Adaptive Fuzzy Sliding Mode Command-Filtered

Available online at www.sciencedirect.
com
Journal of the Franklin Institute 356 (2019) 1880–1898

www.elsevier.com/locate/jfranklin
Adaptive fuzzy sliding mode command-filtered

backstepping control for islanded PV microgrid with
energy storage system
Dezhi Xu a,∗, Yuchen Dai b, Chengshun Yang b, Xinggang Yan c
a School of Internet of Things Engineering, Jiangnan University, Wuxi 214122, Chin
b School of Electric Power Engineering, Nanjing Institute of Technology, Nanjing 211167, China
c School of Engineering and Digital Arts, University of Kent, Canterbury CT2 7NT, UK.
Received 19 June 2018; received in revised form 8 January 2019; accepted 8 January 2019
Available online 14 January 2019
Abstract
This study focuses on the control of islanded photovoltaic (PV) microgrid and design of a controller
for PV system. Because the system operates in islanded mode, the reference voltage and frequency of AC
bus are provided by the energy storage system. We mainly designed the controller for PV system in this
study, and the control objective is to control the DC bus voltage and output current of PV system. First,
a mathematical model of the PV system was set up. In the design of PV system controller, command-
filtered backstepping control method was used to construct the virtual controller, and the final controller
was designed by using sliding mode control. Considering the uncertainty of circuit parameters in the
mathematical model and the unmodeled part of PV system, we have integrated adaptive control in the
controller to achieve the on-line identification of component parameters of PV system. Moreover, fuzzy
control was used to approximate the unmodeled part of the system. In addition, the projection operator
guarantees the boundedness of adaptive estimation. Finally, the control effect of designed controller was
verified by MATLAB/Simulink software. By comparing with the control results of proportion-integral
(PI) and other controllers, the advanced design of controller was verified.
© 2019 Published by Elsevier Ltd on behalf of The Franklin Institute.
∗ Corresponding author.
E-mail address: xudezhi@jiangnan.edu.cn (D. Xu).
https://doi.org/10.1016/j.jfranklin.2019.01.012
0016-0032/© 2019 Published by Elsevier Ltd on behalf of The Franklin Institute.
D. Xu, Y. Dai and C. Yang et al. / Journal of the Franklin Institute 356 (2019) 1880–1898 1881
1. Introduction
At present, the traditional fossil energy has gradually disappeared, and renewable energy
sources play an important role [1,2]. Photovoltaic (PV), a type of renewable energy source,
has been widely applied [3–5]. PV power generation systems can improve energy efficiency
and power quality effectively, reduce carbon emission and energy consumption, and enhance
power system reliability [6,7]. Fig. 1 shows the topology of an islanded microgrid used by
us. A PV power generation system can be composed of PV panels, maximum power point
tracking (MPPT) equipment, DC/DC converter, voltage source converter (VSC), LC filter,
and transformer [8]. A PV microgrid has two operation modes: grid connected mode and
islanded mode [9]. This paper mainly focuses on the control of islanded PV microgrid. The
PV microgrid operation in islanded mode can provide power for the load away from the large
power grid [10]. However, because the islanded PV microgrid lacks reference voltage and
frequency provided by an external power grid, the main power source in the system should
be added to generate voltage and frequency reference in an AC bus [11,12]. For the selection
of main power source, it cannot be a type of unstable power supply such as PV panels or
wind turbines. Thus, in this study, an energy storage system (ESS) was selected as the main
power source of islanded PV microgrid, providing reference voltage and frequency for AC
bus. Because the output power of the system is directly supplied to the load in islanded PV
microgrid, notably the output power of PV system and the rated power of load cannot be
exactly the same [13,14]. Therefore, the energy storage device not only provides the reference
voltage and frequency for the system, but also absorbs the excess power generated by the PV
system and compensates the load power to achieve the rated power when the output power
of PV system is less than the rated power of load [15]. In the control of ESS, V/f control
strategy was used to stabilize the voltage and frequency of AC bus. In this study, mainly a
controller was designed for the VSC in PV system to control the stability of DC side bus
voltage of VSC and the output active and reactive power of the system.
Some related work has been carried out to improve the control effect of islanded mi-
crogrid. Reference [16] established a mathematical model for VSC in islanded microgrid,
and a fractional-order sliding mode controller was designed for VSC. The designed con-
troller achieved the tracking of three-phase voltage to reference voltage at the end of islanded
microgrid. Therefore, the problem of no reference voltage and frequency in the islanded mi-
crogrid is solved. However, in this study, a DC source was used instead of a distributed
energy resource, and the model established in this study is only suitable for some stable
power sources. In addition, reference [16] also does not consider ESS. The work reported in
Fig. 1. Topology of the islanded microgrid.

