You are on page 1of 11

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 60, NO.

1, JANUARY/FEBRUARY 2024 1705

Hybrid Sensor Fault Tolerant Control of


Low Voltage DC Microgrid
M. V. Satya Sai Chandra , Graduate Student Member, IEEE, and Sankarsan Mohapatro , Senior Member, IEEE

Abstract—Low voltage DC (LVDC) microgrids have recently the subject of extensive research and development [4]. Nonlinear
gained popularity. In applications such as rural electrification, loads and sudden, substantial disruptions of the system are
data center power systems, and satellite microgrids stand-alone difficult for these linear control systems to handle [5]. As a
LVDC microgrids have proven to be extremely beneficial. In the
case of standalone LVDC microgrids, voltage control and energy result, many studies have developed sophisticated and nonlinear
management are crucial. In such standalone and safety-critical methods of control.
systems, sensor faults are especially important to address. There- Nonlinear control strategies such as sliding mode control
fore, we present a hybrid sensor fault tolerant control (HSFTC) (SMC) are useful for addressing disturbances and nonlinear
approach with redundancy features at the software and hardware loading in complex systems. Initially, researchers developed
levels. To address voltage sensor failures, an analytical redundancy
component employs a sliding mode observer in conjunction with SMC systems based on equivalent control that use control
adaptive passivity-based control. To overcome battery current sen- signals with varying frequencies [6]. Pulse width modulation
sor failures, a double modular hardware redundancy component is (PWM)-based SMC control techniques were created so that they
included. Both MATLAB simulations and hardware experiments could be easily implemented in hardware [7], [8]. An adaptive
on the microgrid testbed are used to verify the performance of SMC approach for dealing with nonlinear loading conditions is
this HSFTC in a variety of scenarios involving sensor malfunctions
and load disturbances. We observe that the suggested HSFTC can also presented in [9]. Few other works have been reported on
address sensor faults and load disturbances simultaneously. The advanced controllers such as passivity-based control [10] and
effectiveness of our control approach in dealing with sensor mal- Lyapunov-based control [11] for microgrid applications. The
functions is evaluated in comparison to reference control strategies. goal of these control strategies is to address the intermittent
Index Terms—Fault-tolerant control, hardware redundancy, nature of the sources, as well as disturbances in the system,
LVDC microgrid, passivity control, sensor fault. nonlinear loading, and dynamic effects.
Sensors play a crucial role in the operation of the systems,
as they offer the necessary measurements for control and mon-
I. INTRODUCTION itoring. However, sensor faults can occur, resulting in subpar
C MICROGRIDS that operate independently have found performance or even system failure if they are not dealt with
D widespread use in places such as electric aircraft, satellites,
and ships. Research on low voltage direct current (LVDC) micro-
properly. Sensor faults can reduce performance in the afore-
mentioned well-designed controllers. Inaccurate readings from
grids has also increased significantly in recent years, particularly malfunctioning sensors can cause a number of issues, includ-
for use in applications such as rural electricity. [1], [2]. Voltage ing a loss of reference tracking and the accidental triggering
regulation and energy management in isolated microgrids are of safety mechanisms. Therefore, fault-tolerant control (FTC)
difficult tasks because of the lack of grid support. Battery storage must be implemented to lessen the effects of sensor failures
interface control is essential for stable system voltage because and keep the system operational and safe [13]. For essential
of source and load variability. The linear control strategies of the systems such as electric aircraft and satellite microgrids, as
microgrid, such as voltage mode control and cascaded control well as other standalone microgrid systems, FTC is absolutely
[3], have been developed. Primary control mechanisms, such as essential.
droop control and effective power-sharing controls, have been Active FTC schemes are quite popular which include the
fault detection and isolation (FDI) unit and control reconfig-
Manuscript received 18 May 2023; revised 26 July 2023; accepted 5 Septem- uration. Analytical redundancy-based systems and hardware
ber 2023. Date of publication 14 September 2023; date of current version 18 redundancy-based methods are two subsets of this larger group
January 2024. Paper 2023-IACC-0821.R1, presented at the 2022 IEEE Global [14]. To achieve fault tolerance [15], analytical redundancy
Conference on Computing, Power and Communication Technologies, New
Delhi, India, Sep 23–25, and approved for publication in the IEEE TRANSAC- systems make use of state observers, while hardware redundancy
TIONS ON INDUSTRY APPLICATIONS by the Industrial Automation and Control methods make use of redundant hardware units. Active fault
Committee of the IEEE Industry Applications Society [DOI: 10.1109/Glob- tolerance control (FTC) techniques have been researched for
ConPT57482.2022.9938148]. (Corresponding author: Sankarsan Mohapatro.)
The authors are with the School of Electrical Sciences, Indian In- their potential use in mitigating sensor and actuator failures in
stitute of Technology Bhubaneswar, Bhubaneswar 751013, India (e-mail: complex systems [16], [17]. In the control systems literature,
msc11@iitbbs.ac.in; sankarsan@iitbbs.ac.in). active FTC has been investigated for a variety of uses, including
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TIA.2023.3315276. solar water pumping [18], power converter control, and motor
Digital Object Identifier 10.1109/TIA.2023.3315276 drives [19]. Across these applications, FTC can help detect

