DIMACS, Rutgers U January 21, 2013 Michael Caramanis

Power Market Participation of Flexible Loads and Reactive Power Providers: Real Power, Reactive Power, and Regulation Reserve Capacity Pricing at T&D Networks DIMACS, Rutgers U January 21, 2013 Michael Caramanis mcaraman@bu.edu

Outline How can Flexible Loads Provide Fast Reserves How do Distribution Network Injections Differ From Transmission System Bus Injections? Current Market Bidding Rules Motivate Flexible Distributed Loads to Exercise Strategic Behavior Resulting in a Hierarchical Game Conditions for Hierarchical Game to Converge Revised Bidding Rules Remove Strategic Behavior Incentives and Allow ISO/ DNO to Clear Market in a Socially Optimal Manner Detailed Distribution Market Pricing Real and Reactive Power 2

Generation and Demand Share Functional Characteristics that are Key to the Efficient and reliable Operation of the Electricity Grid Characteristic Generation Demand Dispatchability- Schedulability Low/Med/High Flexibility Low/med/high Forecastaility Low/Med/High Wind, Run of Riv /Neuclear, L.E.P,/ HydroFossil No Ramp steady output e.g., nucl, min gen, start up cost and delay/ Inertia and medium storage/high ramplow inertia large storage Wind, Solar. RoR Hydro/reliable fossil/unreliable fossil Voltage Control Synchronous Generators with dynamic Var compensators, DC-AC Converters Capacity Loads, dependent on env. e.g.,light/ Ind. Energy Loads Aluminum idle/schedulable production of electr. energy intensive storable products (gas liquif.) Thermal or work inertia (Allum. Smelter)/Enegy Demand with small storage to capacity ratio (HVAC)/ Large storage to capacity ratio (ice, molten salt, batteries in Evs) Inflexible loads (lighting cooking)/weather dependent/scheduled loads Distributed Power Electronics accompanying EVs. HVAC, Roof top PV. 4

Flexible Loads Require Energy by some deadline => Capable of Regulation Reserves 1 D: 10/1/11 0.8 0.6 S: 9/1/10 0.4 0.2 0-0.2-0.4-0.6-0.8 Instance of PJM Regulation Signal, y(t). Note Constant Average over relatively 5 short period of Time

Today Generating Units are Only Reserve Providers Example of Generator providing Super Fast Reserves: Frequency control and 40MW of Secondary Reserves Frequency Control Secondary Reserves 320MW50MW Source: Courtesy of EnThes Inc., March 2007 6

Outline How can Flexible Loads Provide Reserves How do Distribution Network Injections Differ From Transmission System Bus Injections? Current Market Bidding Rules Motivate Flexible Distributed Loads to Exercise Strategic Behavior Resulting in a Hierarchical Game Conditions for Hierarchical Game to Converge Revised Bidding Rules Remove Strategic Behavior Incentives and Allow ISO/ DNO to Clear Market in a Socially Optimal Manner Detailed Distribution Market Pricing Real and Reactive Power 7

Distribution Network Low Voltage Bus Marginal Cost Based Dynamic Prices (DLMP) Result from Augmenting Transmission System High Voltage Prices (LMP) by Marginal cost of: Line Losses, Reactive Power, Voltage control, Transformer Life Loss LMP at Bus n HV, Bus n LV, n(i) LV, n(k) DLMP at n(i)=m n(i) (LMP at n)+ Where m n(i) =(1+ML at n(i)) 8

Examples of Flexible Loads: State Dynamics Determine Preferences Distributed PHEV Charging t t1 t ˆ t n( i) n( i) n( i) n( i) F F F F F j j j j j x x d x dep time ni () 0 Centralized Pumped Storage Hydro Units psh t psh t1 p t g t r R t n( p) n( p) n( p) n( p) n( p) n( p) n( p) n( p) psh x x p g g x x 0 psh 24 n( p) n( p) 10

