Inexpensive Ancillary Service from Automated Load Tuning Prabir Barooah Mechanical and Aerospace Engineering, University of Florida in collaboration with Prof. Sean Meyn, He Hao, Yashen Lin, Tim Middelkoop, and PPD@UF
Operating the grid is a control problem,.. (non- contingency) balancing currently achieved by > feedforward : scheduling and dispatching generation to match predicted load > feedback : frequency regulation Power (GW) 15 10 5 Day-ahead forecast Hourly schedule Power (GW) 15 10 5 Hour ahead forecast Total dispatch Residual Load following Power (GW) 15 10 5 Realized load Total dispatch Regulation 00 01 02 03 04 Time (h) 00 01 02 03 04 Time (h) 00 01 02 03 04 Time (h) slow, large generators: energy small, fast ramping generators: ancillary services
.., it needs actuators to provide energy and ancillary services Gas Nuclear Turbine Σ Control C Coal Batteries gas turbines Water Pump HVAC Actuation Power Grid H Measurements: Voltage Frequency Phase slow, large generators: engine small, fast ramping generators: flaps
The trouble(?) with solar and wind Potential for tremendous societal benefit : 33% energy from solar and wind will reduce carbon emissions by ~30%, and save ~7 billion $ per year (NREL western wind and solar integration study, phase 2, Sept. 2013) Potential for headache: (Germany, U.S. Pacific Northwest,...) not controllable, (i) Volatile (time varying), (ii) Unpredictable Regulation requirement Load- following requirement Hellman, Resources and transmission planning to achieve a 33% RPS in CAISO..., 2010
Automated Load Tuning to obtain Inexpensive Ancillary Services build more gas turbines? giant batteries? 1. Flexible loads = short- term storage 2. Load Tuning = Demand Response 3. FERC 755 *1980: Schweppe et al.
Automated Load Tuning to obtain Inexpensive Ancillary Services build more gas turbines? giant batteries? Now: dispatchable generation meeting demand Future: tune demand to meet volatile generation* Enabling resource: flexible loads Task: tune loads to provide ancillary service, without causing inconvenience to consumers the loads serve has to be inexpensive to the grid as well as to the consumer has to respect capacity and bandwidth (time- scale) constraints of loads 1. Flexible loads = short- term storage 2. Load Tuning = Demand Response 3. FERC 755 *1980: Schweppe et al.
Automated Load Tuning Intelligence at the nodes grid operator option 1: Hierarchical control architecture C Grid G C C C low bandwidth loads medium bandwidth loads high bandwidth loads option 2: fully distributed (not in this talk)
#1 Source of flexible loads: buildings www.eia.gov HVAC systems in commercial buildings: approx 20% of national electricity consumption (U.S.) Large thermal inertia, so small and fast variation in air flow has little effect on indoor climate Buildings with VAV (variable air volume) systems are particularly well suited for load tuning Next: ancillary service from variable speed loads in commercial buildings
VAV (Variable air volume) HVAC systems chiller Electricity consumers in order: 1. chiller 2. fans 3. pumps (4/2/1) heating
Case study #1: obtain high frequency ancillary service, in [1/(5 min) 1/(8 sec)] Regulation Signal r Fan Controller Regulation Controller + fan speed command u r Fan fan speed v p b+r Desired air flow Building + climate controller Air flow building control system
Case study #1: obtain high frequency ancillary service, in [1/(5 min) 1/(8 sec)] Grid operator Regulation Signal band pass filter r Fan Controller Regulation Controller + fan speed command u r Fan fan speed v p b+r Desired air flow Building + climate controller Air flow Band pass filter to ensure the climate control unit doesn t fight the regulation command
Band pass filter to prevent fight between controllers Magnitude (db) -20-40 -60 0 Regulation Command to Fan Speed Regulation Command to Temperature 10-3 1/600 1/8 Frequency (Hz) power 10 0 Regulation Signal filter r Fan Controller Regulation Controller + fan speed command u r Fan fan power v p b+r Desired air flow Building + climate controller Air flow
Excellent tracking with little impact..... as long as you maintain frequency (time scale) separation Simulation results: 5 Regulation Actual Estimation Power (kw) 0 5 13 14 15 16 17 Time (h) Temperature ( o C) 0.1 0.05 0 0.05 0.1 13 14 15 16 17 Time (h)
Experimental results : Pugh Hall @ UF AHU 2: serves auditorium u b u r u a Regulation controller Fan Plant P ref P f Power (kw) 4 3 2 Fan power 1 0 20 40 60 80 Reference: PJM regd, bandpass filtered [1/(30 s) 1/(120 s) ] 2 1.5 1 Reference signal Measured power deviation Temp ( o F) 74 73.5 73 72.5 Room temperature 72 0 20 40 60 80 Time (min) Power (kw) 0.5 0 0.5 1 1.5 20 21 22 23 24 25 Time (min) Ack: PPD, TM, Yashen,...
