Due to the nonlinearity and complexity of the threephase photovoltaic inverter, we propose an intelligent control based on fuzzy logic and the classical proportionalintegralderivative. The feedback linearization method is applied to cancel the nonlinearities, and transform the dynamic system into a simple and linear subsystem. The system is transformed from abc frame to dq0 synchronous frame, to simplify the state feedback linearization law, and make the closeloop dynamics in the equivalent linear model. The controls improve the dynamic response, efficiency and stability of the threephase photovoltaic grid system, under variable temperature, solar intensity, and load. The intelligent control of the nonlinear characteristic of the photovoltaic automatically varies the coefficients K
_{p}
, K
_{i}
, and K
_{d}
under variable temperature and irradiation, and eliminates the oscillation. The simulation results show the advantages of the proposed intelligent control in terms of the correctness, stability, and maintenance of its response, which from many aspects is better than that of the PID controller.
1. INTRODUCTION
The use of renewable energy sources is rapidly developing, and the application of solar energy focusing photovoltaic systems is becoming increasingly popular. The major challenge in photovoltaics is posed by the instability, nonlinearity, and complexity of the currentvoltage characteristic equation. In this paper, we propose an intelligent control of a threephase photovoltaic gridconnected inverter system, which is essential to consider. The behavior of photovoltaic’s in currentvoltage and voltagepower is described by complex and nonlinear analytical equations, and depends on various levels of solar intensity and various cell temperatures. The characteristic of photovoltaic’s is a complex and nonlinear function, and it is difficult to identify a dynamic model. On the other hand, the Lambert Wfunction can be used to find a mathematical equation that is capable of describing the behavior of photovoltaic model, including all related parameters. The lambert Wfunction has been widely used in many publications and fields to reduce the complexity, and describe the behavior of a photovoltaic system, to convert the nonlinear and complex equation into explicit form. Ref.
[1]
has given many available applications of the Lambert Wfunction, while Ref.
[2]
used it to develop an analytical compact model for the asymmetric lightly doped MOSFET. Furthermore, the feedback linearization method and state feedback law to transform and eliminate the nonlinearity model of photovoltaic system into a simple equivalent linear model, were used to find a direct relation between the output and the control input, by using inverse dynamics
[3
,
4]
. This nonlinear state model transforms the dq0 synchronous frame reference into an equivalent linear system, where the pole placement control loop technique is applied to separate the control, and place the closedloop system pole in the desired location. The system is composed of a photovoltaic array, capacitive DClink, and threephase inverter connected to the grid, which is assumed to be in phase with the inductor current.
The classical PID controller is widely used in 95% of industrial technology. In fact, to assure the stability of the threephase photovoltaic gridconnected with L filters, intelligent adaptive fuzzy logic control and classical PID control using the feedback linearization method, is used to ensure a high quality system. This intelligent control is based on the combination of classical PID, and fuzzy logic controller, and the principle is to find out the fuzzy relationship between the three parameters of PID, error, and error change, by automatically varying the coefficients, K
_{p}
, K
_{i}
and K
_{d}
with variations of radiation, temperature, and load change. This controller combines the advantage of fuzzy control and PID control, and has a high performance under rapidly changing atmospheric conditions. Its effectiveness depends on the experience or knowledge of the right rule, input and output variables, and the membership functions.
Finally, the results obtained from the classical PID controller and intelligent PIDFuzzy logic with feedback linearization method show the advantages of the proposed intelligent control system in terms of its correctness, and the feasibility, stability, and maintenance of its response, which is better than the classical PID controller form many aspects, and which has the advantages of being fast, robust and showing good performance under varying atmospheric conditions.
2. LAMBERT WFUNCTION OF PHOTOVOLTAIC
The Lambert Wfunction is used to find a mathematical equation capable of describing the behavior of a photovoltaic model, including all related parameters. The Lambert wfunction has been widely used in many publications and fields to reduce the complexity and describe the behavior of a photovoltaic system, by converting the nonlinear and complex equation into explicit form
[5]
.
