electronics

Article
Control Design of LCL Type Grid-Connected Inverter
Based on State Feedback Linearization
Longyue Yang 1,2, *, Chunchun Feng 2 and Jianhua Liu 1,2
1 Jiangsu Province Laboratory of Mining Electric and Automation, China University of Mining and
Technology, Xuzhou 221116, China
2 School of Electrical and Power Engineering, China University of Mining and Technology,
Xuzhou 221116, China
* Correspondence: yanglongyue@cumt.edu.cn; Tel.: +86-0516-83885605




Received: 11 July 2019; Accepted: 3 August 2019; Published: 7 August 2019


Abstract: Control strategy is the key technology of power electronic converter equipment. In order to
solve the problem of controller design, a general design method is presented in this paper, which is
more convenient to use computer machine learning and provides design rules for high-order power
electronic system. With the higher order system Lie derivative, the nonlinear system is mapped to
a controllable standard type, and then classical linear system control method is adopted to design
the controller. The simulation and experimental results show that the two controllers have good
steady-state control performance and dynamic response performance.

Keywords: state feedback linearization; grid-connected inverter; LCL filter

1. Introduction
As the main equipment of an AC power system, grid-connected voltage source inverters (VSI)
have been widely used in recent years, such as photovoltaic inverters [1], Pulse Width Modulation
Pulse Width Modulation (PWM) rectifiers [2], static var generators [3], active power filters [4], etc.
The filter is an important part of the inverter, the structure of which directly determines the
mathematical model and control mode of the inverter. Nowadays, a Inductance-Capacitance-Inductance
(LCL) filter, with the advantages of small size, low cost and high harmonic attenuation for high frequency
current, is widely used in voltage source type grid-connected converters. Considering that the LCL
filter is the 3-order system, the damping of the system is small, resulting in a resonant peak of the
grid-side inductance current. This phenomenon will adversely affect the safe and stable operation
of grid-connected system. In order to improve the damping characteristics of the system, additional
system damping is required. Active damping control method is commonly used at present stage [5].
With the state variables feedback, the traditional proportional integral controller can be used to achieve
the grid-side inductance current control.
However, for nonlinear power electronic devices, this design method of the controller is based on
small signal modeling [6] and harmonic linearization. As the coupling and high order terms in the
system are ignored in the Taylor series calculation, the obtained controller is suitable for working at
steady working point with poor performance in other control domains [7].
In order to solve the global control problem, many solutions have been put forward [8,9]. Among
them, the state feedback linearization method that develops from differential geometry [10,11] has
become an effective way to solve the nonlinear power electronics system control problems. Based on
the differential homeomorphism, Lie derivative is used to analyze the numerical relation between
the state variables, the input variables, and the output variables. Then, the necessary and sufficient
conditions for controllability and observability of nonlinear control systems can be established. Choi et

Electronics 2019, 8, 877; doi:10.3390/electronics8080877 www.mdpi.com/journal/electronics

 2 of 16

theElectronics
permanent 2019, 8,magnet
877 synchronous motor [12]. In addition, the drive flux and torque ripple 2 of were
15

suppressed. Yang et al. combined feedback linearization and sliding mode variable structure
control to complete
al. proposed a feedbackthe control of three-phase
linearization direct torque four-leg
controlinverter [13]. Thismagnet
for the permanent method was used to
synchronous
decouple the torque
motor [12]. and stator
In addition, flux flux
the drive of the
andinductive motor
torque ripple by Lascu
were et al. [14].
suppressed. YangYang
et al. etcombined
al. applied
thefeedback
state feedback
linearization and sliding mode variable structure control to complete the control of three-phasethe
control to the Modular Multilevel Converter (MMC) system and analyzed
performance characteristics
four-leg inverter [13]. Thisof the system
method was used[15].to decouple the torque and stator flux of the inductive
Different from the local approximate linearization
motor by Lascu et al. [14]. Yang et al. applied method,
the state feedback the
control feedback
to the Modulardecoupling
Multilevel is
Converter
achieved by (MMC)
adopting system and analyzed
this nonlinear the performance
algorithm characteristics
without ignoring of the system terms.
the higher-order [15]. However,
the commonly used feedback linearization algorithm at the present stage is mostlyisused
Different from the local approximate linearization method, the feedback decoupling achieved by 2-
in the
adopting
order this nonlinear
or 1-order system.algorithm without
The research onignoring the higher-order
the controller design of terms. However,
high-order orthe commonly
other complex
used feedback linearization algorithm at the present stage is mostly used
systems is rare. In addition, when the order of the controlled object is high, whether the state in the 2-order or 1-order
system.control
feedback The research
can be ontransformed
the controllerinto
design of high-order
a simple or other
form has complex
not been systemsby
analyzed is literatures.
rare. In addition,
when the order of the controlled object is high, whether the state feedback control can be transformed
In order to achieve the control of high-order power electronic systems, the design of controller
into a simple form has not been analyzed by literatures.
based on LCL filter type grid-connected inverters is studied in this paper. For the 3-order control
In order to achieve the control of high-order power electronic systems, the design of controller
system, two controller design methods based on state feedback linearization are proposed in this
based on LCL filter type grid-connected inverters is studied in this paper. For the 3-order control
paper. Lie derivative vector field is used to solve the system relationship and design the decoupling
system, two controller design methods based on state feedback linearization are proposed in this paper.
matrix. Through the nonlinear mapping, the nonlinear system can be mapped to a controllable
Lie derivative vector field is used to solve the system relationship and design the decoupling matrix.
standard
Through form. Then, themapping,
the nonlinear classical linear system system
the nonlinear controlcan
method,
be mappedwhichtocan be designed
a controllable easily, is
standard
applicable
form. Then, the classical linear system control method, which can be designed easily, is applicable forand
for the controller design. Finally, the performance of the two controllers is compared
analyzed by simulations
the controller and experiments.
design. Finally, the performance of the two controllers is compared and analyzed by
simulations and experiments.
2. Single Closed-Loop Controller Design Based on State Feedback Linearization
2. Single Closed-Loop Controller Design Based on State Feedback Linearization
2.1. Model of the Three-Phase Three-Leg Grid-Connected Inverter
2.1. Model of the Three-Phase Three-Leg Grid-Connected Inverter
Figure 1 shows the structure of the three-phase three-leg LCL grid-connected inverter. L1 and
Figure 1 shows the structure of the three-phase three-leg LCL grid-connected inverter. L1 and L2
L2 denote the inverter side inductance and the grid-side inductance, respectively.
denote the inverter side inductance and the grid-side inductance, respectively.

L1 L2
ia
ea
ib
+ eb
U dc ic
− ec
C

Figure 1. Schematic diagram of the three-phase three-leg grid-connected inverter.

