Approaches of Setting SAIDI Target Value for Taiwan Power System

Transcription

1 Approaches of Setting SAIDI Target Value for Taiwan Power System S. C. Hsu*, F. C. Lu**, S. L. Chen*, Y. F. Chuang***, C. M. Lin*** and Mark Lauby**** This paper presents a method for setting the annual target on customer s electricity supply reliability for the regional transmission offices of Taiwan Power Company (Taipower). The reliability of customer s electricity supply is measured by the System Interruption Duration Index (SAIDI) and the System Interruption Frequency Index (SAIFI) in Taipower. The analytic hierarchy process (AHP) has been adopted, by which all the influential factors on SAIDI and SAIFI can be presented into a hierarchical structure which describes how these factors affect the annual SAIDI (or SAIFI) level. The evaluation indices that measure the regional status of these influential factors are designed at the lowest layer of the hierarchical model. The relative weightings among the influential factors are derived from the linguistic comparison results obtained by the questionnaire survey conducted on the maintenance and operation engineers of Taipower s 6 regional transmission offices. The evaluation index values of regional transmission offices are multiplied to the weighting of each influential factor from the lowest layer upwards until the objective values at the top of hierarchical structure are calculated. On basis of the relative comparison among these 6 objective values calculated, Taipower s corporate SAIDI (or SAIFI) target is then distributed to be the 6 regional targets. To provide a further insight into the effectiveness and limitation of proposed method, the AHP results on Taipower s corporate target distribution obtained from the calculation are compared with Taipower s SAIDI records at The method after slight modification can be extended to other power engineering management target. * Professor Shi-Lin Chen and Ph.D student Shih-Che Hsu: Department of Electrical Engineering, National Tsing Hua University, 101, Section 2, Kung Fu Road, Hsinchu City, Taiwan 30013, Tel: , Fax: , d897901@oz.nthu.edu.tw. ** Associate Professor Feng-Chang Lu: Department of Electrical Engineering Ta Hwa Institute of Technology, 1, Ta Hwa Road, Chonglin Township, Hsinchu County, Taiwan 30740, Tel: ~2719, Fax: , fclu@ee.thit.edu.tw *** Section Chief Yu-Fa Chuang and Section Head Chi-Ming Lin: Department of System Planning, Taiwan Power Company, 242, Section 3, Roosevelt Road, Taipei City, Taiwan 100, Tel: , Fax: , d02304@taipower.com.tw **** Managing Director Mark Lauby: Electric Power Research Institute, 3412 Hillview Avenue, Palo Alto, California USA, P.O. Box 10412,Tel: , Fax: , mlauby@epri.com I. INTRODUCTION Taipower has been endeavored to improve customer s power supply reliability. Thus reducing SAIDI and SAIFI values has been accounted as one of Taipower s key management targets. Since 2000, Taipower has started with the setting of annual corporate target for the SAIDI (and SAIFI) and also with the assignment of regional targets for each of Taipower s regional offices including regional transmission and distribution offices. The assignment has been based on the reliability performance or the SAIDI (and SAIFI) records of each office in the previous year [1-3], e.g., by multiplying the same percentage to the SAIDI actual levels as next-year s targets for all the regional offices, which are thus unfair to the regional offices already attaining a lower SAIDI (or SAIFI) level and also unfair to the regional offices encountering more inherent difficulty than others in reducing SAIDI and SAIFI levels.. To solve these problems, EPRI S.A. contracted a research project to National Tsing Hua University for the design of a rational procedure in setting Taipower corporate as well as