Figure 1
represents a photovoltaic model with a single diode; under illumination the relation between the current and voltage of the photovoltaic single diode is given by
[6]
:
Photovoltaic model.
where, V
_{th}
=K.T/q is the thermal voltage of the photovoltaic module; I
_{ph}
is the irradiance current (photocurrent); I
_{0}
is the cell reverse saturation current (diode saturation current); q is the electron charge; n is the cell ideality factor; K is the Boltzmann constant; T is the cell temperature; and R
_{s}
and R
_{p}
represent the cell series and shunt resistance, respectively.
The Lambert Wfunction is the inverse function of
Equation (1) is transcendental in nature, and the explicit form
Similarly, the output power of the photovoltaic array can be given in explicit terms, as follows:
Figure 2
presents the characteristic currentvoltage and powervoltage of the photovoltaic cell for different values of temperature and solar incident irradiance.
Photovoltaic cell electrical curves with constant temperature and different values of irradiance.
In the electrical characteristics, two points exist for a specific curve that can generate more power than other points. The maximum output power varies with temperature and irradiation change, so it necessary to extract the maximum available power for any changes.
3. STATE FEEDBACK MODEL OF THE THREEPHASE PHOTOVOLTAIC
Figure 3
shows the general circuit topology configuration of a threephase photovoltaic grid connected inverter connected to the grid connected to the inductance. The system is composed of a photovoltaic array, threephase inverter and capacitive DClink. The photovoltaic array converts solar irradiation into DC current, and the DClink capacitor reduces the harmonic of the DC voltage on the input side of the inverter.
Diagram of threephase gridconnected photovoltaic inverter.
The threephase inverter model of
Fig. 3
given in states space coordinates by:
Park’s transformation is applied to equation (5)
We apply the dq transformation to equation (6), and obtain the statespace dq in equation (7)
where, e
_{d}
, e
_{q}
, e
_{0}
, i
_{d}
, i
_{0}
, v
_{d}
, v
_{q}
and v
_{0}
are the components of the grid voltage, grid current, and inverter output voltage, respectively, and ω is the angular frequency. The power losses are neglected in inverter switches, and the power of the DCinput and ACoutput is given by:
Where, v
_{dc}
and i
_{dc}
are the input voltage and current of the inverter, respectively.
Kirchhoff’s law gives
From equations (7) and (9), the equation of the state model can be grouped as:
 3.1 State feedback linearization controller
The main objective of feedback linearization control is to transform and eliminate the nonlinearity model of the photovoltaic inverter system into a simple equivalent linear model, and find a direct relation between the output and the control input, by using inverse dynamics
[7
,
8]
.
Let us consider a nonlinear control system of the phase state variable vector and phase input vector. The inputoutput state equation of the threephase system is
where,
The outputs are differentiable until the inputs appear for searching for the exact input output. Let r
_{i}
be a small integer such that at least one of the inputs appears in the r
_{i}
th derivation of y
_{i}
. The method is detailed in
[9
,
10]
.
When
For the first output:
We have u
_{2}
input from the first derivation and the relative degree is r
_{1}
=1.
For the second output:
The inputs u
_{1}
and u
_{2}
appear after the second derivation and the relative degree is r
_{1}
=1.
The sum of r
_{1}
+r
_{2}
=3 and this is the order of the system. Then we can give an exact linearization of the system.
From equations (14) and (16), the feedback linearization of the system is
where:
Where, β(x) is a non singular matrix, whose determinant is:
The corposant e
_{d}
of the gridconnected voltage is always different from zero. Then the determinant is not null, and β(x) is nonsingular. The linearization of the control part is
To eliminate the nonlinearity of the system, we substitute equation (21) into (18). So the simple linear relation between outputs yi and the new inputs v
_{i}
is:
The nonlinearity of the system is cancelled, and
Fig. 4
presents the linear equivalent control block diagram of the proposed method.
Feedback linearization control block diagram.