Figure 1. Schematic diagram of the three-phase three-leg grid-connected inverter.
As the model has been analyzed by many literatures, this paper does not repeat the description.
Asmathematical
The the model model
has been analyzed
of inverter by many axis
in dq coordinate literatures, this paper
is given directly as does not repeat the
description. The mathematical model of inverter in dq coordinate axis is given directly as
ω
 
 0 −1/L1 0 0 0 md /L1 
   
i1d i1d
ω
 0 0−1 L1 −1/L0 0 0
0 md Lm1  /L
  
  i1d −ω q  i1d1  
 
    −ω 0 0 0
   
 i1q 1 i1q 
−
  1q 
0 1 L
ω 1 −1/C 0 0
0 q 1  0 i1q  
m L
 
 ucd  i  1/C 0 0 ucd 
 0−1/C 0  0  

ω

 =u  0 1 C 1/C 0 −ω0 −1 C

d 0 0
ucq

ucq

 ucd 

dt    cd

d  u  = 0 0 0 1 C 1/L−2ω 00 0 0 −1 C ω 0  0  
 
i2d i2d
  ucq  

 cq 

 (1)
dt   0 0 0 0 01 L2 1/L 02 0−ω ω 0 0  0  

i2q i2q
 
  i2d i
 0 2d  
 
  − md 0 − mq 0 0 1 L − ω 0 0
  (1)
udc  i  C dc 0 0 2 0 0
  i2q 
udc

Cdc
 2q  "  m mq #T " #  
d
 udc   −
0 0 0 0 0 −1/L 0 e 0   udc 
+  Cdc − Cdc 0 0 2 0 0 d

0 0 0 0 −1/L2 0 0 eq
T
0 0 0 0 0 −1 L2 0  ed 
+  
0 0 0 0 − 1 L2 0 0   eq 
Electronics 2019, 8, 877 3 of 15

h iT
where the state variables X = i1d i1q ucd ucq i2d i2q udc represent the inverter side inductance current,
the filter capacitor voltage, the grid-side inductance current and the DC voltage in the synchronous
reference coordinate system, respectively. ed and eq represent grid voltage in the synchronous reference
coordinate system, respectively. md and mq represent the nonlinear pulse width modulation variables,
respectively. ω represents the fundamental angular frequency of the system.
According to (1), due to the mutual inductance and mutual capacitance, the inductance current
and the capacitance voltage is coupled in the synchronous reference coordinate system.
Usually, the voltage source grid-connected inverter system is a double closed loop structure.
The outer loop controls DC voltage, and the inner loop controls AC current. However, for the different
structure like the PV system, the DC voltage may not need to be controlled. Considering the generality,
only the current inner loop is analyzed. Due to the LCL filter, the system can be regarded as a 2-input
2-output system. At the same time, the existence of LCL filter capacitor, which introduces a pair of
conjugate pure imaginary roots to cause resonance, increases the order of the system. Therefore, it is
necessary to consider the suppression of resonance.

2.2. Single Closed-Loop Control Strategy Based on State Feedback Linearization

Define the LCL filter inverter side inductance current, filter capacitor voltage, and grid-side
iT
inductance current as state variables, written as X = [x1 x2 x3 x4 x5 x6 ]T = i1d i1q ucd ucq i2d i2q .
h

Define the 2 dimensional modulation vector in the synchronous coordinate system as the input variable,
iT
written as U = [u1 u2 ]T = md mq . The deviation of grid-side inductance current (controlled object)
h
iT
is taken as the output variable Y = R(X) = [r1 (X) r2 (X)]T = idre f − i2d iqre f − i2q , in which idre f and
h

iqre f are the reference current.

Therefore, Equation (1) can be expressed as a nonlinear differential equation composed of
polynomials of state variables and input variables:
 .
 X = f[X(t)] + G[X(t)]U

 Y = R(X)
 (2)

ωx2 − x3 /L1
 
 
 
 −ωx 1 − x 4 /L 1 

 x /C + ωx − x /C 
 " #T
1 4 5 0 udc /L1 0 0 0 0
where f(X) =  , G(X) = [ g1 (X) g2 (X)] = .
 
 x2 /C − ωx3 − x6 /C  udc /L1 0 0 0 0 0
 
 x /L + ωx − u /L 
 3 2 6 gd 2  
x4 /L2 − ωx5 − u gq /L2

Equation (2) shows that the LCL-type three-phase three-leg converter is a 2-input 2-output affine
nonlinear system with the state variables number of n = 6. It is nonlinear for the state vector X,
but linear for the input U. Since the state variables are all dq-axis symmetric variables, the number of
vector field pairs is l = 3. The corresponding l − 1 order Lie derivative can be described as


 ad f g1 (X) = L f g1 − L g1 f


ωu
 h u
iT
= 0 L1dc − L1dcC 0 0 0





ad2f g1 (X) = L f [ad f g1 (X)] − Lad f g1 (X) f






 T
ω udc
  2
 udc 2ωu u
= L1 + 2 0 0 L Cdc − L Ldc2 C 0



 L1 C 1 1
(3)

ad f g2 (X) = L f g2 − L g2 f






 h ωu u
iT
= − L1dc 0 0 − L1dcC 0 0





ad2f g2 (X) = L f [ad f g2 (X)] − Lad f g2 (X) f






 T
ω2 udc
 
 udc 2ωudc udc
= 0 + − 0 0 −


 L1 2 L1 C L C
1 L L2 C
1
Electronics 2019, 8, 877 4 of 15

where ad f g denotes Lie bracket operation of vector fields f and g, written as ad f g = L f g − L g f =

(∂g/∂X) f − (∂ f /∂X) g. Then, the corresponding 6-dimensional vector field matrix can be expressed as
h i
D6 = g1 g2 ad f g1 ad f g2 ad2f g1 ad2f g2
ωu ω2 udc
 
 udc − L1dc
udc
0 0 L1 + L2 C 0 
 L1 1

 0 udc ωudc ω udc
2

 udc
0 0 L1 + L2 C

L1 L1
 
1


u 2ωu  (4)
=  0 0 − L dcC 0 0 − L Cdc
 
1 1

 u 2ωudc 
 0 0 0 − L dcC L C 0 
 1 1 
u
 0 0 0 0 − L Ldc2 C 0
 

1
u
 
 0 0 0 0 0 − L Ldc2 C 
1

Obviously, it can be obtained that RankD6 = 6 = n. The rank of the matrix formed by the vector
fields is n in the neighborhood of X0 .  
According to Equations (3) and (4), the elements in g1 g2 ad f g1 ad f g2 ad2f g1 ad2f g2 do not contain
state variables. In other words, they are the constant vector fields, meaning that the result of Lie
brackets operation between two elements is a zero vector. Define Di as the matrix of the former i
column elements in Equation (4). It can be calculated that the 6 vector fields satisfy involution relation.
Therefore, according to the nonlinear control theory, state feedback can be applied to linearize the
inverter system.
A MIMO nonlinear system has a relative degree {r1 , . . . , rm } at a point x0 if

1. L g j Lkf ri (X) , 0 for all 0 ≤ j ≤ m, for all k ≤ ri − 1, for all 0 ≤ i≤ m and for all x in a neighborhood
of x0 .
2. the m × m matrix E(X) is nonsingular at x = x0 .

In Equation (2), the output variable of the inverter is current error r1 (X) = idre f − i2d . In order to
analyze the numerical relation of the output, the Lie derivatives need to be calculated as follows:

∂r1 (X)



 L g1 r1 (X) = ∂X
g1 (X) = 0

∂r1 (X)


L g2 r1 (X) = ∂X g2 (X) = 0





∂r1 (X)

f (X) = − L12 x3 − ωx6 + L12 ed

L f r1 (X) = ∂X





∂(L f r1 (X))






 L g1 L f r1 (X) = ∂X
g1 ( X ) = 0

∂(L f r1 (X)) (5)

L g2 L f r1 (X) = g2 ( X ) = 0

∂X




∂

 ( L r
f 1 ( X )) x −x ω
L2f r1 (X) = L2 x4 + ω x5
f (X) = L5 2 C1 − 2ω
 2 + L2 e q

∂X




∂(L2f r1 (X))


u
L g1 L2f r1 (X) =

g1 (X) = − L Ldc2 C , 0


∂X



 1
∂(L2f r1 (X))


L g2 L2f r1 (X) =



∂X
g2 ( X ) = 0

where L g1 L2f r1 (X) , 0. The relation degree of the output is d1 = 3.