1882 D. Xu, Y. Dai and C. Yang et al. / Journal of the Franklin Institute 356 (2019) 1880–1898
reference [17] discusses the application of islanded PV system in a cluster of houses. In the
article, the interconnection of multiple islanded microgrid is discussed, the transmission of
energy in multiple systems is achieved, and the ESS is connected to the AC side to achieve
energy storage and release. However, proportion-integral (PI) controller was selected to control
VSC. When the system is disturbed, it is difficult for the PI controller to exhibit high ro-
bustness. Moreover, other control methods are not compared in this article. In reference [18],
droop control was applied to an islanded AC-DC microgrid. AC bus and DC bus supplied
power to AC and DC loads, respectively; they were connected using an interlinking con-
verter. However, as in reference [17], the article also fails to compare the proposed control
method with other control methods to show the advanced nature of proposed control method.
In reference [19], the voltage and frequency in an islanded system are controlled by a fully
decentralized control framework. This proposed control method is different from centralized
and distributed control. The authors established a sliding mode observer to estimate the infor-
mation of other distributed generator units and verify the reliability of method. However, this
method should model the control object accurately; therefore, the control strategy still needs
improvement.
In recent years, advanced control has been applied in many engineering practices [20–22].
Advanced control includes predictive control, fuzzy control, neural control, nonlinear con-
trol, and robust control, and it has a better control effect than the traditional PI controller.
Section 2 shows that the mathematical model established by us is a second-order model.
Thus, the virtual controller was constructed by using backstepping control. Backstepping con-
trol is a type of nonlinear control, and it is widely used to control higher-order models
[23]. To reduce the differential process of controller, a command-filter was introduced in the
controller [24]. The work reported in [25] applies command-filtered backstepping control to
nonlinear multiagent tracking problems. The controller was designed, and the stability of the
system was verified based on Lyapunov stability theory. The simulation shows that the de-
signed controller has satisfactory control performance. In the model established for the PV
system, the parameters of resistance, inductance, and capacitance in the circuit are included.
However, the measurement of these parameters cannot be very precise [26]. Therefore, the
adaptive estimation of circuit parameters was considered in the controller. In reference [27],
a command-filtered backstepping controller was designed for compliant actuators robot arms
with parameter estimation. The controller adaptively estimates the parameters, while achiev-
ing the control objectives. The simulation shows that the controller with parameter adaptive
controller reduces the chattering of control torque compared with PI and sliding mode con-
troller. Therefore, the performance of controller is improved. In addition, the incompleteness
of PV system model was also considered; the chattering can easily stimulate the unmodeled
characteristics in the system. To solve this problem, fuzzy control was used to approximate
the unmodeled part of system and improve the performance of controller [28,29]. Sliding
mode control was also introduced in the controller, thus further increasing the robustness of
controller [30].
In brief, in this study, mainly an adaptive fuzzy sliding mode command-filtered backstep-
ping (AFSCB) controller was designed for the PV system. This controls the DC bus voltage
and output power of PV system. The sections in this article are arranged as follows: In
Sections 2 and 3, the topology of entire system is derived, and a mathematical model of
PV system is derived. In Section 4, the design flow of AFSCB controller is described. In
Section 5, the simulation model is built, and the output waveform of the system is provided.
Summary and conclusions are provided in Section 6.
Fig. 2. Structure of islanded microgrid with PV array and ESS.
2. Structure of islanded microgrid with PV array and ESS
The structure of islanded microgrid with PV array and ESS is shown in Fig. 2. The PV
system and ESS are connected to the AC bus directly and supply power to the load. In the
PV system, PV array was used instead of DC source to simulate the actual operation of PV
system. The entire system is in the islanded operation mode, and it is not connected to the
external power grid.
2.1. The structure of ESS
As shown in Fig. 2, in the design of ESS topology in this article, a battery was selected as
the energy storage equipment, and ESS was defined as the main power source of the entire
system to provide reference voltage and frequency for the AC bus. In the DC side of ESS,
the battery is connected to the VSC through a two-way buck-boost circuit. The buck-boost
circuit is controlled by PI control; the control objective is to increase the output voltage of
the battery and achieve constant voltage charging and discharging. The flow chart of control
of buck-boost circuit is shown in Fig. 2.
In the AC side of ESS, VSC was used to turn direct current into three-phase alternating
current. Resistance R2 represents the on-state resistance of VSC; inductance L2 and capacitance
C2 constitute the LC filter. The VSC in ESS is controlled by V/f controller. V/f controller
was selected because we need to provide stable voltage and frequency for the AC bus in
islanded microgrid. The reference voltage and frequency for the V/f controller were specified,
and the controller achieved the tracking of ESS output voltage and frequency to the given
values. Unlike the PV system, the rotation angle ρ in ESS is provided by a voltage-controlled
oscillator (VCO).
2.2. The structure of PV system
The controller of ESS is not what we want to discuss emphatically. We only use PI con-
troller to realize the function of ESS. This article mainly discusses the design and performance
improvement of PV system controller.
Fig. 3. Equivalent circuit of PV cell.
Fig. 4. P-U curve of PV cell.
2.2.1. Structure of PV array

In our simulation, the equivalent circuit of PV cell is used to simulate the output char-
acteristics of PV cell. The equivalent circuit of PV cell is shown in Fig. 3. PV cell can be
considered as a current source. According to Fig. 3, the output current I of the PV cell has
the following relations [31]:

q (U + I Rs ) U + I Rs
I = I ph − Is exp −1 − , (1)
AkT Rsh
where Iph is photo-generated current. q, k are the constants which q = 1.602 × 10−19C and
k = 1.381 × 10−23 J/K . A and T are the characteristic coefficient of the diode and ambient
temperature. It needs to be explained that the output current of a single PV cell is very small,
so we make a number of PV cells in the simulation to make up the 100 kW PV array to
provide power for the system.
2.2.2. Incremental conductance maximum power point tracking

As shown in Fig. 4, the output voltage and power of PV array are not linear when irradiance
and temperature remain constant. Therefore, under fixed irradiance and temperature, an output
voltage Ump corresponding to the maximum power Pmp output of PV array exists. This point
is known as the maximum power point (MPP). MPP changes with the change in irradiance
and temperature; MPP is tracked using an incremental conductance MPPT controller.
From Fig. 4, we know that the derivative of P-U curve is 0 at MPP. The incremental
conductance MPPT algorithm is calculated as follows [31]. The output power P of the PV
array in Fig. 3 can be expressed as
P = U I. (2)
Find the derivative of Eq. (2), one obtains:

dP dI
= I +U . (3)
dU dU
When Eq. (3) is equal to 0, the U at this time is the voltage corresponding to MPP, which
expressed as
I
I +U = 0. (4)
U
2.2.3. AC side
Similarly, we use VSC to connect the DC side and the AC side. R1 represents the turn-on
resistance of VSC, L1 and C1 constitute the LC filter. utabc , uabc and iabc are the three-phase
terminal voltage of VSC and PV system and the three-phase output current, respectively. The
AFSCB controller is designed in dq-frame, therefore, we converted the three-phase variables
usabc and isabc to the dq-frame and provide to the controller. Because of the voltage and
frequency of the AC bus is controlled by ESS, thus the rotation angle ϕ is provided by
phase locked loop (PLL), which is the same as the rotation angle in AC bus. Modulation
signals τ d and τ q in dq-frame is produced by AFSCB controller, then, the signals τ d and τ q
are converted to the abc-frame to form three-phase modulation wave τ abc . Finally, we use
10 kHz triangular wave and three-phase modulation wave τ abc to generate six-pulse signals
to control VSC, so as to achieve control objectives.
3. Dynamic mode of PV system and fuzzy algorithm
3.1. Dynamic mode of PV system
The dynamic model of PV system is established in this section. From PV system in Fig. 2,
by using Kirchhoff’s voltage law, the three-phase voltage and current in the system has the
following relationship:
−
→
d is −
→ → −
L1 = −R1 is − − us + →
ut , (5)
dt
−
→ →
where vector is , − ut and −→
us are the space vector of the three phase variable isabc , utabc and
usabc . Furthermore, the terminal voltage of the VSC −
→
ut can be expressed as −→ut = (udc /2)− →τ,
−
→
where τ is the three-phase PWM modulating wave. So Eq. (5) can be rewritten as:
−
→
d is −
→ → udc −
L1 = −R1 is − −us + →τ. (6)
dt 2
Then, convert Eq. (6) to dq-frame from abc-frame, one obtains
d isd R1 usd udc
= − isd + ωisd − + τd , (7)
dt L1 L1 2 L1
d isq R1 usq udc
= − isq − ωisq − + τq . (8)
dt L1 L1 2 L1
According to the power conservation theorem (ignoring the impedance of Cdc , VSC, and
LC filter), the power in the system has the following relation [32]:
3
udc i pvout = (usd isd + usq isq ), (9)
2
where i pvout in Fig. 2 represents the output current of the buck-boost circuit. Then, the calcu-
lation of the current relation at the capacitance Cdc is calculated as
d udc
Cdc = i pvout − idc . (10)
dt
As a result, by combining Eqs. (9) and (10), we can get the following relation:
d udc 3(usd isd + usq isq ) idc
= − . (11)
dt 2Cdc udc Cdc
Integrating the Eqs. (11), (7), and (8), we get the basic mathematical model of PV sys-
tem. In the design of controller, we consider the adaptive estimation of unmodeled part and
parameter. Therefore, the mathematical model of PV system is rewritten as follows:

d udc 3(usd isd + usq isq )
= η1 − idc + δ1 ,
dt 2udc
d isd udc
= −η2 isd + ωisq − η3 usd + η3 τd + δ2 ,
dt 2
d isq udc
= −η2 isq − ωisd − η3 usq + η3 τq + δ3 , (12)
dt 2
where η1 = 1/Cdc , η2 = R1 /L1 , η3 = 1/L1 , and δ 1 , δ 2 , δ 3 represent the unmodeled nonlinear
parts of the dynamic model of PV system.
3.2. Fuzzy algorithm
Fuzzy algorithm is applied in this paper to deal with the unmodeled part of PV system.
Using the following fuzzy rules to construct a fuzzy system [33]:
I F x1 is Al11 and . . . and xn is Alnn , T hen ud is S l1 ...ln , (13)
where l1 = 1, 2, . . . , m, i = 1, 2, . . . , n. The fuzzy controller can be calculated as

N
n
θi μij (x j )
i=1 j=1
f (x) = = θ T ξ(x), (14)

N
n
μij (x j )
i=1 j=1

n
μij (x j )
where ξ(x) = [ξ1 (x ), ξ2 (x ), . . . , ξN (x )] , in whicn ξi (x) =
T
N
j=1

n , μij (x j ) is the mem-
i=1 μij (x j )
j=1
bership degree of fuzzy system, and θ = [θ1 , θ2 , . . . , θN ]T is the ideal constant weight vector.
According to the fuzzy universal approximation theorem, if y(x) is defined as a function
of compact set , for any constant μ > 0, there exists a fuzzy system satisfying sup f (x)∈ =
|y(x) − f (x)| ≤ μ.
Fig. 5. Structure of the command-filter.
4. Design process of AFSCB controller
4.1. Controller design
In this section, we design the AFSCB controller τ d and τ q for the PV system in dq-frame.
First of all, we approximate the unmodeled part by fuzzy function:
δ1 = θ1T ξ1 (x), δ2 = θ2T ξ2 (x), δ3 = θ3T ξ3 (x). (15)
We define θ˜b = θˆb − θb (b = 1, 2, 3) as an estimated error, where θˆb are the adaptive esti-
mated value of θ b , and δˆb = θˆbT ξb (x) are the fuzzy approximation value of δ b .
For parameters ηa , (a = 1, 2, 3), we define adaptive estimation error as ηã = ηâ − ηa ,
where ηâ are the estimate of ηa .
Then, the tracking error of DC voltage is defined as
e1 = udc − udcre f , (16)
We choose the Lyapunov function as
1 2
V1 = e , (17)
2 1
and the derivative of V1 can be calculated as

