The 6th International Conference on Renewable Power Generation (RPG)

19–20 October 2017

Maximising wind generation through optimised operation of on-load tap changing

transformers in active distribution networks
Daniel Danzerl1, Simon Gill2, Olimpo Anaya-Lara2
1
Wind Energy CDT, Department of Electronic and Electrical Engineering, University of Strathclyde, Glasgow, UK
2
Department of Electronic and Electrical Engineering, University of Strathclyde, Glasgow, UK
E-mail: [email protected]

Published in The Journal of Engineering; Received on 12th October 2017; Accepted on 3rd November 2017

Abstract: On-load tap changing transformers are the most common control device to regulate and maintain distribution network voltage within
required limits. Voltage rise issues on the other hand have become a major factor limiting greater penetration of low carbon generators, par-
ticularly in weak distribution networks. Here, the voltage rise problem is addressed through the application of optimised set-point voltage
technique that aims to improve network hosting capacity to accommodate high wind penetration. It assesses the effectiveness of the technique
on a realistic 289-node UK generic 11 kV distribution network using time-series optimal power flow simulations. The results reveal that when
the tap changer is operated at the optimised set-point voltage, it can lead to greater energy yields. It also shows a reduction in the number of tap
changing operations when the transformer is operated within the optimised deadband allowing for an improved life-span and minimum main-
tenance cost.

1 Introduction are highly susceptible to voltage rise problems when various DG

resources are connected [6]. Power generations from intermittent re-
Centralised voltage control methods in distribution networks have
newable resources such as wind and solar compound the challenge
traditionally relied on on-load tap changing (OLTC) transformers
and undermine the performance of the OLTCs to efficiently regulate
as the most common control device to regulate and maintain
the voltage. These complexities have rendered conventional OLTC
network voltages within required limits [1]. In the UK, the statutory
transformers alone, inefficient and unreliable to control the voltage,
voltage limits are defined in the Electricity Safety, Quality and
limiting DG connection capacities on the network.
Continuity (ESQC) Regulations 2002 [2] and it specifies that, low-
The paradigm shift towards an intelligent and advanced distribu-
voltage (LV) customers’ be supplied at 400/230 V with tolerance of
tion management systems (DMS) requires sophisticated control
+10/−6%, whereas high-voltage (HV) customers, at tolerance of
strategies that optimise existing network assets. Addressing the
±6%. The OLTC operates by moving their tap positions to select
voltage rise problem efficiently will require a well-coordinated
appropriate transformer turns ratio that suits a range of power
DMS that can control transformer tap changers, voltage regulators,
flow conditions on the network. To simplify and automate
power plants, compensators, and loads at the primary substation [7].
voltage control, automatic voltage control (AVC) relays are used
The advent of active network management (ANM) techniques
in conjunction with line drop compensation (LDC) equipment.
offers a feasible solution in this direction in developing an efficient,
The AVC relay continually monitors the network to detect
flexible, and reliable network. It involves real-time monitoring and
voltage variation and initiates a tap change command to the
control of the network and is seen to enhance greater DG connec-
motorised OLTC when the voltages are outside the pre-set
tion capacities without reinforcement. Recent ANM schemes and
limits [3].
trial projects in the UK have investigated control strategies that
The AVC relay operation usually incorporates a time delay
address thermal and power flow congestion problems on the
setting between 10 and 120 s from detecting an out of range
network. However, voltage constraint issues are becoming a more
voltage and starting a tap-change command. The time-delay
complex challenge and one that has not been fully investigated
setting is to avoid unnecessary tap operation during short-term
and trailed in current ANM schemes.
voltage fluctuations on the network. The LDC is used to compen-
In this paper, the voltage rise problem is addressed and mitigated.
sate for voltage drop variations on the line between the transformer
The study investigates an optimised set-point voltage technique that
and loads situated towards the far end of the feeder
aims to improve network hosting capacity to accommodate high
Previous passive operation of the network assumed unidirection-
wind penetration levels. It presents a detailed case-study assessment
al power flows and uncontrollable resources from the transmission
of the benefits and impact on DG capacities using time-series AC
system into distribution networks at the grid supply points (GSP)
optimal power flow (OPF) simulations. It concludes that, when
[4]. However, recent proliferation of distributed generator (DG)
the OLTC is operated at the optimised set-point voltage, it can
connections renders such assumptions invalid and as such have
lead to greater generation levels.
initiated significant changes to the previous passive methods to an
active approach to network management. Operational challenges
such as reverse power flows from the DGs in current active distri- 2 Optimisation problem
bution systems adversely affect the operation of OLTCs to efficient- Previous work reported by Deckmyn et al. [8] involved an area
ly regulate and maintain the voltages [5]. Weak distribution control strategy of OLTC operation using local network measure-
networks, particularly, rural networks with poor X/R characteristics, ment. The authors used voltage information of remote measurement

