Optimal power flow in an unbalanced active

distribution network
Sanat Kumar Paul1, †,
1
National Institute of Technology, Silchar, Assam 788010, India

Abstract: Recently, with the inclusion of distributed generation (DG) in distribution line, the scenario of a distribution
Network (DN) has changed drastically. A proper setting of the real power & voltage magnitude of DG’s is necessary for
optimal power flow & to maintain various network constraints. DN is inherently unbalanced in nature with mutual
coupling between the lines. An optimization problem is formed using convex relaxation & using KCL, KVL equations
for an active unbalanced DN. From the solution of the optimization problem, the settings of the DG’s are found for a
particular time duration & its robustness is verified for fluctuation in loads using Monte Carlo Simulations (MCS). The
method is tested on IEEE-34 bus unbalanced radial DN to find its effectiveness.

Key words: Setting of Distributed generators, Optimal power flow, Optimization in active unbalanced distribution
network technology.

1. Introduction
Programmable Balancing generation & load has been the foremost requirement for electrical power
system. Optimal power flow (OPF) is one of the main tool for this purpose, which has been used
with different objectives, e.g loss minimization, generation cost minimization etc. The main features
of a DN are that, it is inherently unbalanced, high R/X ratio, mutual coupling between lines & finally
it is mostly radial in structure. Also the lines may not have to be three phase lines in all portions.
With the inclusion single phase DG’s, this situation gets further stressed. Various recent research
works in DN has been discussed in[1]. The OPF for unbalanced DN is there in [2-5]. But in all these
papers the effect of coupling is missing, whose presence due to proximity of lines can’t be ignored.
Where as convergence is an issue in[2], DG setting & load fluctuation does not find its presence
in[3,4,5] . And reference [5] suffers from many assumptions. The main contribution of the paper is:

• It provides a setting for DG scheduling in ADN, which can take care of load fluctuations;
• This is one its first kind to the best of the knowledge of the author;
• With this setting, DSO can do hour ahead scheduling just like TSO;
• It takes care of all the attributes of a DN, which are mentioned earlier;
• It uses convex relaxation; but recovery of node voltages after solution does not required.
That also makes it one of its first kind;
• A verification is also done to check its robustness using MCS with load fluctuations.
The rest of the paper is organized as follows: Section II discuses the problem formulation; Section III
for test case scenario; Section IV for results & verification & Section V concludes the paper with
future work.

† Correspondence to: [Link], Email: sanatkumarpaul@[Link]

2. Problem formulation
A. Rotated Conic Constraints & Angular Displacement

The voltage of an ith bus & φth phase is written as

(1)
φ φ φ φ
where φ ∈{a,b,c}, Ci = cosθi ,Si = sinθi So the rotated conic constraints [6], [7] becomes
(2)

Where, N is the number of buses. Now in three phases line there is angular displacement between
the phases. In case of unbalanced case [9] these angles can’t be known a beforehand. Let αib & αic be
the angular displacement between phase a,b, & b,c. Then,

(3)

Where, & represents respectively.

and, (4)
Similar equations are written for “c” phase also. For sub-station bus “0”, which is considered as
slack bus, the voltage phase angle & angular displacement is known. Hence,

B. Charging current
The charging current is calculated as it is done in reference [8],

(5)

where is the charging current of ith bus & Baa, Bab are self & mutual
susceptance of ijth line respectively. Separating the real & imaginary part of ISH, ia

(6)

here j ∈ N(i) means all the lines connected to ith bus. Similarly the imaginary part & equations for ‘b’
& ‘c’ phases are written.

C. Load current
The loads in this case is assumed to be of constant power type. The output of DG’s are considered
as negative load. So the load current is calculated as [8],

(7)
where represents the load current. , means the real & reactive power of
load demand& DG’s output respectively in ith bus & φth phase . Separating the real & imaginary part
of IL, ia ,

Similarly the equation for imaginary part is written.

D. KCL Equation
KCL has been considered separately for the real & imaginary part of the load current. So the KCL
for lth line between ith & jth bus is written as,

∀ ij ∈ lines,l = 1,...,L; φ ∈{a,b,c} (8)

where is the line current of lth line between ith & jth bus for φth phase. L denotes the total number of
lines. Similarly the equation for imaginary part is written.
E. KVL Equation
The KVL equations between ith & jth bus is expressed as,

(9)

Where are the series self & mutual impedance of ijth line. Separating the real
& imaginary part,

(10)

Similar equations can be written imaginary part & for b & c phases.

G. Voltage & Current constraints

At substation the voltage is considered as 1.05pu. So the voltage & current constraints are,

∀ φ ∈{a,b,c} (11)
0.95 ≤ Vjφ ≤ 1.05 ∀ i = 1,...,N – 1 (12)

(13)

Where is the maximum allowable limit of current through ijth line.

H. Objective function

For objective function let’s consider a single line diagram, where represents the series
admittance without mutual coupling. Then the real power flow between the buses are,

So to minimize power loss, can be maximized. Therefore the objective is to maximize

(14)
Hence the final optimization problem becomes,
Maximize:

Subject to:
1. Conic quadratic constraints given by “(2)” ∀ i = 0,...,N − 1; φ ∈{a,b,c};
2. Angular displacement constraints defined by “(3)- (4)” ∀ i = 0,...,N – 1;
3. KCL equality constraint given by “(8)” ∀ lines, l ∀ ij ∈ 1,...,L; φ ∈{a,b,c};
4. KVL equality constraint given by “(10)” ∀ ij lines, l = 1,...,L; φ ∈{a,b,c};
5. Bus voltage limits defined by “(11)-(12)” ∀ i = 0,...,N − 1;
6. Current limit constraints given by “(13)” ∀ ij ∈ lines, l = 1,...,L; φ ∈{a,b,c};

3. Test case scenario

IEEE-34 bus unbalanced system is taken as test case, without voltage regulators, considering loads as
spot load & of constant power type. Three 3φ reactive power support is provided at three location i.e

in Bus number 832, 844 and 848; with ratings {832,844,848}, φ

∈{a,b,c}. Single phase DG’s are considered to be connected at Bus number 832, 844, 848 of 100KW
each in phase a,b,c respectively. This is taken to create more stress in the system unbalance. All the
line bus data are taken from [10]. The problem is solved in GAMS [11] 25.1.3 using KNITRO solver.

