Applied Energy 194 (2017) 343–352

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Optimized hierarchical power oscillations control for distributed

generation under unbalanced conditions
Peng Jin a,1, Yang Li b,⇑,1, Guoqing Li b, Zhe Chen c, Xiaojuan Zhai b
a
State Grid Customer Service Center, Tianjin 300309, PR China
b
School of Electrical Engineering, Northeast Dianli University, Jilin 132012, PR China
c
Institute of Energy Technology, Aalborg University, Aalborg DK-9220, Denmark

h i g h l i g h t s

A hierarchical control structure is proposed to reduce power oscillations for DGs.

Relationship between amplitudes of active and reactive power oscillations is deduced.

Injected negative-sequence currents are obviously restrained by using the proposal.

Test results show that the proposal can effectively limit the oscillations together.

a r t i c l e i n f o a b s t r a c t

Article history: Control structures have critical influences on converter-interfaced distributed generations (DG) under
Received 14 March 2016 unbalanced conditions. Most of previous works focus on suppressing active power oscillations and rip-
Received in revised form 18 May 2016 ples of DC bus voltage. In this paper, the relationship between amplitudes of the active power oscillations
Accepted 17 June 2016
and the reactive power oscillations are firstly deduced and the hierarchical control of DG is proposed to
Available online 24 June 2016
reduce power oscillations. The hierarchical control consists of primary and secondary levels. Current ref-
erences are generated in primary control level and the active power oscillations can be suppressed by a
Keywords:
dual current controller. Secondary control reduces the active power and reactive power oscillations
Distributed generation
Hierarchical control
simultaneously by optimal model aiming for minimum amplitudes of oscillations. Simulation results
Power oscillations suppression show that the proposed secondary control with less injecting negative-sequence current than traditional
Dual current controller control methods can effectively limit both active power and reactive power oscillations.
Unbalanced conditions Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction oscillations have negative effects on DC-link of converters; for

With the rapid development of renewable energy resources, the
concept of converter-interfaced distributed generation (DG) is
playing an important role in power grids [1–3]. The voltage-
sourced converter (VSC) has become a key component of an
energy-conversion device for DGs [4]. Due to inertialess factors
of VSC, grid disturbances have adverse impact on performances
of DG converters [5,6]. Most of DGs are located at the terminals
of a distribution network or microgrid where unbalanced condi-
tions exist owing to single-phase loads and asymmetrical faults
[7,8]. Grid imbalance in a three-phase system leads to double
working frequency power oscillations. For one thing, active power
⇑ Corresponding author.
E-mail address: liyang@nedu.edu.cn (Y. Li).
1
Peng Jin and Yang Li contributed equally to this work and should be considered
co-first authors.
another, reactive power oscillations may result in high power loss
and over-current stress [9]. Consequently, various control struc-
tures have been proposed to enhance the operation performance
of VSC under asymmetrical conditions in recent years.
It’s known that control structures play a very important role in
the DG’s behavior under unbalanced conditions [10]. Notch filters
are adopted to separate the components of positive- and
negative-sequences from the sampling electrical quantities and
the dual PI controllers in dual synchronous rotating frames (SRFs)
are proposed to regulate positive- and negative-sequence compo-
nents respectively [11]. The DC-link voltage ripple and active
power oscillations can be suppressed by means of the dual current
controllers, but reactive power oscillations are significantly ampli-
fied and system dynamics are limited by notch filters. On the con-
trary, to extract symmetrical components, the delay signal
cancellation (DSC) is employed by combining 1/4 working fre-
quency cycle delay and the original AC value [12]. Moreover,

http://dx.doi.org/10.1016/j.apenergy.2016.06.075
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
344 P. Jin et al. / Applied Energy 194 (2017) 343–352

positive- and negative-sequences detection can be replaced when

VT1 VT3 VT5
resonant controllers are adopted and it is demonstrated that the
stationary frame resonant controllers are suitable for extracting VD1 VD3 VD L R
ia 5
symmetric-sequence electrical quantities. The dual SRFs can be A PCC
Udc B ib
Ceq
simplified to one stationary frame and realize zero error tracking. C
VT6 ic
Further, in order to promote DG performances and system dynamic VT4 VT2
tracking, controllers must be able to extract the positive- and
VD 4 VD6 VD2
negative-sequences components and achieve feedback control
[13,14]. Therefore, all these control schemes focus on eliminating
Fig. 1. Equivalent circuit of converter for DG.
the active power oscillations and improving reference tracking,
while there is seldom optimized operation strategy considering
reactive power compensation and DC bus voltage oscillation simul- The converter model of DG is described in Fig. 1 and can be
taneously [15]. decomposed as two separate parts
Besides control structures, it is also very important to generate 8 dIþ
current references under unbalanced conditions. Considering the < Eþ ¼ V þdqþ þ L dqþ
þ jxLIþdqþ þ RIþdqþ
dqþ dt
constraints of DG converter, an accepted constant DC-link bus volt- ð2Þ
: dI
age should be maintained. Based on the input positive- and Edq ¼ V dq þ L dq
dt
 jxLIdq þ RIdq
negative-sequence components, an unbalanced transfer matrix of
where input current {Ia , Ib , Ic } and output voltage of inverter {V a , V b ,
input phase voltages is generally feasible [11]. However, a constant
V c } are also expressed in forms of positive- and negative-sequence
DC bus is obtained at the expense of asymmetrical currents and a
components like three-phase input voltage. With the unbalanced
sharp increase of reactive power oscillations. As discussed in [16],
input voltage, apparent power is given by
several schemes may improve different quality of electric energy at
  