0093-9994 © 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
1706 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 60, NO. 1, JANUARY/FEBRUARY 2024

and compensate for faults, allowing continuous and efficient


operation even in the presence of faults.
For fault detection and control reconfiguration, state-
observer-based analytical redundancy FTC methods have
proven very popular [19]. The estimated states of a observer
act as a reference for the detection of sensor faults for the
corresponding state. Significant efforts have been made toward
developing sensor FTC schemes using popular observers such
as Kalman filters [20] and sliding mode observers (SMOs) [21].
For such FTC systems, SMO is a widely-used set of observers Fig. 1. Schematic of the equivalent LVDC Microgrid.
due to its reliability in delivering accurate estimations [22],
[23], [24]. SMO is used for developing a current sensor FTC TABLE I
in speed sensor-less induction motor drive applications [21]. To SYSTEM PARAMETERS
address voltage and current sensor problems in grid-connected
photovoltaic (PV) systems, FTC schemes based on SMO have
been developed [25]. For the purpose of regulating the voltage
and frequency of three-phase AC microgrids, a fault-tolerant
control scheme is designed and implemented based on optimal
SMO-based current and voltage sensor fault estimation [26]. To
maintain the grid voltage and frequency stability an SMO based
FTC scheme is developed to handle the PV voltage sensor faults
r The HSFTC is rigorously tested and assessed to verify how
and grid current sensor faults in a grid connected DC microgrid
[27]. Such observer-based approaches suffer from computa- well it handles various sensor faults and simultaneous load
tional cost and modeling complexity in the FTC schemes that fluctuations.
handle numerous sensor faults and the inclusion of augmented The remainder of this article is organized as follows: The
fault variables. Therefore, the FTC design becomes more com- system is described in Section II. Sensor faults and fault toler-
plicated and cumbersome at times. ance are discussed in Section III. The proposed fault-tolerant
Fault tolerance is achieved using redundant hardware mod- control technique is described in Section IV. The findings of
ules in hardware redundancy schemes. Hardware redundancy the simulations for the various case studies are discussed in
schemes generally use voter/decider logic to use the extra hard- Section V. The proposed control is experimentally validated
ware component in the case of fault detection [28], [29]. Unlike for various instances in Section VI. Conclusions are offered in
analytical redundancy schemes, they offer the advantage of ease Section VII.
of implementation. The adoption of supplementary hardware,
however, may not always be practical and cost-effective. There- II. SYSTEM DESCRIPTION
fore, a more efficient, feasible, and cost-effective approach in ap- A low voltage DC microgrid analog is considered, as shown
plications such as microgrids may be to combine analytical and in Fig. 1. The system consists of a load connected to a common
hardware-based redundancy techniques. Unfortunately, efforts DC bus, a bidirectional DC-DC converter (BDC), and a battery
toward the application of a combination of both these schemes storage system. The reference DC bus voltage is set at 48 V.
together have been sparse, especially with consideration of The most important step to achieving system voltage tracking
standalone DC microgrids. is controlling the switches of the BDC. For this task, boost
To address these aforementioned research gaps, a hybrid operation of BDC is considered.
sensor fault tolerant control (HSFTC) technique is presented The parameters of the system are shown in Table I.
here for efficiently dealing with voltage and current sensor
faults. This report is an extension of the work in [31] by the
A. Modelling the BDC Converter
same authors. In contrast to previous DC microgrid studies,
this study intends to make use of the benefits of both analytical BDC converters can be modeled using the average state space
and hardware redundancy techniques. Additionally, unlike many model as in (1)-(2).
prior studies, this study attempts to address sensor failures and diL
load fluctuations occurring at the same time. The contributions L = vbat − (1 − u)vdc (1)
dt
of this work are as follows:
r To address voltage and current sensor failures, a hybrid sen- C
dvdc
= (1 − u)iL − id (2)
sor fault tolerant control (HSFTC) was developed, which dt
incorporated an analytical redundancy system based on where inductor current (iL ) and DC bus voltage (vdc ) are the
a sliding mode observer and a double modular hardware state variables of the system, u is the control input to the
redundancy scheme. boost switch, battery voltage (Vbat ) is the input voltage, and
r A variety of sensor faults and changes in system load are id is the disturbance current, which is the output load current
used to test the HSFTC’s resilience. supplied. This current can be modeled as id = vRC , where R is

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
CHANDRA AND MOHAPATRO: HYBRID SENSOR FAULT TOLERANT CONTROL OF LOW VOLTAGE DC MICROGRID 1707

the equivalent load resistance as coupled to the battery storage


systems.
The BDC model is a nonlinear model with affinity between
state variables and control input. The presence of the external
disturbance in the form of id makes the system much more
difficult to control.

