523(&6SHFLDO6HVVLRQ2Q5HQHZDEOH(QHUJ\

Optimal Control for a Renewable-energy-based Micro-grid

Fernando Ornelas-Tellez

Abstract— This paper presents an optimal nonlinear control resources is proposed, which can easily include new sources,
scheme for a renewable-energy-based micro-grid. The optimal loads, storage elements and the connection with the utility
controller synthesis is based on the state-dependent Riccati grid. Simulation results demonstrate the adequate operation
equation approach, where it is considered the minimization
of a performance index (cost functional), which results in an of the micro-grid.
efficient control strategy. The micro-grid is used to integrate the The paper is organized as follows. Section II presents the
energy from renewable resources, with the facility to storage optimal control for SDCF nonlinear system. The proposed
energy, to provide of energy to loads and to provide/consume micro-grid scheme is presented in Section III, where the
energy to/from the utility grid. Simulation results are presented different components, corresponding to energy generation
to assess the performance of the proposed controller for a case
study micro-grid, which is composed of a wind energy system, and consumption, are described. In Section IV, the proposed
a load, a storage element and the connection with the utility optimal control scheme is applied for a case study micro-
grid. grid. Simulation results are presented in V. Finally, Section
VI concludes the paper.
I. I NTRODUCTION
Due the interest in exploiting renewable energy sources, II. O PTIMAL C ONTROL FOR SDCF N ONLINEAR
it is common to integrate such energy resources through S YSTEMS
distributed power generation systems. These systems usually This section describes the optimal tracking control solution
constitute a micro-grid, which is composed of multiple gener- for a class of nonlinear systems which can be presented as a
ation sources (solar, wind, etc.), energy storage devices, local state dependent coefficient factorized (SDCF). The proposed
loads and the connection to the utility grid for exchanging solution considers that all states are measured and available
energy. In general, the functionality and the interconnection for feedback.
between the components of the micro-grid is realized by
means of power converters, which requires of adequate A. State-dependent coefficient factorized nonlinear systems
control techniques to provide of reliability and efficiency to
Let us consider the input-affine nonlinear system
the operation of the micro-grid.
Correctly designed, power electronics and controllers en- ẋ = f (x) + B(x) u, x(t0 ) = x0 (1)
sure that the micro-grid can meet its own as well as the
utilities needs in an economic way. The basic inputs to the y = h(x) (2)
power converter controllers are the set points or references
for the exchanged power among components of the micro- where x ∈ Rn is the state vector, u ∈ Rm is the control
grid, or best levels of local bus voltages and currents deter- input and y ∈ Rp is the system output; the functions f (x),
mined in accordance to energy to be transferred. For instance, B(x) and h(x) are smooth maps of appropriate dimensions.
to achieve energy transfer among multi-sources the authors Consider that functions f (x) in (1) and h(x) in (2) can be
in [1] have proposed an energy router that dynamically decomposed in the state-dependent coefficient factorization
controls energy flow between sources, it is based on an as f (x) = A(x) x and h(x) = C(x) x, respectively [3–6];
idea described by Duindam and Stramigioli in [2]. In [1] then system (1)-(2) results in
the feedback linearization technique and the Proportional-
ẋ = A(x) x + B(x) u (3)
Integral (PI) scheme are used for controlling the energy
flow between different sources. A main drawback of this y = C(x) x. (4)
controller is that it does not work well when the system is
nonlinear and its operating point is different from the point As established in [7, 8], the assumptions f (0) = 0, h(0) = 0,
for which the controller was tuned. f (·) ∈ C 1 and h (·) ∈ C 1 guarantee that the factorization as
This paper proposes an optimal control strategy for the described in (3)–(4) can be carried out. This salient feature
different power converters of a micro-grid. The resulting will be used in this paper to obtain an analytical solution for
controller ensures the stability and an efficient operation of the optimal control via the Differential Riccati equation. Note
the converters. A general scheme for integrating renewable that factorizations A(x) x and C(x) x are not unique [9].
In order to obtain well-defined control schemes, appropriate
1 F. Ornelas-Tellez is with Faculty of Electrical Engineering, Uni-
factorization for these representations should be determined
versidad Michoacana de San Nicolas de Hidalgo, Morelia, Mexico such that controllability and observability properties are
[email protected] fulfilled for system (3) and output (4).