J. Eng., 2017, Vol. 2017, Iss. 13, pp. 2339–2344 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0749 Attribution License (http://creativecommons.org/licenses/by/3.0/)
points to determine optimal tap positions that minimises voltage de- The quadratic cost function of the generators is represented as:
viation on the network. This was achieved through heuristic-based
optimisation methods which treated the tap changer as discrete vari-

NG
ables. A decision-making algorithm is proposed in [9] to find ad- min ai + bi (PGi ) + ci (PGi )2 (7)
equate set-point voltage generated through simulations or i=1
historical performance to control voltage rise using OLTCs. In
[3], the authors discuss voltage regulation using a coordination where ai, bi, and ci represent the cost coefficient of active power
between OLTC and LDC in medium-voltage feeders to improve output PGi. The OPF model assigns high cost values to the swing
DG connection capacities. A similar area-based coordinated bus to discourage grid active power imports from the GSP and
control strategy that controls the voltage set-points of the AVC low cost values to all the DGs to encourage active demands on
relay at the primary substation have been studied in [6] using time- the network be met by the DGs. The nodal power balance (2) can
domain simulations. The authors investigated the dynamic oper- be expanded as:
ation and responses of the AVC relays and tap changer mechanism
NG 


to the system load variations. In [10], the authors discuss an interior  
point (IP) method based on a non-linear complementarity model for PGi − PDi − Vi Vj Yij  cos (uij − di + dj ) = 0 (8)
i=1
OPF problems with load tap changes. The authors initially treat the

NG  
tap positions as continuous variables to identify the upper and lower  
bounds of the transformer using OPF. The complementarity con- QGi − QDi − Vi Vj Yij  sin (uij − di + dj ) = 0 (9)
i=1
straints of the transformer taps are then modelled into the OPF
and solved using IP optimisation method.
The parameter constraint limits in (3) is expanded as:
The proposed control technique in this paper is formulated as an
Generator constraints: Generator voltages and real and reactive
optimisation problem that makes use of mathematical AC OPF ana-
power outputs are constrained by their upper and lower limits as
lysis tools using time-series simulations. It utilises the standard
follows:
ACOPF [11, 12] formulation at each time-step and treats the tap
position as control variables. The optimisation aims to achieve
the following system objectives:
min
VGi ≤ VGi ≤ VGi
max
(10)
min
PGi ≤ PGi ≤ PGi
max
(11)
† maximise wind generation outputs at minimum cost,
† improving network hosting capacity,
Gi ≤ QGi ≤ QGi
Qmin max
(12)
† minimise the impact of DGs on voltage profiles,
† minimise the number of OLTC transformer tap operations. Transformer constraints: The OLTC transformer tap settings are
bounded as follows:
The ACOPF-based problem is modelled as a set of non-linear
equations consisting equality and inequality constraints and is Timin ≤ Ti ≤ Timax (13)
given as follows:
Slack bus constraint: The slack bus voltage is bounded as:
Optimise: f (x, u) (1)
vmin
s ≤ vs ≤ vmax
s (14)
Subject to: g(x, u) = 0 (2)
Security constraints: These include the constraints of voltages at
load buses and thermal power flow limits
h(x, u) ≤ 0 (3)
Vimin ≤ Vi ≤ Vimax (15)
where f represents the objective function, g the physics of the power  
 
