Grid Modeling, Analysis and Simulation of Different ... - Semantic Scholar
Grid Modeling, Analysis and Simulation of Different Scenarios for a Smart Low-Voltage Distribution Grid L. Mihet-Popa, X. Han, H. Bindner
J. Pihl-Andersen*, J. Mehmedalic**
Department of Electrical Engineering Technical University of Denmark Roskilde, Denmark [email protected]
SEAS-NVE*, Dansk Energi** Svinninge*, Copenhagen**, Denmark [email protected], [email protected]
Abstract—This paper presents the modeling, analysis and simulation of a low-voltage distribution grid model based on the real data designed for evaluation of a future smart grid. The grid model is built measuring the distribution lines’ length and considering the cable dimensions and lengths, the grid age, the number of cabinets and customers and the load per customer. The aim of the model is to design, implement and test the proposed configuration and to investigate whether the lowvoltage distribution grid is prepared for the expected future increase of PV penetration, heat pumps and electric cars. The model is implemented in NEPLAN and DIgSILENT Power Factory and different scenarios are developed and analyzed. A time series simulation is conducted for a specific scenario with a comparison between different voltage and load profiles along the feeders. Index Terms—Low-Voltage Distribution Grid; DIgSILENT Power Factory; PV Penetration.
I.
NEPLAN;
INTRODUCTION
One of the main goals in the Danish energy policy is to increase the amount of renewable energy in the energy mix to 30% of the energy in 2025 [1]. Last year, the Danish parliament approved an even more ambitious target: to have renewable supply 35 % of the country’s total energy needs-not just electricity but also heating and transportation-by 2020, and an incredible 100 % by 2050 [2]. The increase in solar penetration will affect operation and design of distribution systems [3]. More than that, due to changes in the way electrical energy is produced and used, distribution network operators must adapt to changing usage patterns: the penetration of renewable energy will continue to grow and electricity is expected to increasingly substitute fossil fuel in areas such as transportation and building heating [1-3]. Increased distribution generation is becoming more important in the current power system. In the future it will rely more on distribution energy resources and on smart-grids [4][5]. In the future smart-grid distribution systems must be flexible and to be able to import/export the power from/to the
grid, to control the active and reactive power flows and to manage the storage of energy [4-9]. This paper presents the design and implementation of a representative low-voltage distribution grid model, based on real data for summer and detached houses, for evaluation of a future smart-grid. The aim of the model is to design, implement and test the proposed configuration and to investigate whether the low-voltage distribution grid is prepared for the expected future increase of PVs, heat pumps and electric cars. The model was implemented in NEPLAN and DIgSILENT Power Factory to study different scenarios. A steady-state and a dynamic analysis of the models have also been presented with a comparison between different voltage profiles. II.
LOW-VOLTAGE GRID MODEL
The aim of this section is to design and build a representative low-voltage grid for summer houses and detached houses. The grid model contains many components, such as: PV systems, EV systems, heat pumps and conventional loads and will be designed and tested to find out whether the proposed low-voltage distribution grid is prepared for the expected future increase of PV penetration, heat pumps and electric cars connected along the feeders. A. Grid Model Setup and Database Building Many representative areas have been looked through for lowvoltage feeders containing the two types of houses (summer houses and detached ones). The length of the low-voltage feeders is measured using a GIS (Geographical Information System) map. The number of customers, number of cable cabinets and the number of customers per cable cabinet for each feeder are collected and the maximum load per customer is calculated using the Velander correlation [16] which assumes that the load along the feeders is normally distributed. Then a grid model was constructed from the data using a statistical percentile. To build a database for the grid model the data from 334 lowvoltage feeders was mapped. The following parameters are taken into consideration: cable dimensions and lengths, the
grid’s age, category, substation no., no. of customers, no. of cabinets and the load per customer. The measured feeder length for the 85 % was 500 m for detached houses and 730 m for summer houses. We consider only two types of cables (with a cross-section of 150 mm2 AL and 95 mm2 AL) because 240 AL and 50 AL cables represent a very small part of the total cable length investigated. Based on measurements a regression analysis could establish a statistical correlation between the consumers’ annual energy consumption and the maximum load of the grid caused by their consumption. The feeder’s aggregate load was calculated using the Velander correlation:
Pmax = α ⋅ w + β ⋅ w
(1) In which Pmax represents the maximum load (measured in kW), w is the annual consumption (measured as MWh), α and β are the Velander constants. We used the following values for Velander constants: α=0.29 and β=2.09 for detached houses and α=0.32 and β=3 for summer houses. B. Grid Model Components The grid model contains conventional residential loads (houses), heat pumps (with 1 and 3 phases), electric cars (1 phase and 3 phases) and PV systems (1, 2 and 3 phases). In the case of one phase appliances it is assumed that all the devices are connected to the same phase. Similarly 2 phase PV systems are assumed to be connected to the same 2 phases. For PV systems 3 parameters were considered: rated voltage, tap position of the 10/0.4 kV transformer and the number of phases (1, 2 or 3). The position of the tap-changer normally depends on the primary voltage level on the 10 kV side. If the distribution transformer is located at the end of a long 10 kV feeder, the tap-changer is used to normalize the voltage level on the secondary side. In this study the tap changer position is used to allow a larger voltage drop/rise in some of the scenarios. The connection of heat pumps and electric cars to the lowvoltage grid, along the different feeders, will cause a voltage drop. C. Scenarios Developed to Design and Test Grid Model The grid model was designed, based on SEAS-NVE’s supply area and was analyzed and tested for various load scenarios through connection of appliances such as heat pumps, EV’s and PV systems. 23 scenarios were developed using 2 different voltage limits, +6%/-10% and +10%/-10%. The first limit is based on DEFU recommendations no. 16/2001 [12]. These limits were used in Denmark during a grace period after EU legislation changed nominal voltage from 220 V to 230 V in order to allow for older 220 V household appliances to be phased out. The +10%/-10% limit is based on the latest version of DEFU recommendation 16, which closely mirrors European norm EN 50160 [13]. It is estimated that the worst case scenarios is a summer day with low load (20 % of max.) with the PV’s connected to the grid.
In the first four scenarios it is investigated how many solar cells can be connected to the grid along the feeder without to exceed the voltage limits defined in [12] and [13], when the PV systems have three-phases and 6 kW each with the residential load at 20 % of the maximum load. In the next five scenarios it is investigated what happens in the same conditions when one phase and 4 kW and two phases and 6 kW PV systems are connected along the feeder. In the next seven scenarios we investigated how many electric cars (3 phases EV of 11 kW and one phase EV of 3.7 kW each) can be connected to the grid, when the residential load is at the maximum load, without exceeding the lower voltage limit. We have investigated in the last five scenarios the connection of heat pumps (3 phases HP of 2.8 kW each and one phase HP of 1.5 kW each), to the grid, evenly distributed along the feeder, with the residential loads at maximum load. In one of these last five scenarios we considered a mixed load case (the worst case scenario) when heat pumps (one phase and three phases) and electric cars are connected simultaneously on the same distribution grid. III.
