Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Available
Available online
online
Energy atatwww.sciencedirect.com
Procedia
www.sciencedirect.com
00 (2018) 000–000
Energy Procedia 00 (2018) 000–000 www.elsevier.com/locate/procedia
ScienceDirect
ScienceDirect www.elsevier.com/locate/procedia

Energy
EnergyProcedia 158
Procedia 00(2019)
(2017)2524–2529
000–000
www.elsevier.com/locate/procedia
10th
International Conference on Applied Energy (ICAE2018), 22-25 August 2018, Hong Kong,
10 International Conference on Applied Energy
th
China(ICAE2018), 22-25 August 2018, Hong Kong,
China
A Combined Central and Local Voltage Control Strategy of Soft
A Combined Central
The 15th and Local
International Symposium Voltage
on District Control
Heating and Strategy
Cooling of Soft
Open Points in Active Distribution Networks
Open Points in Active Distribution Networks
Assessing the feasibility of using the heat demand-outdoor
Peng Lia, Haoran Jiaa, Guanyu Songa,a,*, Mingkun Yaoaa, Chengshan Wangaa, Jianzhong Wubb
a

temperature
Peng Li , Haoran Jifunction
, Guanyu Song
a
for a*,long-term
Mingkun Yao district
, Chengshan heatWang demand forecast
, Jianzhong
Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin 300072, China
Wu
a b
Institute ofofEnergy,
Key Laboratory School
Smart Grid of Engineering,
of Ministry CardiffTianjin
of Education, University, CardiffTianjin
University, CF24 3AA, UK China
300072,
I. Andrića,b,cInstitute
*, A.ofPina
b a
Energy,,School
P. Ferrão a
, J. Fournier
of Engineering,
b
., B.Cardiff
Cardiff University, Lacarrière c
, O. Le Correc
CF24 3AA, UK

