THE IMPACT OF NETWORK SPLITTING ON FAULT LEVELS AND OTHER PERFORMANCE MEASURES

THE IMPACT OF NETWORK SPLITTING ON FAULT LEVELS AND OTHER PERFORMANCE MEASURES C.E.T. Foote*, G.W. Ault*, J.R. McDonald*, A.J. Beddoes *University of Strathclyde, UK EA Technology Limited, UK c.foote@eee.strath.ac.uk SUMMARY The fault level rating of installed equipment in urban areas is a common restriction on the expansion of distributed generation (DG). One option for reducing fault level is to split the network at selected points by, for example, opening bus section circuit breakers that are normally run closed. This paper presents studies performed to evaluate the impact of network splitting on fault levels and various network performance measures such as reliability and power quality. Four generic test networks were used that reflect typical operating conditions and configurations of UK distribution networks, although the results are generally applicable to other networks. The studies provide valuable guidance on the potential benefits and costs of network splitting and so contribute to the debate on how to accommodate more DG on existing networks. Other issues like network operation and maintenance must also be considered when assessing if a particular network splitting option is acceptable from both a network and generation perspective. INTRODUCTION Where fault levels are already high, as is often the case in urban areas, any additional contribution from DG may necessitate the potentially expensive replacement of switchgear. The reconfiguration of distribution networks by splitting them at strategic points and thereby increasing the network impedance and reducing fault level has been identified as one way in which more DG might be accommodated without incurring the expense of upgrading network equipment [1,2]. To help assess the advantages and disadvantages of this approach, a series of studies was performed. Network performance in terms of fault level, reliability and power quality was analysed using a set of generic network models. The objective of the exercise was to identify how fault level can be reduced in power networks without a serious impact on other network performance indicators such as reliability and power quality. This paper provides a review of the methods used and results obtained from studies of network performance under alternative network splitting options. These studies contributed to a broader assessment of network splitting performed in the UK for the Distributed Generation Co-ordinating Group an industry working group that is examining the barriers to the expansion of DG and proposing solutions [3]. Clearly other issues like network operation and maintenance must be considered alongside these performance measures when assessing if a particular network splitting option is acceptable from both a network and generation perspective. TEST NETWORKS AND INPUT DATA Studies were performed on four test networks, which were stylised versions of four commonly used topologies in UK distribution networks. The transformer feeder system (see Figure 1) was a straightforward radial network with parallel supplies to the busbar at each voltage level. The 33kV ring system had two parallel lines at 132kV supplying a 33kV busbar from which a ring configuration at 33kV supplies three more 33kV busbars, which each supply 11kV busbars. The 132kV ring system had a four busbar ring at 132kV that supplies three 33kV busbars that in turn supply 11kV loads. The interconnected system had a double busbar arrangement at 132kV supplying three 33kV busbars, which were interconnected in a triangular arrangement through parallel paths that link together multiple 33kV busbars. Groups of three interconnected 33/11kV substations were formed with triangular arrangement at 11kV and loads were supplied from each of the 11kV busbars. Figure 1 shows the transformer feeder system. Each busbar is represented in the diagram by two buses with a connecting line representing the bus section breaker (represented by a low impedance line). The splitting options for the transformer feeder system are indicated on Figure 1 with numbers in circles next to the appropriate line/breaker or transformer. The odd-numbered splitting options take one of two parallel transformers out of service. The even-numbered splitting options split the busbar on the secondary side of transformers. The other test networks included splitting options that opened the rings or interconnections that provided alternative, parallel paths through the network. The data associated with the networks was agreed on with industry partners as being typical of UK distribution networks but of course does not cover the wide range of different circumstances that are to be found in reality. The studies performed with the test networks give an indication of the relative effect of different network splitting options. Results