2 regional SAIDI (and SAIFI) targets. For a clear presentation, the annual SAIDI due to the forced system interruption of transmission system is taken as an example to present the procedure on setting the annual target for Taipower s 6 regional transmission offices on basis of the annual corporate target. The procedure accounts for the disparity among 6 regions on the regional geographic, load and circuit conditions etc. In the study, scheduled transmission system interruptions have not been taken into account, because, after proper coordination with customers, construction or maintenance work of transmission system commonly do not affect customer s electricity supply. Although the influential factors for the SAIDI or SAIFI due to distribution system interruptions are not the same as for the transmission system interruptions, the same procedure can be applied to the annual target setting of Taipower s distribution offices. Furthermore the designed procedure, after slight modification, can be extended to the target setting for other power engineering management. II. INDICES FOR MEASURING CUSTOMERS POWER SERVICE INTERRUPTION Customers power service interruption can be measured by a variety of reliability indices [4-5]. Among them the SAIDI and the SAIFI with the following definition have been adopted in Taiwan: SAIDI m Ki i= 1 j= 1 C C ij T ij (1) SAIFI C where C i, C ij : Number of customers interrupted T ij : Interruption duration C: Total number of customers served m i=1 C i (2) In Eqs. (1) and (2), C ij refers to different groups of customers power service interruption due to the same system outage (i); T ij follows the same way of definition. For easier presentation of the methodology adopted in the study, only the SAIDI due to transmission forced system interruptions are presented in this paper. This paper aims at analyzing the rational differences of target value on customers interruption duration among regional transmission systems of Taipower. Because the factors that affect the customers interruption duration are intricate, the AHP has been adopted to work out the regional target values with the operation and maintenance engineers of Taipower s 6 regional transmission offices. III. GENERAL DESCRIPTION OF THE METHOD The flowchart of working procedure for this study is shown in Fig. 1. To identify the relative weighting of factors that affect customers power service interruption duration relies heavily on decision makers experiences which have been extracted in this research through questionnaires designed by following the AHP. To assist the questionnaire design, each influential factor is measured by a numerical index after several meetings with the regional transmission maintenance and operation engineers, so to capture the essential characteristics of regional transmission system.

3 Select major influential factors in transmission system that affect regional SAIDI levels Interview with professionals of Taipower, Korea and Japan to select influential factors Eliminate these factors which are common to all regional offices Design a numerical index for each factor Design numerical indices Collect index values for each regional office. Design questionnaire Identify the relative weighting among influential factors Estimate the reasonable disparity duration among regional offices on SAIDI targets Conduct surveys on Taipower s operation and maintenance engineers Analyze the results of survey or questionnaires *Apply AHP to analyze questionnaire result, estimate the relative weighting among factors. Distribute the corporate target to regional offices *Estimate the share of duty (or duty ratio) on corporate s SAIDI reduction ( SAIDI) among regional offices. *Estimate SAIDI s reduction target ( SAIDI) for each regional office. Fig. 1 Flowchart of working procedure in SAIDI target setting for Taipower s regional transmission offices A. Two Types of Model Parameters The AHP proposed by professor Saaty in 1980 [6-10], has been applied in this research to sort out the multiple factors that affect system interruption duration into a hierarchical structure as shown in Fig. 2. The main purpose of AHP is to evaluate two kinds of parameters for each influential factor of Fig. 2: The first type of parameters: Numerical indices are designed to measure the influential factors at the bottom layer of the structure such as those at the third layer of Fig. 2. The index values should be able to portray the regional characteristics of transmission system and express the disparity among regional systems. These data could be collected subjectively from decision makers or could be taken directly from the objective data source. With these measurement indices, the authors of this paper then conducted survey at each regional transmission office to collect data for the indices so to describe the present status for each regional transmission system. As to the index values for the factors at upper layers, an inference mechanism is provided for their evaluation by AHP to be presented below. The second type of parameters: The parameters refer to the weighting, which measures the relative importance among factors at the same layer. They are thus subjective data. These weightings are commonly collected through questionnaire interviewees with decision makers. B. Inference for Values of Numerical Index to Measure Upper Layers Criteria The index values to measure the influential factors at upper layers (or the evaluation criteria at the second layer of fig. 2) are derived from the two types of model parameters at the lower layers:

4 Evaluation Target First Layer The First Evaluation Criterion of Layer 2 The Second Evaluation Criterion of Layer 2 Second Layer The N 2 th Evaluation Criterion of Layer 2 The First Evaluation Criterion of Layer 3 The Second Evaluation Criterion of Layer 3 Third Layer The N 3 th Evaluation Criterion of Layer 3 Fig. 2 AHP structure (three layers as an example) The value of numerical index to measure criteria at layer m N m + 1 i= 1 = Vmi Wm i (3) where Vm i : Value of numerical index for the i th factor at layer m+1. Wm i : Weighting of the i th factor at layer m+1. N m+1 : Total number of factors at layer m+1. C. Acquisition of Weighting Values Referring to Fig. 2, the evaluation criteria listed at the same layer are pairwisely compared by decision makers through a self-administered questionnaire survey. In the linguistic comparison, the degree of relative importance is presented into ratio scale. These ratio values ticked or filled into blanks of questionnaire by decision makers can then be complied into a comparison matrix. The calculation procedure designed of AHP for the weighting of evaluation criteria as presented inside the dash box of Fig. 3 is followed. In the procedure, the eigenvalue and eigenvector for the comparison matrix is calculated to obtain a consistency index and a consistency ratio [6] which evaluate the consistency among pairwise comparisons. If the consistency fulfills the requirement, the eigenvector under evaluation is taken as a preferential vector to represent the weightings of criteria at the same layer; if not, revision of the comparison matrix or redo of the survey are necessitated. IV. MODEL FORMULATION AND PARAMETER EVALUATION A. Formulation of AHP Model The formulation of AHP model shown as an example in Fig. 2 comprises of the following two steps (ref. Fig. 1): Selection of evaluation criteria or factors that affect customers power service interruption due to the forced interruption of transmission system, Categorization of evaluation criteria into a hierarchical structure.

5 Categorize Criteria into Layers Select Evaluation Criteria Build AHP Model, in Form of Hierarchical Structure Single Layer Pairwise Comparison Revising Judgments Build Comparison Matrix Replace that Criterion in Comparison Matrix by the Corresponding Ratio of Elements in Eigenvector Solve Eigenvector & Max. Eigenvalue Find the Max. Absolute Differences between the Criteria of Comparison Matrix and Ratio of Elements in Eigenvector No Calculate Consistency Index & Consistency Ratio Does Comparison Matrix Accord with Consistency Index? Yes Obtain the Relative Weighting among Criterion Repeat The Derivation by Eq. (3) until The Target Value at The Top Layer Is Derived Derive Index Values for Criteria at Upper layer (ref. Eq. (3)) Fig. 3 Flowchart of AHP (single layer as an example) SAIDI The resulted AHP model is shown in Fig. 4 where Fig. 4(a) is derived from Eq. (1): = [( times of service interruption per year) ( duration per service interruption) (4) ( number of customers affected per service interruption)] / (Total number of customers served) As shown, the numerator can be decomposed into three terms. Each is then evaluated individually by Fig. 4(b) ~ (d), which are explained below: Fig. 4(b): The total times of forced system interruption in a regional transmission system could be affected by a variety of reasons; some are common to all the regional transmission systems (e.g. shortage of manpower) and some can tell the disparity among the regional systems (e.g. inferiority in weather condition). Among the latter reasons, some can be

6 controlled by the management, but some are inherent natures of the region, e.g., weather severity, circuit length, customers substation type, system loading condition, geographical condition and network configuration etc., as shown in Fig. 4(b). Only these inherent natures which are uncontrollable in short time (e.g. within a year) by the regional offices, are accounted for in the model. Fig. 4(c) and 4(d): By following the same rule as for Fig. 4(b), the accounted are shown in Fig. 4(c) and 4(d). Among them, customers ineffective support refers to the response time that takes for the customer to respond to the fault (such as to trail energize the feeders to restore power etc.). The overhead/underground ratio is to account for the longer time required for underground line on fault location and repair than for the overhead. B. Design of Numerical Index Table 1 provides numerical indices designed to measure the evaluation criteria, or influential factors, at the bottom layer for each structure of Figs. 4(b) ~ (d). The design tends to make use of the presently exist data of Taipower s regional transmission offices so to reduce data collection effort. System System Interruption Duration Times of Service Interruption per Year Duration per Service Interruption Number of Customer Affected per Service Interruption (a) Times of System Interruption per Year Thunderbolt, Salt or Fog Overhead Circuit Proportion Total Circuit Length Customer s Outdoor Substation Equipment Load Rate Geographical Environment Radial Circuit Proportion Urban Suburb Mountain (b)