Strategic Flexible PHEV Load Behavior E t R t t Fj t Fj, R t,,, n λ n mn ( i ) dn( i ), dn( i ), j, t jt st.. min F F, R t R t Fj, R t Fj t Fj t n( i) n n( i) n( i) n( i) t t1 t t n( i) n( i) n( i) n( i) t j t n( i) n( i) t E t Fj t n( i) n n( i) { E [ m λ d m λ d ] U ( x )} F F F F j j j j j x x d d j d λ ˆ e.g., state dyn of EV dem. Fj t Fj, R t t [ d d ] C Local Constraint n( i) n( i) n( i) up/dn nature of Reg. Res. ˆ 11

Use of Current Bidding Rules to Self Schedule Bid Energy Fj t* Fj, R t* dn( i) dn( i) at a very high price Fj, R t* t E t dn( i) mn ( i) n Bid Energy 2 at energy price ~ λ and Regulation Service Rate at 0. 12

Using the Current Bidding Rules. Bids described on the previous slide, induce the ISO/DSO to almost surely Schedule Energy and Reserves to the * values, and thus effectively self dispatch. c t t R t n( i ) n( ) n( ) d, g, g, t ti,, st.. max ( c c R u d c g r g ) t t t t t t n( i) n( i) n( ) n( ) n( ) n( ) t Fj t * c t g d d n( ) n( i) n( i) n( ) n( i), j n( i) n() i ni () * 2 ( Fj t dni () d ) 2 j R t Fj, R t* gn( ) dn( i) n( ) j, n( i) 0, t c t E, u t n() i n, c t R u t Lossni () R t n ni () & other capacity and ramp constr. for conv. gen. and dem. 13

Hierarchical Game Dynamics Undamped Oscillations when Flex Load Updates Clearing Price Estimates Myopically to Most Recent ex-post ISO/DMO value Convergence to stable equilibrium when Flex Load Updates Clearing Price Estimates Factoring in History, for example sets them Equal to their Time Average 15

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 117 121 125 129 133 137 141 145 149 153 157 161 165 169 173 % Convergence 5.00% UPQB LMP Convergence by Iteration (No Congestion) 4.00% Base Case Discrete Smooth Case Quadratic Case 3.00% 2.00% 1.00% 0.00% Iteration 16

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 117 121 125 129 133 137 141 145 149 153 157 161 165 169 173 % Convergence 5.00% 4.00% UPQB LMP Convergence by Iteration (With Congestion) Base Case Bus 1 Discrete Smooth Case Bus 1 Quadratic Case Bus 1 3.00% 2.00% 1.00% 0.00% Iteration 17

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 117 121 125 129 133 137 141 145 149 153 157 161 165 169 173 % Convergence 3.00% Step Size Impact on UPQB Convergence Discrete Smooth Case Bus 1 (Steepest Descent) 2.50% 2.00% Discrete Smooth Case Bus 1 (Averaging) Quadratic Case Bus 1 (Steepest Descent) Quadratic Case Bus 1 (Averaging) 1.50% 1.00% 0.50% 0.00% Iteration 18

$/MWh 0.20 (UPQB LMP - TCCB LMP) by Iteration 0.10-1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Hour (0.10) (0.20) (0.30) Bus 1 Iterations 148, 151, 154, etc Bus 1 Iterations 149, 152, 155, etc Bus 1 Iterations 150, 153, 156, etc (0.40) (0.50) 19

Under New Bidding Rule allowing Flex Load to Express True Utility, ISO/DNO will Solve c t t R t Fj t Fj, R t W t n( i ) n( ) n( ) n( i ) n( i) n( ) d, g, g, d, d, g t,, i, t,, t t t R t Fj t Fj t n( ) n( ) n( ) n( ) n( i) n( i) t Fj t c t g n( ) d n( i) dn( i) n( ) n( i), j n( i) ni () t ni () t n() i [ t c t n( i) n( i) c g r g U ( x )] s. t. n() i 2 F max j c ( d d ) j c u 0, t 2 E, x d t n t Fj t c t g n( ) d n( i) dn( i) n( ) n( i), j n( i) n() i ni () 2, ( Fj t c t E x t dn() i d ) 0, t n() i n 2 j x x d ˆ n( i) Flex Load Dynamics Fj t Fj t1 Fj t Fj t n( i) n( i) n( i) n( i) & other Local ni ( ) and System constraints 21

