The use of surge arresters in parallel for the lightning protection of pole mounted distribution transformers in Eskom.

Abstract

Eskom (Electricity Supply Commission of South Africa) is the national electrical utility that provides the generation, transmission and distribution of electricity in South Africa. The majority of Eskom's electricity distribution is done with either 11 kV or 22kV electrical overhead networks. An unacceptable number of Eskom's pole mounted power transformers on these networks have been failing over the past six years in the Kwa-Zulu Natal region. The average transformer failure rate for the Distribution Eastern Region was calculated to be 2.4% per annum. International norms seem to indicate a transformer failure rate of between 0.5% and 1.0% per annum as acceptable. The estimated cost of these transformer failures was between R9 million and R13 million per annum for the Eastern Region. Eskom Distribution has seven regions and the total cost of these failures was considerable to the business. These transformer failures contributed an average of 5.3% per month of the Supply Loss Index (SLI) for the Eastern Region, with a maximum contribution of 14.5% per month of the SLI for the region. The SLI is an Eskom performance measure of the unavailability of supply of the networks. The Eskom plant performance database (NAPI) was statistically analysed in detail and a number of field investigations conducted at transformer installations that had failed in the past. Transformer earth electrode resistance measurements were taken in an attempt to identify the cause of these transformer failures. Local transformer manufacturers were consulted and a national transformer refurbishment company's database was analysed during the investigation as part of a holistic approach to the industry related dissertation. The main finding of the NAPI data analysis was that the majority of the transformers failed during lightning storm periods. Another similar Eskom investigation had a sample of failed transformers opened for internal inspection. Signs of lightning damage to either the primary side winding or the primary lead were found. The proposed failure hypothesis was that the transformers required additional lightning protection of the primary side to protect the transformer against lightning. It seemed that the current specification of the Eskom distribution class surge arresters was inadequate to offer sufficient lightning protection of the pole mounted transformers. Practical measures were implemented on existing Eskom 11 kV networks in the Glencoe area as part of an Eskom research project to reduce the high failure rates of transformers. Two experimental networks were established and one control network was used as a reference line. The project implementation was completed at the end of October 1999. The experimental project looked at applying additional primary side lightning protection of the transformers. The main emphasis of the lightning protection on the first network was the use of two distribution class arresters in parallel (double arrester configuration) for each transformer and an understrung conductor connected to the prior structure back from the transformer installation. The second network had the standard single arresters installed on the transformers. A 600mm wood path was placed in series with an earthed down conductor installed on each intermediate woodpole structure to ensure a basic insulation level of 300kV for the network. The control network also had single arresters installed and was a fully insulated network (no earthed down conductor on the woodpoles). Each network was carefully monitored in terms of equipment failures and the performance levels of each network was measured with installed voltage dip recorders near the individual network circuit breakers. The Eskom Lightning Position and Tracking System (LPATS) data was collected and analysed to quantify the lightning activity before the project implementation to that of after project implementation. For the period November 1999 to March 2002 not a single transformer or surge arrester had failed ,due to lightning on the double arrester and understrung conductor configuration experimental network. There were recorded transformer and surge arrester failures on the second experimental network and on the control network. It was found that the practical methods implemented on the networks did not have a negative impact on the performance levels of the networks. Laboratory work was conducted in the high voltage laboratory at the University of Natal, on various metal oxide varistor (MOV) blocks of opened up new and failed surge arresters. This was to determine the effect of MOV blocks in parallel under power frequency and current impulse conditions. In particular, to determine what the effect of parallel MOV blocks with different voltage-current (V-I) characteristics would have on the current sharing and energy absorption capabilities of the individual blocks. The work was performed to simulate the behaviour of two surge arresters in parallel as in the experimental project. The experimental project lines were modeled using the Alternative Transients Program (ATP) simulation package and various parametric studies performed in the single phase conductor simulations. Each network component (such as the line, transformer and surge arrester) and phenomena (such as the effect of corona, the transient earth electrode resistance and voltage flashover) were modeled. The effect of surge arresters in parallel and the use of an understrung conductor arrangement were quantified. A current sharing factor (k-factor) was introduced to quantify the sharing of currents through surge arresters in parallel with different V-I curves. The main finding from the impulse laboratory work was that current sharing between parallel MOV blocks became better at higher currents. This finding was supported by other research work findings, particularly in the field of nuclear fusion research with parallel arresters. The results of the ATP simulations showed that the experimental network with the parallel arresters and understrung conductor arrangement considerably reduced the energy absorbed by the individual arresters. The effect of the double arrester configuration was to reduce the energy absorbed by the individual arresters even with arresters with different V-I characteristics and different manufacturers. The understrung conductor arrangement was found to be the major contributor towards the reduction of the energy absorbed by the arresters. The equivalent circuit of a MOV block for transient studies was proposed and then simulated in ATP. The simulated results were compared to the measured waveforms obtained from the impulse laboratory work. A good agreement between the simulated and measured waveforms was obtained. For existing Eskom networks with high arrester and transformer failure rates, the double arrester (distribution class) configuration would be the most time and cost effective solution. The alternative of using a single station class arrester is not proposed due to the costs involved and the availability of stock. The understrung conductor arrangement did significantly reduce the energy absorbed by the arresters but due to high labour costs and time requirements this would not be recommended for existing networks. It is suggested that Eskom investigate this practical method for new lines to be built in high lightning areas. Even with arresters from different manufacturers, the use of the double arrester configuration would decrease the energy absorbed and hence reduce the risk of failure of the individual arresters protecting the transformer. This means that Eskom field staff can use different manufacturer arresters in parallel. This would be especially for times when replacing failed arresters or a faulty transformer under breakdown conditions and electrical supply has to be restored to the customer as soon as possible. It was proposed to perform arrester matching by ensuring that the both arresters were from the same manufacturer.