1876-6102 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license [609756]

1876-6102 © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/ ).
Peer-review under responsibility of the scientific committee of the Applied Energy Symposium and Forum, REM2016: Renewable Ener gy Integration
with Mini/Microgrid.doi: 10.1016/j.egypro.2016.11.261 Energy Procedia 103 ( 2016 ) 129 – 134 ScienceDirect
Applied Energy Symposium and Forum, REM2016: Renewable Energy Integration with
Mini/Microgrid, 19-21 April 2016, Maldives
Artificial Neural Network Based Fault Detection and Fault
Location in the DC Microgrid
Qingqing Yanga, Jianwei Lia*, Simon Le Blonda, Cheng Wanga
aUniversity of Bath, Claverton Down, Bath, BA2 7AY, United Kingdom
Abstract
In DC microgrid, power electronic devices may suffer from over current during short circuit faults. Since DC bus
systems cannot sustain high fault currents, suitable protection strategy in DC lines is indispensable. This paper presents a novel use of artificial neural network (ANN) for fault detection and fault location in a low voltage DC bus
microgrid system. In the proposed scheme, the faults on DC bus can be fast detected and then isolated without de-
energizing the entire system, hence achieving a more reliable DC microgrid. The neural network is trained based on the different short circuit faults in DC bus to ensure its validity. A microgrid with ring DC bus, which is segmented
into overlapping nodes and linked with circuit breakers, is built in PSCAD/EMTDC to test the performance of the
protection scheme.
© 2016 The Authors. Published by Elsevier Ltd.
Selection and/or peer-review under responsibility of REM2016
Keywords: Artificial neural network; DC microgrid; fault detection; fault location, short circuit fault
1. Introduction
Nowadays, a large amount of renewable powers an d energy storage systems are introduced to the
conventional power system [1]. However, the larg e penetration of distributed generations may challenge
to conventional power generation and distribution syst em, such as the voltage rise, frequency fluctuations,
and protection problems [2]. Therefore, the microgri d which is defined as a low voltage system with
generations and energy storage systems providing elect ricity to local demand is becoming more and more
attractive [3]. Many previous wo rks have investigated the applica tion of AC microgrid. However, DC
microgrid has been proved to have many advantages ove r AC microgrid [2]. Firstly, DC power devices
* Corresponding author. Tel.: +44 7449015799.
E-mail address: [anonimizat] (Jianwei Li), [anonimizat] (Qingqing Yang).
Available online at www.sciencedirect.com
© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/ ).
Peer-review under responsibility of the scientific committee of the Applied Energy Symposium and Forum, REM2016: Renewable Energy Integration with Mini/Microgrid.

130 Qingqing Yang et al. / Energy Procedia 103 ( 2016 ) 129 – 134
such as batteries and solar panels are largely integrated into the AC microgrid, which means a great many
of converters are needed ,resulting in larg einvestment on power conversion system. In addition, the micro
AC power supplies which have varying voltage or frequency from those of the utility grids also need
power conversion [4].Compared with AC system, the requirement sof power converters are reduced in
DC microgrid resulting in higher efficiency and lower investment [5]. Secondly, there is no requirement
for synchronization, and the loss of the reactive power and harmonics in DC microgrid [1]. In addition,
the DC microgrid system is more conducive t osimpler and more efficient power electronic interface [6,
7].Hence, b uilding DC transmission network in microgrid, connecting the distributed power supplies and
loads, has become a new research direction.
Even though DC microgrid has lots of advantages , there are still some challenges in DC microgrid
protection [8, 9]. Generally, it is difficult to locate the fau lt before isolation and to extract useful
information for fault location [10].This paper focuses on DC bus fault detection and fault location .The
rapid detection and location of fault ,based on transient voltage signal or fault current signal ,is very
important in protection scheme. As indicate sin [11-13], the ANN based method is one of the most
efficient method for fault detection and fault loca tion on AC grid and HVDC system because of its
accuracy, robustness and fastness .However, to the authors ’best knowledge, there is no published work
for ANN applied in DC microgrid protection .Hence, this paper introduces the Artificial Intelligence (AI)
based method for fault detection and fault location in a DC microgrid system .A DC microgrid is built in
PSCAD/EMTDC to simulate the DC system under normal and transient conditions to analyze the fault profile and build training set for ANN. Different fault types with various fault resistance
sand fault
locations are studied in the test network. Two neural networks are established for fault detection and fault location respectively
.The faulted segment will be isolated, and the rest system will keep working. The
simulation results show an accuracy performance of the proposed method.
2. System description and modeling
The structure of typical DC microgrid is shown in Figure 1, which mainly consists of wind power
generation, energy storage system, load sand AC main grid connection four parts .The wind turbine is
used as a typical distributed generation connected to the grid. The wind turbines working as a typical
distributed generation are developed based on a simplified model in [14].The converter station is operated
under maximum power point tracking (MPPT) mode to capture wind energy as much as possible [15].
The DC battery energy storage with the bidirectional DC/DC converter is simulated as a storage unit
based on [16]and worked as energy storage unit in this paper. During normal operation, battery works in
the charging statues or as a back -up power supply. However, during island operation, battery energy
storage is used as a slack bus to keep DC voltage stable and steady operation .DC load sdirectly connect to
DC microgrid through DC/DC converter, and AC load s connect to grid by using VSC converter. When
power supply is insufficient, load shedding control strategy will be applied to keep power balancing .DC
microgrid is incorporated into the main AC grid by VSC converter with bi -directional power flow [17].
When the DC microgrid network is in normal operation, the active power balancing is kept by control DC
voltage through VSC converter. But during the AC voltage drop cause by short circuit fault, VSC will lose the stability of controlling DC voltage, but limit the current.
Several kinds of topology such as ring type, radial type, and central
-ring type and general topology can
be used in DC system. Ring topology of DC microgrid is selected in this paper. Even though the investment will be increased using this topology due to the increase of DC transmission line length and
capacity, which results in the increasing quantity of circuit breaker, this kind of topology is more flexible
and robust. During a DC line fault, the circuit breakers operate and cut the fault at both ends of the line to keep stable operation with no power losses. Other part of the DC line will undertake the increase of
transmission capacity. Hence, the ring topology can play the advantages of DC microgrid
for an ideal
networkin g.

