1

Using AC Optimal Power Flow for

DG Planning and Optimisation
Luis F. Ochoa, Member, IEEE, and Gareth P. Harrison, Member, IEEE

term– is left behind.

Abstract—While it is difficult to estimate how much Depending on the particular circumstances of a DG
distributed generation (DG) capacity will be connected to development, such as resource availability, planning consents
distribution systems in the coming years, it is certain that or declared net capacity, it might be possible to have more
increasing penetration levels require robust tools that help assess
than one network integration scheme (i.e., connection point
the capabilities and requirements of the networks in order to
produce the best planning strategies. This work presents an and/or operation strategy) that is economically sound for the
overview of some of the uses to which a tailored AC Optimal DG owner. The economics of different locations becomes
Power Flow (OPF) can be applied to DG planning and even more relevant if not only are infrastructure costs
optimisation. Its application to maximising connectable capacity involved but also distribution connection charges. Indeed,
and energy loss minimisation are presented and discussed. The DNOs could tune the latter to steer DG projects towards
incorporation of a wide range of network constraints, including
specific areas where the technical and economic impacts on
fault levels, voltage step change and N-1 security are also
outlined. The AC OPF technique is extended to a multi-period the system are less onerous or even beneficial. Alternatively,
approach to cater for the variability of demand and renewable bilateral commercial arrangements between DNOs and DG
generation. Advanced smart grid-like control strategies are also owners could also provide win-win situations. However, for
incorporated into this planning tool in order to evaluate the DNOs to determine appropriate locational signals or
potential benefits. Simple test feeders as well as generic UK commercial arrangements they need to investigate how
distribution systems are adopted to demonstrate the
capable their networks are for integrating renewable or
methodology.
conventional DG.
Index Terms—Distributed generation, optimal power flow, While it is difficult to estimate how much distributed
active network management, wind power, distribution networks. generation capacity will be connected to distribution systems
in the coming years, it is certain that increasing penetration
I. INTRODUCTION levels require robust tools that help assess the capabilities and

F OR Distribution Network Operators (DNOs) in fully requirements of the networks in order to produce the best
liberalised electricity markets (e.g., UK), planning the planning strategies. Whatever the particular driver for a given
siting and sizing of Distributed Generation (DG) units is, in DNO, e.g., to allow the connection of more DG capacity or to
many respects, not possible. Due to unbundling rules, DNOs reduce energy losses, these planning tools must take into
cannot invest in generation facilities and are meant to provide account network constraints such as voltage and thermal
DG owners with cost-effective connection means, irrespective limits. The inherent variability of demand and renewable
of the technology or geographical location. In this context, generation (e.g., wind power) is an aspect that must also be
uncertainties due to, for instance, planning consents or considered. In addition, envisaging actively managed
financial support surrounding DG investments pose DNOs networks (as opposed to the current ‘fit and forget’ approach),
with major challenges as to what, where and when to reinforce where Smart Grid-like control schemes employing real-time
the system in order to deliver timely connections without the control and communication systems allow more effective
risk of stranded assets. This lack of certainty and planning management of different network participants, including DG
coordination translates into distribution networks that connect units and voltage regulation devices, need also to be
DG units by adopting a ‘fit and forget’, case-by-case approach accounted for.
where only traditional reinforcements (e.g., new lines or Based on the authors’ previous and ongoing work, this
transformers) are carried out. Thus, any sophisticated solution paper compiles the different DG planning and optimisation
– albeit potentially more cost-effective for society in the long applications to which Optimal Power Flow (OPF) techniques
can be applied. The flexibility provided by a tailored AC OPF
makes it possible to extend the analysis to cater for a number
This work is funded through the EPSRC Supergen V, UK Energy of complex aspects. It can handle multi-periods to deal with
Infrastructure (AMPerES) grant in collaboration with UK electricity network
operators working under Ofgem’s Innovation Funding Incentive scheme – full the variability of demand and renewable generation.
details on http://www.supergen-amperes.org/. Advanced control strategies such as coordinated voltage
The authors are with the Institute for Energy Systems, School of control, adaptive power factor and generation curtailment can
Engineering, University of Edinburgh, Edinburgh, EH9 3JL, U.K. (e-mail:
[email protected], [email protected])
also be incorporated to evaluate potential benefits. More