The controller is designed to obtain the stability of an integral controller, which is added to eliminate the steadystate error due to parameter variations.
where, e
_{1}
=y
_{1}
−y
_{1ref}
, e
_{2}
=y
_{2}
−y
_{2ref}
and y
_{ref}
is the tracking error, y
_{1ref}
is the tracking reference of the grid current reference, y
_{2ref}
the DClink voltage reference, and k
_{ij}
the gain. Then, the output error is:
The controller gains k
_{ij}
are determined by Rollscriterion and to assure the characteristic polynomial of the equation (24), these are Hurwitz polynomials, so the error trackings of e
_{1}
and e
_{2}
converge to zero.
4. DSIGN OF CONTROLLERS
 4.1 Conventional PID controller
The conventional fundamental classical PID controller is often described by the following equation in ideal form for the output voltage of a grid photovoltaic system
[11

13]
:
where, e
_{1}
(t)=y
_{1}
−y
_{1ref}
, e
_{2}
(t)=y
_{2}
−y
_{2ref}
and e
_{3}
=(t)y
_{ref}
is the tracking error, in which V
_{out}
(t) denotes the controller output at time t of the photovoltaic array. K
_{p}
, K
_{i}
and K
_{d}
are known as the proportional constant, integral constant, and derivative constant respectively.
 4.2 Structure of the PIDfuzzy selftuning controller
Figure 5
shows the structure of the control system with the proposed PIDFuzzy logic controller. The principle of fuzzy selftuning PID is first to find the fuzzy relationship between the three parameters of PID and error e
_{i}
(t)(i=1, 2 and 3) and the variation of the error
(i=1, 2 and 3), and adjust the PID parameters with fuzzy rules. Fuzzy inference engines modify three parameters, to be content with the online demands of the control system. The inverse reference model of the response of the controlled object is used in the approach to tuning the PIDfuzzy controller.
The structure of the PIDFuzzy logic selftuning control.
The input parameters are the deviation e
_{i}
(t) and the variation of deviation Δe
_{i}
(t), which are shown in equation (26). The output is the variation of MPPT.
The output parameters are K
_{p}
, K
_{i}
and K
_{d}
, and continuously detect e
_{i}
(t) and Δe
_{i}
(t) according to the fuzzy logic rules and regulations in order to satisfy the desire of the two inputs to the parameters of the controller at any time
[14]
.
 4.3 Fuzzy gain scheduling
According to the methodology proposed by Zhao
[15]
, the three current PID parameters are determined as follows. We assume the range of K
_{p}
is [K
_{p,min}
K
_{p,max}
], the range of K
_{i}
is [K
_{i,min}
K
_{i,max}
], and the range of K
_{d}
is [K
_{d,min}
K
_{d,max}
]. The range of K
_{p}
and K
_{d}
can be normalized into the range, through the linear transformation in equations (27) and (28).
where, K'
_{p}
, K'
_{d,}
and α are constants that are determined by means of the fuzzy mechanism. The K
_{p,min}
, K
_{p,max}
, K
_{d,min}
, K
_{d,max}
are constants adopted to normalize the values of K
_{p}
and K
_{d}
which are given by :
The parameters K
_{p}
, K
_{i}
, and α are determined by a set of fuzzy rules.
If e
_{i}
(t) is A
_{i}
and Δe
_{i}
(t) is B
_{i}
, then K'
_{p}
is C
_{i}
, K'
_{d}
is D
_{i}
, and α=α
_{i}
(i=1, 2…m). Here, A
_{i}
, B
_{i}
, C
_{i}
, and D
_{i}
are fuzzy sets on the corresponding supporting sets, while α
_{i}
is a constant.
The fuzzy rules variables values of the input e
_{i}
(t) and Δe
_{i}
(t) are configured for seven membership functions with NB= Big Negative, NM=Medium Negative, NS=Small Negative, ZO=Zero, PS= Small Positive, PM=Medium Positive, and PB=Big Positive, in the format of fuzzy control rules:

{NB, NM, NS, ZO, PS, PM, PB}
with the truth value of {0,6}
Figure 6
shows the memberships of the input variables e
_{i}
(t) and Δe
_{i}
(t).
Memberships of the input variables e_{i}(t) and Δe_{i}(t).
