Received December 13, 2019, accepted January 23, 2020, date of publication January 29, 2020, date of current

version February 6, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.2970205

A General Fast Power Flow Algorithm for

Transmission and Distribution Networks
YANSONG WANG1 , HAO WU2 , HAILIANG XU 1, (Member, IEEE),
QIANG LI3 , AND SHUNCHAO LIU1
1 Collegeof New Energy, China University of Petroleum (East China), Qingdao 266580, China
2 StateGrid Linyi Power Supply Company, Linyi 276000, China
3 Research Institute, CNOOC, Beijing 100028, China

Corresponding author: Hailiang Xu ([email protected])

This work was supported in part by the National Key Research and Development Project of China under Grant 2018YFB0904800, in part
by the Key Research and Development Project of Shandong Province under Grant 2019GGX103045, and in part by the Fundamental
Research Funds for the Central Universities under Grant 19CX02016A.

ABSTRACT Fast power flow calculation is generally required in static security analysis and power system
optimal planning. And the classical DC power flow algorithm, due to its simple mathematical model,
linearized equation, and rapid solving speed, is widely utilized. However, the classical DC power flow
algorithm has been found to be only suitable for high voltage transmission networks with small branch
impedance ratio. To address this issue, a general fast power flow algorithm is proposed. Utilizing this method,
the classical DC power flow algorithm is firstly performed to obtain the initial values of the branch active
power flow. And the nodal voltage angles are then calculated by the established node-injected reactive power
equations. Secondly, through the identical transformation and approximate treatment of the node power flow
equation, a voltage offset calculation method for PQ nodes is proposed. Finally, the active power flow and
active power loss can be calculated based on the corrected phase angles and node voltages. Simulation studies
on standard IEEE power systems, such as the IEEE 33-bus and IEEE 118-bus systems, etc., were conducted
by the presented algorithm, compared with those by the back/forward sweep, classical DC power flow and
Newton-Raphson algorithms. It is indicated that, the proposed power flow algorithm has the superiority in
satisfactory calculation speed and non-sensitivity on the network impedance ratio.

INDEX TERMS Power flow algorithm, transmission and distribution network, high impedance ratio of
branch, bus voltage amplitude deviation.

I. INTRODUCTION transmission capability [10], [11], operation dispatching [5],

Power system load flow calculation is the basis of steady state and probabilistic load flow analysis [12], etc. However,
operation analysis of power system. The ac power system the classical DC power flow algorithm has been found to be
load flow calculation methods, such as the Newton-Raphson mainly suitable for high voltage transmission networks with
method and PQ decomposition method can be used to obtain small branch impedance, with its calculation error usually
the system’s power flow exactly. But the need for multi- being within 3%-10%, which is only considered acceptable
ple iterative operation, and the considerable calculation bur- for occasions with low precision requirements [13].
den, make them not so applicable in real power system In order to extend the applicable range of the DC power
security analysis. Instead, the classical DC flow algorithm, flow algorithm, a modified DC power flow algorithm based
due to its simple mathematical model, linearized equation, on the equivalent load model of network loss is proposed
rapid solving speed, is widely used in quick calculation in [14], [15], where the branch network loss is allocated and
occasions, such as power system planning [1], [2], security iterated until the equivalent load of each node converges.
and stability evaluation of power system [3]–[5], transaction It is worthwhile to point out that the simplified condition
price calculation of electricity market [6]–[9], analysis of could reduce the calculation accuracy of the classical DC
power flow. As presented in [16], ignoring the calculation of
The associate editor coordinating the review of this manuscript and nodal voltage may lead to the omission and miscalculation
approving it for publication was Jenny Mahoney. of checking results in static security analysis and power

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
23284 VOLUME 8, 2020
Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

market calculation. A normal DC power flow of the sys-

tem is executed to determine voltage angles in [17], with
a novel derivation of voltage amplitudes provided. In [18],
a new method is presented to improve the accuracy of the
DC power flow results while maintaining its computation
efficiency and linear formulation. The proposed method uses
the empirical knowledge of the system, including observed
voltage magnitudes and angles from historical data, to for-
mulate the correction terms. A method to find the linear
power flow with the minimum error is presented in [19],
where the formulation of the independent variables asso- FIGURE 1. The unit equivalent circuit between any two nodes.
ciated with the minimum linearization error is determined
by the distribution of state variables, i.e., v and θ . In [20],
a linear model considering the bus voltage magnitudes for According to Fig. 1, the branch (i, j) power flow equation
radial distribution systems is proposed, while in [21] a can be obtained by
novel DC power flow method with reactive power consid- (
erations is introduced. In [22], a decoupled fast power flow Pij = Vi2 (gi0 + gij ) − Vi Vj (bij sinθij + gij cosθij )
algorithm based on the traditional ac power flow is pro- (1)
Qij = −Vi2 (bi0 + bij ) + Vi Vj (bij cosθij − gij sinθij )
posed, and the voltage amplitude with high accuracy can be
obtained. where Pij and Qij are the active power and reactive powers
It is notable that the above literatures mainly focus on of the branch circuit (i, j); Vi is the voltage amplitude of
improving the DC power flow algorithm from aspects of node i; gij is the conductance parameter of series branch;
network loss calculation, node voltage calculation, power bij is the susceptance parameter of series branch; gi0 is the
flow estimation, etc. As a result, they are only applicable to conductance parameters of equivalent parallel branch; bi0 is
the distribution network with lower impedance ratio. Hence, the susceptance parameter of equivalent parallel branch; θij is
as the main contribution of this paper, a general fast power the phase angle difference between the sending and receiving
flow algorithm is put forward for transmission and distribu- ends of the branch (i, j).
tion networks, which considers the high impedance ratio of The branch power flow equation, i.e., (1) can be written in
the networks. a matrix form as
The paper is organized as follows. Section II describes the
Vj sin θij
 
