Changhong Zhao - PNNL

NREL is a national laboratory of the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, operated by the Alliance for Sustainable Energy, LLC.

Realizing virtual power plants

Changhong Zhao

Control of Complex Systems Workshop – May 23, 2017

Acknowledgments

Andrea Simonetto Sairaj Dhople Emiliano Dall’Anese Christine Chen Swaroop Guggilam

Network Optimized Distributed Energy

Systems (NODES)

Funding agency:

Objective

1

Objective

1

Objective

1

Enable DER coordination to pursue objectives of customers and utility/aggregator

Enable feeder to emulate a virtual power plant providing services to the main grid

Feeder as a virtual power plant

2

- Primary frequency response

- Secondary frequency response

- Follow dispatch signals

Feeder as a virtual power plant

2

Respect electrical limits (e.g., voltage regulation)

Maximize customers’ and

utility/aggregator objectives

- Inertial response

- Primary frequency response

- Secondary frequency response

- Follow dispatch signals

Outline

3

Modeling

Tracking of setpoints at the feeder head

Primary frequency response

Leveraging linear models

Nonlinear AC power flows

4

Leveraging linear models

Nonlinear AC power flows

(Approximate) linear relationships

How to obtain (and update) the model parameters?

Regression-based; e.g., online recursive least squares [Angelosante-Giannakis’09]

Model-based [Baran-Wu’89, Dhople at al’15, Bolognani-Dorfler’15]

4

Leveraging linear models

Example

Bus0 5 10 15 20 25 30 35 40

p.u

.

1.01

1.015

1.02

1.025

1.03

1.035

1.04

1.045

1.05

1.055Voltage Approximation Results for Peak and Off-Peak Solar

Voltage Approximation at 12:00 AM

Actual Voltage at 12:00 AM

Voltage Approximation at 12:30 PM

Voltage Approximation at 12:30 PM

5

Leveraging linear models

Example

For future developments …

Bus0 5 10 15 20 25 30 35 40

p.u

.

1.01

1.015

1.02

1.025

1.03

1.035

1.04

1.045

1.05

1.055Voltage Approximation Results for Peak and Off-Peak Solar

Voltage Approximation at 12:00 AM

Actual Voltage at 12:00 AM

Voltage Approximation at 12:30 PM

Voltage Approximation at 12:30 PM

5

Tracking AGC, ramping, and dispatch signals

6

Controller design principles

1

“Sample and solve”: series of optimization problems, one every

Leverage the time-varying optimization formalism [Simonetto-Leus’14]

7

Controller design principles

1

Leverage the time-varying optimization formalism [Simonetto-Leus’14]

“Sample and solve”: series of optimization problems, one every

Not practical: computation/communication limits; convergence

7

Controller design principles

1

Leverage the time-varying optimization formalism [Simonetto-Leus’14]

“Sample and solve”: series of optimization problems, one every

Not practical: computation/communication limits; convergence

Not accurate: model mismatches (linear)

7

Controller design principles

8

1

Low-complexity online algorithms to find and track optimal solutions

.

Controller design principles

1

Low-complexity online algorithms to find and track optimal solutions

Feedback from the system to cope with model mismatches and promote adaptability

[Jokic at al’09, Dall’Anese at al’15, Bernstein at al’16, Dall’Anese-Simonetto’16, Gan-Low’16]

8

Controller design principles

1

Low-complexity online algorithms to find and track optimal solutions

Feedback from the system to cope with model mismatches and promote adaptability

[Jokic at al’09, Dall’Anese at al’15, Bernstein at al’16, Dall’Anese-Simonetto’16, Gan-Low’16]

Establish analytical results for tracking capabilities

8

Formalizing operational target

9

Targeted setpoint at feeder head:

Formalizing operational target

Targeted setpoint at feeder head:

Targeted problem:

Tracking setpoints

Voltage regulation

Hardware constraints

Cost/objective

9

Formalizing operational target

Targeted setpoint at feeder head:

Targeted problem:

Tracking setpoints

Voltage regulation

Hardware constraints

Cost/objective

(Ass. 1) convex and continuously differentiable

(Ass. 3) Problem is feasible

(Ass. 2) The map is Lipschitz continuous

9

Controller design

10

Primal-dual gradient method for regularized Lagrangian [Koshal et al’11]

+ online + feedback

Controller design

Primal-dual gradient method for regularized Lagrangian [Koshal et al’11]

+ online + feedback

10

Controller design

Primal-dual gradient method for regularized Lagrangian [Koshal et al’11]

+ online + feedback

: unique primal-dual (time-varying) solutions

10

Real-time control strategy

11

[S1] Measure voltages and power at the substation

Real-time control strategy

[S2a] Update dual variables using measurements

11

Real-time control strategy

[S2a] Update dual variables using measurements

11

Real-time control strategy

[S2b] Broadcast dual variables

11

Real-time control strategy

[S3] Compute and command setpoints

11

Convergence

12

Convergence

(Ass. 4) There exist and such that:

12

Convergence

13

Assumption captures measurement errors and slow DER dynamics

(Ass. 5) There exists such that:

Convergence

14

Theorem [Dall’Anese et al’16]. Under current modeling assumptions, if the stepsize is

chosen such that:

then the following holds for the closed-loop system above:

Where and .

Representative results

15

IEEE 37-node test feeder

Real load and solar irradiance data from Anatolia, CA

PQ updated every 300ms

PV inverter mimics first-order system

Representative results

16

Voltage regulation Dispatchable VPP

Representative results

17

The value of fast communication

Representative results

19

9:00 10:00 11:00 12:00 13:00 14:00 15:001.00

1.01

1.02

1.03

1.04

1.05

1.06

Node 2

Node 28

Node 35

Volt

age

mag

nit

ude

[pu]

Outline

3

Modeling

Tracking of setpoints at the feeder head

Primary frequency response

Primary frequency response provided by DERs

inertial response

inertial response

Primary frequency response provided by DERs

Integral

Frequency response provided by DERs

Integral

How to design dk to achieve prescribed droop characteristic at the feeder head?

Outline

3

Modeling

Tracking of setpoints at the feeder head

Primary frequency response

Guggilam, Zhao, Dall’Anese, Chen, and Dhople,

“Primary frequency response with aggregated DERs,”

Thursday 16:20—16:40, Willow A

Validation

28

Power hardware-in-the-loop at NREL:

Algorithms embedded in microcontrollers

100 physical DERs controlled in real time

Real feeder with 2000 customers

Concluding remarks

29

Feeder as virtual power plants to provide services to the main grid

Primary frequency response and setpoint tracking to enhance reliability

Address voltage regulation and pursue optimization objectives

Value of fast communications to enable coordination

Future efforts towards

Convergence to solution of nonconvex problems

Optimization layer to dispatch HVAC systems, EVs, and other DERs

Validation and implementation

Thank you!

Comments?