The output K'
_{p}
and K'
_{d}
use two exponential membership functions; to calculate K'
_{p}
and K'
_{d}
, we need to correct the factors K'
_{p}
and K'
_{d}
in the range. The values of the linguistic variable of K'
_{p}
and K'
_{d}
are shown in
Figs. 9
and
10
.
Membership of K'_{p}.
Membership of K'_{d}.
Membership of α.
Threephase output voltage v_{a,b,c} of gridconnected using PID.
The fuzzy sets C
_{i}
and D
_{i}
may be either Big or Small and are characterized by the grade of the membership functions μ, and the variables μ=(K'
_{p}
or K'
_{d}
) have the following equations:
For Small:
For Big:
To calculate the integral coefficient K
_{i}
, correction of the factor is needed, and considered as a fuzzy number, it has a singleton membership function represented by points in
Fig. 9
, and has 7 linguistic variables, whose values are {α..
_{NB}
, α
_{NM}
, α
_{NS}
, α
_{ZO}
, α
_{PS}
, α
_{PM}
, α
_{PB}
}, their value being 2, 3, 4, 5.
Through the analysis of the characteristic curve of a photovoltaic array, we can track the fuzzy control rules and correction factor of the classical PID controller parameter, for which e
_{i}
(t) and Δe
_{i}
(t) are in different conditions. The fuzzy control rules are shown in
Tables 1
,
2
, and
3
.
The fuzzy control rules for K'p.
The fuzzy control rules for K'_{p}.
The fuzzy control rules for K'd.
The fuzzy control rules for K'_{d}.
The fuzzy control rules for α.
The fuzzy control rules for α.
5. SIMULATIONS AND RESULTS
Theoretical analysis of the proposed hybrid control of the output current and voltage of the threephase photovoltaic system is validated and done by simulation using the Simulink platform. The simulation started when the solar irradiance was at G=1,000 W/m² and T=25℃. Firstly, the threephase gridconnected photovoltaic inverter using the proposed hybrid controller is studied, in comparison with the separated classical PID controller. The inverter is connected to the DC source through a DClink capacitor, with constant voltage reference value used to emulate the output of the maximum power point controller. The DClink capacitor voltage is regulated and controlled by a PIDFuzzy logic control loop, to reduce the oscillation in P
_{r}
at twice the line frequency due to grid imbalances. The output is the reference power, which must be injected into the grid by the inverter. The simulation control of the sinusoidal output current and voltage obtained with the DClink capacitor replaced by a constant voltage source V
_{link}
=400 V. The objective is to analyze the behavior of the current and voltage controller of
Figs.10
,
11
, and
12
, without influence of the DClink and of the rest the system.
Threephase output current i_{a,b,c} of gridconnected using PID.
Plot of the DClink voltage v_{DC}.
Figures 10
, and
11
show the current and voltage, respectively, that are synthesized by the PID controller, and injected into the grid.
Figure 14
shows the plot voltage at the load.
Threephase output voltage v_{a,b,c} of the photovoltaic gridconnected converter.
Plot of the DClink voltage v_{DC} using the proposed control
Due to the structure of the threephase photovoltaic system converter, the output current and voltage are disturbance and unbalanced operating condition as shown in
Figs.10
,
11
, and
12
.
The objective now is to present the results of the simulation with all system parts of the photovoltaic array and DCAC converter working together, relative to the PIDfuzzy controller.
From t = 0 s to t = 0.16 s the system drains energy from the AC grid. From t = 0.16 s, the converter injects active power into the grid. At 0.16 seconds the reference current is stepped, and after this time, the converter operates in a steady state condition. Overall, the controller provides an excellent dynamic response.
Figure 16
shows the behavior of the DClink voltage, which initially is not charged. The photovoltaic capacitor is charged from 0 V to 700 V from the photovoltaic array. During the initial charge, the photovoltaic array supplies its maximum current, approximately near 0.16 s; and when the capacitor is charged, the DCAC starts to supply current to the grid. The output sinusoidal current draining power to charge the capacitor suffers a phase inversion near t=0.16 s, and from this time, the DCAC converter begins to deliver active power to the grid, as in
Figs. 14
and
15
.
