C I R E D 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0113

CIRED2009 Session 5 Paper No 0113

RELIABILITY INDICES: THE DETERMINATION OF REFERENCE VALUES FOR SWISS ELECTRIC DISTRIBUTION NETWORKS

Hans-Heinrich SCHIESSER Jürg BADER LUKAS KÜNG ewz – Switzerland ewz – Switzerland ewz - Switzerland hans.schiesser@ewz.ch juerg.bader@ewz.ch lukas.kueng@ewz.ch

ABSTRACT At the beginning of the year 2008 Switzerland introduced a new law concerning the supply of electricity. This new law governs the opening of the electricity market as in the European Union. A regulator (here the Electricity Commis-sion "ElCom") controls the prices and the continuity of supply of electricity of each supplier and sets the perform-ance level. The reliability of a network will be determined by using internationally employed continuity indices like SAIFI and SAIDI. Reasonable values for such indicators have to be determined first. Therefore, the regulator de-pends on the experience of Swiss power companies, which have already performed such analyses. The paper high-lights the momentary process which leads to appropriate guidelines.

INTRODUCTION ewz (Elektrizitätswerk der Stadt Zürich), a Swiss utility, which supplies electricity mainly in the city of Zürich (ur-ban network), has a continuous 25 years record (1983-2007) of outage data. Customer-related continuity indices such as SAIFI, SAIDI, not delivered energy etc. are published in the company's annual report. In the last few years ewz com-pared its continuity values with four other urban utilities in Switzerland [1]. The benchmark is not only useful to get a feeling about the outage performance of the company but also for the determination of reference values. In the sense of the subsidiary principle the "Association of Swiss Elec-trical Enterprises" (Verband Schweizer Elektriz-itätsunternehmen, VSE) published a set of reference conti-nuity values in the so called "Distribution Code", stratified into four groups of networks: urban, semi-urban, rural and mountainous [2]. The "Distribution Code" contains recom-mendations of the branch for the regulator "Elcom". The paper shows the process of the derivation of minimal reli-ability indices for urban networks. NETWORK, DATA AND DEFINITIONS The urban network of “ewz” covers an area of about 100 km2, consists of 4 HV-, resp. 15 HV/MV-substations and 150 km of HV-cables or transmission lines. The MV-network is composed of 840 km cables and 830 MV/LV-substations, the LV-network of 2000 km cables and 1600 distribution cabins [3]. Energy is delivered to about 260’000 customers (synonymous for meters). Outage data are avail-able for the years 1983-2007. The five urban utilities, which

perform a yearly benchmark, deliver power to between 40'000 to 260'000 customers. The density of the population per hectare (0.01 km2) build up area (buildings, industry, infrastructure) is about 50 to 75 and the number of custom-ers per km LV-grid varies between 70 and 240. Outage data are available for the period 2004-2007. The following con-tinuity indicators are computed: 1) SAIDI (System Average Interruption Duration Index), indicating the duration that energy is not supplied to a customer in a year (min/customer and year); 2) SAIFI (System Average Interruption Frequency Index) revealing the number of times in a year energy is not delivered to a customer (fre-quency/customer and year), resp. MAIFI (interruptions <= 3 min) and 3) CAIDI (Customer Average Interruption Du-ration Index), representing the average time required to restore service to the average customer per interruption (min). The computations are made for unplanned cases with an interruption duration of > 1 sec. For more details see [1] or [3]. A SINGLE URBAN UTILITY The performance of the distribution network in form of customer-related continuity indices are computed in a rou-tine manner at ewz. Due to the long data acquisition period (1983-2007) comparisons with earlier periods can be made. Since yearly values show a large variability from year to year it was decided to use a 5-year moving average for the annual report. Figure 1 compares the yearly values of SAIDI with two moving averages. The single year range varies between 2 and 11 min per customer.

Figure 1: Time-series of yearly values of SAIDI from ewz. Two moving average curves are drawn, moving average 1 is placed in the middle, resp. 2 at the end of the 5-year period. The 5-year average performance for earlier periods are depicted in Figure 2. Over the whole observation period

C I R E D 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0113

CIRED2009 Session 5 Paper No 0113

about 5 min per customer and year have been computed.