Pij /Vi
   
basic concept of the branch equivalent power flow calcula- bij −gij −gi0 
=− Vi − Vj cos θij  (2)
tions. In Section III, the proposed fast power flow algorithm is Qij /Vi gij bij bi0
Vi
presented, while the detailed steps and layout of the proposed
method is provided in Section IV. In Section V, case studies According to (2), when carrying out the elementary trans-
are performed on standard IEEE power systems, such as the formation on the power flow equation matrix, and eliminating
IEEE 33-bus system, IEEE 14-bus system, IEEE 118-bus the row related to reactive power in the matrix, the active
systems, etc., to verify the correctness and feasibility of the power of the branch can be obtained as
proposed method. Finally, some useful conclusions are sum-
Vi Vj sin θij
 
marized in Section VI. rij rij
Pij = + Qij + Vi2 gi0 + bi0 (3)
xij xij xij
Similarly, the reactive power of the branch can be calcu-
II. BASIC CONCEPT OF THE BRANCH EQUIVALENT
lated by
POWER FLOW
−Vi Vj sin θij
 
In the power network, the unit equivalent circuit between any xij Pij 2 xij
 Qij = + − Vi bi0 + gi0 (4)
two nodes can be shown as Fig. 1, where yij = 1 zij = gij + rij rij rij
jbij , denotes the branch admittance.
Note that the power lines and transformers are usually III. THE PROPOSED FAST POWER FLOW ALGORITHM
be represented by π -type equivalent circuits, whereas the In a real power system, the difference of phase angle of node
capacitors and reactors can be equivalent to be series or par- voltage usually includes branch resistance parameters, with
allel branches, respectively. In the distribution network, if the the node voltage’s amplitude offset being around its rated
capacitor is installed in the substation, it can be equivalent value. Especially, the resistance of the distribution network
to a part of the substation node injecting power according to is usually much bigger than that of the transmission net-
the power factor; if the capacitor is installed in the middle of work. In this section, the calculation error with the classical
a line, it can be regarded as a parallel branch to the ground. DC power flow algorithm is firstly discussed. And then an
The parallel branches of the same node can be equivalent to improved fast power flow algorithm is put forward, with the
a branch. possible influencing factors being taken account.

VOLUME 8, 2020 23285

 Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

A. CLASSICAL DC POWER FLOW ALGORITHM Considering that the distribution network contains
The classical DC power flow algorithm is widely used in branches with high impedance ratio, it can be concluded from
power system. For instance, when a power network is in (8) that the error of applying the classical DC power flow
plan, the active power flow distribution is usually concerned, to solve the node voltage phase angle will become larger if
while the voltage amplitude of each node does not need to the branch resistance or reactive power increases. In order to
be considered. In such cases, the DC power flow algorithm is calculate the phase angle of node voltage more accurately,
usually adopted. In other words, when the calculation speed the effect of parallel branch resistance should be taken into
rather than the calculation accuracy is put in priority, the DC consideration. Assume that Vi = Vj = 1, and sin θij = θi −θj ,
power flow algorithm shows a satisfactory performance. then (3) and (4) can be simplified as
In the classical DC power flow algorithm, the paral- θi − θj
  
rij Qij rij
lel branches in the π -type equivalent circuit, as shown P = + + g + b

 ij i0 i0

xij xij xij

in Fig. 1 are usually ignored. And the following assumptions (9)
−(θi − θj ) xij Pij
 
are usually made: ¬ Vi = Vj = 1; ­ sin θij = θi − θj ; ® 
 Qij =

 + − bi0 + gi0
xij
rij = 0. Then, the active power flow equation of the branch rij rij rij
circuit (i, j) can be calculated by Note that the power injected into any node i is always equal
θi − θj to the algebraic sum of the branch power connected to the
Pij = (5) point. As a consequence, the power balance equation of node
xij
i can be obtained as
For a power network with n + 1 nodes (including a slack 
X n X n
θi − θj rij Qij

rij

SP
node), the matrix form of the classical DC power flow equa- P = P = + + g +b

ij i0 i0

 i

tion can be represented as
 x x
ij ij x ij
j=1(j6=i) j=1(j6=i)
n n
−(θi − θj ) xij Pij
 
