You are on page 1of 23

Electric Power Systems Research 209 (2022) 108024

Contents lists available at ScienceDirect

Electric Power Systems Research


journal homepage: www.elsevier.com/locate/epsr

A nonlinear optimal control approach for shipboard AC/DC microgrids


G. Rigatos *, a, M.A. Hamida b, M. Abbaszadeh c, P. Siano d
a
Unit of Industrial Autom. Industrial Systems Institute, Rion Patras, 26504, Greece
b
LS2N, CNRS UMR 6004, Ecole Centrale de Nantes, Nantes 44321, France
c
Dept. ECS Engineering, Rensselaer Polytechnic Inst., NY 12065, USA
d
Dept. of Innovation Syst. University of Salerno Fisciano, 84084, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Shipboard AC/DC microgrids are used for power supply and electric propulsion in vessels. An indicative form of
Shipboard AC/DC microgrids such a microgrid comprises diesel engines or gas turbines that provide power for the rotation of synchronous or
Electric ships asynchronous generators. Next, the AC output voltage of the generators is turned into DC voltage with the use of
Ship power supply
AC to DC converters and is distributed through DC voltage buses to the ship’s compartments. Besides, with the
Ship propulsion
Nonlinear H-infinity control
use of DC to AC inverters voltage excitation is provided to synchronous or asynchronous motors which can be
Taylor series expansion used in turn for the vessel’s propulsion. The dynamic model of the considered shipboard AC/DC microgrid, being
Jacobian matrices initially expressed in a nonlinear and multivariable state-space form, undergoes approximate linearization
Riccati equation around a temporary operating point that is recomputed at each time-step of the control method. The linearization
Global stability relies on first-order Taylor series expansion and on the computation of the associated Jacobian matrices. For the
Differential flatness properties linearized state-space model of the shipboard AC/DC microgrid a stabilizing optimal (H-infinity) feedback
controller is designed. This controller stands for the solution to the nonlinear optimal control problem of the AC/
DC microgrid under model uncertainty and external perturbations. To compute the controller’s feedback gains an
algebraic Riccati equation is repetitively solved at each iteration of the control algorithm. The global stability
properties of the control method are proven through Lyapunov analysis. Finally, to implement state estimation-
based control of the shipboard AC/DC microgrid, without the need to measure its entire state vector, the H-
infinity Kalman Filter is used as a robust state estimator. The article’s method provides one of the few algo­
rithmically simple and computationally efficient solutions for the nonlinear optimal control problem of ship­
board microgrids.

1. Introduction comprises both AC and DC elements [13], [14–16]. For instance diesel
engines and gas turbines provide torque that causes the turn motion of
During the last years, the concept of the all-electric-ship has gained the rotor of synchronous or asynchronous power generators. The gen­
much acceptance, because it exhibits several advantages [1–3]. Actu­ erators give voltage and power to the shipboard power grid. Often the
ally, it enables to develop highly reliable and powerful propulsion sys­ AC voltage at the output of the generators is turned into DC voltage with
tems, that generate less polluting emissions while also keeping noise at the use of AC/DC converters, and next a DC voltage bus is used to
low levels [4–6]. The feasibility of the all-electric-ship approach de­ distribute voltage to the ship’s compartments, as well as to motors that
pends on the implementation of automatic control methods that ensure can be used for the vessel’s propulsion. Usually there are three-phase
the accurate, synchronized and faultless functioning of power genera­ motors of the synchronous or asynchronous type which have to be fed
tors and power electronics [7–9]. Due to the complexity of the dynamics with alternating voltage. Consequently, they are connected to the pre­
of prime-movers, power generators and power electronics that consti­ viously noted DC bus through DC/AC inverters. So far there have been
tute AC/DC shipboard microgrids, the solution of the related nonlinear several results on linear and nonlinear control of shipboard AC/DC
optimal control problem and of the associated power management microgrids [17]-[18]. Under the linear or nonlinear dynamics assump­
problem is a non-trivial task [10–12]. Such a type of microgrids tion model-predictive control schemes of the MPC or NMPC type are

* Corresponding author.
E-mail addresses: grigat@ieee.org (G. Rigatos), mohamed.hamida@ec-nantes.fr (M.A. Hamida), masouda@ualberta.ca (M. Abbaszadeh), psiano@unisa.it
(P. Siano).

https://doi.org/10.1016/j.epsr.2022.108024
Received 4 December 2021; Received in revised form 21 March 2022; Accepted 18 April 2022
Available online 28 April 2022
0378-7796/© 2022 Elsevier B.V. All rights reserved.
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

respectively considered [19–21]. Elaborated methods aiming at advantages. (i) it has improved performance when compared against
achieving control and at optimizing energy management in shipboard other nonlinear control schemes that one can consider for the dynamic
AC/DC microgrids have been developed [22–24]. Nonlinear control of model of hybrid microgrids (such as Lie algebra-based control, differ­
shipboard microgrids is a growing research area, since it is aimed to ential flatness theory-based control, Model-based Predictive Control,
optimize power management and power efficiency in all types of vessels Nonlinear Model-based Predictive Control, Sliding-mode control,
[25–29]. Backstepping control, etc.), (ii) it achieves fast and accurate tracking of
The present article proposes a nonlinear optimal control method for all reference setpoints for the hybrid microgrid under moderate varia­
the dynamic model of a shipboard AC/DC microgrid that comprises tions of the control inputs, (iii) it minimizes the dispersion of energy in
Diesel engines, induction generators and AC/DC converters. This power the implementation of control and in the functioning of this hybrid
system is used for supplying DC voltage to a DC bus that distributes microgrid.
power to the various compartments of the vessel. First, the dynamic In particular, a comparison of the proposed nonlinear optimal (H-
model of the AC/DC microgrid undergoes approximate linearization infinity) control method against other linear and nonlinear control
around a temporary operating point which is recomputed at each time- schemes for the shipboard AC/DC microgrid, shows the following: (1)
step of the control algorithm, and which is defined by the present value unlike global linearization-based control approaches, such as Lie
of the system’s state vector and by the last sampled value of the control algebra-based control and differential flatness theory-based control, the
inputs vector [13]. The linearization process relies on first-order Tay­ optimal control approach does not rely on complicated transformations
lor-series expansion and on the associated Jacobian matrices [30–33]. (diffeomorphisms) of the system’s state variables. Besides, the computed
The model inaccuracies which are due to the truncation of higher-order control inputs are applied directly on the initial nonlinear model of the
terms from the Taylor series are considered to be perturbations which shipboard AC/DC microgrid’s dynamics and not on its linearized
are asymptotically compensated by the robustness of the control algo­ equivalent. The inverse transformations which are met in global
rithm. For the linearized model of the shipboard AC/DC microgrid, a linearization-based control are avoided and consequently one does not
stabilizing feedback H-infinity controller is designed. come against the related singularity problems. (2) unlike Model Pre­
The proposed H-infinity controller implements a solution for the dictive Control (MPC) and Nonlinear Model Predictive control (NMPC),
nonlinear optimal control problem of the AC/DC microgrid, under the proposed control method is of proven global stability. It is known
model uncertainty and external perturbations. Actually, it implements a that MPC is a linear control approach that if applied to the nonlinear
min-max differential game taking place between the controller and the dynamics of the shipboard AC/DC microgrid the stability of the control
model uncertainty terms, in which (i) the controller tries to minimize a loop will be lost. Besides, in NMPC the convergence of its iterative search
quadratic function of the state-vector’s tracking error (ii) the model for an optimum depends on initialization and parameter values selection
uncertainty terms try to maximize this cost function. To compute the and consequently the global stability of this control method cannot be
stabilizing feedback gains of the H-infinity controller, an algebraic always assured. (3) unlike sliding-mode control and backstepping con­
Riccati equation is repetitively solved at each time-step of the control trol the proposed optimal control method does not require the state-
method [34–36]. The global stability properties of the control scheme space description of the system to be found in a specific form. About
are proven through Lyapunov analysis. First, it is demonstrated that the sliding-mode control it is known that when the controlled system is not
H-infinity tracking performance criterion holds. This signifies elevated found in the input-output linearized form the definition of the sliding
robustness against model uncertainty and external perturbations [13], surface can be an intuitive procedure. About backstepping control it is
[37]. Furthermore, under moderate conditions it is proven that the known that it can not be directly applied to a dynamical system if the
H-infinity control scheme is globally asymptotically stable. The pro­ related state-space model is not found in the triangular (backstepping
posed control approach retains the advantages of linear optimal control, integral) form. (4) unlike PID control, the proposed nonlinear optimal
that is fast and accurate tracking of reference setpoints under moderate control method is of proven global stability, the selection of the con­
variations of the control inputs [13]. To implement state troller’s parameters does not rely on a heuristic tuning procedure, and
estimation-based control, without the need to measure the entire state the stability of the control loop is assured in the case of changes of
vector of the AC/DC microgrid, the H-infinity Kalman Filter is used as a operating points (5) unlike multiple local models-based control the
robust state estimator [13], [38]. nonlinear optimal control method uses only one linearization point and
The present article provides one of the few existing solutions to the needs the solution of only one Riccati equation so as to compute the
nonlinear optimal control problem of shipboard AC/DC microgrids stabilizing feedback gains of the controller. Consequently, in terms of
which is of proven global stability while also remaining computationally computation load the proposed control method for the shipboard AC/DC
efficient [36]. Preceding results on the use of H-infinity control to microgrid’s dynamics is much more efficient.
nonlinear dynamical systems were limited to the case of The structure of the article is as follows: in Section 2 the dynamic
affine-in-the-input systems with drift-only dynamics and considered that model of the shipboard AC/DC microgrid is analyzed and the related
the control inputs gain matrix is not dependent on the values of the state-space description is obtained. Besides, linearization of the micro­
system’s state vector. Moreover, in these approaches linearization was grid’s model is performed and the associated Jacobian matrices are
performed around points of the desirable trajectory whereas in the obtained. In Section 3 an H-infinity feedback controller is designed for
present article’s control method the linearization points are related with the shipboard AC/DC microgrid. In Section 4 the global asymptotic
the value of the state vector at each sampling instance as well as with the stability properties of the proposed nonlinear optimal control method
last sampled value of the control inputs vector. The Riccati equation are proven through Lyapunov analysis. Besides, the H-infinity Kalman
which has been proposed for computing the feedback gains of the Filter is introduced as a robust state estimator. In Section 5 the differ­
controller is novel, so is the presented global stability proof through ential flatness properties of the hybrid shipboard microgrid are proven.
Lyapunov analysis. Furthermore, a flatness-based controller is designed. In Section 6 the
The article has a meaningful contribution because it provides one of tracking performance of the nonlinear optimal control method is
the few algorithmically simple and computationally efficient solutions confirmed through simulation experiments. Moreover, the nonlinear
for the nonlinear optimal control problem of shipboard microgrids. The optimal control method is experimentally tested and compared, at the
proposed nonlinear optimal control method exhibits specific simulation level, against flatness-based control for the dynamics of this

2
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 1. Diagram of the shipboard AC/DC microgrid

microgrid. Finally, in Section 7 concluding remarks are stated.