Figure 2: The averaging of different periods in the past shows a SAIDI between 3 and 7 min per customer and year. Often a distinction between short and long (sustained) inter-ruptions are made. For example VDN (Verband der Netzbe-treiber) in Germany defines outages > 3 min as sustained interruptions. Figure 3 depicts the SAIDI for different thresholds of minimum duration. Using the 3 min threshold almost no difference between short and long outages is obtained (computed as 25-year average). First by the 30 min threshold is a decrease visible. A threshold of 3 min can be important for the HV-level for a single year or a few year's observation [3]. Whether this finding is valid also for the other types of networks has not been investigated yet.

Figure 3: Decrease of SAIDI, according the minimum duration of observed interruption. Data from ewz 1983-2007. The avera-ged total SAIDI is also stratified into three voltage levels (LV, MV, HV). The time-series of SAIFI is given in Figure 4. The yearly range varies between 0.01 to 0.5. The upper curve shows the total SAIFI and the lower one the value for long dura-tions. The variation of the percentage of the frequency be-tween events of short and long duration is given in Figur 5. Also for SAIFI a five year average is recommended to smooth the sometimes large year to year differences. An interesting projection of all three reliability indices is shown in Figure 6. The depiction is useful to compare the relativ position of the yearly performance for one utility, the comparison of values from different companies or results on the regional or country scale (s. also Figure 10). The whole range of values is seen on one glance. The curves represent the locations of equal SAIDI (SAIFI * CAIDI), the x-axis

the SAIFI (SAIDI / CAIDI) and the y-axis the CAIDI (SAIDI / SAIFI).

Figure 4: Time-series of yearly values of SAIFI, resp. MAIFI. The lower curve displays the SAIFI > 3 min. The difference between both curves illustrates the MAIFI (<= 3 min duration).

Figure 5: The yearly percentage of MAIFI and SAIFI on the total interruption frequency per customer and year.

Figure 6: Combination of all three customer related reliability indices (SAIFI, SAIDI, CAIDI). The dots represents yearly values for the observation period 1983-2007. The curves indicate loca-tions of equal SAIDI (SAIFI * CAIDI). BENCHMARKING FIVE URBAN UTILITIES Now the questions is: How good or bad is the above shown performance? A benchmark with further urban networks in Switzerland should answer the question. Since 2004 five utilities conduct a yearly benchmark and calculate reliability indices in a harmonized manner. The resulting spectrum of values gives an idea of the range which can be expected in Switzerland. Swiss Urban Average (SUA) values can be

C I R E D 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0113

CIRED2009 Session 5 Paper No 0113

determined if we assume that the five urban networks are representative for the country. Examples of results are shown in Figures 7 to 9.

Figure 7: Yearly SAIDI (2004-2007) and the 4-year average for five urban utilities. SUA (Swiss Urban Average) is the yearly and average mean of all five utilities. For utility 3 a blackout event in 2005 was not considered (extreme event). Figure 7 indicates the variability of SAIDI among the years and between the 5 participating networks. The SUA shows already after 4 years a rather robuste value of about 6 min / customer and year.

Figure 8: The same as in Figure 7 but for the SAIFI. The same can be said in observing the SAIFI (Figure 8). For single companies and years the variability can be large but the SUA already shows rather equal values for all 4 years of about 0.2, meaning that an average Swiss urban customers has to expect an interruption every 5 year.

Figure 9: The average percentage over 4 years (2004-2007) of MAIFI (short duration) and SAIFI (long duration) on the total of interruptions. A threshold of 3 min between short and long interruptions were discussed also among the participating companies. The analysis shows a very different picture from company

to company (Figure 9). On the average (SUA) about 35 % of the events are of short and 65 % are of long duration. For comparing the different networks the three indices pro-jection (Figure 10) is drawn to visualize the relative position of each value against the others. Figure 10 shows also (in addition to Figure 6) the SAIDI-curve of 30 min, which is given as a reference value in the guidelines of VSE (see next chapter). All SAIDI are lower than 15 min, which seems to be a good performance in comparison to other countries [1].

0

20

40

60

80

100

120

140

160

180

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70SAIFI

CA

IDI

SAIDI=2min SAIDI=5min SAIDI=10minUtility 1 SAIDI=15min Utility 2Utility 3 Utility 4 Utility 5SUA SAIDI=30min

Yearly and average SAIDI of 5 urban networks and their average (SUA: Swiss urban average) for 2004-2007

