Received September 3, 2020, accepted September 11, 2020, date of publication September 15, 2020,

date of current version September 25, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.3024178

A Novel Fundamental and Harmonics Detection

Method Based on State-Space Model for Power
Electronics System
DONGDONG CHEN AND LONG XIAO
School of Electronic and Electrical Engineering, Minnan University of Science and Technology, Quanzhou 362700, China
Corresponding author: Long Xiao ([email protected])
This work was supported in part by the Natural Science Foundation of Liaoning Province under Grant 2019KF2307 and Grant
2020KF2105, and in part by the Science and Technology Project of Quanzhou City under Grant 2020C011R and Grant 2019CT003.

ABSTRACT Harmonic current estimation is the key technology of power electronics systems to generate a
harmonics reference current for harmonic control. In this paper, a novel fundamental and harmonic current
detection method based on the state-space model (SSM) is introduced. The proposed SSM harmonics
detection method has the ability to detect the positive sequence component (PSC), negative sequence
component (NSC), and zero-sequence component (ZSC) of both fundamental and each order harmonics
current with fast response and high accuracy separately, including the fundamental component, which is
appealing for harmonics control in power electronic applications, such as active power filter (APF), unified
power quality conditioner (UPQC), virtual synchronous generator (VSG), and so on. The correctness of the
proposed SSM based harmonics detection method has been verified by simulation and experiment result,
and it shows that the PSC, NSC, and ZCS of the grid fundamental and high order harmonic current can
be detected quickly and accurately, which is important for harmonics analysis and control in the power
electronics system.

INDEX TERMS Harmonics detection, state-space model (SSM), positive negative and zero sequence
separation, power quality control.

I. INTRODUCTION harmonic source needs to detect harmonics to control their

With the development of power electronics devices and harmonics output, to avoid a lot of harmonics injected into the
nonlinear load, the power quality problem of the power power grid. Though commercialized of some harmonic sup-
grid has been paid more and more attention [1]–[4]. The pression power electronic converters, such as APF, UPQC,
high harmonic content of the power grid will bring adverse and so on [11], the research of finding an effective real-time
effects such as electromagnetic interference, cable overheat- harmonic detection method is still in progress.
ing, and low power factor, which will seriously affect the In previous studies, many methods for detecting the
safe operation of the power grid. Since the 1980s, researchers power grid harmonics or the fundamental have been pro-
began to pay attention to the problem of harmonics. By far, posed, which can be divided into time-domain method,
some active harmonics elimination methods have been pro- frequency-domain method, and intelligent algorithm based
posed to address this problem, such as active power fil- method. First, the common time-domain methods pro-
ters (APF) and unified power quality conditioner (UPQC), posed by researchers include instantaneous reactive power
etc. [5]–[10]. Among all the active harmonics elimination theory (IRPT) method [10], [11], synchronous reference
methods, harmonic detection is one of the most critical tech- frame (SRF) method [12], the second-order generalized inte-
nology for harmonic governing and control, because con- gral (SOGI) method [13], the cascaded delayed signal can-
trollers of power inverter containing a fast harmonic current cellation (CDSC) techniques [14] and the Kalman filter
detection method will lead a fast harmonics response for the method [15], etc. Traditional IRPT and SRF play an important
power inverter. Also, power electronic equipment as the main role in conventional harmonics detection for APF and UPQC,
and in recent years, IRPT and SRF are improved for more
The associate editor coordinating the review of this manuscript and complex applications with the better performance [16], [17].
approving it for publication was Zhehan Yi . As to the SOGI method for harmonics detection, a lot of
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
170002 VOLUME 8, 2020
D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