Experimental results : Pugh Hall @ UF AHU 2: serves auditorium u b u r u a Regulation controller Fan Plant P ref P f Power (kw) 4 3 2 Fan power 1 0 20 40 60 80 Reference: PJM regd, bandpass filtered [1/(30 s) 1/(120 s) ] 2 1.5 1 Reference signal Measured power deviation Temp ( o F) 74 73.5 73 72.5 Room temperature 72 0 20 40 60 80 Time (min) Power (kw) 0.5 0 0.5 1 1.5 20 21 22 23 24 25 Time (min) Ack: PPD, TM, Yashen,...
Experimental results : Pugh Hall @ UF AHU 2: serves auditorium u b u r u a Regulation controller Fan Plant P ref P f Power (kw) 4 3 2 Fan power 1 0 20 40 60 80 Reference: PJM regd, bandpass filtered [1/(30 s) 1/(120 s) ] 2 1.5 1 Reference signal Measured power deviation Temp ( o F) 74 73.5 73 72.5 Room temperature 72 0 20 40 60 80 Time (min) Power (kw) 0.5 0 0.5 1 1.5 20 21 22 23 24 25 Time (min) Ack: PPD, TM, Yashen,...
How much can this help? A LOT! Resource 35 kw fan motor Ancillary service potential 5 kw 46,000 sq.ft bldg, 75 kw fan power 11 kw All buildings in the U.S. with VAV HVAC systems 6.6 GW 70% of the regulation capacity needed by the US in 2012, in the high frequency range ( ~1/1 min) Inexpensive! 1. software add on, no change in equipment 2. non disruptive to consumer service > Economic incentive? FERC oder 755 (2011) > Load aggregation? Ref: Ancillary Service for the Grid Via Control of Commercial Building HVAC Systems, He Hao, Anupama Kowli, Yashen Lin, Prabir Barooah, Sean Meyn, American Control Conference, June, 2013
Case study #2: obtain medium frequency ancillary service, in [1/(60 min) 1/(5 min)] Medium frequency variation in air flow will affect (i) fan power and (ii) chiller power consumption Good : much more ancillary service! Bad : delay, more complex dynamics, lack of direct actuation ability u fan power [τ 1 τ 2 ] chiller u fan air flow rate m m T la T la small P chiller high freq. variation in air flow rate P chiller! d " low freq. variation in air flow rate
Indirect actuation of chillers without VFDs P r P d To handle transport delay in chiller: 1. Predict reference command (Kalman filter), 2. delay tolerant controller Band-pass Filter + Scaling m az Indoor Climate Controller Kalman Predictor P dp Smith Predictor m ar m ad Fan/Duct/Damper Dynamics m a T Zone Dynamics Power (KW) T (F) 40 20 0 20 Real 40 0 5 10 15 20 80 75 Reference Real Baseline P b P Power Model Regulation Controller Tmix Wmix Closed-loop Building Dynamics Approx 80 GW of ancillary service in the medium frequency [1/(60 min) - 1/(5 min)] range, with about 1 o F temperature deviation ma (kg/s) 70 0 5 10 15 20 8 6 4 2 0 0 5 10 15 20 Time (h) Real Baseline Ref: Low Frequency Ancillary Services from Commercial Building HVAC Systems, Yashen Lin, Prabir Barooah, Sean Meyn, IEEE Smart Grid Comm, October, 2013, Vancouver, Canada
Ancillary service from on/off loads? Residential flexible loads (A/Cs, refrigerators,..) Commercial flexible load (constant volume HVAC, on/off chillers, pool pumps*) Constraint: capacity and bandwidth of loads (time scale of flexibility) Existing work : thermostatic loads (Hiskens, Callaway, Mathieu,...) Case study #3: low frequency ancillary service, in [1/(a few hours) 1/(1 hour)] using on/off resources : pool pumps* (1 GW load in FL) Intelligence at the node: > randomized policy, control changes the transition probabilities p(on has been off for i hours) > optimal control : minimize - zu(x) + D(p p_0) (needs of the grid + needs of the consumer) > closed form solution: family of trans. probs. parameterized by z *FP&L: 780,000 customers enrolled in On- Call
Intelligence at the grid: > task: design z so that power consumption dev. of a collection of pool pumps tracks a reference signal r(t) > mean field limit of a collection of pool pumps: min. phase LTI system simulation with 1 million Markovian pools, 1 kw each r(t) : A regulation signal Ancillary service to the grid from deferrable loads: the case for intelligent pool pumps in Florida, Sean Meyn, Prabir Barooah, Ana Busic and Jordan Ehren, IEEE Con. on Decision and Control, Dec 2013, Florence, Italy.
