Black Start Studies for Micro-grids with Distributed

Generators
1 1 1 2 2
YJ Wang , XZ Dong , B Wang , J Liu , A X Guo

(1 Dept. of Electrical Engineering, Tsinghua University, Haidian District, Beijing 100084, China

2 Electric Power Co. Of Shaanxi, Shaanxi, Xian, 710048, China)

Abstract: These progressive MG technologies and controlling

With the development of distributed generation apparatus have enhanced the stability of distributed
technologies, the reliability of Islanded distribution generators, which make it possible use a distributed
systems containing distributed generators (DG) generator as black start unit when a general or local
increased gradually, which make it possible to use a blackout occurs. If a fast black start recovery can be
distributed generator as black start unit when a general provided and allowing Micro-grid (MG) islanded
or local blackout occurs. In this paper, a low voltage operation to feed local consumers, customer interruption
distribution network with several kinds of distributed times can be significantly reduced.
generators is established based on Matlab® Simulink®
The procedure of power system restoration after a
simulation platform. Through numerical simulations, a
complete or partial black out can be divided into three
basic control strategy for micro-grid system restoration is
stages: black start unit restart, reconfiguration of the
summarized. And through three major steps, a method to
network and restoration of loads. Black start units are
define an automatic procedure for MG restoration can be
generators which are capable of starting on local support
obtained: i) Establish an optimization model to determine
without any external power source from the grid. These
the black start capability; ii) Assess each black start
units are mostly gas turbines or hydro units, and in a few
scheme considering two main factors, low voltage [1]
cases steam units. The DGs with storage devices
distribution network building time and minimum load
provide a new choice for black start unit, since some of
power factor during the black start procedure; iii) verify
them have the black start capability. Up to now, most
relative effective schemes by simulations and chose the
existing black start research work has focused on
best feasible black start scheme.
generating and assessing an optimal black start scheme
KEY WORDS: micro-grid, black start, distributed for high voltage networks (500kV in China). The choice
generators, inverter of black start schemes is a very complicated
semi-structural optimization problem. Expert
I INTRODUCTION system-based approaches, fuzzy logic-based analytic
hierarchy process (AHP), case-based reasoning model
Following the increasing penetration of distributed
and some other optimized algorithms have been applied
generation in medium voltage (10kV) networks,
to achieve reliable power system black start schemes.
micro-grid (MG) technologies have attracted lots of [3,4,10]
However, these models and algorithms are not
attention due to its potential for reducing greenhouse gas
[2]
appropriate and efficient for generating and selecting
emissions and increasing reliability to final consumers.
black start schemes in a MG with DGs. Many factors
The concept of MG is put forward by the Consortium for
considered for choosing high voltage networks black
Electric Reliability Technology Solutions (CERTS) firstly.
start program, such as voltage conversion frequency,
A MG typically contains electrical loads and small
path length, etc., are not considerable in MG black start.
generation systems (fuel cells, micro-turbines, wind
Moreover, in a MG the restoration procedure is much
generators, photovoltaic panels) together with storage
simpler because of the reduced number of controllable
devices (like flywheels and batteries) coexist through an
variables (loads, switches and micro-sources) compared
embedded management and control system.
with high voltage system restoration.
 This paper is organized as follows: Section II A. PQ Inverter Control
describes the simulation model for micro-gird. Section III
A current control technique has been used to perform
discusses the capability of DG as a black start unit and
a PQ inverter control (shown in Figure3): the inverter
the interaction among DGs and local loads. Section IV
transforms the three-phase voltage and current signals
proposes a novel method for generating and assessing
to dq0 frame using synchronous rotating coordinator
MG black start schemes. Section V provides the
transformation (abc to dq block in Figure3). Then
conclusion to the paper.
calculates the active power P and reactive power Q
II MICRO-GRID ARCHITECTURE AND (calculate power block in Figure3). Its output adjusted by
DISTRIBUTED GENERATOR MODELLING a PI regulator according to the given reference values
Pref and Qref (power control block in Figure3).
As mentioned before, a MG shown in Figure 1 involves
Subsequently, produces PWM drive pulses. The inverter
a low voltage electrical grid, loads, different kinds of
current is controlled in amplitude and phase to meet the
controlled and uncontrolled distributed generators and
desired set-points of active and reactive power (current
storage devices. [6,7,9]
control and PWM controller blocks in Figure3). The
Phase-locked loop block (PLL) is designed for frequency
and phase synchronization in synchronous rotating
coordinator transformation.