978-1-4244-6551-4/10/$26.00 ©2010 IEEE

 2

complex problems resulting from other network constraints to provide/absorb 15 MVA. In the original demand-only case,
commonly overlooked by DG studies, such as fault levels, N- the OLTC at the substation has a target voltage of 1.045pu at
1 security and voltage step change, , are also viable within this the secondary bus. The voltage regulator (VR) and OLTCs on
approach. the 33/11-kV rural distribution transformers have a target
This paper, composed mainly by adapted extracts from [1- voltage of 1.03 pu. Voltage limits are ±6 % of nominal,
5], is structured as follows: First, the maximum connectable reflecting UK practice. The total peak demand is 38 MW.
DG capacity problem [1, 6] is presented in section II. Then, Hourly demand for central Scotland in 2003, as well as two
section III addresses the energy loss minimisation problem [2, different wind sites (named here, WP1 and WP2) derived
6]. Section IV discusses the incorporation of further network from UK Meteorological Office measured wind speed data
constraints such as fault levels [3, 6], N-1 security [4, 7], and [23], are adopted. Six wind generation sites are available.
voltage step change [5]. Finally, section V concludes the Buses 1105, 1106, 1108) are considered to all use the WP1
work. For the benefit of the reader, the framework for profile. The network contains a subsea cable connecting the
handling variable resources and demand, as well as the basic ‘mainland’ with an island on which the other three sites lie
mathematical structure for the multi-period AC OPF, (1113, 1114, 1115). The island enjoys approximately the same
including the incorporation of active network management wind resource as the first group but is sufficiently distinct to
schemes, are briefly included in the Appendix. In all the require the use of WP2.
studies the method was coded in the AIMMS optimisation
334 1118 328 342 325 1113
modelling environment [8] and solved using the CONOPT 302 331 332 333
DG
3.14A NLP solver. 329 337 335 336
322 324 1112

1114
interconnector
II. DISTRIBUTION NETWORK CAPACITY ASSESSMENT [1] DG
320 326 1115
Distribution network capacity assessment studies, i.e., the 327 330 1117 DG
VR
evaluation of the spare connection capacity, have been 328 1116
304 321 1111

approached in the literature from different angles, taking subsea cable

GSP
account of technical, economic and environmental issues. In 100 338 318 319 1110
317 1109
[9-11], the siting and sizing of DG units was investigated 341 311
using impact indices. A number of studies have adopted OLTC
314 316
301 1101

metaheuristics techniques, such as genetic algorithms [12-14], 340

313 1107 315 1108
DG
but analytical approaches have also been proposed [15, 16]. 312
303
The use of linear programming was demonstrated in [17, 18], 1102

whereas the non-linear formulation based on the AC Optimal 339 310 1106
LEGEND
1113 Node Index
Power Flow (OPF) problem was adopted in [1, 4-6, 19-21]. DG

The approach proposed by the authors in [1, 6], uses the 305 307 309 1105
DG
non-linear programming (NLP) formulation of a multi-period 306 1103
AC OPF adapted to determine the maximum DG capacity able 308 1104

to be connected to a given network. In addition to effectively Fig. 1 UK GDS EHV1 Network [22] and potential locations for distributed
handling the time-variation of multiple renewable sites and wind power generation.
demand, it also considers a range of active network techniques Maximum 54.6
DG Capacity
to allow maximum absorption of renewable generation 48.8
51.4
53.8
capacity while respecting voltage statutory limits and thermal 46.0
45.2 48.1
constraints. Active network management control algorithms 60 48.1
42.6
including coordinated voltage control of transformers and 45
39.4 44.5 43.0

voltage regulators, adaptive power factor control and energy (MW) 37.4
39.7
32.6 38.3
curtailment are embedded within the formulation. 30
29.6
The objective function corresponding to the maximisation 15
34.1
27.5 10%
of the total active DG capacity p of a set of generators G 5%
0
(indexed by g) across the set of periods M (indexed by m), is PFc 24.3
2%
given by ( ∀m ∈ M ): 0.98 (i) CVC+
CVC
max ∑ pg ,
Unity