‹,(((

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 18,2021 at 20:59:59 UTC from IEEE Xplore. Restrictions apply.
 523(&6SHFLDO6HVVLRQ2Q5HQHZDEOH(QHUJ\

B. Optimal Tracking Controller Synthesis Control System

For many applications, such as power converters,
aerospace, electrical machines, robotics, among others, it is
important that the controller could track a desired trajectory; Subsystem 1 Subsystem 2
then it is required that an output of the closed-loop system

Bus DC
tracks a desired trajectory as close as possible in an optimal
sense and with minimum control effort expenditure [10, 11].
Subsystem N-1 Subsystem N
In order to introduce the trajectory tracking, let us define
the tracking error as
Fig. 1. Micro-grid scheme.
e = r−y
= r − C(x) x (5)
III. T HE M ICRO - GRID BASED ON R ENEWABLE E NERGY
where r is the desired reference to be tracked by the system This section describes a general scheme for the energy
output y. integration from different energy resources in a micro-grid.
The quadratic cost functional J to be minimized, associ- It is considered that the micro-grid is composed by N
ated with system (3), is defined as subsystems, where each one can corresponds to: 1) an
 energy generation system (wind, solar, fuel cells, etc.), 2)
1 ∞ T 
J= e Q e + uT R u dt. (6) an energy consumption system (load, storage, etc), or 3) the
2 t0
connection with the utility grid to exchange energy with the
Therefore, the optimal tracking solution is related to micro-grid. Specifically, this paper considers a micro-grid
determining the control u(t), t ∈ [t0 , ∞), such that the based in renewable energy resources. Fig. 1 illustrates the
criterion (6) is minimized. The optimal tracking solution is connection of the subsystems in the micro-grid, where arrows
established as the following theorem. indicate the energy flow in accordance to the nature of the
Theorem 1: Assume that system (3)-(4) is state-dependent subsystem. The exchange of energy between the subsystems
controllable and state-dependent observable. Then the opti- is considered to be realized through power converters, which
mal control law require a control system to manage the energy. In this sense,
the proposed control strategy for power converters is the
u∗ (x) = −R−1 B T (x) (P (x) x − z(x)) (7) optimal control technique, as described in Section II-B.
For the control of the micro-gird, it is considered that each
achieves trajectory tracking for system (3) along a desired subsystem can be represented as
trajectory r, where P (x) is the solution to the symmetric ẋi = A(xi ) xi + B(xi ) ui
matrix differential equation
yi = C(xi ) xi (12)
T −1 T
Ṗ (x) = −C (x) Q C (x) + P (x) B(x) R B (x) where xi is the state vector of the N -th subsystem, with
T
× P (x) − AT (x) P (x) − P (x) A(x) (8) xi = [xi1 , xi2 , . . . , xij ] , i = 1, . . . , N and j = 1, . . . , ni ,
being ni the order of the corresponding subsystem. Matrices
and z(x) is the solution to the vector differential equation A(xi ), B(xi ) and C(xi ) are of appropriate dimensions. The
 T respective cost functional becomes
ż(x) = − A(x) − B(x) R−1 B T (x) P (x) z(x) 
1 ∞ T 
−C T (x) Q r (9) Ji = ei Qi ei + uTi Ri ui dt. (13)
2 t0
with boundary conditions P (∞) = 0 and z(∞) = 0, Note that system (12) has the form of (3)–(4), therefore
respectively. Control law (7) is optimal in the sense that it the optimal control technique, as developed in the previous
minimizes the cost functional (6), which has an optimal value section, can be used. Indeed, this constitutes a decentralized
function given as control scheme, which easily allows the integration of dif-
ferent energy resources and components, corresponding to
1 T generation, consumption, storage and connection with the
J∗ = x (t0 ) P (t0 ) x(t0 ) − z T (t0 ) x(t0 ) + ϕ(t0 ) (10)
2 utility grid. It is worth noting that the only limitation for
applying the proposed optimal controller is that the system
where ϕ is the solution to the scalar differentiable function
to be controlled admits the structure of (12).
1 1
ϕ̇ = − rT Q r + z T B(x) R−1 B T (x) z (11) IV. O PTIMAL C ONTROL FOR A S INGLE -P HASE
2 2 M ICRO -G RID : A C ASE S TUDY
with ϕ(∞) = 0. This section presents the application of the optimal track-
Proof: For proof details see [12]. ing control scheme for the different power converters used

