were obtained: Tank, bushings, windings, insulation system (oil and paper), tap charger and core. • Identification of failure modes:

Failure modes were detected by the non-operation of the elements described in the previous paragraph. Additionally, transformers that were uninstalled by mistake were categorized in the failure mode of poor execution. It is to note that these uninstalled transformers were found to be in good condition. This category does not fit in the definition of failure mode

but it was included that it leads to the unavailability of the transformer. • Estimation of the failure rate

The next step was to estimate the failure rate of each failure mode through the failure probability of each basic event (cause of failure diagnosed in the laboratory CHEC) compared to the equivalent universe, using the following equation:

(1) where: Xocu: Number of occurrences of the basic event NRe_i: Equivalent number to the installed transformers The following equation was used to calculate the equivalent number proportional (NRe_i) to the installed transformers:

(2) where: NR: Number of repeat nodes (341). nd_NR: Sample number of failed transformers per year (83). NRTF: Number of failed transformers at repeat nodes per year (162). Figure 5 shows an example of the fault tree built for the core. The failure rate is obtained by means of the application of Boolean algebra, equation (3).

Figure 5 - Core fault tree

( ) ( ) ( )1 2 3 4 5 6 7 8P N N N N N N N N N= + + + + + + + (3)

• Ranking of failure modes according to

magnitude. The ranking is done by calculating the Probabilistic Risk Number (PRN) using the following equation:

(4) where: FC: Frequency is obtained by dividing the failure rate by 100. IP: Impact of the failure mode. The following scale was assigned to assess the impact of each of the failure modes of the transformers: 1: Poor execution for transformers in good condition. 2: Transformers than can be repaired. 3:Transformers that can not be repaired. Table 2 shows the results. The Pareto's law was applied and it was decided that two (winding and core) of the seven failure modes were regarded as serious.

Table 2 - PRN calculation of failure modes Failure modes Impact Frequency PRN

Core failure 3 18,85% 0,5656 Tank failure 3 0,57% 0,0171 Winding failure 2 14,85% 0,2971 Dielectric strenght 2 7,43% 0,1485 Bushing failure 2 0,00% 0,0000 Tap charger failure 2 0,00% 0,0000 Poor execution 1 0,57% 0,0057

• Identification of root causes in serious failure

modes. The root causes of failure in the core and winding failure modes were identified. Table 3 shows the number of basic event occurrences that produce failure mode in the winding and the core.

Table 3 - Occurrences in the serious failure mode Root cause of failure

modes Amount of occurrences Winding Core

Overvoltage 1 10 Surge 23 18 Damaged packages 1 5 Mishandling 0 0 Short circuit 1 0

After performing root cause analysis of the serious failure modes it was determined that the dominant causes of failure were surge with a total of 41 occurrences and overvoltage with a total of 11 occurrences. 2) Field Technical visit

Once the critical zone was identified, a technical visit to the field was performed to verify compliance with the standards of transformer installation. This visit was made to the most critical circuit identified as SNA23L15 "Rancho Largo". There was evidence of poor practices of grounding systems where conductors and connectors in ASCR with copper wires were used

causing impedance which would prevent an optimal drainage of an atmospheric event to the SPT, figure 6a. Additionally, the surge-protective device SPD in some cases is connected to the tank of the transformer (see figure 6b) and the tank of the transformer is connected to the grounding system. Even though there is a discharge path, this is set through the tank and represents high impedance to the path of the lightning.

a) b)

Figure 6 - Grounding system and SPD installation In some of the installed transformers, the SPD is mounted on the crosshead of the circuit breaker as shown in Figure 7a. This generates an increase in the electromagnetically induced voltages on the transformer bushings due to the inductive loop between the SPD and the tank of the transformer. The steepness of lightning current di/dt, which is effective during the interval ∆t, determines the height of the induced voltages, and the inductance is proportional to loop area. Inductive loop formation is represented in Figure 7b [11].

(5) Where:

: Steepness of lightning current.

L: Inductance of the loop.

a) b)

Figure 7 - SPD installation and inductive loop

C. Analysis of root cause failure From the results obtained in the above processes it can be inferred that transformers are failing primarily due to overload and overvoltage. However, possible partial discharges in the transformers insulation might be a possible failure cause taking into account the handling, and transport of these transformers through

unpaved roads in mountainous areas that could form bubbles in the oil and the energizing before the set time to remove these bubbles (6-8 hours) [13]. These partial discharges degrade the insulation in the same manner as overload. The hypothesis of partial discharges as a cause of failure is reinforced by examining several transformers diagnosed in the laboratory CHEC. Figure 8 shows blackened oil and with mud as main features which, by normal diagnostic procedures in the laboratory, indicate overload as the cause of failure. However, the laboratory diagnostic results are contradictory because according to the results of the analysis, the transformer chargeability in the eastern part CHEC is on average 5.2 % of the transformer nominal charge which is very low to cause surge in the transformer. Based on the above information, it is concluded that the 41 transformers diagnosed as failed by overload have actually failed because of partial discharges in the insulation.

Figure 8 – Failed transformer oil

V. REMEDIAL ACTION PLAN

Based on the results of root cause analysis, it was determined that the second cause of failure was overvoltage. Using the above information and based on the high keraunic levels near eastern Caldas, it was determined that poor performance of the SPD was a possible failure cause. A simulation of the transferred impulse using the software ATP/EMTP with the following characteristics was performed to verify the above [14]:

• A standard voltage impulse (1.2x50µs) in primary terminals before the SPD with an amplitude of 150kV in order to overcome the basic BIL (95kV) in transformers of 13200V in overhead distribution networks of CHEC.

