CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0035

DIGITAL MODEL OF A DISTRIBUTION MANAGEMENT SYSTEM

FOR THE OPTIMAL OPERATION OF ACTIVE DISTRIBUTION SYSTEMS
Fabrizio PILO Giuditta PISANO Gian Giuseppe SOMA
University of Cagliari – Italy University of Cagliari – Italy University of Cagliari – Italy
[email protected] [email protected] [email protected]

According to the definition, the Distribution Management

ABSTRACT System (DMS) is necessary to control the system by
interacting with the OLTC (On Load Tap Changer) and
The growing interest on environmental issues and the DERs. Many DMS algorithms have been recently proposed
increasing cost of fossil fuels have determined the in the literature. Some of them are focused on voltage
conditions for high shares of Distributed Generation in the regulation, which can be severely affected by DG,
distribution system. Nowadays distribution networks are especially weak rural distribution systems Algorithms for
approaching a critical point whereby the connection of DG voltage regulation are proposed in [3]-[4]. In [3] the DMS
will require an active approach. The paper presents a improves voltage regulation by resorting to generation
Distribution Management System to manage an active curtailment whether all other possible operation setting are
distribution network economically and safely. unsuccessful. In [4] the optimization algorithm is based on
sensitivity indexes to identify the most convenient DG units
INTRODUCTION to inject active and/or reactive power. The objective of the
Deployment of Distributed Generation (DG) into the algorithm is to minimise the amount of curtailed power. In
existing passive distribution networks is reaching a critical [5] and [6] the DMS is based on the optimization of an
point whereby it can no longer be installed in the typical “fit objective function that considers energy losses, line
and forget” fashion without impacting network operation ampacity and the contribution of responsive loads. The
and stability. Connection and management of DG in the algorithm is fast and well suited for real time applications.
network is the first challenge facing Distribution System In this paper the optimization algorithm for the DMS has
Operators (DSOs) in making the transition to active, been improved so that it continuously optimizes the voltage
integrated networks. profile, minimizes losses and optimizes power flows. The
Voltage regulation, fault levels, and line overloads represent paper presents the DMS and its software realization in the
a serious concern to DSOs when high shares of DG are DIgSilent Power Factory® software. Examples derived from
connected to the system. According to the passive approach real world cases are provided to show some features of the
the solution is constituted by major network investments algorithm.
that are not compatible with DSO budgets and it may pose
economic barriers to the DG integration, limiting the ACTIVE DISTRIBUTION NETWORK
amount of additional DG that can penetrate the distribution The aim of ADNs is to increase the share of DG in existing
system [1]-[2]. Taking an active approach allows the DSO distribution networks and keep the system stable and
to relieve network contingencies without allocating great reliable at the same time [1]. A schematic of an active
money amount for network development. Indeed, the active network is presented in Fig. 1.
management of distribution networks enables the DSO to In ADNs power flows through network feeders may be
maximize the use of the existing circuits by taking full modified and nodal voltages regulated in real time by
advantage of generator dispatch, control of transformer taps, controlling DG, FACDS (Flexible AC Distribution System),
voltage regulators, reactive power management and system OLTC, storage devices, etc..
reconfiguration in an integrated way. This active approach
to system operation reduces the negative impact of DG on DMS
the network.
The most comprehensive definition to describe what such
an active approach is of Active Distribution Network,
recently approved by the CIGRE SC C6, is the following: P, Q, V P, Q, V
“Active distribution networks (ADNs) are distribution
networks that have systems in place to control a
P, Q, V

combination of distributed energy resources (generators, P, ±Q

P, ±Q P, -Q
loads and storage). DSOs have the possibility of managing
the electricity flows using a flexible network topology. S WT ∼= PV
DERs (Distributed Energy Resources) take some degree of
responsibility for system support, which will depend on a CHP
suitable regulatory environment and connection agreement.” Figure 1: Main features of active distribution networks.