Similarly, the output r2 (X) = iqre f − i2q is symmetrical. The relation degree of the output r2 (X)
is d2 = 3.
Electronics 2019, 8, 877 5 of 15

The Lie derivatives can be obtained as follows:

∂r2 (X)



 L g1 r2 (X) = ∂X g1 (X) = 0

∂r

 2 ( X )
 L g2 r2 (X) = ∂X g2 (X) = 0




∂r

( X )
f (X) = − L12 x4 + ωx5 + L12 eq

 L r (X) = 2



 f 2 ∂X
∂(L f r2 (X))



 L g1 L f r2 (X) = g1 ( X ) = 0

∂X



∂ (6)
 ( L r ( X ))
f 2
 L g2 L f r2 (X) = g2 ( X ) = 0

∂X



∂

 ( L r
f 2 ( X )) x −x ω
f (X) = L6 2 C2 + 2ωL2 x3 + ω x6 −
 L2 r2 (X) = 2

L2 ed



 f ∂X
∂(L2f r2 (X))


2 r (X) =

L L g1 ( X ) = 0

g 2

 1 f ∂X



∂(L f r2 (X))
2

u

 L g2 L2 r2 (X) = g2 (X) = − L Ldc2 C , 0


f ∂X 1

Based on the above analysis, the decoupling matrix can be established as follows.

 L g1 L2f r1 (X) L g2 L2f r1 (X) 

 
E(X) =  
L g1 L2f r2 (X) L g2 L2f r2 (X) 

" # (7)
−udc /(L1 L2 C) 0
=
0 −udc /(L1 L2 C)

It can be calculated that the matrix is a non-singular matrix. The total relation degree of the system
output variables is d = d1 + d2 = 6 = n. Therefore, in view of coordinate transformation, the original
system can be directly transformed into a controllable linear system of Brunovsky standard type.
Define the state variable of linear standard system after feedback linearization as
Z = [z1 z2 z3 z4 z5 z6 ]T . According to the relative order theory, the relationship between the new
state variables and the original state variables can be expressed as

z1 = r1 (X) = idre f − x5





z2 = L f r1 (X) = − L12 x3 − ωx6 + 1

L2 ed





x5 −x1 ω

L2 x4 + ω x5
2ω
z3 = L2f r1 (X) = − 2 + L2 eq


L2 C


(8)





 z4 = r2 (X) = iqre f − x6


z5 = L f r2 (X) = − L12 x4 + ωx5 + 1
L2 e q






x6 −x2 ω
L2 x3 + ω x6
2ω
z6 = L2f r2 (X) = 2

+ − L2 ed


L2 C

Equation (8) shows that the original system state variable X is mapped into new state variable Z
after the coordinate transformation of the Lie derivative. The coupling and high order terms in the
system are not ignored in the coordinate transformation. The numerical relationship between new
state variables can be calculated as
Electronics 2019, 8, 877 6 of 15

. ∂r1 (X) .



 z1 = ∂X
X = − L12 x3 − ωx6 + L12 ed = z2


. ∂(L f r1 (X)) .

x −x ω
L2 x4 + ω x5 + L2 eq
X = L5 2 C1 − 2ω 2

z2 = = z3

∂X





∂(L2f r1 (X)) .

 . L1 + L2 3ω2
X = − L3ω

z3 = x2 + ( + L2 )x3

∂X

2C L1 L22 C




ω2 udc
+( L3ω + ω3 )x6 − ( L21C +

L2 )ed − L1 L2 C u1


2C


 2 (9)
. ∂r2 (X) . 1
ωx5 + L12 eq = z5

z4 = X = − x +

4

∂X L2



∂(L f r2 (X)) .

.

x −x ω
X = L6 2 C2 + 2ωL2 x3 + ω x6 − L2 ed
2

z5 = = z6

∂X





∂(L2f r2 (X)) .

 . 3ω2
3ω
+ ( LL1 + L2


 z6 = ∂X
X= L2 C x 1 + L2 )x4
L2 C

1 2



ω2 udc
−( L3ω + ω3 )x5 − ( 1

+ L2 ) e q − L1 L2 C u2


2C L22 C


It can be seen that information of z1 is not directly contained in state variable z3 and information
of z4 is not directly contained in state variable z6 . Defining the control variable after the coordinate
transformation as V = [v1 v2 ]T , the relationship between the original nonlinear system control variable
U and V of can be obtained as
U = E−1 (X)[V − Γ(X)] (10)
 L +L 2 
  − L3ω x2 + ( 1 2 2 + 3ω
 
2C L2 )x3
 
  L1 L2 C  
ω2
 
 L f r1 (X)    +( L2 C + ω )x6 − ( L2 C + L2 )ed  
 r1    3ω 3 1 
where Γ(X) =  r2  =   2
2
.
L f r2 (X)    L3ωC x1 + ( L1 +2L2 + 3ω

)x4

L
 
  2 L L
1 2 C 2 
 
  −( 3ω + ω3 )x5 − ( 21 + ω2 )eq  
 
L C 2 L L2 C 2
Substituting Equation (10) into Equation (2), the controller can be expressed as

L1 + L2
1
[−3L1 ωi1q + (3L1 Cω2 +



 u1 = udc L2 )ucd

+(3L1 ω + L1 L2 Cω3 )i2q − (L1 Cω2 + LL12 )ed ] − L1uL2 C v1




dc (11)

= u1 [3L1 ωi1d + (3L1 Cω2 + L1L+2L2 )ucq



 u2

 dc
L L L C
−(3L1 ω + L1 L2 Cω3 )i2d − (L1 Cω2 + L12 )eq ] − 1u 2 v2



dc

The original nonlinear system is transformed into the linear system as follows:
 .
 Z = AZ + BV

 Y = CZ
 (12)

     
0 1 0 0 0 0 
  0 0   1 0 
     
0 0 1 0 0 0 


 0 0 
 
 0 0 
 
  0 0 
0 0 0 0 0 0 
  1 0 
where A =  , B =  , C =  .
    
0 0 0 0 1 0 


 0 0 
 
 0 1 
 
0 0 0 0 0 1   0 0   0 0 
 
    
0 0 0 0 0 0 0 1 0 0
     
The matrix A shows that the dq components of the original system are completely decoupled by
nonlinear mapping. The transfer function of the system is shown in Figure 2. The block diagrams in
the dotted box are the equivalent controller and the original system.
where A = 0 0 0 0 1 0  , B =  0 0  , C =  0 1  .
     