θ̇g = ωg (4)
2. Dynamic model of the shipboard AC/DC microgrid
1 Kb n ( )
ω̇g = Tm − ωg − p irq ψ sd − ird ψ sq (5)
J J J
2.1. Modelling of the ship’s AC/DC microgid
1 M
The ship’s AC/DC microgrid comprises Diesel engines connected to ψ̇ sd = ωdq ψ rq − ψ sd − ird + vd (6)
τs τs
induction generators while these asynchronous generators are con­
nected in turn to an AC to DC converters. The converters provide DC 1 M
ψ̇ sq = − ψ sq − ωdq ψ sd + irq + vq (7)
voltage to a DC bus that distributes power to the vessel’s compartments τs τs
as well as to electric motors which can be used for the ship’s propulsion
(Fig. 1). First, the equations of the AC-DC converter’s dynamics are β ( )
i˙rd = βωg ψ sq + ψ sd + ωdq − ωg irq − γ 2 ird − βvd +
1
v (8)
provided [13]: τs σLr rd
Vdc ( )
Li˙d = − Rid + Lωdq iq + vd − η (1) β
i˙rq = ψ sq + βωg ψ sd − γ2 irq − ωdq − ωg ird − βvq +
1
v (9)
2 1 τs σ Lr rq
Vdc where ψ sq , ψ sd , irq , ird are the stator flux and the rotor currents variables,
Li˙q = − Lωdq id − Riq + vq − η (2)
2 2
vsq , vsd , vrq , vrd are the stator and rotor voltages, Ls and Lr are the stator
1 3 3 and rotor inductances, ωg is the rotor’s angular speed, M is the mutual
Cdc V̇ dc = − Vdc + id η1 + iq η2 (3)
Rc 4 4 inductance. Moreover, denoting as Rs and Rr the stator and rotor re­
sistances the following parameters are defined
where id , iq are the line currents (ia , ib , ic ) after transformation in the dq
reference frame, and equivalently vd , vq are the phase voltages va , vb , vc M2 1− σ Ls
σ =1− β= τs =
after transformation in the dq reference frame. Variable Vdc denotes the Lr Ls Mσ Rs
(10)
( )
DC voltage output of the converter, u1 = ηd and u2 = ηq stand for control Lr 1− σ 1
τr = γ2 = +
inputs. The line losses and the transformer conduction losses are Rr στs στr
modelled by resistance R and the inverter switching losses are modeled
by resitance Rc . Parameter L denotes inductance. Moreover, vq is taken The angle of the vectors that describe the magnetic flux ψ s α and ψ sb in
to be 0. the ab inertial reference frame, is first defined for the stator, i.e. ρ =
( )
Next, the equations of the induction generator’s dynamics are pro­ ψs
tan− 1 ψ b . The angle between the inertial reference frame and the
vided: [13] sa

rotating reference frame is taken to be equal to ρ. Moreover, it holds that

3
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

ψs
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
cos(ρ) = ‖ψsa‖, sin(ρ) = ‖ψb‖, and ‖ ψ ‖ = ψ 2sα + ψ 2sb . Therefore, in the
ψ ⎛ ⎞
R vd x3
rotating d − q frame of the generator, and under the condition of field ⎜
⎜ − x1 + ωdq x2 + −
L L 2L
K1 x 8 ⎟

orientation, there will be only one non-zero component of the magnetic ⎜ ⎟
⎜ R vq x3 ⎟
⎜ ⎟
flux ψ sd , while the component of the flux along the q axis equals 0. ⎜ − ωdq x1 − x2 + − K2 x 9 ⎟
⎜ L L 2L ⎟
Furthermore, the equations of the diesel engine’s dynamics are given ⎜



1 3K 3K
[19–21], [36] ⎜ ⎟
1 2
⎛ ⎞ ⎜ − x3 + x x
1 8 + x x
2 9 ⎟
ẋ1 ⎜ R c C dc 4x 3 4x3 ⎟
⎜ ⎟
Tm ⎜ ẋ2 ⎟ ⎜ x ⎟
(11)
5
Ṫ m = − + Ken fen ⎜ ⎟ ⎜ ⎟
τs ⎜ ẋ3 ⎟ ⎜ 1 Kb np ⎟
⎜ ⎟ ⎜ x10 − x5 − (x9 x6 − x8 x7 ) ⎟
⎜ ẋ4 ⎟ ⎜ J J J ⎟
⎜ ⎟ ⎜ ⎟
This model of the Diesel engine signifies that the produced torque is ⎜ ẋ5 ⎟ ⎜
⎜ ⎟=⎜


1 M
proportional to the inflow of fuel to engine, subject to time-delay effects ⎜ ẋ6 ⎟ ⎜
⎜ ⎟ ⎜ ωdq x7 − x6 − x8 + vd ⎟

τs τs
which are determined by the time constant τs . ⎜ ẋ7 ⎟ ⎜
⎜ ⎟ ⎜


⎜ ẋ8 ⎟ ⎜ 1 M ⎟
⎜ ⎟ ⎜ − x7 − ωdq x6 + x9 + vq ⎟
⎝ ẋ9 ⎠ ⎜ τ τ ⎟
2.2. State-space description of the shipboard AC/DC microgrid ⎜ s s ⎟
ẋ10 ⎜ ( ) ⎟
⎜ β ⎟
⎜ βx5 x7 + x6 + ωdq − x5 x9 − γ 2 x8 − βvd ⎟
The state-vector of the shipboard AC/DC microgrid is defined as ⎜
⎜ τs ⎟

⎜β ( ) ⎟
⎜ ⎟
ẋ = [x1 , x2 , x3 , x4 , x5 , x6 , x7 , x8 , x9 , x10 ]T ⇒ ⎜ x7 + βx5 x6 − γ2 x9 − ωdq − x5 x8 − βvq ⎟
[ ]T (12) ⎜ τs



ẋ = id , iq , Vdc , θg , ωg , ψ sd , ψ sq , ird , irq , Tm ⎜ ⎟
⎝ x10 ⎠

τs
Moreover, the control inputs vector is defined as u1 = vrd , u2 = vrq and u3
⎛ ⎞
= fen . Because of connecting the AC to DC converter to the rotor of the 0 0 0
doubly-fed induction generator one has η1 = K1 ⋅ird and η2 = K2 ⋅irq where ⎜
⎜ 0

0 0 ⎟
K1 and K2 are constants. ⎜



⎜ 0 0 ⎟
Using the previous state-space notation, the state-space equations of ⎜ 0 ⎟
⎜ ⎟
the AC-DC converter are re-written as: ⎜ 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
R vd x3 ⎜ ⎟
ẋ1 = − x1 + ωdq x2 + − K1 x8 (13) ⎜ 0 0 0 ⎟⎛ ⎞
L L 2L ⎜ ⎟ u1
⎜ ⎟
+⎜ 0 0 0 ⎟⎝ u2 ⎠ (23)
⎜ ⎟
R vq x3 ⎜ ⎟ u3
ẋ2 = − ωdq x1 − x2 + − K2 x9 (14) ⎜ 0

0 0 ⎟

L L 2L ⎜ 1 ⎟
⎜ 0 0 ⎟
⎜ σL ⎟
1 3K1 3K2 ⎜ r ⎟
ẋ3 = − x3 + x1 x8 + x2 x9 (15) ⎜



Rc Cdc 4x3 4x3 ⎜ 0 1
⎜ 0 ⎟⎟
⎝ σLr ⎠
Next, the state-space equations of the induction generator are re-written 0 0 Ken
as:
ẋ4 = x5 (16) Equivalently, the shipboard AC/DC microgrid is written in the concise
state-space description
1 Kb np
ẋ5 = x10 − x5 − (x9 x6 − x8 x7 ) (17) ẋ = f (x) + g(x)u (24)
J J J

1 M where x ∈ R10×1 , f(x) ∈ R10×1 , g(x) ∈ R10×3 and u ∈ R3×1 .


ẋ6 = ωdq x7 − x6 − x8 + vd (18)
τs τs
2.3. Approximate linearization of the model of the shipboard AC/DC
ẋ7 = −
1
x7 − ωdq x6 +
M
x9 + vq (19) microgrid
τs τs
The state-space model of the shipboard AC/DC microgrid undergoes
β ( ) 1
ẋ8 = βx5 x7 + x6 + ωdq − x5 x9 − γ 2 x8 − βvd + u (20) approximate linearization around the temporary operating point (x∗ ,u∗ ),
τs σ Lr 1 where x∗ is the present value of the system’s state vector and u∗ is the last
( ) sampled value of the control inputs vector. The linearization process
β 1
ẋ9 = x7 + βx5 x6 − γ2 x9 − ωdq − x5 x8 − βvq + u2 (21) relies on first-order Taylor series expansion and on the computation of
τs σ Lr
the associated Jacobian matrices [13]. The modelling error which is due
Next, the state equations of the diesel engine are re-written as: to truncation of higher-order terms in the Taylor series is considered to
be a perturbation which is asymptotically compensated by the robust­
x10
ẋ10 = − + Ken u3 (22) ness of the control method. The associated linearized model is:
τs
ẋ = Ax + Bu + ̃
d (25)
Consequently, the model of the shipboard AC/DC microgrid is written as
where matrices A and B are the system’s Jacobian which are computed

