NEW METHODS AND REQUIREMENTS FOR PLANNING OF MEDIUM

VOLTAGE NETWORK DUE TO DISTRIBUTED GENERATION

Sami Repo, Hannu Laaksonen and Pertti Järventausta

Tampere University of Technology, Finland
sami.repo@tut.fi

ABSTRACT

The subject of the paper is planning of distribution network in case of integration of

distributed generation on weak medium voltage network. Because network long-term
planning includes major uncertainties and integration of distributed generation has remarkable
effects at these areas, final network plans may be done after the integration agreement has
been concluded. In order to limit the price of integration fee the capability of existing network
should be utilised as efficiently as possible.

The existing planning methods of distribution network are designed for cases where a voltage
drop limits the utilisation of the network. Because the proposed planning methods for a
combined voltage drop and rise cases are based on the worst network conditions, the
consideration of stochastic properties of production are not taken into account at all. The
paper proposes two complementary planning methods for distribution networks including
distributed generation. The conservative “minimum demand − maximum production”
planning principle is extended to consider stochastic nature of distributed generation by using
production curves. The main focus of the paper is on a comparison of proposed planning
methods and on a testing of advantage voltage control to increase the network transfer
capability. The studies are based on real life distribution network and load data.

1. INTRODUCTION

In weak distribution networks over-voltage problems are likely to occur during low demand
periods when there are a large amount of distributed generation (DG) integrated on medium
voltage (MV) network [1]. Due to this network planning based on maximum demand will no
longer be enough. The planner might want to know how often the “minimum demand
maximum production” planning principle actually occurred, when it occurred and how the
operation of the system might be improved at those hours.

An example presented at Figures 1 describes the voltage rise and MV network planning
problems. The test system includes a connection to 110 kV regional network via 110/20 kV
transformer equipped with on-line tap changer controlled by voltage magnitude at secondary
(voltage set-value is 20.4 kV). MV network includes one feeder and equivalent of other
feeders at bus 3. A DG unit is connected to bus 4 at the end of feeder via 20/0.69 kV
transformer. The equivalent of loads is also presented at the bus 4.

The idea of the example is to calculate how much production might be connected to chosen
location. Due to uncertainty of DG unit the “maximum demand - zero production” case is a
relevant study case at MV network planning. The maximum load demand which fulfils
voltage level requirements (MV voltage between 21 and 19 kV) is 2 MW with existing feeder
parameters (R = 3,06964 p.u., X = 1,76314 p.u. and B = 0,001 p.u.). In order to increase loading

1
capability for 3 MW, the parameters of feeder are needed to reduce for values R = 2,06964
p.u., X = 1,06314 p.u. and B = 0,001 p.u. The results of these cases are presented at Figure 2
marked as No production, old and No production, new.

When the output of a DG unit is larger than zero, the capability of MV network is determined
by voltage rise in most cases. A 1 MW production is a very good choice from MV network
point of view, because the existing network may fully utilised independent of load demand. In
order to increase the integration capability of DG unit, MV network enforcements are
required.

The minimum load demand at bus 4 influences also very much for the interconnection
capability. If the minimum load demand is 1 MW, the integration capability of DG unit
increases for 2 MW with existing network and for 2.5 MW with enforced network. However,
it is very important to notice that the integration capability of enforced network is not 3 MW.
This means that cases “maximum load demand – zero production” and “minimum load
demand – maximum production” are not reverse i.e. loading and production capabilities may
be different size at chosen location. The reasons for this are the chosen voltage set-value at
primary substation and the amount of minimum load demand at feeder studied.

Figure 1. Test system.

22

21,5

No production, old
21
No production, new
Production 1 MW, old
20,5
Production 1 MW, new
Voltage [kV]

Production 2 MW, old

20
Production 2 MW, new
Production 3 MW, old
19,5
Production 3 MW, new
Max voltage limit
19
Min voltage limit

18,5

18
0 0,5 1 1,5 2 2,5 3
Load demand [MW]

Figure 2. Voltage and load demand at bus 4.

2
This short example presents the complexity of MV network planning when DG units are
included at the network. The integration capability of a DG unit determined by voltage level
requirements is in general dependent on:
– location and type of DG unit
– location, type and size of other DG units and loads
– correlation of power production and load demand
– network parameters and configuration

The paper will present two complementary methods to calculate the effects of DG integration
on distribution network, where the conservative “minimum demand − maximum production”
planning principle is extended to consider stochastic nature of DG. Both methods are based
on local long-term wind or temperature statistics. Combined planning of distribution network
and DG units based on production statistics gives more flexibility for planning decisions
compared to the “minimum demand − maximum production” planning principle. These
methods are used to clarify the advantages of flexible planning principle and the requirements
on distribution system operation. The paper considers voltage and network transfer capability
issues.