GRID MODEL IMPLEMENTATION
Based on investigations, calculations and assumptions presented in the last section a grid model for implementation of all 23 scenarios has been designed. The model contains an external grid, a distribution substation with a medium voltage transformer (10/0.4 kV) and with ten cabinets (switch boards) at which PV panels, residential loads, EVs and heat pumps are connected. The grid model was implemented in two different tools, NEPLAN and DIgSILENT Power Factory, to study load flow, steady-state voltage stability and dynamic and transient behavior of the power system. These tools have been selected as they have the ability to simulate load flow and RMS fluctuations in the same software environment [14, 15]. In Fig. 1 is presented the implementation of the grid model for a low-voltage distribution system in NEPLAN (a) and DIgSILENT Power Factory (b).
a)
usbar/BB_HV
Ul=10,00000 kV u=1,00000 p.u. phiu=0,00000 deg
P=-47,3023 kW Q=5,5354 kvar cosphi=-0,9932
Trafo 10/0.4
Line
Busbar(1)/BB_LV Ul=0,40087 kV u=1,00217 p.u. phiu=-149,42033 deg
P=-130.589 Q=-16.318 cosphi=-0.992
P=114.96 Q=12.322 cosphi=-0.994
P=-47,.. Q=5,06.. I=68,7..
Line(1)
P=50,4.. Q=-4,1.. I=68,7..
P=-58,.. Q=3,63.. I=79,3..
SingleBusbar(2)/BB1 Ul=0,4..
Line(2)
P=62,1.. Q=-2,3.. I=79,3..
SingleBusbar(3)/BB2 Ul=0,4..
u=1,06.. phiu=-..
P=1,82.. Q=0,45.. I=2,55..
P=-6,0.. Q=-0,0.. I=8,16..
Low-Volta..
PV1
Line(3)
P=-63,.. Q=1,98.. I=81,1..
P=63,8.. Q=-1,9.. I=81,1..
P=-65,.. Q=1,54.. I=83,0..
SingleBusbar(4)/BB3 Ul=0,4..
u=1,13.. phiu=-..
P=1,45.. Q=0,36.. I=1,91..
P=-6,0.. Q=-0,0.. I=7,67..
Low-Volta..
PV2
Skab4
u=1,13.. phiu=-..
P=6,01.. Q=0,00.. I=7,67..
P=1,45.. Q=0,36.. I=1,91..
u=1,13.. phiu=-.. P=1,45.. Q=0,36.. I=1,90..
Low-Volta..
PV1(..
Low-Volta..
P=0,00.. Q=0,00.. I=0,00..
Ul=0.386 u=0.964 phiu=-151.941
Line(6)
P=-68,.. Q=0,66.. I=86,7..
Ul=0,4.. u=1,14.. phiu=-.. P=1,45.. Q=0,36.. I=1,89..
P=69,0.. Q=-0,5.. I=86,7..
SingleBusbar(7)/BB6 P=0,00.. Q=0,00.. I=0,00..
Ul=0,4.. u=1,14.. phiu=-.. P=2,73.. Q=0,68.. I=3,53..
Low-Volta.. Static Ge..
Low-Volta..
P=-71,.. Q=-0,1.. I=90,1..
Line(7) P=72,1.. Q=0,19.. I=90,1..
SingleBusbar(8)/BB7 P=0,00.. Q=0,00.. I=0,00..
Static Ge..
Ul=0,46220 kV u=1,15549 p.u. phiu=-146,24109 deg
P=1,37.. Q=0,34.. I=1,76..
Low-Volta..
P=0,00.. Q=0,00.. I=0,00..
P=-49,.. Q=-0,5.. I=61,7..
Line(8) P=49,6.. Q=0,54.. I=61,7..
P=-26,.. Q=-0,8.. I=33,5..
SingleBusbar(9)/BB8 Ul=0,4..
u=1,15.. phiu=-.. P=1,45.. Q=0,36.. I=1,86..
P=-23,.. Q=-0,0.. I=29,9..
Static Ge.. PV7
Low-Volta.. Static Ge..
P=-23,.. Q=-0,0.. I=29,8..
PV8
D10 L56
P=-4,3.. Q=-0,4.. I=5,46..
u=1,16.. phiu=-.. P=1,45.. Q=-0,3.. I=1,85..
Low-Volta..
P=-23,.. Q=-0,0.. I=29,7..
PV9
P=0,00.. Q=0,00.. I=0,00..
P=106.732 Q=10.187 cosphi=-0.995
P=4,3793 kW Q=0,4664 kvar I=5,4681 A
Static Ge..
Ul=0,46500 kV u=1,16251 p.u. phiu=-146,17348 deg P=1,71.. Q=-0,4.. I=2,19..
P=6,02.. Q=0,01.. I=7,48..
Branches
P=68.498 Q=6.091 cosphi=-0.996
P=-67.940 Q=-5.969 cosphi=-0.996
L67
Low-Volta.. Static Ge.. PV10
Project: Graphic: Grid10 Date: 1/21/2013 PowerFactory 14.1.6
P=-105.883 Q=-9.909 cosphi=-0.996
P=31.155 Q=2.283 cosphi=-0.997
P=-31.037 Q=-2.257 cosphi=-0.997
L89
P=0,00.. Q=0,00.. I=0,00..
Load Flow Balanced Line-Line Voltage, Magnitude [kV] Active Power [kW] Voltage, Magnitude [p.u.] Reactive Power [kvar] Voltage, Angle [deg] Current, Magnitude [A]
L910
L78
Line(9)
P=27,0.. Q=0,87.. I=33,5..
P=-113.996 Q=-12.005 cosphi=-0.995
Skab5
Nodes
Ul=0.394 u=0.984 phiu=-151.727
L45
Static Ge..
SingleBusbar(10)/BB9 Ul=0,4.. SingleBusbar(11)/BB10 P=0,00.. Q=0,00.. I=0,00..
D04
Ul=0.390 u=0.974 phiu=-151.834
P=-66,.. Q=1,10.. I=84,8..
Line(4) Line(5) 67,2.. 1,0.. 4,8..
P=-151.098 Q=-21.773 cosphi=-0.990
Skab1 Skab2
Ul=0.382 u=0.955 phiu=-152.048
Ul=0.378 u=0.946 phiu=-152.154
Ext.
LB1 P=-149.535 Q=-21.262 cosphi=-0.990
P=131.824 Q=16.721 cosphi=-0.992
Skab3 P=65,5.. Q=-1,4.. I=83,0..
SingleBusbar(5)/BB4 Ul=0,4..
P=-151.098 Q=-21.773 cosphi=-0.990
P=140.459 Q=18.989 cosphi=-0.991
L23
P=-47,.. Q=5,53.. I=2,74..
P=47,4.. Q=-5,0.. I=68,7..