4
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

as follows: Ninth row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂xf91 = 0, ∂∂xf92 = 0, ∂∂xf93 =
A = ∇x [f (x) + g(x)u]∣(x∗ ,u∗ ) ⇒A = ∇x f (x)∣(x∗ ,u∗ ) (26) 0, ∂∂xf94 = 0, ∂∂xf95 = βx6 + x8 , ∂∂xf96 = βx5 , ∂∂xf97 = τβs , ∂∂xf98 = − (ωdq − x5 ), ∂∂xf99 = −
∂f9
γ2 , ∂x10 = 0,
B = ∇u [f (x) + g(x)u]∣(x∗ ,u∗ ) ⇒B = g(x)∣(x∗ ,u∗ ) (27)
Tenth row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂fx101 = 0, ∂∂fx102 = 0, ∂∂fx103 =
The linearization approach which has been followed for implementing
∂f10

0, ∂∂fx104 = 0, ∂f10
= 0, ∂f10
= 0, ∂∂fx107 = 0, ∂f10
= 0, ∂f10
= 0, ∂x10 =1
.
the nonlinear optimal control scheme results into a quite accurate model ∂x5 ∂x6 ∂x8 ∂ x9 τs

of the system’s dynamics. Consider for instance the following affine-in-


the-input state-space model 3. Design of an H-infinity nonlinear feedback controller

ẋ = f (x) + g(x)u⇒ẋ = [f (x∗ ) + ∇x f (x)∣x∗ (x − x∗ )] + [g(x∗ ) + ∇x g(x)∣x∗ (x − x∗ )]u∗ + g(x∗ )u∗ + g(x∗ )(u − u∗ ) + ̃
d1 ⇒ ẋ
= [∇x f (x)∣x∗ + ∇x g(x)∣x∗ u∗ ]x + g(x∗ )u − [∇x f (x)∣x∗ + ∇x g(x)∣x∗ u∗ ]x∗ + f (x∗ ) + g(x∗ )u∗ + ̃
d1 (28)

3.1. Equivalent linearized dynamics of the shipboard AC/DC microgrid


where ̃d1 is the modelling error due to truncation of higher order terms
in the Taylor series expansion of f(x) and g(x). Next, by defining A = After linearization around its current operating point, the dynamic
[∇x f(x)∣x∗ + ∇x g(x)∣x∗ u∗ ], B = g(x∗ ) one obtains model for the shipboard AC/DC microgid is written as

ẋ = Ax + Bu − Ax∗ + f (x∗ ) + g(x∗ )u∗ + ̃


d1 (29)
ẋ = Ax + Bu + d1 (31)

Parameter d1 stands for the linearization error in the shipboard AC/DC


Moreover by denoting ̃ d = − Ax∗ + f(x∗ ) + g(x∗ )u∗ + ̃
d1 about the cu­
microgrid ’s model that was given previously in Eq. (25). The reference
mulative modelling error term in the Taylor series expansion procedure
setpoints for the state vector of the aforementioned dynamic model are
one has
denoted by xd = [xd1 ,⋯,xd10 ]. Tracking of this setpoint vector is achieved
ẋ = Ax + Bu + ̃
d (30) after applying the control input u∗ . At every time instant the control
input u∗ is assumed to differ from the control input u appearing in Eq.
which is the approximately linearized model of the dynamics of the (31) by an amount equal to Δu, that is u∗ = u + Δu
system of Eq. (25). The term f(x∗ ) + g(x∗ )u∗ is the derivative of the state
ẋd = Axd + Bu∗ + d2 (32)
vector at (x∗ , u∗ ) which is almost annihilated by − Ax∗ .
The computation of the elements of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ )
The dynamics of the controlled system described in Eq. (31) can be also
gives
written as
First row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂xf11 = − R ∂f1
L , ∂x2 = ωdq , ∂∂xf13
∂f1 ∂f1 ∂f1 ∂f1 ∂f1 ∂f1 ẋ = Ax + Bu + Bu∗ − Bu∗ + d1 (33)
=− K1
2L x8 , ∂x4 = 0, ∂x5 = 0, ∂x6 = 0, ∂x7 = 0, ∂x8 =− K1
2L x3 , ∂x9 = 0, and ∂∂xf101
= 0. and by denoting d3 = − Bu∗ + d1 as an aggregate disturbance term one
Second row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂xf21 = − ωdq , ∂∂xf22 = − obtains
R ∂f2 ∂f2
L , ∂x3 = − K2
2Lx9 , ∂x4 = 0, ∂∂xf25 = 0, ∂∂xf26 = 0, ∂∂xf27 = 0, ∂∂xf28 = 0, ∂∂xf29 = − K2
2Lx3 , ẋ = Ax + Bu + Bu∗ + d3 (34)
∂f2
and ∂x10 = 0.
∂f3 ∂f3
By subtracting Eq. (32) from Eq. (34) one has
3K1
Third row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂x1 = 4Cdc x8 , ∂x2 =
∂f3 ∂f3 ẋ − ẋd = A(x − xd ) + Bu + d3 − d2 (35)
3K2 1
4Cdc x9 , ∂x3 = − Cdc Rc , ∂x4 = 0, ∂∂xf35 = 0, ∂∂xf36 = 0, ∂∂xf37 = 0, ∂∂xf38 = 4C
3K1
dc
x1 , ∂∂xf39 =
3K2 ∂f3
4Cdc x3 , and ∂x10 = 0.
By denoting the tracking error as e = x − xd and the aggregate distur­
Fourth row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂xf41 = 0, ∂∂xf42 = 0, ∂∂xf43 = d = d3 − d2 , the tracking error dynamics becomes
bance term as ̃
0, ∂x4 = 0, ∂∂xf45 = 1, ∂∂xf46 = 0, ∂∂xf47 = 0, ∂∂xf48 = 0, ∂∂xf49 = 0, ∂∂xf104 = 0.
∂f4
ė = Ae + Bu + ̃
d (36)
Fifth row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂xf51 = 0, ∂∂xf52 = 0, ∂∂xf53 =
n n n n The above linearized form of the shipboard AC/DC microgrid’s model
0, ∂x4 = − KJb , ∂∂xf55 = 0, ∂∂xf56 = − Jp x9 , ∂∂xf57 = Jp x8 , ∂∂xf58 = Jp x7 , ∂∂xf59 = − Jp x6 ,
∂f5
can be efficiently controlled after applying an H-infinity feedback con­
and ∂∂xf105 = 0. trol scheme.
Sixth row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂xf61 = 0, ∂∂xf62 = 0, ∂∂xf63 =
0, ∂∂xf64 = 0, ∂∂xf65 = 0, ∂∂xf66 = − τ1s , ∂∂xf67 = ωdq , ∂∂xf68 = M ∂f6 ∂f6
τs , ∂x9 = 0, and ∂x10 = 0. 3.2. The nonlinear H-infinity control
Seventh row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂x1 = 0, ∂x2 = 0, ∂∂xf73
∂f7 ∂ f7

The initial nonlinear model of the shipboard AC/DC microgrid is in


= 0, ∂∂xf74 = 0, ∂∂xf75 = 0, ∂∂xf76 = − ωdq , ∂∂xf77 = − τ1s , ∂∂xf78 = 0, ∂∂xf79 = M ∂f7
τs , and ∂x10 = the form
0.
Eighth row of the Jacobian matrix ∇x f(x)∣(x∗ ,u∗ ) : ∂∂xf81 = 0, ∂∂xf82 = 0, ∂∂xf83 = ẋ = f (x, u) x ∈ Rn , u ∈ Rm (37)

0, ∂∂xf84
= 0, ∂∂xf85 = − βx7 − x9 , ∂∂xf86 = β ∂f8
τs , ∂x7 =− βx5 , ∂∂xf88 =− γ2 , ∂∂xf89 = ωdq − Linearization of the model of the shipboard AC/DC microgrid is per­
x5 , and ∂∂xf108 = 0. formed at each iteration of the control algorithm around its present
operating point (x∗ , u∗ ) = (x(t), u(t − Ts )). The linearized equivalent of

5
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

control input and weight coefficient ρ determines the reward of the


disturbances’ effects. It is assumed that (i) The energy that is transferred
∫ ∞ ̃T ̃
from the disturbances signal ̃ d(t) is bounded, that is 0 d (t)d(t)dt < ∞,
(ii) matrices [A, B] and [A, L] are stabilizable, (iii) matrix [A, C] is
detectable. In the case of a tracking problem the optimal feedback
control law is given by
u(t) = − Ke(t) (41)

with e = x − xd to be the tracking error, and K = 1r BT P where P is a


positive definite symmetric matrix. As it will be proven in Section 4,
matrix P is obtained from the solution of the Riccati equation
( )
2 1
AT P + PA + Q − P BBT − 2 LLT P = 0 (42)
r ρ

where Q is a positive semi-definite symmetric matrix. The worst-case


disturbance is given by
1
̃
d(t) = 2 LT Pe(t) (43)
ρ
Fig. 2. Diagram of the control scheme for the shipboard AC/DC microgrid
The solution of the H-infinity feedback control problem for the ship­
board AC/DC microgrid and the computation of the worst-case distur­
the system is described by
bance that the related controller can sustain, comes from superposition
ẋ = Ax + Bu + L̃
d x ∈ Rn , u ∈ Rm , ̃
d ∈ Rq (38) of Bellman’s optimality principle when considering that the microgrid is
affected by two separate inputs (i) the control input u (ii) the cumulative
where matrices A and B are obtained from the computation of the pre­ disturbance input ̃
d(t). Solving the optimal control problem for u, that is
viously defined Jacobians and vector ̃
d denotes disturbance terms due to for the minimum variation (optimal) control input that achieves elimi­
linearization errors. The problem of disturbance rejection for the line­ nation of the state vector’s tracking error, gives u = − 1r BT Pe. Equiva­
arized model that is described by lently, solving the optimal control problem for ̃d, that is for the worst-
ẋ = Ax + Bu + L̃
d d = ρ12 LT Pe.
case disturbance that the control loop can sustain gives ̃
(39)
y = Cx The diagram of the considered control loop for the shipboard AC/DC
microgrid is depicted in Fig. 2.
where x ∈ Rn , u ∈ Rm , ̃
d ∈ Rq and y ∈ Rp , cannot be handled efficiently if
the classical LQR control scheme is applied. This is because of the ex­ 4. Lyapunov stability analysis
istence of the perturbation term ̃ d. The disturbance term ̃ d apart from
modeling (parametric) uncertainty and external perturbation terms can 4.1. Stability proof
also represent noise terms of any distribution.
In the H∞ control approach, a feedback control scheme is designed Through Lyapunov stability analysis it will be shown that the pro­
for setpoints tracking by the system’s state vector and simultaneous posed nonlinear control scheme assures H∞ tracking performance for the
disturbance rejection, considering that the disturbance affects the sys­ shipboard AC/DC microgrid, and that in case of bounded disturbance
tem in the worst possible manner. The disturbances’ effects are incor­ terms asymptotic convergence to the reference setpoints is achieved.
porated in the following quadratic cost function: The tracking error dynamics for the shipboard AC/DC microgrid is
∫ written in the form [13], [36]
1 T[ T T ]
J(t) = y (t)y(t) + ruT (t)u(t) − ρ2 ̃
d (t)̃
d(t) dt, r, ρ > 0 (40)
2 0 ė = Ae + Bu + L̃
d (44)