7 Duration per System Interruption Overhead / Underground Customer s Ineffective support Geographical Environment Overhead Underground Urban Suburb Mountain (c) Number of Customer Affected per System Interruption Load Transfer Inability Customer Density Lack of Backup Capacity Equipment Load Rate Geographical Environment Radial Circuit Proportion Urban Suburb Mountain Urban Suburb Mountain (d) Fig 4 The AHP structures of influential factors for transmission forced outage: (a) SAIDI can be decomposed into three terms, (b) The influential factors of average times of system interruption per year, (c) The influential factors of average duration per system interruption, (d) The influential factors of average number of customer per system interruption. C. Questionnaire Design and Weighting Evaluation Questionnaires have been designed for acquisition of the weighting for each influential factor at the same layer in the AHP model. Take the factors affecting the load transfer inability for evaluation of the average number of customers per forced service interruption (ref. Fig. 4d) as an example. Fig. 5 shows one of the survey results among 90 engineers of Taipower

8 regional transmission engineers. As shown, the respondent accounted for the network configuration (referring to the radial circuit proportion out of the total circuits) as of essential importance over the lack of circuit backup capacity, and accounted for the geographical environment as of equal importance over the system average loading condition. The survey results of Fig. 5 are then expressed into a comparison matrix by following Satty s scale table of intensity [11]. The matrix is then checked with the consistency among the matrix elements. If the consistency does not fulfill the pre-specified consistency level (e.g. 0.1 in Appendix A), the matrix is then revised until the consistency is met. The process then yields the weighting of each influential factor under evaluation. For the example of Fig. 5, the weighting are evaluated as 0.29, 0.27, 0.17 and 0.27 for the 4 factors, i.e., geographical environment, radial circuit proportion, average equipment load rate and lack of circuit backup capacity, respectively (ref. Appendix A). The same process of Appendix A has applied to each layer of the hierarchical structure of Figs. 4(b) ~ (d). Table 1 Numerical index designed to specify the status of regional transmission system on the influential factors that affect SAIDI Influential Factors Interruption caused by thunderbolt, salt or fog Overhead transmission line proportion Total circuit length Customer s outdoor substation equipment load ratio of peak hour Radial circuit proportion The degree of ineffective support on customer s operation Geographical environment Lack of circuits backup (N-1) operating capability Numerical Index Designed to Specify Regional Status The number of power service interruption times caused by thunderbolt, salt or fog with occurrence at regional transmission system in 2003 The percentage of overhead line in circuit length out of the total circuit length The total circuit length in kilometers of regional transmission system The number of outdoor substations owned by customers The annual peak load / the total capacity of main transformers at regional transmission system in 2003 The radial circuit percentage in circuit length out of the total circuit length Number of customers that cannot cooperate with Taipower on the efficient electricity restoration (e.g. those, due to shortage of manpower, refusing to re-energize load within one minute etc.) Kilometer squares for urban, suburban and mountainous areas in the region (as to Table 2c, the listed are the proportions calculated from the km 2 data) Number of circuits that do not meet (N-1) criterion Importance of A over B Importance of B over A Column A Essential Weak Equal Weak Essential Column B Geographical Environment Geographical Environment Geographical Environment Radial Circuit Proportion Radial Circuit Proportion Equipment Load Rate Radial Circuit Proportion Equipment Load Rate Lack of t Backup Capacity Equipment Load Rate Lack of Backup Capacity Lack of Backup Capacity Fig 5 One of the survey results among 90 engineers of Taipower regional