T10/0.4
L12 P=-139.082 Q=-18.539 cosphi=-0.991
P=123.326 Q=14.499 cosphi=-0.993
P=-122.227 Q=-14.140 cosphi=-0.993
P=-122.227 Q=-14.140 cosphi=-0.993
Ul=10.000 u=1.000 phiu=-0.000
DIgSILENT
L34
External Grid
P=87.630 Q=8.096 cosphi=-0.996
Skab6 Ul=0.375 u=0.938 phiu=-152.260
P=-86.736 Q=-7.900 cosphi=-0.996
Skab8
Skab7 Ul=0.372 u=0.930 phiu=-152.366
P=49.693 Q=4.158 cosphi=-0.997
Ul=0.368 u=0.921 phiu=-152.440
P=-49.394 Q=-4.093 cosphi=-0.997
Skab9 Ul=0.365 u=0.913 phiu=-152.500
Skab10 Ul=0.363 u=0.908 phiu=-152.547
Ul=0.362 u=0.904 phiu=-152.578
Fig. 2. Power flow calculation for scenario 10 using DIgSILENT Power Factory.
Annex:
b) Fig. 1. Comparison between a) NEPLAN and b) DIgSILENT Power Factory implementation for the designed distribution grid.
All components of the single line diagram, presented in Fig. 1, are built with standard blocks from the library. The blue square above the cable cabinet (Fig. 1a) shows the voltage on each phase, while the green squares show the power and current for each phase. Also, the arrows show that the components are active. In DIgSILENT Power Factory implementation (Fig. 1b) the squares above the bus-bars contain the parameters of the cables and below the bus-bars show the voltages of each phase. Also, each component of the cabinet, such as PV systems, heat pumps and EVs has its own square block able to show internal parameters (current, power, power factor). Fig. 1 shows a power flow calculation for scenario no. 4, where it was investigated how many PV systems (3 phase 6 kW each and Un ±10 %) could be connected to the distribution grid, with a residential load at 20 % of maximum load. In this case the voltage increase may not exceed 440 V in the most remote cable cabinet. We have connected 4 PV systems at each cabinets of the feeder (40 PV systems in total), and a maximum voltage (438 V) was observed. A power flow calculation for the scenario 10 is presented in Fig. 2. In this case it was investigated how many electric cars can be connected to the grid before voltage decreases below the minimum limit (Un -10 %). In this scenario, six electric cars (11 kW & 3 phase each) were distributed over the five most remote cable cabinets, two electric cars in the last remote cable cabinet and one electric car in each of the next four cable cabinets. A minimum voltage of 359 V was observed. The residential load in this case is regulated to the maximum with the 10 kV stable voltage on the primary side of the substation with a conversion factor of 25.
IV.
SIMULATION RESULTS
A. Steady-State Analysis In this section a time series simulation is conducted using DIgSILENT Power Factory. To investigate the characteristics of the new components: PVs, EVs and HPs, we used the same grid model proposed in the last section. The daily time series in Fig. 3 (i.e. external grid voltage profile, PV production, EV charging consumption, and the detached house consumption) are the inputs of the simulation model. The data with voltage profile (the dash curve in Fig. 4) and PV production curve (06:00 – 20:00) is collected on 15-05-2012 from a laboratory with real components and renewable energy production (SYSLAB), at the Technical University of Denmark-Risø campus. The capacities of PVs are adopted to 6 kW. EV charging profile, from (18–22), is calculated based on the driving pattern [17] with the rated charging power of 11 kW. The load profiles are calculated by the load pattern, from the aggregated historical data in Denmark, multiplying the peak power using the Velander equation (1). By subtracting the production or adding the consumption new household load profile can be obtained. The time sweep load flow is executed by using DIgSILENT programming language DPL. Fig. 3 shows that at 19:00, the modified load has a peak due to the EV charging (Scenario 10). However, considering the deviation of human behaviors (smoothing effect), the over all peak may be less than the sum of them. At 15:00, the reversed load reach the peak where there is the largest mismatch between load and PV production (Scenario 2). A voltage rise can also be observed. Fig. 4 shows the simulation results for one day with a comparison between voltage profiles in the last node versus voltage profile of the external grid. Using this approach (i.e., generate the profiles outside the power system software and use its traditional function) the complexity of the models can be avoided, but introducing a lot of manual work on modifying the inputs.
The dynamic load model is implemented using a voltage dependency of active power and it is described by ([14], [18]): ( p2 ) ( p3 ) ⎡ ⎛ V ⎞( p1) ⎛V ⎞ ⎤ ⎛V ⎞ P = P0⎢kp1⎜⎜ ⎟⎟ + kp2⎜⎜ ⎟⎟ +(1−kp1 −kp2)⎜⎜ ⎟⎟ ⎥ (2) ⎢⎣ ⎝V0 ⎠ ⎝V0 ⎠ ⎥⎦ ⎝V0 ⎠
Fig. 3. Time series implementation of the load profiles for residential load (based on Velander correlation (1)), EV load and PV load (distinguished from the conventional load with a peak power multiplication factor) based on the load pattern from the aggregated historical data in Denmark.
Where P0 and V0 are the initial values of power and voltage, p1 to p3 are coefficients that define the proportion of each component and kp are coefficients that reflect the dependency of load (kp1=100%, kp2=0). Typically, in dynamic simulation, a simplification is done and all loads are consider to be constant admittance type (p1=2, p2=p3=0). In Fig. 5 are presented the time-series simulation results, with PV systems and loads connected together to the same bus-bar, for 6 days in November 2012. The input data for the simulation model (irradiation and temperature) was measured from the weather station placed on Risø campus with a sampling time of 1 second. The load profile was defined by (2) with the voltage dependency of the active power.
Time series simulation provides an overview of the potential problems (thermal loading and voltage rise in our case). Based on this the concurrency of a group of units can be investigated towards less conservative design.
In Fig. 5 b) are shown the simulation results when 3 PV inverters are connected together with the loads at the last 3 cabinets of the grid model presented in Fig. 1. In this particular case, EVs and HPs are assumed to be an unknown proportion of the total load and the PV systems have the same parameters but the panels have a different orientation and tilt angles. DIgSILENT
Fig. 4. a) Comparison between voltage profile of the external grid and the voltage profiles of the PV’s, EV’s, base loads and all loads connected at the last cabinet to the end of the feeders.
Fig. 5 a) shows a comparison between input data (Tcell between -4 and +6 0C and Gcell around 300 W/m2 during the day) and the output parameters of the model [output power of the panels-Pdc1 +Pdc2 (Pdc_total=2.7 kW) and output power of the inverter-Pac =2.6 kW).
9,00
B. Time-series Simulations using Dynamic Models The dynamic model of the PV System (PV panels and PV inverter) has been built with standard block components from the Power Factory library and also using the dynamic simulation language (DSL). It is based on a single diode equivalent electrical circuit for the PV model, described by an exponential equation [9-11]. The model uses the cell irradiation Gcell and cell temperature Tcell as inputs, measured from a weather station and implemented as look-up tables, as it is shown in the first two graphs of Fig. 5 a).