SP
X X xij
P SP = B0 θ

(6) Q = Q = + + b +g


 i
 ij i0 i0
rij rij rij
j=1(j6=i) j=1(j6=i)
where P SP = [PSP 1 , · · · , Pi , · · · , Pn ] is the net injection
SP SP T
(10)
power of the node, and specified as a known condition; θ =
[θ1 , · · · , θi , · · · , θn ]T represents the voltage phase angle of n According to (10), the node-injected reactive power is
nodes; B0 is an n × n order node admittance matrix based coupled with the active power of the branch, the impedance
on branch reactance xij , and any element of the matrix can be of the series branch and the admittance parameters of the
expressed as parallel branch as well. As for PQ nodes, the node-injected
 n reactive power equation can be written in matrix form as
X 1 1
QSP = G0 θ + P cp + C Q


 − i=j (11)
x xii

B0 (i, j) = k=1 ik (7)
1 ,··· , QSP
i ,··· , QSP
SP
 1 where Q = [QSP n ]
T represents the
− i 6= j

 net injected reactive power into the nodes, and speci-
xij
fied as a known condition; G0 is an n × n order nodal
B. THE PROPOSED VOLTAGE PHASE ANGLE conductance matrix based on the branch resistance rij ;
CALCULATION METHOD CONSIDERING P cp = [Pcp1 , · · · , Pcpi , · · · , Pcpn ]T is defined as the node-
THE EFFECT OF BRANCH RESISTANCE branch power coupling term, which couples the node reac-
According to Fig. 1, the voltage drop generated by the series tive injection power and branch active power; C Q =
branches can be calculated by [CQ1 , . . . , CQi , . . . , CQn ]T denotes the node power-branch
parameter coupling term, which couples the node reactive
V̇j Î = Pj + jQj power and branch parameter.
1V̇ = V̇i − V̇j = 1V + jδV It is notable that any element of the matrix G0 can be
expressed as
Pj rij + Qj xij Pj xij − Qj rij
= +j (8)  n
X −1 1
Vj Vj 

 + i=j
rik rii

where V̇i denotes voltage phasor of node i; Î denotes the G0 (i, j) = k=1 (12)
conjugate of the current flowing through the impedance of

 1

 i 6= j
the series branch; 1V denotes the in-phase component of r ij
the voltage drop, which reflects the difference between the Similarly, any element of the coupling term P cp between
ends of the branch; δV is the quadrature component of the the node reactive power and branch active power can be
voltage drop, indicating the difference of the voltage phase obtained as
angle between the two ends of the branch; Pj and Qj are the Xn
xij
net injection active power and reactive power of node j; Vj is Pcpi = · P0ij (13)
rij
the voltage amplitude of node j. j=1(j6=i)

23286 VOLUME 8, 2020

Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

where P0ij is the initial value of the active power flowing into (17) can be obtained as
the branch (i, j), which can be calculated by the classical DC ωi X
X
power flow algorithm. =− ηij 1 + 1Vj − ηij Vj (i ∈ M) (17)
1 + 1Vi
In addition, any element of the coupling term CQ between j∈M j∈N
the node reactive power and branch parameters can be calcu-
where 1Vi denotes the voltage offset of node i; M =
lated by the following equation.
{1, 2, · · · · · · , m} represents the PQ node sets; ωi , ηij are the
n intermediate variables and can be defined as
X xij
CQi = bi0 + gi0 (14)
rij ωi = Qi cos θi − Pi sin θi . (18)
j=1(j6=i)
ηij = Bij cos θj + Gij sin θj . (19)
Note that in (11), QSP is given, P cp can be obtained by
DC load flow algorithm, and C Q can be obtained by branch Since the node voltage deviation of a power system, i.e., 1V ,
parameter. Consequently, the phase angle of the node voltage is usually within ±10% in its steady-state operation, it can be
can be obtained by solving (11), with the impedance of assumed that 1/(1 + 1Vi ) ≈ 1 − 1Vi , which would result in
parallel branch and branch being taken into account. an error less than ±0.999%.
Similarly, based on (10), the node-injected active power Regarding that the voltage amplitudes of the PV and V
and reactive power equation can be written as θ nodes are known, it is only necessary to modify the node
voltage amplitude of PQ nodes. For m PQ nodes, the left side
P = B0 θ + Qcp + C P
 SP

 of (17) can be linearized and approximated to the matrix form
 n

 X rij as
Q = · Qij


 cpi
xij
−η11 + ω1 . . . ... 1V1
  
j=1(j6=i) (15) −η1j −η1m
n ... ... ... ... ...   ... 