the point of common coupling (PCC) in terms of power oscillations S ¼ ejxt Eþdqþ þ ejxt Edq ejxt Iþdqþ þ ejxt Idq ð3Þ
and current distortions. These strategies show flexible adjustment
of converters under asymmetrical voltage. However, most of them An unbalanced three-phase input voltage {Ea , Eb , Ec } at the PCC
only deal with some specific distortions. It is reasonable to regulate causes double working frequency power oscillations, and instanta-
the amplitudes of active and reactive power oscillations through an neous power of a DG can be expressed as
adjustable parameter of current Ref. [17]. Moreover, the relation-
ship between active and reactive power oscillations is discussed,
PðtÞ ¼ P0 þ Pc2 cosð2xtÞ þ Ps2 sinð2xtÞ ð4Þ
but it is only lagging in the qualitative analysis stage.
Previous studies mainly focus on the reduction of active power
QðtÞ ¼ Q 0 þ Q c2 cosð2xtÞ þ Q s2 sinð2xtÞ ð5Þ
oscillations, but the analytic relationship between active and reac- where P c2 , P s2 , Q c2 , Q s2 caused by unbalanced input voltage appear
tive power oscillations is still unclear. In addition, most work in the as the double working frequency oscillations, and they can be
field of DG operation is under unbalanced grid faults, and seldom expressed as
work focuses on long-term operation under a low voltage unbal- 8
ance factor. However, this working condition widely exists at ter- >
> P0 ¼ Eþdþ Iþdþ þ Eþqþ Iþqþ þ Ed Id þ Eq Iq
<
minals of distribution networks or microgrids with DGs. Pc2 ¼ Eþdþ Id þ Eþqþ Iq þ Ed Iþdþ þ Eq Iþqþ ð6Þ
In this paper, a novel hierarchical control structure is proposed >
>
:
to suppress the power oscillations for DGs under unbalanced con- Ps2 ¼ Eþdþ Iq  Eþqþ Id  Ed Iþqþ þ Eq Iþdþ
ditions. The contribution of this paper is twofold: the relationship 8
between amplitudes of the active power oscillations and the reac- >
> Q ¼ Eþqþ Iþdþ  Eþdþ Iþqþ þ Eq Id  Ed Iq
< 0
tive power oscillations is firstly deduced and a hierarchical control Q c2 ¼ Eþqþ Id  Eþdþ Iq þ Eq Iþdþ  Ed Iþqþ ð7Þ
structure of DGs is proposed to reduce power oscillations; and sig- >
>
: Q ¼ Eþ I  þ Eþ I   E I þ  E I þ
nificant performance improvements from applying the proposed s2 dþ d qþ q d dþ q qþ

scheme is demonstrated in detail.

Instantaneous active and reactive power in (4) and (5) can be
The remainder of this paper is organized as follows. First, the
rewritten as
mechanism of power oscillations is revealed from viewpoint of   
the positive- and negative-sequence current injection. Then, an Pc2
PðtÞ ¼ P0 þ Pv sin 2xt þ tan1 ð8Þ
optimization model for suppression of power oscillations is estab- P s2
lished, and a hierarchical control structure is proposed to reduce   
both the active and reactive power oscillations simultaneously. Q c2
QðtÞ ¼ Q 0 þ Q v sin 2xt þ tan1 ð9Þ
Application of the proposed control scheme is demonstrated using Q s2
simulation tests, and finally the conclusions are made. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where P v ¼ P 2s2 þ P2c2 , Q v ¼ Q 2s2 þ Q 2c2 . From the Eqs. (8) and (9),
2. Mathematical description of converter under unbalanced the instantaneous power can be divided into AC and DC compo-
conditions nents. When a microgrid contains single-phase loads and sources,
active and reactive power oscillations exist simultaneously because
An unbalanced three-phase input voltage {Ea , Eb , Ec } at PCC of unbalance voltage in PCC. Consequently, there is an urgent need
without a zero sequence can be represented as the sum of to reveal the couple relationship between them.
positive- and negative-sequences, such that

Eab ¼ ejxt Eþdqþ þ ejxt Edq ð1Þ 3. Mechanism of power oscillation under unbalanced
conditions
þ 
where Eþ þ  
dqþ ¼ Edþ þ jEqþ , Edq ¼ Ed þ jEq ; +, – respectively denote
the positive and negative-sequence component; x is the angular Based on (6) and (7), the relationship between active and reac-
frequency. tive power oscillations can be derived as followed:
 P. Jin et al. / Applied Energy 194 (2017) 343–352 345

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P2v þ Q 2v ¼ P2s2 þ P2c2 þ Q 2s2 þ Q 2c2 P2s2 þ P2c2
Pv
 2 p ¼ ¼ ð18Þ
¼ Eþdþ Iq  Eþqþ Id  Ed Iþqþ þ Eq Iþdþ P0 P0
 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ Eþdþ Id þ Eþqþ Iq þ Ed Iþdþ þ Eq Iþqþ Q Q 2s2 þ Q 2c2
 2 q ¼ v¼

