PNNL ISU Project Report Integrated Retail and Wholesale (IRW) Power System Operation with Smart Grid Functionality 18 July 2011 Pacific Northwest National Lab Richland, WA Last Updated: 15 July 2011
Project Directors: Leigh Tesfatsion (Prof. of Econ, Math, & ECpE, ISU) Dionysios Aliprantis (Litton Industries Ass t Prof. of ECpE, ISU) David Chassin (Staff Scientist, PNNL/Department of Energy) Research Assoc s: Dr. Junjie Sun (Fin. Econ, OCC, U.S. Treasury, Wash, D.C.) Dr. Hongyan Li (Consulting Eng., ABB Inc., Raleigh, NC) Research Assistants: Huan Zhao (Econ PhD student, ISU) Chengrui Cai (ECpE PhD student, ISU) Pedram Jahangiri (ECpE PhD student, ISU) Auswin Thomas (ECpE M.S. student, ISU) Di Wu (ECpE PhD student, ISU) IRW Project: Integrated Retail/Wholesale Power System Operation with Smart-Grid Functionality Current Government & Industry Funding Support: PNNL/DOE, the Electric Power Research Center (an industrial consortium), and the National Science Foundation Industry Advisors: Personnel from PNNL/DOE, XM, RTE, MEC, & MISO
Retail & Wholesale Power System Operations 3
Meaning of Smart Grid Functionality? For our project purposes: Smart-grid functionality = Market design & resource enhancements permitting more responsiveness to the needs, preferences, and decisions of retail energy consumers. Examples: Introduction of advanced metering and other technologies to support flexible dynamic-price contracting between suppliers ( Load-Serving Entities ) and retail energy consumers integration of distributed renewable energy resources, e.g., consumer-owned photovoltaic (PV) panels 4
Key Project Research Topics Dynamic retail/wholesale reliability and efficiency implications of introducing price-sensitivity of demand for retail customers as realized through Top-down demand response (e.g., emergency curtailment) Automated demand dispatch Price-sensitive demand bidding by demand-side resources Dynamic retail/wholesale effects of increased penetration of consumer-owned distributed energy resources, such as photovoltaic (PV) generation and plug-in electric vehicles Development of agent-based algorithms for smart device implementation (e.g., smart HVAC systems) 5
Primary Project Tool: The IRW Power System Test Bed An agent-based computational laboratory Culture dish approach to complex dynamic systems Permits systematic computational experiments Permits sensitivity testing for changes in physical constraints (e.g., grid configuration), market rules of operation, and participant behavioral dispositions Seams empirically grounded test beds (AMES/GridLAB-D) Market rules based on business practices manuals for restructured North American electric power markets Realistically rendered transmission/distribution networks Retail contracting designs based on case studies (e.g., ERCOT) and pilot studies (e.g., Olympic Peninsula 2007) Open source software release planned. 6
IRW Power System Test Bed: AMES & GridLAB-D Bilateral Contracts x x Wholesale AMES ISU Team Seamed Retail GridLAB-D DOE/PNNL Team 7
IRW Project Year-1 Tasks Task 1) Extension of AMES to include a fully operational two-settlement system (day-ahead forward market & real-time balancing market operating in tandem) 2) Development of basic IRW Power System Test Bed (V1.0): Seaming of AMES (wholesale) and GridLAB-D (retail) test beds with a focus on household loads 3) Development of an initial test case: Resident-occupied house with a Heating-Ventilation-Air-Conditioning (HVAC) system 4) Modeling and implementation of a smart HVAC module to reflect the resident s optimal comfort/cost trade-offs Status Done Done Done In progress 8
IRW Project Year-1 Tasks Continued Task 5) Design & running of electricity contract experiments to study IRW system performance under alternative retail dynamic-price contracts between LSEs & consumers 6) Development of a PV module calibrated by means of a PV experimental lab set-up for implementing retail PV generation in the IRW test bed 7) IRW system impacts of reactive power support by distributed retail PV generation Status In progress In progress In progress 8) IRW system impacts of aggregator-controlled PEV In progress 9
Task 1: Extension of AMES Typical Day-D Market Operator (ISO) Activities 10
Task 2: IRW Test Bed (Version 1.0) Seams AMES (wholesale) & GridLAB-D (retail) with a focus on load from households with HVAC and appliances 11
Task 2: IRW Test Bed Development Seaming of AMES (wholesale) & GridLAB-D (retail) implemented via MySQL database server and a data management program. 12
Task 2: IRW Test Bed Development DMP (Data Management Program): 1. Prepare input files for running GridLAB-D 2. Collect simulation results 3. Receive price signals from the database 4. Send load flow data to the database 13
Task 2: IRW Test Bed Development GridLAB-D 1. Run the simulation with prepared input files 2. Record the daily load profile 14
Task 2: IRW Test Bed Development SQL Database Program: 1. ODBC connection 2. Two tables - price and load 3. Coordinate the data flow 15
Task 2: IRW Test Bed Development Data Transmission Between AMES & GridLAB-D: 1. GenCos and LSEs submit supply offers and demand bids on day D for the Day-Ahead Market on day D+1 2. ISO performs Day-Ahead Market (DAM) clearing 3.DAM price solutions (ex ante LMPs) sent to database, and database transmits them to GridLAB-D 4.Energy consumers with price-sensitive demands determine day D+1 loads based on DAM price signals 5.Loads transmitted to ISO, where they determine realtime market prices (ex-post LMPs) on day D+1 6.Deviations between DAM/RTM loads on day D+1 affect demand bids by LSEs on day D+1 for DAM on D+2. 16