9 transmission engineers. D. Inference for quantification of factors at upper layers Having obtained the weighting of each factor at the same layer and the quantification (or index value) for each factor at the lowest layer, Eg.(3) is then applied to the inference for quantification of the factors at the upper layers. The quantified values at the top layer of Figs. 4(b), (c) and (d) are then substituted into Eg.(4) for the quantified value of SAIDI for each of Taipower s 6 regional transmission systems, denoted by: q j (SAIDI) for j=1,2,,6 respectively. The corporate SAIDI target, or targeted SAIDI reduction ( SAIDI), can thus be rationally distributed in proportion to the resulted 6 SAIDI quantified values (q j (SAIDI)) as the targeted SAIDI reduction for Taipower s 6 regional transmission system: SAIDI for j th region = (Corporate SAIDI) 6 1 q j(saidi) q j(saidi) (5) A. Numerical Index Results V. RESULT ANALYSIS Table 2 shows the data collection results for numerical index designed to specify the status of each regional transmission system on the influential factors listed in Table 1. For easier presentation, the results have been normalized versus the sum of data collected from 6 regional transmission offices. Take the total circuit length of Table 2(a) as an example. The circuit length of Taipei, Shintao, Taichong, Gianan, Gaupin, and Hwadong regional systems are: 2422, 2957, 3055, 3013, 2327 and 1005 km respectively. The normalized for Taipei regional system is thus: 2422 / ( ) = Referring to Table 2(a), among the 6 regions, Shintao and Gaupin regions have more times of power service interruption caused by weather, than the remaining 4 regions in 2003 (ref. Table 1). Also the geographical data in Table 2(a) and (b) indicate that Taichong has the large km 2 of mountainous area than the remaining 5 regions, i.e., Taichong could have a relatively higher probability to be hit by lightning and, after each line tripping in mountainous area, could need a longer time to repair. As for Table 2(c), which refers to the number of customers of each service interruption, the factors accounted are irrelevant to region s km 2 but to Taipower s design criteria towards urban, suburban and mountain areas as well as to customers density of the region. In the latter two aspects, the accounted are the relative proportion of urban, suburban and mountain areas out of the km 2 for the same region. Also, taking Taichong as an example, because Taichong has the largest km 2 for its mountainous area, the relative proportions irrespective for Taichong s urban and suburban areas are thus small, resulting in the least urban proportion and the second least suburban proportion out of the 6 regions (ref. Table 2c). These two factors could lead to a postulation by the model that Taichong has less number of customers interrupted out of Taiwan s 6 regions for service interruption (ref. Table 4 to be detailed later in section V.). Also due to the largest km 2 for the mountain area of Taichong, Taichong has the highest radial circuit proportion out of Taiwan s 6 regions (Table 2a). Taipei has the second highest radial circuit proportion, because Taipei has the second largest km 2 for its suburban out of Taiwan s 6 regions (Table 2a). As to the peaking load factor, among Taiwan s 6 regions, Shintao has the highest and Taichong has the second highest (Table 2a), both of which will result in a high power service interruption times per year postulated for Shintao and Taichong (to be detailed also in Table 4). Because Shintao and Taichong possess the highest and second highest number of industrial customers among Taipower s 6 transmission service regions (referring to the column under customers outdoor substations in Table 2a), their underground

10 circuit proportion out of the total circuit-length are thus the highest (Table 2b). Table 2 Collected numerical index values for regional status description on influential factors listed at bottom layers of 6 regional AHP structure in Fig. 4: For factors affecting (a) average times of service interruption per year; (b) average duration per service interruption; (c) average number of customers per service interruption. (a) First layer times of service interruption per year Second layer Thunderbolt, salt or fog Overhead circuit proportion Total circuit length Customer s outdoor substation equipment load ratio Geographical environment Radial circuit proportion Third Urban Suburb Mountain Taipei Shintao Taichong Gianan Gaupin Hwadong (b) First layer duration per service interruption Second layer Overhead /underground Customer s ineffective support Geographical environment Third layer Overhead Underground Urban Suburb Mountain Taipei Shintao Taichong Gianan Gaupin Hwadong (c) First layer number of customers per service interruption Second layer Load transfer inability Customer density Third layer Geographical environment Radial circuit Factor Urban Suburb Mountain proportion equipment load ratio Lack of backup capacity Urban Suburb Mountain Taipei Shintao Taichong Gianan Gaupin Hwadong Because Huadong is sited at the east coast of Taiwan and the northwest wind is retained by Taiwan s central mountain, and also because Huadong has the least population among Taiwan s 6 regions, Huadong has zero time of service interruption caused by weather, the least circuit length, the least customers outdoor substations, the least peaking load factor (all in Table 2a) and the least underground circuit proportion (Table 2b). B. Results for Weighting of Influential Factors A survey meeting has been conducted at each of Taipower s 6 regional transmission offices by the research team with fifteen experienced engineers who are either the section