2. IDEA OF PROPOSED PLANNING METHODS

2.1 Why “flexible” production?

If the enforcements of MV network are based on the most severe case i.e. on “minimum
demand − maximum production” case, this will introduce a severe economical barrier for DG
interconnection at weak distribution networks. The utilisation of existing MV network may be
improved in the case of DG unit integration if the operation of DG unit is not totally
independent of network conditions [2,3]. The proposed MV network planning methods are
based on fact that “flexible” production capacity may be larger than “fixed” production
capacity, when network constraints occur occasionally.

The flexible planning methods take into account stochastic nature of load demand, power
production and correlation between these. The planning of MV network is not based on a
single worst case but series of possible network conditions. The flexible production capacity
may require a control of DG unit power factor, a control of DG unit voltage or a curtailment
of production during few over-voltage hours in order to avoid over-voltages. The flexible
planning principle may allow integration of larger amount of DG on distribution network with
less network investments and connection charges than with the fixed worst case planning
principle.

2.2 Stochastic planning methods

Load duration curve [4]

In order to calculated the amount of production curtailment an approximate method to
calculated energy not produced is presented here. The method is used to get a rough idea
about energy not produced due to network constraints. The method is based on weather
statistics from plant site or from reference site, characteristics of DG unit (e.g. wind – power
curve), load duration curve of feeder studied, and predefined network transfer capabilities.
Network transfer capability may be a fixed capacity or it may be a flexible one where capacity

3
is dependent on network loading conditions.

The production of DG unit is simulated from yearly or monthly weather statistics, which is
presented in the next chapter. The simulated output power of DG unit is compared to the
network transfer capability in different loading conditions taken from a load duration curve.
The amount of electricity not produced is the sum of differences between the output and the
network transfer capability i.e. a production curtailment. The method is an approximate one
because it does not take into account simultaneous load and production conditions. Seasonal
variation in load and production may be taken into account by doing the analysis in monthly
basis.

The amount of energy not produced gives a clear financial indication what kind of network
investments and production curtailment contracts would achieve the optimum solution at the
specific case. The average value of energy not produced may be easily calculated which may
be used at the comparison of different choices of investments and contracts.

Load-flow simulation based on load and production curves

The accuracy of load duration curve method is not accurate enough for MV network planning.
That is why the load-flow calculations of existing network information systems were chosen
as a starting point for the development of appropriate methodology. The load-flow calculation
at network information systems at Finland are based on load curves [5]. The similar approach
was also chosen for production side.

Load-flow calculation of the network information systems is based on annual unit

consumptions and hourly load curve models of different customer groups. For using the load
curves in the load-flow calculation, customer data is obtained from the customer information
system. Using the load curves the load-flow can estimate the mean voltages and desired
exceeding probability levels using the given parameters. Existing automatic network planning
methods are based on calculation of maximum loading conditions and hence designed for
voltage drop calculations [6]. Consideration of minimum demand periods should be one
planning task in addition to a planning method based on maximum demand.

The application of both the load and the production curves at load-flow calculation makes
possible to simulate the functioning of the distribution system during a chosen period. The
load-flow simulation allows calculation of technical properties of the network (voltages,
power flows, losses etc.) which may also be used for calculation of variable costs and
incomes of a network. The most interesting changes of costs may appear at network losses,
market place fee and use of grid fee. Also the incomes from transfer charges may reduce
when the output of DG unit replaces part of load demand.

The production curves are not accurate in a similar way as load curves are. The load-flow
simulation with production curves is more or less a good guess what might happen. Especially
the results of certain hour would not be accurate due to uncertainty of future wind and
temperature. The stochastic properties of production curve will be valid only at the long-term
studies. The load-flow simulations are more capable for long-term planning in order to see
what kind of network conditions there might exists and to see the differences between voltage
control strategies or network enforcements. When different production curves are used at
load-flow simulations and the simulation results are examined together, the method will
converge towards probabilistic load-flow simulation e.g. Monte-Carlo simulation.

4
3. STOCHASTIC MODELLING OF LOADS AND PRODUCTION

3.1 Load curves

Electricity distribution network planning is based on load-flow calculation which uses

customer group based load curve models as initial data to determine network loading for
every hour of the year. Load curves are statistical by nature i.e. they are estimated to be
normally distributed and represented with mean values and variations. Variation describes the
uncertainty of load modelling. When population of certain customer group increases the total
variation decreases and therefore the load curve gives right result more probably. Estimation
of normally distributed value for load needs sufficient amount of customers to be accurate
enough. The load curve models for 46 different customer groups published by Association of
Finnish Electric Utilities are the most extensive collection of load models collected in Finland
[5].