V˙1 = e1 e˙1 = e1 u˙dc − u˙dcre f

3 u d i d + u q i q
= −c1 e21 + e1 η1 − idc + θ1T ξ1 (x) − u˙dcre f + c1 e1 (18)
2udc
where c1 > 0 is a designed constant. To stabilize V1 , according to Lyapunov asymptotic sta-
bility theory, the virtual controller considering parameter estimation and fuzzy approximation
is selected as
2udc
u i
iˆsd −c1 e1 − θˆ1T ξ1 (x) + u˙dcre f + ηˆ1 idc −
sq sq
d
= . (19)
3usd ηˆ1 usd
In order to reduce the computational complexity of the controller, an command-filter in
the controller introduced in the controller, and the structure of the command-filter is shown
in Fig. 5.
The state equation of the command-filter can be written as [34]:
c

y˙
c
y˙
= σni (20)
ÿc 2ζi SR 2ζ i
SM (yd ) − yc − y˙c
where yd and yc are the input and output signal of command-filter respectively, the waveform
of the two signals should be the same. The σ ni and ζ i are the bandwidth and damping of
the filter. In this paper, we choose σni = 300 and ζ i =0.1. If we design the virtual controller
d
isd as the input of the command-filter, then we can get the derivative of virtual controller
i˙sd
c
through integral part, thereby reducing the amount of calculation of the controller. The
filtering error ε is defined as
3usd ηˆ1 c
ε˙ = −c1 ε + isd − isd
d
. (21)
2udc
Then, redefine the DC voltage tracking error and the current tracking error:
ē1 = e1 − ε, e2 = isd − isd
c
, e3 = isq − isq
d
, (22)
and redefine Lyapunov function to stabilize ē1 simultaneously:
1 2
V2 = ē . (23)
2 1
According to Eqs. (18), (19), (21) and (22), the derivative of V2 can be calculated as
V˙2 = ē1 ē˙1 = ē1 (e˙1 − ε˙ )

3usd isd η1 3usq isq η1
= ē1 + − η1 idc + θ1T ξ (x) − u˙dcre f + c1 ε1
2udc 2udc

3usd ηˆ1 c 3usq isq ηˆ1
− isd − c1 e1 − θˆ1T ξ (x) + u˙dcre f + ηˆ1 idc −
2udc 2udc

3 u η
ˆ 3( u sd isd + usq isq )
ē1 e2 − ē1 θ˜1T ξ (x) + ē1 η˜1 idc −
sd 1
= −c1 ē21 + . (24)
2udc 2udc
Defining the sliding mode surfaces in d-frame and q-frame as
S1 = k1 ē1 + e2 , S2 = k2 e3 . (25)
In order to facilitate the computation, calculating the derivatives of S1 and S2 first as
S˙1 = k1 ē˙1 + e˙2

udc

= k1 ē˙1 + η˜2 − ηˆ2 isd + ωisq + η˜3 − ηˆ3 usd + ηˆ3 − η˜3 τd + θˆ2T − θ˜2T ξ2 (x) − i˙sd
c
,
2
(26)
S˙2 = k2 e˙3

udc

= k2 η˜2 − ηˆ2 isq − ωisd + η˜3 − ηˆ3 usq + ηˆ3 − η˜3 τq + θˆ3T − θ˜3T ξ3 (x) − i˙sq
c
, (27)
2
where k1 > 0, k2 > 0 are the gains of the sliding mode surfaces.
In order to stabilize the entire system, the Lyapunov function has been defined once again
as
1 1 η˜ 2 η˜ 2 η˜ 2 θ˜T θ˜1 θ˜T θ˜2 θ˜T θ˜3
V3 = V2 + S12 + S22 + 1 + 2 + 3 + 1 + 2 + 3 , (28)
2 2 2 r1 2 r2 2 r3 2 λ1 2 λ2 2 λ3
where ra > 0, λb > 0, (a, b = 1, 2, 3) are the designed gains of adaptive law. Combination
Eqs. (24), (26) and (27), the derivative of V3 can be calculated as
˙ ˙ ˙
η˜1 η˙ˆ1 η˜2 η˙ˆ 2 η˜3 η˙ˆ 3 θ˜T θˆ1 θ˜T θˆ2 θ˜T θˆ3
V˙3 = ē1 ē˙1 + S1 S˙1 + S2 S˙2 + + + + 1 + 2 + 2
r1 r2 r3 λ1 λ2 λ3
3usd ηˆ1
= −c1 ē21 + ē1 e2 − c2 S12 − c3 S1 sat (S1 ) − c4 S22 − c5 S2 sat (S2 )
2udc

udc
+ S1 k1 ē˙1 − ηˆ2 isd + ωisq − ηˆ3 usd + ηˆ3 τd + θˆ2T ξ2 (x) − i˙sd
c
+ c2 S1 + c3 sat (S1 )
2
udc 1 1
+ k2 S2 −ηˆ2 isq − ωisd − ηˆ3 usq + ηˆ3 τq + θˆ3T ξ3 (x) − i˙sq
d
+ c4 S2 + c5 sat (S2 )
2 k2 k2

η˜1 ˙ 3(usd isd + usq isq ) η˜2 ˙

+ ηˆ 1 − r1 ē1 − idc + ηˆ 2 − r2 −isd S1 − k2 isq S2
r1 2udc r2
η˜3 ˙

udc
udc
+ ηˆ 3 − r3 S1 −usd + τd + k2 S2 −usq + τq
r3 2 2
θ˜T
˙ θ˜T
˙
θ˜T
˙

+ 1 θˆ1 − λ1 (ē1 ξ1 (x) ) + 2 θˆ2 − λ2 (S1 ξ2 (x) ) + 3 θˆ3 − λ3 (k2 S2 ξ3 (x) ) , (29)
λ1 λ2 λ3
where c2 , c3 , c4 , c5 > 0 are the designed controller parameters. According to Eq. (29), in order
to ensure the convergence of the system, the controllers in d-frame and q-frame are designed
respectively:
2

τd = −k1 ē˙1 + ηˆ2 isd − ωisq + ηˆ3 usd −θˆ2T ξ2 (x)+i˙sd
c
− c2 S1 − c3 sat (S1 ) , (30)
ηˆ3 udc
2

τq = k2 ηˆ2 isq + ωisd + ηˆ3 usq − θˆ3T ξ3 (x) +i˙sqd
− c4 S2 − c5 sat (S2 ) . (31)
k2 ηˆ3 udc
The function sat(.) represents the saturation function which expressed as
⎧
⎨1, S>ψ
sat (S) = S/φ, |S | ≤ ψ (32)
⎩
−1, S < −ψ
where 0 < ψ ≤ 0.5 is the layer of the sliding surface.
According to Eq. (29), the adaptive laws of entire system are designed as follows:

˙ηˆ = r1 Proj ηˆ1 , ē1 3(usd isd + usq isq ) − idc − m1 ηˆ1 ,
1
2udc

˙ηˆ = r2 Proj ηˆ2 , − isd S1 − k2 isq S2 − m2 ηˆ2 ,
2

udc
udc
ηˆ˙ 3 = r3 proj ηˆ3 , S1 −usd + τd +k2 S2 −usq + τq − m3 ηˆ3 ,

2 2
˙ˆ
θ 1 = λ1 Proj θˆ1 , ē1 ξ1 (x) − n1 θˆ1

˙
θˆ2 = λ2 Proj θˆ2 , S1 ξ2 (x) − n2 θˆ2

˙
θˆ3 = λ3 Proj θˆ3 , k2 S2 ξ3 (x) − n3 θˆ3 (33)
where ma > 0, nb > 0 (a, b = 1, 2, 3) are the designed parameters. The function Proj(.) in
(33) represents the projection operator (see in reference [35]), which guarantees that the
adaptive parameter estimation is bounded.
4.2. Stability proof
Substituting Eqs. (30)–(33) into Eq. (29), the V˙3 is computed as

3usd ηˆ1
V˙3 = −c1 ē21 + ē1 e2 − c2 S12 − c3 S1 sat (S1 ) − c4 S22
2udc
3
η˜ T ηâ θ˜bT θˆb
3
− c5 S2 sat (S2 ) − ma a − nb . (34)
a=1
ra b=1
λb
3usd ηˆ1
Define 0 < μ < 1, then, from Eq. (34) we can obtain that −c1 ē21 + 2udc 1 2
ē e≤ −c1 (1 −
3usd ηˆ1 3usd ηˆ1
μ)ē1 − c1 μ|ē1 | + 2udc |ē1 ||e2 |. If we make 2udc |ē1 ||e2 | − c1 μ|ē1 | ≤ 0, that is |ē1 | ≥
2 2 2
3usd ηˆ1 |e2 |

2c1 μudc
. Then we can get the conclusion that −c1 ē21 + 32uusddcηˆ1 ē1 e2 ≤ −c1 (1 − μ)ē21 . Moreover,
we can know that c3 S1 sat(S1 ) ≥ 0, c5 S2 sat(S2 ) ≥ 0 according to the definition in this article.
Therefore, combining with the above conclusions, V˙3 has the following inequalities relation-
ship:

3
ηãT ηâ θ˜bT θˆb
3
V˙3 ≤ −c1 (1 − μ)ē21 − c2 S12 − c4 S22 − ma − nb . (35)
a=1
ra b=1
λb
Furthermore, according to the Young’s inequality, we can obtain that:

θ˜T θ˜b θ T θb
−θ˜bT θˆb = −θ˜bT θ˜b + θb ≤ −θ˜bT θ˜b + b + b
2 2
θ˜bT θ˜b θbT θb
≤− + , (b = 1, 2, 3), (36)
2 2
similarly:
ηãT ηã η T ηa
−ηãT ηâ ≤ − + a , (a = 1, 2, 3), (37)
2 2
Substituting Eqs. (36) and (37) in (35), one obtains:
V˙3 ≤ −c1 (1 − μ)ē21 − c2 S12 − c4 S22
3
ma ηãT ηã ma ηaT ηa
3
nb θ˜bT θ˜b nb θbT θb
− − − −
a=1
2 ra 2 ra b=1
2 λb 2 λb
≤ −αV3 + β (38)
3 ma ηaT ηa a nb θbT θb
where α = min{2c1 (1 − μ), 2c2 , 2c4 , ma , nb}, a, b = 1, 2, 3, β = + a=1 2ra b=1 2λb .
According to the conclusion of (38), and referring to the general solution formula of the
first order differential equation, we can get
t0 t0 t0
− αdx − αdx αdx β
V3 ≤ C e t +e t βe t dx = C e−α(t−t0 ) + . (39)
α
Put (t0 , V(t0 )) into (39), we can get C = V3 (t0 ) − βα . Furthermore, due to α > 0, β > 0, we
can get, βα > 0 and e−α(t−t0 ) < 1. Finally, we can get the conclusion that:

β −α(t−t0 ) β β
V3 (t ) ≤ V3 (t0 ) − e + ≤ V3 (t0 ) + , ∀t ≥ t0 . (40)
α α α
Fig. 6. Signal flow diagram of the AFSCB controller.
Fig. 7. Simulation model of islanded PV microgrid with ESS.
Table 1
Parameters of PV system.
Parameters Value Description

L pv 5 mH PV inductor
C pv 100 μF PV capacitance
Cdc 6 mF DC-link capacitance
R1 2 m Filter resistance
L1 250 μH Filter inductor
C1 500 μF Filter capacitance
Tr1 0.26/25 kV/kV Transformer ratio
In conclusion, it is proved that the system is bounded and stable. The signal flow diagram
of the AFSCB controller is shown in Fig. 6.
5. Simulation results and studies
In this section, we built the 100 kW islanded PV power generation system with ESS in
MATLAB/Simulink, see in Fig. 7, and the whole simulation lasts 2 s. It is worth noting
that we used the per-unit value in the calculation of the controller in order to simplify the
calculation. Then, the output signal calculated by the controller is amplified and to control
the VSC.
The circuit parameters in PV system and ESS are shown in Tables 1 and 2, the parameters
of designed control laws are shown in Table 3. The membership function of the fuzzy system
is defined as: μij = exp[−(x j + l × 2 )2 /3], where l = −5, −4, . . . , 0, . . . , 5, see in Fig. 8.
Table 2
Parameters of ESS.