6,00
3,00
0,00
-3,00
-6,00 -2.78E-5
2.88E+1
5.76E+1
8.64E+1
1.15E+2
[h]
1.44E+2
5.76E+1
8.64E+1
1.15E+2
[h]
1.44E+2
5.76E+1
8.64E+1
1.15E+2
[h]
1.44E+2
PV_3b_MeasFile: Tcell (C)
0,40
0,30
0,20
0,10
0,00
-0,10 -2.78E-5
2.88E+1 PV_3b_MeasFile: Gcell (p.u.)
The PV panels are mounted in three strings: two of them having 18 panels of 165 W each, and the 3rd one having 12 panels of 100 W [9]. The strings of panels are connected to the low-voltage distribution network through a three-phase PV inverter. More details can be found in [9]-[10]. The simulation model was validated using experiments carried out using RISOE experimental facility-SYSLAB ([9][11]).
2,9077 2,2943 1,6809 1,0675
0,4541 -0,1593 -2.78E-5
2.88E+1 PV_Inverter_block: Pac (kW) PV_Inverter_block: Pdc (kW)
Subplot/Diagramm Date: 8/15/2013 Annex: /2
a)
DIgSILENT
2,419 1,909 1,398 0,888 0,378 -0,133 -2.8E-5
2.9E+1 PV_Inverter_block: Pdc1(kW) PV_Inverter_block: Pdc2(kW) PV_Inverter_block(1): Pdc3(kW) PV_Inverter_block(1): Pdc4(kW) PV_Inverter_block(2): Pdc5(kW) PV_Inverter_block(2): Pdc6(kW)
5.8E+1
8.6E+1
1.2E+2
[h]
components placed along the feeders in a future smart-grid distribution network.
1.4E+2
ACKNOWLEDGMENT This work was supported in part by the E.U. Project-Smart Plan, No. 55807/2011-2013.
2,820 2,225 1,630 1,035 0,440 -0,155 -2.8E-5
2.9E+1 PV_Inverter_block(1): Pac1 (kW) PV_Inverter_block: Pac3 (kW) PV_Inverter_block(2): Pac2 (kW)
5.8E+1
8.6E+1
1.2E+2
[h]
1.4E+2
REFERENCES [1]
20,00 15,00
[2]
10,00 5,00 0,00 -5,00 -2.8E-5
[3] 2.9E+1 Load_file2: Pload (kW)
5.8E+1
8.6E+1
1.2E+2
[h]
1.4E+2
Date: 8/16/2013 Subplot/Diagramm(1) Annex: /3
[4]
b) Fig. 5. a) Time series simulation results of the PV system for 6 days with real data implemented as input and with DC and AC powers as outputs; b) Comparison between IN and OUT of the 3 PV inverters, connected to the last 3 cabinets of the grid model together with a dynamic load defined by (2).
[5] [6] [7]
CONCLUSION In this paper we have proposed a representative low-voltage grid for summer and detached houses based on real data measurements. The feeder lengths were measured using GIS maps. The number of customers, cable cabinets and of customers per cable cabinet were collected and the maximum load per customer has been calculated using the Velander correlation. The grid model contains many components, such as: PV systems, EVs, heat pumps and residential loads and was designed and tested, based on 23 scenarios, to find out whether the proposed distribution grid model is prepared for the expected future increase of PV penetration with heat pumps and electric cars connected along the feeders. The developed scenarios clearly have shown that there is room for larger loads if the output voltage from the substation can be optimally set and/or varies according to the type and the size of the load. Also, in a weak distribution line with a high output voltage only 3 phase PV systems should be installed, as 1 phase PVs under the same conditions are more likely to exceed the voltage limits. The low-voltage distribution grid model has been developed and implemented in NEPLAN and DIgSILENT Power Factory to study load flow, steady-state voltage stability and dynamic behavior of the components. The comparison between both simulation tools has shown a good alignment and the possibility to use them for further developments, regarding the integration of smart-grid technologies. It means that this work could be used for development and improvements of the models for different
[8] [9]
[10]
[11]
[12] [13] [14] [15] [16] [17] [18]
"Smart Grid i Danmark", joint report by energinet.dk and Dansk Energi, September 2010. J. Kumagai, ”The Smartest, Greenest Grid-What a little Danish island is showing the world about the future of energy”, IEEE Spectrum, May 2013, pp. 38-43. Y. Riffonneau, S. Bacha, S. Barruel and S. Ploix, “Optimal power management for grid connected PV systems with batteries”, IEEE Transaction on Sustainable Energy, vol. 2, no. 3, pp. 309-320, July 2011. H. Jiayi, J. Chuanwen, and X. Rong, “A review on distributed energy resources and MicroGrid”, ELSEVIER Renewable & Sustainable Energy Reviews, vol. 12, pp. 2472-2483, 2008. “Smart Grid: Reinventing the electric power system”, IEEE Power & Energy Magazine, March 2012. P.C. Loh, L. Zhang and F. Gao, “Compact integrated energy systems for distributed generations”, IEEE Transactions on Industry Electronics, Vol. 5, May 2012. M. Jansen, R. Louie, M. E. Amoli and F. Sami, “Model and simulation of a 75 kW PV solar array”, in Proc. of 2010 IEEE PES Transmission and Distribution Conference and Exposition, pp. 1-5. H. Liu, L. Jin, D. Le and A. A. Chowdhury, “Impact of high penetration of solar photovoltaic generation on power system small signal stability”, in Proc. of 2010 POWERCON, pp. 1-7. C. Koch-Ciobotaru., L. Mihet-Popa, F. Isleifsson and H. Bindner, „Simulation model developed for a small-scale PV-System in a distribution network”, in Proc. of the 7th IEEE SACI Conf., 2012, Timisoara-Romania, May 24-26, pp. 257-261. L. Mihet-Popa, C. Koch-Ciobotaru, F. Isleifsson and H. Bindner, „Development of tools for DER Components in a distribution network”, in Proc. of ICEM Conf., September 2-5, 2012, MarseilleFrance, pp. 1022-1031. Y. Zong, L. Mihet-Popa, D. Kullman, A. Thavlov, O. Gehrke and H. Bindner, „Model Predictive Controller for Active Demand Side Management with PV Self-Consumption in an Intelligent Building”, IEEE PES Innovative Smart Grid Technologies Europe, BerlinGermany, October 14-17. DEFU recommendations no. 16/2010. European Standard-EN 50160. DIgSILENT PowerFactory, Digsilent gmbh, November 2012. NEPLAN, Tutorial manual, 2011. M. Velander, “Methods of operational analysis applied to distribution of electric power”, Journal of Teknisk Tidskrift, Vol. 82, 1952, pp. 293299. [Online].Available: http://www.scb.se/Pages/Product____10616.aspx?Produktkod=TK1101 &displaypublications=true; P. Kundur, Power system stability and control, McGraw-Hill, New York, 1994.
