How does an industry regulator measure and regulate the performance of a distribution company? For U.K. regulator Ofgem, the answer is to monitor the quality of supply and quality of service, and then adjust the level of regulated income accordingly.
Following the 1999 Price Review, failure to achieve performance levels resulted in reductions of up to 0.5% in allowed regulated income for the following five-year period, with increases of a similar magnitude for companies achieving or exceeding targets. From April 2002, Ofgem is able to penalize companies by up to 2% — some £50 million (US$78.9 million) per year for the industry — for failing to meet the required levels of performance. Ofgem could raise this figure further starting in April 2005.
In addition, companies already compensate customers for interruptions of more than 18 hours duration, and a new guaranteed standard requires companies to compensate customers experiencing more than four interruptions of more than 3 hours in a year.
Over the decade, there has been an overall improvement of 1% in SAIFI per year with a larger improvement in SAIDI, some 3% per year. This results in a SAIFI of 0.82 and a SAIDI of 77 min per customer per year averaged across the United Kingdom. The target improvements for the current five-year period vary between companies but average a further 1% to 2% per year for both SAIFI and SAIDI.
While these may seem small, considerable improvements have been made, particularly in terms of improving the worst medium-voltage (MV) circuits in circumstances where aging and wear cause inevitable deterioration in the underlying performance of these existing assets. One company said the underlying fault rate on its underground MV cables has doubled over the last 15 years.
How can a distribution network be designed and operated at an affordable cost to meet the set targets and minimize the disruption to the supply to the customer? Because faults on the MV network cause some 60% of supply interruptions in the United Kingdom, with a further 10% planned interruptions, the emphasis is on improving the MV network.
The Tip of the Iceberg
The first rule for meeting the targets is: When something breaks down, fix it quickly. The larger improvement in SAIDI is the result of investment in customer reporting systems to identify the location of faults, faster response times, the implementation of remote control to promote faster switching times and establishing alternative supplies including mobile generators.
The second rule is: When breakdowns occur repeatedly, remove the cause or replace the faulty item. Thus, policies on tree trimming, surge protection, third-party training, higher maintenance levels and line-reconstruction practices remain important. Each company publishes a list of worst-performing circuits — so-called rogue circuits — and reports on progress annually. United Utilities selects these on the basis of excessive number of interruptions over the previous five years or within the last year, long restoration times or a large number of complaints. The 20 rogue circuits United Utilities selected for improvement in 2000/2001 accounted for about 4.5% of the total customer interruptions (CI) caused by MV faults. The target is to reduce this contribution by at least 20%.
These are the “visible” bits of the iceberg. However, meeting the targets demands an underlying improvement in those sections of the network not so evidently performing poorly, the “invisible part of the iceberg.” United Utilities' MV network consists of 3000 separate circuits fed from about 500 primary substations (33/11 kV, 33/6.6 kV, 132/11 kV). About 900 of the circuits experienced no faults during the last five years. Circuits with fewer faults than the rogue circuits but with more connected customers receive greater attention. Thus, the top 20 circuits account for 13% of the MV CI and 200 circuits account for the top 50% of MV CI.
When looking beyond the rogue circuits, companies consider many factors. The number of faults recorded on a circuit and the resulting CIs are not exact predictions of what will occur in the future. If 25 faults occur at random over five years, the standard deviation “s” is 5. In general, s is somewhat larger because faults tend to occur together in time and location due to incidence of severe weather and repetitive failures. This leads to an accuracy of prediction that is, at best, only about 25%. It is unlikely that the decision of whether to upgrade a circuit with such a high fault rate will depend greatly upon the 25% uncertainty. The CI varies more than the fault rate, and the location of the fault in relation to the position of the protection devices also adds to the variation.
Since the value of SAIFI for United Utilities in 2000/2001 was 0.54, let's consider the benefits of improving circuits with considerably less than 25 faults over five years — say only nine interruptions over five years (s = 3). Here, the percentage variation in the predicted number of faults is more substantial. There is a 20% probability that an average of one fault or less per year will occur and a similar probability that the fault rate will be four or more per year. This means that it is not only necessary to review circuits, which are historically high up the league in terms of faults or CIs, but also those at risk as the fault rate fluctuates. This involves analyzing the theoretical performance and potential for improvement of all circuits using fault rates typical for that type of circuit and locality. If either the theoretical or historical data suggest a need for improvement, then the first task is to resolve the cause of any major differences by examining the fault history.
The next step is to assess a range of possible improvements. In what circumstances should the different techniques be used? How should each technique be implemented? The asset management team has the task of selecting the location of the improvements and determining whether, for example, to refurbish circuits or to add devices improving customer performance. These decisions are made balancing the need to improve individual customer and network performance in light of a five-year capital budget.