Flexible loads providing ancillary service Gas Turbine Disturbances from nature Σ Control C Coal Batteries Water Pump HVAC Actuation Power Grid H Baseline Generation Measurements: Voltage Frequency Phase
Flexible loads providing ancillary service Gas Turbine Disturbances from nature Σ Control C Coal Batteries Water Pump HVAC Actuation Power Grid H Baseline Generation Measurements: Voltage Frequency Phase Intelligence at the nodes: control of the actuators Gas Turbine Σ Control C C A Coal Batteries Water Pump LOAD Actuator feedback loop Power Grid H Measurements: Voltage Frequency Phase
Automated Load Tuning: recap Intelligence at the nodes, respecting capacity and bandwidth constraints grid operator option 1: Hierarchical control architecture C Grid G C C C flexible manufacturing HVAC, heaters, EVs flywheel, batteries option 2: fully distributed? Economic incentive: FERC oder 755 (2011)
Looking into the crystal ball (personal opinion)
Looking into the crystal ball (personal opinion) When the distinction between Generation and Loads diminish (as providers of ancillary service), the current language is less- than- desirable: load following Capacity constraint, ramp rate constraint, energy constraint, regulation reserve, gain Challenges in..., actuation,... for future power systems (?)
Looking into the crystal ball (personal opinion) When the distinction between Generation and Loads diminish (as providers of ancillary service), the current language is less- than- desirable: load following Capacity constraint, ramp rate constraint, energy constraint, regulation reserve, gain Challenges in..., actuation,... for future power systems (?) G(j!)!(= 1/ ) 1. bandwidth 2. PSD \G(j!)
Looking into the crystal ball (personal opinion) When the distinction between Generation and Loads diminish (as providers of ancillary service), the current language is less- than- desirable: load following Capacity constraint, ramp rate constraint, energy constraint, regulation reserve, gain Challenges in..., actuation,... for future power systems (?) G(j!)!(= 1/ ) 1. bandwidth 2. PSD \G(j!) Unintended consequences due to focus on energy efficiency
Parting Comments A sustainable energy future with a high renewable energy portfolio requires substantially large amounts of inexpensive ancillary service Tuning of a large number of loads offers a vast and untapped source of inexpensive ancillary service In principle, can be solved through appropriate automation Many, many, many open research questions: 1. technological 2. social- science related 2. economic Financial support from National Science Foundation, DOE (EPAS, CPS)
References 1. How demand response from commercial buildings will provide the regulation needs of the grid, He Hao, Timothy Middelkoop, Prabir Barooah and Sean Meyn, invited paper, 50th Allerton Conf., Oct. 2012. 2. Ancillary Service for the Grid Via Control of Commercial Building HVAC Systems, He Hao, Anupama Kowli, Yashen Lin, Prabir Barooah, Sean Meyn, American Control Conference, June, 2013 3. Ancillary Service to the Grid through Control of Fans in Commercial Building HVAC Systems, He Hao, Anupama Kowli, Yashen Lin, Prabir Barooah, Sean Meyn, IEEE Transactions on Smart Grid, under review 4. Low Frequency Ancillary Services from Commercial Building HVAC Systems, Yashen Lin, Prabir Barooah, Sean Meyn, IEEE Smart Grid Comm, October, 2013, Vancouver, Canada 5. Ancillary service to the grid from deferrable loads: the case for intelligent pool pumps in Florida, Sean Meyn, Prabir Barooah, Ana Busic and Jordan Ehren, IEEE Con. on Decision and Control, Dec 2013, Florence, Italy. http://humdoi.mae.ufl.edu/~pbarooah/research/pbresearch_buildings.html Classifieds: 1. Workshop at the IEE CDC on Ancillary Service from Flexible Loads to Help the Electric Grid of the Future by Barooah, Hiskens, Kowli, Meyn, Mathieu, December 9, Florence, Italy 2. 2nd Interdisciplinary workshop on Smart Grid design and implementation, Gainesville, FL, March 28-29, 2014: http://ccc.centers.ufl.edu/?q=sg2014 (ECE and College of Business Administration) Acknowledgements: National Science Foundation, DOE (EPAS, CPS)