0 0 0 0 0 1  0 0  0 0 
     
0 0 0 0 0 0   0 1   0 0 
The matrix A shows that the dq components of the original system are completely decoupled
by nonlinear
Electronics mapping.
2019, 8, 877The transfer function of the system is shown in Figure 2. The block diagrams
7 of 15
in the dotted box are the equivalent controller and the original system.

ed i1q i2q
− L1Cω 2 − L1 L2 −3L1ω 3L1ω + L1L2Cω 3
*
i2d u1d i1d icd ucd i2d
v1 d − L1L2C 1 udc udc 1 / sL1 1/ sC 1/ sL2
i2d
ucd
3L1Cω 2 + ( L1 + L2 ) L2

eq i1d i2 d
− L1Cω 2 − L1 L2 −3L1ω 3L1ω + L1L2Cω 3
*
i2q u1q i1q icq ucq i2q
v1q − L1L2C 1 udc udc 1 / sL1 1/ sC 1 / sL2
i2 q
ucq
3L1Cω 2 + ( L1 + L2 ) L2

Figure
Figure2.2.Current
Currentcontrol
controlsystem
systembased
basedononstate feedback
state linearization.
feedback linearization.

For the
the system
systemshown
shownininEquation
Equation (12),
(12), thethe controller
controller V can
V can be designed
be designed to make
to make it stable
it stable at the at
T
the
point [𝑧 𝑧[z]1 z=
point 4 ] [0= 0]T .the
0] [0. As Asfeedback
the feedback decoupling
decoupling of theof the original
original systemsystem hascompleted,
has been been completed,
the
controller
the can can
controller be designed
be designed based on the
based onlinear theory
the linear to achieve
theory the control
to achieve of original
the control system.
of original system.

2.3. Parameters Design

2.3. Parameters DesignofofCurrent
CurrentSingle
SingleClosed
ClosedLoop
LoopController
Controller
Through thethe state
state feedback
feedbacklinearization,
linearization,the
thesystem
systemisistransformed
transformed into
into a 3-order
a 3-order controllable
controllable
standard
standard type.
type. Therefore,
Therefore,atatleast twotwo
least open-loop
open-loopzeroszeros
should be introduced
should into the
be introduced linear
into thecontroller,
linear
controller,
that that is, two
is, two first-order first-order differentiation
differentiation elements.
elements. Considering that Considering that the
the differentiation differentiation
elements may cause
elements
system maythe
noise, cause systemshould
controller noise, contain
the controller
a filter should contain
part (i.e., a filter part
the first-order (i.e.,element).
inertial the first-order
To make
inertial element). To make the controller simple, the transfer function of the system controller
the controller simple, the transfer function of the system controller after linearization can be written after as:
linearization can be written as:
v1 = v2 = −(k2 s2 + k1 s + k0 )/(s + k3 ) (13)

Then the open-loop transfer function and the closed-loop transfer function of the control system
can be represented as follows:
k2 s2 + k1 s + k0
GOL (s) = (14)
s3 (s + k3 )
k2 s2 + k1 s + k0
GCL (s) = (15)
s4 + k3 s3 + k2 s2 + k1 s + k0
Compared with the original open-loop system, the controller which adds two zeros and one pole
as a series correction link turns the original 3-order system into a 4-order one. Therefore, the order
reduction processing of the higher order system is considered.
Assume two poles and two zeros constitute a pair of dipoles. Then, Equation (15) can be
rewritten as
k2 (s2 + 2ζ1 ωn1 s + ω2n1 )
GCL (s) = (16)
(s2 + 2ζ2 ωn2 s + ω2n2 )(s2 + 2ζ3 ωn3 s + ω2n3 )
With dipoles elimination, the system can be transformed to a typical 2-order system. For the
closed-loop transfer function of the 2-order linear system, the parameter design method of the linear
system can be used to construct the restrictive conditions, so as to obtain the parameters range.
Electronics 2019, 8, 877 8 of 15

1. Stability conditions of the system

According to the stability criterion, the closed loop characteristic equation of the original system
should satisfy the Hurtwitz criterion as follows:

∆1 = k2 k3 > 0




 ∆2 = ∆1 − k1 > 0



(17)
∆3 = k1 ∆2 − k0 k32 > 0




 ∆4 = k0 k1 k2 k3 > 0



2. Restrictive conditions of dipoles parameters

The closed loop system (16) can be reduced to a 2-order system with sufficient condition of dipoles
existence. It is necessary to keep the dipole far from the imaginary axis (i.e., ζ2 ωn2 > ζ3 ωn3 ).

ζ2 ωn2
q q
(ζ1 ωn1 − ζ2 ωn2 ) + j(ωn1 1 − ζ21 − ωn2 1 − ζ22 ) < (18)
10

3. Cut-off frequency of the closed loop system

The cut-off frequency (i.e., ωb ) of current control system is generally less than 1/5 of
switching frequency.
Gain( jω) ω=ωb
k2 (19)
= q 2
= 0.707
(ω2n3 −ω2b ) +(2ζ3 ωn3 ωb )2

4. Closed loop amplitude frequency characteristic at zero frequency

The closed loop system should maintain good tracking performance at low frequency band.
The amplitude frequency characteristics satisfy the requirement of Gain( jω) ω=0 ≈ 1.

k2 ω2n1
Gain( jω)|ω=0 = ≈1 (20)
ω2n2 ω2n3

Combined with the above four conditions, the optimal control shown in Equation (12) can be
realized according to the classical control theory.

3. Design of Double Closed Loop Controller Based on Reduced Order State Feedback Linearization
The single closed loop feedback linearization method is used to decouple and simplify the inverter.
The controller designed through this method is simple in structure, but it is too dependent on system
precise model. Furthermore, the design of controller coefficient is complex, which limits the application
of the algorithm.
Considering the objective of control, it is not necessary to configure all open loop poles of LCL type
inverter model to the coordinate origin. Therefore, the system can be divided into two parts, which are
analyzed separately. Then, a control strategy based on reduced order state feedback linearization can
be designed. Consequently, the capacitance voltage in the LCL filter can be adopted as the intermediate
variable of the reduced order feedback line linearization system. At the same time with system
decoupling control, the single state variable active damping strategy is added to the system to achieve
resonance suppression.

3.1. State Feedback of Inverter Side Inductance and Filter Capacitor Subsystem
Define the inverter side inductance current and the filter capacitor voltage as state variables
iT iT
X = [x1 x2 x3 x4 ]T = i1d i1q ucd ucq , modulation variables as input variables U = u1d u1q , filter
h h
h iT
capacitor voltage as output variables Y = R(X) = ucd ucq .
Electronics 2019, 8, 877 9 of 15

The model of the inverter inductance and the filter capacitance subsystem can be written as
ωx2 − x3 /L1
   
   udc /L1 0 
   
 −ωx 1 − x4 /L1  h i  0 udc /L1 
Equation (2), where f(X) =  , G(X) = g (X) g (X) =  .
 x1 /C + ωx4 − i2d /C   1 2 
 0 0 

  
x2 /C − ωx3 − i2q /C 0 0

The grid-side inductance current can be considered as a measurable feedback variable in the system.
The model is a 2-input 2-output affine nonlinear system with 4 state variables (i.e., n = 4).
According to the aforementioned analysis method, the 4-dimensional vector field matrix can be
calculated as follows:
h i
D4 = g1 g2 ad f g1 ad f g2
 
 udc /L1 0 0 −ωudc /L1 
udc /L1 ωudc /L1 (21)
 