J. Pihl-Andersen*, J. Mehmedalic**
Department of Electrical Engineering Technical University of Denmark Roskilde, Denmark [email protected]
SEAS-NVE*, Dansk Energi** Svinninge*, Copenhagen**, Denmark [email protected], [email protected]
Abstract—This paper presents the modeling, analysis and simulation of a low-voltage distribution grid model based on the real data designed for evaluation of a future smart grid. The grid model is built measuring the distribution lines’ length and considering the cable dimensions and lengths, the grid age, the number of cabinets and customers and the load per customer. The aim of the model is to design, implement and test the proposed configuration and to investigate whether the lowvoltage distribution grid is prepared for the expected future increase of PV penetration, heat pumps and electric cars. The model is implemented in NEPLAN and DIgSILENT Power Factory and different scenarios are developed and analyzed. A time series simulation is conducted for a specific scenario with a comparison between different voltage and load profiles along the feeders. Index Terms—Low-Voltage Distribution Grid; DIgSILENT Power Factory; PV Penetration.
I.
NEPLAN;
INTRODUCTION
One of the main goals in the Danish energy policy is to increase the amount of renewable energy in the energy mix to 30% of the energy in 2025 [1]. Last year, the Danish parliament approved an even more ambitious target: to have renewable supply 35 % of the country’s total energy needs-not just electricity but also heating and transportation-by 2020, and an incredible 100 % by 2050 [2]. The increase in solar penetration will affect operation and design of distribution systems [3]. More than that, due to changes in the way electrical energy is produced and used, distribution network operators must adapt to changing usage patterns: the penetration of renewable energy will continue to grow and electricity is expected to increasingly substitute fossil fuel in areas such as transportation and building heating [1-3]. Increased distribution generation is becoming more important in the current power system. In the future it will rely more on distribution energy resources and on smart-grids [4][5]. In the future smart-grid distribution systems must be flexible and to be able to import/export the power from/to the
grid, to control the active and reactive power flows and to manage the storage of energy [4-9]. This paper presents the design and implementation of a representative low-voltage distribution grid model, based on real data for summer and detached houses, for evaluation of a future smart-grid. The aim of the model is to design, implement and test the proposed configuration and to investigate whether the low-voltage distribution grid is prepared for the expected future increase of PVs, heat pumps and electric cars. The model was implemented in NEPLAN and DIgSILENT Power Factory to study different scenarios. A steady-state and a dynamic analysis of the models have also been presented with a comparison between different voltage profiles. II.
LOW-VOLTAGE GRID MODEL
The aim of this section is to design and build a representative low-voltage grid for summer houses and detached houses. The grid model contains many components, such as: PV systems, EV systems, heat pumps and conventional loads and will be designed and tested to find out whether the proposed low-voltage distribution grid is prepared for the expected future increase of PV penetration, heat pumps and electric cars connected along the feeders. A. Grid Model Setup and Database Building Many representative areas have been looked through for lowvoltage feeders containing the two types of houses (summer houses and detached ones). The length of the low-voltage feeders is measured using a GIS (Geographical Information System) map. The number of customers, number of cable cabinets and the number of customers per cable cabinet for each feeder are collected and the maximum load per customer is calculated using the Velander correlation [16] which assumes that the load along the feeders is normally distributed. Then a grid model was constructed from the data using a statistical percentile. To build a database for the grid model the data from 334 lowvoltage feeders was mapped. The following parameters are taken into consideration: cable dimensions and lengths, the
grid’s age, category, substation no., no. of customers, no. of cabinets and the load per customer. The measured feeder length for the 85 % was 500 m for detached houses and 730 m for summer houses. We consider only two types of cables (with a cross-section of 150 mm2 AL and 95 mm2 AL) because 240 AL and 50 AL cables represent a very small part of the total cable length investigated. Based on measurements a regression analysis could establish a statistical correlation between the consumers’ annual energy consumption and the maximum load of the grid caused by their consumption. The feeder’s aggregate load was calculated using the Velander correlation:
Pmax = α ⋅ w + β ⋅ w
(1) In which Pmax represents the maximum load (measured in kW), w is the annual consumption (measured as MWh), α and β are the Velander constants. We used the following values for Velander constants: α=0.29 and β=2.09 for detached houses and α=0.32 and β=3 for summer houses. B. Grid Model Components The grid model contains conventional residential loads (houses), heat pumps (with 1 and 3 phases), electric cars (1 phase and 3 phases) and PV systems (1, 2 and 3 phases). In the case of one phase appliances it is assumed that all the devices are connected to the same phase. Similarly 2 phase PV systems are assumed to be connected to the same 2 phases. For PV systems 3 parameters were considered: rated voltage, tap position of the 10/0.4 kV transformer and the number of phases (1, 2 or 3). The position of the tap-changer normally depends on the primary voltage level on the 10 kV side. If the distribution transformer is located at the end of a long 10 kV feeder, the tap-changer is used to normalize the voltage level on the secondary side. In this study the tap changer position is used to allow a larger voltage drop/rise in some of the scenarios. The connection of heat pumps and electric cars to the lowvoltage grid, along the different feeders, will cause a voltage drop. C. Scenarios Developed to Design and Test Grid Model The grid model was designed, based on SEAS-NVE’s supply area and was analyzed and tested for various load scenarios through connection of appliances such as heat pumps, EV’s and PV systems. 23 scenarios were developed using 2 different voltage limits, +6%/-10% and +10%/-10%. The first limit is based on DEFU recommendations no. 16/2001 [12]. These limits were used in Denmark during a grace period after EU legislation changed nominal voltage from 220 V to 230 V in order to allow for older 220 V household appliances to be phased out. The +10%/-10% limit is based on the latest version of DEFU recommendation 16, which closely mirrors European norm EN 50160 [13]. It is estimated that the worst case scenarios is a summer day with low load (20 % of max.) with the PV’s connected to the grid.
In the first four scenarios it is investigated how many solar cells can be connected to the grid along the feeder without to exceed the voltage limits defined in [12] and [13], when the PV systems have three-phases and 6 kW each with the residential load at 20 % of the maximum load. In the next five scenarios it is investigated what happens in the same conditions when one phase and 4 kW and two phases and 6 kW PV systems are connected along the feeder. In the next seven scenarios we investigated how many electric cars (3 phases EV of 11 kW and one phase EV of 3.7 kW each) can be connected to the grid, when the residential load is at the maximum load, without exceeding the lower voltage limit. We have investigated in the last five scenarios the connection of heat pumps (3 phases HP of 2.8 kW each and one phase HP of 1.5 kW each), to the grid, evenly distributed along the feeder, with the residential loads at maximum load. In one of these last five scenarios we considered a mixed load case (the worst case scenario) when heat pumps (one phase and three phases) and electric cars are connected simultaneously on the same distribution grid. III.