• A surge arrester for overhead distribution networks on zinc oxide of the polymer type for 12kV with MCOV of 10.2 kV as CHEC requires in three-phase, 3-wire installations. 13200 V or single phase 2-wire [9].

Figure 9 shows the ATP-EMTP implemented model. Figure 10 shows the impulse applied across the primary, it also shows the results of the simulation

implementing the discharger installed by CHEC. The remnant voltage for the operation of the SPD is 95kV. It should be noted that the good performance of the SPD depends on the value of the grounding system available on the node [16], [17]. Thus, it is necessary to verify that minimum operating requirements of the surge arrester are met taking into account the current installing conditions in the eastern part of Caldas (grounding system of 10Ω [10]).

Figure 9 - SPD model

(f ile DPS_P_IEEE_1ohm.pl4; x-v ar t) v :IMPULS v :PRIM

0,00 0,05 0,10 0,15 0,20 0,25[ms]0

30

60

90

120

150

[kV]

Figure 10 - Simulation of the SPD model

Additionally, a decrease of inductive loops in the transformer should be considered. This is achieved by landing the SPD directly to the down conductors of the grounding system in the shortest distance possible as shown in Figure 11, and using a single conductor from the grounding point of the tank to the down conductor of the grounding system. Additionally, the SPD have to be installed on the transformer tank, figure 11.

Figure 11 - Decrease in inductive loops

Based on the simulation results, it is recommended to reinforce the grounding system using 5 reinforcements or additional weights as shown in Figure 12. This type of mesh can, by itself and without soil treatment requirements, reach the resistance values required by the system of distribution of the eastern zone. Table 4 shows the simulation results of the mesh. Figure 13

show distribution of contact and surface voltage respectively.

Table 4 -Simulation results of the proposed mesh Type of

mesh Resistance in the mesh

Max step voltage

Max contact voltage

Electrode 17.64 Ω 370.75 V 215.72 V

Figure 12 - Configuration of the SPT mesh

Figure 13 - Diagram of surface voltage

The characteristics and dimensions of the proposed mesh are shown below: • Mesh depth of 2.4 m. • Conductor gauge of 2/0 AWG. • Copper electrodes. Copper is resistant to

corrosion and has high conductivity. • Spacing between parallel conductors between 3

m and 15 m.

VI. CONCLUSIONS

• The results of the analysis of main failure causes, performance and the technical check of transformer installation showed that overload and overvoltage were the dominant causes of failure. The proposed solutions presented on this paper were determined based on these results.

• The onsite inspection and the geographic

location of the failed nodes and of the repeat nodes helped to established that the critical zone was located in mountainous areas and in areas far away from paved roads. In addition, practices of transformer energizing without waiting the recommended oil gasket time led to the hypothesis of oil partial discharges in the insulation as the primary cause of failure, and

lightning, considering the high keraunic level of the area, as the second cause.

VII. REFERENCIAS

[1]. H. Torres, “Experience and first Results of Colombian lightning Location Network”, in Proceedings of the 23th International Conference on Lightning Protection Firenze, Italy: 1996.

[2]. H. Torres, D. Rondon, W. Briceño and L. Barreto, “Lightning Peak Current Estimation Analysis from Field Measurements in Tropical Zones”, in Proceedings of the 23th International Conference on Lightning Protection Florence, Italy: 1996.

[3]. www.paas.unal.edu.co [4]. J. A. Agudelo, Fallas de transformadores de

distribución, Universidad Nacional de Colombia, 2000.

[5]. H. Torres, Evaluación y modificación del transformador apropiado y óptimo para zona tropical, Universidad Nacional de Colombia, 2004.

[6]. Acero, L., Muñoz, E., Román, F., Soluciones al daño de transformadores en la zona de Samaná, Caldas., II Congreso Internacional Sobre Uso Eficiente y Racional de la Energía CIUREE, 2006, V.18, fasc. 54, ISSN: 1692-7052, 2006, pp.36 41.

[7]. H. Torres, Contribución a la solución del problema de falla de transformadores de distribución en Colombia, Universidad Nacional de Colombia – Documento interno Programa de Investigación sobre Adquisición y Análisis de Señales - PAAS- UN, 2008.

[8]. NTP 333: Análisis probabilístico de riesgos: Metodología del “Árbol de fallos y errores”. Tomás Pique Ardanuy. Antonio Cejalvo Lapeña.

[9]. www.chec.com.co [10]. Reglamento Técnico de Instalaciones Eléctricas –

RETIE 2008. [11]. F. R. Campos, Análisis de las fallas en

transformadores causadas por la operación del pararrayos ante sobretensiones externas, Portal de Revistas de la Universidad Nacional, Bogotá-Colombia, 1991.

[12]. Zapata J. Zapata, Estimación de tasas de fallas de componentes en casos de ausencia de datos o cantidades limitadas de datos, Scientia et Technica.

[13]. http://www.partial-dischargeacademy.com/ partial-discharge-causes.

[14]. IEEE Working Group. Modeling of metal oxide surge arresters. USA, Transactions on Power Delivery, Vol. 7 No.1, January 1992.

[15]. NTC 4552-1 Protección contra descargas eléctricas atmosféricas (Rayos). Parte 1: Principios generales. 2008-11- 26.

[16]. Cano Plata, Eduardo y Ramírez C, Samuel. “Sistemas depuesta a tierra: diseñando con IEEE-80 y evaluando con MEF”, Universidad Nacional de Colombia, Departamento de Ingeniería Eléctrica, Electrónica y Computación.

[17]. ANSI / IEEE Std 1410 - 2010 “Guide for Improving the Lightning Performance of Electric Power Overhead Distribution Lines” (IEEE Std 1410 – 2004).