CIRED2009 Session 5 Paper No 0035

CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0035

The DMS aims at finding the optimal operation point of the voltage, ri is the resistance of the i branch, Δt is the time
system avoiding critical contingencies and minimizing the interval between two successive DMS runs. The average
operation cost. The DMS input are real time measurements value of the estimated power, Favg, equal for each network
and the distribution state estimation output [7]. branch in order not to penalise specific paths in the
Furthermore, to safely operate the distribution system, the optimisation process, is used to obtain an estimate of the
central control system would need backup scenarios in case average losses.
of failure in the control or communication systems. The second summation in (1) takes into account the role of
Parts of active network management are: active power dispatch to reduce energy losses and relieve
• real time measurements, network congestions. By dispatching the power from each
• distribution network state estimation, generator, the DMS can modify the line power flows with
• communication technology, positive effects on the system. In the paper it has been
• distribution management system, assumed that DG owners have to be compensated for any
• local DG control systems, power curtailment so that the resort to this control action is
• responsive loads, justified only if the cost of the losses becomes greater than
• energy storage. the cost of power curtailment. The cost of generation
curtailment, CGC, can be calculated with (3).
NDG _ gc NDG _ gc NDG _ gc
DISTRIBUTION MANAGEMENT SYSTEM
The DMS proposed in the paper helps solve the following
CGC = ∑ (
j Δt Pgj − Pgj =
c DG *
) ∑ β j Pgj − ∑ β j Pgj*
(3)
j =1 j =1 j =1
contingencies:
β j = −cDG
j Δt
• voltage regulation problems (typically overvoltages
caused by DG and voltage drops caused by high load), where cjDG is the cost for reducing 1 kWh of the jth DG unit
• line overloads in both standard or emergency production, Δt is the interval between two successive real-
configurations, time network calculations, NDG_gc is the number of the
• loss minimization. controllable generators connected to the distribution
The options to relieve the above mentioned contingencies network, Pgj is the real power output of the jth DG unit and
are: Pgj* is the rated power production in the time interval. Eq.
• network reconfiguration, if permitted by the (3) is the second term of (1), considering that the second
automation level, term of (3) is invariant and does not modify the solution of
• DG generation curtailment and DG ancillary services, the optimization problem.
• demand side response. The third summation in (1) represents the cost for
The optimization algorithm in the DMS has to find the purchasing reactive power from DG; ψj is the cost that the
optimal combination of such operation options to minimize DSO has to pay for the kVARh produced by the DG units.
system costs without causing violations of the technical Finally, the last summation takes into account the cost for
constraints. The objective function (1) takes into shedding the responsive loads in the network. PkDSR is the
consideration the energy losses, the cost of generation power shed from the kth load, NDSR is the number of the
curtailment, the cost of reactive power, and the cost of load responsive loads, γk is proportional to the cost of power
shedding. shedding.
Nbranches N DG _ gc N DG _ gc N DSR In order to linearize the optimization problem, the power
min ∑ δ i Fi + ∑ β j Pjgc + ∑ ψ j Q jgc + ∑ γ k PkDSR (1) flow Fi is expressed by means of two non-negative
i =1 j =1 j =1 k =1 quantities, Xi and Yi, that cannot be both nonzero at the
With some approximation, the objective function can be same time [8]. The optimization problem can be stated as
expressed as a linear combinations of line flows [8], follows:
NDG _ gc NDG _ gc
curtailed power [5], and shed power [6]. In this paper the Nbranches NDSR

optimization algorithm has been modified to take into min ∑ δ (X

i =1
i i + Yi ) + ∑βP
j =1
j j
gc
+ ∑ψQ
j =1
j
gc
j + ∑ γ k PkDSR (4)
k =1
consideration nodal voltage violations and the effect of
reactive power injections. subject to:
The first summation in (1) is proportional to the cost of the ⎡ XP ⎤
⎢ ⎥
energy losses. Fi is the active power flow through the ith ⎡A −A Bg BDSR 0 0 0 0 ⎤ ⎢ YP ⎥
⎢ ⎥ ⎢ gc ⎥ ⎡ P ⎤
branch of the network. δI is a coefficient that allows ⎢0 0 0 tgϕ A −A Bg 0⎥ ⎢ P ⎥ ⎢ ⎥
estimating the cost of energy losses. Eq. (2) gives the ⎢ ⎥ ⎢ P DSR ⎥ ⎢Q ⎥ (5)
⎢0 0 dv dv ⎢ ⎥
0 0 0 0⎥ ⎢ ⎥ = ΔV
approximated value of the cost of energy losses in the ⎢ dP dP ⎥ ⎢ XQ ⎥ ⎢ B ⎥
⎢ ⎥ ⎢q ⎥
network, Closs.
⎢ I −I 0 0 mB −m B 0 I ⎥ ⎢ YQ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ g ⎥
Nbranches
⎛ cl ⋅ Δt ⋅ ri ⋅ Favg ⎞ Nbranches
⎢0 0 I ⎥⎦ ⎢ Q gc ⎥ ⎣ ⎦
q
Closs = ∑
i =1 ⎝
⎜⎜
3 ⋅ Vn2
⋅
⎟⎟ i
⎠
F =
i =1
δ i Fi (2)∑ ⎣ I 0 0 0 mg
⎢
⎢⎣ S ⎥⎦
⎥

cl is the unitary cost of the energy lost, Vn is the nominal

CIRED2009 Session 5 Paper No 0035

CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0035

[ X P ] , [YP ] , ⎡⎣ X Q ⎤⎦ , ⎡⎣YQ ⎤⎦ , [ S ] ≥ 0 (6)