GRID MODEL IMPLEMENTATION
Based on investigations, calculations and assumptions presented in the last section a grid model for implementation of all 23 scenarios has been designed. The model contains an external grid, a distribution substation with a medium voltage transformer (10/0.4 kV) and with ten cabinets (switch boards) at which PV panels, residential loads, EVs and heat pumps are connected. The grid model was implemented in two different tools, NEPLAN and DIgSILENT Power Factory, to study load flow, steady-state voltage stability and dynamic and transient behavior of the power system. These tools have been selected as they have the ability to simulate load flow and RMS fluctuations in the same software environment [14, 15]. In Fig. 1 is presented the implementation of the grid model for a low-voltage distribution system in NEPLAN (a) and DIgSILENT Power Factory (b).
a)
usbar/BB_HV
Ul=10,00000 kV u=1,00000 p.u. phiu=0,00000 deg
P=-47,3023 kW Q=5,5354 kvar cosphi=-0,9932
Trafo 10/0.4
Line
Busbar(1)/BB_LV Ul=0,40087 kV u=1,00217 p.u. phiu=-149,42033 deg
P=-130.589 Q=-16.318 cosphi=-0.992
P=114.96 Q=12.322 cosphi=-0.994
P=-47,.. Q=5,06.. I=68,7..
Line(1)
P=50,4.. Q=-4,1.. I=68,7..
P=-58,.. Q=3,63.. I=79,3..
SingleBusbar(2)/BB1 Ul=0,4..
Line(2)
P=62,1.. Q=-2,3.. I=79,3..
SingleBusbar(3)/BB2 Ul=0,4..
u=1,06.. phiu=-..
P=1,82.. Q=0,45.. I=2,55..
P=-6,0.. Q=-0,0.. I=8,16..
Low-Volta..
PV1
Line(3)
P=-63,.. Q=1,98.. I=81,1..
P=63,8.. Q=-1,9.. I=81,1..
P=-65,.. Q=1,54.. I=83,0..
SingleBusbar(4)/BB3 Ul=0,4..
u=1,13.. phiu=-..
P=1,45.. Q=0,36.. I=1,91..
P=-6,0.. Q=-0,0.. I=7,67..
Low-Volta..
PV2
Skab4
u=1,13.. phiu=-..
P=6,01.. Q=0,00.. I=7,67..
P=1,45.. Q=0,36.. I=1,91..
u=1,13.. phiu=-.. P=1,45.. Q=0,36.. I=1,90..
Low-Volta..
PV1(..
Low-Volta..
P=0,00.. Q=0,00.. I=0,00..
Ul=0.386 u=0.964 phiu=-151.941
Line(6)
P=-68,.. Q=0,66.. I=86,7..
Ul=0,4.. u=1,14.. phiu=-.. P=1,45.. Q=0,36.. I=1,89..
P=69,0.. Q=-0,5.. I=86,7..
SingleBusbar(7)/BB6 P=0,00.. Q=0,00.. I=0,00..
Ul=0,4.. u=1,14.. phiu=-.. P=2,73.. Q=0,68.. I=3,53..
Low-Volta.. Static Ge..
Low-Volta..
P=-71,.. Q=-0,1.. I=90,1..
Line(7) P=72,1.. Q=0,19.. I=90,1..
SingleBusbar(8)/BB7 P=0,00.. Q=0,00.. I=0,00..
Static Ge..
Ul=0,46220 kV u=1,15549 p.u. phiu=-146,24109 deg
P=1,37.. Q=0,34.. I=1,76..
Low-Volta..
P=0,00.. Q=0,00.. I=0,00..
P=-49,.. Q=-0,5.. I=61,7..
Line(8) P=49,6.. Q=0,54.. I=61,7..
P=-26,.. Q=-0,8.. I=33,5..
SingleBusbar(9)/BB8 Ul=0,4..
u=1,15.. phiu=-.. P=1,45.. Q=0,36.. I=1,86..
P=-23,.. Q=-0,0.. I=29,9..
Static Ge.. PV7
Low-Volta.. Static Ge..
P=-23,.. Q=-0,0.. I=29,8..
PV8
D10 L56
P=-4,3.. Q=-0,4.. I=5,46..
u=1,16.. phiu=-.. P=1,45.. Q=-0,3.. I=1,85..
Low-Volta..
P=-23,.. Q=-0,0.. I=29,7..
PV9
P=0,00.. Q=0,00.. I=0,00..
P=106.732 Q=10.187 cosphi=-0.995
P=4,3793 kW Q=0,4664 kvar I=5,4681 A
Static Ge..
Ul=0,46500 kV u=1,16251 p.u. phiu=-146,17348 deg P=1,71.. Q=-0,4.. I=2,19..
P=6,02.. Q=0,01.. I=7,48..
Branches
P=68.498 Q=6.091 cosphi=-0.996
P=-67.940 Q=-5.969 cosphi=-0.996
L67
Low-Volta.. Static Ge.. PV10
Project: Graphic: Grid10 Date: 1/21/2013 PowerFactory 14.1.6
P=-105.883 Q=-9.909 cosphi=-0.996
P=31.155 Q=2.283 cosphi=-0.997
P=-31.037 Q=-2.257 cosphi=-0.997
L89
P=0,00.. Q=0,00.. I=0,00..
Load Flow Balanced Line-Line Voltage, Magnitude [kV] Active Power [kW] Voltage, Magnitude [p.u.] Reactive Power [kvar] Voltage, Angle [deg] Current, Magnitude [A]
L910
L78
Line(9)
P=27,0.. Q=0,87.. I=33,5..
P=-113.996 Q=-12.005 cosphi=-0.995
Skab5
Nodes
Ul=0.394 u=0.984 phiu=-151.727
L45
Static Ge..
SingleBusbar(10)/BB9 Ul=0,4.. SingleBusbar(11)/BB10 P=0,00.. Q=0,00.. I=0,00..
D04
Ul=0.390 u=0.974 phiu=-151.834
P=-66,.. Q=1,10.. I=84,8..
Line(4) Line(5) 67,2.. 1,0.. 4,8..
P=-151.098 Q=-21.773 cosphi=-0.990
Skab1 Skab2
Ul=0.382 u=0.955 phiu=-152.048
Ul=0.378 u=0.946 phiu=-152.154
Ext.
LB1 P=-149.535 Q=-21.262 cosphi=-0.990
P=131.824 Q=16.721 cosphi=-0.992
Skab3 P=65,5.. Q=-1,4.. I=83,0..
SingleBusbar(5)/BB4 Ul=0,4..
P=-151.098 Q=-21.773 cosphi=-0.990
P=140.459 Q=18.989 cosphi=-0.991
L23
P=-47,.. Q=5,53.. I=2,74..
P=47,4.. Q=-5,0.. I=68,7..