Due to the many load points in MV and LV networks, the load data is usually insufficient. For
the estimation of load demand customers’ annual unit consumptions and customer group
based load curves are needed. The complete load model includes mean values and variations
of power demand for every hour of the year. The models are presented as 365*24 tables,
known also as topographies.

3.2 Production curves

Because the operation of DG units is based on

– weather conditions (wind, temperature, water flow),
– industrial or household heat consumption (district heating CHP application)
– electricity market price
also production curves have to be based on the general statistical data describing them.

In order to estimate the network effects of distributed generation better in planning it is

required to create hourly production curves for generators. Because at the planning stage data
from production units behaviour is unknown the establishment of production curves for
production which is more or less stochastic in nature could be based on the compiled statistics
about monthly wind speed [7] and temperature [8] distributions of the planned production site
or some reference site.

Wind power
Wind speed is usually represented with Weibull distribution e.g. in wind atlas files. Typically
the wind atlas includes yearly average wind speed and Weibull distribution’s parameters
(scale and shape factors) of the site in question. Finnish wind atlas contains also information
about monthly wind speeds and therefore it is possible to get the monthly average wind
speeds and Weibull distribution’s parameters from it.

From monthly wind speed distribution it is possible with random sampling to create hourly
wind speed curve for that month. However the randomness of the created hourly wind speed
curve has to be restricted to a certain extent to make wind speed behaviour more realistic i.e.
wind speed of consecutive hours is not allowed to differ more than some chosen maximum
value. With above restriction the hourly wind speed curve will be created several times for the
month in question until the average wind speed of the created curve differs only +/- 0.1 m/s

5
from the original wind speed distribution and it contains wind speed values also from both
ends of the original distribution. The latter condition can be fulfilled for example by checking
that the created wind speed curve’s Weibull distribution’s parameters are close enough to the
original distribution’s parameters.

Establishment of the wind generator’s hourly production curve for real power and production/
consumption curve for reactive power needs also other information than the hourly wind
speed curve. To create real power production curve the power curve of particular wind turbine
is needed with information about turbine’s nominal power and cut-in, cut-off and nominal
power wind speeds. Production/consumption curve for reactive power can be formed from the
real power production curve if generator’s real – reactive power dependence is known. Figure
3 illustrates the creation of the production curves for wind power.

Figure 3. Establishment of the real and reactive power production curves for wind power

CHP, district heating

Electric power production curve of combined heat and power (CHP) unit depends on the
application. In district heating electric power production follows CHP unit’s heat production
with some unit-specific ratio if no heat storage is used. CHP unit’s heat production is aimed to
fulfil customers heat demand which is heavily dependent on the outdoor temperature. From
the compiled statistics about monthly temperature distributions of the planned production site
it is possible to generate daily average temperatures in the same way as hourly wind speeds
were above taken from monthly Weibull distributions. It is assumed here that the monthly
average temperatures are normally distributed.

It is also presumed that it is possible to create combined proportional standard heat demand
curve for the heat customers of the particular area. Proportional heat demand curve could be
for example daily or weekly. With heat demand curve, daily average temperatures and initial
data needed one can determine CHP unit’s nominal heat power. The initial data includes
following things:
– Percentual proportion of unit’s nominal heat power from customers’ peak heat demand
– Percentual proportion of yearly heat demand which is dependent on the outdoor
temperature
– Customers’ total heat demand
– The reference indoor temperature for heat demand (usually +17°C) which is used to
calculate the heat demand number
– First and last day of the summer season (In summer season heat demand is not dependent
from the outdoor temperature and CHP unit is disconnected from the network)

After the CHP unit’s nominal heat power has been calculated it is possible to form the CHP
unit’s hourly heat production curve where summer season is also taken into account. With
unit-specific electric power / heat power ratio can CHP unit’s nominal electric power and

6
hourly electric power production curve be determined. Then production/consumption curve
for reactive power can be formed from the electric power production curve if generator’s
real–reactive power dependence is known. Figure 4 illustrates the creation of the production
curves for district heating CHP unit.