 0 0 
=  
 0 0 −udc /L1 C 0 



0 0 0 −udc /L1 C


From (21), it can be obtained that RankD4 = 4 = n. The calculation result shows that the vector
fields are involution, which satisfies the state feedback linearization condition.
The Lie derivative of the system output (i.e., r1 (X) = ucd ) can be calculated as

∂r1 (X)



 L g1 r1 (X) = ∂X
g1 ( X ) = 0
∂r1 (X)


L g2 r1 (X) = ∂X g2 (X) = 0




∂r1 (X)

f (X) = C1 x1 + ωx4 − C1 i2d


 L f r1 (X) = ∂X (22)
∂(L f r1 (X))

u

g1 (X) = L dcC , 0



 L g1 L f r1 (X) = ∂X
 1
∂(L f r1 (X))


L g2 L f r1 (X) = g ( X ) = 0


∂X 2

In Equation (22), the total relationship of the output variables of the system is d = d1 + d2 = 4,
so the subsystem can be directly transformed to a linear controllable system through coordinate
transformation. The corresponding decoupling matrix can be written as
" #  udc 
L g1 L f r1 (X) L g2 L f r1 (X)  L1 C 0 
E(X) = =  udc
 (23)
L g1 L f r2 (X) L g2 L f r2 (X) 0 L1 C


Then, the new state variables after the feedback linearization can be expressed as
. ∂r1 (X) .



 z1 = ∂X
X = C1 x1 + ωx4 − C1 i2d
. ∂(L f r1 (X)) .

 1+L1 Cω2 ω udc
z2 = X = 2ω C x2 − L1 C x3 − C i2q + L1 C u1d


∂X

(24)

. ∂r2 (X) .
X = C1 x2 − ωx3 − C1 i2q

z3 =

∂X



. ∂(L f r2 (X)) . 1+L1 Cω2

ω udc
X = − 2ω

z4 = C x1 − L1 C x4 + C i2d + L1 C u1q


∂X

The relationship between the new control variable V and the original one U is as (10), where
 r1   2

 L f r1 (X)   2ω x2 − 1+L1 Cω x3 − ω i2q 
Γ(X) =  r2  =  C L1 C C .
2
L r2 (X)   − 2ω x − 1+L1 Cω x + ω i 

f C 1 L1 C 4 C 2d
Combining Equation (23) and Equation (10), the original control variable U can be expressed as
 2
 u1d =
L1 C
v − [ i − 1+LL11CCω ucd − ω
L1 C 2ω
C i2q ]

udc 1d udc C 1q


2 (25)
L1 C L1 C 1 + L Cω ω
[− 2ω
 1
 u1q = v − C i1d − L1 C ucq + C i2d ]


udc 1q udc
 Electronics 2019, 8, 877 10 of 15

Through coordinate transformation, the original nonlinear system is transformed to two linear
systems as follows:
 . " . # " #" # " #
 z1 0 1 z1 0

 d
 Z = . = +

 z2 0 0 z2 v1d

 y =z
 .d "1 . # " #" # " # (26)
 z3 0 1 z3 0

 q
 Z = . = +

 z4 0 0 z4 v1q 11 of 16

 y =z
q 3
Similarly, define the grid-side inductance current of the LCL filter as the state variables 𝑿 =
[𝑥 𝑥3.2.
] = 𝑖 Feedback
State 𝑖 of Grid-Side
, the Inductancevoltage
filter capacitor as the input variables 𝑼 = [𝑢
Subsystem 𝑢 ] = 𝑢 𝑢 ,
and the grid side current
Similarly, define as
thethe variable 𝒀 =
output inductance
grid-side 𝑖 𝑖 of the
current . LCL filter as the state variables
X = [x1 x2 ] = i2d i2q , the filter capacitor voltage as the input variables U = u𝜔𝑥 u2q− 𝑒=⁄𝐿ucd ucq ,
T
h iT h iT h iT
iT(2), where 𝑿 = −𝜔𝑥 − 𝑒 ⁄𝐿 , 𝑮 𝑿 =
The model of subsystem can be written as Equation 2d
h
and the grid side current as the output variable Y = i2d i2q .
1⁄𝐿 0 G ( X ) = g ( X ) g ( X ) = 1 L2 0 
[ ] ωx2 − ed /L2
" #
[𝑔 𝑿 𝑔 The
𝑿 ]model
= 1 2   .
1⁄𝐿
0 of subsystem  0 1 (2),
can be written as Equation L2  where (X) =
−ωx1 − eq /L2
,
Then, the original control1/L
"
variable
2 0 U can be expressed as
#
G(X) = [ g1 (X) g2 (X)] = .
0 1/L2
 2U
Then, the original control variableu d = L2be
can − L2ω i2 qas+ ed
v2 dexpressed
 (27)
u2 q = L2v2 q + L2ω i2 d + eq
u2d = L2 v2d − L2 ωi2q + ed
(
(27)
u2q = L2 v2q + L2 ωi2d + eq
3.3. Parameters Design of Double Closed Loop Controller
3.3. Parameters Design of Double Closed Loop Controller
In the reduced order feedback linearization method, the filter capacitor voltage is the input and
In the reduced order feedback linearization method, the filter capacitor voltage is the input
output of the two
and output of the reduced orderorder
two reduced systems,
systems,respectively.
respectively.Then
Then the doubleloop
the double loop control
control system
system is is
constructed
constructed thatthat
thetheouter
outercontrol
controlloop
loop isis the
thegrid
gridside
side current
current control
control andinner
and the the control
inner control loop is
loop is the
the filter
filter capacitor voltagecontrol.
capacitor voltage control.
The Thecontrol inner
control innerloop is the
loop is thefilter
filter capacitor
capacitor voltage
voltage controlcontrol
and canand can be as
be designed designed as a unit
a unit negative
negative feedback
feedback closedclosed loop system.
loop system. Considering
Considering the resonance
the resonance suppression,
suppression, differential
differential feedback feedback
is needed.
Therefore, a double control loop strategy can be designed, in which the d-axis
is needed. Therefore, a double control loop strategy can be designed, in which the d-axis system system is shown in is
shownFigure 3.
in Figure 3.

i1q 2ω ω i2q
ed −
C C
*
i2d L1C u1d 1 i1d icd 1 ucd 1 i2d
v2d L2 v1d udc
udc sL1 sC sL2