from this analysis are not directly applicable to real distribution networks, which exhibit their own unique characteristics. Implementation of network splitting on a real network would require detailed modelling and analysis of the real network. generation without increasing existing fault level values and threatening existing switchgear. The network models and input data were tuned to produce fault levels similar to those typically seen on UK networks. However, the models only provide an indication of the approximate effect of network splitting on fault levels. The headroom gained by different splitting options gives an indication of their relative impact but these results are not directly transferable to any real networks, which are unique and require specific models and analysis. Example Results for Fault Level Studies The results for three phase faults in the transformer feeder system in TABLE 1 show that splitting options must be implemented at or above the voltage level at which fault level headroom gain is required. Removing a transformer from service (the odd-numbered options) lowers fault levels everywhere below that point. Splitting a busbar (the evennumbered options) has a much more localised impact. Figure 1 Transformer feeder test network Full details on the test networks and input data are available in the final project report [3]. Some features of the input data are as follows: It was assumed that each complete network supplied 150,000 customers and had a total load of 240MW and 75MVAr. Each network was supplied entirely from a single generator at 400kV that represented the grid supply. All lines were assumed to be underground cables because it is in urban areas where fault levels are most likely to pose a problem. Reliability data was collected from the database held by the UK electricity industry. FAULT LEVEL STUDIES Fault levels for three phase and single phase to ground faults were calculated based on the RMS value of initial symmetrical short-circuit current. In each network, for each splitting option, the fault level headroom gain was calculated. This is the reduction in fault level at the bus and thus the additional fault level that could be added by distributed TABLE 1 Fault MVA Base and Headroom gain for 3-phase faults in transformer feeder system. Bus Base Opt1 Opt2 Opt3 Opt4 Opt5 Opt6 400KV 33401 0 0 0 0 0 0 132KV_1 2255 1087 935 0 0 0 0 132KV_2 2255 1087 935 0 0 0 0 132KV_3 1955 879 718 0 0 0 0 132KV_4 1955 879 718 0 0 0 0 33KV_1 585 120 0 235 220 0 0 33KV_2 585 120 0 235 220 0 0 33KV_3 343 48 0 97 71 8 6 33KV_4 343 48 0 97 71 4 6 11KV_1 139 9 0 19 0 58 60 11KV_2 139 9 0 19 0 58 60 The results for the other systems showed that removing a transformer has a greater and more extensive impact than splitting a busbar. Generally, at lower voltages the impact of splitting is more localised. Splitting rings or interconnections has a significant effect on the buses that are left electrically distant from the grid. RELIABILITY STUDIES The impact of network splitting options on the reliability of the four test networks was evaluated. Reliability models were used to determine the expected number of supply failures at 11kV per year. For each of the splitting options, the additional failures per year compared to the base case were calculated. To properly reflect the number of customers that would be affected, failure rates were scaled depending on the voltage level at which a splitting option was implemented. The scaled failure rates were then multiplied by 100 to produce a measure of additional customer interruptions per 100 customers for each splitting option.