second-order filters are adopted, and the number of the fil- In this paper, a novel algorithm based on SSM for dynami-
ter is equal to the number of the harmonics, which makes cally detecting the polluted power gird harmonic components
SOGI method too complex to be implemented, and not engi- is proposed. The convergence of amplitude and phase errors
neering applicable. The CDSC harmonics detection method of SSM harmonic detection is analyzed in detail, and the
is based on a series of DSC operators, thus, a group of range of parameters to ensure the stability and controllability
harmonics can be detected at the same time, and in recent of the proposed state-space model are given by theoretical
years, the researchers have developed some algorithms to analysis. Simulation and experiments show that this method
improve the performance of CDSC [14], [18]. But the CDSC has a good detection effect.
method needs too many DSC operators, which makes the In general, the proposed SSM harmonics detection method
system complex and increases the system computational bur- has the following practical features:
den [15], [18]. In a word, Though effective for harmonics 1) The proposed SSM harmonic detection method can
detection, both SOGI and CDSC methods have disadvantages separately detect out PSC, NSC, and ZSC of each harmonic
of structural complexity and computation intensity. [19], [20]. orders from distorted power grid signal;
The harmonic detection method base on Kalman filter is 2) The method proposed in this paper can detect the har-
developed in recent years, but the Kalman filter methods monics in real-time, so it is suitable to suppress the har-
need prior information of the detected harmonics for better monic amplification effect of the power electronic system in
accuracy and faster response [21]–[23]. dynamic response.
As to frequency-domain methods, discrete Fourier trans- 3) The SSM method is computed in time-domain, and there
form (DFT) and its upgraded methods fast Fourier trans- is not transform computation burden for the system.
form (FFT) are commonly used [24]. In recent researches, For convenience, some common matrices are defined in
many improved DFT methods are studied for more stable, this paper. In is defined for the n-order unit matrix. On is the
faster response, and fewer leakage effects, to achieve a bet- n-order square matrix and Or×s is the matrix with dimension
ter harmonics detection performance. The detection method r × s. Elements of On and Om×n are all zero. En is the n-order
based wavelet-function is another frequency-domain method, square matrix and Er×s is the matrix with dimension r × s.
but it is limited by the large computation burden [25]. Elements of En and Er×s are all 1. Subscripts variable h, s,
Regardless of the time domain method or the frequency and r are all natural numbers. This paper adopts f = 50Hz as
domain method, the harmonic detection method needs to the fundamental frequency of China and most other national
have a certain frequency adaptability to solve the problem of power grids, and the sample frequency of the controller is
fundamental frequency offset. In this regard, the researchers 5kHz in this paper.
also proposed the methods, such as phase-locked loop et.al, The paper is organized as follows. Section II describes
to make the harmonic detection method acquires frequency the proposed harmonic detection method, and the state-space
adaptability [14], [18], [26]. model is established in detail. The theoretical analysis of
With the development of artificial intelligence technol- the proposed SSM harmonics detection method is presented
ogy in recent years, intelligent algorithms have also been in Section III, which include controllability, stability, and
applied in the field of harmonic detection, such as neural error convergence. Experimental results of the proposed har-
networks [27], [28] and particle swarm optimization algo- monic detection method are reported in Section IV. Finally,
rithm [29]. Paper [30] proposed a harmonic current detection Section V concludes the paper.
method based on state observer, but the results of the harmon-
ics obtained by this method have a 90-degree phase difference II. STATE SPACE MODEL CONSTRUCTION FOR
with the real harmonic signals, so it is not applicable to most HARMONICS DETECTION
applications. These intelligent algorithms are not practical for Three-phase hth harmonic vector (current or voltage) ihj_abc
industries because they need a lot of sample training, so they (j = p, n, z) could be expressed as Equation (1), and Ihj_m is
still need to be further explored. the amplitude of the corresponding sequence component.
However, the harmonics detection methods mentioned 
Ihp_m sin(hωN t + ϕhp )
 
above cannot separately detect out the positive sequence 

components(PSC), negative sequence component(NSC), and 
3 + ϕhp ) 
ihp_abc =  Ihp_m sin(hωN t − 2π

  

zero sequence component(ZSC) of each harmonic from the 
3 + ϕhp ) 

I sin(hωN t + 2π

 hp_m

distorted power grid, which limits their application in some 

Ihn_m sin(hωN t + ϕhn )


situations. It is well-known that the three-phase four-wire 


hn_abc =  Ihn_m sin(hωN t + 3 + ϕhn ) 

i 2π
system is common in the power electronics system, and there
 
(1)
Ihn_m sin(hωN t − 2π3 +ϕhn )
exist zero-sequence components in the four-wire system. 

Moreover, an unbalanced grid or load can also bring an 




Ihz_m sin(hωN t + ϕhz )
unbalanced current of negative-sequence and zero-sequence. 
ihz_abc =  Ihz_m sin(hωN t + ϕhz ) 

  

Therefore, a harmonic detection method that can individually 

Ihz_m sin(hωN t + ϕhz )


detect the PSC, NSC, and ZSC of each harmonic is needed to 



meet the requirement of flexible harmonic control. 

VOLUME 8, 2020 170003

 D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

The transformation results of Equation (1) in the αβγ coor- It can be seen from Equation (3) that the PSC ihp−αβγ
dinate system are shown in Equation (3), and the Clarke and the NSC ihn−αβγ on γ -axis are zero, while the ZSC
transformation matrix Tabc−αβγ is shown in Equation (2). ihz−αβγ on α-axis and β-axis is zero. Their derivatives are the
  Equation (5), in which the derivative of ihz−αβγ is nonlinear
1 √ −1/2 −1/2
√ with ihz−αβγ , Because it contains the nonlinear part of the
2
T abc−αβγ = 0 3/2 − 3/2  (2) trigonometric function as the third branch of Equation (5),
3
1/2 1/2 1/2 and it is impossible to establish a state-space model which
Ihp_m sin(hωN t + ϕhp )
  
 is linear time-invariant (LTI), according to the Equation (5).



 ihp_αβγ =  −Ihp_m cos(hωN t + ϕhp ) 
  To meet the requirement of the LTI system, this paper defines
a generalized zero-order vector vhz−αβ in the αβ frame sys-




 0

 tem. The definition rules are shown as Equation (6).
−Ihn_m sin(hωN t + ϕhn )
  

  √ 
π

ihn_αβγ =  −Ihn_m cos(hωN t + ϕhn )  (3) 2
Ihz_m sin(hωt + ϕhz + ) 
   
v
vhz_αβ = hz_α =  2 4 



0 √
vhz_β π 

  2
Ihz_m cos(hωt + ϕhz + )
  

 0 −
2 4




ihz_αβγ =  0 (6)

  

 
Ihz_m sin(hωN t + ϕhz )

In Equation.(5), the γ -axis of PSC and NSC are both 0,
Therefore, iαβγ can be expressed as Equation (4) and so they can be characterized by the α-axis and β-axis. ZSC
Equation (5) in αβγ coordinate. only has γ -axis non-zero and there is a non-linear relation-
ship. Therefore, the αβ axis in the generalized ZSC vector
iαβγ = T abc−αβγ iabc of Equation.(6) can be used to characterize the actual ZSC.
N Obviously, the generalized ZSC definition of Equation.(6) is
linear.
X
= (T abc−αβγ ihp_abc
h=1
According to Equation (6), the relationship between
generalized zero-sequence vector vhz−αβ and the actual
+T abc−αβγ ihn_abc + T abc−αβγ ihz_abc )
zero-sequence component ihz−αβγ on the γ -axis can be
N
X obtained as shown in Equation (7).
= (ihp_αβγ + ihn_αβγ + ihz_αβγ ) (4)  
vhz_α
h=1 ihz_γ = [ 1 1 ] = E1×2 vhz_αβ (7)
vhz_β
hωN Ihp_m cos(hωN t + ϕhp )
  
Therefore, by replacing the nonlinear actual ZSC vector


 ·
hωN Ihp_m sin(hωN t + ϕhp ) 
  
i = with a linear generalized ZSC vector, the PSC, NSC, and ZSC



hp_αβγ
  
in the system can be made linear, so that an LTI system can


 0
be constructed on this basis.