T10/0.4
L12 P=-139.082 Q=-18.539 cosphi=-0.991
P=123.326 Q=14.499 cosphi=-0.993
P=-122.227 Q=-14.140 cosphi=-0.993
P=-122.227 Q=-14.140 cosphi=-0.993
Ul=10.000 u=1.000 phiu=-0.000
DIgSILENT
L34
External Grid
P=87.630 Q=8.096 cosphi=-0.996
Skab6 Ul=0.375 u=0.938 phiu=-152.260
P=-86.736 Q=-7.900 cosphi=-0.996
Skab8
Skab7 Ul=0.372 u=0.930 phiu=-152.366
P=49.693 Q=4.158 cosphi=-0.997
Ul=0.368 u=0.921 phiu=-152.440
P=-49.394 Q=-4.093 cosphi=-0.997
Skab9 Ul=0.365 u=0.913 phiu=-152.500
Skab10 Ul=0.363 u=0.908 phiu=-152.547
Ul=0.362 u=0.904 phiu=-152.578
Fig. 2. Power flow calculation for scenario 10 using DIgSILENT Power Factory.
Annex:
b) Fig. 1. Comparison between a) NEPLAN and b) DIgSILENT Power Factory implementation for the designed distribution grid.
All components of the single line diagram, presented in Fig. 1, are built with standard blocks from the library. The blue square above the cable cabinet (Fig. 1a) shows the voltage on each phase, while the green squares show the power and current for each phase. Also, the arrows show that the components are active. In DIgSILENT Power Factory implementation (Fig. 1b) the squares above the bus-bars contain the parameters of the cables and below the bus-bars show the voltages of each phase. Also, each component of the cabinet, such as PV systems, heat pumps and EVs has its own square block able to show internal parameters (current, power, power factor). Fig. 1 shows a power flow calculation for scenario no. 4, where it was investigated how many PV systems (3 phase 6 kW each and Un ±10 %) could be connected to the distribution grid, with a residential load at 20 % of maximum load. In this case the voltage increase may not exceed 440 V in the most remote cable cabinet. We have connected 4 PV systems at each cabinets of the feeder (40 PV systems in total), and a maximum voltage (438 V) was observed. A power flow calculation for the scenario 10 is presented in Fig. 2. In this case it was investigated how many electric cars can be connected to the grid before voltage decreases below the minimum limit (Un -10 %). In this scenario, six electric cars (11 kW & 3 phase each) were distributed over the five most remote cable cabinets, two electric cars in the last remote cable cabinet and one electric car in each of the next four cable cabinets. A minimum voltage of 359 V was observed. The residential load in this case is regulated to the maximum with the 10 kV stable voltage on the primary side of the substation with a conversion factor of 25.
IV.
SIMULATION RESULTS
A. Steady-State Analysis In this section a time series simulation is conducted using DIgSILENT Power Factory. To investigate the characteristics of the new components: PVs, EVs and HPs, we used the same grid model proposed in the last section. The daily time series in Fig. 3 (i.e. external grid voltage profile, PV production, EV charging consumption, and the detached house consumption) are the inputs of the simulation model. The data with voltage profile (the dash curve in Fig. 4) and PV production curve (06:00 – 20:00) is collected on 15-05-2012 from a laboratory with real components and renewable energy production (SYSLAB), at the Technical University of Denmark-Risø campus. The capacities of PVs are adopted to 6 kW. EV charging profile, from (18–22), is calculated based on the driving pattern [17] with the rated charging power of 11 kW. The load profiles are calculated by the load pattern, from the aggregated historical data in Denmark, multiplying the peak power using the Velander equation (1). By subtracting the production or adding the consumption new household load profile can be obtained. The time sweep load flow is executed by using DIgSILENT programming language DPL. Fig. 3 shows that at 19:00, the modified load has a peak due to the EV charging (Scenario 10). However, considering the deviation of human behaviors (smoothing effect), the over all peak may be less than the sum of them. At 15:00, the reversed load reach the peak where there is the largest mismatch between load and PV production (Scenario 2). A voltage rise can also be observed. Fig. 4 shows the simulation results for one day with a comparison between voltage profiles in the last node versus voltage profile of the external grid. Using this approach (i.e., generate the profiles outside the power system software and use its traditional function) the complexity of the models can be avoided, but introducing a lot of manual work on modifying the inputs.
The dynamic load model is implemented using a voltage dependency of active power and it is described by ([14], [18]): ( p2 ) ( p3 ) ⎡ ⎛ V ⎞( p1) ⎛V ⎞ ⎤ ⎛V ⎞ P = P0⎢kp1⎜⎜ ⎟⎟ + kp2⎜⎜ ⎟⎟ +(1−kp1 −kp2)⎜⎜ ⎟⎟ ⎥ (2) ⎢⎣ ⎝V0 ⎠ ⎝V0 ⎠ ⎥⎦ ⎝V0 ⎠
Fig. 3. Time series implementation of the load profiles for residential load (based on Velander correlation (1)), EV load and PV load (distinguished from the conventional load with a peak power multiplication factor) based on the load pattern from the aggregated historical data in Denmark.
Where P0 and V0 are the initial values of power and voltage, p1 to p3 are coefficients that define the proportion of each component and kp are coefficients that reflect the dependency of load (kp1=100%, kp2=0). Typically, in dynamic simulation, a simplification is done and all loads are consider to be constant admittance type (p1=2, p2=p3=0). In Fig. 5 are presented the time-series simulation results, with PV systems and loads connected together to the same bus-bar, for 6 days in November 2012. The input data for the simulation model (irradiation and temperature) was measured from the weather station placed on Risø campus with a sampling time of 1 second. The load profile was defined by (2) with the voltage dependency of the active power.
Time series simulation provides an overview of the potential problems (thermal loading and voltage rise in our case). Based on this the concurrency of a group of units can be investigated towards less conservative design.
In Fig. 5 b) are shown the simulation results when 3 PV inverters are connected together with the loads at the last 3 cabinets of the grid model presented in Fig. 1. In this particular case, EVs and HPs are assumed to be an unknown proportion of the total load and the PV systems have the same parameters but the panels have a different orientation and tilt angles. DIgSILENT
Fig. 4. a) Comparison between voltage profile of the external grid and the voltage profiles of the PV’s, EV’s, base loads and all loads connected at the last cabinet to the end of the feeders.
Fig. 5 a) shows a comparison between input data (Tcell between -4 and +6 0C and Gcell around 300 W/m2 during the day) and the output parameters of the model [output power of the panels-Pdc1 +Pdc2 (Pdc_total=2.7 kW) and output power of the inverter-Pac =2.6 kW).
9,00
B. Time-series Simulations using Dynamic Models The dynamic model of the PV System (PV panels and PV inverter) has been built with standard block components from the Power Factory library and also using the dynamic simulation language (DSL). It is based on a single diode equivalent electrical circuit for the PV model, described by an exponential equation [9-11]. The model uses the cell irradiation Gcell and cell temperature Tcell as inputs, measured from a weather station and implemented as look-up tables, as it is shown in the first two graphs of Fig. 5 a).