Pref id ref
power current PWM
Qref control control controller
inverter grid
iqref

P Q
PLL
phase ia
id ib
iq ic
calculate ud
abc ua
power uq
to
ub
dp0
u c

Fig.3 Signal processing in PQ inverter

Fig.1 Micro-grid with several DGs, loads and storage devices
B. V/f Inverter Control
Since inverters are usually required to provide AC grid
interface for distributed generators, we can simplify the In this case the inverter is controlled to feed the load
micro-grid simulation model according to the two typical with pre-defined values for voltage and frequency
types of inverter controls (as shown in Figure 2): PQ according to a specific control strategy. The signal
inverter control and V/f inverter control. processing procedure in V/f inverter is consistent with the
process described before instead fref and Vref of Pref and
Qref. The control principle of V/f inverter is based on grid
voltage-oriented vector control and equivalent control
block diagram for voltage and current is shown in Figure
4.

Fig.2 Simulation model for micro-gird

According to the mathematic model in dq coordinate,

PQ and V/f control structures are designed. In this paper,
a full representation of the inverter is required, because
when analysing the capability of DG as a black start unit
and the interaction among DGs and local loads, the initial voltage control current control

fast transients related with commutation of the solid state

switches need to be considered. All this models (DGs, Fig.4 Equivalent control block diagram for voltage and
filters, transmission lines, transformers) can be current
established using MatLab® Simulink® simulation
III CAPABILITY OF DG AS A BLACK START UNIT
platform.
 DGs meet the following two requirements have the The simulation results are shown in Table 1. Bus
possibility to be black start unit: i) the DGs with storage frequency deviation is round 0 in every simulation
devices (batteries or super-capacitors) in the DC bus of condition. Table 2 shows the simulation results when the
their inverters. Those DGs are capable to restart without load connected to Bus1 is a DG with PQ inverter control.
any external power source, feed local auxiliary control The enlighten parts of the tables are not meet criteria
systems and launch generation when a general or local and requirements.
blackout occurs, since the storage devices can provide
TABLE 1 Simulation results
energy in the initial stage; ii) In order to build the LV
network, those DGs’ inverters should operate in V/f power
+ -
Load/kVA THDu/% du /% du /%
control mode. factor

However the black start unit restart procedure is not 1 1 51 11 0

dealt in the work presented in this paper. In this section, 3 1 5.0 2.1 -0.3
we focus on the interaction among DGs and local loads 5 1 1.9 3.8 -0.4
in the initial stage of the black start procedure and 80 1 4.7 0 -8.7
analyse the voltage and frequency stability of DGs under 100 1 0.9 0 -17
an extreme condition. In the simulation system shown in 5 0.83 60 1.0 0
Figure 5, changing the load (connected to Bus1) capacity 10 0.90 23 5.2 -1.3
and power factor, the interaction among DGs and local 20 0.95 0.9 1.0 -0.2
loads can be evaluated by three indices (the V/f inverter 50 0.98 2.1 3.2 -0.4
80 0.99 4.7 0 -8.7
control signal, udref 220 u 2 311V , uqref 0 , Zref 100S ;
100 0.99 0.9 0 -17
1 -0.5 3.5 -0.5 -0.5
L s , Rs ,is the equivalent parameters of distribution lines;
3 -0.75 1.2 -0.2 -0.2
ESS refers to storage devices): 10 -0.90 0.5 -0.2 -0.2
20 -0.95 1.2 -0.3 -0.3
i) Bus (MG Bus in Figure 5) frequency deviation:
80 -0.99 0 -7.5 -7.5
df max( f  fe / fe ), fe 50Hz , f is the 100 -0.99 0 -17 -17

bus voltage frequency during the simulation; TABLE 2 Simulation results

+ -
ii) Bus voltage deviation: Pref/kW Qref/kVar THDu/% du /% du /%
1 0 7.7 2.1 -0.5
du  urms max  U e / U e ,du  urms min  U e / U e ,
3 0 7.6 1.6 -0.8
10 0 10.5 2.5 -0.6
Ue 220V , urms max , urms min are the maximum
25 0 22 6.5 -0.5
and minimum phase RMS voltage of the three
phases during the simulation.
From the results, some characteristics can be
iii) Bus voltage total harmonic distortion:
concluded: the inverter can operate properly with a