11 chief or the department head of the operation or maintenance department in the same regional transmission office under evaluation. Thus total 90 engineers have been surveyed, each ticking the blanks in the same way as in Fig. 5. of the questionnaires, i.e., each answering 3 pieces of questionnaires, based on the 3 AHP sub-models in Figs. 4(b) ~ 4(d). For each AHP sub-model, these 90 copies of questionnaires answered by the engineer are then converted into 90 comparison matrices (in the same form as of Eq.6 matrix), and combined into one comparison matrix by taking the geometric mean of the 90 values for each matrix element. As the geometric mean results are close to the arithmetic mean (ref. Fig. 6) and the ratio scale adopted by AHP is also based on multiplication rather than summation [11]. Although the 6 regions are given equal rights in the estimation of factors weighting, the weighting results can be similar among 6 regions (ref. Fig. 7), or can be disparate (ref. Fig. 8). The weighting calculated by following the AHP such as for that in Appendix A for each influential factor of the three AHP substructures in Fig. 4(b) ~ (d) are shown in Tables 3 (a)~(c). Referring to Table 3(a), among 7 factors under evaluation, weather conditions are reckoned as the most important (0.187) in affecting the times of service interruptions per year, although the differences among 7 factors are not significant. Because of the high circuit density, urban area is given the highest weighting (0.461) among three geographical areas. As to the factors affecting service interruption duration (ref. Table 3b), power restoration after underground cable fault usually takes longer time than overhead; thus the former is given a higher weighting (0.71) than the later (0.29). Due to the same reason, mountain area is given a higher weighting (0.53) than the remaining 2 geographical areas. Table 3(c) gives the weighting results for the factors affecting the number of customers per service interruption. As shown, load transfer inability is given a higher weighting (0.62) than customer density (0.38) and the lack of circuit backup capacity is the main cause for the large number of customers per service interruption. Due to the higher loading factor of circuit, urban area is given higher weighting (0.46) than the other geographical areas in measuring the degree of load transfer inability. Weighting Overhead/Underground (for average duration per service interruption Overhead Geometric Arithmetic Underground Influential Factor Fig. 6 The geometric mean results are close to the arithmetic mean, for the influential factors (taking the line type proportion as an example).

12 Weighting Overhead/Underground (for average duration per service interruption Overhead Underground Influential Factor Taipei Shintao Taichong Gianan Gaupin Huadong Fig. 7 For some factors, the weightings are almost unanimous among 6 regions (taking the line type proportion as an example). Weighing Times of Service Interruption per Year Thunderbolt, Salt or Fog Overhead Circuit Proportion Total Circuit Length Customer s Outdoor Substation Equipment Load Rate Geographical Environment Radial Cir Proportion cuit Taipei Shin Tao Taichong Influential Factor Gia Nan Gau Pin Hua Dong Fig. 8 For some factors, the weightings are disparate among 6 regions (taking the 7 factors affecting outage frequency as an example; because of Huadong s unique weather and loading conditions, the weighting given by Huadong differ from the weighting by other regions). Table 3 The weighting of influential factors calculated by AHP for the three AHP sub-models in Fig. 4: (a) Factors in the sub-model of Fig. 4(b); (b) Factors in the sub-model of Fig. 4(c); (c) Factors in the sub-model of Fig. 4(d). (a) times of service interruptions per year Influential Factors Thunderbolt, salt or fog Overhead circuit proportion Total circuit length Customer s outdoor substation equipment load ratio Geographical environment Radial circuit proportion Weighting Influential Factors Urban Suburb Mountain Weighting Influential Factors Overhead /underground (b) duration per service interruption Customer s ineffective support Geographical environment Weighting