g ∈G
(1) 0.98 (c) no CVC
no Curtailment
It is subject to the constraints presented in the Appendix.
Fig. 2 Full EHV 1 Network: Connectable DG capacity (in MW) with ANM
A. Case Study - Full EHV1 Network
strategies (c: capacitive, i: inductive, and PFc: adaptive power factor control).
Fig. 1 shows the EHV1 Network, a 61-bus 33/11-kV radial
distribution system available in [22]. The interconnector, The aggregate DG capacity that can be connected to the six
treated as a PV bus, has a target voltage of 1.00 pu and is able sites has been evaluated for all control modes and the results
 3

are shown in Fig. 2. Due to the number of potential locations approaches [20, 21], only peak demand and passive operation
and their corresponding proximity to loads, relatively high of the network were considered.
values of connectable DG capacity were found. It is clear that Here, the AC OPF technique used in the previous section is
the greater flexibility offered by adaptive power flow control tailored to minimise energy losses across a given time
(PFc) leads to the largest wind power generation capacities in horizon. The objective function corresponding to the
all cases. Without coordinated voltage control (‘no CVC’ minimisation of the total energy losses of the network over a
case), PFc alone allows a DG penetration of 85% relative to time horizon comprising m periods, m ∈ M , is given by:
peak demand (an increase of around a fifth over passive ⎛ ⎞
management). The allowable penetration reaches 103% when min ∑ ⎜⎝ ∑ f
m∈M l∈L
+ f l ,2,mP ⎟ ⋅τ m
1, P
l ,m
⎠
(2)
applying CVC, while curtailment, permits progressively 1, P
greater integration of wind capacity, reaching 118% where f l ,m and fl ,2,mP are the active power injections at each
penetration for a limit of 2% and 143% at the 10% curtailment end (denoted 1 and 2) of branch l, l ∈ L ; and, τ m is the
level (i.e., doubling capacity). duration of period m. The difference between the net
Given the detailed modeling of each DG site, it is possible injections at each end of the branch defines the energy loss.
to have a breakdown of the connectable generation capacity
across the six sites (Fig. 3). The impact of using coordinated A. Case Study - Full EHV1 Network
voltage control and adaptive power factor control on the The previously described EHV1 Network (Fig. 2), together
available capacity of each location is evident. In the passive with the same data for demand and wind profiles (WP1 and
management case with fixed 0.98 capacitive power factor, WP2), is considered in this analysis. Using the segmented
more than 73% of the total capacity is sited on the mainland demand scenarios (similar to those applied in section II) for
(1108, 1106 and 1105). Adopting the CVC and PFc schemes, the original configuration (no DG) results in an annual
however, it is possible to make much more capacity available consumption of 214 GWh with energy losses of 4.7%
on the island, increasing it by almost three-fold to a 46% share (comparable with typical UK rural networks).
of the total capacity. Fig. 4 presents the minimum percentage energy losses that
can be achieved if wind power generation is optimally
1108

1106

1105
1113

1114

1115

CVC, PFc accommodated under each operating strategy and without

* exceeding voltage or thermal limits. Compared to the original
CVC, 0.98(c)
*
losses significant gains are achieved when optimally
no CVC, PFc accommodated. Assuming a business as usual (BAU)
no CVC, 0.98(c)
**
(no 1115) management of the network unity power factor operation of
the DG units sees energy losses reduced by 40%. If
0 10 20 30 (MW) 40
coordinated voltage control is incorporated, then losses are cut
Fig. 3 Locational breakdown of available DG capacity across the sites. Except by more than a half. From all the studied cases, the adoption
for *, 33/11kV transformers at the DG connection nodes reached their thermal
limits. of both CVC and adaptive power factor control lead to the
lowest losses. Nonetheless, it is clear that, for this particular
III. MINIMISING ENERGY LOSSES [2] network, the largest benefits are brought about by the CVC
Capturing the effects that the variability of both demand scheme, raising the question of the cost effectiveness of using
and (renewable) generation has on total energy losses for a further control mechanisms.
3.04
given horizon is essential as it considers the actual metrics Energy BAU