III. SENSOR FAULTS AND FAULT TOLERANT CONTROL


Fig. 2. Structure of Active Fault Tolerant Control Scheme.
Faults are considered to be unexpected events of deviation
from the regular operation of the system. The incidence of faults
hinders system performance and may have detrimental impacts
on how the system functions. Specifically, sensor and actuator B. Fault Tolerant Control
faults seem to be critical from the system control perspective. Fault Tolerant Control (FTC) has two popular categories:
Therefore, it is extremely desirable to develop control techniques Active Fault Tolerant Control (AFTC) and Passive Fault Tolerant
that can accomplish the control goals while tolerating sensor Control (PFTC). AFTC schemes are capable of handling a wide
faults. variety of faults compared to PFTC schemes. Further details of
the AFTC scheme are discussed as follows:
A. Sensor Fault Description 1) Active Fault Tolerant Control (AFTC): The AFTC
Sensor faults are generally categorized into multiplicative schemes are capable of handling many varieties of faults in the
faults and additive faults. They are mathematically modeled as system. The AFTC scheme has two functional units, namely,
in (3), [13], [30]: the fault detection and isolation (FDI) unit and control recon-
figuration. The structure of the AFTC scheme is shown in
z f = (βz )z + fz (3) Fig. 2. The implementation of AFTC schemes is performed
The actual sensor output is z, whereas the faulty sensor output is using two popular techniques, namely, analytical redundancy
z f . The multiplicative fault component or gain fault is modeled and hardware redundancy [14]. The details of these schemes are
as βz , where 0 ≤ βz ≤ 1. The additive fault component is discussed as follows:
modeled as fz . The sensor faults are categorized into different a) Analytical redundancy: Analytical redundancy schemes
types based on the characteristics of their occurrence. Some of use an extra analytical component that is used to achieve
them are discussed as follows: the FTC based on the mathematical model of the system.
1) Sustained Faults: Sustained faults correspond to the situ- Generally, state observer-based approaches are popularly
ation where the sensor is faulty right from the start of the system. used in analytical redundancy schemes [15]. The state
They can be modeled as in (4): observer is used for detecting faults and for the control
reconfiguration process after fault detection.
z f = (βz ∗ z + f0 ) : t ≥ 0 (4) b) Hardware redundancy: Hardware redundancy schemes
where z f is considered to be the faulty sensor output and z is use extra hardware components to achieve FTC of the sys-
the actual output of the sensor. tem. The purpose of the extra/redundant hardware module
2) Abrupt Faults: Abrupt faults are characterized by the oc- is to maintain the system operation even if one of the
currence of a sensor fault after a time t0 but not from the start. hardware components has a fault. A voter/decision logic
These faults can be modeled as in (5): component is present to detect the fault in one hardware
 module and then switch to the fault-free module.
z : t ≤ t0
z =
f
(5)
(βz ∗ z + f0 ) : t ≥ t0 IV. PROPOSED HYBRID SENSOR FAULT TOLERANT
3) Intermittent Faults: Intermittent faults are characterized CONTROL STRATEGY
by the occurrence of a sensor fault between times t0 and t1 . Our controller is a hybrid configuration of the analytical
These faults can be modeled as in (6): redundancy and hardware redundancy schemes of the FTC. The

⎨ z : t ≤ t0 analytical redundancy scheme is designed based on the sliding
z f = (βz ∗ z + f0 ) : t0 ≤ t ≤ t1 (6) mode observer (SMO) in tandem with adaptive passivity-based

z : t ≥ t1 control to handle DC bus voltage sensor faults. A double modular
hardware redundancy scheme is designed to handle battery
4) Open Sensor Faults: Open sensor faults are characterized current sensor faults, as shown in Fig. 3. The motivation for using
by the sensor giving zero output instead of the actual sensed a hybrid FTC with both analytical and hardware redundancy is
value. between a time t0 and t1 . These faults can be modeled as as follows:
in (7): r The cost of the current sensor is almost 1/5th of that of the

⎨ z : t ≤ t0 voltage sensor (in view of the LEM sensors LA-55P and
z f = 0 : t 0 ≤ t ≤ t1 (7) LV-25P). Therefore, hardware redundancy is more cost-

z : t ≥ t1 effective, especially for current sensors.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
1708 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 60, NO. 1, JANUARY/FEBRUARY 2024

The values of the observer gains are calculated by using the


conditions e = 0 and ė = 0, which are given in (11)
vdc
l1 = k1 (1 − u)
L
vdc
l2 = k 2 (11)
RC
where k1 and k2 are constants that are tunable to enable an
improved observer transient response. The proper choice of
the observer gains is made to maintain the error dynamics of
the observer stable, ensuring proper convergence of observer
estimates to the measured states [24].
2) Adaptive Passivity Based Control: Adaptive passivity-
based control [10] is used to generate the switching control signal
Fig. 3. Proposed Hybrid Sensor Fault Tolerant Control Strategy. to the BDC. This controller is used along with SMO to achieve
an analytical redundancy scheme to handle voltage sensor faults.
The average model of the BDC is given in (12) as follows:
r The estimation of both state variables is computationally 1
ẋ1 =[−(1 − u)x2 + E]
expensive with the need for a bank of observers. The L
observer modeling complexity problem is also a drawback 1
ẋ2 = [(1 − u)x1 − θx2 ] (12)
of analytical redundancy. C
where x1 = iL and x2 = vdc are the state variables. E is the
A. SMO-Based Analytical Redundancy battery voltage, and θ = (1/R0 ) is the output load conductance.
The state vector is defined as x, which is a column vector con-
As a part of this work, an SMO-based analytical redundancy sisting of x1 and x2 . The system’s unknown constant parameter
scheme has been incorporated into the controller to handle volt- vector is represented as Θ, which is a column vector containing
age sensor faults. An adaptive passivity-based controller is used E and θ. The parameter estimate vector for the unknowns is
with the SMO, along with the FDI and control reconfiguration denoted as Θ̂, a column vector consisting of the estimates for Ê
unit. and θ̂. The system given in (12) is represented as in (13).
1) Sliding Mode Observer Design: The sliding mode ob-
server is described bv the observer equations given in [23], [24]. G ẋ + (1 − u)J x + Rx = ζ (13)
The system dynamics are described by the general form given ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
L 0⎥ ⎢ 0 1⎥ ⎢0 0 ⎥ ⎢Ê ⎥
in (8). where G =⎢
⎣ ⎦ , J =⎣ ⎦, R = ⎣ ⎦, and ζ =⎣ ⎦.
0 C −1 0 0 θ̂ 0
ẋ = Ax + Bu (8) The error state variables are defined as x˜1 = x1 − x1d and x˜2 =
x2 − x2d , where x1d and x2d are the desired values of the states.
where x = [iL⎡vdc ]T is the state vector. ⎤The system matrix is The objective of the controller is to maintain the output
0 −(1 − u)/L⎥ voltage (x2 ) at a desired constant value (Vd ). Indirect control
given by A =⎢ ⎣ ⎦. A nonlinear term is is implemented due to the nonminimum value exhibited by the
(1 − u)/C −1/RC
boost converter. The error dynamics are therefore as in (14).
added to the state equation of the converter to obtain the observer
equation as in (9). This term is proportional to the difference G x̃˙ + (1 − u)J x̃ + Rx̃ = ζ − (G x˙d + (1 − u)J xd + Rxd )
between the actual value and the observed value. (14)
The error dynamics are rewritten after the addition of a damping
x̂˙ = Ax̂ + Bu + L sign(y − C x̂) (9)
term on both sides of (14), as shown in (15), (16).
where, x̂ = [iˆL vˆdc ]T denote the observed value of the state G x̃˙ + (1 − u)J x̃ + Rd x̃ = φ (15)
vector, sign(y − C x̂) is the non linear observer term, y is the
output of the system and L = [l1 l2 ]T is the observer gain vector. φ = ζ − (G x˙d + (1 − u)J xd + Rxd − RI x̃) (16)
⎡ ⎤
The observer error dynamics are given in (10) R1 0 ⎥
where RI =⎢
⎣ ⎦, R1 > 0 and Rd = RI + R. RI is the
−(1 − u)vdc 0 0
i˙L = + l1 sign(ȳ) additive damping term added to the error dynamics.
L
A quadratic energy function is applied using the state error
˙ = (1 − u)i̇L vdc vector and the control input. Correspondingly, a parameter adap-
vdc − + l2 sign(ȳ) (10)
C RC tation equation is derived as in (17).
˙
where iL = iˆL − iL , vdc = vˆdc − vdc and y = y − ŷ. Θ̂ = −Γ(Y x̃ − σ Θ̂) (17)