rij

 X 
 C Pi =

gi0 + bi0 . . . −ηij + ωi . . .  ·  1Vi 
   
 −ηi1 −ηim   

 xij
... ... ... ... ... ... 

j=1(j6=i)   
−ηm1 . . . −ηmj . . . −ηmm + ωm 1Vm
where Qcp = [Qcp1 , · · · , Qcpi , · · · , Qcpn ]T denotes the node- X X
ω1 + η1j + η1j Vj
 
branch power coupling term, which couples with the node
injected active power and branch reactive power; C P =
 j∈M j∈N 
X ... X
 
[CP1 , · · · , CPi , · · · , CPn ]T denotes the node power-branch
 
 
 ω + ηij + ηij Vj 
parameter coupling term. = i  (20)
As a result, if the branch reactive power Qij is given, j∈M j∈N
 
 
the phase angle of node voltage can be obtained by solv-

 X ... X 

ing (15), with the impedance of parallel branch and branch ω + ηmj + ηmj Vj 
m
being considered. j∈M j∈N

C. VOLTAGE AMPLITUDE CORRECTION METHOD BASED

The m-order coefficients matrix A and m-dimensional
ON NODAL POWER FLOW EQUATION
independent vector b are defined as
During normal operation conditions, the voltage amplitude of
(
−ηij i 6= j
each node is usually near its rated value. In order to obtain the A(i, j) =
precise value, the assumption, i.e., Vi = Vj = 1 needs to be −ηij + ωi i = j
X X
corrected by node voltage deviation. b(i) = ωi + ηij + ηij Vj (i ∈ M) (21)
As known, the injected power equation of a node can be j∈M j∈N
represented as
Then substituting (21) into (20), the node voltage correc-
n
Pi + jQi X tion equation can be obtained as
= Vj (Gij − jBij )(cos θj − j sin θj ) (16)
V̇i j=1 A · 1V = b (22)
where Pi , Qi are the net injected active and reactive powers where 1V is m-dimensional voltage offset of the PQ nodes
of node i; V̇i , Vj are the node voltage phase and magnitude, and can be obtained by solving (22). Then the voltage ampli-
respectively; Gij , Bij are the real and imaginary parts of the tude of PQ nodes can be modified as
node admittance element, respectively; n is the node number
in the system. V = 1 + 1V (23)
Since the voltage amplitudes of the PV and Vθ nodes are
known, it is only necessary to modify the voltage amplitude where V denotes the corrected voltage amplitude of m PQ
of the PQ nodes. By sorting out the imaginary part of (16), nodes; 1 represents m-dimension all unit column vector.

VOLUME 8, 2020 23287

 Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

IV. STEPS AND ARRANGEMENT OF THE PROPOSED FAST

POWER FLOW CALCULATION
On the basis of the classical DC power load flow algorithm,
the calculation method of node voltage phase angle and the
correction method of node voltage amplitude are proposed
by modifying the assumptions of DC power load flow calcu-
lation. Without iteration, the power flow distribution can be
solved quickly. The implementation steps and process of the
algorithm are as follows.
Step I: initialize the network parameters
According to the network topology, the nodes are num-
bered, and the branch is represented by ij, where i and j
denote the starting and receiving points of the branch, respec-
tively, with i < j. Then the network parameters, including the
resistance, reactance and admittance of each branch, can be
initialized.
Step II: set the initial value of the variables
Since the injected power of PQ and PV nodes are given,
the equilibrium node and the initial voltage amplitude of each
node can be set conveniently.
Step III: calculate the initial power of the branch
Establish the nodal admittance matrix considering the
branch resistance and the nodal admittance matrix neglect-
ing the branch resistance, respectively. The classical DC
power flow algorithm is used to calculate the branch
active power flow, and set as the initial value of the branch
power.
Step IV: calculate the phase angle of the node voltage
The node-branch power coupling term P cp and the node
power-branch parameter coupling term C Q are calculated.
The node conductance matrix G0 can then be formed. Hence,
based on these coupling terms (P cp , C Q ) and node admit- FIGURE 2. Flow chart of the fast power flow algorithm.
tance matrix (G0 ), the node voltage phase angle (θ ) can be
calculated.
Step V: calculate the node voltage offset and correct the power at both ends of the branch, which is given as
voltage amplitude
For m PQ nodes, the m-dimensional coefficient matrix A (
1Pij = Pij − Pji
and the m-dimensional independent vector b can be formed X (24)
by calculating the intermediate variables. The voltage ampli- 1P = 1Pij
tude offset 1V of PQ node is obtained by solving the linear
voltage correction equation. And then the voltage amplitude
V of the PQ node is modified. V. CASE STUDIES AND DISCUSSION
Step VI: calculate the injected reactive power of PV The effectiveness and feasibility of the proposed algorithm
node were tested and validated by simulation studies. Firstly,
As for PV nodes, the injected reactive power is cal- the standard IEEE 33-bus system (open distribution net-
culated. If the reactive power exceeds the limit, the PV work) and IEEE 118-bus system (closed transmission net-
nodes will be converted into PQ nodes, and update the work) were selected as tested examples. And the classical
known input variables of the network. Then return back to DC power plow algorithm, back/forward sweep algorithm,
Step IV. Newton-Raphson algorithm and other algorithms were also
Step VII: calculate the power flow of the branches and carried out for comparison.
the total network power loss Note that, for the open distribution network with high
Based on the obtained node voltage phase angle, i.e., θ and impedance ratio, the back/forward sweep algorithm is
the modified node voltage amplitude V , the branch active and selected as the standard power flow algorithm. And for the
reactive power flow can be obtained. closed power transmission networks with low impedance
Finally, the branch power loss and the whole network ratio, the Newton-Raphson algorithm is chosen as the stan-
power loss can be calculated by the difference of the active dard power flow algorithm. The calculation results of the