4. Results and verification

From the solution of the optimization problem solved in GAMS, the setting for the P&V of the DG’s
are found for minimum loss. To check the robustness of these settings ±5% load fluctuation is
considered. These settings are tabulated in “Table. I”. It can be seen here that the real power setting
of DG is its maximum, which justifies the fact that local disbursing of power instead of feeding from
far substation end will minimize the loss. However with the change in network topology, DG
location & load concentration, the settings need not necessary to be at its maximum, & that can be
found from the solution of the problem. Table 1 shows the PV settings of DG’s for the test case

Table 1. PV Settings of DG’s for the test case

Case Settings Bus Number
19 26 28
Original Load Pg(pu) 1.0 1.0 1.0
V(pu) 0.9561 0.9505 0.9514
0.95 Load Pg(pu) 1.0 1.0 1.0
V(pu) 0.9557 0.9505 0.9512
1.05 Load Pg(pu) 1.0 1.0 1.0
V(pu) 0.9556 0.9505 0.9515

For verification of these results, load flow is done in MATLAB 2019a with a modified
Backward-Forward sweep method. In this method after every iteration voltage angle for DG
connected buses is retained, whereas it’s magnitude is set to the fixed setting. Then the reactive
power is calculated before starting next iteration.

To find a robust setting load flow is performed with the settings for 0.95load & 1.05load. With
first setting, the load flow is run for 20 different cases with any random combination of load
between 0.95 to 1.05. It is found that in all the cases the voltage limit is violated. With 1.05load
setting the load flow is performed for 200 times with each time different random fluctuation of load
between 0.95 to 1.05. And with these second settings it is found that every time the load flow is
converging & also no constraints are being violated. Hence the settings for 1.05 load is a robust one
& can be chosen for operation during that interval.

4. Conclusion
This paper presents an optimization problem based on CP for an unbalanced DN with DG. Its
solution provides a robust setting of PV for DG’s in a particular time interval for a forecasted load &
generation. Then DSO can use these settings for an optimum operation of the network in the same
manner as it done by TSO. Therefore it can be implemented for real system operation. As future
work this formulation can also be used for DN reconfiguration problem.

References
[1] S. Dalhues et al., Research and practice of flexibility in distribution systems: A review, in CSEE Journal
of Power and Energy Systems, vol. 5, no. 3, pp. 285-294, Sept. 2019, doi:
10.17775/CSEEJPES.2019.00170.
[2] S. Bruno, S. Lamonaca, G. Rotondo, U. Stecchi and M. La Scala, Unbalanced three-phase optimal power
flow for smart grids, in IEEE Transactions on Industrial Electronics, vol. 58, no. 10, pp. 4504-4513, Oct.
2011, doi: 10.1109/TIE.2011.2106099.
[3] Y. Han et al., Fully distributed optimal power flow for unbalanced distribution networks based on ADMM,
2016 IEEE International Conference on Power System Technology (POWERCON), Wollongong, NSW,
 2016, pp. 1-6, doi: 10.1109/POWERCON.2016.7754000.
[4] T. R. Alsenani and S. Paudyal, Distributed approach for solving optimal power flow problems in
three-phase unbalanced distribution networks, 2018 Australasian Universities Power Engineering
Conference (AUPEC), Auckland, New Zealand, 2018, pp. 1-6, doi: 10.1109/AUPEC.2018.8757955.
[5] H. Yuan, F. Li, Y. Wei and J. Zhu, Novel linearized power flow and linearized OPF Models for active
distribution networks with application in distribution LMP, in IEEE Transactions on Smart Grid, vol. 9,
no. 1, pp. 438-448, Jan. 2018, doi: 10.1109/TSG.2016.2594814.
[6] R. A. Jabr, Radial distribution load flow using conic programming, in IEEE Transactions on Power
Systems, vol. 21, no. 3, pp. 1458-1459, Aug. 2006, doi: 10.1109/TPWRS.2006.879234.
[7] R. A. Jabr, R. Singh and B. C. Pal, Minimum loss network reconfiguration using mixed-integer convex
programming, in IEEE Transactions on Power Systems, vol. 27, no. 2, pp. 1106-1115, May 2012, doi:
10.1109/TPWRS.2011.2180406.
[8] P. Gangwar, S. N. Singh and S. Chakrabarti, ”Network reconfiguration for unbalanced distribution
systems,” TENCON 2017 - 2017 IEEE Region 10 Conference, Penang, 2017, pp. 3028-3032, doi:
10.1109/TENCON.2017.8228381.
[9] Zhai, H.F. Yang, M. Chen, B. Kang, N.. (2018). Dynamic reconfiguration of three-phase unbalanced
distribution networks. International Journal of Electrical Power Energy Systems. 99. 1-10.
10.1016/[Link].2017.12.027.
[10] D. T. Feeders, “IEEE PES distribution system analysis subcommittee,” OnlineAvailable:
[Link] 2011.