used by DNOs. Modeling DG plants as firm generation (to Losses

2.81
2.55
some extent a less complex optimisation problem) has been (%) 2.79
adopted for loss analyses using Tabu Search [24] or Genetic 3
2.30
2.69
Algorithm (GA)-based multiobjective approaches [13]. As for
2.24
variable (renewable) generation, the optimal allocation of DG 2

plants based on impact indices (including losses) was 2.17

1 0.98 (i)
previously proposed by the authors in [11], and extended to a
GA-based multiobjective formulation in [14]. Energy losses 0
Unity

were also considered in [25], where it was presented a multi- no CVC 0.98 (c)

resource GA-based multiobjective technique that catered for CVC PFc

some aspects of active network management through the use
Fig. 4 Percentage of energy losses considering wind power generation –
of a linearised OPF. Energy loss minimisation was also business as usual operation and two different Smart Grid strategies.
studied in [26] through the optimal mix of statistically-
modelled renewable sources considering a passive approach to In terms of installed capacity, Fig. 5 shows the total values
manage the network. An AC OPF-like (reduced gradient) found for each of the analysed cases. Due to the variable wind
method applied to a (power) loss minimisation problem was availability for the different demand levels, critical scenarios
proposed in [19]. However, in this and other OPF-based
 4

such as minimum and peak demand do not normally coincide considering a three-phase fault at the MV busbar of the
with maximum wind potential. For this reason, more capacity substation. The initial symmetrical short-circuit power at the
than for firm generation can be connected to the network. It MV busbar resulting from the contributions of both the
can also be seen that, again, generation capacity is strongly upstream grid and the DG capacity connected to each feeder
related to the reduction of losses. It is worth pointing out that has to be limited to the design short-circuit capacity of the
while in most cases the CVC scheme only allows a marginal network, SCmax. The IEC 60909 standard is adopted for the
increase in capacity, when DG units are operated at 0.98 corresponding calculations.
capacitive power factor, the gain is much more significant. The fault level constraint incorporated into the single-
This is primarily due to the ability of the CVC scheme to period AC OPF formulation is as follows:
alleviate voltage rise problems. As for the PFc scheme, whilst 64 upstream grid
4744 DG fault contribution
8 644 7448
it does provide lower losses, it is also clear that, for this cmax ⋅Vb 1 2
cmax ⋅ Vb21
network, similar gains can easily be achieved by setting the +∑ ≤ SCmax (3)
Z SYS MV + ZT g∈G Z (b ,b ) + Z g
operation of the generators to unity or capacitive power factor. g 1