  


 0 −hωN 0 By combining Equation (6) and Equation (7), the positive,

negative, and zero sequence vectors are transformed into αβ
  
0 0  ihp_αβγ


= hωN
coordinate system, which can get rid of the dependence of

  

0 0 0

state vectors on the γ -axis and facilitate the establishment of




−hωN Ihn_m cos(hωN t + ϕhn ) the appropriate state-space model.

  


The h-th sub-state vector xh contains the PSC, NSC, and

 ·
hωN Ihn_m sin(hωN t + ϕhn ) 
  
i =

ZSC of h-th harmonics. The h sub-state vector xh (n) in the



hn_αβγ  
digital control system discrete domain is defined as Equa-

 0
  (5) tion (8).


 0 hωN 0
  h iT
xh (n) = xhp_αβ (n)T xhn_αβ (n)T xhz_αβ (n)T
  
−hωN 0 0  ihn_αβγ

=
 


   


  " #
i


 0 0 0 
 hp_α

 
 xhp_αβ (n) =
i
   
" hp_β #
 


 0 

 ·
   ihn_α (8)

 i = 0  
 xhn_αβ (n) =
i
 
hz_αβγ  
" hn_β #

 
ϕ

cos(hω + )

hω I t
 
 N hz_m N hz 
vhz_α
 

 
 xhz_αβ (n) =
   
0
 
vhz_β



  
0

= 
 


   
i hz_γ
  Therefore, at the moment (n +1), the PSC of the sub-state
 hωN Ihz_m cos arcsin Ihz_m vector xhp _αβ (n+1) of the corresponding detected h-th order

170004 VOLUME 8, 2020

D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

harmonic component can be calculated from xhp _αβ (n) as And Equation (13), as shown at the bottom of the next
shown in Equation (9), where Ts is the sampling time and page, is available with the combination of Equation (3), Equa-
Rothp is the corresponding rotation transformation matrix. tion (4), and Equation (6).
In which, the dimensions of the matrix F, G, and H are
xhp_αβ (n + 1) 2 × 6 N , 1 × 6N, and 3 × 6 N respectively. And matrix F and
Ihp_m sin((n + 1)Ts hω0 + ϕhp )
 
G with 6N columns are shown in Equation(14).
=
−Ihp_m cos((n + 1)Ts hω0 + ϕhp )  
N

Ihp_m sin((nTs hω0 + ϕhp ) + nTs hω0 )
  
 z }| {
    
=

−Ihp_m cos((nTs hω0 + ϕhp ) + nTs hω0 ) F =  I 2 I 2 O2 · · · I 2 I 2 O2 



  
  

cos(Ts hω0 ) −sin(Ts hω0 )
=  
sin(Ts hω0 ) cos(Ts hω0 ) 


Ihp_m sin(nTs hω0 + ϕhp ) G =  (O1×2 O1×2 E1×2 ) · · · (O1×2 O1×2 E1×2 ) 
  
  

×

−Ihp_m cos(nTs hω0 + ϕhp )

 | {z }

N
= Rot hp xhp_αβ (n) (9) (14)
Similarly, for the NSC and the ZSC, Equation (10) as the Then, substitute Equation (14) into Equation (10) and Equa-
following can be established. tion (11), Equation (15) and Equation (16) can be get, where
 " # λ is a constant coefficient, and λ ∈(0,1).
 xhn_αβ (n + 1) = cos(hω 0 T s ) sin(hω 0 T s )
xhn_αβ (n)

h λFx(n) +
i λFx(n)



 −sin(hω0 Ts ) cos(hω0 Ts ) 
 xhp_αβ (n + 1) = Rot hp xhp_αβ (n) −

 = Rot hp xhp_αβ (n) − λFx(n) + λ I 2 O2×1 iαβγ (n)

 

" = Rot hn xhn_αβ (n)


i hn xhn_αβ (n) − λFx(n)

#
 cos(hω 0 T s ) −sin(hω 0 T s ) 
 xhn_αβ (n h+ 1) = Rot
xhz_αβ (n + 1) = xhz_αβ (n)

 




 sin(hω 0 T s ) cos(hω 0 T s )
 +λ I 2 O2×1 iαβγ (n)


= Rot hz xhz_αβ (n) (15)
(10) xhz_αβ (n + 1)

  
G G
Therefore, combining Equation (8), Equation (9) and = Rot hz xhz_αβ (n) − λ x(n) + λ x(n)
Equation (10), the transformation matrix Roth of sub-state G G
 
vector xh (n) to xh (n+1) can be obtained. Roth is a six-order G
= Rot hz xhz_αβ (n) − λ x(n) + λ O2 E2×1 iαβγ (n)
 
square matrix, as shown in Equation (11). G
   (16)
Rot hp xhp_αβ (n)
xh (n + 1) =  Rot hn   xhn_αβ (n)  = Rot h xh (n) On the basis of the above analysis, the system state-space
Rot hz xhz_αβ (n) model for N harmonics detection is re-established as
(11) Equation (17).
x(n + 1)
To detect the PSC, NSC, and ZSC of 1st ∼ N th harmonics,    
the whole state-space model should contain N sub-state vec- 
 Rot 1 K  
Rot 2  K 

   
tors, the state space defined as x(n), which containing N sub-   
..  .. 
  
