Power Flow Efficiency of a DC Distribution Grid
A test case with a High Penetration of Renewable Energy Resources
in Different Operational Modes
TuanDat Mai, Jeroen Tant, Johan Driesen
ESAT/Electa, KU Leuven, Heverlee, BELGIUM.
tuandat.mai, jef.beerten, jeroen.tant, johan.driesen@esat.kuleuven.be
Abstract—A models of a DC local distribution grid with
penetration of distributed energy resources (DER) is studied
that are expected to increase significantly in the future.
Because the existence AC distribution networks are not
optimally designed for bidirectional energy flows caused by
the distributed energy resources (RES, storage energy
systems, PHEVs), a replacement of delivering DC power in
conductors for existing AC network is assumed. Therefore, a
DC distribution grid with participation of renewable energy
resources will be presented in different operational modes to
evaluate the efficiency of power control solution using DC
signal in terms of power flow and total losses. A comparison
with AC power flow is the same grid is also introduced to
clarify the viable alternatives of a low-voltage DC grid for
bidirectional power flow in future.
Keywords- DC grid; efficiency; distribution; energy flow;
I.
INTRODUCTION
Electricity was discovered from 19th century and quickly
developed to be one of major energies, which is widely
used. Development of electric power has been started both
in AC and DC power in 1880s. Due to advantages of AC
power transmission compared to DC power, AC electric
networks are deployed in all of the architecture of power
network in 20th century.
Since the increasing demand of energy requires having
more power generation while conventional generations
cause to badly impact on environment, more ‘green’ and
‘distribution’ energy sources such as PV and wind turbines
have been installed in utility grid. Nevertheless, AC local
distribution networks are not designed to adopt DC power
generation and draw bidirectional power flow. Additionally,
development of power electronics is recently getting
achievements that AC-DC energy conversion is more
flexible and more efficient. The success of recent HVDC [1]
projects proved the efficiency in terms of long-distance DC
power transmission, while DC loads in datacenters are good
example of DC power‘s efficiency in local areas [2], [3].
A DC local distribution grid is a suitable solution that
PV arrays, PHEVs and energy storage systems without ACDC round-trip can work together in a low voltage utility
grid. The connection of DER to a DC distribution bus does
require neither frequency synchronization nor reactive
power compensation. It hence gives one of advantages of
DC bus compared to AC systems. However, the architecture
of HVDC transmission and MV/LV DC distribution has
slightly differences. While a HVDC exists in ‘mesh’
topology and requires a tight AC/DC interconnection, a DC
distribution grid is mostly constituted by radical topology.
On the other hands, strengthening grid reliability is the key
requirement for HVDC while coupling on-site DC
generation, DC loads, energy storage systems, PHEVs and
efficiency in a local network is the highest considerations
for DC utility. Offering more flexibility with a lower price
for DERs and DC-power electrical devices is also a
consideration in DC distribution grid.
Previous researches concluded that a reduction of
approximation 5-20% in electricity charges depending on
bus voltage could be expected when data centers with an AC
distribution bus migrates to a DC distribution system [2],
[4].
In order to achieve a lower price of energy use,
efficiency of power a one of benchmarks. Therefore, a DC
distribution grid with participation of PV generation will be
presented in different operating cases to evaluate the
efficiency of power control solution using DC signal in
terms of power flow and total losses. A comparison with AC
power flow is the same grid is also introduced to clarify the
viable alternatives of a low-voltage DC grid for bidirectional
power flow in future.
II.
ASSUMPTION AND MODEL
A. Model and role of converters:
Power converters including AC/DC, DC/AC and DC/DC
types are required in order to control power absorbed by
loads in an optimal way in a DC power architecture. Due to
different characteristics of them, several models of
converters and loads are assumed.
1) Active Front-End Converter: provides a bidirection
energy link between an AC bulky power source and a DC
feeder. It converts AC power to DC power when energy
production grenerated by local DERs is in lack.
Contraditonally, it tranfers energy back to the AC utility if
there is a redundacy of energy generated by DERs.
2) Three-phase rectifier: is used to connect a AC power
gerator to a DC feeder such as wind generators. The DC
voltage at the DC feeder is regulated depending on
operational modes of DC power circuitry. Due to small
inertia of low-power wind generators, it is not necessary to
account a bidirection-flow converter which converts DC
power back to mechanical power for inertia reserves.
3) Uni-direction DC converter:converts DC energy
captured by PV arrays to a suitable DC voltage level and
supplies to DC feeders. Non-rechargable DC power sources
such as PV could be connected to DC bus by using this type
of converter to benefit from advanced efficiencies and
power control.