ð19Þ
Q0 Q0
þ Eþdþ Id þ Eþqþ Iq  Ed Iþdþ  Eq Iþqþ
 2 The active and reactive power oscillation rates p and q are the
þ Eþqþ Id  Eþdþ Iq þ Eq Iþdþ  Ed Iþqþ ð10Þ ratio of the corresponding power oscillation amplitude and the
average power. Therefore, a control strategy suppressing both
Given A ¼ Eþ  þ   þ  þ þ  þ 
dþ I d þ Eqþ I q , B ¼ Ed I dþ þ Eq I qþ , C ¼ Edþ I q  Eqþ I d ,
amplitudes and rates of power oscillations is needed.
D ¼ E þ  þ
q I dþ  Ed I qþ , (10) can be simply denoted as follows:
4. Hierarchical control scheme of DG converter
P2v þ Q 2v ¼ ðC þ DÞ2 þ ðA þ BÞ2 þ ðA  BÞ2 þ ðC þ DÞ2
n
2 4.1. A primary control of DG
¼ 2ðA2 þ B2 þ C 2 þ D2 Þ ¼ 2 ðEþdþ Id Þ þ 2Eþdþ Id Eþqþ Iq
2 2 2
þ ðEþqþ Iq Þ þ ðEd Iþdþ Þ þ 2Ed Iþdþ Eq Iþqþ þ ðEq Iþqþ Þ The hierarchical control consists of two levels. The difference
2 2 2 between primary control and secondary control lies in current ref-
þ ðEþdþ Iq Þ  2Eþdþ Iq Eþqþ Id þ ðEþqþ Id Þ
þ ðEq Iþdþ Þ erences of the controller. The active power oscillations can be sup-
o n
2 þ  2 2
 2Eq Iþdþ Ed Iþqþ þ ðEd Iþqþ Þ ¼ 2 ðEdþ Id Þ þ ðEþqþ Iq Þ pressed in the primary level by matrix as follows [11]:
2 3 2 31
2 2 2 2 2
Eþdþ Eþqþ Ed Eq 2 3
þ ðEd Iþdþ Þ þ ðEq Iþqþ Þ þ ðEþdþ Iq Þ þ ðEþqþ Id Þ þ ðEq Iþdþ Þ Iþd0þ P0
o 6 þ 7 6 7
þ ðEd Iþqþ Þ
2 6I 7 6 Eþqþ Eþdþ Eq Ed 7 6 7
ð11Þ 6 q0þ 7 6 7 6 Q0 7
6  7¼6 7 6 7 ð20Þ
6 Id0 7 6 E Ed Eþqþ þ 7 6 7
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þ 2
4 5 6
4 q
Edþ 7
5
4 Ps2 5
where Pv ¼ P 2s2 þ P2c2 , Q v ¼ Q 2s2 þ Q 2c2 , Eþ ¼ Eþ dþ þ Eqþ , 
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Iq0 Ed Eq Eþdþ Eþqþ Pc2
2 þ 2
E ¼ E 2
d þ Eq , I ¼
 2 þ
Iþdþ þ I qþ , I ¼

I 2  2
d þ I q . P v and Q v
where Ps2 ¼ Pc2 ¼ 0. The active power oscillations can be eliminated
are the amplitudes of active and reactive power oscillations, respec- in ideal state. However, the negative-sequence currents are injected
tively. Eq. (11) can be simplified as
and reactive power oscillations cannot be reduced. Once active
P2v þ Q 2v ¼ 2ðE Iþ Þ þ 2ðEþ I Þ
2 2
ð12Þ power oscillations are suppressed in primary control, constant
DC-link voltage is kept for equivalent capacitance of converter.
In (12), power oscillations are divided into two parts: positive- But only four power coefficients in (6) and (7) can be controlled
sequence current and negative-sequence voltage constitute one and negative-sequence currents are injected when the DG is in pri-
part, and negative-sequence current and positive-sequence voltage mary control. Eq. (12) indicates that negative-sequence currents
constitute another. 2
could lead to additional power oscillations term 2ðEþ I Þ and reac-
In the normal state, the amplitude of negative-sequence electric
tive power oscillations are enlarged.
quantity is much less than that of positive-sequence electric quan-
tity. Consequently,
4.2. Secondary control of DG
Eþdþ Iþdþ þ Eþqþ Iþqþ  Ed Id þ Eq Iq ð13Þ
Once primary control achieves stable and reactive oscillations
Eþqþ Iþdþ  Eþdþ Iþqþ  Eq Id  Ed Iq ð14Þ exceed a certain range, secondary control is activated. Secondary
control corrects power quality by resetting current references, so
P 0 in (6) and Q 0 in (7) can be approximated as both active and reactive power oscillations can be limited simulta-
neously. The optimal current references of secondary control can
P0  Eþdþ Iþdþ þ Eþqþ Iþqþ ð15Þ
be obtained as follows:
8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Q 0  Eþqþ Iþdþ  Eþdþ Iþqþ ð16Þ >
> min F ¼ P2s2 þ P2c2 þ Q 2s2 þ Q 2c2
>
>
>
>
The positive-sequence currents can be written as >
< s:t: P0 ¼ Pref
" # " #1  ð21Þ
 > Q 0 ¼ Q ref
Iþdþ Eþdþ Eþqþ P0 >
>
>
 ð17Þ >
>
Iþqþ Eþqþ Eþdþ Q0 >
: P2s2 þP 2c2

Q 2s2 þQ 2c2
¼0
P2ref Q 2ref

Eq. (17) indicates that the positive-sequence currents can be where P ref and Q ref are power references of the DG. The aim of (21)
approximately regarded as constants if P0 and Q 0 are determined. is to reduce amplitudes of active and reactive power oscillations
2
Accordingly, 2ðE Iþ Þ in (12) can also be regarded as a constant simultaneously, and the optimized power oscillations can be pro-
and P v þ Q v reaches the minimum when I
2 2 
d and I q are set to zero.
portional to power references by the constraints. Power coefficients
2
Therefore, 2ðEþ I Þ in (12) constitutes additional increment of the in (21) are the functions of ½Iþ þ  
dþ ; I qþ ; I d ; I q , and the optimized cur-

power oscillations if the negative-sequence currents are injected. rent references can be obtained through optimization.
Power quality of a DG is not only affected by the oscillation The optimization that contains equality constraints can be
amplitudes but also by P0 and Q 0 . Thus, two indexes to measure transformed into unconstrained problem by Lagrange multiplier
power oscillations are defined as follows: method as follows:
346 P. Jin et al. / Applied Energy 194 (2017) 343–352