=  .  − λ . 
 
state vectors. Thus, the entire state-space model of entire N
   
detected harmonics is shown in Equation (12), in which, the 

  Rot N −1   K  

 
Rot is a square matrix and its order is 6N . 
Rot N K

 
Bu
 
Rot 1
 Rot 2   Bu 
×x(n) + λ  ...  iαβγ (n)
 
..
   
x(n + 1) =  .  x(n)
 
   
 Rot N −1   Bu 
Rot N Bu
= Rot · x(n) (12) = Ax(n) + Biαβγ (n) (17)

According to the control theory, the state-space model where matrix K and Bu are shown as Equation (18).
 
variables x(n) shown in Equation (12) are not controlled F  
I 2 O2×1
by the input vector iαβγ (n), that is, the state-space model F
K =  G  Bu = I 2 O2×1 (18)
  
is uncontrollable. Therefore, it is necessary to improve the
O2 E2×1
state-space model shown in Equation (12). G

VOLUME 8, 2020 170005

 D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

Therefore, the standard form of a state-space model for har- illustrated in Figure.1. And in this paper, the calculation is
monic detection is established as Equation (19). executed when N increase from 1 to 13, for the harmonics
( in the power electronics system mainly contains these order
x(n + 1) = Ax(n) + Biαβγ (n) harmonics.
(19)
y(n) = I 6N x(n)
In the Equation (19), y(n) is the output of the state-space
model, A is the system matrix of the state space, B is the
input control matrix, and I6 N is the output control matrix.
Matrix A, B, and I6 N are time-independent constant coef-
ficient matrices, that is, the detection system based on the
presented state-space model above is a linear time-invariant
system.

III. PERFORMANCE ANALYSIS OF THE PROPOSED STATE

SPACE MODEL
A. CONTROLLABILITY ANALYSIS OF PROPOSED FIGURE 1. Rank(Mc ) under different λ and N for controllability analysis.
HARMONIC DETECTION SYSTEM
In Equation (19), the improved state-space model is estab-
Figure.1 shows the change of rank(MC )/6 concerning as
lished for 1st ∼ N th PSC, NSC, and ZSC harmonics detec-
N and λ changes, and it can be found that the desired value
tion, although the state variables are controlled by input
range of λ which makes rank(MC ) =6N , will decrease as N
variables iαβγ (τ ), it still needs to analyze its robustness to
increases. This also means that the higher order of harmonics
ensure that the model is practical.
to be detected, the smaller the value range of the λ which
In this paper, the controllability of the proposed state-space
makes the system controllable. In Figure.1, the parts above
model is analyzed.
curve st are all uncontrollable regions of the system, while
For a given rated angular frequency ωN and sampling
the areas below curve st are controllable regions under N .
period Ts , matrices A and B are affected by the detected
When N ≤6, the system is under control in the full range of
harmonics number N and constant-coefficient λ. So in this
λ ∈(0,1). while N = 13, the control of the system has been
paper, the effect of N and λ on the performance of the built
reduced to λ ∈(0, 0.18).
state-space model is mainly analyzed. According to PBH
criterion [34], which is the controllability criterion of the
B. STABILITY ANALYSIS OF PROPOSED HARMONIC
linear system, the controllability matrix MC of the proposed
DETECTION SYSTEM
system is shown in Equation(20).
System stability is the premise of its application. In this paper,
M C = B AB A2 B · · · A6N −1 B
 
(20) the Lyapunov Equation is used to judge the stability of multi-
input and multi-output systems [34], [35]. The Lyapunov
The sufficient and necessary condition for the detection sys- Equation is shown in Equation (21).
tem to be controllable is the rank(MC ) of the controllability
matrix rank(MC ) = 6 N . However, the analytic solution to AT PA − P = −S (21)
the rank of the matrix MC is very complex, especially when
the matrix dimension is high due to large N . Therefore, in this The matrices P and S in this equation are 6N -order square
paper, the relationship between rank(MC ) and N and λ of a matrices according to the analysis in Section.II. The neces-
matrix is solved by computer numerical calculation, which is sary and sufficient condition of the proposed system global

 N
P

  I 2 × ihp_αβ (n) + I 2 × ihn_αβ (n) 
iαβ (n)  h=1
iαβγ (n) = = N

iγ (n)  P 
ihz_γ (n)
h=1
 N
P

 I 2 × xhp_αβ (n) + I 2 × xhn_αβ (n) + O2 × xhz_αβ (n)