4) Three-phase inverter:converts the DC power from a
feeder to three phase AC power that supplies power to
rotating mechanical loads such as AC motors. In fact it can
be considered as a variable-frequency drive without a
rectifier stage thanks to the exsitence of DC power in the
feeders.
5) Single-phase inverter: functionates as one output
phase of a three phase inverter which supplies for separate
single-phase AC loads such as house-hold appliances.
6) Bidirectional DC converter: steps up or steps down
DC voltage and links power to a DC power terminal. The
power terminal can be considered as a loads for charging a
battery energy storage/PHEVs ; as well as a power supply
for discharge a battery energy storage/PHEVs.
B. Proposed Modes of Operation in a DC grid
The DC voltage at internal distribution buses is passively
controlled depending on the two factors – difference
between the total renewable energy production and total
demand; and the situation of bulky power interconnection.
The operation of the proposed DC local distribution grid is
hence categorized in four modes which are slightly modified
compared to [5] and represented inFig. 1.
1) Mode 1- islanding mode with battery discharging
The DC grid operates in islanding mode and the DC bus
voltage is regulated by battery discharging, which means the
generated PV power is less than local load demand. Slack
bus voltage is regulated at 95% of DC bus voltage in grid
mode.
2) Mode 2- grid-connected mode
The DC grid operates with a power interconnection to a
higher power distribution system via an Active Front-End
converter. The output voltage of the AFE is kept as a slack
bus while generated DERs’ power is less than the demand.
Converters for PV arrays and rotating generators work in
MPPT and bi-directional DC/DC converter for energy
storage does not work. PHEVS can be simultaneously
charged in this mode.
Fig. 1. Operational modes of the proposed DC grid
3) Mode 3- grid-connected mode with battery charging
The DC grid is connected to a higher power distribution
system via an AFE converter similarly to Mode 2. The
output of AFE is still kept as slack bus while generated PV
power is greater than the local load demand; hence they are
controlled in droop control. Battery storage systems are
charged if they are not full. The charging current is also
controlled to keep power flow nearly equal to zero at slack
bus, which minimize loss of AC/DC conversion.
4) Mode 4- islanding mode with DC voltage regulation
The DC grid operates in islanding mode. DC bus voltage
is regulated by converters for DERs. As MPPT power of
distribution generators is greater than the actual demand of
loads, the redundancy power is for charging battery energy
storage. If they are fully-charged, they are switched to burst
mode and the node generated the highest power works as
slack-bus at the highest acceptable voltage in this mode.
Slack bus voltage is kept at 105% compared grid-mode at
the bus where PV produces the highest power.
The coefficient difference between these modes is
defined at 5%, which doesn’t cause the malfunction during
switching from one mode to another mode.
C. Assumption of losses in DC networks:
Energy losses are main factors which is influence to the
efficiency of a system. Losses in the proposed network can
be classified into 3 groups including converter losses, losses
of wiring and losses of non-MPPT operation of DERs.
1) Converter losses: are refered to losses of power
electronic switches. It can be categorized to static losses,
switching losses and driving losses [6]. For the ease of
calculation in next steps, it is assumed that all converters
operate at 10kHz switching frequency at 25oC.
a) Static losses:
On-state losses is actually conduction losses caused by
power switches and reverse diodes which is dependent on
the load current, the junction temperature and the duty
cycles.
Blocking losses accounts for a small share of the total
power dissipation and it can be neglected in case of low
voltage and low environmental temperature.
b) Switching losses:
Switching losses consist of turn-on losses and turn-off
losses. They are dependent on load current, electric load
type, DC link voltage, freewheeling diode, junction
temperature and switching frequency. All of these
Fig. 2. Efficiency of different power converters
parameters are considered as constants except the DC link
voltage and the load current.
c) Driving losses: can be neglected since conveters
operates around 400V at DC bus and switching frequency
of 10kHz.
By taking account the loss specification of converters,
efficiency of different converters is introduced in Fig. 2.
2
line line
2) Branch losses: can be calculated as r ·I
produced or consumed in a DC power system, the loads
modeled in DC is of the same real power magnitude[7].

ZwireAC   rXLPE  j·xXLPE  length 


Z wireDC   rXLPE  j·xXLPE  length 

where
rXL PE , x XL PE : resistance and reactance values of
where
rline is resistance value of the branch and I line is current
the cable specified in datasheet
length : length of the feeder
carried by the branch.