Abstract
a
IN+ Center for Innovation, Technology and Policy Research - Instituto Superior Técnico, Av. Rovisco Pais 1, 1049-001 Lisbon, Portugal
Abstract b
Veolia Recherche & Innovation, 291 Avenue Dreyfous Daniel, 78520 Limay, France
With the increasingc penetration of distributed generation (DG), the risk of voltage violations in active distribution networks (ADNs)
Département Systèmes Énergétiques et Environnement - IMT Atlantique, 4 rue Alfred Kastler, 44300 Nantes, France
Withbecome
has the increasing
a major penetration
concern forofthedistributed generation
system operator. Soft(DG),
openthepointrisk(SOP)
of voltage violations
is a flexible in active
power distribution
electronic device networks
which can(ADNs)
realize
has become
accurate a major
active concern power
and reactive for the flow
system operator.
control. ThisSoft
paperopen point (SOP)
proposes is a flexible
a combined power
central electronic
and local devicestrategy
operation which of canSOPs
realize
to
accuratevoltage
realize active and reactive
control power The
in ADNs. flow active
control.power
This paper
of SOPs proposes a combined
is centrally adjustedcentral
basedandon local
the operation
informationstrategy
and of SOPs to
forecasting
realize
throughout
Abstractvoltage control inwhich
the network, ADNs. aimsThe active power
to maintain of SOPs
the voltage is centrally
within the limits adjusted basedoptimization.
in the global on the information
And the andlocalforecasting
control of
throughout
reactive power the based
network, which aims
on real-time to maintain the
measurements canvoltage
rapidly within
respondthetolimits in the global
the frequent voltageoptimization. And the
violations caused localfluctuations
by the control of
reactive
ofDistrict power
DG outputs.
heating based on real-time
Thenetworks
potential are measurements
benefits of SOPs
commonly arecan
fullyrapidly
addressed in the respond
explored to as
to reduce
literature thepower
frequent
one thevoltage
oflosses violationsvoltage
and improve
most effective caused byforthedecreasing
profile
solutions offluctuations
ADNs. By the
of DG outputs.
applying
greenhouse convex
gas The potentialfrom
relaxation,
emissions benefits
the of SOPs
original are fullyThese
mixed-integer
the building sector. explored
nonlinear
systemstoprogramming
reduce
requirepower losses
high(MINLP) andmodel
investments improve voltage
is converted
which profile of effectively
intothrough
are returned an ADNs. By
the heat
applying
solved convex
sales. mixed-integer
Due to the relaxation,
changed the
second-order original
climatecone mixed-integer
programming
conditions nonlinearrenovation
(MISOCP)
and building programming
model. Case (MINLP)
studies
policies, model
on the
heat PG&E
demand is converted
in69-node into could
an effectively
distribution
the future system
decrease,
solved
are mixed-integer
conducted to verifysecond-order
the cone
effectiveness
prolonging the investment return period. programming
of the proposed (MISOCP)
method. model. Case studies on the PG&E 69-node distribution system
are
Theconducted
main scope to verify
of thisthe effectiveness
paper is to assessofthe
thefeasibility
proposed method.
of using the heat demand – outdoor temperature function for heat demand
Copyright © 2018 Elsevier Ltd. All rights reserved.
© 2019 The Authors. Published by Elsevier Ltd. Lisbon (Portugal), was used as ath case study. The district is consisted of 665
forecast. The district of Alvalade, located in
Copyright
Selection
buildings ©that
and 2018 Elsevier
peer-review
vary inarticle
bothLtd.
under All rights reserved.
responsibility
construction of the
andscientific committee of the 10 International Conference on and
Applied
threeEnergy
This is an open access under the CC period
BY-NC-ND typology. Three weather scenarios (low, medium,
license (http://creativecommons.org/licenses/by-nc-nd/4.0/) high) district
Selection
(ICAE2018). and
renovation under peer-review
scenarios under
were developed responsibility
(shallow,of the scientific committee
intermediate, deep). To–ofestimate
the 10th International Conference on Applied Energy
Peer-review responsibility of the scientific committee of ICAE2018 The 10ththe error, obtained
International heat demand
Conference on Applied values were
Energy.
(ICAE2018).
compared with results from a dynamic heat demand model, previously developed and validated by the authors.
Keywords:
The results active distribution
showed network
that when (ADN);
only distributed
weather changegeneration (DG); soft
is considered, the open
marginpointof(SOP);
error local
couldvoltage control; mixed
be acceptable for -integer second-
some applications
order cone programming
Keywords:
(the erroractive distribution network (ADN); distributed generation (DG); soft open point (SOP); local voltage control; mixed -integer second-
in annual demand was lower than 20% for all weather scenarios considered). However, after introducing renovation
order cone programming
scenarios, the error value increased up to 59.5% (depending on the weather and renovation scenarios combination considered).
The value of slope coefficient increased on average within the range of 3.8% up to 8% per decade, that corresponds to the
1. Introduction
1.decrease in the number of heating hours of 22-139h during the heating season (depending on the combination of weather and
Introduction
With thescenarios
renovation increasing penetration
considered). Onof thedistributed
other hand,generation (DG), the
function intercept risk offor
increased voltage violations
7.8-12.7% in active
per decade distribution
(depending on the
With the
coupled
networks increasing
scenarios).
(ADNs) has penetration
Thebecome
values suggestedof distributed
a major could befor
concern generation
used thetosystem
modify (DG), thethe
operator risk
[1].ofIn
function voltage
currentviolations
parameters in active
for the scenarios
distribution distribution
considered,
networks, and
voltage
networks
improve the
violations (ADNs)
accuracy
generally hasofbecome
are heat a major
demand
mitigated concern for
byestimations.
dispatching the system
various operatorsuch
VAR devices [1]. In current
as the distribution
on-load tap changernetworks,
(OLTC) voltage
and
violations generally are mitigated by dispatching various VAR devices such as the on-load tap changer (OLTC) and
© 2017 The Authors. Published by Elsevier Ltd.
Peer-review under responsibility of the Scientific Committee of The 15th International Symposium on District Heating and
Cooling.
* Corresponding author. Tel.: +86 158 2283 1879; fax: +86 22 27892810.
*E-mail
Keywords: address:
Corresponding gysong@tju.edu.cn.
Heat demand;
author.Forecast;
Tel.: +86Climate
158 2283change
1879; fax: +86 22 27892810.
E-mail address: gysong@tju.edu.cn.
1876-6102 Copyright © 2018 Elsevier Ltd. All rights reserved.
Selection
1876-6102and peer-review
Copyright under
© 2018 responsibility
Elsevier Ltd. All of the scientific
rights reserved. committee of the 10th International Conference on Applied Energy (ICAE2018).
Selection and peer-review under responsibility of the scientific committee of the 10th International Conference on Applied Energy (ICAE2018).
1876-6102 © 2017 The Authors. Published by Elsevier Ltd.
Peer-review under responsibility of the Scientific Committee of The 15th International Symposium on District Heating and Cooling.
1876-6102 © 2019 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of ICAE2018 – The 10th International Conference on Applied Energy.
10.1016/j.egypro.2019.01.410
 Peng Li et al. / Energy Procedia 158 (2019) 2524–2529 2525
2 Author name / Energy Procedia 00 (2018) 000–000

Nomenclature
Sets  𝑉𝑉𝑖𝑖 q,min ,𝑉𝑉𝑖𝑖 q,max set points of the dead-zone in 𝑄𝑄 − 𝑉𝑉 curve
continuous/binary variables used in the
Ωb set of all branches 𝑎𝑎 𝑡𝑡,𝑖𝑖,𝑛𝑛 , 𝑑𝑑𝑡𝑡,𝑖𝑖,𝑛𝑛 
piecewise linearization of 𝑄𝑄 − 𝑉𝑉 curve
Indices Parameters
𝑖𝑖ǡ 𝑗𝑗 indices of nodes, from 1 to 𝑁𝑁N 𝑁𝑁T total periods of the time horizon
– indices of time periods, from 1 to 𝑁𝑁T 𝑁𝑁N total number of the nodes
Variables 𝑃𝑃𝑡𝑡,𝑖𝑖 ,𝑄𝑄𝑡𝑡,𝑖𝑖  active/reactive power consumption
LOAD LOAD