It should be noted that modelling of this sort simply provides the mathematical expectation of failures specifically for the stylised models developed. Caution must be taken in applying the results of these studies beyond their explicit purpose, which is a comparative analysis of the reliability implications of network splitting. It is likely that splitting options will be implemented along with auto-close schemes that will restore supplies very quickly in the event of a failure of the single remaining supply circuit after a network splitting option is implemented. Thus, the additional failures due to the splitting options are likely to be Short Duration Interruptions (SDIs). However, if switching times are greater than 3 minutes then in the UK regulatory system the additional failures will count as Customer Interruptions (CIs) and contribute to Customer Minutes Lost (CMLs). The time customers are without supply will depend on the time required to switch the alternative supply back in. Interpretation of the results as either Customer Interruptions or Short Duration Interruptions can be undertaken through a simple allocation of the additional interruptions caused by the splitting option as SDIs if an auto-changeover scheme is assumed. Alternatively if supply changeover cannot be achieved in a 3-minute interval then the additional interruptions caused by the splitting option are assumed to be additional CIs. The reliability results assumed a 5-minute switching time to restore supplies after each failure for the calculation of additional customer minutes lost for each network splitting option. The average value of Customer Interruptions for DNOs in 2001/2 was 87.40 and the average value of Customer Minutes Lost was 83.70. The average number of Short Duration Interruptions was 75. By making assumptions about the implementation of network splitting options, the impact on the reliability indices of a typical DNO was estimated. Example Results for Reliability Calculations Network splitting never has a positive impact on reliability. The results for the 33kV ring system in TABLE 2 suggest that removing one of the 132/33kV transformers will have the greatest impact on reliability. The effect of Options 5 and 6 is much less than on the transformer feeder system because the 33kV ring links the buses on the high-voltage side of the 33/11kV transformers, thereby retaining two parallel 33kV cables to this point. In the transformer feeder system, the buses on the HV side of the 33/11kV transformers are not linked and Options 5 and 6 mean each 11kV bus is supplied through a series combination of a single cable and single transformer. The linking of buses on the HV side of transformers while taking one of those transformers out of service could be worthy of further investigation. POWER QUALITY STUDIES Power quality is increasingly important to electricity consumers and network operators. Distortions to the electricity supply waveform can produce nuisance effects such as flickering lights, result in higher losses and inefficiencies in the supply system, and cause malfunction of sensitive equipment, which can impose high costs if industrial or commercial systems are interrupted as a result. Three aspects of power quality were investigated in this set of studies: the effects of harmonic interference injected at 11kV; the effects of voltage flicker introduced at 11kV; and the effect of voltage dips on the grid supply. Harmonics A harmonic current source was connected to an 11kV bus and the effect on voltage measured on a set of selected buses. The harmonic source was tuned to produce a total harmonic distortion (THD) of around 4% at the bus where it was connected, which is the planning level specified for 11kV systems in UK Engineering Recommendation G5/4 [4]. Values of THD lower than the Base indicated improvements in power quality. Higher values of THD indicated deterioration in power quality. The results gave an indication of the influence of network splitting on harmonics. In reality, harmonic problems can be exacerbated with resonance in the network where the combination of cable capacitance and transformer inductance can provide a low-attenuation path for harmonics of particular frequencies. This will depend entirely on the network in question and no such effects were captured in these studies. Example Results of Harmonics Studies. The harmonics results for the 132kV ring system in TABLE 3 show a similar pattern of results as in the other systems. Options 3, 5 and 6 have the greatest effect on THD at 11kV. TABLE 2 Effect on a typical DNO of splitting options in the 33kV ring system Opt 1 Opt 2 Opt 3 Opt 4 Opt 5 Opt 6 Opt 7 Customer Interruptions / Year / 100 101.50 87.40 146.68 87.43 87.86 87.86 87.40 customers Short Duration Interruptions (assuming sub 89.10 75.00 134.28 75.03 75.46 75.46 75.00 3 minute switching time) Customer Minutes Lost / Year (assuming 5 minutes switching time) 84.41 83.70 86.66 83.70 83.72 83.72 83.70