3) Losses due to the non-MPPT operation of RESs: is
defined as the difference between power produced at MPPT
condition and the actual generation power. These losses
occur only in Mode 4 when all of battery energy storage
systems are fully-charged and few converters switch to
burst mode.
Z wire : total impedance
E. Assumption of Power Line Communication (PLC):
It is assumed that the demand at loads and the production at
each generator are known. The data can be transferred and
accessible by converters in the DC infrastructure via PLC.
III.
POWER FLOW AND EFFICIENCY CALCULATION
Due to these factors, total losses at each node vary
depending on the percentage of loads at certain conditions
and mode of operations of the DC grid.
Start
Demand power
RES production
D. Assumption of differences in AC network model and
DC network model:
In an AC distribution system, three phase network is
commonly used. One of three feeders is sequentially
contributed and supply AC power to households. However,
demand at each house is not balance that leads to non-zero
current in the neutral conductor [7]. Therefore, losses and
voltage drops are greater than a three-phase balance load.
As a consequence, the lowest loss case in AC system,
which is three phase-balance load, is assumed and can be
represented by its equivalent single line diagram. The three
phase active power is defined as the following.
P3  3·Vl l ·Ilinecos   

where

Mode of operation
Loss calculation at
converters
Power flow calculation
NR method
Total loss at nodes and lines
P3 : active power
Efficiency calculation
Vl  l , I line : line to line voltage and line current
cos   : power factor of load
Stop
In DC networks, the power at generators and loads is
calculated by
PDC  V DC ·I line · pol 

where

P D C : power
V D C , I l in e : voltage from positive/negative
polarity to neutral and line current
pol : number of polarity (1 for mono-polar and
2 for bipolar)
In both AC and DC networks, cable EXVB 1kV is used
for distributing power. While AC systems uses 3 conductors
as three AC phase power and one conductor as neutral
phase, DC system twins 2 conductors as the positive
conductors and double 2 other conductors as the
negative/neutral conductor. Since there is no reactive power
Fig. 3 Flow chart of power flow
and efficiency calculation
TABLE 1. PARAMETERS IN DISTRIBUTION GRID
Power demand
Length of the main feeder
Cable type
Transformer
Power electronic switch
Energy storage
108.08 kW
725 m
EXVB 1kV_ 4x150 (main feeder)
EXVB 1kV_ 4x95 (main feeder)
EXVB 1kV_ 4x16 (to loads/PVs)
ABB EcoDry Basic
250kVA, 24/0.4KV
IGBTs 1200V, forced air cooling
AFE: 400V/250 KVA
DC/DC converter: 400V/10KVA
DC/AC converter: 400V/10KVA
DC/DC bi-directional:
400V/100kVA
Pb-acid battery, 34 packs 12V
Fig. 4. Schematic diagram of the DC semi-urban network [15]
for the
scenario of high-density penetration of PV
installation [13].
Due to having a suitable DC bus available in the
Four cases corresponding to 4 proposed operation modes
distribution architecture saves losses and chance of failure
are evaluated in this paper.
[2], the standard DC bus voltage is defined at 400 V to cope
with household electric appliances, conventional converters
- Test case 1: The DC local grid is operating under islanding
and distribution cable infrastructure.
mode while production from PV is less than the total
demand Ppv < Pdemand. Voltage drop in Test case 1 is
Defining the power difference of demand and production
shown in Fig. 6.
is firstly processed in the main routine. If mismatch demand
of the DC local grid is fulfilled by storage systems either by
- Test case 2: The AC/DC converter at bus 1 is running
AC bulky power, DC grid operates at Mode 1 or 2
to supply extra power from AC bulky system to the DC grid
depending on the condition of interconnection. If the total
while production from PV is less than the total demand Ppv
< Pdemand. As a consequence, converter at bus 1 is chosen as
produced power by PVs is greater than the local loads’
slack bus, there by controlling the voltage on the network.
demand, DC grid operates at Mode 3 or 4, while the extra
Other converters control power.
power is highly contributed for battery charging.
Afterwards losses corresponding to each converter are
calculated to define the certain amount of injected power of
generators and real consumed power of local loads. Outage
converters are disconnected from terminals. Data for the
next step of calculation are power at distribution nodes,
which is called Point of Loads (POL). If DC grids operates
under Mode 1 or 4, the status of grid coupling converterAFE converter ‘s outage are considered as stations without
AC grid connection.