‹,(((

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 18,2021 at 20:59:59 UTC from IEEE Xplore. Restrictions apply.
 523(&6SHFLDO6HVVLRQ2Q5HQHZDEOH(QHUJ\

Ls x 1 Rs
Control System
μ1 μ̄1
+
Wind
PW PL +
Load Vi DC Bus
Energy
−

Bus DC
−

Utility
PG PB μ̄1 μ1
Storage
Grid
Fig. 4. AC-DC power converter circuit.

Fig. 2. The case study micro-grid.

it is assumed that an AC voltage is obtained by a wind-
driven electrical generator. The switches formed by a diode
iBus in parallel to a transistor are controlled in its gates by a
switching signal μ, which denotes the switch position taking
+
+ values in the finite set {−1, 1}. The notation μ̄ denotes the
logic complement of μ. It is well known that when the
VCBus CBus RBus ⇒ DC Bus switching frequency is sufficiently high, the model describing
− iCBus the behavior of the circuit can be represented by its averaged
iRBus − model as [13, 14]
ẋ11 = −xb u/Ls − Rs x11 /Ls + Vi /Ls (16)
where x11 is the inductor current, xb is the bus voltage, Rs is
Fig. 3. DC bus circuit and an equivalent representation. the resistance of the inductor and the associated impedance
of the source and switches and Ls is the inductance, Vi =
Vp sin (ωt) is the voltage of an AC source, with ω = 2πf ,
in the micro-grid to efficiently distribute the generated re- where f is the frequency in Hertz and Vp is the amplitude of
newable energy. Without of generality, this paper proposes the input voltage. For this average model, u ∈ [0, 1] becomes
the control for a single-phase micro-grid composed of the the duty cycle, which is used to generate the switching
following subsystems: a DC bus, a wind system, a load, signal μ for the circuit in Fig. 4 by means of a pulse-
a storage element and the connection with the utility grid. width modulation (PWM) strategy. Note that for system (16),
It is worth pointing out that additional resources could be variable x11 represents the average value of the current.
integrated to the micro-grid and its respective control system The generated electrical power (PW ), is considered to be
can be synthesized by using the proposed control scheme. known from a maximum power point tracking algorithm,
The components of the micro-grid and its optimal con- then the current x11 can be calculated from the knowledge
troller design are designed as follows. of the voltage (Vi ) as
A. DC Bus PWRM S
x11RM S =
A capacitor, as the depicted in Fig. 3, is used as the ViRM S
DC bus.. The continuous-time voltage dynamics of the which will be used as the current reference when synthesized
charge/discharge of this device is given as the optimal controller.
VCBus 1 1) Controller Synthesis: For space limitation, only the
V̇CBus = − + iBus (14) controller synthesis for this subsystem will be described,
RCBus CBus CBus
meanwhile for the other ones the optimal controllers are
where VC is the capacitor voltage, iBus is the current in designed in a similar way.
the capacitor, C is the capacitance and RC is the respective For i = 1 in (12), subsystem (16) can be presented as
internal resistance of the capacitor. For easy of notation,
system (14) is represented by ẋ1 = A(x1 ) x1 + B(x1 ) u1 + E1 (17)

xb 1 y1 = C(x1 ) x1 (18)
ẋb = − + ib (15)
RCBus CBus CBus where x1 = x11 , A(x1 ) = [−Rs /Ls ], B(x1 ) = −xb /Ls ,
where xb = VCBus and ib = iBus . Current ib is calculated C(x1 ) = 1 and E1 = Vi /Ls .
as the sum of the input and output currents in the bus. Matrices for the performance index (13) are selected as

