Power Flow
Power Flow
Power Flow
• System model
• Equations
• Solution techniques:
– Newton-Raphson
– Fast Decoupled
• Examples:
– Matlab
– PSAT
• Optimization approach
1
Power Flow Model
• The steady state operating point of a power
system is obtained by solving the “power
flow” equations.
• Power flow system model corresponds to the
steady state model.
• Generator:
– Generates and injects power P in the system
while keeping the output voltage V constant
within active and reactive power limits (capability
curve):
2
Power Flow Model
– Thus, it’s modeled as a PV bus:
3
Power Flow Model
• Slack bus:
– The phasor model needs a reference bus.
– A “large” generator is typically chosen as the
reference bus, as it should be able to take the
power “slack”:
4
Power Flow Model
– Hence, if Q reaches a limit:
5
Power Flow Model
• Load:
– Loads are typically connected to the
transmission system through ULTC
transformers.
– Thus, most loads in steady state represent a
constant power demand in the system, and
hence are modeled as a PQ bus:
6
Power Flow Model
• Transmission system:
– AC transmission lines and transformers in
steady state are basically modeled using the
following model:
7
Power Flow Model
– Hence:
8
Power Flow Model
– Thus, for an N bus system interconnected through an ac
transmission system, an NxN bus admittance matrix can be
defined:
9
Power Flow Equations
• In steady state, a system with n generators G
and m loads L can then be modeled as:
10
Power Flow Equations
• Hence, the power injections at each node
are defined by:
11
Power Flow Equations
• This yields two equations per node or bus:
12
Power Flow Equations
• These equations are referred to as the
power mismatch equations.
• The equations are typically subjected to
inequality constraints representing control
limits:
13
Power Flow Solution
• The power flow equations are known can be
represented as:
14
Power Flow Solution
• A “robust” NR technique may be used to solve
these equations:
1. Start with an initial guess, typically Vi0=1,
δi0=0, QGi0=0, Pslack0=0 (flat start).
2. At each iteration k (k = 0,1,2,…), compute
the “sparse” Jacobian matrix:
15
Power Flow Solution
3. Find ∆ zk by solving the following linear set
of equations (the sparse matrix Jk is
factorized and not inverted to speed up the
solution process):
16
Power Flow Solution
5. Stop when:
17
Example 1
• For the following system:
18
Example 1
– The Ybus matrix is:
19
Example 1
– The mismatch equations are then:
20
Example 1
21
Example 1
22
Example 1
– Power flow equations in MATLAB format (PFeqs.m), where
lambda is used to simulate constant power factor load
changes: function F = function(z)
%
% 3-bus example Power Flow equations
% Flat start: z0=[0 0 0 0 0 0];
%
global lambda
d2=z(1);
d3=z(2);
P1=z(3);
Q1=z(4);
Q2=z(5);
Q =z(6);
23
Example 1
– This 6 equations and 6 unknowns, i.e. δ2, δ3, P1, Q1,
Q2, and Q, can be solved using MATLAB’s fsolve()
routine:
>> global lambda; lambda=1;
>> zo=fsolve(@PFeqs,[0 0 0 0 0 0],optimset('Display','iter'))
Norm of First-order
Iteration Func-count f(x) step optimality CG-iterations
1 7 0.945696 18 1
2 14 0.000661419 0.705901 0.0205 1
3 21 9.98638e-018 0.0257331 2.52e-009 1.76
Optimization terminated successfully:
Relative function value changing by less than OPTIONS.TolFun
zo =
24
Example 1
– Observe that as the load (lambda) is increased,
convergence problems develop until the equations fail
to converge at lambda ≈ 21+:
>> lambda=20;
>> zo=fsolve(@PFeqs,[0 0 0 0 0 0],optimset('Display','iter'))
Norm of First-order
Iteration Func-count f(x) step optimality CG-iterations
1 8 396.67 1 360 0
2 15 115.235 8.55116 106 3
.
.
.
31 218 2.40413e-006 0.0020814 0.00659 3
32 225 1.46753e-006 0.00091283 0.00345 3
Optimization terminated successfully:
Relative function value changing by less than OPTIONS.TolFun
zo =
25
Example 2
• For a 3-area sample system:
100 MW
150 MW
AREA 1 AREA 2
50 Mvar
R = 0.01 p.u.
