ECEN 615

Methods of Electric Power

Systems Analysis
Lecture 21: Economic Dispatch,
Optimal Power Flow, Linear Programming

Prof. Tom Overbye

Dept. of Electrical and Computer Engineering
Texas A&M University
[email protected]
 Announcements
• Read Chapter 8
• Read the Chapter 3 appendices (3A covers optimization with constraints,
3B covers linear programming, 3D covers dynamic programming, and 3E
convex optimization
• An excellent book on optimization is Linear and Nonlinear Programming
by Luenberger and Ye (the 5th edition came out in 2021)
• Homework 6 is now due on Thursday Nov 17 but it counts as two regular
homeworks.

2
 Power System Economic Dispatch
• Generators can have vastly different incremental operational costs
– Some are essentially free or low cost (wind, solar, hydro, nuclear)
– Because of the large amount of natural gas generation, electricity prices are very
dependent on natural gas prices
• Economic dispatch is concerned with determining the best dispatch for
generators without changing their commitment
• Unit commitment focuses on optimization over several days. It is
discussed in Chapter 4 of the book and was briefly covered last lecture

3
Variation in Natural Gas Prices and Generation
Sources

Source:
www.eia.gov/dnav/ng/h
ist/rngwhhdm.htm

4
 Power System Economic Dispatch
• Economic dispatch is formulated as a constrained minimization
– The cost function is often total generation cost in an area
– Single equality constraint is the real power balance equation
• Solved by setting up the Lagrangian (with P D the load and PL the losses,
which are a function the generation)
m m
L(PG ,  )   Ci ( PGi )   ( PD  PL (PG )   PGi )
i 1 i 1

• A necessary condition for a minimum is that the gradient is zero. Without

losses this occurs when all generators are dispatched at the same marginal
cost (except when they hit a limit)

5
 Power System Economic Dispatch
m m
L(PG ,  )   Ci ( PGi )   ( PD  PL ( PG )   PGi )
i 1 i 1
L(PG ,  ) dCi ( PGi ) PL ( PG )
   (1  ) 0
PGi dPGi PGi
m
PD  PL ( PG )   PGi 0
i 1

• If losses are neglected then there is a single marginal cost (lambda); if

losses are included then each bus could have a different marginal cost

6
Economic Dispatch Penalty Factors
Solving each equation for  we get
dCi ( PGi ) PL ( PG )
  (1  0
dPGi PGi
1 dCi ( PGi )
 
 PL ( PG )  dPGi
 1  P 
 Gi 

Define the penalty factor Li for the i th generator

1 The penalty factor
Li  at the slack bus is
 PL ( PG ) 
 1  P  always unity!
 Gi 
7
Economic Dispatch Example
58 MW A
56 MW 39 MW A
39 MW 78 MW
39% 21% 29 Mvar
1.05 pu MVA 0.99 pu MVA

3 4
1 1.00 pu
130.0 MW
72 MW 0.0000 52 MW 147 MW 46 MW
92.5 MW
slack

AGC ON 39 Mvar
-0.0825
A A

53% 46% A
A AGC ON
MVA MVA
48% 38%
MVA
MVA

54 MW
20 MW
71 MW A
47 MW 60%
1.04 pu 112 MW MVA
107 MW 0.96 pu

2 5
39 MW 181.9 MW 127 MW
20 Mvar -0.0274 39 Mvar
AGC ON
Total Hourly Cost: 5916.04 $/h Load Scalar: 1.00
Total Area Load: 392.0 MW MW Losses: 12.44 MW
Marginal Cost ($/MWh): 0.00 $/MWh

Case is GOS_Example6_22; use Power Flow Solution

Options, Advanced Options to set Penalty Factors 8
 Optimal Power Flow (OPF)
• OPF functionally combines the power flow with economic dispatch
• SCOPF adds in contingency analysis
• Goal of OPF and SCOPF is to minimize a cost function, such as operating
cost, taking into account realistic equality and inequality constraints
• Equality constraints
– bus real and reactive power balance
– generator voltage setpoints
– area MW interchange