system enforced through the power flow equations, and h the par- Sij  ≤ Sijmax (16)
ameter limits of the system. x represents a vector of state variables
and it includes the real power generation of the slack bus PG(slack), The ACOPF formulation in this paper involves a two-stage opti-
load bus voltages Vi, reactive power generation QGi, and line misation process that treats the tap positions as discrete and continu-
thermal flows Si and can be expressed as: ous control variables similar to the methods discussed in [10]. The
initial stage is a discrete method which models the voltage step ratio
x = [PG(slack) Vi QGi Si ] (4) as discrete variables that can vary between a certain maximum vmax s
and minimum vmin s set-point voltage by a fixed step-size Δv. The
AVC model manually moves up or down by one step size Δv at a
u represents the vector of control variables and includes generator time and locks the transformer tap position to a pre-determined
real power PGi, generator voltage VGi, and transformer tap voltage set-point. Each optimisation cycle requires a physical move-
changer Ti and can be expressed as: ment of the tap position to the desired set-point, where the voltage is
held fixed at the slack bus to connect the multiple DG schemes. The
control logic can be expressed mathematically as:
u = [PGi VGi Ti ] (5)
⎧
⎪
⎨ vs + Dv if vk − vref , vs , vmax min
= vmax
 Dm and s ; vs s
To maximise wind generation capacities PGi, the objective function vs = vs if vk − vref  , Dm; vmin
s = vmax
s
f takes the form in (6): ⎪
⎩
vs − Dv if vk − vref . Dm and vs . vmax min
s ; vs = vmax
s
(17)


NG
Maximise PGi (6) where vs represents the set-point voltage at the slack bus, vk the
G=1 regulated voltage quantity, vref is the reference voltage, and Δm

This is an open access article published by the IET under the Creative Commons J. Eng., 2017, Vol. 2017, Iss. 13, pp. 2339–2344
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0749
represents the AVC relay deadband. The ACOPF discrete model
investigates a narrowed deadband that aims to optimise DG capaci-
ties within the specified network constraints. Within this narrowed
deadband, lies an optimum set-point voltage. In the second stage,
the narrowed AVC deadband is modelled into the OPF and treats
the transformer tap positions as continuous variables. The continu-
ous method assumes a small tap step ratio Δv and models the dis-
crete switches as approximated continuous variation of the
set-point voltage vs. In this case, the ACOPF can choose
optimum set-point values between the narrowed upper maximum
Fig. 2 Active demand profile (2 weeks)
limit vmax min
s(optimised) and lower minimum limits vs(optimised) that maxi-
mises the objective function at each iteration. This can be repre-
sented mathematically as:

s(optimised) ≤ vs(optimised) ≤ vs(optimised)

vmin max
(18)