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
CHANDRA AND MOHAPATRO: HYBRID SENSOR FAULT TOLERANT CONTROL OF LOW VOLTAGE DC MICROGRID 1709

 
Γ1 0
where Γ = , Γ1 , Γ2 > 0 is the gain matrix to adjust
0 Γ2
 T
the convergence rate of the parameters. Θ̂ = Ê θ̂ is the
parameter estimation vector of the system.  σ > 0 is a constant
−1 0
design parameter and Y = . In the present work,
0 x2 d
the input battery voltage is considered to be constant. Therefore,
Ê is considered to be 24 V.
The system incorporates an adaptation law block and a con-
troller dynamics block. The adaptation law block utilizes system
state information (x), and the desired voltage (x2d ) in order to
estimate the unknown parameter vector (Θ̂) based on a specified
relationship in (17). This estimation is continuously updated us- Fig. 4. Flowchart of SMO based analytical redundancy to handle voltage
ing the system’s response and the desired output. The dynamics sensor faults.
of the desired states are also considered in the development of
the control scheme for the system. The dynamics of the desired
state x1d are given by (18). Algorithm 1: Double Modular Hardware Redundancy Al-
2
gorithm.
Vd Step 1: Obtain the measurements of the battery current
ẋ1d = −Γ1 θ̂x̃1 − Γ2 Ê x̃2 x2d (18)
Ê (IL ) through both the current sensors as I1 and I2 .
Step 2: Compare both the readings of the current sensors.
where x1d = (vref )2 θ̂/Ê. The choice of the parameters Γ1 and
If | I1 − I2 |< IT hreshold , then its fault free condition.
Γ2 of the adaptation law is done in such a way to achieve desired
Step 3: If I1 − I2 ≥ IT hreshold then current sensor-2 is
updation of the unknown parameters θ̂ and Ê such that the
faulty. Fault status flag of current sensor-1 is high
overall control of the states is achieved effectively.
(S2 = 1). Output of current sensor-1 (I1 ) is fed back to
In addition, the controller dynamics block takes system states
the controller.
(x), the estimated parameter vector (Θ̂), and the desired output
Step 4: If I2 − I1 ≥ IT hreshold then current sensor-1 is
voltage (Vd ) as inputs. Its primary purpose is to generate the
faulty. Fault status flag of current sensor-1 is high
control input (u) for the plant based on these inputs. The control
(S1 = 1). Output of current sensor-1 (I2 ) is fed back to
input is then defined as per (19).
the controller.
1
u=1+ Lẋ1d − Ê − R1 x̃1 (19)
x2d TABLE II
PARAMETERS OF THE PROPOSED HSFTC
where R1 is the additive damping term chosen as mentioned in
(16). The dynamics of desired state x2d is given by (20).
1
ẋ2d = (1 − u)x1d − θ̂x2d (20)
C
Overall, the adaptation law block and the controller dynamics
block work in tandem to achieve the desired control objectives.
The adaptation law block continually refines the estimated pa-
rameter vector, while the controller dynamics block utilizes this current sensors. If the faulty sensor is detected, then the con-
estimated vector, along with system states and desired output troller feedback is routed from the healthy current sensor. The
voltage, to determine the appropriate control input. double modular hardware redundancy is implemented as in
3) Flowchart of SMO Based Analytical Redundancy Scheme: Algorithm 1.
The measured state is compared with the estimated states of the
observer and is used to generate the residuals as r = x − x̂.
V. RESULTS AND DISCUSSION
Based on the residuals, the sensor fault is detected, and control
reconfiguration is performed, as shown in Fig. 4. The performance of the proposed Hybrid FTC control is ana-
lyzed in simulation using the MATLAB/SIMULINK platforms.
B. Double Modular Hardware Redundancy The parameters of the simulated HSFTC are listed in Table II.

To handle the current sensor faults, a double modular hard-


ware redundancy scheme is used, as shown in Fig. 3. This A. Performance of the System Without Proposed HSFTC
system consists of two identical current sensors (CS-1 and CS-2) The performance of the system with only the adaptive
measuring the same battery current. There is a voter/decision passivity-based controller [10] is observed under the situation
logic to detect the faulty sensor in coordination with the two of a faulty voltage sensor.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
1710 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 60, NO. 1, JANUARY/FEBRUARY 2024

Fig. 5. DC Bus voltage of the microgrid under the presence of faulty voltage Fig. 7. DC Bus voltage with sustained voltage sensor fault using proposed
sensor. HSFTC.

