Computation of Maximum loadability limit in

Matpower
Camille Hamon
May 19, 2017

Contents
1 Examples 2
1.1 IEEE 9 bus system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 With reactive power limits . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Without reactive power limits enforced . . . . . . . . . . . . . . . . . 2
1.1.3 Comparison with the Matpower CPF function . . . . . . . . . . . . 3

2 Notations 5

3 The maximum loadability limit problem 5

4 Implementation in Matpower 6
4.1 Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.2 Defining dispatchable loads . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.3 Adjusting the Matpower case . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.3.1 Power constraints of dispatchable loads . . . . . . . . . . . . . . . . . 7
4.3.2 Power constraints on generators . . . . . . . . . . . . . . . . . . . . . 7
4.3.3 Voltage constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.3.4 Branch flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.3.5 Decision variable α and additional constraints . . . . . . . . . . . . . 8
4.3.6 Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.3.7 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1
 1 Examples
1.1 IEEE 9 bus system
1.1.1 With reactive power limits
The following code looks for the maximum loadability point limit in a the direction of load
increase defined by dir mll.
1 % t h e IEEE 9
2

3 % MATPOWER
4 % Copyright ( c ) 2015 −2016 , Power Systems E n g i n e e r i n g Research
Center (PSERC)
5 % by C a m i l l e Hamon
6 %
7 % This f i l e i s p a r t o f MATPOWER.
8 % Covered by th e 3− c l a u s e BSD L i c e n s e ( s e e LICENSE f i l e f o r
details ) .
9 % See http : / /www. p s e r c . c o r n e l l . edu /matpower/ f o r more i n f o .
The printed results are shown in Figure 1. The first block of information gives information
about the stability margin to the maximum loadability limit (MLL) point and the type of
bifurcation at the MLL point. In this example, the MLL is a limit-induced bifurcation that
occurs when generator 2 reaches its reactive power limit. The second block of information
gives the participation of all buses in the direction of load increase, the loads and the voltage
magnitudes at all buses at the MLL. The third block of information shows the reactive power
generation at all generators as compared to their limits, and the voltage magnitude at the
generator buses as compared to their set points. A flag REF indicates which generator is the
slack. In the case of a limit-induced bifurcation, a flag LIB indicates the generator which
generator caused the LIB when reaching its reactive power limit.

1.1.2 Without reactive power limits enforced

In this example, the reactive power limits are not enforced in the search for the MLL point.
This is done by setting the flag use qlim to 0.
1

2 % Loading t h e system and d e f i n i n g p a r a m e t e r s

3 mpc = l o a d c a s e ( ’ c a s e 9 ’ ) ; % l o a d i e e e 9
The results are shown in Figure 2. Since the reactive power limits are not enforced, the
maximum loadability limit point is now a saddle-node bifurcation point (SNB). It can be
seen that the reactive power generation at the generator at bus 2 is 341.64 Mvar which is
larger than its limit (300 Mvar).

2
 =======================================
Maximum loadability problem
=======================================
The stability margin is 339.53 MW
The type of bifurcation is limit-induced bifurcation (LIB).
Generator responsible for LIB: Gen 2 connected at bus 2

--------------------------------------------------
Bus nb Direction Load at MLL Voltages
------ --------- ----------- --------
1 0 0.00 1.000
2 0 0.00 1.000
3 0 0.00 1.000
4 0 0.00 0.867
5 0 90.00 0.806
6 0 0.00 0.852
7 1 439.53 0.663
8 0 0.00 0.819
9 0 125.00 0.784

=============================================================
Reactive power production and voltages at the generators
=============================================================
Bus nb Qgen Qmin Qmax Vm Vref
-------- ------- ------ ------ ----- -----
1 305.15 -300.00 9999.00 1.0000 1.0000 REF
2 300.00 -300.00 300.00 1.0000 1.0000 LIB
3 255.64 -300.00 300.00 1.0000 1.0000

Figure 1: Results of the search for the MLL in the IEEE 9 bus system with reactive power
limit enforced.

1.1.3 Comparison with the Matpower CPF function

The continuation power flow is implemented in the Matpower function runcpf. It does
not consider reactive power limits. The following code runs the continuation power flow in
the same direction of load increase as in the previous section (load increase at bus 7).
1

2 %% With r e a c t i v e power l i m i t s ( d e f a u l t )
3 % Running th e s e a r c h f o r MLL with v e r b o s e s e t t o 1 t o p r i n t t he
results .
4 r e s u l t s m l l = maxloadlim (mpc , d i r m l l , ’ v e r b o s e ’ , 1 ) ; %

3
 5

6 %% Without e n f o r c e m e n t o f th e r e a c t i v e power l i m i t s .
7 % The r e a c t i v e power l i m i t s a r e not e n f o r c e d i n th e f o l l o w i n g
example .
8 r e s u l t s m l l = maxloadlim (mpc , d i r m l l , ’ v e r b o s e ’ , 1 , ’ u s e q l i m ’ , 0 ) ;
9

10 % The same MLL p o i n t can be o b t a i n e d by MATPOWER i m p l e m e n t a t i o n o f

t h e CPF.
11 define constants ;
12 m p c t a r g e t = mpc ;
13 n o n z e r o l o a d s = m p c t a r g e t . bus ( : ,PD) ˜= 0 ;

=======================================
Maximum loadability problem
=======================================
The stability margin is 342.37 MW
The type of bifurcation is saddle-node bifurcation (SNB).