B. Wind Energy System Q1 = 1000 and R1 = 50.

For wind energy generation we use the circuit of the Note that Q is a parameter weighting the performance of
full bridge power converter as displayed in Fig 4, where the tracking error e, meanwhile R is a parameter weighting

‹,(((

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 18,2021 at 20:59:59 UTC from IEEE Xplore. Restrictions apply.
 523(&6SHFLDO6HVVLRQ2Q5HQHZDEOH(QHUJ\

LL x21 LSt −x31

μ2 μ̄2 μ3
+ + + +

DC Bus CL x22 RL DC Bus CSt x32 RSt

− − − −
μ̄2 μ2 μ̄3

Fig. 5. DC-AC power converter circuit with a resistive load. Fig. 6. DC-DC power converter.

Li x41 Lo x43
the control effort expenditure; hence these parameters are μ4 μ̄4
used to establish a trade-off between the state variables per- + +

formance and control effort. In addition, these parameters are +

used to obtain the solution of the corresponding differential DC Bus Co x42 VS
equations (8) and (9) (with their respective subscript i = 1). −
Finally the reference vector used in the control scheme (7)- − −
(9) becomes μ̄4 μ4
√
r1 = r11 = x11RM S 2 sin (ωt) Fig. 7. Inverter and LCL filter for the utility grid connection.

where r11 is the desired value for the inductor current. The
optimal controller for (17) results in The dynamical average model is described by
u∗1 (x1 ) = ū1 − R1−1 B T (x1 ) (P1 x1 − z1 ) (19) ẋ31 = x32 /LSt − xb u3 /LSt
ẋ32 = −x31 /CSt − x32 /(RSt CSt ) (21)
where ū1 = Vi /xb , which is included to cancel the constant
term E1 in (17). y3 = xb x31
where x31 is the inductor current, x32 is the capacitor
C. Load in the Micro-grid voltage, LSt is the inductance, CSt is the capacitance, RSt
To provide from energy to the subsystem corresponding models the internal resistance of the capacitor and u3 is the
to the load, we are considering a DC-AC converter [15], as control input (duty cycle for the PWM).
displayed in Fig. 5. The averaged model in its state space The power to be transferred in the DC-DC converter (PS )
representation is given as [13, 14] is calculated directly from the bus voltage and the inductor
current, which establishes the reference for PS as PSref .
ẋ21 = −x22 /LL + xb u2 /LL Parameters for this system and for the controller are given
ẋ22 = x21 /CL − x21 /(RL CL ) (20) in Table I.
y2 = x22 E. Utility grid connection
where x21 is the inductor current, x22 is the output capacitor This section focuses on the component (converter plus
voltage, LL is the inductance, CL is the capacitance, RL is filter) to transfer/extract energy toward/from the utility grid
the resistive load and u3 is the control input (duty cycle for through a LCL filter as shown in Fig. 7. This is achieved by
the PWM). controlling of the power PG calculated as PG = VS x43 .
For this subsystem, its respective optimal control scheme The average model for this system is given as
is synthesized similarly as the developed in the previous ẋ41 = −x42 /Li + xb u4 /Li
section. For simulation purposes, the load RL is calculated
ẋ42 = (x41 − x43 )/Co
in accordance to the knowledge in advance of the dissipated
power by the load (PL ). The subsystem and controller ẋ43 = (x42 − x44 )/Lo (22)
parameters are given in Table I. ẋ44 = ω x45
ẋ45 = −ω x44
D. Energy Storage
y4 = x43 (23)
Fig. 6 shows the electrical circuit used for achieving
energy exchange between the bus and the storage element, where x41 , x43 are the currents across the inductors Li and
which is modeled by the capacitor CSt in parallel with RSt . Lo , respectively; x42 is the voltage in the capacitor Co ; the
This storage component could correspond to a rechargeable parameters Li , Lo and Co in the circuit are the inductances
battery, a capacitors bank, etc. and capacitance, respectively; VS is the utility grid voltage,