𝑃𝑃𝑡𝑡,𝑗𝑗𝑗𝑗 , 𝑄𝑄𝑡𝑡,𝑗𝑗𝑗𝑗 active/reactive power flow of branch 𝑆𝑆𝑖𝑖SOP capacity limit of SOP
𝐼𝐼𝑡𝑡,𝑖𝑖𝑖𝑖 ǡ𝐼𝐼2,𝑡𝑡,𝑗𝑗𝑗𝑗 branch current magnitude and its square 𝑅𝑅𝑖𝑖𝑖𝑖 ,𝑋𝑋𝑖𝑖𝑖𝑖  resistance/reactance of branch
𝑈𝑈𝑡𝑡,𝑖𝑖 ǡ𝑈𝑈2,𝑡𝑡,𝑖𝑖  node voltage magnitude and its square 𝑉𝑉 max min
,𝑉𝑉  upper/lower limit of statutory voltage
max min
𝑃𝑃𝑡𝑡,𝑖𝑖 ǡ𝑄𝑄𝑡𝑡,𝑖𝑖  total active/reactive power injection 𝑉𝑉thr ,𝑉𝑉thr  upper/lower limit of desired voltage range
SOP SOP
𝑃𝑃𝑡𝑡,𝑖𝑖 ǡ𝑄𝑄𝑡𝑡,𝑖𝑖  active/reactive power injection by SOP 𝐴𝐴SOP𝑖𝑖  loss coefficient of SOP
DG
𝑃𝑃𝑡𝑡,𝑖𝑖 ǡ 𝑄𝑄𝑡𝑡,𝑖𝑖DG active/reactive power capacity of DG 𝛼𝛼, 𝛽𝛽 weight coefficients
SOP,L
𝑃𝑃𝑡𝑡,𝑖𝑖  active power losses of SOP  
switchable capacitor banks (CBs), which usually provide a slow response and a discrete voltage regulation. However,
it is difficult to meet the requirement of the fast voltage control by such conventional VAR devices when DGs fluctuate
frequently in ADNs [2].However, the rapid development of power electronic technologies provides opportunities for
the further optimization of ADNs’ operation. Soft open point (SOP) is a novel power electronic device to realize the
flexible connection between feeders [3]. SOP can accurately realize the fast power flow control and continuous voltage
regulation. Thus, it is of significance to study the strategies of SOPs for voltage control in ADNs.
In general, the central control for SOPs is usually adopted in ADNs. The central control relies on the information
and forecasting of the whole system to realize global optimization, which heavily aggravates the computation and
communication burdens and consequently might hinder fast response in large networks. Whereas, the local control
method, compared with central control, has significant advantages of non-communication, high computation
efficiency and strong reliability, which is suitable for the real-time response to the fluctuations.
Thus, a combined central and local voltage control method of SOPs in ADNs is proposed in this paper. The active
power of SOPs is centrally adjusted based on the information and forecasting throughout the network, which aims to
maintain the voltage within the limits in the global optimization. And the local control of reactive power based on
real-time measurements can rapidly respond to the frequent voltage violations caused by the fluctuations of DG
outputs. The potential benefits of SOPs are fully explored to reduce power losses and improve voltage profile of ADNs.
By applying convex relaxation, the original mixed-integer nonlinear programming (MINLP) model is converted into
an effectively solved mixed-integer second-order cone programming (MISOCP) model. Finally, the effectiveness of
the combined method is validated on the PG&E 69-node distribution system.
2. Combined central and local operation strategy of SOPs
Day-ahead Schedule Inter-day Control
Determine the active power control strategies of SOPs

Adjust reactive power

Tune parameters of local curve of SOPs based on the local
voltage measurement N
2
Send active power strategies and local curves to SOPs SOP1

Fig. 1. Schematic of the combined central and local operation strategy of SOPs.

The schematic of the combined strategy of SOPs is shown in Fig.1. Based on the acquired information, the active
power control strategies and parameters of local control curves for SOPs can be determined by DMS in the day-ahead
schedule. While in the inter-day operation, SOPs adjust the reactive power compensation in real-time, based on the
local voltage measurements to maintain both the system losses and the voltages at the desired level.
2526 Peng Li et al. / Energy Procedia 158 (2019) 2524–2529
Author name / Energy Procedia 00 (2018) 000–000 3