i2q ed 1 + L Cω 2
− L2ω 1
L1C

k0 s + 1

Figure 3. Block diagram of the reduced order feedback decoupling control system.
Figure 3. Block diagram of the reduced order feedback decoupling control system.
In Figure 3, the VSI system is transformed to three series integration links. The outer control
In Figure
loop 3, the VSI
is the grid-side system is
inductance transformed
current to three
control, and series integration
the conventional links.
proportional The outer
integral control
controller
loop isiscompatible.
the grid-side inductance current control, and the conventional proportional
The inner control loop is the filter capacitor voltage control. In order to track high integral
controller is compatible. The inner
frequency ripple, a proportional controlcan
controller loop is the filter
be adopted capacitor
to improve voltage
the response control.
speed of theIn order to
system.
trackDue
high frequency
to the influence ripple, a proportional
of the inner control loop oncontroller
the originalcan be structure,
system adopted the
to outer
improve thecontrol
current response
loop is no longer a precise feedback. A feedback term deviation appears. However, compared
speed of the system. Due to the influence of the inner control loop on the original system structure, with the
the outer current control loop is no longer a precise feedback. A feedback term deviation appears.
However, compared with the forward channel gain, the deviation, which can be equivalent to a
feed forward interference, is smaller. Therefore, it can be compensated by a PI controller.
Define the outer loop controller parameters as k2 and k3, the inner loop controller parameter as
k1. Then, the control system can be represented as Figure 4.
 In Figure 3, the VSI system is transformed to three series integration links. The outer control
loop is the grid-side inductance current control, and the conventional proportional integral
controller is compatible. The inner control loop is the filter capacitor voltage control. In order to
track high frequency ripple, a proportional controller can be adopted to improve the response
Electronics 2019, 8, 877 11 of 15
speed of the system. Due to the influence of the inner control loop on the original system structure,
the outer current control loop is no longer a precise feedback. A feedback term deviation appears.
However,
forward compared
channel gain, with the forward
the deviation, channel
which can begain, the deviation,
equivalent which interference,
to a feed forward can be equivalent
is smaller.to a
feed forwarditinterference,
Therefore, is smaller.
can be compensated by aTherefore, it can be compensated by a PI controller.
PI controller.
Define
Define theouter
the outerloop
loopcontroller parametersasask2kand
controller parameters k3 ,kthe
2 and inner
3, the loop
inner controller
loop parameter
controller as k1 . as
parameter
Then, the control system can be represented as Figure
k1. Then, the control system can be represented as Figure 4. 4.

k0 s
* i2d
i2d k2 s + k3 1 1 1
k1
s s s s

Equivalentblock
Figure4.4.Equivalent
Figure block diagram
diagram of
ofcontrol
controlsystem.
system.
Open loop and closed loop transfer function of the system can be written as
Open loop and closed loop transfer function of the system can be written as
k1 (k2 s + k3 )
GOL (s) = 2 2 (28)
s (s + k0 k1 s + k1 )

k1 (k2 s + k3 )
GCL (s) = (29)
s4 + k0 k1 s3 + k1 s2 + k1 k2 s + k1 k3
In the double closed loop current control system shown in (29), the numerator order of the
transfer function is quite different from the denominator order. The parameters coupling degree is
low. Therefore, the parameters can be designed directly according to the system characteristics. As the
parameters design method was analyzed before, this section does not repeat the description.

4. Simulations and Experiments

4.1. Simulation and Experimental Environment

In order to verify the effectiveness of the proposed algorithm, the three-phase static var generator
is used as the controlled object.
Matlab/Simulink simulation software (2010b, MathWorks, Inc., Natick, Massachusetts 01760 USA)
is used to carry out numerical simulation analysis of three-phase three-leg grid-connected inverter
based on feedback linearization. IGBT model in MATLAB/Simulink simulation software is selected
as switch, with internal resistance Ron = 1 mΩ, snubber resistance Rs = 500 kΩ, snubber capacitance
Cs = inf.
Furthermore, in order to verify the feasibility of the algorithm, a 50 kW prototype of three-phase
three-leg inverter is built and the experiment is carried out. The Controller chip of the prototype is
TMS320F28335 (Texas Instruments, Inc., Dallas, Texas 75243 USA). The IGBT module is SKM150GB12V
(Semikron, Ltd., Nuremberg, Germany), with the switching frequency of 10 kHz.
A photograph of the test rig is shown in Figure 5.
Based on the system parameters shown in Table 1, the control parameters of the two feedback
linearized systems are selected, respectively.

1. Based on Equations (17)–(20), selecting the cut-off frequency of the open-loop transfer function
as 750 Hz, the single closed-loop control system parameters can be designed as: k3 = 5000,
k2 = 107 π, k1 = 200k2 , k0 = 104 k2 . At this time, the corresponding phase margin is Pm = 45◦ ,
and the closed-loop bandwidth is 1 kHz.
2. Similarly, the parameters of the double closed loop control system can be designed as: k0 = 2 × 10−4 ,
k1 = 108 , k2 = 5 × 103 , k3 = 102 k3 . Then, the corresponding open loop cut-off frequency is 680 Hz,
phase margin is Pm = 43◦ , and closed-loop bandwidth is 1 kHz.
capacitance Cs = inf.
Furthermore, in order to verify the feasibility of the algorithm, a 50 kW prototype of three-
phase three-leg inverter is built and the experiment is carried out. The Controller chip of the
prototype is TMS320F28335 (Texas Instruments, Inc., Dallas, Texas 75243 USA). The IGBT module is
SKM150GB12V (Semikron, Ltd., Nuremberg, Germany), with the switching frequency of 10 kHz.
Electronics 2019, 8, 877 12 of 15
A photograph of the test rig is shown in Figure 5.

Oscilloscope
Current probes Inverter

Inverter Inverter

Switching power supply

IGBTs and drivers

Control board 13 of 16
Load resistance
Grid Voltage
Inverter

Inverter ug 380 V
Grid-side inductor L2 0.2 mH
Inverter side
L1 0.3 mH
inductor Inverter

Filter capacitor C 20 uF
Switching frequency
Inductance f 10 kHz

1. Based on Equations (17)–(20), selecting the cut-off frequency of the open-loop transfer function
as 750 Hz, the single Figure 5. Photograph
controlof
Photograph
closed-loop the
the experimental
ofsystem
experimental test
test rig.
parameters rig.
can be designed as: k3=5000,
k2=10 π, k1=200k2, k0=10 k2,,,. At
7 4 this1.time, the corresponding
Parameters phase margin is Pm=45°, and the
Based on the system parameters Table
shown in Tablefor1,the
thesystem.
control parameters of the two feedback
closed-loop bandwidth is 1 kHz.
linearized systems are selected,Variables
respectively. Symbols Value
2. Similarly, the parameters of the double closed loop control system can be designed as: k0=2*10-4,
DC Bus Voltage udc 3 650 V
Grid ×10−41. Parameters
k0 = Table
2Voltage k1 = 108 k2for
=u5the
g×10system. 2
k3 =10380k3V
k1=10 , k2=5*10 , k3=10 k 8 ,
3 , , 2 3 . Then, the corresponding
Grid-side inductor
Variables L2
Symbols 0.2 mH
Value 
Inverter side inductor L1 0.3 mH Pm = 43
open loop cut-off frequency
DC Bus isVoltage
680 Hz, phaseudcmargin Pm=43°
is 650 V , and closed-loop
Filter capacitor C 20 uF
bandwidth is 1 kHz. Switching frequency f 10 kHz
The corresponding frequency domain characteristics of two control systems are shown in
Figure 6. corresponding frequency domain characteristics of two control systems are shown in Figure 6.
The

Bode Diagram
Bode Diagram 60
60 Open Loop
Open Loop 40
40 Closed Loop
Closed Loop
Magnitude (dB)
M agnitude (dB )

20
20
0
0
-20
-20
-40
-40
-60
-60 0
90
Phase (deg)

0 -90
P hase (deg)

-90
-180
-180
-270
-270 1 2 3 4
1 2 3 4 10 10 10 10
10 10 10 10
Frequency (Hz)
Frequency (Hz)

(a) (b)
Figure
Figure 6.
6. Bode
Bodediagrams
diagramsofofthe
theclosed-loop
closed-loopcontrol systems:
control (a)(a)
systems: Single closed-loop
Single control
closed-loop system,
control (b)
system,
double closed-loop
(b) double control
closed-loop system.
control system.