- Test case 3: DC local distribution grid is still in gridconnected operation. Due to production from PV is greater
The power flow calculation is mainly involved by
Matpower, MatACDC [8], [9], [10], [11] and some
additional routines to obtain real power solutions of the DC
networks. The file format of test case is compatible to
MatACDC, a free MATLAB based open source program for
AC/DC power flow analysis[12]. Before involving NewtonRaphson power flow calculation, input data is converted into
pu. The solutions of power and voltage drop are used to
calculate losses of the distribution grid and the efficiency as
well. The main routine stops after converting pu result to SI
units.
IV. TEST CASES AND RESULTS
A typical Belgian household distribution network is
introduced in Fig. 4 and Table 1. It consists of 62
loads/generators except node 1 connected to the AC bulky
power system. Energy storage systems are placed at the
intersection or at one end of the feeder. At each generator or
loads, there is a power converter to control power flow and
DC voltage. 48 PV sources are integrated with household’s
loads and provides from 1.81 to 4.40 kW of electric power.
The PV peak power of 10-kW per three houses is assumed
Fig. 5. Voltage drop in AC architecture and DC architecture
Fig. 6. Branch losses at different cases
than the total demand Ppv> Pdemand in this case, converter at
bus 1 still controls DC voltage as a slack bus while others
converters control power production/injection. It is similar
to case 2. Redundancy power produced by PVs is partly
transferred back to the bulky system through the DC/AC
converter at bus 1.
- Test case 4: There is not energy link between the DC
local grid and the AC bulky system. The DC local
distribution is under islanding operation while there a power
redundancy of PVs production as compared to total demand.
Therefore, the generator producing the highest power
operates as a slack-bus, while other converters operate as
controlling power. At bus 61, the converter operates in nonMPPT condition and reduces their efficiency since batteries
are fully-charged.
When the DC grid is connected to the AC bulky system
in Case 2 and Case 3, voltage at bus 2 is kept at 400 V.
Although the lowest voltage is 394.21V in Mode 2, it drops
less than in AC case in which voltage is 378.81V. Thanks to
production support from PVs, voltage at loads is around
400V in Case 3. There is also a very small voltage drop in
Mode 1 and Mode 4 shown in Fig. 5 and Fig. 8.
In terms of losses, converter losses mostly contribute in DC
grids and they are the highest in Case 3 compared to other
cases, which can be seen in Fig. 7. Since most of loads
operate as nearly-zero power nodes, power transferred
between these nodes is minimized. Branch losses are shown
in
Fig. 6 branch losses in case 2 are the higher than other 3
DC cases because most of power draws from slack bus to
distribution load. However, it is small compared to AC case
thanks to the wiring structure and DC power characteristic.
Fig. 8. Differences of voltage drop compared to base voltage
The total required power in Case 2 is approximately the
least values compared to other 3 cases. However, Case 2 has
less efficiency gain than Case 3 and 4. Efficiency in Case 2
and Case 4 is subsequently equal to 82.09%, 84.08% and
84.11%. It is consequence that around 19kW of PV
production is either sent back to grid or charged battery in
energy storage systems.
For the comparative study shown in Fig. 9 the efficiency
rating of the DC grids is not completely more superior to
AC grids; even there are several inverters and constantpower electronic loads such as variable frequency drives and
adjustable lighting systems in AC systems. AC and DC
distribution systems can have the same merit when the loads
are equal in ratio, the distribution grid is connected to the
AC bulky system and there are few of inverter loads in AC
case. The difference of efficiency in AC case with inverters,
Mode 2, Mode 3 and Mode 4 is slightly around 2 % in total.
The efficiency of these operating cases gains from 0.4 to
2.5% compared to AC case with inverter, while much lower
than AC case, which is roughly 92% if there is not any DC
on-site generation at distribution network.
In fact, a new technology transformers applied in this
paper would be competitive to increase the overall
efficiency. Due to lower efficiency in battery discharging,
the efficiency of 76.70% in Case 1 is the worst in this
comparative result.
V. CONCLUSION
In this paper, a test case of a Belgian DC local
distribution grid with a participation of DERs and storage
energy systems have been introduced in different operating
mode including islanding and grid-connected. Both voltage
drop, total losses and efficiency of the DC architecture is
Fig. 7. Contribution of demand and losses in total required power of different test cases
power flow calculation MatACDC. The DC powerflow
analysis is partly inherited from his previous PhD works.
REFERENCES
Fig. 9. Efficiency of power in different test modes
evaluated in simulation and compared each other. The paper
also investigates AC power distribution architecture with a
high efficiency transformer and compares it to the promising
DC power distribution alternative.
Efficiency is highly depends on working load compared
to the rated capacity of converters, as well as modes of
operations of DC networks. AC distribution grid is more
advantage where the penetration of DERs and DC-powered
loads is low.