Figure 4. Establishment of the electric power production curve for district heating CHP unit

3.3 Shortcoming of production curves

Wind power
The above introduced method to compose production curve for wind turbines includes many
simplifications which make it somewhat inaccurate. In creation of hourly wind speed curves
following things have not been taken into account:
– height dependence of original wind atlas files wind speed data
– wind speed data in wind atlas files is cleared from terrain roughness and surrounding
obstacles
– wind speed curve is not necessarily based on measured data from planned production site
– hourly modelling of wind speed with some average value hides the varying nature of wind
and therefore wind speed curve with e.g. 10 minutes intervals could be better
– wind speed variation of consecutive hours is not based on real mathematical/physical
models and transition probabilities like for example Markov chains or ARMA models
– possible error resulting from Weibull fitting because Weibull distribution is usually only
appropriate in cases where average wind speed is quite high (> 7 m/s)

One way to get wind atlas files without surface wind measurements is to create them with so
called mesoscale model which is often more realiable when terrain is complex and surface
wind measurements are made low (< 50 m). Problem is that it needs heavy and expensive
computation which also requires an expert to carry it out.

In addition with hourly wind speed curves inaccuracies the hourly production curve for wind
turbines has also other simplifications like:
– it doesn’t question the reliability of power curve (temperature and air pressure corrections,
turbulence dependency, impact of other wind turbines in wind farms etc.)
– it doesn’t take the shadowing effect into account when wind turbines in wind farms are
considered
– it doesn’t take the limitations in availability into account (failures, maintenance etc.)

Because of all the above mentioned inaccuracies wind turbines yearly production (MWh) and
utilization period of maximum load (h) will become somewhat overoptimistic.

7
CHP, district heating
Also the method to compose electricity production curve for district heating CHP unit’s
includes many simplifications which make it inaccurate. Those simplifications and
assumptions include among others:
– Assumption of normally distributed monthly/daily average temperatures
– Hourly variation of temperature during a day is not taken into account
– Assumption of consecutive days maximum allowed difference in average temperatures
– Assumption of proportional standard daily/weekly heat demand curve (can be quite hard
to determine in practice)
– Estimation of customers’ total heat demand
– The limitations in availability are not taken into account (failures, maintenance etc.)

4. ADVANCED VOLTAGE CONTROL OF DISTRIBUTION SYSTEM

The control of voltage level is typically restricted be standard EN 50160 [9], which sets limits
for voltage level of LV network. However, the voltage control devices of distribution system
are situated on the HV/MV substation (e.g. on-line tap changer and reactive power
compensation capacitors) and along the MV network (e.g. off-line tap changers of MV/LV
transformers and reactive power compensation capacitors). Typically the continuous control
of voltage level has realised by on-line tap changer and other devices are used for
management of long-term voltage level and reactive power flow.

Many DG units are capable for continuous control of voltage level. This opportunity is,
however, remarkably restricted by interconnection contracts which set very narrow limits for
free of charge power factor, and sometimes even lost by using DG units at unity power factor.
It seems like similar interconnection contracts are applied for produces and consumers in case
of free of charge power factor. From MV network voltage control point of view, it would be
beneficial if a DG unit would take part in the voltage control of MV network. The voltage
control of MV network produced by a DG unit could be an ancillary service produced by a
DG unit for a network company. An ancillary service contract could replace network
reinforcement when interconnection of DG unit causes occasional voltage rise problems. The
network transfer capability may be increased by controlling the reactive power of the DG unit
[10-16].

The proposed voltage control concept includes two hierarchical levels. The first level is local
level and the second one is co-ordinated level. Figure 5 explains the basic idea of the
proposed voltage control method. The local voltage control is a traditional one, where local
controller maintains constant voltage or power factor at DG unit terminal. The controller
should be able to control excitation of synchronous machine, power factor correction of
induction machine, or reactive power production / consumption of frequency converter. This
control level must be the fastest one. The co-ordinated level introduces a system wide
perspective for voltage control of distribution network.

8
Figure 5. Proposed voltage control of distribution system.

4.1 Local voltage control

The local voltage control is naturally based on local voltage. The voltage control may be
implemented for example by
– control of HV/MV substation secondary voltage by on-line tap changer
– control of MV/MV substation secondary voltage by on-line tap changer (voltage boost /
reduction transformer)
– automatic voltage regulation (AVR) of synchronous generator
– control of reactive power flow into or out from DG unit by frequency changer or inverter
– continuous voltage control by static var compensator (SVC) or STATCOM
– power factor control (cosφ-control) by automatic or manual reactive power compensation
capacitors and reactors

The idea of local voltage control in case of DG is to reduce the voltage rise when necessary
by controlling the flow of reactive power into a DG unit. If a DG unit like synchronous
generator or frequency changer is capable for continuous voltage control that may be applied
to directly control terminal voltage.