6,00
3,00
0,00
-3,00
-6,00 -2.78E-5
2.88E+1
5.76E+1
8.64E+1
1.15E+2
[h]
1.44E+2
5.76E+1
8.64E+1
1.15E+2
[h]
1.44E+2
5.76E+1
8.64E+1
1.15E+2
[h]
1.44E+2
PV_3b_MeasFile: Tcell (C)
0,40
0,30
0,20
0,10
0,00
-0,10 -2.78E-5
2.88E+1 PV_3b_MeasFile: Gcell (p.u.)
The PV panels are mounted in three strings: two of them having 18 panels of 165 W each, and the 3rd one having 12 panels of 100 W [9]. The strings of panels are connected to the low-voltage distribution network through a three-phase PV inverter. More details can be found in [9]-[10]. The simulation model was validated using experiments carried out using RISOE experimental facility-SYSLAB ([9][11]).
2,9077 2,2943 1,6809 1,0675
0,4541 -0,1593 -2.78E-5
2.88E+1 PV_Inverter_block: Pac (kW) PV_Inverter_block: Pdc (kW)
Subplot/Diagramm Date: 8/15/2013 Annex: /2
a)
DIgSILENT
2,419 1,909 1,398 0,888 0,378 -0,133 -2.8E-5
2.9E+1 PV_Inverter_block: Pdc1(kW) PV_Inverter_block: Pdc2(kW) PV_Inverter_block(1): Pdc3(kW) PV_Inverter_block(1): Pdc4(kW) PV_Inverter_block(2): Pdc5(kW) PV_Inverter_block(2): Pdc6(kW)
5.8E+1
8.6E+1
1.2E+2
[h]
components placed along the feeders in a future smart-grid distribution network.
1.4E+2
ACKNOWLEDGMENT This work was supported in part by the E.U. Project-Smart Plan, No. 55807/2011-2013.
2,820 2,225 1,630 1,035 0,440 -0,155 -2.8E-5
2.9E+1 PV_Inverter_block(1): Pac1 (kW) PV_Inverter_block: Pac3 (kW) PV_Inverter_block(2): Pac2 (kW)
5.8E+1
8.6E+1
1.2E+2
[h]
1.4E+2
REFERENCES [1]
20,00 15,00
[2]
10,00 5,00 0,00 -5,00 -2.8E-5
[3] 2.9E+1 Load_file2: Pload (kW)
5.8E+1
8.6E+1
1.2E+2
[h]
1.4E+2
Date: 8/16/2013 Subplot/Diagramm(1) Annex: /3
[4]
b) Fig. 5. a) Time series simulation results of the PV system for 6 days with real data implemented as input and with DC and AC powers as outputs; b) Comparison between IN and OUT of the 3 PV inverters, connected to the last 3 cabinets of the grid model together with a dynamic load defined by (2).
[5] [6] [7]
CONCLUSION In this paper we have proposed a representative low-voltage grid for summer and detached houses based on real data measurements. The feeder lengths were measured using GIS maps. The number of customers, cable cabinets and of customers per cable cabinet were collected and the maximum load per customer has been calculated using the Velander correlation. The grid model contains many components, such as: PV systems, EVs, heat pumps and residential loads and was designed and tested, based on 23 scenarios, to find out whether the proposed distribution grid model is prepared for the expected future increase of PV penetration with heat pumps and electric cars connected along the feeders. The developed scenarios clearly have shown that there is room for larger loads if the output voltage from the substation can be optimally set and/or varies according to the type and the size of the load. Also, in a weak distribution line with a high output voltage only 3 phase PV systems should be installed, as 1 phase PVs under the same conditions are more likely to exceed the voltage limits. The low-voltage distribution grid model has been developed and implemented in NEPLAN and DIgSILENT Power Factory to study load flow, steady-state voltage stability and dynamic behavior of the components. The comparison between both simulation tools has shown a good alignment and the possibility to use them for further developments, regarding the integration of smart-grid technologies. It means that this work could be used for development and improvements of the models for different
[8] [9]
[10]
[11]
[12] [13] [14] [15] [16] [17] [18]
"Smart Grid i Danmark", joint report by energinet.dk and Dansk Energi, September 2010. J. Kumagai, ”The Smartest, Greenest Grid-What a little Danish island is showing the world about the future of energy”, IEEE Spectrum, May 2013, pp. 38-43. Y. Riffonneau, S. Bacha, S. Barruel and S. Ploix, “Optimal power management for grid connected PV systems with batteries”, IEEE Transaction on Sustainable Energy, vol. 2, no. 3, pp. 309-320, July 2011. H. Jiayi, J. Chuanwen, and X. Rong, “A review on distributed energy resources and MicroGrid”, ELSEVIER Renewable & Sustainable Energy Reviews, vol. 12, pp. 2472-2483, 2008. “Smart Grid: Reinventing the electric power system”, IEEE Power & Energy Magazine, March 2012. P.C. Loh, L. Zhang and F. Gao, “Compact integrated energy systems for distributed generations”, IEEE Transactions on Industry Electronics, Vol. 5, May 2012. M. Jansen, R. Louie, M. E. Amoli and F. Sami, “Model and simulation of a 75 kW PV solar array”, in Proc. of 2010 IEEE PES Transmission and Distribution Conference and Exposition, pp. 1-5. H. Liu, L. Jin, D. Le and A. A. Chowdhury, “Impact of high penetration of solar photovoltaic generation on power system small signal stability”, in Proc. of 2010 POWERCON, pp. 1-7. C. Koch-Ciobotaru., L. Mihet-Popa, F. Isleifsson and H. Bindner, „Simulation model developed for a small-scale PV-System in a distribution network”, in Proc. of the 7th IEEE SACI Conf., 2012, Timisoara-Romania, May 24-26, pp. 257-261. L. Mihet-Popa, C. Koch-Ciobotaru, F. Isleifsson and H. Bindner, „Development of tools for DER Components in a distribution network”, in Proc. of ICEM Conf., September 2-5, 2012, MarseilleFrance, pp. 1022-1031. Y. Zong, L. Mihet-Popa, D. Kullman, A. Thavlov, O. Gehrke and H. Bindner, „Model Predictive Controller for Active Demand Side Management with PV Self-Consumption in an Intelligent Building”, IEEE PES Innovative Smart Grid Technologies Europe, BerlinGermany, October 14-17. DEFU recommendations no. 16/2010. European Standard-EN 50160. DIgSILENT PowerFactory, Digsilent gmbh, November 2012. NEPLAN, Tutorial manual, 2011. M. Velander, “Methods of operational analysis applied to distribution of electric power”, Journal of Teknisk Tidskrift, Vol. 82, 1952, pp. 293299. [Online].Available: http://www.scb.se/Pages/Product____10616.aspx?Produktkod=TK1101 &displaypublications=true; P. Kundur, Power system stability and control, McGraw-Hill, New York, 1994.