Qingqing Yang et al. / Energy Procedia 103 ( 2016 ) 129 – 134 131
Fig. 1. The configuration of DC microgrid
3. ANN based fault detection and fault location methodology
Two types of fault sare considered in the fault detection method in DC system, which are pole -to-pole
fault and pole -to-ground fault [8]. In general, pole -to-pole fault is caused by external mechanical stress
which can be regarded as permanent fault. Pole -to-ground fault happens with high probability which is a
kind of temporary fault due to a branch drop or lightning. The appropriate line protection strategy is
conducive to reduc ethe loss of the system and avoid the damage to the whole DC system .The protection
strategies and post fault recovery capability of the system must be considered .
Compared with high voltage direct current (HVDC )system ,low voltage DC system (LVDC) for power
system is a relatively new concept [10].During faults, a complete route will be developed through anti –
parallel diode in VSC, so that converter station will rel ease active power to the fault point. This may cause
over current on DC bus and transmission line with a very high value and the change of current direction .
Fig. 2.Direction of currents under short circuit fault
The basic structure of the simulated system is shown in figure 2. DC circuit breaker swhich play a
considerable role in DC fault isolation are installe d at both end of the line to break the fault within

132 Qingqing Yang et al. / Energy Procedia 103 ( 2016 ) 129 – 134
milliseconds. The direction of current before and afte r fault can be clearly seen. On different faulted
segment, the directions of current are varying.
Fig. 3.DC current under short circuit fault
The current signal give sthe appropriate information regarding the different power system condition
(figure 3) .During a fault, currents on every terminal will increase rapidly to a very large value which may
result in severe damage to the electronic devices. It can be also obtained that the magnitude of fault
current changes under different fault location sand different fault resistances. The slope sof the current
increase and peak magnitude are difference as well as can be seen in figure 3 (a) .It can be concluded that
the same fault location appears the same rate of curr ent rise. Hence, the current signals on both end of the
line are used as input sin the designed ANNs.
Fig. 4.Artificial neural network model
Figure 4 shows the developed ANN modules for the fault detection and fault location on the bus
segments of the simulated system. The artificial inte lligence algorithm will help the system make a correct
decision to tell whether there is a fault and where the fault occurs. As indicated in the DC fault analysis,
samples of the current waveforms obtained are select ed for ANN training under different conditions to
detect fault sand locate the faults. ANNs will consider the di rection, slope of cu rrent increase and the
information obtained from DC current signal s. Two ANNs are designed in the protection scheme. One
ANN is to detect the fault and the other one is to determine the fa ult location on the dc bus segments with
the same input data. After the detection of fault, circuit breakers are used to isolate the bus segment after fault
s.

Qingqing Yang et al. / Energy Procedia 103 ( 2016 ) 129 – 134 133
4. Simulation test
Fig. 5.DC current of transmission line
Simulation is completed based on the DC microgrid model introduced above using PSCAD/EMTDC as
aplatform. The power in network for wind turbine, battery energy storage, loads and AC grid are 22 kW,
10kW, 30 kW and 18kW respectively. The fault is set at 1.5 s last 0.01s on the DC bus segment between wind turbine and AC grid. The current data detected from each terminal of the bus segments are shown in figure 5. The surge of current and the change of positive/negative value obviously displayed.
The DC current
son both end of the line are select ed as input data. Multilayer feed -forward ANN is
used in this paper .Fault detection and fault location processes are completed in two ANNs. With the same
input vectors applied in two different ANNs, the fault can be accurately detected and the location of fault on bus segment can be measured.
Table 1. The results based on ANN method
Fault type Fault resistance Fault location Measured location Error
Pole-to-pole 0.7 56% 55.32% 0.68%
Pole-to-ground 0.65 78% 78.09% 0.09%
Pole-to-pole 2.5 25% 25.25% 0.25%
Pole-to-ground 6 47% 46.84% 0.16%
Pole-to-pole 1.2 89% 88.82% 0.18%
Pole-to-ground 0.25 11% 11.03% 0.03%
A sampling rate of 5kHz is used to acquire signals and current data windows of 20 samples are
obtained at each side of the source, which are applied as ANN input. Therefore ,the designed ANN
receives 40 input vectors to the neural network in figure 4.Different cases are considered for neural
network training including the situation in different fault resistance sand fault location s. 20 samples of
cases with fault resistance of 0.1 ¡, 0.5¡,1¡, 2 ¡, 10 ¡and fault location of 10%, 30% 50%, 70%,
90% in each bus segments are detected for data traini ng. Fault detector and locator are trained with its
corresponding 2250 cases. Results are presented in table 1. In the result, the faults can be detected
correctly on each bus segments through the designed ANN with 100% accuracy. With the accurate
detection, the fault on each segment will be isolated by circuit breakers at the terminal svery fast. In
addition, the error for fault location is detected within 1%.