THDu max(
¦U 2
h
u100%) .
certain capacity of active load (in this case, 3 kW to
UI 80kW); this capacity decreases with inductive reactive
load increased and increases with capacitive reactive
load increased; No reactive power oscillations, voltage
MG Bus Bus 1
V/f control
i bus surge and other stability or synchronization problem
ESS VSC
Lf Ls Rs
Cf occur among the DGs in parallel operation.
u bus Load
IV BLACK START SCHEME

Based on the analyses shown in Section III, a black

Fig.5 Simulation system
start scheme using a DG with storage devices in the DC
bus of its V/f control inverter as a black start unit can be
developed through three major steps: i) Establish an
2 Load (30kW  1kVar)
optimization model to determine the black start capability;
ii) Assess each black start scheme considering two main 3 PQ DG 20kW 1kVA d P22  Q32 d 20kVA
factors, low voltage distribution network building time and 4 Load (15kW  3kVar)
minimum load power factor during the black start
5~7 Bus
procedure; iii) test relative effective schemes by
simulations and chose the best feasible black start
8 Load 0 d P8 d 100kW,-1kVar d Q8 d 30kVar
scheme. The application of this novel method will be Transf-
9 (0.8kW  1.5kVar)
illustrated by the following example. And before the black ormer
start procedure begins, it is required to disconnect of all
TABLE 4 Branch information table
loads in order to avoid large frequency and voltage
deviations when energizing the network. Each action for Num Conecting Transmission
MG restoration is carried out after the transients of the Node limitations
previous action. 1 1ǃ5 I d 200A
2 2ǃ5 I d 200A
A. Optimization Model
3 3ǃ6 I d 200A
Preparation for the establishment of optimization 4 4ǃ6 I d 200A
model includes: i) generate the network topology 5 5ǃ7 I d 200A
descripting the physical connections of the available 6 6ǃ7 I d 200A
generators, loads and other facilities in the MG; ii) 7 7ǃ9 I d 200A
establish node information table, at least the node 8 7ǃ8 I d 200A
number, type (power, load, transformer or bus), injected
power, restrictions should be involved; iii) establish The optimization model to determine the black start
branch information table, at least the branch number, capability described as:
connecting nodes, transmission power limitations should n
F min ¦ D i Fi ( X , U c )
be involved. i 1
(1)
s.t. X f ( x, U c , U d )
Figure 6 shows the network topology of the MG shown
g ( X ,U c ,U d ) 0
in Figure2, Table 3 shows the node information of the
h( X , U c , U d ) d 0
network, and Table 4 shows the branch information of
the network. The DG in Node 1 is the black start unit and Where, F is the overall objective of the optimization
its output power balances the demand and supply in the related to some optimization objective indexes Fi. In this
[1-14]
system. case, Fi includes: F1, maximize load restoration capacity
index; F2, minimize low voltage transformer (Node 9 in
B1 B3 7 9 Figure 6) start-up time index (When building the LV
1 5
1 5 7 network, it is necessary to energize the low voltage

2 8 transformer as soon as possible, since the earth

6
2
8
connection is performed at the transformer neutral point
[11]
); F3, minimize load inductive reactive power index. N
3
3 6
B2 is the number of indexes. D i is the weight of each index.
4 4
In this case, D1 10, D 2 3, D3 1 . X is state variable
Fig.6 Topology of the MG shown in Fig.2
vector which involves the bus voltage vector, current
TABLE 3 Node information table vector, etc. Uc is controllable variable vector which
involves active and reactive output power of generators,
Injected
Num Type restrictions switch status, etc. Ud is uncontrollable variable vector. x
Power
are characteristic parameters of the system determined
1 V/f DG 3kVA d P12  Q12 d 80kVA by parameters of devices, network structure of the grid,
etc. The state variable vector X is determined by
controllable variable vector Uc, uncontrollable variable Switch status: binary variables xi, i=1,2,..9;
vector Ud and characteristic parameters of the system x.
g and h are equality and inequality constraints of the Load in Node 8: real variables P9 ,0 d P9 d 100 ,
restoration. To simplify the model, Uc only includes
switch status xi and inverters reference command values Q9 , 1 d Q9 d 30.