Threephase output currents i_{a,b,c} of the photovoltaic gridconnected converter.
Plot of the DClink voltage.
Output power of the photovoltaic array delivered to the AC grid.
6. CONCLUSIONS
This paper presented an intelligent hybrid based on PID control, and a fuzzy logic controller was proposed to control and stabilize the threephases of the output current and voltage of the photovoltaic gridconnected converter. The system is complex and nonlinear, and a Lambert Wfunction and feedback linearization method was proposed. In the analysis, a fully mathematical approach was used to gain insight into the behavior of the photovoltaic model related to the onediode equivalent circuit. The feedback linearization method and state feedback law transform and eliminate the nonlinearity model of the threephase photovoltaic into a simple linear model. Nowadays, it has become evident that complex and nonlinear problems need intelligent systems to be solved. Recently, more than 95% control loops still use the PID controllers. They are used with fuzzy logic control to overcome the problems in the unknown mathematical model of the system. A PIDFuzzy logic hybrid controller was applied to the control power system. Furthermore, fuzzy logic controller is implemented as a gain scheduler to automatically tune the K
_{p}
, K
_{i}
and K
_{d}
parameters for the PID controller. The system was regulated and controlled, and the results show the active power reduced in oscillation, and the current and voltage stabilized. As the results confirmed, initially the DClink voltage capacitor is not charged, and during the charge, the threephase photovoltaic system supplies a maximum current; and when the capacitor charges, the DCAC starts to supply the grid. Compared with conventional controllers, this ensures minimum peak values in the gridinjected currents during unbalanced voltage sags. The proposed intelligent hybrid controller gave fast response and stability without delay, eliminating the oscillation around the operating points. From the simulation results, we concluded that the performance of the intelligent controller is better than the ones obtained with the classical PID controller, since the response time in the transitional state was shortened, and the fluctuations in the steady state were considerably reduced.
Corless R. M.
,
Gonnet G. H.
,
Hare D. E. G.
,
Jeffrey D. J.
,
Knuth D. E.
1996
J. Adv. Comput. Math.
5
329 
Abebe H.
,
Cumberbatch E.
,
Morris H.
,
Tyree V.
,
Numata T.
,
Uno S.
2009
J. Semiconduct. Technol. Sci.
9
225 
Arifa J.
,
Rayb S.
,
Chaudhuria B.
2013
International Journal of Electrical Power & Energy Systems
[DOI: ]
45
87 
Johansen T. A.
,
Hunt K. J.
2000
IEEE Publication in Decision and Control Conference
5
4467 
Hwang C.
,
Cheng Y.
2005
American Control Conference
6
4283 
Louzazni M.
,
Aroudam E.
,
Yatimi H.
2013
International Journal of Electrical and Computer Engineering
3
568 
Charlet B.
,
Lévine J.
1989
Systems & Control Letters
[DOI: ]
13
143 
Van Der Schaft A. J.
1984
Systems & Control Letters
[DOI: ]
5
27 
Tall I. A.
2010
Systems & Control Letters
[DOI: ]
59
429 
Monaco S.
,
NormandCyrot D.
2006
Systems & Control Letters
http://dx.doi.org/10.1016/j.sysconle.2005.04.016
55
71 
Chaouachi A.
,
Kamel R. M
,
Nagasaka K.
2010
Sol. Energy
[DOI: ]
84
2219 
Louzazni M.
,
Aroudam E.
2014
IEEE Publication, Intelligent Systems: Theories and Applications (SITA14), 2014 9th International Conference on
Rabat
[DOI: ]
1 
6
Louzazni M.
,
Aroudam E.
2014
IEEE publication, Intelligent Energy and Power Systems (IEPS), 2014 IEEE International Conference on
Kiev
[DOI: ]
279 
284
Wang Z.
,
Gong Z.
,
Wang Y.
,
Wei G.
2009
IEEE 9th International Conference on Electronic Measurement & Instruments
913 
Zhao Z.
,
Tomizuka M.
,
Isaka S.
1993
IEEE Transactions on Systems Man and Cybernetics
[DOI: ]
23
1392 