4.2. Steady
4.2. Steady State
State Control
Control Performance
Performance
Setting VSI
Setting VSI output
outputcurrent i2 =
currentasas i2 50 A, A,
= 50 the the
three-phase output
three-phase current
output on theon
current basis
the ofbasis
two of
control
two
methods is shown in Figure 7.
control methods is shown in Figure 7.

Ia
Ia 100
100 Ib
Ib
Ic
Ic
50
50
I /A

0
I /A

0
2
2

-50
-50

-100
-100
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
 double closed-loop control system.

4.2. Steady State Control Performance

Setting VSI output current as i2 = 50 A, the three-phase output current on the basis of two
Electronics 2019, 8, 877 13 of 15
control methods is shown in Figure 7.

Ia
Ia 100
100 Ib
Ib
Ic
Ic
50
50

I /A
0
I /A

2
2

-50
-50

-100
-100
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4 t/s
t/s

(a) (b)
Figure 7. Steady-state
Figure current
7. Steady-state waveforms
current waveformsofof voltage sourceinverters
voltage source inverters(VSI)
(VSI) based
based on on
twotwo control
control
schemes:
schemes: (a) (a) Outputcurrent
Output current of the
thesingle closed-loop
single control
closed-loop system,
control (b) output
system, (b)current
outputof current
the double
of the
14 of 16
closed-loop control system.
double closed-loop control system.
According to Figure 7, both the two inverter systems can track the reference current accurately
According to Figure 7, both the two inverter systems can track the reference current accurately
with small current distortion. Compared to single closed loop control, the output ripple of double
with small current distortion. Compared to single closed loop control, the output ripple of double
closed loop control system is smaller.
closed loop control system is smaller.
The prototype is adopted to carry out the experiment. The output current of the system based
The prototype is adopted to carry out the experiment. The output current of the system based on
ontwotwo kinds
kinds of control
of control methodsmethods
is shownisin shown
Figure 8.in
TheFigure
output8. The Total
current output current
Harmonic Total Harmonic
Distortion (THD)
Distortion
with the two(THD) with
control the two
methods control
are 4.36% andmethods are 4.36%The
1.57%, respectively. andcurrent
1.57%, respectively.
ripple The loop
of single closed current
ripple of system
control single based
closedonloop control
feedback system based
linearization onThe
is larger. feedback linearization
main reason is that theiscontrol
larger.system
The main
is
reason is that the control system is highly dependent on the accuracy
highly dependent on the accuracy of the system model. However, the sampling and control delays, of the system model.
However,
which existtheinsampling and control
real systems, cause the delays, whichofexist
inaccuracy in real systems,
the feedback signal. Incause the inaccuracy
addition, of the
the nonlinear
feedback
magnetization curve of the inductance and the equivalent series resistance of the capacitance also the
signal. In addition, the nonlinear magnetization curve of the inductance and
equivalent series accuracy.
affect the model resistanceItof the capacitance
reduces the dampingalso affect
effect of thethe model point
resonance accuracy. It reduces
and increases the the
current ripple.
damping effect of Thethereduced order
resonance double
point andclosed loop controller
increases the current takes the filter
ripple. The capacitor voltagedouble
reduced order as
the intermediate
closed loop controller variable.
takesTo some
the filterextent, the measure
capacitor voltagereduces the coupling relationship
as the intermediate variable. Toamong the
some extent,
control variables and improves the robustness of the system. The steady
the measure reduces the coupling relationship among the control variables and improves the state control performance
is better. of the system. The steady state control performance is better.
robustness

(a) (b)
Figure
Figure8.8.Experimental
Experimentalsteady-state
steady-state current waveforms:(a)
current waveforms: (a)Output
Outputcurrent
current
ofof
thethe single
single closed-loop
closed-loop
control
control experiment,(b)
experiment, (b)output
output current
current of
of the
thedouble
doubleclosed-loop
closed-loopcontrol experiment.
control experiment.

4.3. Dynamic Control Performance

4.3. Dynamic Control Performance
In this paper, a dynamic test is carried out in the form of virtual load. The grid-connected current
In this paper, a dynamic test is carried out in the form of virtual load. The grid-connected
is set as 25 A at the initial time, and then doubles at 0.3 s. The output current of the system is shown in
current
Figureis9.set as 25 A at the initial time, and then doubles at 0.3 s. The output current of the system
is shown in Figure
When 9. current changes, the output current of the two systems fluctuates with little
the load
When the
overshoot. load
After current
a brief changes,
transient the
process, theoutput currentaof
system reaches the
new two state
steady systems fluctuates
within with little
one fundamental
overshoot.
period. Compared to the direct feedback linearization control, the reduced order feedback linearizationone
After a brief transient process, the system reaches a new steady state within
fundamental period. Compared to the direct feedback linearization control, the reduced order
feedback linearization control lacks one closed loop zero, which reduces the response speed.
However, the transition process is relatively smooth. Therefore, it can be seen that the two control
schemes both have fast response and small overshoot.
 In this paper, a dynamic test is carried out in the form of virtual load. The grid-connected
current is set as 25 A at the initial time, and then doubles at 0.3 s. The output current of the system
is shown in Figure 9.
When the load current changes, the output current of the two systems fluctuates with little
Electronics
overshoot.2019, 8, 877a brief transient process, the system reaches a new steady state within
After 14 ofone
15

fundamental period. Compared to the direct feedback linearization control, the reduced order
feedback linearization
control lacks one closedcontrol lacks
loop zero, onereduces
which closedthe
loop zero, speed.
response whichHowever,
reduces the
thetransition
responseprocess
speed.
However, the transition process is relatively smooth. Therefore, it can be seen that the two control
is relatively smooth. Therefore, it can be seen that the two control schemes both have fast response and
schemes both have fast response and small overshoot.
small overshoot.
100
100 Ia
Ia
Ib
Ib
Ic
Ic
50 Iref(amp)
50 Iref(amp)

I /A
0
I /A

2
2

-50
-50

-100
-100 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7
0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 t/s
t/s
15 of 16
15 of 16
(a) (b)

(c)(c) (d)
(d)
Figure
Figure
Figure 9.9.9.
Dynamic
Dynamic
Dynamic current
currentwaveforms
current waveformsof
waveforms VSI
ofofVSI based
VSIbased on
basedonon two
two two control
control
control schemes:
schemes:
schemes: (a)(a)
(a) Output
Output
Output current
current
current of
of of
thethe
single closed-loop
single closed-loop control
closed-loop control system,
control system, (b)
system, (b) output
(b) output current
output current of
current of the
of the double
the double closed-loop
double closed-loop control
closed-loop control system,
control system, (c)
system, (c)
partial
(c) enlarged
partial
partial enlarged detail
enlarged ofof
detail (a),
detail of (d)
(a),
(a), partial
(d)(d) enlarged
partial
partial detail
enlarged
enlarged ofofof(b).
detail
detail (b).
(b).

The experimental conditions

The
The experimentalconditions
experimental conditions are
arethe
the
are same
thesameasas
same those
asthose of the
those of simulation.
ofthe The experimental
the simulation.
simulation. The resultsresults
The experimental
experimental are
results
shown
areare
shownin Figure
inin 10.
Figure Similar to the simulation results, it is shown that both of the two control schemes
shown Figure10.10.Similar
Similartotothe
thesimulation
simulationresults,
results, itit isis shown
shown that
that both of the two
two control
control
have good dynamic performance.
schemes
schemeshave good
have gooddynamic
dynamicperformance.
performance.