G ¼ F þ k1 ðP0  Pref Þ þ k2 ðQ 0 where Q i ðDk Þ is the objective function of new trust-region problem;
! Bk is the approximate matrix and a positive definite matrix.
P2s2 þ P2c2 Q 2s2 þ Q 2c2
 Q ref Þþk3  ð22Þ The radius of trust region r k is calculated in each iteration on the
P2ref Q 2ref basis of the solution in the last iteration. The updation is based on
where k1 , k2 and k3 are Lagrange multipliers, respectively. In order
rk as follows:
to solve unconstrained optimization, partial derivatives of the F i ðX k Þ  F i ðX k þ Dk Þ
objective function G is derived and optimization is equivalent to rk ¼ ð27Þ
Q i ð0Þ  Q i ðDk Þ
nonlinear equations as follows:
8 In the k iteration, Bk can be calculated as
@ðP2s2 þP 2c2 Þ @ðQ 2s2 þQ 2c2 Þ
>
> @F
þ k1 @I@Pþ0 þ k2 @I
@Q 0
þ þ 2
k3
 Qk23 ¼0 8
>
> @Iþ Pref @Iþ @Iþ
>
>
dþ dþ dþ dþ ref dþ
< B  Bk Dk DTk Bk þ yk yTk ; r P 0:01
>
> @F
>
> þ þ k1
@P 0 @Q 0
þ k2 @I þ þ 2
k3 @ðP 2s2 þP 2c2 Þ

2 2
k3 @ðQ s2 þQ c2 Þ
¼0 Bk ¼
k DTk Bk Dk yT D k k
ð28Þ
>
> @Iþ @Iþ @Iþ :
k
> @Iqþ
>
qþ qþ P ref qþ Q 2ref qþ
Bk ; rk < 0:01
>
>
>
> @P @Q k @ðP2s2 þP 2c2 Þ
 Qk23
@ðQ 2s2 þQ 2c2 Þ
> @I@Fd þ k1 @Id0 þ k2 @Id0 þ P23
< @I @I
¼0
ref d ref d where yk ¼ rFðxkþ1 Þ  rFðxk Þ, Bk is a positive definite matrix, and
@ðP 2s2 þP 2c2 Þ @ðQ 2s2 þQ 2c2 Þ
ð23Þ
>
> @F
þ k1 @I@P0 þ k2 @I
@Q 0 k3
 þ 2  Qk23 ¼0 B0 is an identity matrix.
>
> @I @I @I
>
>
q q q P ref q ref q
If rk is less than 0.01, this means that Q i does not have the ten-
>
>
>
> P0  P ref ¼ 0 dency toward F i at X k . In this case, this iteration is inappropriate;
>
>
>
> its results are not updated (i.e., X kþ1 ¼ X k ),and the radius of trust
>
> Q 0  Q ref ¼ 0
>
> region is reduced to half in the next iteration ðrkþ1 ¼ rk =2Þ. If the
>
: þP
P 2 2
Q 2 þQ 2
s2
P2
c2
 s2Q 2 c2 ¼ 0 solution of quadratic model comes toward the original problem,
ref ref
there are two different cases. If rk 2 ½0:01; 0:5, the results are
updated with the step ðX kþ1 ¼ X k þ Dk Þ, but the radius of trust
4.3. Quasi Newton-Trust Region (QNTR) method region is not changed ðr kþ1 ¼ r k Þ. On the contrary, if rk P 0:5, the
results are updated ðX kþ1 ¼ X k þ Dk Þ and the trust region is
The optimal references of positive and negative-sequence cur- enlarged ðr kþ1 ¼ minð2  r k ; rmax ÞÞ within the upper limit of rk [18].
rents can be obtained by solving nonlinear equations in (23) Therefore, the Quasi Newton-trust region method can be
through iterative approaches. Considering that Newton’s method described as follows:
is very sensitive to the value of initial approximation, it may obtain
an infeasible solution even with a very good initial guess. So the
Step #1: Given X 0 ; B0 , r max > 0, e ¼ 103 , r k0 2 ½0; rmax , k ¼ 0.
QNTR method is employed for doing so here. Further, the QNTR
Step #2: If krF i ðX k Þk2 < e, then stop;
method can be summarized as follows.
Step #3: Solve (26) using Dogleg method in [19] to get Dk , and
The set of nonlinear Eq. (23) can be regarded as the following
compute rk by (27).
optimization problem:
Step #4: Calculate approximate matrix Bkþ1 , and r kþ1 is revised
8
> F 1 ðXÞ ¼ f 1 ðXÞ  A1 ¼ 0 by the value range of rk .
>
>
>
> F 2 ðXÞ ¼ f 2 ðXÞ  A2 ¼ 0 Step #5: update k, go to Step #2.
>
>
>.
>
>
< ..
4.4. Start flow of hierarchical power oscillation control
min
>
> F i ðXÞ ¼ f i ðXÞ  Ai ¼ 0 ð24Þ
>
>
>
> .. When three-phase voltage of PCC is unbalance, primary control
>
>
>
> . adopting current references in (20) starts to reduce power oscilla-
:
F n ðXÞ ¼ f n ðXÞ  An ¼ 0 tions at the beginning. Once primary control achieves stable and
subject to X 2 Rn reactive ripple exceeds a certain range, secondary control starts.
The flowchart of secondary control is illustrated in Fig. 2.
Eq. (24) can be simplified to a quadratic model formed by the Taylor As illustrated in Fig. 2, the process determines the state of
series of the function F i , as follows: primary control firstly. If the difference between the current
(
S i ðDk Þ ¼ F i ðX k Þ þ DTk rF i ðX k Þ þ 12 DTk r2 F iðX k ÞDk reference and the actual measured value is less than 0.04 A, the
min ð25Þ primary control becomes stable. If primary control is in steady
s:t: kDk k2 < r k
state, amplitude of reactive oscillations and reactive oscillation
The Taylor series of the original function can be an suitable rep- rate are calculated. Once the amplitude of reactive power
resentation around the current iterate X k , and the trust region step oscillations is greater than 500 Var and the power oscillation ratio
Dk can minimize S i ðDk Þ within radius r k . is not less than 10%, secondary control that reduces active and
The QNTR method starts from an initial guess reactive power oscillations simultaneously will be started
X 0 ¼ ½Iþ þ    immediately after a time delay.
d0þ ; Iq0þ ; Id0 ; Iq0  to the optimal X . For the k-th iteration, a
P2s2 þP2c2 Q 2s2 þQ 2c2
step Dk is calculated to make F i ðX k þ Dk Þ close to F i ðX  Þ. If this If Qref is equal to 0, the constraint P 2ref
 Q 2ref
¼ 0 should be
attempt is reached, the solution is updated from ðX k Þ to ðX k þ Dk Þ. ignored in (21). Accordingly, the flowchart of secondary control
Because the Hessian matrix is very complex and it must be recal- is illustrated in Fig. 2.
culated in each iteration, second-order Taylor series r2 F i ðX k Þ is dif-
ficult to obtain. In order to maintain convergence of the QNTR 5. Control structure of DG
method and reduce the computational complexity of Hessian
matrix, an approximate matrix is constructed. 5.1. Dual current PI controller
The new sub-problem of QNTR method represented as
(
Q i ðDk Þ ¼ F i ðX k Þ þ DTk rF i ðX k Þ þ 12 DTk Bk Dk In order to control a voltage source converter, current refer-
min ð26Þ ences which are calculated by primary control and secondary con-
s:t: kDk k2 < ck
trol should be translated into voltage references as follows:
 P. Jin et al. / Applied Energy 194 (2017) 343–352 347