Task 2 Test bed operation 17
Task 3: Initial Test Case Development Resident-occupied house with HVAC system and background load 18
Task 4: A Smart HVAC Controller (Heating-Ventilation-Air-Conditioning) Inputs include: Preferences of a household resident trying to achieve optimal daily trade-offs between comfort and costs Structural home attributes (e.g., square footage & insulation level) Electricity prices (e.g. fixed regulated price, marketbased LMPs) Other forcing terms (outdoor temp, solar radiation, control actions) State equations for a two-dimensional state vector x(t) consisting of (1) Indoor air temp T a (t) and (2) indoor mass temp T m (t), e.g. for furniture and walls. 19
Task 4 Continued. An ETP Modeling of a Smart Residential HVAC System The house resident has a bliss temp (e.g., 77 o F) Using a discretized form of ETP state equations, HVAC sets its status levels (from cooling to heating) to achieve optimal comfort/cost trade-offs for the resident over time, conditional on forecasted prices, outdoor temp s, & other forcing terms. Status levels are derived via dynamic closed-loop control. 20
Task 5: Initial HVAC Experiments. Resident cares only about comfort. HVAC status updated every 5 minutes. 21
Resident has a balanced concern for comfort and cost. HVAC status updated every 5 minutes. 22
Resident cares most about cost. HVAC status updated every 5 minutes. 23
Next Steps for Task 5 Explore alternative forms of dynamic pricing contracts between LSEs and retail consumers Explore alternative ways for LSEs to forecast price-sensitive loads Simulate retail load from multiple consumer types Permit households to have PV panels. Explore reactive power support from PV generation 24
Task 6: Photovoltaic Generation Modeling Modeling, Analyzing and Control of Large-Scale Distributed PV Generation 1. Develop PV generation model 2. Apply practical weather model in analysis 3. GridLAB-D serves as the experiment platform 4. Utilize controllable load to mitigate intermittency 5. Study the impact of PV generation on wholesale energy and ancillary service markets 25
Task 6: Photovoltaic Generation Modeling Solar Energy Conversion Nellis SEGS Photovoltaic (PV) power plant 14 MW capacity with 22% capacity factor 72,000 solar panels; $100 million; build cost $7.14/Watt 0.57 km 2 land use; 5.2 W/m 2 Concentrated Solar Power (CSP) plant World s largest solar energy plant 354 MW capacity with 21% capacity factor 936,384 mirrors; build cost of a CSP $2.5~$4/Watt 6.5 km 2 land use; 11.4 W/m 2 By 2009, U.S. had 1.25 GW PV capacity and 432.5 MW CSP capacity 26
Task 6: Photovoltaic Generation Modeling Characteristics of Distributed PV Generation Static system without inertia Fast variation of power output with large amplitude High repetitive output pattern due to sun diurnal cycle Large land use Ota-Pal town pilot project in Japan 2.16 MW, 1.65 km 2 land use 27
Task 6: Photovoltaic Generation Modeling Challenges and Approaches C: Solar radiation data with high temporal and spatial resolution A: Generate realistic cloud pattern to obtain the synthesized data C: Energy conversion model to convert the received radiation to electric power, considering environmental factors A: Establish a small-scale PV generation system and validate the MPPT surface model C: Interdependency between feeder geographic topology and distributed PV generation A: Superimpose the moving cloud pattern on the taxonomy feeder topology map 28
Task 6: Photovoltaic Generation Modeling Generate Cumulus Cloud Pattern Cumulus cloud Fractal surface 20% coverage 40% coverage 29
Task 6: Photovoltaic Generation Modeling Generate Solar radiation Components of solar radiation Global radiation on a tilted surface Horizontal beam radiation Conversion factor from horizontal beam to tilted beam radiation Horizontal diffuse radiation Tilted angle in degree 30
Task 6: Photovoltaic Generation Modeling Determine the Radiation Components Theoretically compute the clear sky horizontal beam radiation Read in the hourly horizontal diffuse radiation from the TMY data Synthesize the radiation along with the shading condition determined by the cloud pattern 31
Task 6: Photovoltaic Generation Modeling Simulated Solar Radiation 3.5 Km by 3.5km area Define the area installed with PV systems Determine the PV system density Obtain the total available solar radiation in the defined area 32
Task 6: Photovoltaic Generation Modeling Simulated Solar Radiation Single point radiation W/m 2 Average radiation second 33
Task 6: Photovoltaic Generation Modeling PV Model PV panel circuit 34
Task 6: Photovoltaic Generation Modeling Power Curves of PV Panel 35
Task 6: Photovoltaic Generation Modeling PV Energy Conversion Model Calculated result for KC200GT module Experiment result for ASP 140 Module 36
Task 6: Photovoltaic Generation Modeling PV Experimental Setup LICOR 200 pyranometer RTD sensor Hall effect current sensor Kyocera 135 W PV panel GE temperature and humidity sensor NI ENET 9205 sampling card 37
Task 6: Photovoltaic Generation Modeling Feeder and PV Installation 38