(MW) where Z SYS MV is the equivalent impedance of the upstream

21.1
21 20.4 20.6
CVC
grid (at the secondary side) and ZT is the short-circuit
20.9
20.4
impedance of the substation transformer(s). The second term
19 18.9 on the left hand side of (3) corresponds to the fault level
no CVC 18.7 contributions from the distributed generators. Z (bg ,b 1 ) is the
18.0
line impedance from the DG connection point to the MV
17
PFc 0.98 (c) Unity 0.98 (i) busbar, while Z g represents the (equivalent) impedance of the
Fig. 5 Total DG capacity – business as usual operation and two different generator(s) g. The simplified formulation of (3) implies only
Smart Grid strategies. one DG connection per feeder; however, it can be tailored to
cater for multiple DG configurations.
IV. INCORPORATING MORE NETWORK CONSTRAINTS For planning and indicative studies it is common to
This section presents the underlying concepts behind the consider the fault current contribution at the network terminals
incorporation of more complex network constraints of a given DG plant, I k" ,bg , to be a multiple of the
(irrespective of the objective function), such as fault levels, N-
1 security and voltage step change. Although, they can be corresponding rated current, I rg , called contribution factor
implemented within a multi-period framework, for simplicity ( CF = I k" ,bg / I rg ). To explore the effects that different fault
and the fact that these constraints are more important during
level contributions from DG have on the total generation
worst-case scenarios, the formulation is presented for the
capacity that can be accommodated, this (linear) dependency
single-period analysis (index m is not used).
is taken into account. Consequently, Z g will be a function of
A. Fault Levels [3, 6] the contribution factor CF.
Only a few capacity assessment studies have considered
fault level contributions from DG [10, 11, 13, 18]. Fault level B. N-1 Security [4, 7]
constraints were formally included in the OPF approach The analysis of N-1 contingencies (e.g., loss of a line) can
presented in [21], where the corresponding derivatives with also be incorporated in the AC OPF (also known as Security
respect to the OPF variables were obtained by expressing the Constrained OPF). It ensures that no limits are exceeded even
Z bus as a function of new capacity. It used a highly complex during contingencies that might occur in the system. In
practical terms, multiple network topologies are
series of matrix operations to enact this and works well in
simultaneously analysed.
meshed networks. Its formulation does, however, make it very
Apart from the multiple periods (if taken into account),
difficult to integrate with other analyses such as security
multiple topologies, K, due to N-1 contingencies require
constraints [4, 27], voltage step change [5] or multiple time
different sets of power flow variables. Thus, all the affected
periods [1, 28].
equations need to be adapted to cater for N-1 security. For
Here, a simplified approach to cater for fault contributions
example, the voltage limits at each bus (see Appendix) will be
that avoids the use of derivatives is adopted, allowing better
integration within optimisation modelling environments. The re-formulated considering the index k: Vb− ≤ Vb , m , k ≤ Vb+ .
three-phase fault at the substation is calculated according to
C. Voltage Step Change [5]
the IEC 60909 standard with the calculation of fault level
contributions from DG based on the corresponding rated Most DG studies do not consider a particular requirement
currents and the generation technology. The design short- of the distribution networks: voltage step constraints on loss
circuit capacity of the network provides the constraint. of a generator, which is a quite distinct issue from voltage
With distribution networks typically radial in rise. Voltage step changes occur when a DG is started up or
configuration, the maximum fault level will be obtained when suddenly disconnected from the network, and limits are
 5

typically placed on the maximum step change allowed.

Voltage step constraints are incorporated using a security Fig. 6 (bottom) shows the discrete values following the
constrained OPF-like formulation (see previous subsection), allocation of the original data into a series of 7 bins covering
where the contingencies considered are outages of generators specific ranges ({0}, (0,0.2pu], (0.2pu,0.4pu],…,
rather than lines. Thus, the voltage step constraint to be (0.8pu,1.0pu), {1.0}) in which the mean values (e.g., 0.3pu for
incorporated into the single-period AC OPF is: the (0.2pu,0.4pu] range) characterised each new hour. The
Vb − VS+ ≤ Vb , g ' ≤ Vb + VS+ (4) aggregation process groups hours in which the same
combination of demand and generation occur. For instance,
where for an outage of generator g, indexed by g ' , the
the arrows point to hours where demand is 0.7pu and wind is
contingency voltage Vb, g ' at bus b must differ by no more than zero; these conditions occur for a total of 18 hours in this
VS+ from the pre-outage voltage Vb . In practical terms, extra particular week. This will constitute a period to be evaluated
topologies – similar to the N-1 security constraint – are along with other combinations of different total duration in the
evaluated considering (4) on top of other network constraints. optimisation problem. References [1, 2] provide a more
detailed treatment of the framework.
V. CONCLUSIONS B. Multi-Period AC Optimal Power Flow
The AC Optimal Power Flow technique is widely accepted Although the objective function will vary according to the
and mainly used to solve the economic dispatch problem. It study, i.e., either DG capacity maximisation or energy loss
has been demonstrated by a number of studies that it provides minimisation, the mathematical structure to consider basic
a platform where other network analyses can be carried out. network constraints, such as voltage and thermal limits,
This paper has compiled the authors’ previous and ongoing remains the same, as well as the multi-periodicity, in terms of
work on DG Planning and Optimisation using the AC OPF demand/generation combinations. This multi-periodicity is
technique. The way this approach can handle real-life achieved by providing each combination, m, with a different
complexities such as variable demand and generation, set of power flow variables with a unique, inter-period set of
different network constraints, and Smart Grid-like control generation capacity variables is used throughout the analysis.
schemes, proves its versatility and robustness. Given an objective function, the basic multi-period AC
It is important to point out that increasing the level of OPF formulation (single topology) will be subject to a range
complexity will result in longer computing times. Ultimately, of constraints (more details in [1, 2]). Voltages at each bus b
solving extremely large problems, in terms of the number of (B, set of buses) are constrained by maximum/minimum levels
variables and constraints, will depend on the capability of the Vb( + , − ) :
commercial or non-commercial solvers available.
Vb− ≤ Vb , m ≤ Vb+ (5) ∀b ∈ B
VI. APPENDIX [1, 2] Thermal constraints on the flow at each end of lines and
transformers, l (L, set of lines):
A. Framework for Handling Variable Resources and Demand
To diminish the number of periods to be evaluated whilst
(f l ,m ) +( f
(1,2), P 2 (1,2), Q 2
l ,m ) =(f ) l
+ 2
∀l ∈ L (6)
(1,2), P (1,2), Q
preserving the behaviour and inter-relationships between where f l ,m and f l ,m are the active and reactive power
resource and demand, it is used a process of discretisation and injections at each end of the branch and f l + is the branch
then aggregation according to the characteristics of ‘similar’
apparent power flow limit.
periods. To illustrate this, Fig. 6 (top) presents a week-long
Active and reactive power injections into each end of the
snapshot of hourly demand and wind power data for central
lines are governed by Kirchhoff’s voltage law:
Scotland in 2003 [23].
f l (1,2),( = fl ,(1,2), m ( Vm , δ m ) ∀l ∈ L
P ,Q ) KVL ( P , Q )
1.0
Wind Demand ,m (7)
m ( Vm , δ m ) (Vm , δ m )
KVLP KVLQ
0.8 where f l ,(1,2), and f l ,(1,2) are standard
0.6
Kirchhoff voltage law expressions.
(p.u.)