13 Influential Factors Overhead Underground Urban Suburb Mountain Weighting Influential Factors (c) number of customers per service interruption Load transfer inability Customer density Weighting Influential Factors Geographical environment Radial circuit proportion equipment load ratio Lack of backup capacity Urban Suburb Mountain Weighting Influential Factors Urban Suburb Mountain Weighting C. Rational Distribution of Corporate SAIDI Target to Regional Offices Having obtained the index values for measuring the status disparity among 6 regional transmission systems (ref. Table 2) and the weighting of each factor in the three hierarchical structures under evaluation (ref. Table 3), the final stage of AHP is to estimate the objective at top of the AHP structure (ref. Fig 4a ) by applying Eq. (4). For this purpose, each term in the bracket of Eq. (4) must be estimated first. Take Taipei region as an example; the three terms evaluated are irrespectively at 0.183, 0.169, and 0.174, where = ( ) Namely, it is a sum of multiplication of regional index value for each of total 9 factors (in Fig. 4b) with its corresponding weighting. In the same way, and are calculated. The rational SAIDI under normalization for Taipei region is and the rational distribution of Taipower s corporate SAIDI target to Taipei region is then calculated as 17.60%= / (ref. Table 4). The relative proportions among Taipower s 6 regions on the regional SAIDI performance of year 2003 are also given in Table 4. As shown in Table 4, among Taipower s transmission regions, Shintao, Taichong and Hwadong are irrespectively the first, the second and the last in their rankings according to the rational service interruption times per year from high to low (measured as 0.197, 0.192, and 0.092). The highest for Shintao is due to its severe weather and high peaking load factor (measured at and respectively in Table 2a ); the second highest for Taichong, due to its high radial circuit proportion and high peaking load factor (measured at and in Table 2a); the last for Hwadong due to its good weather, least circuit length, least customers outdoor substations, least peaking load factor and least radial circuit proportion ( measured at 0, 0.068, 0.015, and respectively, also in Table 2a). Referring to Table 4, Gaupin and Taichong have relatively larger duration per power service interruption measured at and respectively due to customers ineffective support and high underground cable proportion for Gaupin (measured at and respectively in Table 2b), and high km 2 for mountain area and also high underground cable proportion (measured at and respectively in Table 2b). As to the number of customers per service interruption, Shintao and Taichong are both higher than the other regions (measured at and in Table 4) due to their insufficient circuit backup capacity, high peaking load factor and high radial circuit proportion (ref. Table 2c) The resulted order according to the rational SAIDI distribution evaluated from high to low in Table 4 is: Taichong > Shintao > Taipei > Gaupin > Gianan > Hwadong This order has been compared with the following arranged according to th actual SAIDI performance of year 2003: Shintao > Taipei > Taichong > Gaupin > Gianan > Hwadong The rational order may not have to accord with the performance order, and can serve,