The current commands appear as DC in rotate frame, and there is

no need to build a tracking controller for AC signal. PWM signal
Calculate current reference is generated to control converter.
[Id , Iq , Id , Iq ]
5.2. Sampling of positive- and negative -sequence measurement
N
max Id Idm , Iq Iqm , Id Idm , Iq Iqm Positive- and negative-sequence currents should be separated
from original sampling current, and dual current controller can
Y work precisely. The monitored current is denoted in the stationary
Calculate amplitude of reactive oscillations frame as follows:
Qs22 +Qc22 Iabm ¼ ejxt Iþdqmþ þ ejxt Idqm ð30Þ
jxt
Y By multiplying Iabm by e , (30) can be rewritten in positive
Qs22 +Qc22 ? synchronous reference frame as follows:

Idqm ¼ Iþdqmþ þ ej2xt Idqm ð31Þ

N
Y In (31), the positive-sequence current is DC signal and negative-
Qref =0 sequence current appears as double working frequency AC in pos-
itive reference frame. In the same way, the negative-sequence
N appears as DC in the negative reference frame, whereas the
Y positive-sequence appears as double working frequency AC. Notch
q* 10% filters are adopted to separate positive- and negative -sequence
currents and they can remove a narrow band of frequencies from
N the signal path. The transfer function of notch filters can be
described as
After time delay, call secondary control
s2 þ x21
GðsÞ ¼ ð32Þ
s2 þ 2nx1 s þ x21
where x1 ¼ 200p, the damping ratio n ¼ 0:25. With the increase of
Fig. 2. Flowchart of secondary control. digital control, the digital notch filters adopting sampled data the-
ory should be design to meet the requirement. The bilinear trans-
8  form is widely used for translating continuous domain transfer
V dþ ¼ Eþdþ  ðK p þ Ksi Þðidþ  idmþ Þ þ xLiqþ
þ þ þ
>
> function into z-domain transfer function. The desired transfer func-
>
>
>
< V  ¼ Eþ  ðK p þ K i Þðiþ  iþ Þ  xLiþ tion of notch filters in z-domain can be expressed as follows:
qþ qþ s qþ qmþ dþ
ð29Þ
>
>
> V d ¼ Ed  ðK p þ Ksi Þðid  idm Þ  xLiq
  
1  ð2cosxTÞz1 þ z2
>
> GðzÞ ¼ ð33Þ
:  1 þ n sin xT  ð2cosxTÞz1 þ ð1  n sin xTÞz2
V q ¼ Eq  ðK p þ Ksi Þðiq  iqm Þ þ xLid
  

where the sampling frequency T ¼ 10 kHz. According to transfer

where K p is the proportional gain, K i is the integral gain. The sub- function, a recursive difference equation can be expressed as
script m represents the measured value. The terms xLiqþ , xLidþ ,
þ þ
follows:
xLiq , þxLid are inserted to decouple dq axes dynamics.
 
E0 ðqÞ ¼ ½C 1 Ei ðqÞ þ C 2 Ei ðq  1Þ þ C 3 Ei ðq  2Þ  C 5 E0 ðq  1Þ
As illustrated in Fig. 3, the dual current controller adjusts
 C 6 E0 ðq  2Þ=C 4 ð34Þ
positive- and negative-sequence electrical quantity separately.

Ed
Id + -
Ed Eq + + Sa
PI
-
+ +dq + S
I dm ωL +
P Sb
I qm 0L
ωL + W
Pref Current - - abc M Sc
Iq + -
reference of PI + +
Eq +
primary
Qref control and + Ed
secondary PI +
Id - - -dq
control -
I dm ωL

ωL
I qm
- + abc
Ed Eq PI
+ -
Iq +
Eq

Fig. 3. Control structure of the dual current controller.