‹,(((

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 18,2021 at 20:59:59 UTC from IEEE Xplore. Restrictions apply.
 523(&6SHFLDO6HVVLRQ2Q5HQHZDEOH(QHUJ\

TABLE I
S UBSYSTEMS AND CONTROLLERS PARAMETERS .

Parameter Value Parameter Value

CBus 0.1 F RBus 1 MΩ
Ls 1 mH Vp 150 V
Rs 2.2 Ω f 60 Hz
R1 50 Q1 1000
LL  1 mH  CL 50 μF
Vo2 /2 √
RL Ω Vo 120 2 V
PLRM S
R2 1 Q2 2000
r1 Vo sin(ωt) V ω 2 π f rad/s
LSt 195 μH CSt 50 F
RSt 1 MΩ R3 10
Q3 1 r3 PSref (a) Power control for electrical energy from the wind system.

Li 2 mH Lo 833 μH
Co 1μF R4 0.1
Q4 1000 r4 PGref VS /2302

represented by an exosystem modeled by the linear oscillator

in the last two√equations of (22), with initial condition for
x45 (0) = 230 2 V . The remaining initial conditions are
settled to zero values.
The power (PGref ) to be transferred/extracted for this
inverter is calculated such that the voltage in the bus xb
is maintained in a constant value. In this case, the optimal
control is synthesized to maintain xbref = 425 V . This
voltage value is selected for an appropriate operation of (b) Power control for the load.
the inverters. Thus, when voltage xb is the increasing from
Fig. 8. Power control for PW and PL .
its reference value xbref , energy must be transferred to
the utility grid, otherwise, when decreasing from xbref ,
energy must be extracted from the utility grid. This control
procedure allows to have uncertainties, disturbances and Numerical simulations are used to evaluate the effectiveness
losses in the different subsystems, which constitutes a robust ot the the proposed micro-grid, the power converters and
control strategy, meanwhile the optimal control scheme will their respective controllers. The simulation results illustrate
guarantee the adequate and efficient operation of the whole the adequate operation of the whole micro-grid to manage
micro-grid, as will be shown in the simulation results. renewable energy.
By considering all subsystems, the bus current ib is
calculated as ib = x11 − x21 + x31 − x41 . R EFERENCES
V. S IMULATION R ESULTS [1] A. Sanchez-Squella, R. Ortega, R. Grino, and S. Malo,
The system and controller parameters used for the sim- “Dynamic energy router,” IEEE Control Systems,
ulation are given in Table I. Simulations are performed by vol. 30, no. 6, pp. 72–80, 2010.
using Mathematica R
. [2] V. Duindam, A. Macchelli, S. Stramigioli, and
Figures 8 and 9 displays the electrical power control for H. Bruyninckx, Modeling and Control of Complex
the subsystems in the case study micro-grid. The references Physical Systems. Berlin, Germany: Springer-Verlag,
signals for each subsystem are selected with the purpose to 2009.
illustrate the effectiveness of the proposed optimal tracking [3] J. R. Cloutier, C. N. D’Sousa, and C. P. Mracek,
control technique. “Nonlinear regulation and nonlinear H∞ control via
Finally, the voltage level control for the DC bus is shown the state-dependet Riccati equation technique: Part 1,
in Fig. 10 theory,” in Proc. of the First Int. Conf. on Nonlinear
Problems in Aviation and Aerospace, Daytoba Beach,
VI. C ONCLUSIONS FL, USA, May 1996.
An optimal nonlinear control has been proposed to achieve [4] J. D. Pearson, “Approximation methods in optimal
trajectory tracking of the electrical power in a micro-grid control I. Sub-optimal control,” Journal of Electronics
through power converters, which are used to exchange energy and Control, vol. 13, no. 5, pp. 453–465, 1962.
in a micro-grid. The optimal controller can be synthesized [5] K. D. Hammett, C. D. Hall, and D. B. Ridgely, “Con-
for nonlinear systems which admit the SDCF representation. trollability issues in nonlinear state dependent Riccati