The significance of the negative sign in the cost function’s term that is where in the shipboard AC/DC microgrid’s case L ∈ R10×10 is the
associated with the perturbation variable ̃ d(t) is that the disturbance disturbance inputs gain matrix. Variable ̃
d denotes model uncertainties
tries to maximize the cost function J(t) while the control signal u(t) tries and external disturbances of the microgrid’s model. The following
to minimize it. The physical meaning of the relation given above is that Lyapunov equation is considered
the control signal and the disturbances compete to each other within a
1
min-max differential game. This problem of min-max optimization can V = eT Pe (45)
2
be written as minu max̃J(u, ̃d).
d
The objective of the optimization procedure is to compute a control where e = x − xd is the tracking error. By differentiating with respect to
signal u(t) which can compensate for the worst possible disturbance, time one obtains
that is externally imposed to the shipboard AC/DC microgrid. However, 1 1
the solution to the min-max optimization problem is directly related to V̇ = ėT Pe + ePė⇒
2 2
the value of parameter ρ. This means that there is an upper bound in the (46)
1 T 1 T
disturbances magnitude that can be annihilated by the control signal. ̃
V̇ = [Ae + Bu + Ld] Pe + e P[Ae + Bu + L̃
d]⇒
2 2

3.3. Computation of the feedback control gains V̇ =


1[ T T
e A + uT BT + ̃
T ]
d LT Pe+
2
(47)
For the linearized system given by Eq. (39) the cost function of Eq. 1
+ eT P[Ae + Bu + L̃d]⇒
(40) is defined, where coefficient r determines the penalization of the 2

6
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

1 1 1 T T 1 T 1 T̃
V̇ = eT AT Pe + uT BT Pe + ̃d L Pe+ V̇ ≤ − e Qe + ρ2 ̃
d d (59)
2 2 2
(48) 2 2
1 T 1 1
e PAe + eT PBu + eT PL̃ d Eq. (59) shows that the H∞ tracking performance criterion is satisfied.
2 2 2
The integration of V̇ from 0 to T gives
The previous equation is rewritten as ∫T ∫ ∫T
( ) 1 T 1
1 ( ) 1 1 V̇(t)dt ≤ − ‖ e ‖2Q dt + ρ2 ‖̃
d‖2 dt⇒
V̇ = eT AT P + PA e + uT BT Pe + eT PBu + 2 0 2
(60)
0 0
2 2 2 ∫T ∫T
( ) (49)
1̃T T 1 T ̃ 2V(T) + ‖ e ‖2Q dt ≤ 2V(0) + ρ2 ‖̃d‖2 dt
+ d L Pe + e PLd 0 0
2 2
Moreover, if there exists a positive constant Md > 0 such that
Assumption: For given positive definite matrix Q and coefficients r and ρ ∫∞
there exists a positive definite matrix P, which is the solution of the ‖̃
d‖2 dt ≤ Md (61)
following matrix equation 0
( )
2 1
AT P + PA = − Q + P BBT − 2 LLT P (50) then one gets
r ρ ∫∞
‖ e ‖2Q dt ≤ 2V(0) + ρ2 Md (62)
Moreover, the following feedback control law is applied to the system 0

∫∞
u=−
1 T
B Pe (51) Thus, the integral 0 ‖ e ‖2Q dt is bounded. Moreover, V(T) is bounded
r and from the definition of the Lyapunov function V in Eq. (45) it be­
comes clear that e(t) will be also bounded since e(t) ∈ Ωe =
By substituting Eq. (50) and Eq. (51) in V̇ one obtains ⃒
[ ( ) ] {e⃒eT Pe ≤ 2V(0) + ρ2 Md }. According to the above and with the use of
1 2 1
V̇ = eT − Q + P BBT − 2 LLT P e+ Barbalat’s Lemma one obtains limt→∞ e(t) = 0.
2 r ρ After following the stages of the stability proof one arrives at Eq.
( ) (52)
1 T (59) which shows that the H-infinity tracking performance criterion
+e PB − B Pe + eT PL̃
T
d⇒ holds. By selecting the attenuation coefficient ρ to be sufficiently small
r
( ) and in particular to satisfy ρ2 <‖ e ‖2 /‖ ̃
d ‖2 one has that the first de­
Q

V̇ = −
1 T 1
e Qe + eT PBBT Pe −
1 T
e PLLT
Pe rivative of the Lyapunov function is upper bounded by 0. This condition
2 r 2ρ2
(53) holds at each sampling instance and consequently global stability for the
1 T control loop can be concluded.
− e PBBT Pe + eT PL̃
d
r
4.2. Robust state estimation with the use of the H∞ Kalman Filter
which after intermediate operations gives
1 T 1 T The control loop has to be implemented with the use of information
V̇ = − e Qe − e PLLT Pe + eT PL̃
d (54)
2 2ρ2 provided by a small number of sensors and by processing only a small
number of state variables. Actually, there is no need to measure the turn
or, equivalently speed of the rotor of the induction generator and the magnetic flux at the
1 T 1 T stator of the generator. To reconstruct the missing information about the
V̇ = − e Qe − e PLLT Pe+ state vector of the shipboard AC/DC microgrid it is proposed to use a
2 2ρ2
(55) filtering scheme and based on it to apply state estimation-based control
1 1 T T
d+ ̃
+ eT PL̃ d L Pe [13], [35]. By denoting as A(k), B(k) and C(k) the discrete-time equiv­
2 2 alents of matrices A, B and C of the linearized state-space model of the
Lemma: The following inequality holds [13], [35] system, the recursion of the H∞ Kalman Filter, for the model of the
shipboard AC/DC microgrid, can be formulated in terms of a measure­
1 T ̃ 1̃ T
e Ld + dL Pe −
1 T 1 T̃
e PLLT Pe ≤ ρ2 ̃
d d (56) ment update and a time update part
2 2 2 ρ2 2 Measurement update:
( )2 [ ]− 1
1
D(k) = I − θW(k)P− (k) + CT (k)R(k)− 1 C(k)P− (k)
Proof: The binomial ρα − b is considered. Expanding the left part of − 1 (63)
ρ K(k) = P− (k)D(k)CT (k)R(k)
the above inequality one gets x − (k) + K(k)[y(k) − Ĉx − (k)]
̂x (k) = ̂

ρ2 a 2 +
1 1 1
b2 − 2ab ≥ 0⇒ ρ2 a2 + 2 b2 − ab ≥ 0⇒ Time update:
ρ2 2 2ρ
(57) ̂x − (k + 1) = A(k)x(k) + B(k)u(k)
1 2 1 2 2 1 1 1 2 1 2 2 (64)
ab − b ≤ ρ a ⇒ ab + ab − b ≤ ρa P− (k + 1) = A(k)P− (k)D(k)AT (k) + Q(k)
2ρ2 2 2 2 2ρ2 2
where it is assumed that parameter θ is sufficiently small to assure that
d and b = eT PL and the
The following substitutions are carried out: a = ̃ the covariance matrix P− (k)− 1 − θW(k) + CT (k)R(k)− 1 C(k) will be pos­
previous relation becomes itive definite. When θ = 0 the H∞ Kalman Filter becomes equivalent to
1̃T T 1 1 T 1 T̃ the standard Kalman Filter. One can measure only a part of the state
d L Pe + eT PL̃
d− e PLLT Pe ≤ ρ2 ̃
d d (58) vector of the shipboard AC/DC microgrid, and estimate through filtering
2 2 2ρ2 2
the rest of the state vector elements. For instance, one can measure state
Eq. (58) is substituted in Eq. (55) and the inequality is enforced, thus variables x1 = id , x2 = iq , x3 = Vdc , x4 = θg , x8 = ird , x9 = irq and x10 =
giving [13], [35] Tm and can estimate through filtering the rotor’s turn speed x5 = ωg . and

7
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

the magnetic flux of the stator of the induction generator that is x6 = ψ sd After intermediate operations and regrouping of terms one arrives at the
(while due to the field orientation condition x7 = ψ sq is 0). following binomial
To elaborate on the matrices which appear in the Measurement update ( ) [ ( ) ]
3R 3L 3vd 3R 3L 2 3vq
part and in the Time update part of the H-infinity Kalman Filter, the − + x12 + x1 + ẏ1 − + y21 + y1 − y2
following can be noted: Matrix R(k) ∈ R7×7 is the measurement noise 2 2Rc Cdc 2 2 2Rc Cdc Rc Cdc 2
covariance matrix, that is the covariance matrix of the measurement =0
error vector of the system. Matrix P− (k) ∈ R10×10 is the a-priori state (69)
vector estimation error covariance matrix of the system, that is the
covariance matrix of the state vector estimation error prior to receiving Eq. (69) is a binomial of x1 . The maximum solution for x1 is retained.
the updated measurement of the system’s outputs. Matrix W(k) ∈ R10×10 Thus, state variable x1 is a differential function of the flat outputs of the
is a weight matrix which defines the significance to be attributed by the system, which can be denoted as
H-infinity Kalman Filter in minimizing the state vector’s estimation x1 = h1 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (70)
error, relatively to the effects that the noise affecting the system may
have. Actually, W is a diagonal matrix and the elements on its diagonal By substituting Eq. (70) into Eq. (67) one obtains
are given the value 10− 3 . Finally, matrix D(k) ∈ R10×10 stands for a ( )
modified a-posteriori state vector estimation error covariance matrix, y1 − 3L h1 (y1 , ẏ1 , y2 .ẏ2 , y3 , ẏ3 )2 + y22
x23 = 4
Cdc (71)
that is the covariance matrix of the state vector estimation error after 2
receiving the updated measurement of the system’s outputs. Besides,
about the process noise covariance matrix Q ∈ R10×10 that appears in the Therefore, state variable x3 is also a differential function of the flat
filter’s Time-Update part it holds that this is also a diagonal matrix and outputs of the system, or
the elements on its diagonal are given the value 10− 4 . Conclusively, the x3 = h3 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (72)
H-infinity Kalman Filter retains the structure of the typical Kalman Fil­
ter, that is a recursion in discrete time comprising a Time update part Next, from the first row of the state-space model of the shipboard AC/DC
(computation of variables prior to receiving measurements) and a microgrid given in Eq. (13) one solves with respect to x8 , thus obtaining
Measurement update part (computation of variables after measurements vd
ẋ1 + RLx1 − ωdq x2 −
have been received). There is a modified a-posteriori state vector esti­ x8 = − K1
L
(73)
x
mation error covariance matrix, which in turn takes into account a 2L 3

weight matrix that defines the accuracy of the state estimation under the
which signifies that state variable x9 can be written as a differential
effects of elevated noise.
function of the flat outputs of the system, or
5. Flatness-based control for the shipboard microgrid x8 = h8 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (74)