Asynchronous generators consume reactive power when they are operated. Power factor
correction capacitors are typically used at DG unit terminal to compensate the consumption of
reactive power. Manually switched capacitors are connected during a start-up process of DG
unit and they are typically sized to fully compensate unit’s no-load consumption of reactive
power. Automatic compensation devices are capable for more accurate control by controlling
the power factor of DG unit within specified limits. The voltage rise of asynchronous
generator may be limited by introducing an additional input signal (voltage measurement) for
power factor controller. If the voltage is higher than specified maximum limit, then the
control limits of power factor are ignored and the amount of reactive power compensation is

9
reduced (or manually switched capacitor is disconnected) in order to reduce terminal voltage
and to continue the operation of DG unit. Otherwise the voltage relay would trip the unit if the
terminal voltage is higher than relay upper voltage setting.

The reactive power flows at MV network should be considered carefully at these cases by the
network company. The additional reactive power flow would increase costs of MV network
operation by increasing the losses of MV network, increasing the need for reactive power
compensation capacitors at HV/MV substation and increasing the number of on-line tap
changer operations. However, the consumption of reactive power at the location of DG unit
may be an interesting choice for network enforcement especially when voltage rise is
expected to happen occasionally.

4.2 Co-ordination of voltage control

Figure 6 presents the basic idea of distribution network planning when co-ordination of
voltage control is used. The available voltage drop margin is the minimum of voltage drop
margins of all feeders. The voltage drop margin is the difference between the minimum of
feeder voltages and the voltage minimum limit. Similarly there exist available voltage raise
margin corresponding to the maximum of feeder voltages and the voltage maximum limit.

The idea of co-ordination is to utilise the voltage drop and raise margins if these exists. If the
load demand is low, the voltage drop margin is large which could be utilised by reducing the
voltage set-value of HV/MV substation. This will result as reduction of voltage drop margin
but simultaneously it will increase the voltage rise margin, which may be utilised by
increasing the output of DG unit. When load demand decreases, there is a greater chance for
voltage rise problem due to high DG output, but also a possibility to decrease voltage setting
because voltage drop is less than during high load demand. The need for voltage resetting is
mainly determined by output power of DG units. The minimum voltage setting for light
loading condition is determined by voltage drop of feeders not including DG units.

Figure 6. Available voltage rise and drop margins.

10
The co-ordinated voltage control of MV network is based on resetting of voltage set-value at
primary substation. The capability for reduction of voltage set-value must be evaluated by
state estimation of MV network in order to avoid voltage drop problems at radial feeders. This
is done with distribution management system (DMS) which is based on [17]
– load curve based load-flow calculation (based on information on customer and network
information database)
– measurements and device status information via SCADA from the HV/MV substation and
the controlled DG units
– other measurements (e.g. on-line hourly load measurements).
The DMS is applied for on-line load-flow calculation of the distribution system and for
checking the state of system after a possible resetting of voltage set-value. DMS uses load
curves to estimate the voltage level of the system [18]. DMS uses also feeder current
measurements from HV/MV substation and other measurements to reset load curves in order
to fit the results of load-flow calculation and the actual measurements [19].

The co-ordination of distribution system voltage control enables also proper consideration of
conflicting network conditions which may appear when large amount of DG is connected to
one of the feeders of weak distribution network. There may occur network conditions where
the voltage setting of HV/MV substation should be raised and reduced simultaneously. In that
case the amount of production curtailment may be minimised when both the voltage drop and
raise margins are set to zero. This arrangement will also minimise the time when production
curtailment is needed and allows the management of power quality from voltage level point of
view of all customers.

4.3 Production curtailment

Similarly like the demand side management is capable for reduction of load demand during
peak hours of load demand to reduce voltage drop, the production curtailment is capable for
reduction of power production during over-voltage hours. The opportunity for production
curtailment may also benefit the production company although this may mean waste of wind,
solar etc. power. In case of CHP production there may be needed a heat storage device which
is also very beneficial when a CHP unit is operated at power markets. The storage of
electrical energy may also be applied to store otherwise wasted electrical energy, to maximise
the profit of electrical energy sold or to minimise the operational costs of production unit.

The production curtailment may be realised easily at hydro plants and relatively easy at CHP
plants by resetting set-value of turbine controller. CHP plants are more complicated because
the heat demand of the plant usually dominates the operation of the plant. The heat storage
device makes possible to separate the demand of heat and electricity from each other to some
extent without loss of energy if efficiency of heat storage is not considered. Otherwise CHP
plant should be operated to produce heat only or with reduced electric / heat power ratio
which would reduce efficiency of CHP plant.