( Vref 380Vˈf ref = 50Hz , Pref 20kWˈQref = 0 ). Equality Other given parameters include:
constraints are satisfied unless the black start unit P1e 50, P2 30, Q2 1, P3 20, P4 15
overloaded (exceed 80kW in this case). Inequality Q4 3, P9 0.8, Q9 1.5
constraints h include: The optimal solution:

Generator output power restriction: Pi min d Pi d Pi max , Switch status: x4 =0, xi =1,i z 4 ;

as for DG with inverter, S 2 d Pi 2  Qi2 d Si2max ;

i min
Load capacity in Node 9: P9 69kW, Q9 9.5kVar ;

In this case:
Black start unit output power: P1 80kW, Q1 12kVar ;
G1 =F4 P2 ˜ x1 x2  P4 ˜ x1 x5 x6 x4  P8 ˜ x1 x5 x8
 P9 ˜ x1 x5 x9  P3 ˜ x1 x5 x6 x3 , 80kW d G1 d 80kW Maximum load restoration capacity: 101kVA

B. Black Start Scheme

Inductive reactive load blow 10% of the active load
(otherwise the THDu of the inverter output will be Part A has optimize the final status of each switch and
output power of each load and generator. This part deals
intolerable): Qsum d 0.1Psum ;
with the sequence of nodes connected to the low voltage
network.
In this case:
In the case shown before, Node 2,3,8 and 9 (in Figure
Q2 ˜ x1 x2  Q4 ˜ x1 x5 x6 x4  Q8 ˜ x1 x5 x8  Q9 ˜ x1 x5 x9
6) need to be restored. There is a total 24 different
d 0.1u ( P2 ˜ x1 x2  P4 ˜ x1 x5 x6 x4  P8 ˜ x1 x5 x8  P9 ˜ x1 x5 x9 )
sequences. Check every sequence by the restrictions
Transmission limitation of each line: Iij min d Iij d Iij max , proposed in Part A (generator output power restriction
and load power factor limitation). Therefore, feasible
this will be checked in the next stage; black start schemes can be selected. Then assess these
In this way, we can avoid the calculation of sate feasible schemes with two indexes:
variable vector X. Low voltage transformer start-up time: t;
And optimization objective indexes calculated as: Maximum power factor during the restoration: p.
F1 ( P22  Q22 ) ˜ x1 x2  ( P42  Q42 ) ˜ x1 x5 x6 x4  ( P82  Q82 ) ˜ x1 x5 x8 The feasible black start schemes of the MG shown in
Low voltage transformer can be equivalent to the Figure 2 revealed in Table 5.
inductive reactive load. Here we amplify the capacity of TABLE 5 Feasible black start schemes
the transformer equivalent load in order to guarantee the Restoration
recovery of low voltage transformer. Sequence t p

F1 changes to F1 :
’ 2ǃ3ǃ8ǃ9 5 times of switch action 0.03

F1 ' ( P22  Q22 ) ˜ x1 x2  ( P42  Q42 ) ˜ x1 x5 x6 x4  ( P82  Q82 ) ˜ x1 x5 x8 2ǃ3ǃ9ǃ8 6 times of switch action 0.03
 k ( P92  Q92 ) ˜ x1 x5 x9 , k 100 2ǃ8ǃ3ǃ9 3 times of switch action 0.17

Quantify the two indexes with different weight, the best

F3 Q2 ˜ x1 x2  Q4 ˜ x1 x5 x6 x4  Q8 ˜ x1 x5 x8  Q9 ˜ x1 x5 x9
feasible black start scheme can be selected (in this case,

F2 will be used as an assessment index in next step. the best scheme is Scheme 3 shown in Table 5).