Task 7: Reactive Power Support by PV The objective of the VAR problem is optimal generation of reactive power from local resources that leads to: Minimizing power distribution loss Increasing maximum transfer capability Decreasing transmission line congestions Reactive power can be supplied from: Static resources (capacitor banks) Dynamic resources: any inverter-based equipment such as PV systems, fuel cells, or micro-turbines 39
Reactive Power for House in California 40
Real Power Loss Case 1: Voltage regulator set-point at 1.0417 (pu) and no reactive power injection capability Case 2: Voltage regulator set-point at 1.0417 (pu) and reactive power injection capability Case 3: Voltage regulator set-point at 1.0278 (pu) and reactive power injection capability R1-12.47-4: Representative of a heavily populated suburban area in San Francisco, CA. This is composed of 652 houses, and heavy commercial loads. Around 1 KW reduction on average 41
Total Power from Substation (Cases 1 & 2) 42
Conservative Voltage Reduction In case 2, voltages increase due to reactive power injection Demand is increased (voltage dependent loads) Loss reduction benefit is neutralized So voltages at substation should be decreased based on the CVR concept! CVR: operating a distribution system in the lower half of the acceptable voltage range (114 120 V) Leads to significant energy savings ANSI Standard Voltage Range 114-126 V V O L T A G E 126 125 124 123 122 121 120 119 118 117 116 115 114 122 (V) National Average Customer Voltage CVR area Upper Half Lower Half 43
Average Quantities for 3 Cases 44
Potential Impacts of Aggregator-Controlled PEVs on Distribution Systems Di Wu dwu@iastate.edu home.eng.iastate.edu/~dwu/ 45
Background and assumptions PEV load: the same as a traditional (i.e., uncontrollable) load being served whenever a PEV is plugged in managed charging during off-peak hours by PEV aggregators Aggregators considered herein: wish to maximize their profits from energy trading 46
Assessment tools Distribution feeder under study: R1-12.47-4 Travel pattern: obtained from 2009 National Household Travel Survey Simulation Platform: GridLAB-D 47
Evaluation method Three PEV penetration Levels: 10%, 25%, and 50% Spatial diversity and temporal diversity Stochastic analysis: 100 Monte Carlo simulations for each penetration level 48
Spatial diversity In each simulation, the number of vehicles per residence is randomly generated based on the probability mass function. Veh/HH 0 1 2 3 4 5 6 Prob. 0.087 0.323 0.363 0.144 0.053 0.019 0.01 The probability for a random vehicle to be a PEV is (by definition) the PEV penetration level. 49
Temporal diversity Considered control strategy: Maximize the profit from energy trading using the forecast information about wholesale electricity price PEV load repository: 141011 daily PEV load curves developed using the simulation method with the travel pattern obtained from the 2009 NHTS, and certain assumed drivetrain parameters and charging circuits The daily power consumption of individual PEVs in the distribution feeder is randomly selected from the repository. 50
Preliminary results Simulation results: asset loading customer voltage loading between phases system losses 51
Loading The daily peak of the power consumption of this feeder is below 60% of its rating (5334 kva) except several days of a year. Due to the large amount of remaining capacity, adding PEV load during off-peak hours will not overload either substation transformer or primary trunks that are far away from the customer end. Average apparent power consumption at substation 52
Loading (continued) The assets closest to customers are sensitive to overloading because: no redundant capacity is reserved for reliability purposes they do not benefit much from spatial and temporal diversity Average apparent power consumption for transformer 14 53
Voltage level The ANSI standard C84.1: 5% from the nominal value (114~126 V) Without PEVs and under mild penetration levels: no violation 50% penetration level: low-voltage violations Violation detected: 9 out of 100 simulations Minimum voltage within the distribution system in one simulation 54
Phase unbalance Actual distribution feeders are typically unbalanced in nature Aggregator-controlled PEVs could exacerbate this issue during the charging period Average apparent power consumption in each phase in one simulation 55
Losses The added PEV load will increase the system s ohmic losses. For the feeder under study, it is found that the distribution losses are less than one percent of the overall power consumption. The additional losses from charging PEVs are negligible. The losses could be higher for other distribution systems. 56
Recommendations The negative impacts could be mitigated by modifying the control strategy incorporating the performance metric of interest (e.g., asset loading, customer voltage, or system losses) into the optimization constraints or the objective function. The desired power consumption of each individual PEV from the optimization results can be realized by modulating the charging power magnitude or adjusting the charging duration. However, this can reduce the aggregators profits from energy trading. Without suitable incentives (or regulations), aggregators (especially those who do not have an interest in maintaining the distribution system s reliability) will have no motivation to modify their control strategy. The associated impact in battery lifespan from frequent charging cycles is another concern. 57