109
104

where [A] is the node to branch incidence matrix, [Bg] and

107 103
[BDSR] are binary matrixes introduced to insert DSR and GC 111 108 106
105 101

into power flow equations, tgϕ is referred to DSR loads (the 110

DSR loads are considered with a constant power factor), [S]

99 102
118
is a non-negative vector of slack variables to transform 100

inequality constraints into equality constraints, PDSR is the

114
112
98
vector of the shedding powers, Pgc is the vector of the 119
117
116
113
115
curtailed powers, P and Q are the nodal powers. WT 2MVA
97
96
WT 2 MVA
The main novelty introduced in the paper is that DG can 54
52
53
85 93
91
help regulate voltage by injecting reactive power. For this 64
88
94
95

51
reason the optimization problem is subject to constraints on 61
62
47
49
84
81
92

60 65 63
50
voltage that are expressed according to (7). 58 WT1.5MVA
78
WT1.5MVA
⎛ dv ⎞
57 59
N DG _ gc
dv
ΔV = ∑ ⎜
83
ΔPk + ΔQk ⎟
55 79
(7) 56
⎜
k =1 ⎝ dP k dQ k ⎟ 45 75
74 77 80
82
86

⎠ 43
44
48
66 70 72
71 73
76 87
90
89

where ΔV is the sum of the voltage deviations with

41 67 69
46
40 68

42 GT 9MVA
dv dv
reference to the nominal voltage, and are the 1

dP k dQ k 24 8
6
7
25 2 9 10 HV/MV substation
sensitivity indexes calculated according to [4]. 34
5 11
MV/LV trunk node
26 12
Finally, the active and reactive powers generated from the 27
GT 1.5MVA
13
MV/LV lateral node

kth generator have to comply with the capability curve of the

28
29 14 15 MV/LV Automatic Sectionalized node
30
31 16 DSR Load
19
generator (for synchronous machines). The capability curve 33
32
GT 1.5MVA
17
18 DG unit
20
is approximated with a piecewise linear to maintain the 36
34
GT 1.5MVA
Trunk branch
linear formulation. Assuming that Nseg is the number of 37
35
Emergency connection
Lateral branch
straight lines used to approximate the generator capability 38
39
23
22
21

curve, for each generator Nseg inequality constraints have to

be considered. Eq. (8) gives a general formulation of the Figure 2: The benchmark network
constraint in the equality form. section are used due to the high load density. The increase
Pk + m gjk Qk + S jk = q gjk j = 1...N seg , k = 1...N DG _ gc (8) of the energy demand in this area may easily lead to
overloads and major investments might be necessary.
mjk and qjk are the slope and the intercept of the jth line used
Typically DG in such networks determines an increasing of
to approximate the capability curve of the kth generator.
the short circuit level and can contribute to sharpen voltage
A similar approach is also used to take into account the
regulation problems in the rural area. In both areas DG may
constraints caused by the rated ampacity of the lines.
or not contribute to loss reduction depending on position,
production, and load demand.
SIMULATION STUDIES Five typologies of loads have been considered: residential,
The studies performed with DIgSILENT PowerFactory®. industrial, tertiary, agricultural, and public lighting. The
The active management was tested on an MV distribution loads in the nodes 2, 11, 12, 13, 18, 19, 25, 32, 91, 112, and
network model that is used as benchmark in the 118 (marked in Fig. 2) participate to the DSR offering a
SMARTGRID Research Project that involves ENEL and 8 total or partial load shedding service that can be used by the
Italian Universities. The test network has been created by DSO to solve critical network conditions. Moreover, two
merging two portions of real Italian distribution networks typologies of generators have been taken into account: wind
(Fig. 2). One primary substation feeds 118 MV substations turbine (WT) and gas turbine (GT). The annual medium
(52 trunk nodes and 66 lateral nodes) that deliver power to active power delivered to the MV nodes is about 18.2 MW,
the MV and LV customers. The network is radial with divided in 11.9 MW for the urban feeder and 6.3 MW for
emergency tie connections. Two areas can be identified in the rural one. The unitary cost of Joule losses is 0.22
the picture. In the upper part of the network there are long €/kWh. Three 1.5 MVA GTs are installed in the urban area;
overhead lines feeding small loads. The cross section of the two 2 MVA and two 1.5 MVA WTs are connected to the
conductors is relatively small because of the low load rural part of the network. A dispatchable 9 MVA gas
density; as a consequence of the load growth voltage drop turbine is also connected to the rural portion of the network.
problems have to be expected. The DG in this area may In the simulation model the DMS interacts with the
severely affect voltage regulation causing overvoltages. In distribution systems through the DIgSILENT Power
the lower part of the picture, urban/industrial loads have to Factory® that is used to perform the load flow calculations.
be supplied. Here underground cables with bigger cross The whole procedure may be described with the following