Average 2004-2007: U1-U5, SUA

U1U2 U3

U4

U5

SUA

Figure 10: All three indices (individual years and average) for five utilities and for the SUA (Swiss Urban Average) in comparison. GUIDELINES FOR SWISS NETWORKS The working group "minimum reliability" within the VSE had to evaluate guideline values for the internationally used indices SAIDI and SAIFI. Those values were then recom-mended to the recently introduced "Elcom". In 2009 the network operators are obliged to deliver for the first time reliability data to the "Elcom". The working group was depending on results as shown above. A few values were also available for semi-urban and rural networks (not dis-cussed here). The future will show whether the guidline values are reasonable or whether they have to be adjusted. After five years of collection an evaluation should be per-formed. Table 1 shows the so far fixed guidelines as well as the criterions to determine types of network as it is pub-lished in the "Distribution Code" [2]. Another problem is the maximum duration of an interrup-tion. The guideline values are indicated in the last column of Table 1. A tight threshold of 4 hours per event is pro-posed for urban networks. Some companies prefer a restor-ing time of 6 hours. For ewz the percentage of customers exeeding both limits were computed. Table 2 shows that about 2 % of the affected customers experienced durations longer than 4 hours, and about 0.6 % longer than 6 hours. Assuming a penalty of CHF 200.- per customer over the limit an amount of CHF 126'000, resp. 42'000 would be due for the company. In relation to other costs, for both cases, it seems to be a minor problem.

C I R E D 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0113

CIRED2009 Session 5 Paper No 0113

Table 1: Reliability recommandations for Swiss electrical net-works according to the "Distribution Code" of the VSE (Associa-tion of Swiss Electrical Enterprises).

Table 2: The costs of downtime limit of 4 or 6 hours per event, assuming a fine of CHF 200.- if the limit is exceeded. Data are from ewz, averaged over the period 1983-2007. A more important problem is the definition of an "extreme event" also called "Act of God". In the "Distribution Code" a long list is given. From country to country the definitions differ remarkably [4] and comparisons seems to be almost impossible. A better way would be to use a more objective method as used in [1] or proposed by the IEEE [5]. MONITORING NETWORK RELIABILITY

Figure 11: Example of monitoring SAIDI. The situation on 22nd November 2008 is shown. An example of monitoring the performance of reliability, as used by ewz, is given in Figur 11. The columns of the actual year (2008) are growing, according the outages in real time and can be compared instantaneously with previous years. In addition amongst others, histograms are produced for the

5-year average of SAIDI as well as the energy lost per cus-tomer. DISCUSSION AND CONCLUSIONS ewz started to collect outage data in a systematic manner years befor anybody was talking about liberalisation of the power market. Therefore the company can deliver reason-able reliability values for urban networks. A benchmark with other urban networks provides countrywide values which were then used to establish guidelines to be recom-mended in the first run to the regulator. This process shows that the regulator relies very much on the information of the companies which should be regulated in the long run. In-stead using exact number of reliability indices like SAIDI or SAIFI to judge a network operator, the suggestion is made to use a smoothed number which is averaged over several years to lower the yearly variation. This variation can be already significant with only one big case in the record (not an extreme event but e.g. longer outage of a substation) and which can happen to any of the network providers (5-year average is thought to be reasonable). This smoothed value should be compared with other networks within a few ranges, e.g. urban SAIDI: 0-10 (min/customer and year) is a very good performance, 10-30 reasonable and over 30 means "have a good look to the network". This approach would exclude the influence of a stochastic "bad luck" se-ries of outages. Also there would be less temptation for a network provider to embellish the data (e.g. duration, num-ber of customer). Time-series of yearly results, as shown in the last chapter, might help the regulator to put the actual year in a better perspective for each provider. This will be important if a bonus and malus system is intended to be introduced. REFERENCES [1] L. Küng, H.H. Schiesser, R. Cettou, 2007, "Continuity of

supply: Benchmarking five urban electric distribution utili-ties in Switzerland", Proceedings 19th Int. Conf. on Electric-ity Distribution, CIRED2007, Vienna, 4pp.

[2] VSE, 2008, Distribution Code Schweiz, Merkur Access II, Branchenempfehlung Strommarkt Schweiz, Technical Re-port, 58pp.

[3] H.-H. Schiesser, J. Bader, 2005, "Continuity of supply in a Swiss urban electrical network - a spatial analysis", Pro-ceedings 18th Int. Conf. on Electricity Distribution, CIRED2005, Torino, 4pp.

[4] CEER, 2005, Third benchmarking report on quality of electricity supply, Council of European Regulators, Ref: C05-QOS-01-03, 167pp.

[5] IEEE, 2004, Guide for Electric Power Distribution Reliabil-ity Indices, Transmission and Distribution Committee, IEEE Std 1366- 2003, 26pp.