‹,(((

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 18,2021 at 20:59:59 UTC from IEEE Xplore. Restrictions apply.
 523(&6SHFLDO6HVVLRQ2Q5HQHZDEOH(QHUJ\

ear feedback controllers and compensators: a state-

dependent Riccati equation approach,” Computational
Optimization and Applications, vol. 37, no. 2, pp. 177–
218, 2007.
[10] B. D. O. Anderson and J. B. Moore, Optimal Control:
Linear Quadratic Methods. Englewood Cliffs, NJ,
USA: Prentice-Hall, 1990.
[11] M. Athans and P. L. Falb, Optimal Control: An Intro-
duction to the Theory And Its Applications. New York,
NY, USA: McGraw Hill, 1966.
[12] F. Ornelas-Tellez, J. J. Rico, and R. Ruiz-Cruz, “Op-
timal tracking for state-dpendent coeficient factorized
(a) Power control for energy storage. nonlinear systems,” Asian Journal of Control, vol. 16,
no. 3, pp. 1–14, 2013,DOI: 10.1002/asjc.761.
[13] R. Ortega, A. Lorı́a, P. J. Nicklasson, and H. Sira-
Ramı́rez, Passivity-based Control of Euler-Lagrange
Systems: Mechanical, Electrical and Electromechanical
Applications. Berlin, Germany: Springer-Verlag, 1998.
[14] H. J. Sira-Ramirez and R. Silva-Ortigoza, “Control
design techniques in power electronics devices,” Berlin,
Germany, 1996.
[15] G. Escobar, D. Chevreau, R. Ortega, and E. Mendes,
“An adaptive passivity-based controler for unity power
factor rectifier,” Transactions on control system tech-
nology, vol. 9, no. 4, pp. 637–644, 2001.

(b) Power control of the exchanged energy with the utility grid.

Fig. 9. Power control for PS and PG .

Fernando Ornelas-Tellez was born in Patzcuaro,

Michoacán, México. He received the B. Sc. degree
from Instituto Tecnológico de Morelia (ITM) in
2005, the M.Sc. and D.Sc. degrees in electrical
engineering from the Advanced Studies and Re-
search Center of the National Polytechnic Insti-
tute (CINVESTAV-IPN), Guadalajara Campus, in
Fig. 10. Bus voltage control. 2008 and 2011, respectively. Since 2012 he has
been with Universidad Michoacana de San Nicolas
de Hidalgo, where he is currently a professor
of Electrical Engineering graduate programs. His
equation control,” Journal of Guidance, Control and research interest centers on neural control, direct and inverse optimal control,
Dynamics, vol. 21, no. 5, pp. 767–773, 1998. passivity and their applications to robots, electrical machines and power
[6] M. Vidyasagar, Nonlinear Systems Analysis, 2nd ed. electronics. He is an IEEE member.
Englewood Cliffs, NJ, USA: Prentice-Hall, 1993.
[7] T. Cimen, “State-dependent Riccati equation (SDRE)
control: A survey,” in Proc. of the 17th World Congress,
The Int. Feredation of Automatic Control, Seoul, Korea,
July 2008, pp. 3761–3775.
[8] J. R. Cloutier, “State-dependent riccati equation tech-
niques: An overview,” in Proc. of the 1997 American
Control Conf., 1997., vol. 2, jun 1997, pp. 932–936.
[9] H. T. Banks, B. M. Lewis, and H. T. Tan, “Nonlin-

‹,(((

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 18,2021 at 20:59:59 UTC from IEEE Xplore. Restrictions apply.