9
 OPF, cont.
• Inequality constraints
– transmission line/transformer/interface flow limits
– generator MW limits
– generator reactive power capability curves
– bus voltage magnitudes (not yet implemented in Simulator OPF)
• Available Controls
– generator MW outputs
– transformer taps and phase angles
– reactive power controls
Two Example OPF Solution Methods
• Non-linear approach using Newton’s method
– handles marginal losses well, but is relatively slow and has problems
determining binding constraints
– Generation costs (and other costs) represented by quadratic or cubic
functions
• Linear Programming
– fast and efficient in determining binding constraints, but can have
difficulty with marginal losses.
– used in PowerWorld Simulator
– generation costs (and other costs) represented by piecewise linear
functions
• Both can be implemented using an ac or dc power flow
11
 OPF and SCOPF Current Status
• OPF (really SCOPF) is currently an area of active research, with ARPA-E
having an SCOPF competition (see gocompetition.energy.gov)
• A 2016 National Academies Press report, titled “Analytic Research
Foundations for the Next-Generation Electric Grid,” recommended
improved AC OPF models
– I would recommend reading this report; it provides good background on power
systems include OPF
– It is available for free at www.nap.edu/catalog/21919/analytic-research-foundations-
for-the-next-generation-electric-grid

12
 OPF and SCOPF History
• A nice OPF history from Dec 2012 is provided by the below link, and
briefly summarized here
• Prior to digital computers economic dispatch was solved by hand and the
power flow with network analyzers
• Digital power flow developed in late 50’s to early 60’s
• First OPF formulations in the 1960’s
– J. Carpienterm, “Contribution e l’étude do Dispatching Economique,” Bulletin
Society Francaise Electriciens, 1962
– H.W. Dommel, W.F. Tinney, “Optimal power flow solutions,” IEEE Trans. Power
App. and Systems, Oct. 1968
• “Only a small extension of the power flow program is required”
https://www.ferc.gov/sites/default/files/2020-05/acopf-1-history-formulation-testing.pdf (by M Cain, R. O’Neill, A. Castillo)
13
OPF and SCOPF History
• A linear programming (LP) approach was presented by Stott and Hobson
in 1978
– B. Stott, E. Hobson, “Power System Security Control Calculations using Linear
Programming,” (Parts 1 and 2) IEEE Trans. Power App and Syst., Sept/Oct 1978
• Optimal Power Flow By Newton’s Method
– D.I. Sun, B. Ashley, B. Brewer, B.A. Hughes, and W.F. Tinney, "Optimal Power
Flow by Newton Approach", IEEE Trans. Power App and Syst., October 1984
• Follow-up LP OPF paper in 1990
– O. Alsac, J. Bright, M. Prais, B. Stott, “Further Developments in LP-based Optimal
Power Flow,” IEEE Trans. Power Systems, August 1990

14
OPF and SCOPF History
• Critique of OPF Algorithms
– W.F. Tinney, J.M. Bright, K.D. Demaree, B.A. Hughes, “Some Deficiencies in
Optimal Power Flow,” IEEE Trans. Power Systems, May 1988
• Hundreds of other papers on OPF
• Comparison of ac and dc optimal power flow methods
– T.J. Overbye, X. Cheng, Y. San, “A Comparison of the AC and DC Power Flow
Models for LMP Calculations,” Proc. 37th Hawaii International Conf. on System
Sciences, 2004

15
 Key SCOPF Application: Locational Marginal
Prices (LMPs)
• The locational marginal price (LMP) tells the cost of providing electricity
to a given location (bus) in the system
• Concept introduced by Schweppe in 1985
– F.C. Schweppe, M. Caramanis, R. Tabors, “Evaluation of Spot Price Based Electricity
Rates,” IEEE Trans. Power App and Syst., July 1985
• LMPs are a direct result of an SCOPF, and are widely used in many
electricity markets worldwide
–