0.4
Kirchhoff’s current law describes the active and reactive
0.2
nodal power balance, ∀b ∈ B :
0.0

1.0
1 25 49 73 97 121
d0.7-w0.0
145 16
∑1,2 pbL,m + dbPηm = g∈G∑|β =b pg ωm + x∈X∑|β =b px,m (8)
l∈L| β = b l b g x
0.8

0.6 ∑ q L
b, m + d ηm =
Q
b ∑ pg ωm tan(φg , m ) + ∑ qx , m (9)
(p.u.)

l∈L| βl1,2 = b g ∈Gb | β g = b x∈ X | β x = b

0.4

( p , q )b , m
L
0.2
Here are the total power injections into lines at b;
0.0
1 25 49 73 97
Winter: 1st Week of January
121 145 16
and db( P ,Q ) are the peak active or reactive demands at the same
Fig. 6 (Top) Winter week hourly demand and wind power for central
Scotland, 2003 [36]. (Bottom) Discretised data processed before aggregating
bus. In period m, ηm is the demand level relative to peak and
the coincident hours of each demand-generation scenario.
 6

ωm is the generation level relative to nominal capacity as REFERENCES

dictated by the variable (renewable) resource in that period. [1] L. F. Ochoa, C. Dent, and G. P. Harrison, "Distribution network capacity
assessment: Variable DG and active networks," IEEE Trans. on Power
The distribution network has external connections x (X, set Systems, vol. 25, no. 1, pp. 87-95, Feb. 2010.
of external sources) at the GSP (and interconnectors) through [2] L. F. Ochoa and G. P. Harrison, "Minimising energy losses: Optimal
which power can be exported/imported subject to flow accommodation and smart operation of renewable distributed
constraints : generation," IEEE Trans. on Power Systems, under review.
[3] L. F. Ochoa and G. P. Harrison, "Exploring the effects of fault level
px− ≤ px , m ≤ px+ ⎪⎫ contributions from DG on the hosting capacity of radial networks," IET
⎬ ∀x ∈ X (10) Proceedings Generation, Transmission & Distribution, under review.
qx− ≤ qx , m ≤ qx+ ⎪⎭ [4] C. Dent, L. F. Ochoa, G. P. Harrison, and J. W. Bialek, "Efficient secure
AC OPF for network generation capacity assessment," IEEE Trans. on
The GSP is taken as the reference (slack) bus b0 with the Power Systems, vol. 25, no. 1, pp. 575-583, Feb. 2010.
voltage angle set at zero, i.e., δ b0 , m = 0 . [5] C. Dent, L. F. Ochoa, and G. P. Harrison, "Network distributed
generation capacity analysis using OPF with voltage step constraints,"
IEEE Trans. on Power Systems, vol. 25, no. 1, pp. 296-304, Feb. 2010.
C. Incorporating Active Network Management [6] L. F. Ochoa, C. J. Dent, and G. P. Harrison, "Maximisation of
In addition to voltage and thermal constraints, variables intermittent distributed generation in active networks," in Proc. 2008
CIRED Seminar: SmartGrids for Distribution, pp. 4.
and constraints derived from ANM schemes must also be [7] L. F. Ochoa, A. Keane, C. Dent, and G. P. Harrison, "Minimising
incorporated in the method: transmission reactive support required by high penetration of distributed
1) Coordinated Voltage Control (CVC) wind power generation," in Proc. 2009 8th International Workshop on
In each period the secondary voltage of the OLTC will be Large-Scale Integration of Wind Power into Power Systems, pp. 5.
[8] J. Bisschop and M. Roelofs, "AIMMS - The user’s guide," Paragon
treated as a variable, rather than a fixed parameter, while Decision Technology, 2006.
maintaining its value within the statutory range: [9] P. Chiradeja and R. Ramakumar, "An approach to quantify the technical
benefits of distributed generation," IEEE Trans. on Energy Conversion,
Vb−OLTC ≤ VbOLTC , m ≤ Vb+OLTC (11) vol. 19, no. 4, pp. 764-773, Dec. 2004.
2) Adaptive Power Factor Control (PFc) [10] L. F. Ochoa, A. Padilha-Feltrin, and G. P. Harrison, "Evaluating
distributed generation impacts with a multiobjective index," IEEE Trans.
Here, it is envisaged that DG provides a dispatchable on Power Delivery, vol. 21, no. 3, pp. 1452-1458, July 2006.