14 if sufficient index data have been collected, as guideline for setting the regional SAIDI target. Table 4 Rational distribution of corporate SAIDI target to 6 regional offices as compared to the SAIDI record in year 2003 Regional office (1) times of service interruption per year (2) duration per service interruption (3) number of customers per service interruption Normalized regional SAIDI (1) (2) (3) Rational SAIDI proportion (%) Proportion for SAIDI records in 2003 (%) Taipei Shintao Taichong Gianan Gaupin Hwadong Note: Sum of normalized regional SAIDI s is VI. CONCLUSION A method is presented in this paper which is able to rationally distribute Taipower s corporate SAIDI target to the 6 regional transmission offices of the company. The methodology is designed on basis of the AHP, which encompasses: (1) the selection of influential factors and formulation of these factors into a hierarchical structure, (2) the design of numerical index and collection of index values for description of the regional status on those influential factors at the bottom layer of the hierarchical structure, (3) the design of questionnaire and conduction of survey on total 90 engineers of Taipower s regional transmission offices for linguistic comparison among the factors so to evaluate the weighting of each factor, as well as (4) the evaluation for a rational distribution of corporate SAIDI target to the 6 regional transmission offices of Taipower. The numerical data and results as well as a variety of insight analysis have been presented, which demonstrated that, with sufficient historical data collected for the regional status description, the method presented can be an effective and convincing tool in the distribution of corporate SAIDI target to the regional offices of the company, The same method, after slight modification, can be applied to other power engineering management target. VII. ACKNOWLEDGEMENTS The authors are grateful Anthony Yung-Tien Chen, the director of system planning department; Tzong-Yih Guo and Der-Hwa Huang, both the deputy directors of system planning department; Ch-An Lin and Ui-Wn Huang, the division director and section chief of transmission department all of Taipower. The assistance from professionals and seniorities of Taipei, Shintao, Taichong, Gianan, Gaupin and Hwadong regional offices are also acknowledged. VII. REFERENCES [1]Implementation Methods to Achieve System Reliability Targets Set for Regional Transmission Offices, Transmission Department of Taipower, [2]Forced System Interruption Statistics on Transmission and Substation Equipments, Transmission Department of Taipower, [3]Rules for Processing Work Outage Statistic Data of Distribution Feeders, Distribution Construction Team of Taipower, [4]R. Billinton, and J. E. Billinton, "Distribution System Reliability Indices", IEEE Trans. On

15 Power Delivery, Vol. 4, No. 1, January 1989, pp.561~568. [5]C. A. Warren, D. J. Pearson and M. T. Sheehan, A Nationwide Survey of Recorded Information Used for Calculating Reliability Indices, IEEE Trans. on Power Delivery, Vol.18, No.2, April, 2003, pp.449~453. [6]T. L. Saaty, The Analytic Hierarchy Process: Planning, Priority Setting, Resource Allocation (Vol. I), RWS Publications, U.S.A., [7]T. L. Saaty, Fundamentals of Decision Making and Priority Theory with the Analytic Hierarchy Process (Vol. VI),RWS Publications, U.S.A., [8]T. L. Saaty, Axiomatic Foundation of the Analytic Hierarchy Process, Management Science, Vol. 32, No. 7, July 1986, pp.841~855. [9]T. L. Saaty, and M. Takizawa, Dependence and Independence: from Linear Hierarchies to Nonlinear Networks, European Journal of Operations Research, Vol. 26, 1986, pp.229~237. [10]T. L. Saaty and L. G. Vargas, Inconsistency and Rank Preservation, Journal of Mathematical Psychology, Vol. 28, No. 2, June 1984, pp.205~214. APPENDIX A. EXAMPLE OF REVISING JUDGMENT Following Satty s scale table of intensity, the survey results of Fig. 5 can be transformed into the following comparison matrix: Geo. Rad. Avg. Cir Geo. Rad. A = 1/ (6) 1 1/ 2 1 1/ 2 Avg. 1 1/ Cir. The largest eigenvalue (λ max ) of this matrix = 4.443, the normalized eigenvector (W major ) = [ ], the consistency ratio (CR) = With the acceptable CR pre-specified at 0.1, this result is unacceptable and then the revising judgment is required. To proceed the revision, an ideal matrix denoted by X, is formulated from the normalized eigenvector (W major ), which due to its absolute consistency (i.e., CR=0), can be taken as the reference to revise matrix A. The ideal matrix X so evaluated is: / / / X = 0.32 / / / 0.22 = (7) 0.17 / / / / / / The absolute differences between matrix X and matrix A are: X - A = (8) The maximum value in this matrix is 1.54 at the second row and the 4 th column, so its corresponding element of matrix A (i.e., 3 and 1/3) should be replaced first by the elements of matrix X at the same position (0.32/0.22 and 0.22/0.32). The modified comparison matrix (A 1 ) is:

16 A 1 = 1/ / 0.22 = (9) 1 1/ 2 1 1/ / The largest eigenvalue (λ max ) of A 1 is , the normalized eigenvector (W major ) = [ ], the consistency ratio (CR) = < 0.1. The 4 factors expressed in the comparison matrix, are then derived as [W geo W rad W avg W cir ] = [ ].