The output control of wind power unit is probably the most complicated due to different kind
of technical implementations of these units. However, the wind power curtailment would
probably be the most important of all unit types due to location and stochastic mode of
operation of these units. The simplest method of wind power curtailment would be
disconnection of required number of units from problematic area. The order of disconnection

11
may be realised by voltage relay settings either by setting different upper voltage limits or by
setting different tripping delays for units. The output of fixed speed and stall controlled wind
mills may not be directly controlled hence the production curtailment must be realised by
disconnecting units. The output of variable speed and pitch controlled wind mill may be
controlled by frequency converter and blade angle control hence the unit may not be
disconnected which makes possible to control the output of the unit continuously. The
efficiency of wind mill will however be reduced at that case.

The possibility for production curtailment should be considered in the interconnection

contract. The control of DG unit may benefit both parties when network investments are
avoided. The control of DG unit output power is beneficial, if voltage rise problem appears
occasionally, e.g. during light loading and high production. The probability of this kind of
network condition is very rare and may be evaluated based on e.g. load curves and wind
statistics or measurements.

5. EXAMPLE

5.1 Test case

Simulation examples are done with MATLAB based load-flow calculations. Examples will
present a comparison of methods when different control concepts to avoid over-voltages are
tested. Advantages and disadvantages of planning methods are discussed based on a real life
distribution network of Fortum Sähkönsiirto Oy in south-west Finland including windmills
and combined heat and power units.

The examined MV network (Figure 7) consists of one HV/MV substation, which feeds five
MV feeders: Kasnäs (71 km), Byholmen (53 km), Dalsbruk (2,7 km), Dal (13 km) and
Dragsfjärd (42,5 km). In this paper network effects of distributed generation is studied in a
case where variable speed wind turbine is connected in Kasnäs feeder (node 9) and CHP unit
for district heating (without heat storage) is connected in Dalsbruk feeder (node 34). In
example simulations four 750 kW wind turbines are modelled as one 3 MW generator. Wind
turbines are connected to the Kasnäs feeder 22 km away from the HV/MV substation and they
are equipped with permanent magnet generators and frequency converters at stator circuit,
which allows power factor control between 0.92–1.00 inductive or capacitive. CHP unit’s
nominal electrical power is 1 MW and it is equipped with synchronous generator, which
allows reactive power control within PQ-diagram limits.

Calculations were carried out with an equivalent network, where some of the branch
conductors of the medium voltage are omitted. There is a 44 km long sea cable between nodes
6, 7 and 18, which produces about 1 MVAr of reactive power. To show the influence of this
long sea cable the example calculations in this paper are made with and without the sea cable.
The network effects of wind turbine and CHP unit with different voltage control methods
(local and co-ordinated) is studied in a situation where all feeders are radial operated and
voltage at the HV/MV substation is aimed to be 20.4 kV. In following simulation examples
the power factor of wind turbine was constantly held at extreme value (cos(ϕ)=0.92ind) to
limit voltage rise at Kasnäs feeder. Meanwhile the power factor of CHP unit was held at
cos(ϕ)=1 if maximum allowed voltage limit at Dalsbruk feeder wasn’t exceeded.

12
Figure 7. The studied equivalent MV network of Fortum Sähkönsiirto Oy

5.2 Comparison of planning and control methods

The Table 1 presents the advantages of flexible planning principle and voltage control with
wind mill at node 9. The present situation where fixed network planning principle is used with
unity power factor at DG unit site is very conservative and the utilisation of network
capability is very low. The fixed planning principle is suitable for traditional distribution
network planning but it is not capable for considering system wide aspects when DG is
integrated on distribution network. The management of distribution system when DG units
are integrated on it becomes anyhow more complicated than the management of traditional
distribution network. We think the active management and control (intelligent or active
network) would be a part of distribution system in future.

Table 1. The network transfer capabilities [kW] with wind mill.

Load demand Unity power factor Local voltage control Co-ordinated control
[kW] Fixed Flexible Fixed Flexible Fixed Flexible
364 850 850 1120 1120 2000 2000
473 850 900 1120 1250 2000 2080
572 850 950 1120 1300 2000 2160
690 850 1050 1120 1350 2000 2240
825 850 1150 1120 1550 2000 2320
992 850 1300 1120 1750 2000 2400
1174 850 1500 1120 2000 2000 2500

13
Table 2. Energy not produced due to network constraints at existing network with wind mill.
Network planning Voltage control Fixed Flexible
method method Energy not produced Energy not produced
[% of annual energy] [% of annual energy]
Load duration curve Unity power factor 48.5 41.4
Local voltage control 38.2 30.4
Co-ordinated control 14.5 10.4
Load-flow simulation Unity power factor 48.6 40.4
Local voltage control 38.0 23.3
Co-ordinated control 14.3 0

The planning method based on load duration curve is developed to calculate the amount of
energy not produced due to network constraints. That is why network planning methods are
compared only by this measure. The Table 2 shows the amounts of energy not produced when
3 MW wind mill is connected to node 9. The total available production is 9109.1 MWh which
is calculated from simulated wind. The difference between fixed and flexible planning
principles may also be seen from this table. Similarly the effect of voltage control methods
may be seen. The calculation results of energy not produced are similar with network
planning methods except two values at flexible planning principle. These are due to properties
of stochastic production curve, which would require simulation of series of production curves
in order to see the expected functioning of the system.