Fig. 8. Power balance with sustained voltage sensor fault using proposed
HSFTC.

Fig. 6. Impact of faulty voltage sensor on the load power.

Two situations of voltage sensor faults, sustained faults and


abrupt faults, are considered with multiplicative fault gain βz =
0.8 and additive bias fault of f0 = −2. Under such a situation
of a faulty voltage sensor, it is seen from Fig. 5 that the DC
bus voltage is highly deviating with a voltage of approximately
58 V instead of the 48 V reference. Additionally, it is a factor of
concern that the power consumed by the load is approximately
350 W, as seen in Fig. 6, which is approximately 110 W more
than its rated value of 240 W. Therefore, the presence of a faulty Fig. 9. DC Bus voltage with abrupt voltage sensor fault using proposed
voltage sensor is very problematic in view of the voltage tracking HSFTC.
and power balance criteria of the microgrid.
objective is achieved, as shown in Fig. 9. The transients in
B. Performance of the System With Voltage Sensor Faults With bus voltage due to the occurrence of an abrupt sensor fault
the Proposed HSFTC settle quickly in approximately 0.24 seconds with an overshoot
of approximately 1.2%. Power balance of the system is also
As discussed earlier, the SMO-based analytical redundancy
achieved, as shown in Fig. 10. The transients in power also settle
scheme present in our controller is used to handle voltage sensor
down in approximately 0.26 s with an overshoot of 2.2%.
faults. Two types of voltage sensor faults, namely, sustained and
abrupt sensor faults, are considered, which are discussed in this
section. C. Performance of the System With Current Sensor Faults
With the Proposed HSFTC
1) Sustained Voltage Sensor Gain Fault: A sustained voltage
sensor fault with multiplicative fault component βz = 0.8 and As discussed earlier, a double modular hardware redundancy
additive bias fault of f0 = −2 is considered. Under such a fault scheme present in our controller is used to handle the current
situation, it is observed in Fig. 7 that the DC bus voltage tracking sensor faults. Two identical current sensors are used to measure
with our HSFTC is proper at approximately 48 V. The power the battery current (IL ). These sensors have status flags S1 and
balance is also satisfied, as shown in Fig. 8. S2 , respectively, to indicate a corresponding sensor being faulty.
2) Abrupt Voltage Sensor Gain Fault: An abrupt voltage 1) Sustained Current Sensor Gain Fault: The current sensor-
sensor fault with multiplicative fault component βz = 0.8 and 1 is considered to have a sustained fault with a multiplicative
additive bias fault of f0 = −2 is considered to occur at t = 2 fault component βz = 0.8 and an additive bias fault of f0 =
seconds. In such a situation, the system’s voltage tracking −2. Under such circumstances, our controller is very capable

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
CHANDRA AND MOHAPATRO: HYBRID SENSOR FAULT TOLERANT CONTROL OF LOW VOLTAGE DC MICROGRID 1711

Fig. 10. Power balance with abrupt voltage sensor fault using proposed Fig. 13. DC bus voltage & battery current with load disturbances, intermittent
HSFTC. voltage sensor fault and open current sensor fault using the proposed HSFTC.

Fig. 11. DC Bus voltage and battery current with proposed HSFTC under the Fig. 14. Power variations with load disturbances, intermittent voltage sensor
situation of sustained current sensor fault. fault and open current sensor fault using the proposed HSFTC.

D. Performance of the System Under Load Disturbance With


the Proposed HSFTC
The aspect of handling load disturbances by the proposed FTC
is discussed here. The load disturbance of 20% increment and
20% decrement of the base load is considered at instants t =
4 s and t = 8 s, respectively. Along with the load disturbances,
an intermittent voltage sensor fault occurs between t = 2 s and
t = 6 s. Additionally, an open current sensor fault for current
sensor-1 is considered to be occurring between t = 3 s and t =
Fig. 12. DC Bus voltage and battery current with abrupt current sensor fault
7 s. The performance of the proposed HSFTC in terms of the
using proposed HSFTC. system voltage profile and power profile is shown in Figs. 13
and 14, respectively.
The transient effect in the bus voltage due to the intermit-
tent voltage sensor fault is settled down in approximately 0.24
of maintaining the DC bus voltage at approximately 48 V, as seconds with an overshoot of approximately 1.2%. The voltage
shown in Fig. 11. The battery current is also observed to track the transients due to load change are also effectively handled within
reference current of approximately 10 A without any deviation a settling time of approximately 0.16 seconds with an overshoot
caused by the sensor fault. Power balance has been achieved of approximately 2.2%. The load power transients are observed
without any influence of the current sensor fault. with an overshoot of approximately 4% with a settling time of
2) Abrupt Current Sensor Gain Faults: Current sensor-1 is approximately 0.15 seconds. Therefore, the load disturbances
considered to have an abrupt fault at t = 2 seconds with mul- are handled effectively by the our FTC, which demonstrates a
tiplicative fault component βz = 0.8 and additive bias fault of good transient response even under the presence of an intermit-
f0 = −2. In this situation, we observe that the system voltage tent voltage sensor fault and open current sensor faults.
and battery current are maintained at their references, despite
the current sensor-1 becoming faulty at t = 2 seconds, as shown
in Fig. 12. The transients in the bus voltage due to the abrupt E. Performance of the System Under Simultaneous Load
Change and Sensor Faults With the Proposed HSFTC
current sensor fault are settled in approximately 0.11 seconds
with an overshoot of approximately 0.8%. Therefore, the voltage The aspect of handling simultaneous events of load change
transient recovers, and power balance is also achieved. and sensor faults by our FTC is further investigated.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
1712 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 60, NO. 1, JANUARY/FEBRUARY 2024

Fig. 15. DC bus voltage with simultaneous load disturbance, abrupt voltage Fig. 17. Experimental setup of the considered LVDC Microgrid.
sensor fault, and abrupt current sensor fault at t = 2 s using the proposed
HSFTC.