23288 VOLUME 8, 2020

Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

FIGURE 3. Comparison of node voltage phase angle in the IEEE 33-bus

system. FIGURE 4. Comparison of node voltage phase angle in the IEEE 14-bus
system.

standard power flow algorithms are then selected as the ref-

erence values.
Definite δi(j) to be the relative error between the calculated
and the referred power flow. And suppose σ to represent the
total deviation between the calculated and the referred value
of power flow, i.e.,
αi(j) − αi(j)
AC
δi(j) = (25)
αi(j)
AC
v
u n
u1 X
σ =t (δi(j) )2 (26)
n FIGURE 5. Comparison of node voltage phase angle in the IEEE 30-bus
i=1
system.
where αi(j) represents the calculated power flow of node i or
branch (i, j), while αi(j)
AC denotes the power flow obtained by

the standard power flow algorithm; and n is the number of rated value, i.e., 1p.u., and the deviation of the node voltage,
nodes. i.e., 1V is within ±10%. However, the real distribution net-
works are usually with tree structures, and the node voltage
A. ANALYSIS OF THE PHASE ANGLE CORRECTION OF THE of the tree branch will decline with the distance increasing
NODE VOLTAGE from the first node. The node voltage offset at the end of the
In order to study whether the proposed node voltage phase branch, which is far from the head of the power supply, may
angle calculation method is applicable to the distribution net- exceed −10%, such as the nodes numbered 14 to 18, 28 and
work with high impedance ratio, the IEEE 33-bus distribution 33 in the IEEE 33-bus system.
network was studied, with its node number being shown In order to validate whether the proposed algorithm is also
in Appendix. Note that in such network, about 78% of the suitable for the calculation of transmission networks with
branch impedance ratio is greater than 1, and almost 10% of low impedance ratio, testes on the IEEE 14-bus and IEEE
the branch impedance ratio is greater than 3. The calculation 30-bus transmission systems were then performed. And the
results were compared with those from the classical DC calculation results are compared with those from the classical
power flow algorithm and the back/forward sweep algorithm DC power flow algorithm, as shown in Figs. 4 and 5.
as shown in Fig. 3. Furthermore, the IEEE 118-bus transmission system is
As can be seen from Fig. 3, for the distribution network studied with the proposed method and the classic DC flow
with high impedance ratio, the calculated node voltage phase algorithm. Considering that the nodes and routes are rel-
angle with the classical DC power flow algorithm deviates atively more, the results obtained by the Newton-Raphson
greatly from those by the back/forward sweep algorithm, algorithm are regard as the reference values, as shown
indicating that the classical DC power flow algorithm is not in Fig. 6.
suitable for the distribution network with high impedance Based on the tests aforementioned, the calculation errors
ratio. In contrast, the calculation results obtained through of the phase angle are then summarized in Tab. 1, where the
the proposed algorithm are quite similar to those with the ‘‘average relative error’’ denotes the average value of all δi(j)
standard algorithm, which means that the proposed algorithm in the tested system.
is applicable in such high impedance ratio networks. As can be seen from Tab. 1, compared with the reference
It is notable that the node voltage phase angle calcula- value of the standard Newton-Rapson algorithm, the average
tion method assumes that the node voltage amplitude is its relative error and total deviation degree of the voltage phase

VOLUME 8, 2020 23289

 Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

FIGURE 8. Comparison of node voltage amplitude in the IEEE 14-bus

system.
FIGURE 6. Comparison of voltage phase angle in the IEEE 118-bus system.

TABLE 1. Calculation errors of phase angle in example system.

FIGURE 9. Comparison of node voltage amplitude in the IEEE 30-bus

system.