5.1. Differential flatness properties of the shipboard AC/DC microgrid Besides, from the second row of the shipboard AC/DC microgrid given in
Eq. (14) one solves with respect to x9 . This gives:
It will be proven that the shipboard AC/DC microgrid model is vq
ẋ2 − ωd qx1 + RLx2 −
differentially flat, with flat outputs y1 = 3L 2 2 Cdc 2
4 (x1 + x2 ) + 2 x3 , y2 = x2 and
x9 = − K2
L
(75)
x
y3 = x4 . The differential flatness property signifies that all state vari­ 2L 3

ables and the control inputs of the system can be written as differential which signifies that state variable x8 can be written as a differential
functions of the flat outouts. It holds that: function of the flat outputs of the system, or
3L x9 = h8 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (76)
ẏ1 = (x1 ẋ1 + x2 ẋ2 ) + Cdc x3 ẋ3 ⇒
2
( )
3L R v d K1 Next, from the fourth row of the state-space model of the system one has
ẏ1 = x1 − x1 + ωdq x2 + − x3 x8 +
2 L L 2L
( ) (65) x5 = ẋ4 ⇒x5 = ẏ2 (77)
3L R vq K2
+ x2 − ωdq x1 − x3 + − x3 x9 +
2 L L 2L which signifies that state variable x5 is also a differential function of the
(
1 3K1 3K2
) flat outputs of the system.
+Cdc x3 − x3 + x1 x8 + x2 x9
Cdc Rc 4Cdc 4Cdc 1 K1 np
ẋ5 = − x10 − x2 − (x9 x6 − x8 x7 ) (78)
J J J
After intermediate operations and simplifications of terms the previous
relation gives Besides, using the field orientation concept one obtains x7 = ψ sq = 0,
3R ( 2 ) 1 2 3vd 3vd thus the fifth row of the state-space model of the system becomes
ẏ1 = − x + x22 − x + x1 + x2 (66)
2 1 Rc 3 2 2 1 K1 np
ẋ5 = − x10 − x2 − x9 x6 (79)
Cdc 2
J J J
Besides, from the definition of the flat output y1 = 3L 2 2
4 (x1 +x2 ) + 2 x3 one
can solve for x23 , thus obtaining Moreover, from the sixth row of the state-space model of the shipboard
( 2 ) AC/DC microgrid, given in Eq. (18) one has
y1 − 3L x1 + x22
x23 = 4
(67) 1 M
Cdc
2 ẋ6 = ωdq x7 − x6 + x8 + vd (80)
τs τs
By substituting Eq. (67) into Eq. (66) one obtains
while using again the field orientation concept ψ sq = 0 one obtains
[ ]
3R ( 2 ) 1 2 3L ( 2 ) 3vd 3vq
ẏ1 = − x1 + x22 − y1 − x1 + x22 + x1 + x2 (68) 1 M
2 Rc Cdc 2 2 4 ẋ6 = − x6 + x8 + vd (81)
τs τs

8
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Additionally, from the seventh row of the state-space model given in Eq. define feasible setpoints for the state-variables of the shipboard AC/DC
(19) one has microgrid. Actually, one can define setpoints for the flat outputs of the
system yi i = 1, 2, 3 and using the relations that connect state variables
ẋ7 = −
1
x7 + ωdq x6 +
M
x9 + vq (82) xi i = 1, 2, ⋯, 10 to the flat outputs and their derivatives can advance to
τs τs the computation of setpoints xdi i = 1, 2, ⋯, 10.

Using again the field orientation concept x7 = ψ sq = 0 one obtains


5.2. Design of a flatness-based controller for the shipboard AC/DC
M microgrid
0 = ωdq x6 + x9 + vq (83)
τs
By proving that the state-space model of the shipboard AC/DC
From Eq. (83) one can solve with respect to x6 which gives microgrid is differentially flat, it can be ensured that this system can be
M written in the input-output linearized form, or equivalently in the ca­
− τs x9 − vd
nonical Brunvsky form. By differentiating successively the flat output y1
x6 = (84)
ωdq with respect to time one has

The above relation signifies that state variable x6 is a differential func­ 3L ( 2 ) Cdc 2
y1 = x + x22 + x⇒
tion of the system’s flat outputs, or 4 1 2 3
3L Cdc
x6 = h6 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (85) ẏ1 = (x1 ẋ1 + x2 ẋ2 ) + x3 ẋ3 ⇒
4 2
( )
Returning to Eq. (79) and by solving with respect to x10 one obtains 3L R vd K1
ẏ1 = x1 − x1 + ωdq x2 + −
4 L L 2L
x3 x8 + (94)
ẋ5 + KJb x2 + nJp x3 x6 ( )
x10 = (86) 3L R v q K2
1
J
+ x2 − ωdq x1 − x3 + − x3 x9 +
4 L L 2L
( )
which signifies that state variable x10 is a differential function of the Cdc
+ x3 −
1
x3 +
3K1
x1 x8 +
3K2
x2 x9
system’s flat outputs or 2 Cdc Rc 4Cdc 4Cdc

x10 = h10 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (87) By differentiating once again in time one has:

From the eighth row of the state-space model of the shipboard AC/DC ÿ1 = −
3R 3 3
2x1 ẋ1 + Lωdq (ẋ1 x2 + x1 x˙2 ) + vd ẋ1 −
microgrid, given in Eq. (20) one solves with respect to control input u1 . 4 4 4
This gives 3k1
− (ẋ1 x3 x8 + x1 ẋ3 x8 + x1 x3 ẋ8 )−
[ ] 8
β ( )
u1 = σ Lr ẋ8 + βx5 x7 − x6 − ωdq − x5 x9 + γ 2 x8 + βvd (88) 3 3R 3
τs − Lωdq (ẋ2 x1 + x2 ẋ1 ) − 2x2 ẋ2 + vd ẋ2 −
4 4 4
(95)
which signifies that u1 is a differential function of the system’s flat −
3K2
(ẋ2 x3 x9 + x2 ẋ3 x9 + x2 x3 ẋ9 − )
outputs, or 8
1 3K1
u1 = hu1 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (89) −
2Rc
2x3 ẋ3 +
8
(ẋ3 x1 x8 + x3 ẋ1 x8 + x3 x1 ẋ8 )+

3K2
In a similar manner, from the ninth row of the state-space model of the + (ẋ3 x2 x3 + x3 ẋ2 x9 + x3 x2 ẋ9 )
shipboard AC/DC microgrid, given in Eq. (21) one solves with respect to 8
the control input u2 . This gives By substituting the derivative variables in the previous relation one
[ ] obtains
β ( )
u2 = σ Lr ẋ9 − x7 − βx5 x6 + γ2 x9 + ωdq − x5 x8 + βvq (90)
τs 3R 3 3
ÿ1 = − 2x1 f1 + Lωdq (f1 x2 + x1 f2 ) + vd f1 −
2 4 4
which signifies that u2 is a differential function of the system’s flat
3k1
outputs, or − (f1 x3 x8 + x1 f3 x8 + x1 x3 [f8 + g8 u1 ])−
8
u2 = hu2 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (91) 3 3R 3
− Lωdq (f2 x1 + x2 f1 ) − 2x2 f2 + vd f2 −
4 4 4
Finally, from the tenth row of the state-space model of the shipboard (96)
3K2
AC/DC microgrid, given in Eq. (22), one solves with respect to the −
8
(f2 x3 x9 + x2 f3 x9 + x2 x3 [f9 + g9 u2 ] − )
control input u3 . Thus 1 3K1
[ ] − 2x3 f3 + (f3 x1 x8 + x3 f1 x8 + x3 x1 [f8 + g8 u2 ])+
1 x10 2Rc 8
u3 = ẋ10 + (92)
Ken τs 3K2
+ (f3 x2 x3 + x3 f2 x9 + x3 x2 [f9 + g9 u2 ])
8
which signifies that u3 is a differential function of the system’s flat
outputs, or By re-grouping terms one obtains the following input-output linearized
form
u3 = hu3 (y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 ) (93)
ÿ1 = ̃f 1 + ̃
g11 u1 + ̃
g12 u2 + ̃
g13 u3 (97)
As a result of the previous analysis, all state variables and control inputs
of the shipboard AC/DC microgrid can be written as differential func­ Next, one differentiates in time the flat output y2 = x2 . This gives
tions of this system’s flat outputs and the entire system is differentially
flat. The proof of differential flatness is also an implicit proof of the
system’s controllability. One can use the differential flatness property to