Fig. 16. Power variation with simultaneous load disturbance, abrupt voltage
sensor fault, and abrupt current sensor fault at t = 2 s using the proposed
HSFTC. Fig. 18. DC bus voltage of the considered system with abrupt voltage sensor
fault at t = 2 s using the proposed HSFTC.

A load increment of approximately 20%, along with an abrupt


voltage sensor fault and an abrupt current sensor fault of current
sensor-1, is considered at the same instant t = 2 s. The voltage
tracking and system power balance aspects of the proposed FTC
are observed under such a situation, as shown in Figs. 15 and 16.
The transient effect in the bus voltage due to the simultaneous
occurrence of sensor faults and load change settles after ap-
proximately 0.22 seconds with an undershoot of approximately
1.2%. The load power transients are observed with an overshoot
of approximately 2.5% with a settling time of approximately
0.21 seconds. Despite the occurrence of sensor faults and load
changes at the same time, our HSFTC is able to handle the
voltage and power response effectively. Fig. 19. Power balance in the considered system with abrupt voltage sensor
fault at t = 2 s using the proposed HSFTC.
VI. EXPERIMENTAL VALIDATION
This section describes the experimental validation of the
A. Performance of the System With Voltage Sensor Faults With
proposed HSFTC algorithm on the laboratory-scale LVDC mi-
crogrid testbed, as shown in Fig. 17. The setup consists of battery Proposed HSFTC
storage of 24 V and 100 Ah capacity interfaced to a Semikron An abrupt voltage sensor fault is considered to be occurring at
DC-DC buck-boost converter stack with a base resistive load t = 2 s with a gain fault component of β = 0.8 and an additive
of 240 W (12 Ohm, 12 A). A dSPACE 1103 interface is used fault component of −2. Under these conditions, the performance
to build the control logic and generate the PWM signals for of the FTC in terms of the voltage and power profiles is observed
the converter. To implement the closed-loop control for the in Figs. 18 and 19, respectively.
system, LEM voltage sensors (LV-20) and LEM current sensors Our controller therefore achieves reference tracking and
(LA-55P) are used. power balance in the system despite the presence of an abrupt
The performance of the proposed HSFTC under different voltage sensor fault. The battery supplies a current of approx-
situations is validated on the laboratory testbed. imately 10.4 A. The slight difference of approximately 10 W

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
CHANDRA AND MOHAPATRO: HYBRID SENSOR FAULT TOLERANT CONTROL OF LOW VOLTAGE DC MICROGRID 1713

Fig. 20. DC bus voltage of the considered system with abrupt current sensor Fig. 22. DC bus voltage of the considered System with load disturbances,
fault at t = 2 s using the proposed HSFTC. intermittent voltage sensor fault and open current sensor fault using the proposed
HSFTC.

Fig. 21. Power balance in the considered system with abrupt current sensor
fault at t = 2 s using the proposed HSFTC. Fig. 23. Power balance in the considered system with load disturbances,
intermittent voltage sensor fault and open current sensor fault using the proposed
HSFTC.

between the battery power and the load power accounts for
the overall losses of the system. The transient bus voltage due C. Performance of the System Under Load Disturbance With
to the occurrence of an abrupt sensor fault settles quickly in the HSFTC
approximately 0.25 seconds with an overshoot of approximately The aspect of handling load disturbances by the HSFTC is
4%. The transients in power also settle down in approximately discussed here. The load disturbance of 20% increment and 20%
0.26 s with an overshoot of 4.1%. decrement of the base load is considered at around t = 4 s and
t = 8 s, respectively.
Along with the load disturbances, an intermittent voltage
B. Performance of the System With Current Sensor Faults With sensor fault occurs between t = 2 s and t = 6 s. Additionally,
the Proposed HSFTC
an open current sensor fault of current sensor-1 is considered
Our HSFTC is further tested in its ability to handle the current to occur between t = 3 s and t = 7 s. The performance of the
sensor faults acting alone. An abrupt current sensor fault is FTC in terms of the system voltage profile and power profile is
considered to be occurring at t = 2 s with a gain fault component shown in Figs. 22 and 23, respectively. The transient effect in the
of β = 0.8 and an additive fault component of −2. Under these bus voltage due to the intermittent voltage sensor fault settles in
conditions, the performance of the FTC in terms of the voltage approximately 0.24 seconds with an overshoot of approximately
and power profiles is shown in Figs. 20 and 21, respectively. The 3%. The voltage transients due to load change are also effectively
transient bus voltage due to the occurrence of an abrupt current handled within a settling time of approximately 0.18 seconds
sensor fault is settled in approximately 0.15 seconds with an with an overshoot of approximately 3.2%. The load power
overshoot/undershoot of approximately 2%. The transients in transients are observed with an overshoot of approximately 8.5%
power also settle down in approximately 0.18 s with an overshoot with a settling time of approximately 0.2 seconds. Therefore,
of 3%. Even under the presence of the abrupt current sensor the HSFTC performs well by maintaining the control objectives
fault, the voltage is maintained at 48 V, and the power balance even under the presence of intermittent voltage sensor faults and
is maintained in the system. open current sensor faults.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
1714 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 60, NO. 1, JANUARY/FEBRUARY 2024

Fig. 26. Comparison of DC bus voltage with sensor fault using proposed
HSFTC and other control strategies.