150 MW X = 0.15 p.u. 150 MW
60 Mvar 1.02 ∠0 V2 ∠δ2 56 Mvar
AREA 3 V3 ∠δ3
50 MW
40 Mvar
50 Mvar
100 MW
26
Example 2
– MATLAB-based PSAT can be used to solve the power flow
problem, in which case the system can be drawn using
SYMULINK (ThreeArea.mdl):
Area 1
Area 2
Area 3
27
Example 2
– This diagram is transformed by PSAT into the following MATLAB
database that represents the system (ThreeArea_mdl.m):
Bus.con = [ ...
1 138 1 0 1 1;
2 138 1 0 1 1;
3 138 1 0 1 1;
];
Line.con = [ ...
1 2 100 138 60 0 0 0.01 0.15 0 0 0 0 0 0 1;
1 3 100 138 60 0 0 0.01 0.15 0 0 0 0 0 0 1;
3 2 100 138 60 0 0 0.01 0.15 0 0 0 0 0 0 1;
];
Shunt.con = [ ...
3 100 138 60 0 0.5 1;
2 100 138 60 0 0.5 1;
];
SW.con = [ ...
1 100 138 1.02 0 999 -999 1.1 0.9 1.5 1 1 1;
];
PV.con = [ ...
3 100 138 1 1 0.001 0 1.1 0.9 1 1;
2 100 138 1 1 0.001 0 1.1 0.9 1 1;
];
PQ.con = [ ...
3 100 138 0.5 0.4 1.2 0.8 1 1;
2 100 138 1.5 0.56 1.2 0.8 1 1;
1 100 138 1.5 0.6 1.2 0.8 1 1;
];
Bus.names = {... 28
'Area 1'; 'Area 2'; 'Area 3'};
Example 2
29
Example 2
• Running the Power Flow yields the following
“Static Report” (ThreeArea_01.txt) :
POWER FLOW REPORT
P S A T 2.1.6
File: c:\G\COURSES\MEngPowerEng\ECE6613PD\Lectures\PowerFlow\ThreeArea.mdl
Date: 17-Sep-2013 16:45:42
NETWORK STATISTICS
Buses: 3
Lines: 3
Generators: 3
Loads: 3
SOLUTION STATISTICS
Number of Iterations: 5
Maximum P mismatch [p.u.] 0
Maximum Q mismatch [p.u.] 0
Power rate [MVA] 100
30
Example 2
POWER FLOW RESULTS
31
Fast Decoupled Solution
• If only the unknown bus voltage angles
and magnitudes are calculated using NR's
method (the generator reactive powers
and active slack power are evaluated
later):
32
Fast Decoupled Solution
33
Fast Decoupled Solution
• Assuming:
– (δi - δk) < 10o, then:
34
Fast Decoupled Solution
• Thus, the linear step equations may be
decoupled and reduced to
35
Fast Decoupled Solution
• The Fast Decoupled iterative method is then
defined as follows:
1. Start with an initial guess, typically δi0=0, Vi0=1.
2. Solve for ∆δk →
3. Update →
4. Solve for ∆Vk →
5. Update →
6. Compute unknown generator powers and check for
limits.
7. Repeat process for k=1,2,…, until convergence.
36
Fast Decoupled Solution
• This technique requires of a relatively large
number of iterations as compared to the
robust NR method.
• It is significantly faster, as there is no need to
re-compute the Jacobian matrix every
iteration.
• It is sensitive to initial guesses and there is
no guarantee of convergence, particularly for
systems with large transmission system
resistances.
37
Optimization Approach
• The power flow problem can be restated as an
optimization problem, with complementarity
constraints to represent generator limits:
38
Optimization Approach
• The complementarity ⊥ constraints are used to represent
the switching control conditions when the voltage
regulator reaches or comes out of a limits, and can be
represented as:
39
Optimization Approach
• This problem can be solved using
optimization techniques (e.g. fsolve in
MATLAB), such as interior point methods.
• More about this approach can be found in:
M. Pirnia, C. A. Cañizares, and K.
Bhattacharya, “Revisiting the Power Flow
Problem Based on a Mixed Complementarity
Formulation Approach,” IET Generation,
Transmission & Distribution, vol. 7, no. 11,
November 2013, pp. 1194-1201.
40
Optimization Approach
• Solution technique:
41
Optimization Approach
– Lagrange-Newton method:
42