Current Observer Based Predictive Decoupled Power

Control Grid-Interactive Inverter

Zhen Zhang Student Member, IEEE, Adepoju Webster Oluwafemi, Student Member, IEEE,
Mohsen Hosseinzadehtaher, Student Member, IEEE, Mohammad B. Shadmand, Member, IEEE
Power Electronics & Autonomous Systems (PEAS) Research Group, Dept. of Electrical & Computer Engineering,
Kansas State University, Manhattan, KS, USA
[email protected]; [email protected]; [email protected]; [email protected]

Abstract— This paper presents a sensor-less current model the control scheme of the inverters and may put grid resiliency
predictive control (MPC) scheme via a full state observer based at risk.
current estimator. The grid interactive inverters’ control schemes The observer-based methods that are common in the
require current and voltage sensors. Elimination of current sensor literature are focused on reducing voltage sensors in grid-
enhances the inverter reliability. This paper leverages the inherent interactive inverters [9, 10]. A few instances of a current
characteristics of MPC towards robust current sensor-less grid observer are novel and well-regarded, but there is little explicit
interactive inverter with LC filter. The observer for inductor attention to current controlled grid interactive inverter
current is developed based on the existing capacitor voltage applications [11]. The obscurity lies in operation of current
measurement. The estimator dynamic state-space matrices are
control without using a current sensor. The proportional-integral
then obtained through reconstruction of the inverter model with
(PI) controller is common in the inverters at the grid edge. The
the capacitor voltage and inductor current being the state
variables. The controller objectives are to regulate active and
Proportional Resonant (PR) controller is an extension of the PI
reactive power in a decoupled manner. The theoretical controller and has good operability for grid-tied applications due
expectation, controller performance, and accuracy of the current to its inherent tuning coordinated to the grid frequency [12-18].
estimation are verified by conducting a real-time simulation via Both PI and PR controllers require precise tuning, which dictates
Typhoon HIL. the effectiveness of the controller in terms of transient response
and steady state error [17, 18]. Furthermore, predictive control
Keywords –smart inverter, current sensor-less, full state as well as observer have more computational burden than what
observer, model predictive control, grid–interactive inverter. is required for PI and PR controllers. This underlines the
challenges in achieving current sensor-less control using PI or
I. INTRODUCTION PR control as having worse dynamic response due to the
As the trend of the inverter-dominated grid rises, the focus additional computational loop needed, compared to predictive
on robust inverter control techniques increases for both the grid- control.
forming as well as the supportive grid-following inverter [1, 2]. Observer-based model predictive control (MPC) provides a
The motivation for a current sensor-less predictive control simple and robust solution to the aforementioned issues. The
scheme is based on concerns for reliability of inverters at the advantages of predictive control are that there is no need for any
grid edge. In grid interactive inverters, current sensors are tuning of controller gains, as well as having fast dynamic
utilized for the controller feedback to regulate the injected response to transients, due to the inherent calculation-by-step
current into the grid. Sensing techniques include resistive- nature of predictive control [19, 20]. These characteristics of the
current sensing, inductor resistance-based sensing, current MPC are leveraged in this paper to develop a reliable and robust
transformer-based sensing, and hall-effect sensors among other observer based predictive decoupled power control for inverters
techniques [3], all of which are subject to either high power at the grid-edge. The observer is built depending on the output
dissipation or noisy measurement [4]. The issue at hand is that filter – “LC” in this study – and utilizes the capacitor voltage
current sensors, specifically, are known to have noisy sensor to observe the inductor current without any use of current
measurements in comparison to voltage sensors. In the effort to sensor. Furthermore, the study attempts to demonstrate that the
minimize reliance on sensors, the priority is to first consider quality of the controller was not compromised by this technique
current sensors, as they are more vulnerable in reliability than unlike classical control schemes for grid interactive inverters,
voltage sensors. This is further emphasized in the application of showing only minimal difference between observer-based
predictive control for grid-tied inverters [5-8]. The prediction predictive control and sensor-based predictive control.
accuracy of the inverter output current in horizon of time for The rest of this paper is organized as follows: Section II
predictive control schemes heavily depends on the reliable and proposes the control operation concept. Section III provides the
accurate measurement of the output current sensors. From a predictive control formulation and the proposed observer model.
greater perspective, the reliability of the whole inverter The derivation of the observed inductor current based on the
operation as well as resiliency of high inverter penetrated grid proposed observer is also presented. Further, section IV presents
depends on the number of sensors; each sensor is mission- a real-time simulation of the proposed observer showing a close
critical, and the failure of any single sensor will cause failure in match between the measured and observed inductor current

978-1-7281-4436-8/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: Carleton University. Downloaded on August 02,2020 at 21:50:54 UTC from IEEE Xplore. Restrictions apply.
 Fig. 1. Proposed current observer-based predictive decoupled power control scheme.