--------------------------------------------------
Bus nb Direction Load at MLL Voltages
------ --------- ----------- --------
1 0 0.00 1.000
2 0 0.00 1.000
3 0 0.00 1.000
4 0 0.00 0.846
5 0 90.00 0.776
6 0 0.00 0.830
7 1 442.37 0.620
8 0 0.00 0.793
9 0 125.00 0.753

=============================================================
Reactive power production and voltages at the generators
=============================================================
Bus nb Qgen Qmin Qmax Vm Vref
-------- ------- ------ ------ ----- -----
1 348.43 -300.00 9999.00 1.0000 1.0000 REF
2 341.64 -300.00 9999.00 1.0000 1.0000
3 293.07 -300.00 9999.00 1.0000 1.0000

Figure 2: Results of the search for the MLL in the IEEE 9 bus system without enforcement
of the reactive power limits.

4
14 Q P = m p c t a r g e t . bus ( n o n z e r o l o a d s ,QD) . / m p c t a r g e t . bus (
n o n z e r o l o a d s ,PD) ;
15 m p c t a r g e t . bus ( : ,PD) = m p c t a r g e t . bus ( : ,PD) +2∗ d i r m l l ∗ m p c t a r g e t .
baseMVA ;
16 m p c t a r g e t . bus ( n o n z e r o l o a d s ,QD) = Q P . ∗ m p c t a r g e t . bus (
n o n z e r o l o a d s ,PD) ;
17 % Run t h e CPF with matpower
18 [ r e s u l t s , ˜ ] = r u n c p f (mpc , m p c t a r g e t ) ;
The results are shown in Figure 3. Comparing with the results in Figure 2, it can be seen
that both algorithms lead to very similar results; the small differences can be explained by
the step length of the CPF algorithm.
================================================================================
| Bus Data |
================================================================================
Bus Voltage Generation Load
# Mag(pu) Ang(deg) P (MW) Q (MVAr) P (MW) Q (MVAr)
----- ------- -------- -------- -------- -------- --------
1 1.000 0.000* 482.24 362.02 - -
2 1.000 -50.453 163.00 355.52 - -
3 1.000 -55.236 85.00 305.48 - -
4 0.839 -19.338 - - - -
5 0.766 -37.631 - - 90.00 30.00
6 0.822 -58.708 - - - -
7 0.605 -79.871 - - 442.07 154.72
8 0.784 -57.915 - - - -
9 0.743 -38.424 - - 125.00 50.00
-------- -------- -------- --------
Total: 730.24 1023.02 657.07 234.72

Figure 3: Results of the Matpower implementation of continuation power flow.

2 Notations
If not indicated otherwise, the notations are the same as in Matpower’s manual.

3 The maximum loadability limit problem

Finding the maximum loadability limit from base-case loading Pd0 ∈ Rnb in a load increase
direction d ∈ Rnb can be done by solving the following optimisation problem

5
 max α (1a)
z
subject to Pd = Pd0 + αd (1b)
Qd = diag(tan φ0 )Pd (1c)
g(z) = 0 (1d)
zmin ≤ z ≤ zmax (1e)
where
 
Θ
Vm 
T T
 
 Pg  ,
z = [x α] =  (2)

 Qg 
α
where x is defined in Equation (6.5) in Matpower’s manual and the function g is defined
is Equations (4.2) and (4.3) in Matpower’s manual. Equation (1c) ensures that the power
factor of the loads is kept constant to cos(φ0i ), the value defined in the base case. Note that
tan φ0 ∈ Rnb . The limits (1e) are as follows:
θi = θiref , i ∈ Iref (3)
i i,ref
vm = vm , i ∈ Iref (4)
i i,ref
0 ≤vm ≤ vm , i ∈ {1, . . . , ng }\Iref (5)
qgi,min ≤qgi ≤ qgi,max , i ∈ {1, . . . , ng }\Iref (6)
pig = pi,0
g , i ∈ {1, . . . , ng }\Iref (7)
α ∈ [0, +∞]. (8)
where pi,0
g is the active power generation at bus i in the base case. Equations (3) and (4)
lock the voltage angle and magnitude of the reference buses to their set points. Equations
(5) and (6) limit the terminal voltages and reactive power generation at the PV buses. Note
that in theory, complementarity constraints
i i,ref
(vm − vm ) · (qgi − qgi,max ), i ∈ {1, . . . , ng }\Iref (9)
i i,ref
should be included. However, in practice, solvers will set vm = vm if qgi is smaller than
qgi,max and let vm
i
be free if qgi has reached its limit, since doing so maximizes loadability.