9
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

R x2 ⎛ ⎞
ÿ2 = − ωdq ẋ1 − ẋ2 − (ẋ3 x9 + x3 ẋ9 )⇒ z1
L 2L
(98) ⎛ ⎞

⎜ z2 ⎟
⎞⎜ ⎟
R x2 zm 1 0 0 0 0 0 0 ⎜ ⎟
ÿ2 = − ωdq ẋ1 − ẋ2 − f2 x9 + x3 [f9 + g9 u2 ] ⎜ 1m ⎟ ⎜ z3 ⎟
L 2L ⎜ z3 ⎟ = ⎝ 0
⎝ ⎠ 0 1 0 0 0 0 ⎠⎜ ⎟
⎜ z4 ⎟ (109)
zm 0 0 0 0 1 0 0 ⎜ ⎟
⎜ z5 ⎟
By re-grouping terms one obtains the following input-output linearized 5
⎝ z6 ⎠
form z7
ÿ2 = ̃f 2 + ̃
g21 u1 + ̃
g22 u2 + ̃
g23 u3 (99)
The stabilizing feedback control is computed as follows
Moreover, by differentiating the flat output y3 = x4 with respect to time ( ) ( )
v1 = z̈1,d − K11 ż1 − ż1,d − K12 z1 − z1,d
one obtains ( ) ( )
v2 = z̈3,d − K21 ż3 − ż3,d − K22 z3 − z3,d
( ) (110)
ẏ3 = ẋ4 ⇒ÿ3 = ẋ5 (100) ( ) ( )
v3 = z(3) ¨ − K32 ż5 − ż5,d − K33 z5 − z5,d
5,d − k3,1 z̈5 − z5,d

By substituting ẋ5 the previous condition becomes


This type of feedback control results into the following tracking error
1 Kb np
ÿ3 = x10 − x2 − (x9 x6 − x8 x7 ) (101) dynamics
J J J
( )
( ) ( )
and with the use of the field orientation condition x7 = psisd = 0 and by z̈1 − z̈1,d + K11 z˙1 − ż1,d + K12 z1 − z1,d = 0
differentiating once again in time one gets ( )
( ) ( )
z̈3 − z̈3,d + K21 ż3 − ż3,d + K22 z3 − z3,d = 0 (111)
1 Kb np
y(3)
3 = ẋ10 − ẋ2 − (ẋ9 x6 + x9 ẋ6 ) (102) ( ) ( )
J J J ( ) ( )
z(3) (3)
5 − z5,d + k31 z̈5 − z5,d¨ + K32 ż3 − ż5,d + K33 z5 − z5,d = 0
Next, by substituting in the previous relation the time-derivative terms
one gets or equivalently
[ ]
1 x10 Kb np
(103) ë1 + K11 ė1 + K12 e1 = 0
(3)
y3 = − + Ken u3 − f2 − ([f9 + g9 u2 ]x6 + x9 f6 )
J τs J J
ë2 + K21 ė2 + K22 e2 = 0 (112)
Next, by re-grouping terms one gets the input-output linearized form e(3)
3 + k31 ë3 + K32 ė3 + K33 e3 = 0

(3)
y3 = ̃f 3 + ̃
g31 u1 + ̃
g32 u2 + ̃
g33 u3 (104) The associated characteristic polynomials are

p1 (s) = s2 + K11 s + K12


Consequently, the dynamics of the shipboard AC/DC microgrid is writ­
p2 (s) = s2 + K21 s + K22 (113)
ten in the following input-output linearized form
p3 (s) = s3 + k31 s2 + K32 s + K33
⎛ ⎞
⎜ ÿ1 ⎟
⎛ ⎞ ⎛
̃f 1 g11 ̃
̃ g12 ̃
g13
⎞⎛ ⎞
u1 The above noted feedback gains are chosen according to the require­
⎜ ⎟
⎜ ÿ2 ⎟ = ⎝ ̃f ⎠ + ⎝ ̃ g21 ̃ g23 ⎠⎝ u2 ⎠
g22 ̃ (105) ment the previously given characteristic polynomials to be Hurwitz
⎜ ⎟ 2
⎝ (3) ⎠
y ̃f 3 g31 ̃
̃ g32 ̃
g33 u3 stable, that is to have poles in the left complex semi-plane.
State estimation for the model of the shipboard AC/DC microgrid
3

which has been written in the canonical Brunovsky form is performed


The concise description for Eq. (105) in the for Y F + Gu
̃=̃ ̃ provides also using the Derivative-free nonlinear Kalman Filter. This consists of the
the control inputs that should be applied to the initial nonlinear dy­ Kalman Filter recursion applied to the canonica Brunovsky form and of
̃ − 1 [Y inverse transformations based on differential flatness theory, which
namics of the system, that is u = G F].
̃− ̃
provide estimates for the state variables of the initial nonlinear model of
Next, by defining the virtual control inputs vi = ̃f i + ̃
gi1 u1 + ̃
gi2 u2 + the shipboard AC/DC microgrid [13],[1]. The measurable outputs are
gi3 u3 where i = 1, 2, 3, the input-output linearized model of the ship­
̃ the flat outputs of the system y1 , y2 and y3 .
board AC/DC microgrid is written in the form
Remark 1. It is possible to analyse an extended state-space model for
(106) the shipboard microgrid in which (a) an elaborated model of the diesel
(3)
ÿ1 = v1 ẏ2 = v2 y3 = v3
engine-based prime mover will be considered (b) the output of the DC
or equivalently, by defining the new state vector converter wil be used to feed a DC/AC inverter which in turn will pro­
[ ]T vide control inputs to an induction motor-based electric propulsion
Z = [z1 , z2 , z3 , z4 , z5 , z6 , z7 ]T ⇒Z = y1 , ẏ1 , y2 , ẏ2 , y3 , ẏ3 , ÿ3 (107) system of the vessel. The proposed nonlinear optimal control method is
scalable and its application is not constrained by the system’s dimen­
sionality. It is possible (i) to consider a centralized control scheme in
one has the canonical Brunovsky form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞ which the control inputs for the individual components of the shipboard
ż1 0 1 0 0 0 0 0 0 z1 0 0 0 microgrid will be computed from the solution of one single Riccati
⎜ ż2 ⎟ ⎜ 0 0 0 0 0 0 0 0 ⎟⎜ z2 ⎟ ⎜ 1
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ 0 0⎟⎟⎛ ⎞ equation of high dimensionality, (ii) to consider a decentralized control
⎜ ż3 ⎟ ⎜ 0 0 0 1 0 0 0 0 ⎟⎜ z3 ⎟ ⎜ 0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ 0 ⎟ v1 scheme in which some components of the shipboard microgrid will be
⎜ ż4 ⎟ = ⎜ 0 0 0 0 0 0 0 0 ⎟⎜ z4 ⎟ + ⎜ 0 0⎟ ⎝ ⎠
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ 1 ⎟ v2 controlled independently with the proposed nonlinear optimal control
⎜ ż5 ⎟ ⎜ 0 0 0 0 1 0 0 0 ⎟⎜ z5 ⎟ ⎜ 0 0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v3 method. In the latter case the distributed processing of matrices of
⎝ ż6 ⎠ ⎝ 0 0 0 0 0 1 0 0 ⎠⎝ z6 ⎠ ⎝ 0 0 0⎠
smaller dimensionality and the parallel solution of the associated Riccati
ż7 0 0 0 0 0 0 0 0 z7 0 0 1
equations will take place. In both cases (i) and (ii) global stability cre­
(108)
dentials are ensured for the shipboard microgrid. In particular it is noted
with measurement equation that there exist results about (a) nonlinear optimal control of the com­
plete (high-dimensional) model of the ship’s diesel engine and (ii)

10
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

nonlinear optimal control of an electric propulsion system of the vessel in Matlab. The sampling period was set to Ts = 0.01 sec. To implement
that consists of an induction motor being fed by a three-phase DC/AC the nonlinear optimal control scheme, the algebraic Riccati equation
inverter. These results confirm the reliability of the proposed nonlinear which appears in Eq. (50) had to be solved at each time-step of the
optimal control method [13]. control algorithm with the use of Matlab’s aresolv() function. The ob­
tained results are depicted in Fig. 3 to Fig. 18. It can be noticed that in all
Remark 2. There is a clear contribution of this research work
cases fast and accurate tracking of reference setpoints was achieved
comparing to past approaches for solving the nonlinear optimal (H-in­
under moderate variations of the control inputs. The real values of the
finity) control problem. One can point out the advantages of the
state vector of the shipboard AC/DC microgrid are printed in blue, the
nonlinear optimal control method against Nonlinear Model Predictive
estimated state variables provided by the H-infinity Kalman Filter are
Control (NMPC). In NMPC the stability properties of the control scheme
plotted in green, while the associated setpoints are shown in red colour.
remain unproven and the convergence of the iterative search for an
It can be noted that the transient performance of the control method
optimum often depends on initialization and parameter values’ selec­
depends of the selection of parameters r, ρ and Q which appear in the
tion. It is also noteworthy that the nonlinear optimal control method is
Riccati equation of Eq. (50). Actually, relatively small values of r result
applicable to a wider class of dynamical systems than approaches based
in elimination of the tracking error while relatively large values of
on the solution of State Dependent Riccati Equations (SDRE). The SDRE
matrix Q result in fast convergence to the reference setpoints. Moreover,
approaches can be applied only to dynamical systems which can be
coefficient ρ affects the robustness of the control loop. The smallest
transformed to the Linear Parameter Varying (LPV) form. Besides, the
value of ρ for which one can obtain a valid solution from the algebraic
nonlinear optimal control method performs better than nonlinear
Riccati equation of Eq. (50) is the one that provides the control loop
optimal control schemes which use approximation of the solution of the
with maximum robustness. It is also pointed out that the use of the H-
Hamilton-Jacobi-Bellman equation by Galerkin series expansions. The
infinity Kalman Filter as a robust state estimator has allowed to imple­
stability properties of the Galerkin series expansion -based optimal
ment feedback control by using measurement of a subset of the state
control approaches are still unproven.
vector elements of the shipboard AC/DC microgrid. To elaborate on the
tracking performance and on the robustness of the proposed nonlinear
6. Simulation tests
optimal control method for shipboard microgrids the following
Tables are given: (i) Table 1 which provides information about the ac­
6.1. Results on nonlinear optimal control
curacy of tracking of the reference setpoints by the state variables of the
ship microgrid’s state-space model in the disturbance-free case, (ii)
The global stability properties of the proposed nonlinear optimal
Table 2 which provides information about the robustness of the control
control scheme for the shipboard AC/DC microgrid, as well as the pre­
method to parametric changes in the model of the ship microgrid
cise tracking of reference setpoints by the state variables of this power
(change in parameter Kb in the generator’s rotor), (iii) Table 3 which
system has been further confirmed through simulation experiments.
provides information about the precision in state variables’ estimation
Indicative values for the parameters of the model of the shipboard
that is achieved by the H-infinity Kalman Filter, (iv) Table 4 which
microgrid are: (i) for the AC/DC converter L = 0.25H, R = 0.15Ohm, Rc
provides the convergence times of the ship microgrid’s state variables to
= 0.52Ohm, Cdc = 0.21F, ωdq = 1.0, vd = 1.0, vq = 0.0. (ii) for the
the associated setpoints.
Doubly Fed Induction Generator J = 20 kgr⋅m2 , kb = 0.1, np = 12, M =
0.4H, Lr = 0.3H, Ls = 0.3H, Rs = 0.15Ohm, Rr = 0.15Ohm (iii) for the
diesel engine τs = 4.65sec, Ken = 1.5. The simulation code was written