power factor scheme, hence the power angle of each [11] L. F. Ochoa, A. Padilha-Feltrin, and G. P. Harrison, "Evaluating
generator, φg , m , is considered as a variable. In practice DG distributed time-varying generation through a multiobjective index,"
IEEE Trans. on Power Delivery, vol. 23, no. 2, pp. 1132-1138, Apr.
will be required to operate within a certain range of power 2008.
factors ( φg( + , − ) ); the following constraint applies: [12] A. Silvestri, A. Berizzi, and S. Buonanno, "Distributed generation
planning using genetic algorithms," in Proc. 1999 IEEE Power Tech, pp.
φg− ≤ φg , m ≤ φg+ (12) 257.
[13] G. Celli, E. Ghiani, S. Mocci, and F. Pilo, "A multiobjective
3) Energy Curtailment evolutionary algorithm for the sizing and siting of distributed
Power curtailment is formulated here by adding a negative generation," IEEE Trans. on Power Systems, vol. 20, no. 2, pp. 750-757,
May 2005.
generation (or positive demand) variable ( pgcurt
, m ) at the same [14] L. F. Ochoa, A. Padilha-Feltrin, and G. P. Harrison, "Time-series based
location of each DG unit; solely affecting the constraints maximization of distributed wind power generation integration," IEEE
Trans. on Energy Conversion, vol. 23, no. 3, pp. 968-974, Sep. 2008.
related to active and reactive nodal power balance. Thus, (8) [15] C. Wang and M. Hashem Nehrir, "Analytical approaches for optimal
and (9) are, adapted by adding terms ∑ pgcurt ,m and placement of distributed generation sources in power systems," IEEE
g ∈Gb Trans. on Power Systems, vol. 19, no. 4, pp. 2068-2076, Nov. 2004.
[16] N. Acharya, P. Mahat, and N. Mithulananthan, "An analytical approach
∑p curt
g ,m tan (φg , m ) , respectively. for DG allocation in primary distribution network," International
g ∈Gb Journal of Electrical Power & Energy Systems, vol. 28, no. 10, pp. 669-
678, Dec. 2006.
To examine the impact of different allowed levels of
[17] S. N. Liew and G. Strbac, "Maximising penetration of wind generation
curtailment on overall DG capacity, the total amount of in existing distribution networks," IEE Proceedings Generation,
curtailed energy from each DG will be restricted to a Transmission & Distribution, vol. 149, no. 3, pp. 256-262, May 2002.
[18] A. Keane and M. O'Malley, "Optimal utilization of distribution networks
curtailment factor λcurt , a percentage of the potential energy
for energy harvesting," IEEE Trans. on Power Systems, vol. 22, no. 1,
that could have otherwise been delivered by each DG. The pp. 467-475, Feb. 2007.
following constraint follows: [19] N. S. Rau and Y.-H. Wan, "Optimum location of resources in distributed
planning," IEEE Trans. on Power Systems, vol. 9, no. 4, pp. 2014-2020,
⎡ ⎤
∑ , mτ m ≤ λcurt ⎢ ∑ pg ωmτ m ⎥
Nov. 1994.
pgcurt ∀g ∈ G (13) [20] G. P. Harrison and A. R. Wallace, "Optimal power flow evaluation of
m∈M ⎣ m∈M ⎦ distribution network capacity for the connection of distributed
where τ m is the duration of period m. The curtailment generation," IEE Proceedings Generation, Transmission & Distribution,
vol. 152, no. 1, pp. 115-122, Jan. 2005.
variables pgcurt
, m need to be limited to the output of g at the [21] P. N. Vovos and J. W. Bialek, "Direct incorporation of fault level
constraints in optimal power flow as a tool for network capacity
corresponding period: analysis," IEEE Trans. on Power Systems, vol. 20, no. 4, pp. 2125-2134,
, m ≤ ωm pg
pgcurt ∀g ∈ G (14) Nov. 2005.
[22] Distributed Generation and Sustainable Electrical Energy Centre. United
Kingdom Generic Distribution System (UK GDS). [Online]. Available:
http://www.sedg.ac.uk/
[23] T. Boehme, J. Taylor, A. R. Wallace, and J. W. Bialek, "Matching
 7