4 Implementation in Matpower
4.1 Inputs
We assume that the following inputs are available:

6
 1. A Matpower base case mpc.

2. A load increase direction d ∈ Rnb , with the assumption that only nonzero loads in mpc
can be increased.

4.2 Defining dispatchable loads

A new Matpower case mpc vl is built where all nonzero loads in mpc are transformed to
dispatchable loads:
1 mpc vl = l o a d 2 d i s p (mpc)
Dispatchable loads are described in Section 6.4.2 of Matpower’s package. Note that,
conceptually, this means that the active parts of the nonzero loads in Pd enter Pg in (2) and
their reactive parts enter Qg .

4.3 Adjusting the Matpower case

4.3.1 Power constraints of dispatchable loads
In Matpower, dispatchable loads are assumed to maintain a constant power factor. This
power factor is determined by the values in the columns PMIN and QMIN corresponding to the
dispatchable loads in the field mpc vl.gen. When using the function load2disp, these fields
are set to the opposite of the active and reactive power parts of the nonzero loads in the base
case mpc. When solving the optimisation problem (1) with Matpower, the columns PMIN
and QMIN also constrain the maximum values of dispatchable loads. Since we are looking
for maximum loadability limits, upper values on loads must not be binding. Therefore, the
values PMIN of rows in mpc vl.gen corresponding to dispatchable loads is set to a very large
negative values and the values QMIN of these dispatchable loads are changed accordingly to
keep the constant power factor defined in the base case.

4.3.2 Power constraints on generators

The maximum loadability limit problem differs from a standard AC OPF as defined in
Equations (6.1) to (6.4) in Matpower’s manual in the following way

1. Active power production at PV buses is locked to the base case values. Therefore, the
columns PMIN and PMAX of these generators are set to their active power production in
the base case.

2. Generators at reference buses pick up the loads during the load increase process. The
assumption here is that the reference buses correspond to strong generators that do not
reach their power limits. Therefore, active and reactive power limits of the generators
at reference buses are set to large values.

7
4.3.3 Voltage constraints
Similarly to power constraints discussed above, voltage constraints in the maximum load-
ability problem differ from a standard AC OPF as follows

1. Voltage magnitudes at the reference buses are locked at their set point values, see
Equation (4).

2. Voltage magnitudes at all PV and PQ buses are not constrained from below. The
reason for this is that the maximum loadability limit point typically corresponds to
an operating point that is outside the normal allowed band on bus voltage magnitude.
Therefore, to find this point, lower limits on bus voltage magnitudes of PV and PQ
buses must be relaxed.

3. For the same reason, upper bounds on voltage magnitudes of PQ buses are also relaxed.

4. Upper bounds on voltage magnitudes of PV buses are set to the voltage set point of
the corresponding generators.

4.3.4 Branch flows

For the same reason as for bus voltages, the branch flow constraints must be relaxed so that
they do not become binding during the optimisation process.

4.3.5 Decision variable α and additional constraints

Compared to the standard AC OPF formulation in Matpower, the formulation in (1)
contains an additional decision variable α, see (2), and the additional constraints (1b).
Note that constraints (1c) does not exist in the standard AC OPF formulation either, but
Matpower assumes a constant power factor model for the dispatchable loads, see also
Section 4.3.1. Therefore, there is no need to define the additional constraint (1c). The
decision variable α and the constraint (1b) are defined using a callback function for the
formulation stage.
Following Section 7.2 in Matpower’s manual, we create a callback function called
userfcn direction mll formulation(om,args) in which we define the new variable and
constraints. To define the new constraint, recall that when transforming the loads to dis-
patchable loads, the nonzero loads of the base case enter the set of generators, see Section
4.2. Matpower has a feature to only define the additional constraints on the relevant
components of the decision variable z in (2). For Equation (1b), the relevant component of
z are Pg (since the dispatchable loads enter Pg ) and α.

4.3.6 Costs
When solving an OPF with dispatchable loads in Matpower, the costs of dispatchable
loads is added to that of the generator production costs. Since the objective function of the

8
maximum loadability limit is just α (see (1a)), we are not interested in keeping track of the
cost of dispatchable loads and generators. These costs are therefore set to zero.
The cost for α is defined in the callback function introduced in Section 4.3.5. It is set to -1
since the objective in (1a) is to maximize α, whereas Matpower performs a minimisation.

4.3.7 Functions
The functions in Table 1 are used to set up and solving the maximum loadability limit
problem with Matpower.

Name Description
maxloadlim Main function for solving the maximum load-
ability limit problem.
postproc maxloadlim Post-processing function that transforms the
dispatchable loads back to normal loads.
prepare maxloadlim Prepares the Matpower case before the AC
OPF is run, see Section 4.3.
print maxloadlim Pretty prints the results from the maximum
loadability problem.
userfcn direction mll ext2int Converts the vector of generator production
changes to internal indexing.
userfcn direction mll formulation Callback function to define new variables,
constraints and costs, see Section 4.3.5.

Table 1: Functions used for the maximum loadability limit problem