= h=1 
PN

O1×2 × xhp_αβ (n) + O1×2 × xhn_αβ (n) + E1×2 × xhz_αβ (n)
h=1
 
F
= x(n) = Hx(n) (13)
G

170006 VOLUME 8, 2020

D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

stability is that when S is a positive definite matrix, the exis-

tence of positive definite matrix P makes equation (20) true.
For the convenience of analysis, this paper sets S as the
identity matrix I6N of order 6N . Since S is a real symmetric
matrix, and it is easy to prove that P is also a real symmetric
matrix, the sufficient and necessary condition for P to be
positive definite at this time is that its eigenvalues are all
FIGURE 3. Convergence cure of magnitude error with the change of λ
greater than 0. Therefore, the determination of whether P and h.
is positive definite is transformed into the determination of
whether all the eigenvalues of P are greater than 0, which is
conducive to computer numerical calculation and judgment.
Figure.2 shows the relationship between the matrix P posi-
tive property and the constant-coefficient λ under different N
conditions. The ordinate ‘‘1’’ in Figure.2 testifies the positive
definiteness of P, which shows that the detection system is
globally stable. The y-coordinate ‘‘0’’ means that P is inde-
terminate, that is, the system is not globally stable. As you
can see, the range of the extraction system’s global stability
decreases as N increases. When N = 1, any λ ∈(0,1) system
is globally stable; when N = 13, reduce the value range of
for global stability of the system to λ ∈(0, 0.11).

FIGURE 4. (a, b) Convergence curve of 1Maghp,n (n) and 1Maghz (n) with
different λ when h = 1 (c, d) Convergence curve of 1Maghp,n (n) and
1Maghz (n) with different h when λ equal to 0.1.

on the convergence speed of 1Maghi is not monotonous. With

the increase of h, the convergence speed first accelerates,
and then slows down due to the occurrence of overshoot
FIGURE 2. Positive definiteness of P with different λ and N. after the increase. However, h has less effect on the rate of
convergence of amplitude error, and the less time change in
the convergence of amplitude error to 0 during the change
C. ERROR CONVERGENCE ANALYSIS OF PROPOSED of h.
HARMONIC DETECTION SYSTEM Different from the amplitude error, the convergence pro-
In addition to controllability and stability, the error perfor- cess of the phase error of the positive and negative sequence
mance of the system is also an important performance indi- components, and the phase error of the zero sequence compo-
cator. According to Equations (3), (6) and (19), the amplitude nents are quite different. In Figure.3, with the increase of λ,
value error 1Maghi and phase error 1ϕhi can be obtained the phase error 1ϕhp,n of positive and negative sequence com-
from Equations (22) and (23), as shown at the bottom of the ponents gradually accelerates the convergence rate, while the
next page, respectively, for hth harmonic, and the subscript convergence process of the error 1ϕhz of zero sequence com-
‘‘i’’ of each variable is ‘‘p’’, ‘‘n’’ or ‘‘z’’, which represent ponent is accompanied by the super-harmonic and gradually
PSC, NSC, and ZSC respectively. decaying oscillations, and its convergence speed near 0 has a
To simplify the analysis process, it is assumed that the input complex relationship with λ. Similarly, when h increases, due
signal only contains h-th harmonic ihi_αβγ (τ ), and according to the occurrence of an overshoot, 1ϕhp,n convergence rate
to the Equation (21) and (22), it can get λ and h as to 1Maghi appears to accelerate first and then decrease. For 1ϕhz , except
and 1ϕhi convergence process respectively, the influence of h = 1, the convergence process of 1ϕhz is accompanied
as shown in Figure.3 and Figure.4. In Figure.3, as for the by certain oscillation, and with the increase of h, its initial
positive and negative sequence component amplitude error convergence curve gradually tends to the negative half axis
1Maghp,n , and negative sequence components as amplitude of the ordinate. Similar to the amplitude error, h has little
error 1Maghz (τ ), although the expression is different, they influence on the speed of the convergence of the phase error
converge in the same way. As you can see, the amplitude error to 0.
1Maghi always converges to 0 and the convergence speed Unlike the 1Maghi (n), the convergence process of 1ϕhz (n)
accelerates with the increase in λ. While the effeteness of h is quite different from 1ϕhp (n) and 1ϕhn (n). As shown

VOLUME 8, 2020 170007

 D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

TABLE 1. Simulation parameter.

FIGURE 5. Test signal simulation waveform.

in Figure.4(a, c), the convergence time of 1ϕhp (n), and

1ϕhn (n) decrease with the increase of λ, and it is almost
unaffected by h. However, this regular does not always apply
to 1ϕhz (n). As shown in Figure.4(b), the convergence time of
1ϕhz (n) first increases and then decreases with the increase
of λ. But similar to the 1ϕhp (n) and 1ϕhn (n), the convergence Based on the harmonic detection of the state space
rate of 1ϕhz (n) is also almost unaffected by h, as shown model proposed in this paper, the simulated fundamen-
in Figure.4(d). tal and 5th harmonic PSC, NSC, and ZSC are shown in
Given that the detection system is controllable and globally Figure 6. It can be seen that the method proposed in this
stable, it can be found from Figure.3 and Figure.4. When paper can achieve error-free detection of amplitude and
λ =0.06, the corresponding amplitude error and phase error phase. Moreover, each phase sequence can be completely
1ϕhz has a smaller convergence curve overshoot, a faster con- extracted, and the extraction time of each phase sequence
vergence rate, and an ideal convergence curve, so λ =0.06 is can be controlled within a fundamental period. There is no
adopted in this paper, and with this situation, the convergence obvious overshoot phenomenon in the process of detecting
time of 1Maghi (n) and 1ϕhi (n) is almost equal to one funda- dynamics.
mental period from Figure.3 and Figure.4. Figure.7 shows the simulation comparison results of fun-
damental and 5-th harmonic detection results based on the
IV. SIMULATION AND EXPERIMENT RESULTS second-order generalized integrator (SOGI), discrete Fourier
A. SIMULATION RESULTS transform (DFT), and the state space method proposed in
The simulation is carried out in MATLAB/Simulink, and the this paper. It can be seen that, compared with SOGI and the
main parameter of simulation is shown in Table.1. proposed SSM method, the fundamental detected by SOGI is
Taking the detection of the fundamental, the 3-rd harmonic, polluted by the 5-th harmonic, and the 5-th harmonic detected
the 5-th harmonic, and the 7-th harmonic as an example, by this method has a lot of amplitude and phase error. And the
Table 1 list the composition of the tested input signal iabc . proposed SSM method and DFT have the same precision of
Its waveform is shown in Figure.5, and the distorted signal is steady-state extraction and slightly different dynamic effects.
added at 0.5s. It can be seen from the simulation results that the state-
The harmonic detection method proposed in this paper is space model method proposed in this paper has almost the
adopted to detect the input signal shown above, and the results same dynamic and steady-state performance as the DFT and
are shown in Figure.6. the proposed state-space model method can detect the PSC,