It can be observed from Fig. 7 that, compared with the

voltage amplitude reference value of back/forward sweep
algorithm, the voltage amplitude calculation error by the
fast power flow algorithm proposed in this paper is slightly
smaller than that of the classical DC power flow algorithm,
especially in the high number nodes.
In order to demonstrate the applicability of this algorithm
in transmission network, the IEEE 14-bus system and IEEE
30-bus systems are then used as tested networks. The node
voltage amplitudes obtained by the fast power flow algorithm
FIGURE 7. Comparison of node voltage amplitude in the IEEE 33-bus
system.
were compared with those by the standard Newton-Raphson
algorithm, as shown in Figs. 8 and 9.
Further studies are carried out on the voltage amplitude
of each PQ nodes of the IEEE 118-bus transmission system,
angles by the proposed algorithm in all the IEEE 14-bus comparing the proposed fast power flow algorithm and the
system, IEEE 30-bus system, IEEE 33-bus system, IEEE classical DC algorithm, as depicted in Fig. 10. From Fig. 10,
118-bus systems are smaller than that of the classical DC it can be concluded that the voltage amplitude of each node
power flow algorithm. In other words, the fast power flow obtained by fast power flow algorithm is consistent very
algorithm presented in this paper can obtain a more accuracy well with the reference value that acquired by the standard
power flow result. Newton-Raphson algorithm.
To quantitatively evaluate the correctness of the proposed
B. ANALYSIS OF THE NODE VOLTAGE AMPLITUDE algorithm, the node-voltage calculation errors compared with
CALCULATION RESULTS the back/forward sweep algorithm and Newton-Raphson
The proposed algorithm is firstly performed on the IEEE algorithm were summarized in Tab. 2. As shown, the max-
33-bus system with high impedance ratio, and the results are imum relative error, minimum relative error, average relative
compared with those from the classical DC power algorithm error and total deviation of node voltage amplitude by the
and back/forward sweep algorithm, as shown in Fig. 7. fast power flow algorithm are all under the level of 10−3 ,

23290 VOLUME 8, 2020

Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

FIGURE 10. Comparison of node voltage amplitude in the IEEE 118-bus FIGURE 12. Active and reactive power flow in the IEEE 14-bus system.
system.

TABLE 2. Calculation error of voltage amplitude in example system.

FIGURE 13. Active and reactive power flow in the IEEE 30-bus system.

FIGURE 11. Active and reactive power flow in the IEEE 33-bus system.

FIGURE 14. Active and reactive power flow in the IEEE 118-bus system.

indicating that the proposed method has a satisfactory cal- Furthermore, tests are carried out on the IEEE 14-bus
culation accuracy. system, IEEE 30-bus system and IEEE 118-bus systems, with
the results obtained by the standard Newton-Raphson algo-
C. ANALYSIS OF THE CALCULATION RESULTS OF THE rithm being the reference values, as shown in Figs. 12 to 14.
ACTIVE AND REACTIVE POWER FLOW OF As shown in the figures, the calculation results acquired by
BRANCH CIRCUITS the proposed method coincide very well with those from the
Firstly, to validate the effectiveness of the proposed method, Newton-Raphson algorithm.
the fast power flow algorithm is applied to calculate the Similarly, to give a qualitatively comparison, the calcula-
active and reactive power flow of the IEEE 33-bus system, tion results in Fig. 14 are further handled and summarized in
compared with the back/forward sweep algorithm, as shown Tab. 3. It shows that compared with the classical DC power
in Fig. 11. It can be seen from the figure that the calcu- flow algorithm, the total deviation of the calculated active
lation results of active power flow by proposed algorithm power flow calculated by the fast power flow algorithm is
are in accordance very well with the reference values by much smaller. And the error of the active power flow in heavy
the back/forward sweep algorithm, so do that of the reactive active power branch is also smaller than that of the classical
power flow. DC power flow algorithm. Besides, the calculation results of

VOLUME 8, 2020 23291

 Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

TABLE 3. Calculation error of active and reactive power flow in IEEE

118-bus system.

FIGURE 15. Topology diagram of IEEE 33-bus system with renumbered

node number.

are much smaller than those of the network loss iterative

algorithm in [14].
To illustrate the in computational efficiency of the pro-
posed algorithm, comparison is performed between the pro-
posed algorithm and the back/forward sweep algorithm by
taking the IEEE 33-bus system as an example of distribution
network. Meanwhile, the IEEE 118-bus system is selected as
TABLE 4. Comparison of network loss in IEEE 118-bus system. an example to compare the computational efficiency between
the standard Newton-Raphson algorithm and the proposed
one. The results are summarized in Tab. 5, where the com-
putational efficiency of the proposed algorithm in both trans-
mission and distribution networks can be verified.