16
Example MISO LMP Contour, 11/11/2022

LMPs are now widely visualized

using color contours; the first use
of LMP color contours was
presented in [1]

[1] T.J. Overbye, R.P. Klump, J.D. Weber, “A

Virtual Environment for Interactive
Visualization of Power System Economic and
Security Information,” IEEE PES 1999 Summer
Meeting, Edmonton, AB, Canada, July 1999

Image: api.misoenergy.org/MISORTWD/lmpcontourmap.html

17
Example LMP Contour: 10/27/2020
Note the
wide range
in LMPs
including
some negative
values!

This is just
the real-time
market; most
electricity
is not traded
here.
18
ERCOT LMPs, Nov 11, 2022 at 4:25 pm

19
 OPF Problem Formulation
• The OPF is usually formulated as a minimization with equality and
inequality constraints
Minimize F(x,u)
g (x, u) 0
h min h(x, u) h max
u min u u max

where x is a vector of dependent variables (such as the bus voltage

magnitudes and angles), u is a vector of the control variables, F(x,u) is
the scalar objective function, g is a set of equality constraints (e.g., the
power balance equations) and h is a set of inequality
constraints (such as line flows)
20
Two Bus with Unconstrained Line
With no overloads
Total Hourly Cost : 8459 $/h Transmission line is
the OPF matches Area Lambda : 13.02 $/MWh
not overloaded
the economic A

dispatch 81%
MVA

Bus A 13.02 $/MWh Bus B 13.02 $/MWh

slack

300.0 MW 300.0 MW
203.0 MW 397.0 MW
AGC ON AGC ON

Marginal cost of supplying power to each bus

(locational marginal costs)

21
Two Bus with Constrained Line

Total Hourly Cost : 9514 $/h

Area Lambda : 13.26 $/MWh
A

100%
MVA

Bus A 13.43 $/MWh Bus B 13.08 $/MWh

slack

380.0 MW 300.0 MW
260.9 MW 419.1 MW
AGC ON AGC ON

With the line loaded to its limit, additional load at Bus A must
be supplied locally, causing the marginal costs to diverge.

22
 Three Bus (B3) Example
• Consider a three bus case (Bus 1 is system slack), with all buses
connected through 0.1 pu reactance lines, each with a 100 MVA limit
• Let the generator marginal costs be
– Bus 1: 10 $ / MWhr; Range = 0 to 400 MW
– Bus 2: 12 $ / MWhr; Range = 0 to 400 MW
– Bus 3: 20 $ / MWhr; Range = 0 to 400 MW
• Assume a single 180 MW load at bus 2

23
B3 with Line Limits NOT Enforced

60 MW 60 MW
Bus 2 Bus 1
10.00 $/MWh

0.0 MW 10.00 $/MWh

120 MW 180.0 MW
120%
0 MW
60 MW Line between
120%
Total Cost 60 MW 120 MW Bus 1and Bus 3
1800 $/hr
Bus 3 10.00 $/MWh
180 MW
is overloaded;
all buses have
0 MW
the same
marginal cost
24
B3 with Line Limits Enforced

20 MW 20 MW
Bus 2 Bus 1
10.00 $/MWh

60.0 MW 12.00 $/MWh

100 MW 120.0 MW
100%
0 MW
80 MW
100% 100 MW
LP OPF changes
Total Cost 80 MW
1920 $/hr generation to
Bus 3 14.00 $/MWh
remove violation.
180 MW
Bus marginal
0 MW costs are now
different.
25
Verify Bus 3 Marginal Cost