TABLE 3 Total Harmonic Distortion for base case and network splitting options in 132kV ring system Bus Base Opt1 Opt2 Opt3 Opt4 Opt5 Opt6 Opt7 400kV 0.002 0.002 0.002 0.002 0.002 0.002 0.002 0.002 132kV_1 0.015 0.031 0.016 0.016 0.015 0.014 0.015 0.015 132kV_2 0.015 0.031 0.014 0.016 0.015 0.014 0.015 0.015 132kV_3 0.177 0.356 0.182 0.185 0.177 0.163 0.177 0.169 132kV_4 0.177 0.356 0.182 0.185 0.177 0.163 0.177 0.188 33kV_1 0.165 0.222 0.166 0.295 0.165 0.155 0.165 0.164 33kV_2 0.165 0.222 0.166 0.295 0.165 0.155 0.165 0.164 11kV_1 4.318 4.814 4.326 5.372 4.318 7.123 1.070 4.281 11kV_2 4.318 4.814 4.326 5.372 4.318 7.123 7.567 4.281 Voltage Flicker Flicker was investigated using simple dynamic simulation. UK Engineering Recommendation P28 [5] describes how an assessment of flicker severity can be performed by simulation and calculation. The associated British Standard on flicker (61000-3-3) [6] describes an analytical method for evaluating short-term flicker (P st ) that can be applied to regularly shaped voltage changes. This method is appropriate to the analysis performed in these studies and is outlined in Equation 1. 3.2 ( 2.3*( Fi * di ) ) 3.2 i st = (1) Tp P F i is the shape factor associated with voltage change i as described in the standards. Only step changes in voltage were considered so the shape factor was equal to 1.0 for all changes. d i is the size of voltage change i, expressed as an absolute percentage of the nominal voltage. T p is the total observation period, which should be 600 seconds for evaluating P st. A source of flicker was simulated by connecting and disconnecting shunt admittances at the 11kV_2 bus in each network model to shift the voltage up and down. This was tuned to give a short-term flicker value of around 0.7-0.8 for the base case network configuration. The analytical method is not recommended if the time duration between voltage changes is less than one second, so changes were implemented every two seconds, giving 300 voltage changes in the 600 second observation time. Measurements were taken at selected buses and the short-term flicker value calculated for the base case and the different network splitting options for each network topology. Larger values of flicker represent deterioration in power quality. Engineering recommendation P28 sets an absolute maximum of P st = 1.0 at 132kV and below. in flicker at and below the points where transformers were taken out of service. TABLE 4 Short-term flicker (P st) for base case and network splitting options for interconnected system. Bus Base Opt1 Opt2 Opt3 Opt4 Opt5 400kV 0.003 0.003 0.003 0.003 0.003 0.003 132kV_1 0.043 0.095 0.056 0.046 0.045 0.044 132kV_2 0.043 0.095 0.029 0.046 0.045 0.044 33kV_1 0.176 0.253 0.182 0.289 0.184 0.164 33kV_2 0.148 0.221 0.150 0.227 0.203 0.150 33kV_3 0.238 0.322 0.244 0.341 0.260 0.286 33kV_4 0.221 0.303 0.227 0.330 0.236 0.194 11kV_1 0.689 0.823 0.695 0.832 0.728 0.715 11kV_2 0.827 0.969 0.833 0.974 0.869 0.846 Voltage Dips Voltage dip studies were performed using the same steadystate, load flow models as used for fault level analysis. A 10% voltage dip on the 400kV system was simulated by pulling the voltage at the 400kV bus down to 0.9pu. The load flow was re-solved but transformer tap settings were locked. This simulates the immediate response to a voltage dip on the 400kV network, before automatic transformer tap changers have time to react and correct the voltage. Lower voltage levels indicate a poorer response to the voltage dip in the grid supply. Example Results of Voltage Dip Studies. The voltage dip results showed that voltage is pulled down throughout the system when there is a voltage dip on the 400kV grid supply. The results for the 33kV ring system in TABLE 5 show that the increased impedance of the network in Options 1, 3, 5 and 7 exacerbates voltage dips. The buses affected most strongly are those below the splitting point. The symmetrical nature of the model means that the even-numbered splitting options have no impact on the voltage levels when the dip is imposed at 400kV. The tap position of the 33/11kV transformers in Option 3 is such that the 11kV buses have a higher per unit voltage than the 33kV buses. TABLE 5 Voltage dips for base case and network splitting options on 33kV ring system. Bus Base Opt1 Opt2 Opt3 Opt4 Opt5 Opt6 Opt7 400kV 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 132kV_1 0.8545 0.768 0.8545 0.8435 0.8545 0.8524 0.8545 0.8521 132kV_2 0.8545 0.768 0.8545 0.8435 0.8545 0.8524 0.8545 0.8521 33kV_1 0.8171 