3 UK generic distribution test network

The proposed strategy is applied to a realistic 11 kV UK generic
distribution system (UKGDS) to assess its effectiveness. Fig. 1 Fig. 3 Normalised wind profile (2 weeks)
shows a single-line representation of a 289-node radial distribution
network developed in IPSA with full network parameters given in
ratings are considered as additional thermal constraints on the
[13]. The model comprises a mix of urban and rural sections con-
network. However, thermal limits on the network are deliberately
sisting of underground cables followed by overhead lines charac-
relaxed to enable the voltage limits to operate as binding constraints
terised by varied X/R ratios and medium conductor lengths. The
to control the generators. Half-hourly time-series load profiles are
primary substation supplies three 11 kV feeders and is linked to a
connected on all secondary buses and consists a mixture of residen-
33 kV distribution system represented as a source of real and react-
tial, industrial, and commercial loads. These are aggregated values
ive power. The primary substation is equipped with two identical
scaled from a single load profile (peak loads) provided in [13] with
33/11 kV OLTC transformers, each rated at 22 MVA connected
minimum demands set at 25% of the peak load. Fig. 2 presents a
in parallel to regulate the network voltage to a pre-defined target.
2-week long snapshot of half-hourly demand data.
Here, the OLTCs at the 33/11 kV substation controls the second-
The use of historic wind resource time-series, such as normalised
ary nominal voltage to a reference set-point and has a maximum and
output of a nearby wind farm, is used to estimate potential gener-
minimum tap settings of +4 to −14%, respectively, in 19 steps with
ation. Fig. 3 shows a 2-week sample of normalised wind generation
a step size of 1.0%. The AVC relay bandwidth is allowed an output
profile. A suit of eight ‘non-firm’ distributed wind generator
voltage deviation of ±3% of the nominal. The secondary bus vol-
(DWG) schemes of varied capacities are connected to the
tages are constrained at current DNO operational limits and
network at nodes 1244, 1144, 1105, 1191, 1120, 1310, 1358, and
allowed to vary within a permissible range of ±3% of the
1387 and assumed to operate an ANM scheme.
nominal. The medium circuit conductor lengths with varied MVA
The DWGs are modelled and connected to the network as (P, Q)
nodes and operated at unity power factor mode with no voltage
control capabilities. The principle of access rule for connecting
the multiple generators and sharing of curtailment is via a technical
best arrangement. Here, all the DG schemes are assumed to have
equal priority and access to the network. In this rule, the OPF
engine optimally shares limited network capacity by assigning
greater curtailment in order of generator with the most impact on
voltage rise constraint. The model is deployed in Matpower [14]
where simulation studies are carried over 1 month period at half-
hourly resolution consisting a total of 1440 time-steps.

4 Results
To demonstrate and quantify the effectiveness of the proposed strat-
egy to mitigate the voltage rise problem, five scenarios for connect-
ing the multiple DG schemes have been investigated and are
summarised in Table 1. Each scenario configures the set-point

Table 1 Summary of AVC set-point voltages

Scenario % kV p.u.

1 +3 11.33 1.03
2 +2 11.22 1.02
3 +1 11.11 1.01
4 0 11.00 1.00
5 −1 10.89 0.99
Fig. 1 11 kV radial distribution network (HV_UG_OHb) [13]

J. Eng., 2017, Vol. 2017, Iss. 13, pp. 2339–2344 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0749 Attribution License (http://creativecommons.org/licenses/by/3.0/)
Fig. 4 Energy generated and curtailed at +3% set-point voltage Fig. 6 Energy generated and curtailed at −1% set-point voltage

voltage of the AVC relay to a pre-determined fixed target value to

assess the impact on the DGs and adequacy of the technique to miti- 4.3 Scenario 5: set-point voltage fixed at −1% of nominal
gate the problem. Each scenario provides a quantitative assessment Here, the set-point voltage is further lowered to −1% of the nominal
of renewable energy yields and curtailment levels required to main- voltage. Simulation results presented in Fig. 6 show a significant
tain the network limits when the strategies are implemented. improvement in generation capacities, allowing the remotely con-
In this study, −2 and −3% set-point voltages are ignored as a nected generators greater network access and subsequent wind
result of undervoltage conditions occurring towards the ends of yields. Generator H is now observed to be generating up to a
the feeders. The OPF failed to converge due to breach of lower maximum of 36% of its installed capacity.
voltage limits at certain time-steps. A sample of the scenarios The total energy realised across the investigated set-point vol-
studied and corresponding simulation results obtained are presented tages are presented in Fig. 7. It is seen that, the total generation
in the following sections. across the DG schemes increases by lowering the slack bus set-
point voltage. Operating the AVC at high set-points significantly
4.1 Scenario 1: set-point voltage fixed at +3% of nominal reduces generation capacities. For example, at +3% set-point
voltage, a total energy yield of 1071 MWh was realised out of
This scenario represents a worse-case operation condition of the
12,851 MWh available energy representing a total generation cap-
AVC set-point voltage. In this study, the set-point voltage at the
acity of 8.3% across the DGs. In the case of −1% set point
slack bus is raised and held fixed at the upper maximum limit
voltage, a total generation of 10,909 MWh was realised out of the
(+3% of nominal) which serves as a reference point to the rest of
total available energy representing an improved generation capacity
the secondary bus voltages. The ACOPF reveals poor DG perfor-
of 85% across the DG schemes.
mances in terms of generation levels (Fig. 4) with the biggest size
Generator performance profile across the investigated set-point
generator Gen A (which is electrically closer to the primary sub-
voltages are presented in Fig. 8. From here, it is observed that,
station) generating a maximum capacity of 19%. In comparison
to Gen H (which is the smallest size and electrically the most fur-
thest away), the generator is observed to be generating ∼1% of its
rated capacity. All the DGs are observed to experience significant
curtailment due to reduced voltage margins and substantial
impact on voltage rise constraints from active generations.