19 MW 19 MW
Bus 2 Bus 1
10.00 $/MWh

62.0 MW 12.00 $/MWh

100 MW 119.0 MW
81% 100%
0 MW
81 MW One additional MW
81% 100%
Total Cost 81 MW 100 MW of load at bus 3
1934 $/hr raised total cost by
Bus 3 14.00 $/MWh
181 MW 14 $/hr, as G2 went
up by 2 MW and G1
0 MW
went down by 1MW
26
 Why is bus 3 LMP = $14 /MWh
• All lines have equal impedance. Power flow in a simple network
distributes inversely to impedance of path.
– For bus 1 to supply 1 MW to bus 3, 2/3 MW would take direct path from 1 to 3,
while 1/3 MW would “loop around” from 1 to 2 to 3.
– Likewise, for bus 2 to supply 1 MW to bus 3, 2/3MW would go from 2 to 3, while
1/3 MW would go from 2 to 1to 3.

27
 Why is bus 3 LMP $ 14 / MWh, cont’d
• With the line from 1 to 3 limited, no additional power flows are
allowed on it.
• To supply 1 more MW to bus 3 we need
– PG1 + PG2 = 1 MW
– 2/3  PG1 + 1/3  PG2 = 0; (no more flow on 1-3)
• Solving requires we up PG2 by 2 MW and drop PG1 by 1 MW -- a net
increase of $24 – $10 = $14.

28
Both lines into Bus 3 Congested

0 MW 0 MW
Bus 2 Bus 1
10.00 $/MWh

100.0 MW12.00 $/MWh

100 MW 100.0 MW
100% 100%
0 MW
100 MW For bus 3 loads
100% 100%
Total Cost100 MW 100 MW above 200 MW,
2280 $/hr
Bus 3 20.00 $/MWh the load must be
204 MW
supplied locally.
Then what if the
4 MW
bus 3 generator
opens?
29
Both lines into Bus 3 Congested

An infeasible example can be created by opening the generator

at Bus 3 with the Bus 3 load above 200 MW. There is no way
to serve the load without overloading a transmission line.

30
LP OPF Solution Method
• There are different OPF solution techniques. One common approach uses
linear programming (LP)
• The LP approach iterates between
– solving a full ac or dc power flow solution
• enforces real/reactive power balance at each bus
• enforces generator reactive limits
• system controls are assumed fixed
• takes into account non-linearities
– solving a primal LP
• changes system controls to enforce linearized constraints while minimizing cost

31
 Quick Coverage of Linear Programming
• LP is probably the most widely used mathematical programming
technique
• It is used to solve linear, constrained minimization (or maximization)
problems in which the objective function and the constraints can be
written as linear functions

32
 Example Problem 1
• Assume that you operate a lumber mill which makes both construction-
grade and finish-grade boards from the logs it receives. Suppose it
takes 2 hours to rough-saw and 3 hours to plane each 1000 board feet of
construction-grade boards. Finish-grade boards take 2 hours to rough-
saw and 5 hours to plane for each 1000 board feet. Assume that the saw
is available 8 hours per day, while the plane is available 15 hours per
day. If the profit per 1000 board feet is $100 for construction-grade and
$120 for finish-grade, how many board feet of each should you make
per day to maximize your profit?

33
Problem 1 Setup
Let x1 =amount of cg, x 2 = amount of fg
Maximize 100 x1  120 x2
s.t. 2 x1  2 x2 8
3x1  5 x2 15
x1, x2 0
Notice that all of the equations are linear, but they are
inequality, as opposed to equality, constraints; we are
seeking to determine the values of x1 and x2

34
Example Problem 2 (Nutritionist Problem)
• A nutritionist is planning a meal with 2 foods: A and B. Each ounce
of A costs $ 0.20, and has 2 units of fat, 1 of carbohydrate, and 4 of
protein. Each ounce of B costs $0.25, and has 3 units of fat, 3 of
carbohydrate, and 3 of protein. Provide the least cost meal which has
no more than 20 units of fat, but with at least 12 units of
carbohydrates and 24 units of protein.

35