q
 x 2 (n) + x 2 (n), i = p, n
hi_α hi_β
1Maghi (n) = Ihi_m − q (22)
 2(x 2 (n) + x 2 (n)), i = z
hi_α hi_β
 x (n) cos (hω nT ) + x
hi_α 0 s hi_β (n) sin (hω0 nTs )

 q , i=p
2 2 (n)




 x hi_α (n) + x hi_β
 −xhi_α (n) cos (hω0 nTs ) + xhi_β (n) sin (hω0 nTs )


1ϕhi = ϕhi − arcsin q , i=n (23)
x 2 (n) + x 2 (n)

 hi_α hi_β
xhi_α (n) cos (hω0 nTs + π/4) + xhi_β (n) sin (hω0 nTs + π/4)


,

i=z


 q
2 (n) + x 2 (n))

 2(xhi_α hi_β

170008 VOLUME 8, 2020

D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

FIGURE 6. Simulation results of the proposed harmonics detection FIGURE 8. Picture of (a) the prototype and (b) programmable AC source.
method.

FIGURE 7. A simulation comparison of three detection methods (State FIGURE 9. Experiment waveform of the input signal.
Space Model, DFT, SOGI).

NSC, and ZSC separately of each fundamental and harmonics controller is adopted in built prototype to sample the input sig-
which the DFT cannot achieve. nal and implement the proposed harmonic detection method,
then the calculated result is output by DA chip.
B. EXPERIMENT RESULTS The experimental results based on the proposed SSM har-
To verify the effectiveness and the engineering practi- monics detection method are shown in Figure.10-12. And
cability of the proposed harmonic detection methods in Figure.10, Figure.11 and Figure.12 are the fundamental,3-th
this paper, built in the laboratory prototype as shown in harmonic, 5-th harmonic and 7-th harmonic detection results
Figure.8(a). of PSC, NSC, and ZSC respectively. Similar to the simulation
And the related validation experiment has been car- results, the proposed algorithm can be used to accurately
ried in the laboratory. The distortion test signal as shown detect out the PSC, NSC, and ZSC separately of the funda-
in Fig.9 is produced by the programmable AC source as mental component and the 5-th harmonic component from the
shown in Fig.8(b) based on Table 2. The DSP + FPGA distorted power grid signal.

VOLUME 8, 2020 170009

 D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

TABLE 2. Experiment parameter.

FIGURE 11. The experiment of the NSC detection results.

FIGURE 10. The experiment of the PSC detection results.

Figure.13 shows the comparison results of three harmonic

detection methods including SSM, DFT, and SOGI. Taking
phase A as an example, the dynamic process of detection
for fundamental and 5th harmonics is shown in Figure.13.
No distortion occurs in the extracted fundamental harmonic
by SSM and DFT, compared with severe distortion with
SOGI. For the extraction results of 5th harmonic, the detec- FIGURE 12. The experiment of the ZSC detection results.
tion speed of SSM and DFT are nearly the same, and both
are faster than that of SOGI. In addition, the 5th harmonic not only has a similar steady-state and dynamic performance
detected by SOGI has a large phase deviation, which is shown as DFT, but also can separate the PSC, NSC, and ZSC of each
in Figure.13. Based on the above experimental results, SSM harmonic.