Variable in this case includes:

 A basic sequence of actions can be applied to reduce of a Black Start Decision Supporting System for Isolated
computation time: Power Systems[J]. Power Systems, IEEE Transactions
on, 2013,28(3):2202-2210.
a) Black start unit start and energize certain capacity
[4] Datta S, Kolluri S, He T, et al. Development of an
load directly connected with it;
Alternative Black- Start Plan for System Restoration:
b) Energize the low voltage bus; Power Systems Conference and Exposition, 2006. PSCE
c) Energize other controllable generators and loads; '06. 2006 IEEE PES, Atlanta, GA, 2006[C].Oct. 29
2006-Nov. 1 2006.
d) Energize the remain systems;
[5] Dariani A, Rastegar R. EMTP-based black start analysis
C. Simulation Verification for an isolated power system: Electric Power and Energy

Simulate the best feasible black start scheme and Conference (EPEC), 2010 IEEE, Halifax, NS,

verify that the magnitude, frequency and waveform of 2010[C].25-27 Aug. 2010.

voltage and current at each bus meet the criteria and [6] Mehrtash A, Peng W, Goel L. Reliability Evaluation of

requirements. If so, the scheme is chosen as the black Power Systems Considering Restructuring and

start scheme, otherwise, verify the second best scheme Renewable Generators[J]. Power Systems, IEEE

(Scheme 1 shown in Table 5) until a reliable scheme Transactions on, 2012,27(1):243-250.

found. [7] Costa P M, Matos M A. Reliability of distribution

networks with microgrids: Power Tech, 2005 IEEE
In the example, the Scheme 3 shown in Table 5 is the Russia, St. Petersburg, 2005[C].27-30 June 2005.
best reliable black out scheme after simulation [8] Yue S, Yu J, Liang B. Research on based assessment
verification. (Deviation of bus voltage: -13.7% to 2.1%; method of power system black-start effective schemes of
Deviation of voltage frequency: 0; Maximum THDu: 15%; ancillary services: Electric Utility Deregulation and
Maximum phase RMS current: 85A). Restructuring and Power Technologies, 2008. DRPT
V CONCLUSIONS 2008. Third International Conference on, Nanjuing,
2008[C].6-9 April 2008.
This paper established a simulation system of a MG
[9] Jinwei L, Jianhui S, Xiangzhen Y, et al. Study on
using MatLab® Simulink® simulation platform, and
microgrid operation control and black start: Electric
analyzed the capability of a DG as a black start unit.
Utility Deregulation and Restructuring and Power
From the results obtained, a novel method to develop a
Technologies (DRPT), 2011 4th International
black start scheme for the MG was proposed. This
Conference on, Weihai, Shandong, 2011[C].6-9 July
method simplified the optimization problem of MG black
2011.
start. An example was given to illustrate the method and
[10] Saraf N, McIntyre K, Dumas J, et al. The Annual Black
its validity and effectiveness was verified by simulation.
Start Service Selection Analysis of ERCOT Grid[J].
VI ACKNOWLEDGMENTS Power Systems, IEEE Transactions on,
This work is supported by Project of Shaanxi Electric 2009,24(4):1867-1874.
Power Research Institute, SGCC (Research of [11] Li B, Li Y, Ma T. Research on earthing schemes in LV
self-healing control technology in smart distribution grid). microgrids: Advanced Power System Automation and

 Protection (APAP), 2011 International Conference on,

VII REFERENCES Beijing, 2011[C].16-20 Oct. 2011.
[1] Weijia L, Zhenzhi L, Fushuan W. Analysis and
optimisation of the preferences of decision-makers in
black-start group decision-making[J]. Generation,
Transmission & Distribution, IET, 2013,7(1):14-23.
[2] Feltes J W, Grande-Moran C. Black start studies for
system restoration: Power and Energy Society General
Meeting - Conversion and Delivery of Electrical Energy
in the 21st Century, 2008 IEEE, Pittsburgh, PA,
2008[C].20-24 July 2008.
[3] Yi-Ting C, Chih-Wen L, Yi-Jen W, et al. Development