348 P. Jin et al. / Applied Energy 194 (2017) 343–352

where Ei is the input, E0 is the output, and q is sampling numbers. The frequency and phase responses of notch filters are shown in
Coefficients in (34) can be expressed as: C 1 ¼ C 3 ¼ 1, Fig. 4. Obviously, the theoretical attenuation at the notch fre-
C 2 ¼ C 5 ¼ 2 cos xT, C 4 ¼ 1 þ n sin xT, C 6 ¼ 1  n sin xT. quency is large, round off within discretization process weakens
the actual realization of this amount of attenuation, but filtering
capability of digital notch filters is maintained.
0
The amplitude of negative-sequence electric quantity is much
G(z) less than the magnitude of the positive-sequence electric quantity.
Magnitude [dB]

When the positive-sequence current Iþ dqmþ is measured by notch

-50
filters, the negative-sequence current I dqm can be separated, as
G(s) illustrated in Fig. 5.
-100
5.3. Improved phase locked loop (PLL) technology

-150 In order to ensure the accuracy of phase measurement, phase

0 100 200 300 400 locked loop is widely adopted in tracking phase. 100 Hz AC under
Frequency [Hz]
synchronous reference frame plays negative effects when three-
(a) Frequency response of notch filters
phase voltage is in an unbalanced state. If measurement of
90 100 Hz AC voltage delays 1/4 frequency cycle and combines with
60
the original value, negative effects of 100 Hz AC can be eliminated.
G(s)
As illustrated in Fig. 6, no filters are adopted in this structure of PLL
G(z)
and phase lags which produces by inertia of filters does not exist.
Phase [deg.]

30

0
6. Case studies
-30

-60 In order to verify the performance of the proposed scheme, sim-

ulation studies are carried out in MATLAB/SIMULINK environment
-90
0 100 200 300 400 with a simulation time step of 2 ls. Parameters of the DG are given
Frequency [Hz] in Table 1, the power circuit of the converter and control structure
(b) Phase response of notch filters are depicted in Figs. 1 and 3, respectively. Unbalanced three-phase
input voltage at PCC ½Ea ; Eb ; Ec  ¼ ½341 sinðxt þ 90 Þ; 291 sin
Fig. 4. Bode diagram of the notch filters. ðxt  30 Þ; 311 sinðxt þ 210 Þis illustrated in Fig. 7.

I am I dm Notch - + I dm
Filter
I bm abc I dm

I cm I qm
e2 j t

dq Notch
Filter I qm
I qm

Fig. 5. Separation method for unbalanced current measurement.

Ea
+ Ed
Eb S abc Delay Ts/4
+
1/2

& Eq
Ec H + +
dq Delay Ts/4 1/2

1 +
PI
S
-1 + 0 Eq _ ref 0

+ Ed
abc Delay Ts/4
+
1/2

+ + Eq
dq Delay Ts/4 1/2

Fig. 6. Structure diagram of improved PLL.

 P. Jin et al. / Applied Energy 194 (2017) 343–352 349

Table 1 -10
Parameters of the DG.

Parameters Value -12

Fundamental frequency fs = 50 Hz
-14
Switching frequency fsw = 10 kHz

id [A]
DC bus voltage Udc = 800 V
Capacitance of DC bus C1 = 8800 lF -16
Filter inductance L = 5 mH
Rated power S = 12 kVA -18
Load parameter Zl = 10 X
-20
0 0.3 0.6 0.9 1.2 1.5
Time [S]
400 (a) Unfiltered d axis current
16
200
14
Uabc [V]

Iq [A]
12

-200
10

-400
0 0.04 0.08 0.12 0.16 0.2 8
Time [S] 0 0.3 0.6 0.9 1.2 1.5
Time [S]
Fig. 7. Unbalance voltage at PCC. (b) Unfiltered q axis current
Fig. 9. Unfiltered dq current.

8000
special reciprocal relationship between active and reactive power
oscillations in (12).
6000
Secondary control is not activated until t ¼ 0:77 s. As shown in
P [W]

Fig. 8(b), the amplitudes of active and reactive power oscillations

4000
are less than 450 VA though the active power oscillations are
amplified. Furthermore, the amplitudes of active and reactive
2000 power oscillations are proportional to respective power references
and the power oscillation rate is limited to 5.6%. In addition, the
0
0 0.3 0.6 0.9 1.2
injection of the negative-sequence current is significantly reduced
Time [s] under secondary control as illustrated in Fig. 9(a).
Fig. 9 also shows unfiltered dq current of the DG. Amplitude of
(a) Dynamic active power response of the DG
100 Hz AC component is large when secondary control is not acti-
8000 vated. After secondary control starts, AC component is greatly
weakened. After separation by notch filters, the positive- and
negative-sequence dq currents in Fig. 10 reflect characteristics of
6000
power oscillation amplitudes. The positive- sequence current is
Q [Var]

mainly proportional to the average output active power and reac-

4000
2000
0
0 0.3 0.6 0.9
Time [s]
(b) Dynamic reactive power response of the DG
Fig. 8. Dynamic power response from primary control to secondary control.
The process for stabilizing the active and reactive power oscilla-
tions is illustrated in Fig. 8. Active and reactive power references of
the DG are P ref ¼ 8 kW and Q ref ¼ 6 kVar. At the beginning of the
simulation, primary control of the DG starts to eliminate active
power oscillations. As illustrated in Fig. 8(a), the active power
oscillations are effectively suppressed by primary control, but the
amplitude of reactive power oscillations is up to 780 Var. The large
amplitude oscillations of reactive power can be explained by the
tive power, so it changes slightly when the control mode switches.
The variation of negative-sequence current mainly determines the
change of power oscillation amplitudes.
The simulations results show that primary control suppresses
1.2
active power oscillations at the cost of reactive power oscillations
and negative-sequence current. In contrast, secondary control
simultaneously reduces active and reactive power oscillations.
The voltage unbalance factor eu [20] is defined as
U
eu ¼ ð35Þ
Uþ
where U þ and U  are respectively the positive- and negative-
sequence voltages, and they can be obtained by symmetrical com-
ponents calculation.
In the previous case, eu is set to 4.6%. Increase the voltage unbal-
ance factor to 9.5% at PCC and repeat the simulations. Active and
reactive power references of the DG are Pref ¼ 5 kW and
Q ref ¼ 3 kVar. Due to the increase of voltage unbalance factor,
350 P. Jin et al. / Applied Energy 194 (2017) 343–352

