Resurrecting A Load Research Program Susan Romer NSTAR 2006 AEIC Load Research Conference September 2006 1 NSTAR Territory •Massachusetts’ Largest Investor-owned Electric And Gas Utility •Transmitting & Delivering Electricity & Natural Gas For More Than 100 Years •Serve @1.4 M Residential & Business Customers In Over 100 MA Communities Electric NSTAR Electric Gas Residential 970,000 160,000 Business 225,000 20,000 Total 1,195,000 180,000 NSTAR Gas Combined Gas & Electric •@ 3,200 employees 2 NSTAR - Electric • Three Electric Operating Companies – Boston Edison Company (BECO) – Commonwealth Electric (COMM) – Cambridge Electric (CAMB) • • • 4,959 MW – Peak Load (08/02/06-4 PM) 23,236 GWh – 2005 Annual Energy 17% Share of ISO-New England Load • Load And Customers Served by 13 Competitive Suppliers as of July 2006 – 47% of load • (Industrial – 77%; Commercial – 28%; Residential 16%) – 18% of customers • (Industrial – 59%; Commercial – 26%; Residential 17%) 3 Where’s Load Research Load Research 1999 Sales?? 2005 Load Services 2000 Energy Supply 1993-1998 Homeless 1993 Dismantled 1980-1993 Rates PURPA Baby Finance & Accounting > Energy Supply > Load Services > Load Research – 1 Analyst > Load Settlement & Load Response – 3 Engineers 4 Detour Impact 1992 - 10 Year Rate Freeze Loss of employees with knowledge 1993 - MA Electric Deregulation No senior management sponsor 1999 - Merger 2000> - Management Changes No field commitment Load Research “Those Meters Are Only Used For Research” Data not available when needed Load data gateway shifted to others 5 Regulatory Requirements • Continuous Load Research • Meet 95/10 Sample Results Accuracy Levels • Most Northeast Utilities Perform Load Research At 95/5 Accuracy Levels • Many Northeast Utilities Use Same Month Profiles For Settlement Month 6 Opportunities • 2004 Rate Case and No Load Research Data • Awareness of ISO Settlement Dispute B/W Two Other Parties • Understanding of How Interchange Point Malfunction Can Impact the Calculation of System Loads • 2005 Energy Policy Act DST Change Fear of Financial Exposure = Senior Management Sponsorship 7 Load Research Supports • Estimation of Competitive Supplier Loads and ICAP Values • Development of Rates Charged to Customers • Unbilled Energy Sales And Unbilled Revenues Balances • Billed/Unbilled Sales & Substation Forecasts • Accrual Of Purchased Power Expense As Compared to Actual • Development of Rates Charged to Customers • Local Network Service Billing for Wholesale Customers • Regional Network Transmission Billing for ISO • Development of Engineering Loss Studies 8 Load Settlement & Load Research •Determine suppliers’ load obligations from load research for the purposes of ISO-New England energy market settlement •A contributing factor to unaccounted for energy (UFE) Metering Error – Meter Failures Load Profiling Error** – Sample Estimates Accounting Error – Billing Estimates Distribution Loss Estimation Error – Incorrect Loss Factors Theft Un-metered Energy – Company Use •Revenue Annually @$2 Billion Profiled Loads 55% TOU Loads 45% 9 Load Research Program Assessment Observations • Age of the Samples - Over 9 Years Old • Significant Data Loss – 47% for BECO • Sample Sizes All Less than 30 Per Stratum • Sample Bias Caused By Installation Procedures – Over 50% Indoor Meters – “Don’t Really Need THOSE Meters” • Relative Precisions Averaged from 20-30% 10 Sample Augmentation Plan We Need To Consider Options To Improve The Sample Results Until The New Samples Are Installed. Techniques • Ratio Estimation – Takes Advantage Of The Correlation Of The Variable Of Interest With Another Variable To Increase Precision • Post Stratification – Can Be Used Within A Sample Study, But Can Also Be Utilized To Post Stratify Transferred Data From Another Utility With Similar Population Characteristics • Load Data Transfer – A Transfer of Load Data From One Distribution Company To Another Distribution Company For Certain Samples. Can Utilize Post Stratification To Fit The Transferred Data To The Sample 11 Preliminary Test Results • 2 Primary Goals – To Improve the Current Sample Results in the Interim – To Use Ratio Estimation as an Expansion Methodology • Tested 2 Samples for July 2005 – Commonwealth Residential R-1 – Commonwealth Small General Service G-1 • Looked At Relative Precisions – Ratio Estimation Versus Mean-Per-Unit – Ratio Estimation Versus Ratio Estimation Post-Stratified • Did Not Look At Load Data Transfer, Yet 12 Residential R-1 Relative Precisions MPU Vs. Ratio Estimation Mean Per Unit vs. Ratio Estimation Commonwealth Residential General Service R1 July 2005 40.0 Improved Average Relative Precision From 16.33 to 13.10 35.0 30.0 Precision % 25.0 20.0 15.0 10.0 5.0 0.0 1 25 49 73 97 121 145 169 193 217 241 265 289 313 337 361 385 409 433 457 481 505 529 553 577 601 625 649 673 697 721 MPU Ratio 13 Small GS G-1 Relative Precisions MPU vs. Ratio Estimation Mean Per Unit vs. Ratio Estimation Commonwealth Small General Service G1 July 2005 40.0 Improved Average Relative Precision From 25.72 to 11.17 35.0 30.0 Precision % 25.0 20.0 15.0 10.0 5.0 0.0 1 25 49 73 97 121 145 169 193 217 241 265 289 313 337 361 385 409 433 457 481 505 529 553 577 601 625 649 673 697 721 MPU Ratio 14 Post Stratification Assignment Distribution Residential R-1 Small General Service G-1 Strata Original Assignment Post Stratified Assignment Strata Original Assignment Post Stratified Assignment 1 20 21 1 28 36 2 25 23 2 24 21 3 25 30 3 27 26 4 25 21 4 28 24 15 Residential R-1 Relative Precisions Ratio Estimation Post-Stratified Ratio Estimation vs. Ratio - Post Stratified Commonwealth Rate Residential General Service R1 July 2005 40.0 Improved Average Relative Precision From 13.10 to 10.97 35.0 30.0 Precision % 25.0 20.0 15.0 10.0 5.0 0.0 1 25 49 73 97 121 145 169 193 217 241 265 289 313 337 361 385 409 433 457 481 505 529 553 577 601 625 649 673 697 721 Ratio Ratio-Post Stratified 16 Small GS G-1 Relative Precisions Ratio Estimation Post Stratified Ratio Estimation vs. Ratio - Post Stratified Commonwealth Small General Service G1 July 2005 40.0 Improved Average Relative Precision From 11.17 to 9.83 35.0 30.0 Precision % 25.0 20.0 15.0 10.0 5.0 0.0 1 25 49 73 97 121 145 169 193 217 241 265 289 313 337 361 385 409 433 457 481 505 529 553 577 601 625 649 673 697 721 Ratio Ratio-Post Stratified 17 Relative Precision Ranges Reduction in Variability in the Precision Values Residential Small General Service Mean Per Unit 10.4 - 34.3 17.9 - 38.3 Ratio 07.9 – 29.2 05.9 – 30.8 Ratio Post Stratified 06.5 – 20.2 06.5 – 25.0 18 New Sample Designs Objectives • Achieve Design Accuracy Of +/- 5% At The 95% Confidence Level For All Hours Of The Year • Stay Within Budget Constraints • Use Stratifying Variables That Will Achieve The Targeted Accuracies For All Hours Of The Year • Utilize A 100% (Certainty) Stratum In The Design Process When Feasible • Have Longevity Of At Least Five Years • Represent Target Populations Optimally • Have A Minimum Of 30 Sample Points In Each Stratum • Minimize Sample Bias In The Installation Procedures 19 Current and Proposed Program • • • Do We Need A Separate Sample For R-1 Water Heating (WH)? No Do We Need A Separate Samples For R-2? Yes Do We Need Separate Samples For BECO’s G-2 Commercial And G-2 Industrial? No • Proposed • Current Samples BECO COMM 109 R-1-WH 102 102 R-2 99 99 R-3 138 90 140 TOTAL R-1 64 349 292 G-0 G-1 0 115 120 124 359 G-2 0 G-2C 90 90 G-2I 90 90 743 310 328 1381 13 Sampled Rates – @2002 Sample Meters • 100 CAMB New Samples 11 100% Monitored Rates (TOU) – @4400 Meters • @225 Interchange Meters BECO COMM CAMB TOTAL R-1 220 176 160 556 R-2 82 88 80 250 R-3 161 165 165 491 G-1 200 165 198 563 G-2 160 823 160 594 603 2020 20 Program Objectives • Have a quality, compliant, and maintainable load research program • Lower risks of disputes in regulatory proceedings and in ISO settlements • Minimize loss of data • Improve the predictability of costs of operations, maintenance, & future capital expenditures 21 Load Research Roadmap Load Research Re-Engineer To Monthly Production Use Same Month Profiles For Settlement Build Sample Tracking System Recast Analysis Redesign and Install New Samples 22 Next Steps • Sharing Performance Metrics – Between Metering, Meter Data Management, & Load Services – Sample Installation Rate; Data Retrieval Rate; Interval Meter Check • • • • • • Installing The New Samples – 2006 and 2007 Build a Sample Tracking System Improve Data Validation Procedures Re-Engineer Load Profile Production to Monthly For Settlement Implement A Process Of Load Study Cycling And Replacement Cross Training – Cloning the Old Fashion Way 23 Now, Who’s Driving the Bus? Data Heaven Load Research 24