170010 VOLUME 8, 2020

D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

[5] H. R. Baghaee, M. Mirsalim, G. B. Gharehpetian, and H. A. Talebi,

‘‘Unbalanced harmonic power sharing and voltage compensation of micro-
grids using radial basis function neural network-based harmonic power-
flow calculations for distributed and decentralised control structures,’’ IET
Gener., Transmiss. Distrib., vol. 12, no. 7, pp. 1518–1530, Nov. 2017.
[6] F. R. Blanquez, C. A. Platero, E. Rebollo, and F. Blazquez, ‘‘Novel rotor
ground-fault detection algorithm for synchronous machines with static
excitation based on third-harmonic voltage-phasor comparison,’’ IEEE
Trans. Ind. Electron., vol. 63, no. 4, pp. 2548–2558, Apr. 2016.
[7] D. Chen and G. Chen, ‘‘An optimized reference current detection method
for active power filter,’’ IEICE Electron. Express, vol. 14, no. 6, pp. 1–8,
2017.
[8] A. S. Masoum, N. Hashemnia, A. Abu-Siada, M. A. S. Masoum, and
S. M. Islam, ‘‘Online transformer internal fault detection based on instan-
taneous voltage and current measurements considering impact of harmon-
ics,’’ IEEE Trans. Power Del., vol. 32, no. 2, pp. 587–598, Apr. 2017.
[9] D. Chen, H. Chen, Y. Hu, and G. Chen, ‘‘A novel serial structure repetitive
control strategy for shunt active power filter,’’ COMPEL Int. J. Comput.
Math. Elect. Electron. Engine, vol. 38, no. 1, pp. 199–215, 2019.
[10] F. Harirchi and M. G. Simoes, ‘‘Enhanced instantaneous power theory
FIGURE 13. Experimental comparison of three detection methods:SSM, decomposition for power quality smart converter applications,’’ IEEE
DFT, and SOGI. Trans. Power Electron., vol. 33, no. 11, pp. 9344–9359, Nov. 2018.
[11] L. Ji-Quan and Z. Mao-Song, ‘‘A new method for harmonic detection based
on instantaneous reactive power theory,’’ Elect. Meas. Instrum., vol. 49,
V. CONCLUSION no. 10, pp. 29–32, Oct. 2012.
In this paper, a novel harmonics detection method based [12] P. Xiao, K. A. Corzine, and G. K. Venayagamoorthy, ‘‘Multiple reference
frame-based control of three-phase PWM boost rectifiers under unbalanced
on the state-space model is presented. The proposed SSM and distorted input conditions,’’ IEEE Trans. Power Electron., vol. 23,
method can detect the harmonics component of the polluted no. 4, pp. 2006–2017, Jul. 2008.
power grid with fast speed and high accuracy. Moreover, [13] P. Rodriguez, A. Luna, I. Candela, R. Mujal, R. Teodorescu, and
F. Blaabjerg, ‘‘Multiresonant frequency-locked loop for grid synchroniza-
the proposed SSM harmonics detection method has the ability tion of power converters under distorted grid conditions,’’ IEEE Trans. Ind.
of PSC, NSC, and ZSC separation for each detected har- Electron., vol. 58, no. 1, pp. 127–138, Jan. 2011.
monics order. Therefore, it can be used in complex industrial [14] Y. F. Wang and Y. W. Li, ‘‘Three-phase cascaded delayed signal cancella-
tion PLL for fast selective harmonic detection,’’ IEEE Trans. Ind. Electron.,
applications, including single-phase and three-phase asym- vol. 60, no. 4, pp. 1452–1463, Apr. 2013.
metric applications. And the corresponding simulation model [15] A. Bagheri, M. Mardaneh, A. Rajaei, and A. Rahideh, ‘‘Detection of grid
is built according to the theoretical analysis, the results show voltage fundamental and harmonic components using Kalman filter and
generalized averaging method,’’ IEEE Trans. Power Electron., vol. 31,
the effectiveness of the proposed SSM harmonics detection no. 2, pp. 1064–1073, Feb. 2016.
method. Moreover, the comparison with the traditional har- [16] J. Oliveira Pinto, R. Macedo, V. Monteiro, L. Barros, T. Sousa, and
monics detection method proves that the proposed approach J. Afonso, ‘‘Single-phase shunt active power filter based on a 5-Level
converter topology,’’ Energies, vol. 11, no. 4, p. 1019, Apr. 2018.
has high precision, fast response and has the ability of PSC, [17] L. Hadjidemetriou, E. Kyriakides, and F. Blaabjerg, ‘‘A robust synchro-
NSC, and ZSC separation. The prototype is designed in nization to enhance the power quality of renewable energy systems,’’ IEEE
the laboratory, and the experimental results show that the Trans. Ind. Electron., vol. 62, no. 8, pp. 4858–4868, Aug. 2015.
[18] Y. F. Wang and Y. W. Li, ‘‘A grid fundamental and harmonic component
proposed SSM harmonics detection method is practical in detection method for single-phase systems,’’ IEEE Trans. Power Electron.,
engineering. This novel detection method is meaningful for vol. 28, no. 5, pp. 2204–2213, May 2013.
harmonics control of the power electronics system. [19] M. Chen, L. Peng, B. Wang, and W. Wu, ‘‘Accurate and fast harmonic
detection based on the generalized trigonometric function delayed signal
cancellation,’’ IEEE Access, vol. 7, pp. 3438–3447, 2019.
ACKNOWLEDGMENT [20] J. Alfonso-Gil, E. Peréz, C. Ariño, and H. Beltran, ‘‘Optimization algo-
Dongdong Chen would like to thank his girlfriend Y. Wang rithm for selective compensation in a shunt active power filter,’’ IEEE
Trans. Ind. Electron., vol. 62, no. 6, pp. 3351–3361, Dec. 2014.
for her support, encouragement, and company in the author’s [21] G. W. Chang, C. I. Chen, R. C. Hong, and H. M. Li, ‘‘Extended real model
work. Wish to hold your hand and grow old together with you. of Kalman filter for time-varying harmonics estimation,’’ IEEE Trans.
Power Del., vol. 25, no. 1, pp. 17–26, Dec. 2009.
[22] P. Nayak and B. N. Sahu, ‘‘A robust extended Kalman filter for the esti-
REFERENCES mation of time varying power system harmonics in noise,’’ in Proc. IEEE
[1] E. Hossain, M. R. Tur, S. Padmanaban, S. Ay, and I. Khan, ‘‘Analysis and Power, Commun. Inf. Technol. Conf. (PCITC), Oct. 2015, pp. 635–640.
mitigation of power quality issues in distributed generation systems using [23] T. Zang, Y. Wang, H. Sun, and Z. He, ‘‘Variable parameter Kalman filter
custom power devices,’’ IEEE Access, vol. 6, pp. 16816–16833, 2018. based dynamic harmonic state estimation for power systems with wind
[2] B. Chen, G. Pin, W. M. Ng, T. Parisini, and S.-Y.-R. Hui, ‘‘A fast- energy integration,’’ in Proc. IEEE Conf. Energy Internet Energy Syst.
convergent modulation integral observer for online detection of the funda- Integr. (EI2), Nov. 2017, pp. 1–5.
mental and harmonics in grid-connected power electronics systems,’’ IEEE [24] D. A. Gudovskiy and L. Chu, ‘‘An accurate and stable sliding DFT com-
Trans. Power Electron., vol. 32, no. 4, pp. 2596–2607, Apr. 2017. puted by a modified CIC filter [tips & tricks],’’ IEEE Signal Process. Mag.,
[3] K. Zhou, Y. Yang, F. Blaabjerg, and D. Wang, ‘‘Optimal selective harmonic vol. 34, no. 1, pp. 89–93, Jan. 2017.
control for power harmonics mitigation,’’ IEEE Trans. Ind. Electron., [25] D. Chen, ‘‘An improved harmonics detection method based on sliding
vol. 62, no. 2, pp. 1220–1230, Feb. 2015. disrete Fourier transform for three-phase grid-tie inverter system,’’ IEICE
[4] R. Haider, C. H. Kim, T. Ghanbari, and S. B. A. Bukhari, ‘‘Harmonic- Electron. Express, vol. 16, no. 3, pp. 1–10, 2019.
signature-based islanding detection in grid-connected distributed gener- [26] S. Wang, A. Etemadi, and M. Doroslovački, ‘‘Adaptive cascaded delayed
ation systems using Kalman filter,’’ IET Renew. Power Gener., vol. 12, signal cancellation PLL for three-phase grid under unbalanced and dis-
no. 15, pp. 1813–1822, Nov. 2018. torted condition,’’ Electr. Power Syst. Res., vol. 180, pp. 1–13, Mar. 2020.