Fig. 24. DC bus voltage of the considered system with load disturbance, abrupt
voltage sensor fault, and abrupt current sensor fault at t = 2 s using the proposed E. Comparative Analysis
HSFTC.
The performance of our controller under the situation of a
faulty voltage sensor is compared with several popular control
strategies, such as cascaded control [3], sliding mode control
[6], adaptive sliding mode control [9], and adaptive passivity
control [10]. In this comparison, a sustained voltage sensor fault
is considered with a multiplicative fault component βz = 0.8
and an additive bias fault of f0 = −2. As shown in Fig. 26,
the popular control strategies fail to deliver the voltage tracking
objective under the situation of a voltage sensor fault. Our
controller outperforms these popular control strategies with
effective voltage tracking even under sensor faults.

VII. CONCLUSION
The effect of sensor faults on the control performance of
an LVDC Microgrid has been thoroughly investigated. To ef-
fectively handle the presence of sensor faults in the system, a
hybrid FTC control has been implemented. This controller has
Fig. 25. Power balance in the considered system with load disturbance, abrupt
voltage sensor fault, and abrupt current sensor fault at t = 2 s using the proposed analytical = and hardware redundancy components to handle
HSFTC. voltage sensor and current sensor faults, respectively. The goal
of DC bus voltage regulation and power balance in the system is
effectively achieved even in the presence of voltage and current
sensor faults. The performance of our controller is verified under
D. Performance of the System Under Simultaneous Load different sensor fault situations. We observed that the proposed
Change and Sensor Faults With the HSFTC control outperforms some of the popular controls in terms of
fault tolerance, voltage tracking, and transient response for load
The HSFTC is finally tested for handling simultaneous fault disturbances and sensor faults.
events. The events of load increase (of 20%), abrupt voltage
sensor fault, and abrupt current sensor fault of current sensor-1
REFERENCES
are considered to occur at t = 2 s.
The occurrence of the events at the same time is considered [1] A. Jhunjhunwala and P. Kaur, “Solar energy, DC distribution, and mi-
crogrids: Ensuring quality power in Rural India,” IEEE Electrific. Mag.,
as a severe situation for managing with a fail safe system. vol. 6, no. 4, pp. 32–39, Dec. 2018.
Even under such a situation, the performance of our FTC is [2] M. V. Satya Sai Chandra, B. D. B, L. V. Kumar, and S. Mphapatro, “Voltage
evaluated positively, as shown in Figs. 24 and 25. In spite of control and energy management of solar PV fed stand-alone low voltage
DC microgrid for rural electrification,” in Proc. 21st Nat. Power Syst.
the three simultaneous events occurring at t = 2 s, our FTC Conf., 2020 pp. 1–6.
manages the voltage profile and the power profiles decently well. [3] S. Bacha, I. Munteanu, and A. I. Bratcu, Power Electronic Converters
The transient effect in the bus voltage due to the simultaneous Modelling and Control: With Case Studies, vol. 454. Berlin, Germany:
Springer, 2014.
occurrence of sensor faults and load change settles in approxi- [4] F. Cingoz, A. Elrayyah, and Y. Sozer, “Optimized settings of droop
mately 0.22 seconds with an undershoot of approximately 4%. parameters using stochastic load modeling for effective DC microgrids
The load power transients are observed with an overshoot of operation,” IEEE Trans. Ind. Appl., vol. 53, no. 2, pp. 1358–1371,
Mar./Apr. 2017.
approximately 8.5% with a settling time of approximately 0.23 [5] R. Bhosale, R. Gupta, and V. Agarwal, “A novel control strategy to achieve
seconds. Therefore, the HSFTC is observed to handle a wide SOC balancing for batteries in a DC microgrid without droop control,”
variety of sensor fault situations and loading situations. IEEE Trans. Ind. Appl., vol. 57, no. 4, pp. 4196–4206, Jul./Aug. 2021.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.
CHANDRA AND MOHAPATRO: HYBRID SENSOR FAULT TOLERANT CONTROL OF LOW VOLTAGE DC MICROGRID 1715

