Load Modeling for Power System Studies

Transcription

1 Load Modeling for Power System Studies WECC Load Modeling Task Force Dmitry Kosterev Bonneville Power Administration Transmission Planning

2 Timescale of Interest Our primary interest is the fast dynamic response of loads, not projections of future demand. cycle minute day 10 years second hour year Time, Log scale

3 o o o o o o Motivation: Delayed Voltage Recovery Voltage (pu) Reality Seconds 30 seconds Simulations using old load model

4 Motivation: Oscillation Damping Malin Voltage, August Event Voltage [kv] Time [sec] Inter-area power oscillations are more pronounced during on-peak summer loads

5 WECC-Wide Effort Supported by all major WECC utilities and grid operators, including staff time from: Bonneville Power Administration Pacific Gas and Electric Company Southern California Edison California ISO Arizona Public Service Idaho Power Company Puget Sound Energy Northern California Power Agency WECC Staff Additional research funding is provided by California Energy Commission (CEC) Public Interest Energy Research (PIER) Transmission Reliability Program (TRP)

6 Project Deliverables Load Model Structure Load Model Data Load Model Sensitivity and Validation Studies System Performance Studies and Develop Mitigation Plans

7 Load Model Structure Today 115-kV 230-kV M Static 20% Proposed 115-kV 230-kV Transformer Feeder Equivalent M M M Load Model Components AC Air-Conditioners Static

8 Residential Air-Conditioners Represent significant part of system load during on-peak summer loads (estimated up to 33%) Major cause of slow voltage recovery observed in Southern California and Desert Southwest Contributing factor to poorly damped power oscillations on California Oregon Intertie

9 Residential Air-Conditioners Compressor Motor R 240 V AC S Thermal Relay C

10 Residential Air-Conditioners Compressor Motor

11 Air-Conditioner Models Conventional 3-phase motor models do not represent dynamic behavior of 1-phase air-conditioner compressor motors. BPA, SCE, EPRI tested more than 30 A/C units Project team researched and developed new models: Positive sequence phasor model Bernie Lesieutre (LBNL) and John Undrill (on contract with BPA) preferred option Performance Model BPA and PNNL second choice Hybrid Model SCE / EPRI Solutions AC Model and composite load model in PSLF 17.0

12 Load Model Data - I Distribution Equivalent Data Substation transformer data actual data estimated based on an expected load factor Feeder equivalent data estimated to meet distribution network performance at design condition (methodology developed by PG&E)

13 Load Model Data - II Load component database Model parameters for various electrical end-uses equipment testing manufacturer s data Component granularity define groups that exhibit different behavior (topological, electrical data, inertia, driven load) Model parameter aggregation and sensitivities Mapping between end-users and models

14 Load Model Data - IIIa Load composition information Substation Load Class Data Residential Commercial: -office -retail -warehouse - Industrial Agricultural: -irrigation -processing Confirm with utilities Separation End-User Composition: Installed Capacity Chiller Direct expansion A/C Refrigeration Residential 1-ph A/C... Florescent Lighting Resistive Time Date Temp Rh End-User Composition: Operating Capacity Chiller Direct expansion A/C Refrigeration Residential 1-ph A/C... Florescent Lighting Resistive Validate total with SCADA

15 Load Model Data - IIIb Rules of Association / Model Aggregation End-User Composition: Operating Capacity Chiller Direct expansion A/C Refrigeration Residential 1-ph A/C... Florescent Lighting Resistive Model Aggregation Load Model Components Motor A 3φ Motor B 3φ Motor C 1φ... Electronic Resistive

16 Load Model Validation Studies Challenges of load model validation: Load composition is constantly changing Large disturbances may not occur during loading conditions of interest Most disturbances are not large enough to extrapolate the load behavior for most planned for disturbances Lack of dynamic measurements Validate the process of load model data construction by validating data points

17 Load Model Validation Studies New Load Model Reproduces Delayed Voltage Recovery Phenomenon Voltage (pu) Seconds reality simulations Time( sec ) 0.6 vbug VALLEYSC [COI 4217MW][PDCI 2002MW][Path 15 62MW][BORAH W 1286MW] [N JOHNDAY 6015MW][LC GEN 2839MW][UC GEN 0MW][MP-SL 187MW] [S.LUGO 5083MW][N.LUGO 1586MW][N.SONGS 798MW][S.SONGS 492MW] [SCIT 12922MW][WOR 5418MW][EOR 4073MW][IPP DC 1648MW][PATH MW] MOST RECENT VERSION OF THE MDF USED. ALL COMMENTS FROM THE TSS & OC REVIEW INCLUDED. PDCI IS LIMITED TO 2000 MW DURING SUMMER 2004 DUE TO CONSTRUCTION valley04.chf AT SYLMAR. Add Oak Valley Load to Vista 115 (63%) + Devers 115 (37%) = 187 MW SCE P/Q = 25:1 Palo Verde - Devers 500 kv Line at 5 Section Model LADWP Load = 6130MW, MVAR; SDG&E Load = 4936MW, 628 MVAR IID Load = MW, MVAR Page july 24 valley fault Thu Dec 07 13:00:

18 Load Model Sensitivities Need to understand and characterize load model sensitivities No model is complete until the uncertainties are characterized John Hauer Extreme sensitivity in load model parameters may indicate the we are close to the edge

19 System Performance Studies 3-hase fault Hassayampa Palo Verde Normal clearing Time( sec ) 0.2 a vbus PNPKAPS b vbus PNPKAPS c vbus PNPKAPS d vbus PNPKAPS Baseline simulation 20% of a/c tripped by UV relay 30% of a/c tripped by UV relay 60% of a/c tripped by UV relay TASMO MODEL; OUTPUT GENERATED :52:05 SWINGBUS 1520 FOR FC : :17:4F Current file selected from 4 different files a:ac000.chf b:ac020.chf c:ac030.chf d:ac060.chf Page 1 ltage recovery in arizona Wed Apr 11 01:43:

20 Where are we? Develop Models Quantify Impact Develop Solutions Monitor Performance