VOLUME 8, 2020 170011

 D. Chen, L. Xiao: Novel Fundamental and Harmonics Detection Method Based on State-Space Model for Power Electronics System

[27] W. Yao, Z. Teng, Y. Gao, and Q. Tang, ‘‘Measurement of power system DONGDONG CHEN received the B.S. degree
harmonic based on adaptive kaiser self-convolution window,’’ IET Gener., in electronic information engineering from Zhe-
Transmiss. Distrib., vol. 10, no. 2, pp. 390–398, Feb. 2016. jiang University, Hangzhou, China, in 2013, and
[28] J. Na, J. Yang, X. Ren, and Y. Guo, ‘‘Adaptive online estimation of time- the Ph.D. degree in electrical engineering from
varying parameter nonlinear systems,’’ in Proc. 32nd Chin. Control Conf., the College of Electrical Engineering, Zhejiang
Xi’an, China, 2013, pp. 4570–4575. University, in 2018.
[29] S. K. Singh, A. K. Goswami, and N. Sinha, ‘‘Harmonic parameter esti- He is currently a Vice Professor with the
mation of a power signal using FT-RLS algorithm,’’ in Proc. 16th Int.
Minnan University of Science and Technology.
Conf. Harmon. Qual. Power (ICHQP), Bucharest, Romania, May 2014,
His research interests include power electronics
pp. 157–161.
[30] Y. Hu, Y. Shao, R. Yang, X. Long, and G. Chen, ‘‘A configurable system design, power converter digital control,
virtual impedance method for grid-connected virtual synchronous gen- and optimal design.
erator to improve the quality of output current,’’ IEEE J. Emerg. Sel.
Topics Power Electron., vol. 8, no. 3, pp. 2404–2419, Sep. 2020, doi:
10.1109/JESTPE.2019.2918386.
[31] C.-S. Park, ‘‘Fast, accurate, and guaranteed stable sliding discrete Fourier
transform,’’ IEEE Signal Process. Mag., vol. 32, no. 4, pp. 145–156, LONG XIAO received the B.S. degree in optoelec-
Jul. 2015. tronic information engineering from the Huazhong
[32] S. O. Gilani, Y. Ilyas, and M. Jamil, ‘‘Power line noise removal from ECG University of Science and Technology, Wuhan,
signal using notch, band stop and adaptive filters,’’ in Proc. Int. Conf. China, in 2012, the B.S. and M.S. degrees in
Electron., Inf., Commun. (ICEIC), Honolulu, HI, USA, Jan. 2018, pp. 1–4.
software engineering from the Wuhan University
[33] IEEE Recommended Practice and Requirements for Harmonic Control in
Electric Power Systems, IEEE Standard 519-2014 (Revision IEEE Std 519-
of Science and Technology, Wuhan, in 2012 and
1992), 2014, pp. 1–29. 2014, respectively, and the Ph.D. degree in elec-
[34] Q. Wang, Q. Zhang, and Q. Wang, ‘‘Lyapunov equations with positive trical engineering from the College of Electri-
definite solutions for Continuous-and discrete-time descriptor systems,’’ cal Engineering, Zhejiang University, Hangzhou,
in Proc. IEEE Int. Conf. Control Autom., Guangzhou, China, May 2007, China, in 2019.
pp. 681–685. He is currently a Vice Professor with the Minnan University of Science
[35] X. Liu, ‘‘Stability analysis of switched positive systems: A switched linear and Technology. His research interests include power semiconductor device
copositive Lyapunov function method,’’ IEEE Trans. Circuits Syst. II, Exp. modeling and driving, power converter digital control, and optimal design.
Briefs, vol. 56, no. 5, pp. 414–418, May 2009.

170012 VOLUME 8, 2020