Lb 1 mH Battery inductor
Cb 150 μF Battery capacitance
Cedc 6 mF DC-link capacitance
R2 3 m Filter resistance
L2 300 μH Filter inductor
C2 500 μF Filter capacitance
ω 100π rad/s Nominal angular frequency
Tr2 0.45/20 kV/kV Transformer ratio
Table 3
Parameters of designed control laws.

c1 , c2 , c3 , c4 , c5 1, 300, 360, 1200, 1000 Gains of AFSCB controller
r1 , r2 , r3 30, 20, 5 Gains of adaptive laws
m1 , m2 , m3 0.01, 0.01, 0.01 Gains of adaptive laws
λ1 , λ2 , λ3 50, 100, 100 Gains of adaptive fuzzy laws
n1 , n2 , n3 0.01, 0.01, 0.01 Learning rate of fuzzy systems
k1 , k2 1.2, 1.2 Gains of sliding mode surface
Membership function degree
0.8
0.6
0.4
0.2
0
-10 -5 0 5 10
x
Fig. 8. Membership function of fuzzy system.
a b
1000 60
Irradiance(W/m 2)
Temperature(° C)
800 50
600 40
400 30
200 20
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time(s) Time(s)
Fig. 9. The curves of irradiance and temperature.

a b
173 9
8.5
172.5
8
172
7.5
171.5 7
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time(s) Time(s)
c 4002
4001.5
4001
4000.5
4000
0 0.5 1 1.5 2
Time(s)
Fig. 10. Parameters adaptive estimation curves.
a b
0.01 0.01
0.005 0.005
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time(s) Time(s)
c
0.015
0.01
0.005
0
0 0.5 1 1.5 2
Time(s)
Fig. 11. Fuzzy adaptive estimation curves.
In ESS, the output voltage of ESS was used to track the reference voltage (450 V), and then
the voltage through the transformer was increased to stabilize the AC bus voltage at 20 kV. In
PV system, to observe the change in the output power of PV system under different irradiance
and temperature, the irradiance and temperature curves are shown in Fig. 9. At t = 1.6 s, the
irradiance and temperature remain stable.
Fig. 10 shows the parameter adaptive estimation curves. As shown in Fig. 10, the estimated
curve fluctuates near the real value and shows a convergent trend. Fig. 11 shows the fuzzy
a
100
75
100
50 100
96.7
25 91.2
0.45 0.5 0.55 1.75 1.8 1.85
0
b
100
75
99.6
50 100
99.1
98
25 98.6 1.75 1.8 1.85
0.45 0.5 0.55
0
c
100
75
100.3
50 100.3
99.8
99.8
25 99.3 99.3
0.45 0.5 0.55 1.75 1.8 1.85
0
0 0.5 1 1.5 2
Time (s)
Fig. 12. Output active power curve of PV system (a) under PI control, (b) under SCB control, (c) under AFSCB
control.
adaptive estimation curves. The estimation of fuzzy system actually compensates the dynamic
changes in the system. The fuzzy system estimated the unmodeled part of the system and
improved the performance of the system by matching with parameter adaptive.
Fig. 12 shows the output active power waveform of PV system under different control
methods. The controller in PV system intervened at t=0.05 s. When t=0.4 s, MPPT controller
starts to work, and the system works at the MPP. When t=0.6C1.6 s, the output power varies
with irradiation and temperature, and at t=1.6C2 s, the output power of the system is stable
at 100 kW. At t=0C1.8 s, the load is 42 kW; at t=1.8C2 s, the load mutation to 128 kW.
Fig. 12(a) shows that the output power under PI controller has large chattering; in addition,
the load mutation affects the output power at t=1.8 s. Fig. 12(b) shows that the sliding mode
command-filtered backstepping (SCB) controller has better performance than the PI controller.
The output power curve under SCB controller is smoother than that under PI controller, but
there is still chattering. Moreover, when the load was mutated at t=1.8 s, the impact still
exists. Fig. 12(c) shows the output power curve under AFSCB controller. AFSCB has the best
dynamic characteristics compared with PI and SCB controller. Notably, parameters adaptive
and fuzzy system further weaken the chattering of output power and eliminate the effect of
load switching on output power.
a b
40
20
100
0
50 -20
-40
0 -60
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time(s) Time(s)
c
80.001
SOC (%)
80
0 0.5 1 1.5 2
Time(s)
Fig. 13. (a) The power absorbed by the load, (b) output power of ESS, (c) the SOC of battery.
500
250
0
0 0.5 1 1.5 2
Time(s)
Fig. 14. DC bus voltage of PV system under AFSCB controller.
Fig. 13(a) shows the load absorbed power, and Fig. 13(b) shows the output power of ESS.
The battery is absorbing excess power from the PV system when Pbat is negative. When Pbat
is positive, it represents that the PV system cannot supply the rated power of load needed,
and ESS discharges to compensate the power shortage. Fig. 13(c) shows the state of charge
(SOC) of the battery.
Fig. 14 shows the DC bus voltage of PV system under AFSCB controller. Under the
effect of AFSCB controller, udc achieved the tracking of given reference voltage udcre f =
500 V regardless of the changes in irradiance and temperature. Fig. 15 shows the three-phase
voltage and current of AC bus at t=1.85C1.95 under AFSCB controller. Under the control of
controller, the voltage and current are stabilized. Fig. 16 shows the total harmonic distortion
(THD) of AC bus current under PI, SCB, and AFSCB controller. The current controlled
by AFSCB controller has the minimum THD, further verifying the control performance of
AFSCB controller.
a 104
2
1
0
-1
-2
b 5
-5
1.85 1.9 1.95
Time(s)
phase-a phase-b phase-c
Fig. 15. Three-phase (a) voltage and (b) current of AC bus under AFSCB controller.
a b
Fundamental (50Hz) = 4.297 , THD= 4.19% Fundamental (50Hz) = 4.295 , THD= 1.62%
Mag (% of Fundamental)
2 1
1 0.5
0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Frequency (Hz) Frequency (Hz)
c
Fundamental (50Hz) = 4.292 , THD= 0.53%
0.3
0.2
0.1
0
0 200 400 600 800 1000
Frequency (Hz)
Fig. 16. THD of AC bus current (a) under PI controller, (b) under SCB controller, (c) under AFSCB controller.
6. Conclusion
In this study, a 100 kW PV system with ESS was established in MATLAB/Simulink en-
vironment software. In the VSC in ESS, a traditional PI based V/f controller was used to
stabilize AC bus voltage, and reference voltage and frequency were provided for the PV sys-
tem. Moreover, ESS implemented the storage and release of energy. The AFSCB controller
designed in this paper takes full account of the state and condition of the PV system, and de-
signs control strategies one by one for the problems that may occur during the operation. The
AFSCB controller is designed based on backstepping control method, and a command-filter
was introduced to eliminate the differential expansion of traditional backstepping control. To
increase the robustness of system, a sliding mode control was also introduced in the controller.
In addition, to increase the dynamic characteristics of the system, adaptive estimation of the
parameters was considered. Furthermore, to reduce the effect of unmodeled part of the system
on the control effect, fuzzy function was used to approximate the unmodeled part of system.
The simulation results show that the system under AFSCB controller has better dynamic char-
acteristics and higher power quality compared with PI and SCB controller. In general, this
paper explores the direction for the future development of photovoltaic system controllers. In