TABLE. I: SWITCHING STATES OF INVERTER WITH VOLTAGE VECTOR The state space representation of (3) is given by (4),
Switching States S1 S2 S3 S4 ij ‫ݒ‬௜௡௩
ª 1º ª1 º
1st State 0 0 1 1 0 0 0 − » 0»
2nd State 1 1 0 0 0 0 d ªiL º « L ªiL º « L ªvinv º
« » = « » « »+« » « »
dt v 1 vc ¼ « −1 » ¬ig ¼ (4)
 ¬ c¼ « » ¬
3rd State 1 0 0 1 1 +VDC
0 N 0 
«
¬C »¼ x(t ) «¬ »
 C¼
4th State 0 1 1 0 -1 -VDC
x (t )  u (t )
A B
using Typhoon HIL 402 platform. Section V is the concluding In order to predict the next step of the inductor current or
remarks of this paper. capacitor voltage in horizon of time, Euler forward method is
II. PROPOSED CONTROL OPERATION CONCEPT used. The discretized state-space equations are given by,
x[k + 1] = Ad x[k ] + Bd u[ k ]
As shown in Fig. 1, the considered inverter topology is a (5)
single-phase grid-interactive inverter with a second-order LC y = Cd x[k ]
filter. Throughout this paper, the DC source, inverter, and ªiL [k + 1] º ªiL [k ] º ªvinv [k ]º
capacitor voltages are respectively denoted asܸ஽஼ ǡ ‫ݒ‬௜௡௩ and ‫ݒ‬௖ «v [k + 1]» = Ad «v [k ]» + Bd «i [k ] » (6)
while ݅௅ represents the inductor current. The possible switching ¬ c ¼ ¬ c ¼
 ¬g
 ¼
states and their corresponding inverter bridge output voltage are x[k ] x[k ] u[k ]
listed in Table I. The output voltage of the inverter bridge, ‫ݒ‬௜௡௩ where A d , B d and C d can be found from (4).
is given by,
vinv = ϕ (t )VDC This paper uses Second Order Generalized Integrator
(1) (SOGI) to generate the orthogonal signal required for creating
ϕ (t ) ∈ {1, 0, −1}
the artificial rotating reference frame dq. This method is suitable
where ϕ(t) is calculated by: for generating the reference signal because it has good dynamic
ϕ(t) = s1 (t)s4 (t) − s2 (t)s3 (t) (2) behavior along with attenuation of high-frequency harmonics
[21, 22].
Using Kirchhoff’s voltage and current law and considering
the state variables ݅௅ and ܸ஼ coupled with the input parameters, The grid voltage phase is detected using a combination of
ܸ௜௡௩ and݅௚ , the differential equations for both the filter inductor SOGI and Phase-Locked Loop (PLL). The dq components of the
current iL and the capacitor voltage vc are given by, output grid voltage vd and vq are obtained for calculating the
diL reference current components id and iq . The resulting dq current
L = vinv − vc
dt components are then subjected to inverse Park transformation to
(3)
dv obtain its stationary frame equivalent, iref .The dq components of
C c = iL − ig
dt the reference current are given by,

