MATPOWERs Extensible Optimal Power Flow

Architecture
Ray Zimmerman, Cornell University
Carlos Murillo-Snchez, Universidad Autonoma de Manizales
Robert J. Thomas, Cornell University

IEEE PES General Meeting 2009

July 30, 2009
Calgary, Alberta, Canada

PSERC
Outline

MATPOWER Overview

Extensible OPF Formulation

Standard Extensions

Software Architecture

Example: Adding Reserves

Outline

MATPOWER Overview
What does MATPOWER do?
MATPOWER History
MATPOWER Package

Extensible OPF Formulation

Standard Extensions

Software Architecture

Example: Adding Reserves

What does MATPOWER do?

DC power flow DC optimal power flow (OPF)

AC power flow BPMPD (MEX)
Primal-Dual Interior Point Method
Newton
(PDIPM)
Gauss-Seidel
AC optimal power flow (OPF)
Fast decoupled
Primal-Dual Interior Point Method
functions to compute ... (PDIPM) (pure Matlab & MEX)
derivatives of power flow equations MINOS (MEX)
generation costs successive LPs (BPMPD MEX)
linear shift factors (PTDFs, LODFs) Optimization Toolbox (fmincon, constr)
 MATPOWER 4.0
- refactored OPF (all using gen. form.)
- pure-Matlab PDIPM solver
MATPOWER History - OPF with reserves
- userfcn callbacks
- multiple solvers for DC OPF
- LODFs
1st work with Matlab power flow code for PowerWeb
- support for interior point fmincon
- based on code from Joe Chow & Chris DeMarco

MATPOWER 3.2
1st PF and OPF code of my own - version 2 case format
- based on Opt Tbx, constr() - gen capability curves
- branch angle diff lims
- PTDFs
1st public MATPOWER release - TSPOPF
- not widely publicized
- PWL costs
MATPOWER 3.0
- MINOS-based OPF (gen. form.)
MATPOWER 1.0 - fmincon-based OPF
- in-house successive LP-based OPF - DC PF & OPF
- multiple gens/bus
- Gauss-Seidel PF
MATPOWER 2.0 - improved DP de-commitment
- fast decoupled PF - automated tests
- successive LP-based OPF - separate disp. load output section
- options vector - option for active power line lims
- Users Manual - option to enforce Qg lims in PF

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
MATPOWER Package

Open source Matlab code available at: http://www.pserc.cornell.edu/matpower/

No GUI (graphical user interface)

Set of functions you can run from Matlab command line or include in your own
programs

Example:

>> result = runopf(case300);

or
>> mpc = loadcase(case300);
>> mpc.bus = scale_load(1.1, mpc.bus);
>> result = runopf(mpc);

Primary focus on research and education applications

Outline

MATPOWER Overview

Extensible OPF Formulation

Standard Formulation
Generalized Formulation
User-Defined Costs
User-Defined Constraints

Standard Extensions

Software Architecture

Example: Adding Reserves

Standard OPF Formulation

min f (x)
x
subject to
g(x) = 0
h(x) 0
xmin x xmax
Standard OPF Formulation
ng
!" #
min fP (pi ) + fQ (qi )
i i
,V,P,Q
i=1
subject to
gP (, V, P ) = 0
gQ (, V, Q) = 0
hf (, V ) 0
ht (, V ) 0
ref i ref , i = iref
vimin vi vimax , i = 1 . . . nb
pmin
i p i p max
i , i = 1 . . . ng
qimin qi qimax , i = 1 . . . ng
Generalized Formulation

min f (x) + fu (x, z)

x,z
subject to
g(x) = 0
h(x) 0
xmin x xmax
! "
x
lA u
z
zmin z zmax
Generalized Formulation

min f (x) + fu (x, z)

x,z
subject to
g(x) = 0
h(x) 0
xmin x xmax
! "
x
lA u
z
zmin z zmax additional variables
Generalized Formulation

min f (x) + fu (x, z) additional costs

x,z
subject to
g(x) = 0
h(x) 0
xmin x xmax
! "
x
lA u
z
zmin z zmax additional variables
Generalized Formulation

min f (x) + fu (x, z) additional costs

x,z
subject to
g(x) = 0
h(x) 0
xmin x xmax
! "
x
lA u additional constraints
z
zmin z zmax additional variables
User-Defined Costs
! "
1 T x
fu (x, z) = w Hw + C T w r=N
2 z

wi wi

mi
ri
ri

ri ri
ki ki ki ki

mi fdi (ui + ki ), ui < ki !
, if di = 1
u = r r wi = 0, ki ui ki fdi () =
2 , if di = 2
mi fdi (ui ki ), ui > ki
User-Defined Constraints

additional linear restrictions on all optimization variables

! "
x
lA u
z

inequality constraints

equality constraints if l=u

Outline

MATPOWER Overview

Extensible OPF Formulation

Standard Extensions
piece-wise linear costs
dispatchable (price sensitive) loads
generator reactive capability constraints
branch angle difference limits