3. Voltage control problem formulation with SOPs

In this section, a voltage control model with SOPs is built, which realizes the objective economic efficiency and a
desired voltage profile of ADNs. The widely used 𝑄𝑄 − 𝑉𝑉 curve [4] is adopted to realize the local reactive power
control of SOPs.
3.1. Mathematic description of combined strategy of SOPs
1) Objective function
A linear weighted combination of minimum total power losses and voltage deviations is taken as the objective
function, which is formulated as follows.
min 𝑓𝑓 = 𝛼𝛼𝑓𝑓L + 𝛽𝛽𝑓𝑓V (1)
where the power losses 𝑓𝑓L and the extent of voltage deviation 𝑓𝑓V are formulated as:
𝑓𝑓L = ∑𝑁𝑁T ∑ 2 ∑𝑁𝑁N SOP,L
𝑡𝑡=1 𝑖𝑖𝑖𝑖∈Ωb 𝑅𝑅𝑖𝑖𝑖𝑖 𝐼𝐼𝑡𝑡,𝑖𝑖𝑖𝑖 + 𝑖𝑖=1 𝑃𝑃𝑡𝑡,𝑖𝑖 (2)
𝑓𝑓V = ∑𝑁𝑁 T ∑𝑁𝑁N 2 max
𝑡𝑡=1 𝑖𝑖=1|𝑉𝑉𝑡𝑡,𝑖𝑖 − 1| : (𝑉𝑉𝑡𝑡,𝑖𝑖 ≥ 𝑉𝑉thr ||𝑉𝑉𝑡𝑡,𝑖𝑖 ≤ 𝑉𝑉thr )
min
(3)
Equation (3) indicates the threshold function reflecting the extent of voltage deviation. The weight coefficients 𝛼𝛼
and 𝛽𝛽 in equation (1) can be determined by analytic hierarchy process (AHP) and satisfy 𝛼𝛼 + 𝛽𝛽 = 1
2) System power flow constraints
∑𝑗𝑗𝑗𝑗∈Ωb(𝑃𝑃𝑡𝑡,𝑗𝑗𝑗𝑗 − 𝑅𝑅𝑗𝑗𝑗𝑗 𝐼𝐼𝑡𝑡,𝑗𝑗𝑗𝑗 2 ) + 𝑃𝑃 = ∑
𝑡𝑡,𝑖𝑖 𝑖𝑖𝑖𝑖∈Ωb 𝑃𝑃𝑡𝑡,𝑖𝑖𝑖𝑖 (4)
2
∑𝑗𝑗𝑗𝑗∈Ωb(𝑄𝑄𝑡𝑡,𝑗𝑗𝑗𝑗 − 𝑋𝑋𝑗𝑗𝑗𝑗 𝐼𝐼𝑡𝑡,𝑗𝑗𝑗𝑗 ) + 𝑄𝑄𝑡𝑡,𝑖𝑖 = ∑𝑖𝑖𝑖𝑖∈Ωb 𝑄𝑄𝑡𝑡,𝑖𝑖𝑖𝑖 (5)
𝑉𝑉𝑡𝑡,𝑖𝑖2 − 𝑉𝑉𝑡𝑡,𝑗𝑗
2 + (𝑅𝑅2 + 𝑋𝑋 2 )𝐼𝐼 2 = 2(𝑅𝑅 𝑃𝑃
𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑡𝑡,𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑡𝑡,𝑖𝑖𝑖𝑖 + 𝑋𝑋 𝑄𝑄
𝑖𝑖𝑖𝑖 𝑡𝑡,𝑖𝑖𝑖𝑖 ) (6)
2
𝐼𝐼𝑡𝑡,𝑖𝑖𝑖𝑖 𝑉𝑉𝑡𝑡,𝑖𝑖2 = 𝑃𝑃𝑡𝑡,𝑖𝑖𝑖𝑖
2 2
+ 𝑄𝑄𝑡𝑡,𝑖𝑖𝑖𝑖 (7)
DG + 𝑃𝑃 SOP − 𝑃𝑃 LOAD
𝑃𝑃𝑡𝑡,𝑖𝑖 = 𝑃𝑃𝑡𝑡,𝑖𝑖 𝑡𝑡,𝑖𝑖 𝑡𝑡,𝑖𝑖 (8)
DG SOP LOAD
𝑄𝑄𝑡𝑡,𝑖𝑖 = 𝑄𝑄𝑡𝑡,𝑖𝑖 + 𝑄𝑄𝑡𝑡,𝑖𝑖 − 𝑄𝑄𝑡𝑡,𝑖𝑖 (9)
Constraints (4) and (5) represent the active and reactive power balance of node i at period 𝑡𝑡 .The Ohm’s law over
branch ij at time 𝑡𝑡 is expressed as (6). The current magnitude of each line can be determined by (7). Constraints (8)
and (9) indicate the total active and reactive power injection of node 𝑖𝑖 at period 𝑡𝑡.
3) Secure operation constraints
(𝑉𝑉 min)2 ≤ 𝑉𝑉𝑡𝑡,𝑖𝑖2 ≤ (𝑉𝑉 max )2, 𝐼𝐼𝑡𝑡,𝑖𝑖𝑖𝑖 2 ≤ (𝐼𝐼 max )2
(10)
4) SOP operation constraints
The operation constraints of SOPs mainly include the central control of the active power of SOPs and the local Q-
V control of reactive power of SOPs.
Central control of the active power of SOPs:
SOP + 𝑃𝑃 SOP + 𝑃𝑃 SOP,L + 𝑃𝑃 SOP,L = 0
𝑃𝑃𝑡𝑡,𝑖𝑖 𝑡𝑡,𝑗𝑗 𝑡𝑡,𝑖𝑖 𝑡𝑡,𝑗𝑗 (11)
2 2 2 2
𝑃𝑃𝑡𝑡,𝑖𝑖SOP,L = 𝐴𝐴SOP
𝑖𝑖 √(𝑃𝑃𝑡𝑡,𝑖𝑖
SOP SOP , SOP,L
) + (𝑄𝑄𝑡𝑡,𝑖𝑖 ) 𝑃𝑃𝑡𝑡,𝑗𝑗 SOP
= 𝐴𝐴𝑗𝑗SOP√(𝑃𝑃𝑡𝑡,𝑗𝑗 SOP
) + (𝑄𝑄𝑡𝑡,𝑗𝑗 ) (12)
Local 𝑄𝑄 − 𝑉𝑉 curve control of the reactive power of SOPs:
SOP SOP
𝑄𝑄𝑡𝑡,𝑗𝑗
𝑄𝑄𝑡𝑡,𝑖𝑖
SOP,max = 𝜑𝜑(𝑉𝑉𝑡𝑡,𝑖𝑖 ), SOP,max = 𝑔𝑔(𝑉𝑉𝑡𝑡,𝑗𝑗 ) (13)
𝑄𝑄𝑡𝑡,𝑖𝑖 𝑄𝑄𝑡𝑡,𝑗𝑗
1.0 𝑉𝑉𝑡𝑡,𝑖𝑖 ϵ[0, 0.9)
1 𝑉𝑉𝑖𝑖q,min
q,min 𝑉𝑉𝑡𝑡,𝑖𝑖 + q,min 𝑉𝑉𝑡𝑡,𝑖𝑖 ϵ[0.9, 𝑉𝑉𝑖𝑖 q,min )
0.9−𝑉𝑉𝑖𝑖 𝑉𝑉𝑖𝑖 −0.9
𝜑𝜑(𝑉𝑉𝑡𝑡,𝑖𝑖 ) = 0 𝑉𝑉𝑡𝑡,𝑖𝑖 ϵ[𝑉𝑉𝑖𝑖 q,min,𝑉𝑉𝑖𝑖 q,max ) (14)
1 𝑉𝑉𝑖𝑖q,max
𝑉𝑉𝑡𝑡,𝑖𝑖 + 𝑉𝑉𝑡𝑡,𝑖𝑖 ϵ[𝑉𝑉𝑖𝑖q,max ,1.1)
𝑉𝑉𝑖𝑖q,max−1.1 1.1−𝑉𝑉𝑖𝑖q,max