134 Qingqing Yang et al. / Energy Procedia 103 ( 2016 ) 129 – 134
5. Conclusions
An artificial neural network based fault detection and fault location method has been presented in this
paper. The detailed modeling of DC microgrid including wind turbine, battery energy storage system, loads and AC grid is simulated in PSCAD/EMTDC. Through analyzing faults of the modeled system, DC current signals are used as inputs in the presented method. The results demonstrated that any types of DC faults can be accurately and fast detected. In addition, the fault location can be detected within 1% error. The ANN based method can be regarded as a very efficient method for DC microgrid.
6. Copyright
Authors keep full copyright over papers published in Energy Procedia
References
[1] Lu X, Sun K, Guerrero JM, Vasquez JC, Huang L. State -of-charge balance using adaptive droop control for distributed energy
storage systems in DC microgrid applications. Industrial Electronics, IEEE Transactions on. 2014;61:2804 -15.
[2] Kakigano H, Miura Y, Ise T, Momose T, Hayakawa H. Fundamental characteristics of DC microgrid for residential houses with
cogeneration system in each house. Power and Energy Society General Meeting -Conversion and Delivery of Electrical Energy in
the 21st Century, 2008 IEEE: IEEE; 2008. p. 1 -8.
[3] Planas E, Andreu J, Gárate JI, de Alegría IM, Ibarra E. AC and DC technology in microgrids: A review. Renewable and
Sustainable Energy Reviews. 2015;43:726 -49.
[4] Kakigano H, Miura Y, Ise T. Low -voltage bipolar -type DC microgrid for super high quality distribution. Power Electronics,
IEEE Transactions on. 2010;25:3066 -75.
[5] Ito Y, Zhongqing Y, Akagi H. DC microgrid based distribution power generation system. Power Electronics and Motion
Control Conference, 2004 IPEMC 2004 The 4th International: IEEE; 2004. p. 1740 -5.
[6] Salomonsson D, Sannino A. Low -voltage DC distribution system for commercial power systems with sensitive electronic loads.
Power Delivery, IEEE Transactions on. 2007;22:1620 -7.
[7] Salomonsson D, Söder L, Sannino A. Protection of low -voltage DC microgrids. Power Delivery, IEEE Transactions on.
2009;24:1045 -53.
[8] Park J -D, Candelaria J. Fault detection and isolation in low -voltage DC -bus microgrid system. Power Delivery, IEEE
Transactions on. 2013;28:779 -87.
[9] Laaksonen HJ. Protection principles for future microgrids. Power electronics, IEEE transactions on. 2010;25:2910 -8.
[10] Park J -D, Candelaria J, Ma L, Dunn K. DC ring -bus microgrid fault protection and identification of fault location. Power
Delivery, IEEE Transactions on. 2013;28:2574 -84.
[11] Mirzaei M, Ab Kadir M, Moazami E, Hizam H. Review of fault location methods for distribution power system. Australian
Journal of Basic and Applied Sciences. 2009;3:2670 -6.
[12] Zhang N, Kezunovic M. Transmission line boundary protection using wavelet transform and neural network. Power Delivery,
IEEE Transactions on. 2007;22:859 -69.
[13] Chang C, Kumar S, Liu B, Khambadkone A. Real -time detection using wavelet transform and neural network of short -circuit
faults within a train in DC transit systems. Electric Power Applications, IEE Proceedings -: IET; 2001. p. 251 -6.
[14] Li J, Gee AM, Zhang M, Yuan W. Analysis of battery lifetime extension in a SMES -battery hybrid energy storage system using
a novel battery lifetime model. Energy. 2015;86:175 -85.
[15] Yang Q, Gu C, Le Blond S, Li J. Control scheme for energy storage in domestic households. Power Engineering Conference
(UPEC), 2014 49th International Universities: IEEE; 2014. p. 1 -6.
[16] Li J, Zhang M, Yang Q, Zhang Z, Yuan W. SMES/battery Hybrid Energy Storage System for Electric Buses. 2016.[17] Salomonsson D. Modeling, control and protection of low
-voltage dc microgrids. 2008.