Software Architecture

Example: Adding Reserves

Piece-wise Linear Generation Costs


m1 (x x1 ) + c1 , x x1 given the sequence of points

m2 (x x2 ) + c2 , x1 < x x2
c(x) = .. .. (xj , cj ), j = 0...n

. .

where mj is the slope of
mn (x xn ) + cn , xn1 < x
segment j
c cj cj1
mj = , j = 1...n
cn xj xj1

y add a new variable y and, for

each segment, a new linear
constraint on y

c2 y mj (x xj ) + cj , j = 1...n

c1 use y in place of c(x) in the

c0
x0
x cost function
x1 x2 xn
Dispatchable (price sensitive) Loads

modeled as negative generator

$/MW $
c (total cost)
(marginal benefit)
1 MW
p (injection)
p2 p1

p1

1
2
p1
p2 2
MW p2
p (load)

with an additional constant power factor constraint

Generator Reactive Capability Constraints

Instead of simple box constraints ...

q
q1max

qmax

q2max

p
q2min

qmin

q1min
p1 pmin pmax p2
Generator Reactive Capability Constraints

Instead of simple box constraints ...

q
q1max

qmax

q2max

p
q2min

qmin

q1min
p1 pmin pmax p2
Outline

MATPOWER Overview

Extensible OPF Formulation

Standard Extensions

Software Architecture
Overview of Execution - Callbacks
Adding Variables
Adding Constraints

Example: Adding Reserves

Overview of Execution

load data

convert to internal indexing

set up problem formulation

run optimization

convert results back to external indexing

print results (optional)

save results (optional)

Overview of Execution Callbacks

load data

convert to internal indexing

set up problem formulation

run optimization

convert results back to external indexing

print results (optional)

save results (optional)

Overview of Execution Callbacks

load data Modifying the Formulation

convert to internal indexing Option 1 externally supply

complete constraint matrix A,
cost coeff matrix N, etc.
set up problem formulation
(taking into account internal
conversions)
run optimization
Option 2 modify formulation
convert results back to external indexing directly in a callback function

print results (optional)

save results (optional)

Software Architecture - Variables
AC OPF DC OPF
Utilizes an OPF-Model object (OM) to
manage variable and constraint indexing
Va Va
Variables are added in named blocks, with
dimension, initial value and bounds, e.g.
om = add_vars(om, 'Pg', ng, Pg0, Pmin, Pmax); x= Pg
Vm
Portions of optimization variable x or limit y

shadow prices can be accessed by name, x= z

w/o need to keep track of explicit indexing Pg

Qg
name description
y
Va bus voltage angles
Vm bus voltage magnitudes
z
Pg generator real power injections
Qg generator reactive power injections
y CCV helper variables for pwl costs
z other user defined variables
Software Architecture - Constraints

Constraints added in named blocks, with A, l, u and block column names,

e.g.
om = add_constraints(om, 'Res', Ar, lr, ur, {'Pg', 'R'});
Va Vm Pg Qg y R
! "
x
lA u
z
A = 0 0 A1 0 0 A2 AC OPF
! "
Pg
l r Ar ur
R
# $
! " Pg Ar = A1 A2
lr A1 A2 ur
R

Constraint multipliers can be

accessed by name (e.g , 'Res') A = 0 A1 0 A2 DC OPF
w/o need to keep track of explicit
indexing
Va Pg y R
Outline

MATPOWER Overview

Extensible OPF Formulation

Standard Extensions

Software Architecture

Example: Adding Reserves

Example Adding Reserves

Jointly optimize the allocation of both energy and reserves

Reserve requirements are set of fixed zonal quantities

New reserve variable: 0 ri rimax

!
Additional reserve cost: fu (x, z) = ci ri
iU

Reserve constraints: pi + ri pmax

i , i U
!
ri R k , k
iZk
Adding Reserves Code

name description

om OPF model object, already includes standard OPF setup

ng number of generators
R name for new reserve variable vector
Rmin lower bound on R, all zeros
Rmax upper bound on R, based on ramp rates
Pmax capacity of generators
I identity matrix (ng x ng)
Az zone definitions, Az(i,j) = 1, iff gen j lies in zone i
Rreq vector of reserve requirements for each zone
Rcost cost coefficients for R

Ar = [I I];
om = add_vars(om, 'R', ng, [], Rmin, Rmax);
om = add_constraints(om, 'Pg_plus_R', Ar, [], Pmax, {'Pg', 'R'});
om = add_constraints(om, 'Rreq', Az, Rreq, [], {'R'});
om = add_costs(om, 'Rcost', struct('N',I,'Cw',Rcost), {'R'});
Goals & Applications

Make it as simple as possible for students and researchers to solve problems

that require variations of a power flow or OPF formulation, without having to
rewrite the parts that are shared with a standard formulation.

To be able to easily extend and modify an optimal power flow formulation to

include new variables, constraints and/or costs.

Example applications:
co-optimize energy and reserves
add environmental costs (e.g. CO2, SOx, NOx) or constraints
contingency constrained OPF

MATPOWER 4 available soon at: http://www.pserc.cornell.edu/matpower/