the future work, we will further verify the effectiveness of the designed controller through
experiments.
Acknowledgment
This work is supported by National Natural Science Foundation of China (61503156) and
National Key Research and Development Program (2016YFD0400301), Open Research Fund
of Jiangsu Collaborative Innovation Center for Smart Distribution Network, Open Research
Fund of Jiangsu Collaborative Innovation Center for Smart Distribution Network, Nanjing
Institute of Technology (XTCX201806), National first-class discipline program of Food Sci-
ence and Technology (JUFSTR20180205), and State Grid Zhejiang Province Technology
Project (SGTYHT/17-JS-201).
References
[1] B.M. Eid, N.A. Rahim, J. Selvaraj, et al., Control methods and objectives for electronically coupled distributed
energy resources in microgrids: a review, IEEE Syst. J. 10 (2) (2016) 446–458.
[2] D. Lew, M. Asano, J. Boemer, et al., The power of small: the effects of distributed energy resources on system
reliability, IEEE Power Energy Mag. 15 (6) (2017) 50–60.
[3] Y.K. Wu, J.H. Lin, H.J. Lin, Standards and guidelines for grid-connected photovoltaic generation systems: a
review and comparison, IEEE Tran. Ind. Appl. 53 (4) (2017) 3205–3216.
[4] P. Li, X. Yu, J. Zhang, et al., The H∞ control method of grid-tied photovoltaic generation, IEEE Trans. Smart
Grid 6 (4) (2015) 1670–1677.
[5] C. Kuo, J. Chen, S. Chen, et al., Photovoltaic energy conversion system fault detection using fractional-order
color relation classifier in microdistribution systems, IEEE Trans. Smart Grid 8 (3) (2017) 1163–1172.
[6] X. Liu, M. Shahidehpour, Y. Cao, et al., Microgrid risk analysis considering the impact of cyber attacks on
solar PV and ESS control systems, IEEE Trans. Smart Grid 8 (3) (2017) 1330–1339.
[7] M.E.B. soǧlu, B. Çakir, Hybrid global maximum power point tracking approach for photovoltaic power opti-
misers, IET Renew. Power Gen. 12 (8) (2018) 875–882.
[8] F. Jaramillo-Lopez, G. Kenne, F. Lamnabhi-Lagarrigue, Adaptive control for a class of uncertain nonlinear
systems: application to photovoltaic control systems, IEEE Trans. Autom. Control 62 (1) (2017) 393–398.
[9] J. Li, Y. Liu, L. Wu, Optimal operation for community-based multi-party microgrid in grid-connected and
islanded modes, IEEE Trans. Smart Grid 9 (2) (2018) 756–765.
[10] M. Cucuzzella, G.P. Incremona, A. Ferrara, Decentralized sliding mode control of islanded AC microgrids with
arbitrary topology, IEEE Trans. Ind. Electron. 64 (8) (2017) 6706–6713.
[11] L.G. Meegahapola, D. Robinson, A.P. Agalgaonkar, et al., Microgrids of commercial buildings: strategies to man-
age mode transfer from grid connected to islanded mode, IEEE Trans. Sustain. Energy 5 (4) (2014) 1337–1347.
[12] S. Sajadian, R. Ahmadi, Model predictive control of dual-mode operations z-source inverter: islanded and
grid-connected, IEEE Trans. Power Electron. 33 (5) (2018) 4488–4497.
[13] D. Wu, F. Tang, T. Dragicevic, et al., Autonomous active power control for islanded AC microgrids with
photovoltaic generation and energy storage system, IEEE Trans. Energy Convers. 29 (4) (2014) 882–892.
[14] J.Y. Kim, J.H. Jeon, S.K. Kim, et al., Cooperative control strategy of energy storage system and microsources
for stabilizing the microgrid during islanded operation, IEEE Trans. Power Electron. 25 (12) (2010) 3037–3048.
[15] D. Wu, F. Tang, T. Dragicevic, et al., A control architecture to coordinate renewable energy sources and energy
storage systems in islanded microgrids, IEEE Trans. Smart Grid 6 (3) (2015) 1156–1166.
[16] M.B. Delghavi, S. Shoja-Majidabad, A. Yazdani, Fractional-order sliding-mode control of islanded distributed
energy resource systems, IEEE Trans.Sustain. Energy 7 (4) (2016) 1482–1491.
[17] R.R. Jha, S.C. Srivastava, M. Kumar, Development of control schemes for a cluster of PV-integrated houses in
islanded mode, IET Renew. Power Gen. 11 (7) (2017) 903–911.
[18] V. Mortezapour, H. Lesani, Adaptive primary droop control for islanded operation of hybrid ACCDC MGs, IET
generation, Transm. Distrib. 12 (10) (2018) 2388–2396.
[19] G. Lou, W. Gu, L. Wang, et al., Decentralised secondary voltage and frequency control scheme for islanded
microgrid based on adaptive state estimator, IET Gen. Transm. Distrib. 11 (15) (2017) 3683–3693.
[20] X.J. Su, X.X. Liu, P. Shi, et al., Sliding mode control of hybrid switched systems via an event-triggered
mechanism, Automatica 90 (4) (2018) 294–303.
[21] Y. Zhao, Y. Shen, A. Bernard, et al., Evaluation and comparison of current biopsy needle localization and
tracking methods using 3D ultrasound, Ultrasonics 73 (1) (2017) 206–220.
[22] D.S. Chen, Z.Q. Wang, J. Li, Finite-time state tracking control with unmeasured state and various boundaries,
Int. J. Innov. Comput. Inf. Control IJICIC 14 (5) (2018) 1617–1631.
[23] L. Wang, L. Cao, L. Zhao, Non-linear tip speed ratio cascade control for variable speed high power wind
turbines: a backstepping approach, IET Renew. Power Gen. 12 (8) (2018) 968–972.
[24] J.A. Farrell, M. Polycarpou, M. Sharma, et al., Command filtered backstepping, IEEE Trans. Autom. Control
54 (6) (2009) 1391–1395.
[25] G. Cui, S. Xu, F.L. Lewis, et al., Distributed consensus tracking for non-linear multi-agent systems with input
saturation: a command filtered backstepping approach, IET Control Theory Appl. 10 (5) (2016) 509–516.
[26] J. Liu, H. An, Y. Gao, et al., Adaptive control of hypersonic flight vehicles with limited angle-of-attack,
IEEE/ASME Trans. Mechatron. 23 (2) (2018) 883–894.
[27] Y. Pan, H. Wang, X. Li, et al., Adaptive command-filtered backstepping control of robot arms with compliant
actuators, IEEE Trans. Control Syst. Technol. 26 (3) (2018) 1149–1156.
[28] X.J. Su, F.Q. Xia, J.X. Liu, et al., Event-triggered fuzzy control of nonlinear systems with its application to
inverted pendulum systems, Automatica 94 (8) (2018) 236–248.
[29] J. Liu, C. Wu, Z. Wang, et al., Reliable filter design for sensor networks in the type-2 fuzzy framework, IEEE
Trans. Ind. Inform. 13 (4) (2017) 1742–1752.
[30] C. Chiang, Y. Chen, Incremental fuzzy sliding mode control of pneumatic muscle actuators, Int. J. Innov.
Comput. Inf. Control IJICIC 14 (5) (2018) 1917–1928.
[31] G.C. Hsieh, H.I. Hsieh, C.Y. Tsai, et al., Photovoltaic power-increment-aided incremental-conductance MPPT
with two-phased tracking, IEEE Trans. Power Electron. 28 (6) (2013) 2895–2911.
[32] A. Ajami, A.M. Shotorbani, M.P. Aagababa, Application of the direct Lyapunov method for robust finite-time
power flow control with a unified power flow controller, IET Gen. Transm. Distrib. 6 (9) (2012) 822–830.
[33] Q. Shen, P. Shi, Y. Shi, Distributed adaptive fuzzy control for nonlinear multiagent systems via sliding mode
observers, IEEE Trans. Cybern. 46 (12) (2016) 3086–3097.
[34] W.X. Yan, J. Huang, D.Z. Xu, Adaptive command-filtered backstepping control for linear induction motor via
projection algorithm, Math. Probl. Eng. (2016) 1–13.
[35] Z. Cai, M.S. De Queiroz, D.M. Dawson, A sufficiently smooth projection operator, IEEE Trans. Autom. Control
51 (1) (2006) 135–139.