Fig. 3. Tracking of setpoint 1 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference setpoints
(red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their refer­
ence setpoints

11
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 4. Tracking of setpoint 1 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Fig. 5. Tracking of setpoint 2 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference setpoints
(red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their refer­
ence setpoints

6.2. Results on flatness-based control (green lines). It can be noticed that under flatness-based control fast and
accurate tracking of setpoints is also achieved. As it can be concluded
The nonlinear control problem for the dynamic model of the ship­ from Eq. (112) the rate of elimination of the tracking error and the
board microgrid was also treated with the previously described flatness- transients of the state variables depend on the selection of the control­
based control method. The obtained results are given in Fig. 19 to ler’s feedback gains. It is noted that, to compute the flatness-based
Fig. 22. The convergence of the flat outputs of the robotic system, that is control inputs that should be applied to the initial nonlinear model of
of y1 , y2 , y3 as well as of the output of the AC/DC converter x3 = Vdc to the robot one has to solve Eq. (105) for the control inputs vector u =
the targeted setpoints is shown. The real values of the above noted flat [u1 , u2 , u3 ]T . To this end matrix inversions are needed. Thus, this pro­
outputs are depicted in blue while the targeted setpoints are plotted in cedure may result in more abrupt variations of the control inputs, while
red. State estimation was performed with the use of Kalman Filtering the appearance of singularities cannot be excluded.

12
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 6. Tracking of setpoint 2 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Fig. 7. Tracking of setpoint 3 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference setpoints
(red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their refer­
ence setpoints

7. Conclusions control) method has been applied to a vessel’s AC/DC microgrid model
consisting of Diesel engines, induction generators and AC/DC con­
To implement the all-electric-ship concept the solution of the verters. First, the dynamic model of the shipboard AC/DC microgrid has
nonlinear control problem for shipboard AC/DC microgrids has become undergone approximate linearization with the use of first-order Taylor
a necessity. In this article a novel nonlinear optimal control (H-infinity series expansion and through the computation of the associated

13
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 8. Tracking of setpoint 3 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Fig. 9. Tracking of setpoint 4 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference setpoints
(red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their refer­
ence setpoints

14
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 10. Tracking of setpoint 4 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Fig. 11. Tracking of setpoint 5 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference
setpoints (red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their
reference setpoints

15
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 12. Tracking of setpoint 5 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Fig. 13. Tracking of setpoint 6 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference
setpoints (red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their
reference setpoints

16
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 14. Tracking of setpoint 6 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Fig. 15. Tracking of setpoint 7 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference
setpoints (red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their
reference setpoints

17
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 16. Tracking of setpoint 7 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Fig. 17. Tracking of setpoint 8 for the shipboard AC/DC microgrid (a) convergence of state variables x1 = id , x2 = iq , x3 = Vdc and x5 = ωg to their reference
setpoints (red line: setpoint, blue line: real value, green line: estimated value), (b) convergence of state variables x6 = ψ sd , x8 = ird , x9 = irq and x10 = Tm to their
reference setpoints

18
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Fig. 18. Tracking of setpoint 8 for the shipboard AC/DC microgrid (a) control inputs u1 to u3 applied to shipboard AC/DC microgrid, (b) tracking error e1 , e2 , e3 , for
state variables Vdc , ωg , Tm of the shipboard AC/DC microgrid

Table 1
Tracking RMSE for the shipboard microgrid in the disturbance-free case
RMSEx1 RMSEx2 RMSEx3 RMSEx5 RMSEx6 RMSEx8 RMSEx9 RMSEx10

setpoint1 0.0002 0.0004 0.0001 0.0024 0.0001 0.0003 0.0020 0.0023


setpoint2 0.0005 0.0012 0.0002 0.0005 0.0001 0.0002 0.0005 0.0006
setpoint3 0.0004 0.0001 0.0002 0.0020 0.0001 0.0006 0.0002 0.0024
setpoint4 0.0009 0.0018 0.0005 0.0012 0.0006 0.0003 0.0007 0.0006
setpoint5 0.0018 0.0040 0.0009 0.0014 0.0003 0.0005 0.0007 0.0022
setpoint6 0.0021 0.0044 0.0009 0.0003 0.0013 0.0011 0.0009 0.0007
setpoint7 0.0016 0.0033 0.0007 0.0018 0.0014 0.0008 0.0010 0.0007
setpoint8 0.0023 0.0045 0.0008 0.0028 0.0022 0.0020 0.0008 0.0006

Table 2
Tracking RMSE for the shipboard microgrid in the case of disturbances
Δa% RMSEx1 RMSEx2 RMSEx3 RMSEx5 RMSEx6 RMSEx8 RMSEx9 RMSEx10

0% 0.0005 0.0012 0.0002 0.0005 0.0001 0.0002 0.0005 0.0006


10% 0.0006 0.0014 0.0003 0.0006 0.0001 0.0003 0.0006 0.0006
20% 0.0007 0.0016 0.0003 0.0008 0.0001 0.0003 0.0007 0.0006
30% 0.0008 0.0018 0.0003 0.0009 0.0001 0.0004 0.0007 0.0006
40% 0.0009 0.0020 0.0004 0.0011 0.0001 0.0004 0.0008 0.0006
50% 0.0009 0.0022 0.0004 0.0012 0.0001 0.0005 0.0009 0.0006
60% 0.0010 0.0024 0.0004 0.0014 0.0001 0.0005 0.0010 0.0006

Table 3
4
RMSE of estimation for the shipboard microgrid provided by the H-infinity KF × 10−
RMSEx1 RMSEx2 RMSEx3 RMSEx5 RMSEx6 RMSEx8 RMSEx9 RMSEx10

setpoint1 0.0042 0.0050 0.0048 0.1058 0.5414 0.0153 0.2504 0.0036


setpoint2 0.0049 0.0047 0.0050 0.1643 0.8424 0.0426 0.3150 0.0049
setpoint3 0.0035 0.0042 0.0033 0.0755 0.3890 0.0182 0.1431 0.0036
setpoint4 0.0045 0.0059 0.0045 0.1939 0.9340 0.0404 0.4066 0.0036
setpoint5 0.0009 0.0017 0.0008 0.0839 0.8761 0.0010 0.1934 0.0010
setpoint6 0.0031 0.0030 0.0025 0.0811 0.7959 0.0042 0.1567 0.0031
setpoint7 0.0036 0.0017 0.0038 0.0398 0.7025 0.0096 0.0750 0.0042
setpoint8 0.0027 0.0034 0.0031 0.0900 0.8765 0.0198 0.1225 0.0029

19
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Table 4
Convergence time (sec) for the shipboard microgrid’s state variables
Ts x1 Ts x 2 Ts x3 Ts x5 Ts x 6 Ts x8 Ts x9 Ts x10

setpoint1 8 10 5 8 1 1 5 0.2
setpoint2 9 10 8 8 1 1 4 0.2
setpoint3 9 10 5 8 1 3 6 0.2
setpoint4 10 10 4 8 2 2 5 0.2
setpoint5 9 10 8 9 3 2 2 0.2
setpoint6 10 10 5 8 3 3 3 0.2
setpoint7 9 10 5 7 2 2 3 0.2
setpoint8 5 10 5 7 1 1 1 0.2

Cdc 2
Fig. 19. Flatness-based control of the shipboard microgtid: Convergence of flat outputs z1 = 3L
4 (id + iq ) + 2 Vdc , z3 = iq , z6 = ωg and of converter’s output Vdc (blue
2 2

lines) to targeted setpoints (red lines) and KF-based state estimation (green lines) (a) when tracking setpoint 1 (b) when tracking setpoint 2

Jacobian matrices. The linearization point which was updated at each proposed control method enables fast and accurate tracking of reference
sampling instant was defined by the present value of the system’s state setpoints under moderate variations of the control inputs. By solving the
vector and by the last sampled value of the control inputs vector. For the optimal control problem for the shipboard AC/DC microgrid one ach­
approximately linearized model of the AC/DC microgrid a stabilizing H- ieves minimization of energy dispersion by the ship’s power system and
infinity feedback controller has been designed. consequently the ship’s autonomy and operational capacity is improved.
The H-infinity control represents a min-max differential game in By proving differential flatness properties for the shipboard micro­
which the controller tries to minimize a quadratic cost function of the grid, the solution of the associated control and state estimation problems
state vector’s tracking error, while the model uncertainty and external became also possible with flatness-based control and flatness-based state
perturbation terms try to minimize this cost function. To select the estimation methods. A comparison between the proposed nonlinear
feedback gains of the H-infinity controller which stabilize the control optimal control method and flatness-based control for shipboard
loop an algebraic Riccati equation had to be repetitively solved at each microgrids has been also performed. Future research work can be related
time-step of the control algorithm. The global stability properties of the with sliding mode control for the shipboard microgrid dynamics. As
control scheme have been proven through Lyapunov analysis. To clearly explained in the manuscript the dynamic model of the shipboard
implement state estimation-based feedback control without the need to microgrid is differentially flat. This means that through successive dif­
measure the entire state vector of the shipboard AC/DC microgrid the H- ferentiation of its flat outputs it can be transformed into the input-output
infinity Kalman Filter has been used as a robust state estimator. As linearized form. For the latter state-space description there is a sys­
previously explained, the nonlinear optimal control method exhibits tematic procedure to define sliding surfaces and based on them to design
advantages comparing to other nonlinear optimal control schemes one a sliding mode controller with global stability properties.
can consider for the dynamics of shipboard microgrids. Actually, the

20
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Cdc 2
Fig. 20. Flatness-based control of the shipboard microgtid: Convergence of flat outputs z1 = 3L 2 2
4 (id + iq ) + 2 Vdc , z3 = iq , z6 = ωg and of converter’s output Vdc (blue
lines) to targeted setpoints (red lines) and KF-based state estimation (green lines) (a) when tracking setpoint 3 (b) when tracking setpoint 4