{−1.0 𝑉𝑉𝑡𝑡,𝑖𝑖 ϵ[1.1, 1.2]

The 6-point broken line constitutes 𝑄𝑄 − 𝑉𝑉 curve, represented by mathematical expression 𝜑𝜑(𝑉𝑉𝑡𝑡,𝑖𝑖 ) and 𝑔𝑔(𝑉𝑉𝑡𝑡,𝑖𝑖 ). For
simplicity, 𝜑𝜑(𝑉𝑉𝑡𝑡,𝑖𝑖 ) is taken as an example to be explained, shown as constraint (14). [𝑉𝑉𝑖𝑖 q,min , 𝑉𝑉𝑖𝑖 q,max] is the dead-zones
 Peng Li et al. / Energy Procedia 158 (2019) 2524–2529 2527
4 Author name / Energy Procedia 00 (2018) 000–000

of the curve where inverters don’t supply reactive power. The voltage limits for maximum reactive power provision
and absorption are selected as 0.9 p.u. and 1.1 p.u. To determine 𝑄𝑄 − 𝑉𝑉 curve, only two parameters 𝑉𝑉𝑖𝑖q,min and 𝑉𝑉𝑖𝑖 q,max
are required to be set.
The reactive power outputs of SOPs should satisfy their own capacity constraints, described as (15).
SOP,max 2
SOP , SOP,max 2
𝑄𝑄𝑡𝑡,𝑖𝑖 = √(𝑆𝑆𝑖𝑖SOP)2 − (𝑃𝑃𝑡𝑡,𝑖𝑖 ) 𝑄𝑄𝑡𝑡,𝑗𝑗 = √(𝑆𝑆𝑗𝑗SOP)2 − (𝑃𝑃𝑡𝑡,𝑗𝑗
SOP
) (15)
As a consequence, constraints (1)-(15) form the optimization model of voltage control with SOPs. It is essentially
a large-scale MINLP problem, which requires to be solved accurately and efficiently.
3.2. Conversion to an MISOCP Model
In this section, the original MINLP model is converted into an MISOCP model using the convex relaxation. First,
let 𝑈𝑈2,𝑡𝑡,𝑖𝑖 and 𝐼𝐼2,𝑡𝑡,𝑖𝑖𝑖𝑖 denote the quadratic terms 𝑉𝑉𝑡𝑡,𝑖𝑖 2 2
and 𝐼𝐼𝑡𝑡,𝑖𝑖𝑖𝑖 . Linearized functions are expressed as follows:
∑𝑗𝑗𝑗𝑗∈Ω𝑏𝑏 (𝑃𝑃𝑡𝑡,𝑗𝑗𝑗𝑗 − 𝑅𝑅𝑗𝑗𝑗𝑗 𝐼𝐼2,𝑡𝑡,𝑗𝑗𝑗𝑗 ) + 𝑃𝑃𝑡𝑡,𝑖𝑖 = ∑𝑖𝑖𝑖𝑖∈Ω𝑏𝑏 𝑃𝑃𝑡𝑡,𝑖𝑖𝑖𝑖 (16)
∑𝑗𝑗𝑗𝑗∈Ω𝑏𝑏 (𝑄𝑄𝑡𝑡,𝑗𝑗𝑗𝑗 − 𝑋𝑋𝑗𝑗𝑗𝑗 𝐼𝐼2,𝑡𝑡,𝑗𝑗𝑗𝑗 ) + 𝑄𝑄𝑡𝑡,𝑖𝑖 = ∑𝑖𝑖𝑖𝑖∈Ω𝑏𝑏 𝑄𝑄𝑡𝑡,𝑖𝑖𝑖𝑖 (17)
𝑈𝑈2,𝑡𝑡,𝑖𝑖 − 𝑈𝑈2,𝑡𝑡,𝑗𝑗 + (𝑅𝑅2𝑖𝑖𝑖𝑖 + 𝑋𝑋𝑖𝑖𝑖𝑖2 )𝐼𝐼2,𝑡𝑡,𝑖𝑖𝑖𝑖 = 2(𝑅𝑅𝑖𝑖𝑖𝑖𝑃𝑃𝑡𝑡,𝑖𝑖𝑖𝑖 + 𝑋𝑋𝑖𝑖𝑖𝑖 𝑄𝑄𝑡𝑡,𝑖𝑖𝑖𝑖 ) (18)
2 2
𝐼𝐼2,𝑡𝑡,𝑗𝑗𝑗𝑗 𝑈𝑈2,𝑡𝑡,𝑖𝑖 = 𝑃𝑃𝑡𝑡,𝑖𝑖𝑖𝑖 + 𝑄𝑄𝑡𝑡,𝑖𝑖𝑖𝑖 (19)
min )2 ≤ 𝑈𝑈 max 2 max 2
(𝑉𝑉 2,𝑡𝑡,𝑖𝑖 ≤ (𝑉𝑉 ) , 𝐼𝐼2,𝑡𝑡,𝑖𝑖𝑖𝑖 ≤ (𝐼𝐼 ) (20)
𝑁𝑁T 𝑁𝑁N SOP,L
∑ ∑ ∑
𝑓𝑓L = 𝑡𝑡=1( 𝑖𝑖𝑖𝑖∈Ωb 𝑅𝑅𝑖𝑖𝑖𝑖 𝐼𝐼2,𝑡𝑡,𝑖𝑖𝑖𝑖 + 𝑖𝑖=1 𝑃𝑃𝑡𝑡,𝑖𝑖 ) (21)
Then (19) can be relaxed to the following second-order cone constraint:
T
‖[2𝑃𝑃𝑡𝑡,𝑖𝑖𝑖𝑖 2𝑄𝑄𝑡𝑡,𝑖𝑖𝑖𝑖 𝐼𝐼2,𝑡𝑡,𝑖𝑖𝑖𝑖 − 𝑈𝑈2,𝑡𝑡,𝑖𝑖 ] ‖ ≤ 𝐼𝐼2,𝑡𝑡,𝑖𝑖𝑖𝑖 + 𝑈𝑈2,𝑡𝑡,𝑖𝑖 (22)
2
Auxiliary variable 𝐴𝐴𝑡𝑡,𝑖𝑖 is introduced to linearized constraint (3). Some relevant constraints are added as follows.
𝑓𝑓V = ∑𝑁𝑁 T ∑𝑁𝑁N 𝐴𝐴
𝑡𝑡=1 𝑖𝑖=1 𝐴𝐴𝑡𝑡,𝑖𝑖 , 𝑡𝑡,𝑖𝑖 ≥ 0 (23)
max 2 2
𝐴𝐴𝑡𝑡,𝑖𝑖 ≥ 𝑈𝑈2,𝑡𝑡,𝑖𝑖 − (𝑉𝑉thr ) , 𝐴𝐴𝑡𝑡,𝑖𝑖 ≥ −𝑈𝑈2,𝑡𝑡,𝑖𝑖 + (𝑉𝑉thr ) min (24)
The expressions of curves 𝜑𝜑(𝑉𝑉𝑡𝑡,𝑖𝑖 ) and 𝑔𝑔(𝑉𝑉𝑡𝑡,𝑗𝑗 ) have been exactly modeled based on piecewise linearization [5]. For
simplicity, 𝜑𝜑(𝑉𝑉𝑡𝑡,𝑖𝑖 ) is taken as an example to be explained below. Continuous variables 𝑎𝑎 𝑡𝑡,𝑖𝑖,𝑛𝑛 and integer variables
𝑑𝑑𝑡𝑡,𝑖𝑖,𝑛𝑛 are introduced as follows.
𝑉𝑉𝑡𝑡,𝑖𝑖 = 0.8𝑎𝑎 𝑡𝑡,𝑖𝑖,1 + 0.9𝑎𝑎 𝑡𝑡,𝑖𝑖,2 + 𝑎𝑎 𝑡𝑡,𝑖𝑖,3 𝑉𝑉𝑖𝑖 q,min + 𝑎𝑎 𝑡𝑡,𝑖𝑖,4 𝑉𝑉𝑖𝑖 q,max + 1.1𝑎𝑎 𝑡𝑡,𝑖𝑖,5 + 1.2𝑎𝑎 𝑡𝑡,𝑖𝑖,6 (25)
𝜑𝜑(𝑉𝑉𝑡𝑡,𝑖𝑖 ) = 𝑎𝑎 𝑡𝑡,𝑖𝑖,1 + 𝑎𝑎 𝑡𝑡,𝑖𝑖,2 − 𝑎𝑎 𝑡𝑡,𝑖𝑖,5 − 𝑎𝑎 𝑡𝑡,𝑖𝑖,6 (26)
𝑎𝑎 𝑡𝑡,𝑖𝑖,1 ≤ 𝑑𝑑𝑡𝑡,𝑖𝑖,1, 𝑎𝑎 𝑡𝑡,𝑖𝑖,6 ≤ 𝑑𝑑𝑡𝑡,𝑖𝑖,5 (27)