Adaptive Fuzzy Sliding Mode Command-Filtered - Backstepping Control For Islanded PV Microgrid With - Energy - Storage - System

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Adaptive Fuzzy Sliding Mode Command-Filtered - Backstepping Control For Islanded PV Microgrid With - Energy - Storage - System

Uploaded by

Copyright:

Available Formats

Available online at www.sciencedirect.

Journal of the Franklin Institute 356 (2019) 1880–1898

Fig. 1. Topology of the islanded microgrid.

Fig. 2. Structure of islanded microgrid with PV array and ESS.

2. Structure of islanded microgrid with PV array and ESS

Fig. 3. Equivalent circuit of PV cell.

Fig. 4. P-U curve of PV cell.

2.2.1. Structure of PV array

2.2.2. Incremental conductance maximum power point tracking

Find the derivative of Eq. (2), one obtains:

3. Dynamic mode of PV system and fuzzy algorithm

3.1. Dynamic mode of PV system

3.2. Fuzzy algorithm

I F x1 is Al11 and . . . and xn is Alnn , T hen ud is S l1 ...ln , (13)

where l1 = 1, 2, . . . , m, i = 1, 2, . . . , n. The fuzzy controller can be calculated as

Fig. 5. Structure of the command-filter.

4. Design process of AFSCB controller

4.1. Controller design

V˙2 = ē1 ē˙1 = ē1 (e˙1 − ε˙ )

S˙1 = k1 ē˙1 + e˙2

η˜1 ˙ 3(usd isd + usq isq ) η˜2 ˙  

4.2. Stability proof

Substituting Eqs. (30)–(33) into Eq. (29), the V˙3 is computed as

3usd ηˆ1 |e2 |

Fig. 6. Signal flow diagram of the AFSCB controller.

Fig. 7. Simulation model of islanded PV microgrid with ESS.

Parameters Value Description

5. Simulation results and studies

Parameters Value Description

Parameters Value Description

Fig. 9. The curves of irradiance and temperature.

Fig. 10. Parameters adaptive estimation curves.

Fig. 11. Fuzzy adaptive estimation curves.

You might also like

η˜1 ˙ 3(usd isd + usq isq ) η˜2 ˙