Cdc 2
Fig. 21. Flatness-based control of the shipboard microgtid: Convergence of flat outputs z1 = 3L 2 2
4 (id + iq ) + 2 Vdc , z3 = iq , z6 = ωg and of converter’s output Vdc (blue
lines) to targeted setpoints (red lines) and KF-based state estimation (green lines) (a) when tracking setpoint 5 (b) when tracking setpoint 6

21
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

Cdc 2
Fig. 22. Flatness-based control of the shipboard microgtid: Convergence of flat outputs z1 = 3L 2 2
4 (id + iq ) + 2 Vdc , z3 = iq , z6 = ωg and of converter’s output Vdc (blue
lines) to targeted setpoints (red lines) and KF-based state estimation (green lines) (a) when tracking setpoint 7 (b) when tracking setpoint 8

Declaration of Competing Interest degradtion model, IEEE Transactions on Transortation Electrification 5 (4) (2019)
1306–1318.
[10] Z. Jin, L. Meng, J.M. Guerrero, R. Han, Hierarchical and control design for a
The authors of the article “A nonlinear optimal control approach for shipboard power system with DC distribution and energy stoarge aboard future
shipboard AC/DC microgrids” declare that to their knowledge no con­ more-electric ships, IEEE Transactions on Industrial Informatics 14 (2) (2018)
flict of interest exists with third parties about the content, results and 709–720.
[11] T. Van Vu, D. Gonsoulin, F. Diaz, C.S. Edrington, T. El-Mezyani, Predictive control
methods of the above noted manuscript. for energy management in ship power systems under high-power ramp-rate loads,
IEEE Transactions on Energy Conversion 32 (2) (2017) 788–799.
Acknowledgment [12] Z. Hou, Z. Song, H.F. Hofmann, J. Sun, Adaptive model-predictive control for
hybrid energy storage energy management in all-electric ship microgrids, Energy
Conversion and Management 198 (2019) 111929–111939.
The research work entitled ”A nonlinear optimal control approach [13] G. Rigatos, Intelligent Renewable Energy Systems: Modelling and Control,
for shipboard AC/DC microgrids” has been partially supported by Grant Springer, 2016.
[14] A. Maqsood, K. Corzine, Integration of z-source breakers into zonal DC ship power
Ref. ”CSP contract 040322” - ”Nonlinear control, estimation and fault system microgrids, IEEE Journal of Emergung and Selected Topics in Power
diagnosis for electric power generation and electric traction / propulsion Electronics 5 (1) (2017) 269–277.
systems” of the Unit of Industrial Automation of the Industrial Systems [15] K. Satpathi, A. Ukil, S.S. Nag, J. Poou, M.A. Zagradnik, DC marine power system:
transient behavior and fault management aspects, IEEE Transactions on Industrial
Institute. Informatics 15 (4) (2019) 1911–1924.
[16] S. Yousefizadeh, J.D. Bedtsen, N. Vafamand, M.H. Khooban, T. Dragisevic,
References F. Blaabjerg, EKF-based predictive stabilization of shipboard DC microgrids with
uncertain time-varying load, IEEE Journal of Emerging and Selected Topics in
Power Electronics (2021).
[1] G. Rigatos, Nonlinear control and filtering using differential flatness approaches:
[17] X. Zhoxin, Z. Tiamli, L. Huaimin, J.M. Guerrero, C.L. Su, J.C. Vasquez, Coordinated
Applications to electrommechanical systems, Springer, 2015.
control of a hyvrid electric ferry shipboard microgrid, IEEE Tranactions on
[2] M.M. Mandani, M.H. Khooban, A. Masoudian, T. Dragisevic, Model predictive
Transportation Electrification 5 (3) (2019) 828–839.
control of DC-DC converter to mitigate the effects of pulsed power loads in naval
[18] F. Bu, H. Liu, W. Huang, Y. Hu, M. Degano, C. Gerada, K. Rajashekara, Induction
DC microgrids, IEEE Transactions on Industrial Electronics 66 (7) (2017)
machine-based starter/generator systems, IEEE Industrial Electronics Magazine
5678–5685.
(2020) 4–19.March
[3] M.H. Khooban, M.H. Ghisornejad, H. Farizadeh, A. Masoudian, J. Dragisevic,
[19] A. Haseltalab, R.R. Negenborn, Model predictive maneuvering control and energy
A new nonlinear hybrid control approach for DC-DC converters in zero-emission
management for all-electric autonomous ships, Applied Energy 251 (2014)
ferry ships, IEEE Transactions on Power Electronics 35 (6) (2020) 5832–5841.
113308–113334.Elsevier
[4] A. Accetta, M. Pucci, Energy management system in DC microgrid of smart ships:
[20] A. Haseltalab, M.A. Botto, R.R. Negenborn, Model predictive DC voltage control for
Main gen-set fule consumption minimization and fault compensation, IEEE
all-electric ships, Control Engineering Practice 90 (2019) 133–147.Elsevier
Transactions on industry Applications 55 (5) (2019) 2097–3113.
[21] A. Huseltalab, M.A. Botto, R.R. Negenborn, On-board voltage regulation for all-
[5] Z. Jin, G. Salligui, R. Cuzner, L. Meng, J.C. Vasquez, J.M. Guerrero, Next
electric DC-ships. Proc. IFAC Conference on Control Applications in Marine
generation shopboard DC power systems, IEEE Electrification Magazine (2016)
Systems, Robotics and Vehicles, IFAC CAMS 2018, Opatija, Croatia, Spe., 2018.
45–57.
[22] Y. Liu, W. Xu, J. Zhu, F. Blaabjerg, Sensorless control of standalone brushless
[6] Z. Li, Y. Xu, S. Feng, X. Zhang, X. Feng, Robust coordination of a hybrid AC/DC
doubly-fed induction generator feeding unbalanced loads in a ship shaft power
multi-energy ship microgrid with flexible voyage and thermal loads, IEEE
generation system, IEEE Transactions on Industrial Electronics 66 (1) (2019)
Transactions on Smart Grids 11 (4) (2020) 2782–2793.
739–749.
[7] Z.X. Xiao, M.M. Li, H.W. Fang, Y.Z. Guan, T. Liu, L. Hou, J.M. Guerrero, Operation
[23] J. Li, F. Liu, Y. Chen, C. Shao, G. Wang, Y. Hou, S. Mei, Resielience control of
control for improing energy efficiency of shipboard microgrid including bow
shipboard DC power systems, IEEE Transacions on Power Systems 37 (8) (2018)
thruusters and hybrid energy storages, IEEE Tranactions on Transportation
6675–6685.
Electrification 6 (2) (2020) 856–868.
[24] J. Hou, Z. Sang, H.F. Hofmann, J. Sun, Control strategy for battery-flywheel hybrid
[8] A. Frances-Roger, A. Anvan-Moghadelam, E. Rodriguez-Diaz, J.C. Vasquez, J.
energy storage in electric shipboard microgrids, IEEE Transactions on Industrial
M. Guerrero, J. Uceda, Dynamic assessment of COTS converters-based DC
Informatics 17 (2) (2021) 1088–1100.
mmicrogrid power systems in electric ships, IEEE Transactions in Industrial
[25] L. Xu, J.M. Guerrero, A. Lashab, B. Wei, N. Bazmahammadi, J. Vasquez, M.
Informatics 14 (12) (2018) 5518–5529.
A. Abusarrah, A review of DC shipboard microgrids part i: Power architectures
[9] S. Fang, B. Gou, Y. Wang, Y. Xu, C. Shang, H. Wang, Optimal hierarchical
energy storage and power converters, IEEE Transactions on Power Electronics 37
management of shipboard multi-battery energy storage system using a data-driven
(5) (2022) 5155–5172.

22
G. Rigatos et al. Electric Power Systems Research 209 (2022) 108024

[26] F. Hardan, R. Norman, P. Tricoli, Control and operation of a ship AC/DC microgrid [32] M. Basseville, I. Nikiforov, Detection of abrupt changes: Theory and Applications,
transient propulsion and maneuvering load conditions, Intl. Journal of Electric Prentice-Hall, 1993.
Power and Energy Systems 139 (2022) 107823–107934.Elsevier [33] G. Rigatos, Q. Zhang, Fuzzy model validation using the local statistical approach,
[27] D. Celik, H.E. Meral, Multi-objective control scheme for operation of parallel Fuzzy Sets and Systems 60 (7) (2009) 882–904.Elsevier
inverter-based microgrids during asymmetrical grid faults, IET Renewable Power [34] G.G. Rigatos, Modelling and control for intelligent industrial systems: adaptive
Generation 14 (13) (2020) 2487–2498. algorithms in robotcs and industrial engineering, Springer, 2011.
[28] D. Celik, Lyapunov-based harmonic compensation and charging three-phase shunt [35] G. Rigatos, K. Busawon, Robotic Manipulators and Vehicles: Control, estimation
active power filter in eletrical vehicle applications, Intl Journal of Electric Power and filtering, Springer, 2018.
and Energy Systems 136 (2022) 107564–107582.Elevier [36] G. Rigatos, E. Karapanou, Advances in applied nonlinear optimal control,
[29] M.U. Mattaraf, Y. Guan, Y. Tertiche, C.L. Su, M. Nasr, J.C. Vasquez, J.P. Guerrero, Cambridge Scholars Publications, 2020.
Adaptive power management of hierarchical controlled hybrid shipboard [37] G.J. Toussaint, T. Basar, F. Bullo, h∞ optimal tracking control techniques for
microgrid, IEEE Access 4 (2022) 1–15. nonlinear underactuated systems. Proc. IEEE CDC 2000, 39th IEEE Conference on
[30] G. Rigatos, P. Siano, C. Cecati, A new nonlinear h-infinity feedback control Decision and Control, Sydney Australia, 2000.Dec.
approach for three-phase voltage source converters, Electric Power Components [38] B.P. Gibbs, Advanced Kalman Filtering, Least Squares and Modelling: A practical
and Systems (2015).Taylor and Francis handbook, J. Wiley, 2011.
[31] G.G. Rigatos, S.G. Tzafestas, Extended kalman filtering for fuzzy modelling and
multi-sensor fusion, Mathematical and Computer Modelling of Dynamical Systems
13 (2007) 251–266.Taylor & Francis

23

You might also like