[6] D. K. Fulwani and S. Singh, Mitigation of Negative Impedance Instabil- [25] S. Saha et al., “Diagnosis and mitigation of voltage and current sensors
ities in DC Distribution Systems(SpringerBriefs in Applied Sciences and malfunctioning in a grid connected PV system,” Int. J. Elect. Power Energy
Technology). Berlin, Germany: Springer, 2017. Syst., vol. 115, 2020, Art. no. 105381.
[7] Y. Zhao, W. Qiao, and D. Ha, “A sliding-mode duty-ratio controller for [26] E. Shahzad et al., “Sensor fault-tolerant control of microgrid using robust
DC/DC buck converters with constant power loads,” IEEE Trans. Ind. sliding-mode observer,” Sensors, vol. 22, no. 7, pp. 1–38, 2022.
Appl., vol. 50, no. 2, pp. 1448–1458, Mar./Apr. 2014. [27] S. Saha, T. K. Roy, M. A. Mahmud, M. E. Haque, and S. N. Islam, “Sensor
[8] S. Singh, D. Fulwani, and V. Kumar, “Robust sliding-mode control of fault and cyber attack resilient operation of DC microgrids,” Int. J. Elect.
DC/DC boost converter feeding a constant power load,” IET Power Elec- Power Energy Syst., vol. 99, pp. 540–554, 2018.
tron, vol. 8, no. 7, pp. 1230–1237, 2015. [28] S. Mehdi and A. Mehdi, “New hardware redundancy approach for making
[9] G. Mustafa, F. Ahmad, R. Zhang, E. Ul Haq, and M. Hussain, “Adaptive modules tolerate faults using a new fault detecting voter unit structure,”
sliding mode control of buck converter feeding resistive and constant power IET Circuits, Devices Syst., vol. 14, no. 7, pp. 980–989, 2020.
load in DC microgrid,” Energy Rep., vol. 9, pp. 1026–1035, 2023. [29] E. Moradi, H. Iman-Eini, and S. Sheikhaei, “A simple hardware-based
[10] A. M. Dissanayake and N. C. Ekneligoda, “Adaptive passivity based fault-tolerant method for cascaded H-bridge converters,” IEEE Trans. Ind.
control of DC-DC power electronic converters,” in Proc. IEEE Appl. Power Electron., vol. 69, no. 10, pp. 9711–9720, Oct. 2022.
Electron. Conf. Expo., 2019, pp. 2984–2988. [30] S. Das and M. Manohar, “A resilient current sensor fault tolerant strategy
[11] A. Abedi, B. Rezaie, A. Khosravi, and M. Shahabi, “DC-bus voltage for vector-controlled induction motor drive,” IEEE Trans. Emerg. Sel.
control based on direct Lyapunov method for a converter-based stand-alone Topics Power Electron., vol. 11, no. 4, pp. 4313–4320, Aug. 2023.
DC micro-grid,” Electric Power Syst. Res., vol. 187, 2020, Art. no. 106451. [31] M. V. S. S. Chandra and S. Mohapatro, “Hybrid sensor fault toler-
[12] J. Wu and Y. Lu, “Adaptive backstepping sliding mode control for ant control of low voltage DC microgrid,” in Proc. IEEE Glob. Conf.
boost converter with constant power load,” IEEE Access, vol. 7, Comput., Power Commun. Technol., 2022, pp. 01–06, doi: 10.1109/Glob-
pp. 50797–50807, 2019. ConPT57482.2022.9938148.
[13] Y. Wang, D. Zhou, S. J. Qin, and H. Wang, “Active fault-tolerant control
for a class of nonlinear systems with sensor faults,” Int. J. Control. Autom.
Syst., vol. 6, no. 3, pp. 339–350, 2008.
[14] M. S. Mahmoud and Y. Xia, Analysis and Synthesis of Fault-Tolerant
Control Systems. Hoboken, NJ, USA: Wiley, 2014.
[15] M. V. Satya Sai Chandra and S. Mohapatro, “Active sensor fault tolerant M. V. Satya Sai Chandra (Graduate Student Mem-
control of bus voltage in standalone low voltage DC microgrid,” Elect. ber, IEEE) was born in Srikakulam, India, in 1993. He
Eng., vol. 105, pp. 1079–1092, 2023. received the B.Tech. degree from Jawaharlal Nehru
[16] D. Li, Y. Wang, J. Wang, C. Wang, and Y. Duan, “Recent advances in Technological University, Kakinada, India, in 2014,
sensor fault diagnosis: A review,” Sensors Actuators A: Phys., vol. 309, and the M.Tech. degree in instrumentation and con-
2020, Art. no. 111990. trol systems from the National Institute of Technol-
[17] T. H. Lee, C. P. Lim, S. Nahavandi, and R. G. Roberts, “Observer-based ogy, Calicut, India, in 2017. He is currently working
H∞ fault-tolerant control for linear systems with sensor and actuator toward the Ph.D. degree in electrical engineering
faults,” IEEE Syst. J., vol. 13, no. 2, pp. 1981–1990, Jun. 2019. with DEI Laboratory, Indian Institute of Technol-
[18] Z. Ben Safia, M. Kharrat, M. Allouche, and M. Chaabane, “TS fuzzy fault- ogy Bhubaneswar, Bhubaneswar, India. His research
tolerant tracking control of a PV pumping system based on an induction interests include fault-tolerant control applications,
motor,” IETE J. Res., vol. 69, no. 7, pp. 4826–4835, 2023. advanced control of DC microgrids, and energy management of DC microgrids.
[19] A. A. Amin and K. M. Hasan, “A review of fault tolerant control sys-
tems: Advancements and applications,” Measurement, vol. 143, pp. 58–68,
2019.
[20] G. H. B. Foo, X. Zhang, and D. M. Vilathgamuwa, “A sensor fault detection
and isolation method in interior permanent-magnet synchronous motor
drives based on an extended Kalman filter,” IEEE Trans. Ind. Electron., Sankarsan Mohapatro (Senior Member, IEEE) was
vol. 60, no. 8, pp. 3485–3495, Aug. 2013. born in Cuttack, India, in 1983. He received the B.E.
[21] Y. Zuo et al., “Current sensor fault-tolerant control for speed-sensorless in- degree from the National Institute of Science and
duction motor drives based on the SEPLL current reconstruction scheme,” Technology, Berhampur, India, in 2005, and the M.Sc.
IEEE Trans. Ind. Appl., vol. 59, no. 1, pp. 845–856, Jan./Feb. 2023. (Eng.) and Ph.D. degrees in electrical engineering
[22] C. Edwards and C. P. Tan, “Sensor fault tolerant control using sliding mode from the Indian Institute of Science, Bangaluru, India,
observers,” Control Eng. Pract., vol. 14, no. 8, pp. 897–908, 2006. in 2008 and 2011, respectively. He is currently an
[23] S. Drakunov and V. Utkin, “Sliding mode observers. Tutorial,” in Proc. Associate Professor with the School of Electrical Sci-
IEEE 34th Conf. Decis. Control, 1995, vol. 4, pp. 3376–3378. ences, Indian Institute of Technology, Bhubaneswar,
[24] P. Bhartiya, N. Rathore, and D. Fulwani, “A tutorial on implementation India. His research interests include high voltage ap-
of sliding mode observer for DC/DC power converters using FPGA,” in plications for reducing air pollution, DC microgrids,
Proc. IECON Proc. Ind. Electron. Conf., 2014, pp. 4153–4159. and renewable energy.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on April 22,2024 at 14:07:52 UTC from IEEE Xplore. Restrictions apply.

You might also like