(a)(a) (b)
(b)
Figure
Figure
Figure 10.
10.10. Experimentaldynamic
Experimental
Experimental dynamiccurrent
dynamic currentwaveforms:
current waveforms:(a)
waveforms: (a)Output
(a) Outputcurrent
Output current
current of
ofof the
thethe single
single closed-loop
closed-loop
single closed-loop
control
control experiment,
experiment, (b)
(b) output
output current
current of
of the
the double
double closed-loop
closed-loop control
control experiment.
experiment.
control experiment, (b) output current of the double closed-loop control experiment.
5. Conclusions
5. Conclusion
5. Conclusion
In
In order
order to tosolve
solvethe theproblem
problemofofpoor
poorglobal
globalcontrol
controlperformance
performance of of
thethe
controller designed
controller designedby by
In order
small signal to solve theand
modeling problem of poor
harmonic global control
linearization, two performance
universal of the
controller controller
design methodsdesigned
based by
small signal modeling and harmonic linearization, two universal controller design methods based
small signal modeling and harmonic linearization, two universal controller design
on state feedback linearization are proposed for the LCL type VSI. Relation degree of the VSI system methods based
on state feedback linearization are proposed for the LCL type VSI. Relation degree of the VSI
onandstate
thatfeedback linearization
of the reduced are are
order model proposed forrespectively.
analyzed, the LCL type VSI. Relation
Subsequently, degree matrixes
the coupling of the VSI
system and that of the reduced order model are analyzed, respectively. Subsequently, the coupling
system and thatby
are designed ofadopting
the reduced order
the Lie model are
derivative. analyzed,
Through the respectively. Subsequently,
nonlinear transformation, thethe coupling
original
matrixes are designed by adopting the Lie derivative. Through the nonlinear transformation, the
matrixes
system isare designed
mapped to thebyBrunovsky
adoptingstandard
the Lie type,
derivative.
and theThrough
new systemthelinear
nonlinear transformation,
controller can be used to the
original system is mapped to the Brunovsky standard type, and the new system linear controller
original system is mapped to the Brunovsky standard type, and the new system linear controller
can be used to complete the control of the original complex system. Then, the parameters design
can be used to complete the control of the original complex system. Then, the parameters design
ideas of two kinds of controllers are given. The proposed single closed-loop current controller has a
ideas of two kinds of controllers are given. The proposed single closed-loop current controller has a
simple structure, but it is highly dependent on the exact model of the system. The double closed
simple
loop structure,
controller but sets itthe
is highly dependent
filter capacitor on theasexact
voltage modelloop
the inner of the system.
control The double
variable, closed
by which a
loop controller sets the filter capacitor voltage as the inner loop control variable, by which a
Electronics 2019, 8, 877 15 of 15

complete the control of the original complex system. Then, the parameters design ideas of two kinds
of controllers are given. The proposed single closed-loop current controller has a simple structure,
but it is highly dependent on the exact model of the system. The double closed loop controller sets
the filter capacitor voltage as the inner loop control variable, by which a reduced order feedback
linearization control can be achieved step by step, to improve the adaptability and robustness of the
system. The simulation and experimental results show that the two controllers have good steady-state
control performance and dynamic response performance. In practical application, the robustness of
linear control system based on reduced-order feedback is better.

Author Contributions: L.Y. wrote the paper and designed the control method; C.F. contributed to the conception
of the study and designed the simulation model; J.L. helped to perform the analysis with constructive discussions.
Acknowledgments: This work is supported by the National Natural Science Foundation of China (Grant No.
51607179) and the Fundamental Research Funds for the Central Universities (2017QNB01).
Conflicts of Interest: No potential conflict of interest was reported by the authors.

References
1. Farokhian, F.M.; Roosta, A.; Gitizadeh, M. Stability analysis and decentralized control of inverter-based ac
microgrid. Prot. Control Mod. Power Syst. 2019, 4, 65–86. [CrossRef]
2. Yang, H.; Zhang, Y.; Liang, J.; Gao, J.; Walker, P.; Zhang, N. Sliding-mode observer based voltage-sensorless
model predictive power control of pwm rectifier under unbalanced grid condition. IEEE Trans. Ind. Electron.
2018, 65, 5550–5560. [CrossRef]
3. Xu, X.; Casale, E.; Bishop, M. Application guidelines for a new master controller model for statcom control in
dynamic analysis. IEEE Trans. Power Deliv. 2017, 32, 2555–2564. [CrossRef]
4. Bosch, S.; Staiger, J.; Steinhart, H. Predictive current control for an active power filter with lcl-filter. IEEE Trans.
Ind. Electron. 2018, 65, 4943–4952. [CrossRef]
5. Liu, B.; Wei, Q.; Zou, C.; Duan, S. Stability analysis of lcl-type grid-connected inverter under single-loop
inverter-side current control with capacitor voltage feedforward. IEEE Trans. Ind. Inform. 2017, 14, 691–702.
[CrossRef]
6. Jamshidifar, A.; Jovcic, D. Small-signal dynamic dq model of modular multilevel converter for system studies.
IEEE Trans. Power Deliv. 2016, 31, 191–199. [CrossRef]
7. Bao, X.; Zhuo, F.; Tian, Y.; Tan, P. Simplified feedback linearization control of three-phase photovoltaic
inverter with an lcl filter. IEEE Trans. Power Electron. 2013, 28, 2739–2752. [CrossRef]
8. Lamperski, A.; Ames, A.D. Lyapunov theory for zeno stability. IEEE Trans. Autom. Control 2012, 58, 100–112.
[CrossRef]
9. Hwang, C.L.; Fang, W.L. Global fuzzy adaptive hierarchical path tracking control of a mobile robot with
experimental validation. IEEE Trans. Fuzzy Syst. 2016, 24, 724–740. [CrossRef]
10. Kim, D.E.; Lee, D.C. Feedback linearization control of three-phase ups inverter systems. IEEE Trans.
Ind. Electron. 2010, 57, 963–968.
11. Akhtar, A.; Nielsen, C.; Waslander, S.L. Path following using dynamic transverse feedback linearization for
car-like robots. IEEE Trans. Robot. 2017, 31, 269–279. [CrossRef]
12. Choi, Y.S.; Han, H.C.; Jung, J.W. Feedback linearization direct torque control with reduced torque and flux
ripples for ipmsm drives. IEEE Trans. Power Electron. 2015, 31, 3728–3737. [CrossRef]
13. Yang, L.Y.; Liu, J.H.; Wang, C.L.; Du, G.F. Sliding mode control of three-phase four-leg inverters via state
feedback. J. Power Electron. 2014, 14, 1028–1037. [CrossRef]
14. Lascu, C.; Jafarzadeh, S.; Fadali, S.M.; Blaabjerg, F. Direct torque control with feedback linearization for
induction motor drives. IEEE Trans. Power Electron. 2016, 32, 2072–2080. [CrossRef]
15. Yang, S.; Wang, P.; Tang, Y. Feedback linearization based current control strategy for modular multilevel
converters. IEEE Trans. Power Electron. 2017, 33, 161–174.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).