The Ice Pick
To discover what lies below the tip of the iceberg, a tool is needed to dig below the surface. The GROND Network Reliability Program targets the task of estimating the reliability of MV networks. The program has its own database copied from the master company databases and updated as needed. This provides a modularity and independence from the other IT systems and enables other companies to readily implement the program.
The model uses average fault rates per kilometer of line for both persistent and transient faults. A user can set these rates for overhead line and underground cable according to the type of circuit. The user can check the validity of the model by comparing it with the recorded fault data. In one area, the location of faults on a group of MV circuits for a one-year period suggested a geographical bias to the distribution of faults not automatically represented by the model. However, examination of the fault distribution over a five-year period showed no geographical bias. Studies on this group of circuits, therefore, were carried out using the average rates over five years for underground and overhead lines. For this case, there was satisfactory agreement between the CI recorded and the CI predicted.
In other cases, groups of faults might indicate a fault hot spot. Links to the faults database are used to display details of each fault on the screen. If, for example, the higher fault rate is due to overhead lines passing through wooded areas, this local anomaly is incorporated within the study by assigning a higher fault rate to these conductors. Alternatively, for more specific or repetitive faults, it may be more appropriate to tackle the cause of the faults than to carry out further modeling.
Users can evaluate the effect of a range of possible improvements. Capabilities exist to modify the network diagram by changing or adding devices and customers, to move Normally Open Points, add new lengths of line, change the type of line, add new Supply Points or change the response time of devices.
Results are in terms of SAIFI, CI, SAIDI and CML caused by persistent and transient faults. The results further separate for sustained outages and short outages ( 3 min) and automated switching(< 3 min). The number of interruptions are viewable on the network diagram and tabulated in a security report. To aid understanding of circuit reliability, the protection zones and the isolation sections can be highlighted on the screen, along with the method of restoration for a fault in any selected section. An additional output is a statistical estimate of the number of customers experiencing more than a set number of interruptions per annum.
A company needs to assess the potential and cost of improving the network. This ensures that targets set by the regulator after discussion with the company are realistic and consistent with the allowed capital expenditure. The regulator on behalf of the customers would like to set steadily improving targets while only approving minimal capital investment. However, as circuits improve by adding additional protection or speeding up response and repair times, further improvement becomes more costly. Furthermore, the rapid introduction of large amounts of new technology with an uncertain lifetime leads to increasing maintenance costs.
The figure on the right illustrates how the effect of adding additional protection to circuits follows a law of diminishing returns. The illustration depicts the potential improvement over the whole of the company MV network made by investing in additional protection. The value of SAIFI for each MV circuit has been estimated, and the improvement achieved by adding a single circuit breaker or recloser to each circuit in the optimal position. The results are ranked in order of decreasing reduction in the total CI. The figure (Note: log scale) shows the results for all circuits (~500) showing a reduction in CI > = 100.
The same tool used for the overall policy studies is used by the asset management team in assessing particular schemes and by the design engineer for detailed studies. In the past, detailed schemes were designed by engineers based in the locality (United Utilities formerly had six regions, each with its own design team) with knowledge of the network in their own region. The centralization of the design function means that many engineers inevitably lack local knowledge of the network. Therefore, the program has been designed to aid the engineer in understanding how a circuit and its neighboring circuits perform.
The figure above shows the placement of additional protection on a circuit consisting of about 17 km (10.5 miles) of overhead line and 10 km (6.2 miles) of underground cable with an existing recloser and two sets of fuses. The facility to identify the optimum location of protection devices was used, which provided an optimum position for further protection at the first underground cable/overhead line section. This gave a predicted reduction in CI from 1693 to 1032 and CHL from 3476 to 2265. The optimum position for a second device is for a circuit breaker at the first main fork. This would produce further estimated savings of 297 CI and 548 CHL. Only the first device was added to the circuit.
As industry regulators demand higher standards of network performance on behalf of the customers, it is necessary for distribution companies to assess the potential for improvement and associated costs for parts of the network not currently labeled “badly performing.” This requires tools to analyze the reliability of the network and the effect of deteriorating fault rates along with policies for improvement. United Utilities used the GROND program to carry out such studies both to investigate long-term policy issues and for detailed MV design studies.
Peter Leather is electricity planning policy manager at United Utilities, which distributes electricity to 2.2 million customers in the northwest region of England. He is a graduate electrical engineer from Cambridge University and a member of the IEE.
Paul Clarkson is a professional electrical engineer and a graduate of Liverpool University. He was formerly a distribution engineer with Norweb. He now is an independent consultant specializing in database and software systems. He is the author of the initial GROND program.
Robin Hodgkins graduated with a degree in mathematics from Cambridge University, England. After working with the UKAEA, Brown University, and English Electric, he was appointed head of the Theoretical Section at the Electricity Council Research Centre, now EA Technology. Later, he was principal mathematician and project manager of a computer simulation of distribution networks. He is now an independent consultant and co-developer of the GROND Network Reliability Program.