2
Negative sequecne current [A]

6000
Iq -
1.5

1
4000

P [W]
0.5

0
Id- 2000
-0.5

-1
0 0.3 0.6 0.9 1.2
Time [s] 0
0 0.3 0.6 0.9 1.2 1.5
(a) Negative-sequence dq currents Time [S]
(a) Dynamic active power response of the DG
Positive sequence current [A]

20
15 5000
10 Iq+
5 4000
0
-5 3000

Q [var]
-10 Id+
-15 2000
-20
-25
0 0.3 0.6 0.9 1.2
1000

Time [s]
0
(b) Positive-sequence dq currents 0 0.3 0.6 0.9 1.2 1.5
Time [S]
Fig. 10. Positive- and negative-sequence dq currents. (b) Dynamic reactive power response of the DG
Fig. 12. Dynamic power response under the scheme introduced in [21].

6000
15
P [W]

4000

5
Iabc [A]

2000

-5
0
0 0.3 0.6 0.9 1.2 1.5
Time [S]
(a) Dynamic active power response of the DG -15
5000 1.4 1.42 1.44 1.46 1.48 1.5
Time [S]
4000 (a) Dynamic current response of the DG under secondary control

3000
Q [var]

15

2000

1000 5
Iabc [A]

0
0 0.3 0.6 0.9 1.2 1.5
Time [S]
-5
(b) Dynamic reactive power response of the DG

Fig. 11. Dynamic power response under high voltage unbalance factor.
-15
1.4 1.42 1.44 1.46 1.48 1.5
Time [S]
reactive power oscillation rate is close to 31.7% under primary con-
(b) Dynamic current response of the DG under the scheme introduced in [21]
trol. Once secondary control is activated, active and reactive power
oscillation rates are limited to a reasonable level at the cost of a Fig. 13. Simulation waves of three–phase current response.
slight increase in active power oscillation rate (see Fig. 11).
Fig. 12 shows power responses of negative-sequence distur-
bance compensation method introduced in [21]. The active power much severer. Therefore, the overall level of power oscillations in
oscillations caused by this method is slightly less than the scheme scheme introduced in [21] are greater than that in the proposed
proposed in this paper, but the reactive power oscillations are scheme. As shown in Fig. 13, the output current of proposed
 P. Jin et al. / Applied Energy 194 (2017) 343–352 351

secondary control is better than scheme introduced in [21]. Fur- 6000

thermore, the current unbalance factor (eI ¼ I =Iþ ) of secondary

control is approximately 2.6% compared with 12.40% of the scheme
in [21]. 4000
From the above results, secondary control causes active power

P [W]
oscillations, and active power oscillations lead to voltage ripple
in DC link bus of the converter. The relationship between the DC
2000
voltage ripple and the active power oscillations can be expressed
as [22]:
Pv
V ripple ¼ ð36Þ 0
2C 1 U dc x 0 0.4 0.8 1.2 1.6
Time [S]
where C 1 is equivalent capacitance of DC-link bus. The amplitudes (a) Dynamic active power response of the DG
of active power oscillations which are caused by secondary control
in above two cases are 446 W and 600 W, respectively. By using
(36), DC voltage ripple is about 0:1 V and 0:136 V respectively.
Therefore, the voltage ripple has slight influence on the DC-link bus. 4000
Fig. 14 shows the DC bus voltage response under voltage unbal-
ance factor 4.6% and 23%. The DC voltage undergoes the process 3000

Q [var]
2000

805
1000
803
DC bus voltage [V]

0
0 0.4 0.8 1.2 1.6
801 Time [S]
(b) Dynamic reactive power response of the DG
799
Fig. 16. Power response of the DG when voltage of PCC undergoes a transient dip.
797

from a transient dip to recovery when the DG starts. The amplitude

795
0 0.5 1 1.5 of voltage ripple can be ignored under secondary control in case of
Time [S]
low voltage unbalance factor. Furthermore, the amplitude of volt-
(a) DC bus voltage under low voltage unbalance factor age ripple is relatively smaller under 23% voltage unbalance factor.
805
Load and power variations caused by non-predictable energy
often exist in the terminal of distribution network, and voltage of
803 PCC may change suddenly with shown in Fig. 15. Fig. 16 shows
DC bus voltage [V]

the power response of the DG when voltage of PCC undergoes a

801 transient dip. The DG worked in the state of stable secondary con-
trol before voltage of PCC changed. Then, unbalanced three-phase
799 input voltage at PCC undergoes a transient dip when t is equal to
0.3 s, and active and reactive powers decrease obviously due to
797 reduction of voltage. Therefore, the controller of the DG reset cur-
rent references to satisfy fix power control and switch to primary
795 control mode. At the same time, active power oscillations are sup-
0 0.5 1 1.5 pressed and reactive power oscillations are enlarged. Until t is
Time [S]
equal to 1.11 s, the DG undergo smooth transition to secondary
(b) DC bus voltage under 23% voltage unbalance factor
control again.
Fig. 14. DC bus voltage of the DG converter.