The result reveals that the efficiency gain of DC power
distribution architecture is not much higher than in AC case.
Compared to previous research, DC efficiency on data
centers which consist of all electronic DC loads would
achieve 5-20% efficiency gain while the efficiency gain of a
DC local distribution grid in this paper is limited from 0.4 to
2.5%. In fact, the calculated result matches to the newest
conclusion on high efficiency on data centers given by APC
in [14]. It can be explained that the efficiency can be
improved in DC grids by increasing number of DC
electronic loads and by keeping a link between AC bulky
system and DC local distribution grid. AC distribution grids
also have its merit if modern- high efficient distribution
transformer is applied in the conventional AC systems.
Alternatively, gain of efficiency is highly varies depending
on the efficiency-curves of each components in DC
networks.
However, the future DC local distribution grid unlikely
depends on the efficiency gain in steady state power flow.
DC power distribution holds the most advantage for the
connection of emerging technologies for on-site power
generation and energy storage as a significant amount of this
equipment delivers power in the form of DC or alternatively
as high frequency AC, which then requires an intermittent
DC conversion. The efficient issues of primary reserves and
power flow redistribution via power electronic conversions
in transient are looking forward.
VI. ACKNOWLEGMENTS
The authors gratefully acknowledge Jef Berteen, a
colleague who developed an open source code for AC/DC
[1]
H. Knaak, “Modular Multilevel Converters and HVDC / FACTS : a
success story The modular multilevel technology,” in Proceedings
of the 14th European Conference on Power Electronics and
Applications (EPE 2011), 2011, no. Lcc, pp. 1 – 6.
[2]
P. T. M. Vaessen, “DC power distribution for server farms,” 2010.
[3]
“Onboard DC Grid - Breakthrough for DC Technology,” no. April
2012. p. 1, 2012.
[4]
S. Levy, “Why DC Power is the Next Wave in Power Generation ,
Storage and Applications Prepared by.” Tennessee Solar Energy
Association, p. 49, 2009.
[5]
L. Zhang, T. Wu, Y. Xing, K. Sun, and J. M. Gurrero, “Power
control of DC microgrid using DC bus signaling,” in 2011 TwentySixth Annual IEEE Applied Power Electronics Conference and
Exposition (APEC), 2011, pp. 1926–1932.
[6]
“Application Manual Power Semiconductors,” 2011.
[7]
M. Starke, L. M. Tolbert, and B. Ozpineci, “AC vs . DC
Distribution : A Loss Comparison,” in IEEE/PES Transmission and
Distribution Conference and Exposition, 2008, pp. 1–7.
[8]
R. D. Zimmerman, C. E. Murillo-Sánchez, and R. J. Thomas,
“MATPOWER : Steady-State Operations , Systems Research and
Education,” IEEE transaction on Power Systems, vol. 26, no. 1, pp.
12–19, 2011.
[9]
M. Padhee, P. K. Dash, L. S. Member, K. R. Krishnanand, and P. K.
Rout, “A Fast Gauss-Newton Algorithm for Islanding Detection in
Distributed Generation,” IEEE Transactions on Smart Grid, vol. 3,
no. 3, pp. 1181–1191, 2012.
[10]
E. Veilleux, S. Member, B. Ooi, and L. Fellow, “Power Flow
Analysis in Multi-Terminal HVDC Grid,” in Power Systems
Conference and Exposition (PSCE), 2011 IEEE/PES, 2011, pp. 1–7.
[11]
K. Purchala, L. Meeus, D. Van Dommelen, and R. Belmans,
“Usefulness of DC Power Flow for Active Power Flow Analysis,”
in IEEE Power Engineering Society General Meeting, 2005, p. 6.
[12]
“http://www.esat.kuleuven.be/electa/teaching/matacdc/,” 2011. .
[13]
A. Woyte, V. Van Thong, R. Belmans, and J. Nijs, “Voltage
Fluctuations on Distribution Level Introduced by Photovoltaic
Systems,” IEEE Transactions on Energy Conversion, vol. 21, no. 1,
pp. 202–209, 2006.
[14]
W. Paper, N. Rasmussen, and J. Spitaels, “A Quantitative
Comparison of High Efficiency AC vs . DC Power Distribution for
Data Centers,” White Paper. APC - Schneider Electric, p. 21, 2012.
[15]
J. Tant, G. S. Member, F. Geth, D. Six, P. Tant, and J. Driesen,
“Multi-Objective Battery Storage to Improve PV Integration in
Residential Distribution Grids,” IEEE Transaction of Sustainable
Energy, 2012.