renewable electricity generation with demand," Scottish Executive,

Edinburgh, Feb. 2006.
[24] K. Nara, Y. Hayashi, B. Deng, K. Ikeda, and T. Achizawa, "Optimal
allocation of dispersed generators for loss minimization," Electrical
Engineering in Japan, vol. 136, no. 2, pp. 1-8, 2001.
[25] A. Alarcon-Rodriguez, E. Haesen, G. W. Ault, J. Driesen, and R.
Belmans, "Multi-objective planning framework for stochastic and
controllable distributed energy resources," IET Renewable Power
Generation, vol. 3, no. 2, pp. 227-238, Jun. 2009.
[26] Y. M. Atwa, E. F. El-Saadany, M. M. A. Salama, and R. Seethapathy,
"Optimal renewable resources mix for distribution system energy loss
minimization," IEEE Trans. on Power Systems, vol. 25, no. 1, pp. 360-
370, Feb. 2010.
[27] C. J. Dent, L. F. Ochoa, G. P. Harrison, and J. W. Bialek, "Efficient
secure AC OPF for distributed generation uptake maximisation," in
Proc. 2008 16th Power Systems Computation Conference (PSCC'08),
pp. 7.
[28] L. F. Ochoa, A. Keane, C. Dent, and G. P. Harrison, "Applying active
network management schemes to an Irish distribution network for wind
power maximisation," in Proc. 2009 International Conference on
Electricity Distribution (CIRED), pp. 1-4.

BIOGRAPHIES
Luis F. Ochoa (S’01-M’07) is a Research Fellow
in the School of Engineering, University of
Edinburgh, U.K. He obtained his BEng degree from
UNI, Lima, Peru, in 2000, and the MSc and PhD
degrees from UNESP, Ilha Solteira, Brazil, in 2003
and 2006, respectively.
His current research interests include network
integration of distributed energy resources and
distribution system analysis. Dr. Ochoa is also a
member of the IET.

Gareth P. Harrison (M’02) is a Senior Lecturer in

Energy Systems in the School of Engineering,
University of Edinburgh, U.K.. His current research
interests include network integration of distributed
generation and analysis of the impact of climate
change on the electricity industry.
Dr. Harrison is a Chartered Engineer and
member of the Institution of Engineering and
Technology, U.K.