7. Conclusion
400
This paper focuses on power oscillations of DGs under unbal-
300
anced voltage. The analytic relationship between active and reac-
200
tive power oscillations is revealed, and a hierarchical control
100 scheme is proposed to control power oscillations of DGs. Based
Uabc [V]

0 on the optimal model, the amplitudes of the power oscillations

can be precisely controlled by the injected currents whose refer-
-100
ences are calculated by using QNTR method. Simulation results
-200
show that both active and reactive power oscillations of the DG
-300 can be accurately limited and the injected negative-sequence cur-
-400 rents are reduced effectively. Therefore, the proposal befits power
0.27 0.28 0.29 0.3 0.31 0.32 0.33
Time [S]
oscillations reduction for DGs under unbalanced working condi-
tions. In addition, secondary control can achieve other control tar-
Fig. 15. Abrupt changes of voltage in PCC. gets by resetting the objective function and constraints of
352 P. Jin et al. / Applied Energy 194 (2017) 343–352
optimization in (21) according to different operational require-
ments, and it can easily be extended to multi-DGs in the distribu-
tion network.
It should also be pointed out that the proposed method is suit-
able for grid-connected converter connecting with a variety of
energy forms, such as batteries, photovoltaic panels and fuel cells.
Therefore, it has a very important significance on the energy sys-
tem of urban development.
Acknowledgements
This work was supported in part by the Science and Technology
Development Plan of State Grid Corporation of China under Grant
No. NEDU20150002.
References
[1] Wang J, Conejo AJ, Wang C, Yan J. Smart grids, renewable energy integration,
and climate change mitigation–future electric energy systems. Appl Energy
2012;96():1–3.
[2] Roskilly AP, Taylor PC, Yan J. Energy storage systems for a low carbon future–in
need of an integrated approach. Appl Energy 2015;137:463–6.
[3] Esmaili M, Firozjaee EC, Shayanfar HA. Optimal placement of distributed
generations considering voltage stability and power losses with observing
voltage-related constraints. Appl Energy 2014;113:1252–60.
[4] Blaabjerg F, Chen Z, Kjaer SB. Power electronics as efficient interface in
dispersed power generation systems. IEEE Trans Power Electron 2004;19
(5):1184–94.
[5] Chen Z, Guerrero JM, Blaabjerg F. A review of the state of the art of power
electronics for wind turbines. IEEE Trans Power Electron 2009;24(8):1859–75.
[6] Ashabani S, Mohamed Y. A flexible control strategy for grid connected and
islanded microgrids with enhanced stability using nonlinear microgrid
stabilizer. IEEE Trans Smart Grid 2012;3(3):1291–301.
[7] Pruitt KA, Braun RJ, Newman AM. Evaluating shortfalls in mixed-integer
programming approaches for the optimal design and dispatch of distributed
generation systems. Appl Energy 2013;102:386–98.
[8] Mena R, Hennebel M, Li YF, Zio E. Self-adaptable hierarchical clustering
analysis and differential evolution for optimal integration of renewable
distributed generation. Appl Energy 2014;133:388–402.
[9] Moran L, Ziogas P, Joos G. Design aspects of synchronous PWM rectifier—
Inverter systems under unbalanced input voltage conditions. IEEE Trans Ind
Appl 1992;28:1286–93.
[10] Sannino A, Bollen M, Svensson J. Voltage tolerance testing of three-phase
voltage source converters. IEEE Trans Power Del Apr. 2005;20(2):1633–9.
[11] Song HS, Nam K. Dual current control scheme for PWM converter under
unbalanced input voltage conditions. IEEE Trans Ind Electron 1999;46
(5):953–9.
[12] SaccomandoG, Svensson J. Transient operation of grid-connected voltage
source converter under unbalanced voltage conditions. In: Proc IEEE Ind Appl
Conf; 2001. p. 2419–24.
[13] Yuan X, Allmeling J, Merk W, Stemmler H. Stationary frame generalized
integrators for current control of active power filters with zero steady state
error for current harmonics of concern under unbalanced and distorted
operation conditions. IEEE Trans Ind Appl 2002;38(2):523–32.
[14] Zmood DN, Holmes DG. Stationary frame current regulation of PWM inverters
with zero steady-state error. IEEE Trans Power Electron May 2003;18
(3):814–22.
[15] Etxeberria-Otadui I, Viscarret U, Caballero M, Rufer A, Bacha S. New optimized
PWM VSC control structures and strategies under unbalanced voltage
transients. IEEE Trans Ind Electron Oct. 2007;54(5):2902–14.
[16] Rodriguez P, Timbus AV, Teodorescu R, Liserre M, Blaabjerg F. Flexible active
power control of distributed power generation systems during grid faults. IEEE
Trans Ind Electron 2007;54(5):2583–92.
[17] Wang F, Duarte J, Hendrix M. Design and analysis of active power control
strategies for distributed generation inverters under unbalanced grid faults.
IET Gener, Trans Distrib 2010;4(8):905–16.
[18] Nocedal J, Wright SJ. Trust-region methods, numerical optimization. second
ed. New York: Springer; 1999. p. 66–100 [ch. 4].
[19] Conn AR, Gould NIM, Toint PL. Global convergence of the basic algorithm. In:
Trust-region methods. Philadelphia, PA: SIAM; 2000. p. 115–67 [ch. 6].
[20] Jouanne AV, Banerjee B. Assessment of voltage unbalance. IEEE Trans Power
Del Oct. 2001;16(4):782–90.
[21] Rioual P, Pouliquen H, Louis JP. Regulation of a PWM rectifier in unbalanced
network state using a generalized model. IEEE Trans Power Electron 1996;11
(3):495–502.
[22] Yazdani A, Iravani R. A unified dynamic model and control for the voltage
sourced converter under unbalanced grid conditions. IEEE Trans Power Del Jul.
2006;21(3):1620–9.