𝑎𝑎 𝑡𝑡,𝑖𝑖,𝑛𝑛 ≤ 𝑑𝑑𝑡𝑡,𝑖𝑖,𝑛𝑛 + 𝑑𝑑𝑡𝑡,𝑖𝑖,𝑛𝑛−1 ,𝑛𝑛 = 2,3,4,5 (28)
𝑎𝑎 𝑡𝑡,𝑖𝑖,𝑛𝑛 ≥ 0, 𝑑𝑑𝑡𝑡,𝑖𝑖,𝑛𝑛 ∈ {0,1} (29)
∑6𝑛𝑛=1 𝑎𝑎 𝑡𝑡,𝑖𝑖,𝑛𝑛 = 1,∑5𝑛𝑛=1 𝑑𝑑𝑡𝑡,𝑖𝑖,𝑛𝑛 = 1 (30)
As for the nonlinear product terms 𝑎𝑎 q𝑡𝑡,𝑖𝑖,3 𝑉𝑉𝑖𝑖q,min , 𝑎𝑎 q𝑡𝑡,𝑖𝑖,4 𝑉𝑉𝑖𝑖 q,max , integer variables 𝑐𝑐𝑖𝑖,1 and 𝑐𝑐𝑖𝑖,2 are introduced.
𝑎𝑎 𝑡𝑡,𝑖𝑖,3 𝑉𝑉𝑖𝑖q,min = 0.90𝑎𝑎 𝑡𝑡,𝑖𝑖,3 + 0.01𝑎𝑎 𝑡𝑡,𝑖𝑖,3 𝑐𝑐𝑖𝑖,1 ,0 ≤ 𝑐𝑐𝑖𝑖,1 ≤ 20 (31)
𝑎𝑎 𝑡𝑡,𝑖𝑖,4 𝑉𝑉𝑖𝑖q,max = 0.90𝑎𝑎 𝑡𝑡,𝑖𝑖,4 + 0.01𝑎𝑎 𝑡𝑡,𝑖𝑖,4 𝑐𝑐𝑖𝑖,2 ,0 ≤ 𝑐𝑐𝑖𝑖,2 ≤ 20 (32)
Binary variables 𝑙𝑙 𝑖𝑖,1,𝑚𝑚 , 𝑙𝑙 𝑖𝑖,2,𝑚𝑚 are introduced as 𝑎𝑎 𝑡𝑡,𝑖𝑖,3 𝑐𝑐𝑖𝑖,1, 𝑎𝑎 𝑡𝑡,𝑖𝑖,4 𝑐𝑐𝑖𝑖,2 are still the nonlinear product terms.
𝑎𝑎 𝑡𝑡,𝑖𝑖,3 𝑐𝑐𝑖𝑖,1 = ∑4𝑚𝑚=0 2𝑚𝑚 𝑎𝑎 𝑡𝑡,𝑖𝑖,3 𝑙𝑙𝑖𝑖,1,𝑚𝑚 (33)
𝑎𝑎 𝑡𝑡,𝑖𝑖,4 𝑐𝑐𝑖𝑖,2 = ∑4𝑚𝑚=0 2𝑚𝑚 𝑎𝑎 𝑡𝑡,𝑖𝑖,4 𝑙𝑙𝑖𝑖,2,𝑚𝑚 (34)
Auxiliary variables 𝑤𝑤𝑡𝑡,𝑖𝑖,1,𝑚𝑚 = 𝑎𝑎 𝑡𝑡,𝑖𝑖,3 𝑙𝑙 𝑖𝑖,1,𝑚𝑚 , 𝑤𝑤𝑡𝑡,𝑖𝑖,2,𝑚𝑚 = 𝑎𝑎 𝑡𝑡,𝑖𝑖,4 𝑙𝑙 𝑖𝑖,2,𝑚𝑚 are further introduced as follows.
 𝑎𝑎 𝑡𝑡,𝑖𝑖,3 − (1 − 𝑙𝑙 𝑖𝑖,1,𝑚𝑚 )M ≤ 𝑤𝑤𝑡𝑡,𝑖𝑖,1,𝑚𝑚 ≤ 𝑎𝑎 𝑡𝑡,𝑖𝑖,3 (35)
 𝑎𝑎 𝑡𝑡,𝑖𝑖,4 − (1 − 𝑙𝑙 𝑖𝑖,2,𝑚𝑚 )M ≤ 𝑤𝑤𝑡𝑡,𝑖𝑖,2,𝑚𝑚 ≤ 𝑎𝑎 𝑡𝑡,𝑖𝑖,4 (36)
 0 ≤ 𝑤𝑤𝑡𝑡,𝑖𝑖,1,𝑚𝑚 ≤ 𝑙𝑙 𝑖𝑖,1,𝑚𝑚 M, 0 ≤ 𝑤𝑤𝑡𝑡,𝑖𝑖,2,𝑚𝑚 ≤ 𝑙𝑙 𝑖𝑖,2,𝑚𝑚 M (37)
The operation constraints of SOP in (12) can be transformed into the rotated quadratic cone constraints:
SOP,L SOP,L
2 2 𝑃𝑃𝑡𝑡,𝑖𝑖 𝑃𝑃𝑡𝑡,𝑖𝑖
SOP
(𝑃𝑃𝑡𝑡,𝑖𝑖 SOP
) + (𝑄𝑄𝑡𝑡,𝑖𝑖 ) ≤2 (38)
√2𝐴𝐴SOP
𝑖𝑖 √2𝐴𝐴SOP
𝑖𝑖
2528 Peng Li et al. / Energy Procedia 158 (2019) 2524–2529
Author name / Energy Procedia 00 (2018) 000–000 5