0.7658 0.8171 0.7749 0.8171 0.8092 0.8171 0.8084 33kV_2 0.8171 0.7658 0.8171 0.775 0.8171 0.8092 0.8171 0.8084 33kV_3 0.7751 0.7202 0.7751 0.7302 0.7751 0.7638 0.7751 0.7855 33kV_4 0.7751 0.7202 0.7751 0.7302 0.7751 0.7638 0.7751 0.6977 11kV_1 0.7983 0.7784 0.7983 0.7906 0.7983 0.755 0.7983 0.7734 11kV_2 0.7983 0.7784 0.7983 0.7906 0.7983 0.755 0.7983 0.7734 Example Results of Flicker Studies. The flicker results for the interconnected system in TABLE 4 show little change across the network splitting options. Options 1 and 3, where transformers are taken out of service, have the greatest impact. In the other systems, the results showed deterioration CONCLUSIONS The results of the studies showed that while network splitting may deliver some benefits in fault level headroom gain, it also

has associated costs in terms of poorer reliability and power quality. The trade-off between the costs and the benefits will depend on what value is attached to each of the performance criteria and the voltage level at which they are measured. This may vary from DNO to DNO and may change as a result of changes in regulatory policy or practice. An assessment of the results suggests that the greatest fault level headroom gains are to be had with the splitting options that remove transformers from service. As would be expected, this also has a greater negative impact on reliability and power quality quite significantly in some instances. In contrast, the bus section splitting options seem to have a relatively benign effect on reliability and power quality but can still deliver some fault level headroom gain. The 33kV ring system demonstrated the potential value of closing the bus section breakers on the HV side of transformers. With these breakers open, the switching out of one transformer leaves load being supplied through a single, series combination of a cable and a transformer, exposing it to the failure of either component. Closing the bus section breaker leaves two parallel cables to supply the one remaining transformer, reducing the risk of loss of supply to customer. The studies provide valuable guidance on the potential benefits and costs of network splitting and so contribute to the debate on how to accommodate more DG on existing networks. However, network splitting also influences other issues, such as operational, maintenance and control factors. The quantitative assessment of network splitting presented here is only part of the overall assessment that is required. Further issues that could also be explored with modelling and analysis include: Uneven sharing of load between transformers as a result of splitting bus sections with consequent impacts on transformer operating performance and condition Protection co-ordination, which could be affected by changes in network topologies and impedances Transient stability of distributed generation, which will be affected by the different network impedance Clearly other issues such like network operation and maintenance must be considered alongside simple performance measures when assessing if a particular network splitting option is acceptable from both a network and generation perspective. These issues are fully discussed, and full results of the quantitative analysis are presented, in the final report produced from this project [3]. ACKNOWLEDGEMENTS This work was supported by the UK Department of Trade and Industry through their Technology Programme in New and Renewable Energy. Industry partners, through the Technical Steering Group of the Distributed Generation Coordinating Group, provided a great deal of assistance and data. REFERENCES [1] Collinson, F. Dai, A. Beddoes, J. Crabtree, 2003, Solutions for the connection and operation of Distributed Generation, DTI New and Renewable Energy Programme, Reference K/EL/00303/00/01/REP [2] X. Wu, J. Mutale, N. Jenkins, G. Strbac, 2003, An Investigation of Network Splitting for Fault Level Reduction, Tyndall Centre for Climate Change Research Working Paper 25 [3] DTI Distributed Generation Programme (Contractor: EA Technology); The Performance of Networks Using Alternative Network Splitting Configurations ; Contract Number: DG/CG/00031/00/00, URN Number: 04/1475; 2004 [4] Engineering Recommendation G5/4, 2001, Planning levels for harmonic voltage distortion and the connection of non-linear equipment to transmission systems and distribution networks in the United Kingdom, Energy Networks Association [5] Engineering Recommendation P28, 1989, Planning limits for voltage fluctuations caused by industrial, commercial and domestic equipment in the United Kingdom, Energy Networks Association [6] British Standard BS EN 61000-3-3, 1995, Electromagnetic compatibility (EMC) Part 3: Limits, British Standards Institution