CIRED2009 Session 5 Paper No 0035

CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0035

steps: Generation Curtailment Simulation

1. the time horizon is divided into intervals (in the paper The second simulation assumed that the DMS was able to
the duration of the time interval is set to 1 hour); control the active power produced by the generator by
2. at the beginning of the time interval the DMS gathers means of a control law more sophisticated than a simple on-
data from the network and from the distribution state off approach. In this case the algorithm is able to find the
estimator, that could be essential whether the number optimal reduction of active production. Furthermore, should
of measurement devices was too small. In this paper, more curtailment options be available, the optimization
the network data are the electrical quantities as they finds the optimal curtailment share among different
are calculated by the load flow; generators with the final goal of minimizing losses and
3. whether some constraints were not complied with, the costs. In this case the overvoltage caused by excessive
DMS calculates the set point for generators, which power generation during the off-peak hours is reduced by
might be committed to curtail active power and/or imposing a set-point to the biggest GT according to the
modify the production of reactive power, and loads, hourly demand and the status of the network. Fig. 3 shows
which might be requested to reduce power demand. the amount of active power curtailed to avoid voltage
Network reconfiguration is not allowed in the regulation problems. The total energy curtailed is 37
application proposed in the paper. MWh/day.
4. The new set points are hold until the end of the time
interval, when new data are gathered from the network Active and Reactive Power Dispatch Simulation
and used for a new optimization. The third simulation study simulates a fully functioning
The central control is also integrated with delayed local DMS that dispatches both active and reactive power to
voltage controls on the not dispatchable DG units (e.g. wind resolve voltage regulation. In Fig. 3 the amount of the
turbines). When the DMS was not able to eliminate an active power curtailment is reported. The reactive power
overvoltage in one node, the DG local control commands injection allows reducing the voltage in some nodes with a
the disconnection of the generator. smaller generation curtailment. The reduction of the energy
In order to show the positive role of the DMS in the production for the generator is 6 MWh/day. The general
network operation with high share of DG, the results of remark is that such a control might be very attractive for
three simulations are presented in the following subsections. DG owners. In fact, a 3 MVA GT connected to the network
Connect and Forget Simulation can produce 65 MWh/day running at the nominal power
(2.7 MW with unitary power factor). That power represents
In this study the current worldwide accepted practice, the the maximum power that can be accepted by the DSO
“connect and forget” policy, has been analysed. No central without charging the DG owner with high connection costs
control of generation and load has been used and only local caused by the major network reinforcements imposed by the
voltage controls are available to disconnect DG in case of application of the connect and forget rule. The application
high overvoltages or network faults. Local voltage controls of an on-off generation curtailment allows integrating a 9
should be only seldom used because, according to the MVA GT into the system. With the active power production
“connect and forget” policy, the operation problem is solved control, the GT can run at nominal power during the most
at the planning stage, by limiting the integration of remunerative hours of the day; the generation curtailment
renewable energy sources and DG. imposed by the DMS causes a 16% reduction of energy in
In a typical day, the test network experiences some the less remunerative hours of the day (with reference to a
overvoltages. In particular, whether all the above described nominal power production of 194 MWh/day).
generators run according to their daily production curves, The control of reactive power allows a smaller reduction of
the rural portion of the network needs a voltage regulation
during the off peak hours (from 10:00 pm to 7:00 am). 100%
Because of the high overvoltage, the 9 MVA GT installed in
the rural area has to be disconnected during the entire off 80%
peak hours. Normally, the DSO does not allow the
connection of a DG unit in such conditions because of the
60%
“connect and forget” rule. It requires that the DG owner
pays very high connection costs to reinforce the network or Nominal Power Production
40%
more likely the DG will be asked to be connected to a Power Production with GC
higher voltage level. By adopting the same assumptions, it
Power Production with GC (cosphi=1)
can be argued that the maximum allowable GT in the node 20%

67 has a rated power of 3 MVA. The first conclusion is that

even a drastic generation curtailment has the potentiality to 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24[h]
be convenient for a DG owner because the switch off time
very often will fall in the less convenient hours. Figure 3: The energy curtailed by the DMS with reference to
9 MVA GT unit connected at node 67 in Fig. 2

CIRED2009 Session 5 Paper No 0035

CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0035

1.06
the active power generated. The reduction with reference to without active management
the maximum producible energy is now only 3%, which 1.06
with active management

means that about 188 MWh/day could be generated and Voltage upper limit

injected in the network without network reinforcements. .]