SOP,L SOP,L
2 2 𝑃𝑃𝑡𝑡,𝑗𝑗 𝑃𝑃𝑡𝑡,𝑗𝑗
SOP SOP
(𝑃𝑃𝑡𝑡,𝑗𝑗 ) + (𝑄𝑄𝑡𝑡,𝑗𝑗 ) ≤2 SOP SOP (39)
√2𝐴𝐴𝑗𝑗 √2𝐴𝐴𝑗𝑗
Now, after the conic relaxation, the original MINLP model is converted into an MISOCP model to realize a rapid
and accurate calculation.
4. Case study
In this section, the effectiveness of the proposed method is verified on the modified PG&E 69 -node system, as
shown in Fig. 2(a). Four photovoltaic generators (PVs) are integrated at node 33,35,52 and 54, with a capacity of
1000kVA each. All the PVs are operated at a unit power factor of 0.9. Two groups of SOPs with a capability of
1000kVA are installed. It is assumed that the loss coefficient of each inverter for SOP is 0.02.
1

59 60 61 62 63 64 65 66 67 68 69
0.8

Profile(p.u.)
40 41
57 58
0.6

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
0.4
PV PV SOP1
55 56
DC AC 0.2
42 43 44 45 46 47 48 49 50 51 52 53 54 AC DC

0
39
0 2 4 6 8 10 12 14 16 18 20 22 24
36 37 38 PV PV
DC AC
Time(h)
SOP2
28 29 30 31 32 33 34 35 AC DC Load PV

Fig. 2. (a) Structure of the modified PG&E 69-node system. (b) Daily operation curve of PVs and loads

The daily PVs and loads operation curves are shown in Fig.2(b) [6]. The upper and lower limits of statutory voltage
range are set as 1.10 p.u. and 0.90 p.u. And the desired voltage range is set from 0.98 p.u. to 1.02 p.u. The weight
coefficients α and β are set to 0.7 and 0.3 by AHP.
Q/Q m a x Q/Q m a x
(p.u.) (p.u.)
1 1

1
0.9 0.98 1.1 V(p.u.) 0.9 1 1.1 V(p.u.)

-1 -1
SOP1 at node 27 SOP1 at node 54

Q/Q m a x Q/Q m a x
(p.u.) (p.u.)
1 1

1.03 1.03
0.9 0.95 1.1 V(p.u.) 0.9 0.99 1.1 V(p.u.)

-1 -1
SOP2 at node 35 SOP2 at node 48

Fig. 3. Local Q-V curve of SOPs

The control parameters of two SOPs are optimally tuned by the proposed method. The optimization results of local
𝑄𝑄 − 𝑉𝑉 control curves for each SOP are shown in Fig. 3.
Three scenarios are adopted to verify the effectiveness of the proposed strategy in ADNs:
Scenario I: There is no control strategy conducted on SOPs, and the initial operation state of ADNs is obtained.
Scenario II: The proposed combined central and local control strategy is conducted on SOPs.
Scenario III: The outputs of SOPs are regulated by central control strategy to realize global optimization.
Table 1. Optimization results of three scenarios.
Scenario I Scenario II Scenario III
Power losses(kWh) 1758.7 1311.6 1250.5
Minimum voltage of ADN (p.u.) 0.9351 0.9694 0.9701
Maximum voltage of ADN(p.u.) 1.0460 1.0254 1.0252

The outputs of SOPs in Scenario II are shown in Fig. 4 and Fig. 5. The optimization results of three scenarios are
listed in Table 1. Compared with Scenario I, the proposed control strategy in Scenario II effectively mitigate voltage
deviation and reduce power losses of the whole network. It can be seen that the proposed strategy has a similar
 Peng Li et al. / Energy Procedia 158 (2019) 2524–2529 2529
6 Author name / Energy Procedia 00 (2018) 000–000

performance with the central control strategy in Scenario III, which has the most optimal use of active and reactive
outputs of SOPs. Considering the proposed combined control strategy is based on less measurement information, it
could reduce the computational burden as well as achieve the near globally optimal solution.

Fig. 4. (a) Active outputs of SOP1. (b) Reactive outputs of SOP1.

Fig. 5. (a) Active outputs of SOP2. (b) Reactive outputs of SOP2.

Fig. 6 shows the voltage profiles of the nodes PVs connected to in three scenarios. Compared with Scenario I,
voltage profiles are significantly improved in Scenario II.

Fig. 6. (a) Voltage profile at node 35. (b) Voltage profile at node 52.

5. Conclusion
This paper proposes a combined central and local operation strategy of SOPs to realize voltage control in ADNs.
By applying convex relaxation, the original MINLP model is converted into an effectively solved MISOCP model.
The optimization results show that by adopting the proposed combined control strategy, the potential benefits of SOPs
are fully explored to reduce power losses and improves voltage profile of ADNs.
Acknowledgements
This work was conducted in cooperation of APPLIED ENERGY UNiLAB-DEM: Distributed Energy & Microgrid.
UNiLAB is an international virtual lab of collective intelligence in Applied Energy.
References
[1] Wang X, Wang C, Xu T, et al. Optimal voltage regulation for distribution networks with multi-microgrids. Applied Energy 2018; 210: 1027-
36.
[2] Ferreira P D, Carvalho P, Ferreira L A. Distributed energy resources integration challenges in low-voltage networks: voltage control limitatio n s
and risk of cascading. IEEE Transactions on Power Systems 2013; 4(1): 82-88.
[3] Wang C, Song G, Li P, et al. Optimal siting and sizing of soft open points in active electrical distribution networks. Ap plied Energy 2017; 189:
301-309.
[4] Karagiannopoulos S, Aristidou P, Hug G. Hybrid approach for planning and operating active distribution grids. IET Generat ion Transmission
& Distribution 2016; 11(3): 685-95.
[5] Zhao J, Li Y, Li P, et al. Local voltage control strategy of active distribution network with PV reactive power optimization. Proceedings of
IEEE PES General Meeting, 2017; 1-5.
[6] Long C, Wu J, Thomas L. Optimal operation of soft open points in medium voltage electrical distribution networks with distributed generat ion.
Applied Energy 2016; 184: 427-437.