VI. CONCLUSION
Through analyzing the factors that affects the accuracy of the
classical DC power flow algorithm, a fast power flow algo-
rithm is proposed which adopts the linear equation structure
of the classical DC power flow algorithm. While maintaining
the characteristics of fast calculation speed, the proposed
method improves the calculation accuracy and shows non-
sensitivity on the impedance ratio. Conclusions can be sum-
TABLE 5. Comparison of computational efficiency in IEEE systems. marized as follows.
(1) The node-injected reactive power matrix based on the
branch reactive power flow equation is re-established
with the branch active power and impedance parame-
ters being considered, which makes up for the neglect
of resistance and parallel branch in the classical DC
power flow equation.
(2) The proposed approaches for calculating the voltage
phase angle and correcting the voltage offset are both
based on linearized equations, and thus the calcula-
tion process does not include iteration. Consequently,
the algorithm can be easily applied to the on-line quasi-
real-time power flow analysis.
reactive power flow obtained by the proposed algorithm also (3) The proposed fast power flow algorithm does not
have a high accuracy. depend by the network topologies and branch
impedance ratios, and is suitable for both the dis-
D. ANALYSIS OF THE BRANCH NETWORK LOSS tribution and the transmission networks. Test results
In order to analyze the branch network loss based on the fast on the standard IEEE power systems indicate that,
power flow algorithm, the IEEE118-bus system is selected. the proposed method has satisfactory accuracy and
Again, the results by the Newton-Raphson algorithm are good engineering application prospect as well.
regarded as reference values. Then the proposed algorithm (4) It should be noted that, the proposed algorithm does
is compared with that based on network loss equivalent load not consider the case of PV nodes in medium and low
model in [14]. The operational efficiency and calculation voltage distribution network. With more and more DGs
error of branch network loss are concluded in Tab. 4. It can connected to the distribution network, the influence of
be seen that the proposed fast power flow algorithm saves PV operation mode needs to be considered and will be
more computing resources since it does not need the iterative studied in our future work.
calculation. And the maximum relative error, the minimum
relative error, the mean relative error and the total deviation of APPENDIX
the branch network loss, achieved by the proposed algorithm See Figure 15.

23292 VOLUME 8, 2020

Y. Wang et al.: General Fast Power Flow Algorithm for Transmission and Distribution Networks

ACKNOWLEDGMENT [22] J. Yang, N. Zhang, C. Kang, and Q. Xia, ‘‘A state-independent linear power
The authors would like to thank all the reviewers for their flow model with accurate estimation of voltage magnitude,’’ IEEE Trans.
Power Syst., vol. 32, no. 5, pp. 3607–3617, Sep. 2017.
valuable and insightful comments, which have improved their
manuscript substantially.