4.2 Scenario 3: set-point voltage fixed at +1% of nominal

In this scenario, the set-point voltage is lowered and held fixed at
+1% of the nominal. The ACOPF results show improved generation
levels of the DGs with generators A, B, and C operating at
maximum installed capacities. Subsequently, the OPF allows an
improved access for the rest of the DGs to generate onto the
network. Generator H is now observed to be generating up to a
maximum of 27% of its rated capacity when compared with scen-
ario 1 and is shown in Fig. 5 Fig. 7 Total energy generation

Fig. 5 Energy generated and curtailed at +1% set-point voltage Fig. 8 Generator performance profile

This is an open access article published by the IET under the Creative Commons J. Eng., 2017, Vol. 2017, Iss. 13, pp. 2339–2344
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0749
the DGs generation capacities decrease as we move further along
the feeders and away from the primary substation. Generators
located closer to the substation showed greater penetration levels
across all voltage set-points when compared with the remotely
located DGs. For instance, at +3% set-point, it seen that generators
A, B, and C are most favoured with greater network access due to
their strong location. Conversely, the rest of the DGs are least
favoured (with highly restricted or no access to the network) due
to their weak remote location. Comparatively, it is seen that the
OPF allows an improved access for all DGs schemes when the
slack bus set-point voltages are lowered to levels that reduce their
impact on voltage rise constraint.
Fig. 10 Changes in set-point voltage at full range
5 Discussion
The optimisation engine controls and dictates generation capacities
according to DGs active power impact on local voltage rise at their
point of connection. Setting the AVC at high-voltage set-points ad-
versely limits the network’s headroom capacity to cope with high
DG penetration. In effect, active generation from the DGs
become highly susceptible to pushing the network’s voltage profiles
up to the maximum limits forcing the OPF to heavily curtail genera-
tions to levels that satisfy the constraint. In such extreme cases, gen-
erators located on the stronger sections of the network (with
minimum impact on voltage rise) are allowed access, whereas the
remotely connected DGs have no access as seen in Fig. 8.
Therefore, to influence fairness and greater connection capacities
for all DGs across the various network location points, there is a
Fig. 11 Changes in set-point voltage at optimised deadband
scope to lower the voltage settings at the slack bus to enhance
DG voltage margins and allow greater active generation capabil-
ities. By doing this, the voltage profiles and network headroom cap- number of tap changing operation of the OLTC transformer, enhan-
acity can be improved to accommodate high wind penetration cing the life-span and reducing the cost of maintenance.
levels. Fig. 9 shows single time-step profiles of maximum bus
voltage magnitudes recorded during the optimisation. Here, by
comparing the voltage profiles of selected set-point voltages (scen- 6 Conclusion
arios 1, 3, and 5), it seen that operating the AVC at −1% set-point This paper has presented a constraint optimisation strategy that con-
voltage improves the voltage profile and mitigates the voltage rise figures the set-point voltage of OLTC transformers at distribution
problem. Voltage profile at −1% showed the least number of primary substation. It investigates an optimum set-point voltage
points at which the voltage magnitudes reach the upper maximum control technique of the tap-changer transformer with an overall
limits implying reduced curtailments when compared with +3 and aim to improve network hosting capacity to accommodate greater
+1%. To compare the actions of the tap operation with respect to wind connections at weak distribution networks. The results
changes in set-point voltages, the tap positions are treated as con- obtained demonstrated that, configuring the AVC relays at lower
tinuous variables. A base-case is initially investigated where the set-point voltage significantly improves the network hosting capaci-
slack bus set-point voltage is modelled to vary between the full ties. By doing so, the DG voltage margins can be improved and
set of minimum and maximum operational limits their active power impact on the constraint effectively minimised.
−3% ≤ vs ≤ +3%. In this study, the optimised deadband for the It also shows that, controlling the transformer within a narrowed
OLTC is established at 0% ≤ vs (optimised) ≤ −1%. Fig. 10 (base- optimum deadband effectively reduces the number of tap changing
case) shows high variation and changes in set-point voltage of the operations allowing for an improved life-span and reduced mainten-
AVC which controls the OLTC indicating a potential increase in ance cost.
the number of tap operation of the OLTC transformer.
Conversely, Fig. 11 shows a significant reduction in changes in
voltage variation when the AVC is operated within the narrowed 7 Acknowledgment
optimised deadband. This indicates a potential reduction in the This work was funded by the UK’s Engineering and Physical
Science Research Council (EPSRC), project reference number
EP/G037728/1.

