MAINTENANCE AND OUTAGE REPAIRING ACTIVITIES IN ELECTRICITY NETWORKS

Transcription

1 MAINTENANCE AND OUTAGE REPAIRING ACTIVITIES IN ELECTRICITY NETWORKS Thomas Weyman-Jones Department of Economics, Loughborough University, UK Júlia Boucinha Catarina Feteira Inácio EDP Distribuição, Lisboa, Portugal PORTO, September

2 BACKGROUND EDP Distribuição DEA applications For benchmarking purposes International comparisons between European distribution operators; Comparison of performance between different network regions across the Portuguese mainland (14 networks areas and 40 regions); More recently, DEA models have been applied to a number of activities developed by the company, in order to evaluate possible efficiency gains. 2

3 REGIONAL NETWORKS 3

4 ISSUES IN DEA MODELS Some of the problems that can be found in DEA modelling are: Directions of improvement; Negative data; Undesirable outputs and inputs; Efficiency measures which are dependent on units of measurement instead of % scores. 4

5 DIRECTIONS OF IMPROVEMENT Movement to the production frontier need not be along an input or output ray from the origin; Directional Distance Function measure introduced by Chambers, Chung and Färe (1996). y CRS VRS Z 1 = [x 1, y 1 ] βg z 1 z 0 g Z 0 = [x 0, y 0 ] 0 L x 5

6 DIRECTIONAL DISTANCE A producer is observed with the input and output combination: Z 0 = [x 0, y 0 ]; This lies below the efficient production frontier, represented here by either of the CRS or VRS frontiers. Moving this producer to the more efficient point: Z 1 = [x 1, y 1 ] requires that the producer moves along the line L connecting these vectors; The direction of this line is chosen by the researcher to address particular issues in the application; It is given by the elements of vector g. 6

7 DEA MODEL VRS version is: max β s.t. X Y λ λ x y β β g g x y λ 0 j λ j = 1 7

8 APPLICATIONS OF RELATED IDEAS Negative data, Silva Portela et al (2004); Undesirable outputs Färe & Grosskopf (2005); Slacks-based measurement with a % efficiency transformation, Tone (2001). 8

9 MEASURES AND ORIENTATION USED IN THIS STUDY Radial Input oriented Slacks-Based Measurement Input oriented quantifies the input reduction which is necessary to become efficient, holding the outputs constant. Non-oriented quantifies necessary improvements when both inputs and outputs can be improved simultaneously. 9

10 VARIABLES Inputs reflect the resources used Maintenance and Outage Repairing Costs Quality of Supply Indices: Supply Interruptions (Minutes of lost load) Complaints per Client Number of Incidents (LV and Clients Installations) Outputs reflect the activity level Clients (LV+MV) Lines (LV+MV) 10

11 FRONTIER FOR ONE INPUT AND ONE OUTPUT Input: Costs Output: Lines CRS VRS Lines Costs 11

12 FRONTIER FOR ONE INPUT AND TWO OUTPUTS Input: Costs Output: Clients, Lines 4.5 Lines/ Costs (km per 1000 ) CRS Client/ Costs (clients per 1000 ) 12

13 RESULTS DEA VRS % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Inputs: Costs, Interruptions, Complaints/cl., Incidents All efficient networks are Pareto-efficient (slacks=0) Ouputs: Clients, Lines 13

14 RESULTS DEA VRS COMBINING DATA % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Inputs: Costs, Interruptions, Complaints/cl., Incidents Improvement in overall efficiency from 84% to 91% All efficient networks are Pareto-efficient (slacks=0) Ouputs: Clients, Lines 14

15 EFFICIENCY IN 2005 DEA - VRS Inputs: Costs, Interruptions, Complaints/cl., Incidents Ouputs: Clients, Lines 100% 90% 80% 70% 60% 50% 40% 30% 16 networks below average 5 networks above average but not 100% efficient 19 networks 100% efficient 20% 10% 0% 15

16 RESULTS DEA VRS individually Combining Data 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Inputs: Costs, Interruptions, Complaints/cl., Incidents Ouputs: Clients, Lines Comparison shows that considering 2004 individually the performance of networks is better than combining data. 16

17 RESULTS DEA VRS individually Combining Data 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Inputs: Costs, Interruptions, Complaints/cl., Incidents Ouputs: Clients, Lines Comparison shows the results for 2005 remain fairly similar. 17

18 NEGATIVE OUTPUTS Considering Quality of Supply variables as negative outputs instead of inputs, we may conclude: The same network regions come as 100% efficient; Average efficiency level is slightly lower. 18

19 RESULTS FROM DIFFERENT MEASURES (2005) IO-Radial IO-SBM NO-SBM 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Input: Costs, Interruptions, Complaints/cl., Incidents Comparison shows that results are similar. Ouputs: Clients, Lines 19

20 COMPARISON OF RESULTS FROM DIFFERENT SBM MODELS 100% Non-oriented versus input oriented Non Oriented - SBM 90% 80% 70% 60% 60% 70% 80% 90% 100% Input Oriented - SBM 20

21 Efficiency Rankings The consistency of the efficiency rankings may be shown through the Spearman s rank correlation coefficient: IO RADIAL vs IO SBM IO RADIAL vs NO SBM IO SBM vs NO SBM Spearman 97% 97% 99% Efficiency rankings between different models are highly correlated; 19 best performers (100% efficient) are the same in all the measurements. 21

22 CONCLUSIONS (1) Used a variety of directional distance function measures work is still ongoing; Strong positive correlation between the rankings on all measures used; The results are very similar in all models. 22

23 CONCLUSIONS (2) Use of DEA models to evaluate: Efficiency of different network regions; Set targets for less efficient regions; Find benchmarks for each network region, in order to identify best practices. Average level of efficiency has improved between 2004 and

24 APENDIX 24

25 Min such that DEA MODEL The technical input efficiency rating of a network being measured is θ*, where θ* is the optimal value of θ in: CRS-Model = ( ( slacks ) ( network output network weight ) ( network output+ slacks) ( network input network weight) ( ) = θ network input all θ ε for each output and input network weights - slacks 0 VRS-Model - add the contraint: ( = 1 network weight) network being measured network being measured Network is Pareto-efficient if θ*=1, and Slacks=0 : 25