5.3 Analysis of load-flow simulation results

Simulations with sea cable

Some network effects of wind mill and CHP unit is gathered in Table 3 with different voltage
control methods. Simulations have also been done in situation where only wind mill is
connected to network to see the effects of CHP unit and wind mill more clearly.

Table 3. The network effects of wind mill and CHP unit with different voltage control
methods (with sea cable)
Co-
Local Co-
Local ordinated
No control ordinated
control control
production (wind + control
(wind) (wind +
CHP) (wind)
CHP)
Energy from HV network [MWh] 26 375 19 329 14 564 17 406 12 609
Energy from HV network
[MVArh] -7 407 -4 882 -5 054 -3 691 -3 823
Transmission charges [€] 87 211 62 350 43 049 55 775 38 075
Distribution losses [MWh] 571 566 551 779 767
Energy not produced [MWh],
Wind mill – 2 135 2 172 0 0
Energy not produced [MWh],
CHP unit – – 0 – 0
On-load tap changer’s hourly
operations [number] 44 140 114 1 923 1 956

14
It can be seen from Table 3 that produced reactive power mainly from long sea cable at
Kasnäs feeder is not compensated. This is due to fact that reactive power is not charged from
each HV/MV substation separately. Reactive power charging is based on the sum of few
HV/MV substations’ reactive power production or consumption. Therefore reactive power
produced in this studied MV network can be consumed by some other MV network of the
same distribution network operator. Table 3 also shows that reactive power feeding from
HV/MV substation to HV network decreases with all voltage control methods when
generation is connected into the examined MV network. The increment of reactive power
consumption raises the operational costs of MV network by increasing network losses and by
increasing requirements on reactive power compensation capacity.

When comparing voltage control methods where only wind mill is connected to MV network
one can see that transmission charges (market place and use of grid fees) are 6575 € lower
with co-ordinated control than with local control. However the distribution losses are 213
MWh higher with co-ordinated control. With assumption that price for distribution losses is
30 €/MWh the costs of distribution losses are 6390 € higher with co-ordinated control.
Although in this way the co-ordinated control seems a bit more favourable (6575 € - 6390 € =
185 €) than the local control one must also notice that the number of on-load tap changer’s
hourly operations is much higher with co-ordinated control which may increase the need for
maintenance of on-load tap changer.

Simulations without sea cable

Here the same simulations as above have been done without long sea cable. Table 4 clearly
shows that distribution losses are much lower without the long sea cable i.e. higher losses in
case when sea cable is connected into MV network are result from the transfer of reactive
power in MV network. When the long sea cable is not considered the operation of DG units
would increase network losses in every situation which wasn’t the case in above simulations
with the sea cable. This is remarkable result because there is a common believe that DG
would decrease network losses. Table 4 shows also that the network transfer capability might
be increased by compensating the reactive power of long sea cable locally.

Table 4. The network effects of wind mill and CHP unit with different voltage control
methods (without sea cable)
Co-
Local Co-
Local ordinated
No control ordinated
control control
production (wind + control
(wind) (wind +
CHP) (wind)
CHP)
Energy from HV network [MWh] 26 073 18 144 13 311 17 268 12 470
Energy from HV network
[MVArh] 273 3 626 3 510 4 209 4 084
Transmission charges [€] 86 218 57 990 39 451 55 357 37 640
Distribution losses [MWh] 270 491 484 640 629
Energy not produced [MWh],
Wind mill – 1 025 985 0 0
Energy not produced [MWh],
CHP unit – – 0 – 0
On-load tap changer’s hourly
operations [number] 47 260 182 1 130 1 161

15
6. CONCLUSIONS

The proposed planning and voltage control methods are tested and compared with a real-life
distribution system. The test example clearly shows the capability of flexible planning method
to increase the network transfer capability when that is compared to traditional “minimum
load demand – maximum production” planning principle. The proposed co-ordinated voltage
control method may be applied to further increase the network transfer capability. The
proposed voltage control method is developed for cases where the permissible voltage
variation range is exceeded occasionally. When the integration of DG unit is considered at
weak distribution network the combination of flexible planning method and co-ordinated
voltage control method may produce great advantages for both the network and the
production companies. The capability of these methods may be evaluated by the load-flow
simulations.