8 References

[1] Lakervi E., Holmes E.J.: ‘Electricity distribution network design’

(2007, 2nd edn.)
[2] Department of Trade and Industry: ‘Guidance on the Electricity
Safety, Quality and Continuity regulations 2002’. Rep. Number
URN 02/1544, October 2002, pp. 1–47
[3] Viawan F.A., Sannino A., Daalder J.: ‘Voltage control with on-load
tap changers in medium voltage feeders in presence of distributed
generation’, Electr. Power Syst. Res., 2007, 77, (10), pp. 1314–1322
[4] Vovos P.N., Kiprakis A.E., Wallace A.R., ET AL.: ‘Centralized and dis-
tributed voltage control: impact on distributed generation penetra-
Fig. 9 Maximum voltage profile tion’, IEEE Trans. Power Syst., 2007, 22, (1), pp. 476–483

J. Eng., 2017, Vol. 2017, Iss. 13, pp. 2339–2344 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0749 Attribution License (http://creativecommons.org/licenses/by/3.0/)
 [5] Kulmala A.: ‘Active voltage control in distribution networks includ- [9] Xu T., Wade N.S., Davidson E.M., ET AL.: ‘Case-based reasoning for
ing distributed energy resources’. PhD thesis, 2014 coordinated voltage control on distribution networks’, Electr. Power
[6] Kulmala A., Mäki K., Repo S., ET AL.: ‘Active voltage level Syst. Res., 2011, 81, (12), pp. 2088–2098
management of distribution networks with distributed generation [10] Li B., Bai X., Wei H.: ‘An interior point method based on nonlinear
using on load tap changing transformers’. PowerTech, 2007, pp. complementarity model for OPF problems with load tap changing
455–460 transformers’. Asia-Pacific Power Energy Eng. Conf. APPEEC, 2010
[7] Strbac G., Jenkins N., Hird M., ET AL.: ‘Integration of operation of em- [11] Carpentier J.: ‘Optimal power flows’, Int. J. Electr. Power Energy
bedded generation and distribution networks’. Manchester Centre for Syst., 1979, 1, (1), pp. 3–15
Electrical Energy, 2002, pp. 1–95 [12] Dommel H., Tinney W.: ‘Optimal power flow solutions’, IEEE Trans.
[8] Deckmyn C., Vandoorn T.L., Meersman B., ET AL.: ‘A coordinated Power Appar. Syst., 1968, PAS-87, (10), pp. 1866–1876
voltage control strategy for on-load tap changing transformers with [13] UKGDS: ‘UK generic distribution system’, 2015. [Online]. Available
the utilisation of distributed generators’. 2016 IEEE Int. Energy at https://github.com/sedg/ukgds
Conf. ENERGYCON 2016, 2016 [14] Zimmerman R.D., Murillo-Sanchez C.E.: ‘Matpower 6.0b1’, vol.
22203, February 2017

This is an open access article published by the IET under the Creative Commons J. Eng., 2017, Vol. 2017, Iss. 13, pp. 2339–2344
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0749