Complex Bid ISO/DNO Market Clearing Achieves Hierarchical Game Equilibrium Theorem: First order Optimality Conditions Complementary Slackness, and Feasibility Conditions Coincide if we combine Hierarchical game problems and compare to ISO/DSO problem, Except when Flex loads dominate in a Distr. Location (competitive assumption fails?) F j d t n( i ) 1 ISO ( d d ) ( c t F j t ) E t n( i) n( i) n() i n j R t t Fj, R t t, j t, j t, nn( i) dni ( ) n( i) n( i) ni ( ) 0 22

Additional term in ISO/DNO problem F j d t n( i ) 1 ISO ( d d ) ( c t F j t ) E t n( i) n( i) n() i n j R t t Fj, R t t, j t, j t, nn( i) dni ( ) n( i) n( i) ni ( ) 0 Becomes negligible, i.e. 0 as F j d F j, R d t ni () d t c t n( i) n( i) 0 or as the relative size of flex load Reg. Res. Transactions 0 23

$/MWh $/MWh Impact of Competitiveness Assumption 0.20 100 0.10 50-1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Hour - (50) (0.10) (0.20) UPQB Price Spread - TCCB Price Spread UPQB IFL Costs - TCCB IFL Costs (100) (150) (0.30) (200) (0.40) (250) 24

Conclusion Flexible Loads at Distribution Level may participate in Expanded ISO/DNO Centrally Cleared Power Market bringing significant benefits, particularly w.r.t. Sustainable Renewable Generation Integration to the Grid Expanded ISO/DNO-Operated Power Market Clearing is Practical from Information and Computational Tractability Point of view. Inclusion of Other Important Distribution Network Costs, such as Reactive Power Compensation and Voltage Control is also Practical. 25

Ex. of Var. Speed HVAC - PV Collaboration: Action in the small by Distr. Flex. Loads 27

Load and Other Resources May Participate fully in Future Distribution Markets: Centralized Dispatchable Generation Distributed Consumers Centralized Non-Dispatchable Generation e.g., Wind Parks Market Distributed Generation Centralized Prosumers (e.g., cogeneration) Distributed Prosumers Current market: Only Centralized generation is a full market participant. All others communicate their capabilities and needs without feedback and response Centralized Dispatchable Generation Distributed Consumers Centralized Non-Dispatchable Generation e.g. Wind Parks Market Distributed Generation Centralized Prosumers(e.g., cogeneration) Distributed Prosumers Future Market: Many more non-dispatchable cedtralized generators, distributed generators and prosumers. On the distributed side, feedback renders Non-Dispatchable generation and distributed consumersproducers (prosumers) full market participants 28

Example of a Radial Distribution Network: One Medium Voltage Branch is Shown with three feeders, each with three building loads. Substation is the Slack Bus 29

Distribution Market Problem formulation: Minimize Utility Loss, Real and React. Power Cost (incl Losses), Asset Life Loss, and Volt. Control Cost s.t. Load Flow, Capac., Volt. Magnitude Const Minimize tm c P u P P C C Q c V 2 2 f f gi di di m, n m, n,, (1), (1) 1 i M M P P P V g m m m m b b m i i f f Subject to mn, mn, 2 30

Distribution Market Problem formulation (cont.) 15000 15000 f exp, f bm, H 383 273 fbm, k k S, f H A 2 f 1, f 2, f b, m b, m b, m b, m b, m V V V, b A b b b 0 b, m 31

Distribution Market Benefits Marginal Losses Reflected in DLMPs=>Demand Adaptation Reactive Power Pricing motivates Dual Use of Power Electronics whose presence is expected to Become Ubiquitous while accompanying Distributed Clean Generation (e.g., PV) installations and Flexible Loads (e.g., EVs, Heat Pumps) Marginal Voltage Control Cost Reflected in DLMPs =>Demand Adaptation Distribution Asset Degradation Marginal Costs Reflected in DLMPs =>Demand Adaptation Full Distr. Net Price Unbundling =>Distr. Net Rent 32