1.05
.u
Fig. 3 reports the energy curtailed by the DMS during the p
[ 
e
g 1.05
day. In Fig. 4 the improvement of the voltage profile in the a
lt
o
V
critical area of the network depicted in Fig. 2 is shown. The 1.04
voltage profile refers to the first hour of the day that is
1.04
characterised by low load and possible high production.
1.03
Relieve of overloads during emergency network 66 68 69 71 72 74 78 81 85 97
configurations Rural feeder: TOP nodes

The benchmark network does not experience in the Figure 4: Voltage profile improvement in a critical hour of
examined day line overloads. Anyway, in order to validate the day (the numbers in the x axis correspond to the node
the capability of DMS of reducing overloads, some numbering used in Fig. 2)
emergency network configurations originated by line faults share without violations of the constraints on nodal voltages
have been also simulated. Faults in the underground cables and line currents. This goal is achieved by minimizing the
close to the primary substation 1 in the benchmark network cost of system operation, which is expressed in terms of
(Fig. 2) are particularly dangerous and, if they happened in cost of energy losses, cost of curtailed energy, cost of
the peak hours of the day, serious overloads can cause the reactive support, and cost of shed energy. The use of linear
intervention of the protections and the enlargement of the programming allows reducing the computing burden so that
faulty area. The most severe fault condition is represented the algorithm can be used in real time applications. The case
by a line fault in the line between the primary substation studies presented in the paper confirm the effectiveness of
and node 2. In that situation, the lines 31-32, 20-21, 22-23, the active approach in distribution systems.
and 23-39 would be overloaded because the emergency
connection is used to supply the loads downstream the REFERENCES
faulty line. The most critical overloads would be caused by [1] S. N. Liew, G. Strbac, “Maximising penetration of
a fault at 12 pm. The only solution is to reduce power wind generation in existing distribution networks”, IEE
demand by using the load responsive loads in the area. In Proc. Generation, Transmission and Distribution, vol.
that case, the majority of responsive loads in the network 149, no. 3, 256-262, 2002.
(node 2, 11,12,13, 18, 32) were shed by the DMS. The [2] P. Djapic, C. Ramsay, et al., “Taking an active
amount of shed power was equal to 50% for those nodes approach”, IEEE power & energy magazine, Vol. Vol.
except the node 32 that reduced power demand by 13.6%. 5, Issue 4, July/August 2007.
Globally, the DSR policy shed 1.1 MWh in the examined [3] F. Kupzog, H. Brunner, W. Pruggler T. Pfajfarm, A.
day. At the same time, the active power generated by the Lugmaier, ”DGDemoNet Concept – A new algorithm
big 9 MVA GT in the rural area is curtailed by 7% so that it for active Distribution Grid Operation facilitating high
produces 7.535 MWh at unitary power factor (voltage DG penetration”, 2007.
regulation is not necessary). [4] Q. Zhou, J.W. Bialek, “Generation curtailment to
manage voltage constraints in distribution networks”,
CONCLUSIONS IET Generation, Transmission and Distribution, 2007.
[5] G. Celli, M. Loddo, F. Pilo, “Distribution Network
The passive operation of distribution networks poses
Planning with Active Management”, in Proc. of 6th
intolerable barriers to the integration of DG. Voltage
World Energy System Conf., Torino (Italy), July 2006.
regulation, energy losses and line overloads are the most
[6] G. Celli, F. Pilo, G. Pisano, G. G. Soma, “Optimal
common problems that arise from the integration of DG in a
planning of active networks”, accepted for 16th Power
passive operated network. Active management of
Systems Computation Conference, 2008.
distribution networks allows postponing massive network
[7] C. Muscas, F. Pilo, G. Pisano, S. Sulis, “Optimal
reinforcement by controlling the power injected or drawn by
Placement of Measurement Devices in Electric
generators or loads. In this paper a Distribution
Distribution Systems”, IEEE IMTC 2006, 2006.
Management System for active distribution networks is
[8] A. Abur, G. Celli, M. Loddo, F. Pilo, “On-line network
presented. The optimization algorithm implemented in the
reconfiguration for loss reduction in distribution
DMS allows operating a distribution network with high DG
networks with Distributed Generation”, CIRED 2005.

CIRED2009 Session 5 Paper No 0035