REFERENCES YANSONG WANG received the B.S. degree in

[1] W. Wu, Z. Hu, Y. Song, G. Sansavini, H. Chen, and X. Chen, ‘‘Trans- electrical engineering from Shandong University,
mission network expansion planning based on chronological evaluation Jinan, China, in 1988, and the M.S. and Ph.D.
considering wind power uncertainties,’’ IEEE Trans. Power Syst., vol. 33, degrees in electrical engineering from the China
no. 5, pp. 4787–4796, Sep. 2018. University of Petroleum (East China), Qingdao,
[2] X. Qiu, Z. Zhang, and Q. Wei, ‘‘An improved differential evolution algo- China, in 1998 and 2005, respectively.
rithm for transmission network planning,’’ in Proc. 4th Int. Conf. Electric She is currently a Professor with the China
Utility Deregulation Restructuring Power Technol. (DRPT), Weihai, China, University of Petroleum (East China). Her current
Jul. 2011, pp. 1246–1249. research interests include power quality analysis,
[3] R. Billinton and R. N. Allan, Reliability Evaluation of Engineering Sys- harmonic suppression, power grid optimization
tems. New York, NY, USA: Springer, 1992. planning, power network fault diagnosis, and power load forecasting.
[4] M. K. Enns, J. J. Quada, and B. Sackett, ‘‘Fast linear contingency analysis,’’
IEEE Trans. Power Syst., vol. PWRS-101, no. 4, pp. 783–791, Apr. 1982.
[5] M. Chavez-Lugo, C. R. Fuerte-Esquivel, C. A. Canizares, and
V. J. Gutierrez-Martinez, ‘‘Practical security boundary-constrained
DC optimal power flow for electricity markets,’’ IEEE Trans. Power Syst., HAO WU received the B.S. and M.S. degrees
vol. 31, no. 5, pp. 3358–3368, Sep. 2016. in electrical engineering from the China Univer-
[6] F. Li and R. Bo, ‘‘DCOPF-based LMP simulation: Algorithm, comparison sity of Petroleum, Qingdao, China, in 2015 and
with ACOPF, and sensitivity,’’ IEEE Trans. Power Syst., vol. 22, no. 4, 2018, respectively. Since 2018, she has been with
pp. 1475–1485, Nov. 2007. State Grid Linyi Power Supply Company, Linyi,
[7] T. J. Overbye, X. Cheng, and Y. Sun, ‘‘A comparison of the AC and DC China. Her research interests include power mar-
power flow models for LMP calculation,’’ in Proc. Int. Conf. Syst. Sci., kets and power system optimization, and power
Big Island, HI, USA, Jan. 2004, pp. 1–9. flow algorithm.
[8] A. Saxena, S. N. Pandey, and L. Srivastava, ‘‘DC-OPF based allocation
of wheeling prices for varying contribution of producers and customers,’’
in Proc. IEEE 1st Int. Conf. Power Electron., Intell. Control Energy Syst.
(ICPEICES), Delhi, India, July. 2016, pp. 1475–1485.
[9] M. B. Nappu, ‘‘LMP-lossless for congested power system based on
DC-OPF,’’ in Proc. Makassar Int. Conf. Electr. Eng. Inform. (MICEEI), HAILIANG XU (Member, IEEE) received the B.S.
South Sulawesi, Indonesia, Nov. 2014, pp. 194–199. degree in electrical engineering from the China
[10] H. Wang, S. Mu, Z. Zhang, and J. Zhang, ‘‘An improved DC power University of Petroleum (East China), Qingdao,
flow model with virtual transmission channel for restoring energy Internet China, in 2008, and the Ph.D. degree in electrical
power flow solvability,’’ in Proc. IEEE Conf. Energy Internet Energy Syst. engineering from Zhejiang University, Hangzhou,
Integr. (EI), Beijing, China, Nov. 2017, pp. 1–5. China, in 2014.
[11] R. Surya, N. Janarthanan, and S. Balamurugan, ‘‘A novel technique for con- Since 2018, he has been an Associate Profes-
gestion management in transmission system by real power flow control,’’ sor with the China University of Petroleum (East
in Proc. Int. Conf. Intell. Comput., Instrum. Control Technol. (ICICICT), China). His current research interests include wind
Kannur, India, July. 2017, pp. 1349–1354. power generation, microgrid, and power quality.
[12] Z. Hu and X. Wang, ‘‘A probabilistic load flow method considering branch
outages,’’ IEEE Trans. Power Syst., vol. 21, no. 2, pp. 507–514, May 2006.
[13] K. Purchala, L. Meeus, D. Van Dommelen, and R. Belmans, ‘‘Usefulness
of DC power flow for active power flow analysis,’’ in Proc. IEEE Power
Eng. Soc. Gen. Meeting, San Francisco, CA, USA, Jun. 2005, pp. 1–6. QIANG LI received the B.S. and M.S. degrees
[14] W. Hongfu, T. Xianghong, and Z. Zhiqiang, ‘‘An improved DC power flow in electrical engineering from the China Univer-
algorithm with consideration of network loss,’’ in Proc. Int. Conf. Power sity of Petroleum (East China), Qingdao, China,
Sys. Technol., Chengdu, China, Oct. 2014, pp. 455–460. in 2006 and 2009, respectively.
[15] J. W. Simpson-Porco, ‘‘Lossy DC power flow,’’ IEEE Trans. Power Syst., Since 2017, he has been a Senior Engineer with
vol. 33, no. 3, pp. 2477–2485, May 2018. the Research Institute, CNOOC, Beijing, China.
[16] Y. Qi, D. Shi, and D. Tylavsky, ‘‘Impact of assumption on DC power flow His current research interests include offshore oil
model accuracy,’’ in Proc. North Amer. Power Symp. (NAPS), Champaign, field power grid and power solution for offshore
IL, USA, Sep. 2012, pp. 1–6. oil filed from onshore power grid.
[17] T. Kyriakidis, R. Cherkaoui, and M. Kayal, ‘‘A DC power flow extension,’’
in Proc. IEEE PES ISGT Eur., Lyngby, Denmark, Oct.2013, pp. 1–5.
[18] S. Lu, N. Zhou, N. P. Kumar, N. Samaan, and B. B. Chakrabarti, ‘‘Improved
DC power flow method based on empirical knowledge of the system,’’ in
Proc. IEEE PES T&D, New Orleans, LA, USA, Apr. 2010, pp. 1–6.
SHUNCHAO LIU received the B.S. degree in
[19] Z. Yang, K. Xie, J. Yu, H. Zhong, N. Zhang, and Q. X. Xia, ‘‘A general
formulation of linear power flow models: Basic theory and error analysis,’’ engineering from the School of Electrical Engi-
IEEE Trans. Power Syst., vol. 34, no. 2, pp. 1315–1324, Mar. 2019. neering, Fuzhou University, in 2018. He is
[20] S. Bolognani and S. Zampieri, ‘‘On the existence and linear approximation currently pursuing the master’s degree in elec-
of the power flow solution in power distribution networks,’’ IEEE Trans. trical engineering with the China University of
Power Syst., vol. 31, no. 1, pp. 163–172, Jan. 2016. Petroleum (East China). His current research inter-
[21] S. M. Fatemi, S. Abedi, G. B. Gharehpetian, S. H. Hosseinian, and est includes comprehensive energy optimization
M. Abedi, ‘‘Introducing a novel DC power flow method with reac- planning.
tive power considerations,’’ IEEE Trans. Power Syst., vol. 30, no. 6,
pp. 3012–3023, Nov. 2015.

VOLUME 8, 2020 23293