The stochastic load-flow simulation is based on similar computational approach than existing
commercial network information systems and distribution management systems have.
Load-flows are computed hourly and they are based on stochastic load and production curves.
Especially load-flow simulations are very useful when the functioning of distribution system
is needed to be analysed when DG units are going to be integrated on network. Simulations
may be used to examine both the technical and the economical issues. The integration of DG
unit on distribution system will change operational costs like market place fee, use of grid fee
and costs of network losses. Also income of the network company will change remarkably
when the production of DG unit replaces load demand at customer side. At the worst case
there will be maintenance and operational costs about the network connection, but no at all
income when DG unit follows load demand on customer side.

ACKNOWLEDGEMENT

The authors are grateful to the partners at Tekes (Finnish Technology Agency) project
“Integration of distributed generation into electricity distribution network” for the interesting
real life research topic, comments, funding and research co-operation. The research has also
obtained funding from the Academy of Finland at project “Power system security assessment
and management”.

REFERENCES

[1] N. Jenkins, et al., 2000, Embedded generation, IEE Power and Energy series 31,
London, UK, 273 p.
[2] C.L. Masters, “Voltage rise the big issue when connecting embedded generation to long
11 kV overhead lines”, Power Engineering Journal, February 2002, pp. 5-12.
[3] W.J. Laycock, “Intelligent networks”, Power Engineering Journal, February 2002, pp.
25-29.
[4] J. Matevosyan, Wind power in areas with limited export capability, Licentiate Thesis,
Department of Electrical Engineering, Royal Institute of Technology, Stockholm,
Sweden, 2004, 105 p.
[5] A. Seppälä, Load research and load estimation in electricity distribution, VTT
Publications 289, Espoo, Finland, 1996, 118 p. [On-line] Available:
http://www.enease.fi/asepthes.pdf

16
[6] E. Lakervi and E.J. Holmes, 1995, Electricity distribution network design, IEE Power
series 21, London, UK, 325 p.
[7] VTT Processes, Wind energy statistics in Finland - Monthly statistics. [On-line]
Available: http://www.vtt.fi/pro/pro2/tuulitilastot/monthly.htm
[8] Finnish meteorological institute, Climatological statistics of Finland 1971-2000, 2002,
(in finnish).
[9] Voltage characteristics of electricity supplied by public distribution systems, Standard
EN 50160, 2000.
[10] S. Repo, et al., “Ring Operation of Distribution System in Case of Distributed
Generation ― Voltage Control”, Proceedings of Nordic Wind Power Conference,
Göthenburg, Sweden, March 2004.
[11] S. Repo, et al., “A Case Study of a Voltage Rise Problem Due to a Large Amount of
Distributed Generation on a Weak Distribution Network”, Proceedings of IEEE Power
Tech, Bologna, Italy, August 2003.
[12] S. Repo, et al., “A case study of a voltage rise problem due to a large amount of
distributed generation on a weak distribution network”, Proceedings of PowerTech,
Bologna, Italy, June 2003.
[13] S. Repo, et al., “A method to increase the integration capacity of distributed generation
on weak distribution network”, Proceedings of 17th International Conference on
Electricity Distribution (Cired), Barcelona, Spain, May 2003.
[14] E. Bompard, et al., “Voltage control in radial systems with dispersed generation”,
Proceedings of IEEE PowerTech, Budapest, Hungary, September, 1999.
[15] S.N. Liew and G. Strbac, “Maximising penetration of wind generation in existing
distribution network”, IEE Proceedings in Generation, Transmission and Distribution,
Vol. 149, No. 3, May 2002.
[16] J.-H. Choi, and J.-C. Kim, “Advanced voltage regulation method of power distribution
systems interconnected with dispersed storage and generation systems (revised)”, IEEE
Transactions on Power Delivery, Vol. 16, No. 2, April 2001, pp. 329-334.
[17] Administrator Manual, Distribution Management System Open++ Opera v.3.3, 2002.
[18] P. Verho, et al., “An intelligent support system for distribution network management”,
International Journal of Engineering Intelligent Systems for Electrical Engineering and
Communications, Volume 4, Number 4, December 1996, pp. 219-227.
[19] A. Seppälä and M. Lehtonen, ”Distribution load estimation - a function for the electric
utility in the de-regulated electricity market”, Proceedings of Energy Management and
Power Delivery, November 1995.

17