Authorized licensed use limited to: Carleton University. Downloaded on August 02,2020 at 21:50:54 UTC from IEEE Xplore. Restrictions apply.
 2 ( Pref vd [k ] + Qref vq [k ]) Start
id [k ] =
vd [k ]2 + vq [k ]2
(7)
2 ( Pref vq [k ] − Qref vd [k ]) Measure
iq [k ] = 2
vd [k ] + vq [k ] 2 vC
iref ( k ) = id (k ) sin(θ g (k )) + iq ( k ) cos(θ g (k )) (8)
Full State Observer
III. PROPOSED OBSERVER DESIGN AND PREDICTIVE CONTROL ª 1 º
ªi º ª º
0 − − m1 » ˆ 1
d ªiˆL º « L m1 » ªvinv º
The major contribution of this paper is to design a robust and « » = « »« L » +«L
dt ¬vˆc ¼ « 1 vˆ « » «v »
accurate observer to estimate the inductor current for one step − m2 » ¬ c ¼ «¬ 0 m2 »¼ ¬ c ¼
ahead in horizon of time by using only the already measured ¬« C ¼»
output voltage as the observer input. The observer based MPC
flowchart is shown in Fig 2. The general state equations for the Predicted Current Calculation
observer are given by,
ª T º
xˆ = Axˆ + Bu + My ªiˆL (k + 1) º «
1 − s − Ts m1 » ˆ
L
ªT º
ªi (k ) º « s Ts m1 » ª vinv º
y = y − yˆ (9) « »=« »« L »+« L » «v »
¬ vˆc ( k + 1) ¼ « Ts vˆ (k )
1 − Ts m2 » ¬ c ¼ «¬ 0 Ts m2 »¼ ¬ c ¼
y = vc , yˆ = vˆc «¬ C »¼
where x , u , and y are the state variables, the input, and the
output of the system respectively. M is the observer gain matrix. Cost Function Calculation
vc and vˆc are the measured and estimated output voltage of the G = iˆL [k + 1] − iref
capacitor filter. vc is the input to the observer. vc provides a s[k + 1] = arg min(G )
closed-loop feedback path and is given by,
vc = vc − vˆc (10)
Apply switching state
In order to obtain a perfect estimate of the inductor current, corresponding to
the steady-state value for vc should be zero. The closer vc is to the optimal state
zero, the more accurate the state estimate becomes; a zero value
for vc means the observer is completely matched with the Next time step
measurement value.
The given system in (6) is only observable if the Fig. 2. Observer-based MPC flowchart
observability matrix represented in (11) is full rank [23, 24],
where C is the output matrix and determines the relationship ª −1 − m º ˆ
d ªiˆL º « 0 L 1 ªi º ª1 m1 º ª vinv º
between the system state variables and the system output, and A »« L »+« L »« »
« »= (13)
is the matrix which describes the system. dt ¬ vˆc ¼ « 1 − m2 » ¬ vˆc ¼ ¬« 0 m2 ¼» ¬ vc ¼
¬ C ¼
ªC º
O( A, C ) = « » (11) It should be mentioned that there is a trade-off between
¬CA ¼ good dynamic performance and good noise rejection in tuning
From the discrete nature of the observer, it is assumed that the observer gains [25]. Therefore, in setting the observer gains,
there is no instantaneous changes in the grid current for a it is best to achieve a compromise between those two
sufficiently small sampling time. The closed-loop equation of characteristics. Appling Euler forward technique on (12) gives
the observer is given by, the resulting equations shown by,
diˆL 1 iˆL [ k + 1] − iˆL [ k ] vinv [ k ] − vˆc [ k ]
= (vinv − vˆc ) + m1 (vc − vˆc ) = + m1 (vc [ k ] − vˆc [ k ])
dt L Ts L
(12) (14)
dvˆc iL vˆc [ k + 1] − vˆc [ k ] iˆL [ k ]
= + m2 (vc − vˆc ) = + m2 ( vc [ k ] − vc [ k ])
ˆ
dt C Ts C
where m1 and m2 are the observer gains, iˆL and vˆc are the Readjusting (14) into state space representation results in the
predicted inductor current and capacitor voltage respectively. expression given by,
The expression vc − vˆc is the observer error. The state variables ª Ts º
1 − − Ts m1 » ˆ ª Ts º
ªiˆL [k + 1] º « L ªi [k ] º Ts m1 » ªvinv [k ]º
are iL and vc . The state space model for the observer is given by, « » = « »« L » + « L (15)
¬vˆc [k + 1]¼ « Ts vˆ [k ] « » «v [k ] »¼
1 − Ts m2 » ¬ c ¼ ¬« 0 Ts m2 ¼» ¬ c
¬« C ¼»
The desired poles of the system are set to zero such that,

Authorized licensed use limited to: Carleton University. Downloaded on August 02,2020 at 21:50:54 UTC from IEEE Xplore. Restrictions apply.
 ª Ts º power independently along with its corresponding reference
«1 − − Ts m1 » current. Finally, the functionality of the proposed control
L
det( zI − « » ) = ( z − 0)( z − 0) (16) method is compared with a conventional stationary reference
« Ts 1 − Ts m 2 »» frame PR based current controller in this study.
¬« C ¼
A. Proposed Observer Performance
The observer gains m1 and m 2 are found by co-efficient In this case study, the observer performance is evaluated.
matching of the closed-loop and desired closed-loop equations Observer gains were obtained through first calculating from
on the left and right sides as depicted in (16). The inputs to the (16). Observer gains were set to m1 = 400, and m2 = 95000. Fig.
controller are the filter capacitor voltage and the observed 3a shows the observed inductor current at steady-state condition.
inductor current. The output of the controller is the set of Fig. 3b shows the measured inductor current by sensor. In order
switching states for the corresponding time step. The switching to show the observer performance intuitively, the measured and
states are dictated by the sets available in Table I. Once the filter observed currents are both shown in Fig. 3c. Fig. 3d is the
inductor current is estimated via the observer, it is used as an generated current reference corresponding to the power
input to the predictive control to obtain its next state prediction reference. Root-mean-square error (RMSE), which is given by,
for all feasible switching states given in Table I. These predicted
currents for all switching states are compared with the reference n ( iˆL , j − iL , j ) 2
current in a cost function G subject to minimization, the cost
RMSE = ¦
j =1 n
(18)
function G is formulated as, is calculated to show the accuracy of the observer compared to
G = iˆL [k + 1] − iref the sensor current value mathematically, and is found to be
(17) 6.3%. As it is seen in Fig. 3c, there is negligible difference
s[k + 1] = arg min(G ) between observed and measured values.
where G is the cost vector, which is a function of the control Another metric for evaluating the observer performance,
actions (switching sequences) defined in Table I. Thus, the along with performance of the current control, is FFT harmonic
selected switching sequence for k+1 is the control action which analysis. Fig. 4 shows the spectrum of harmonics of the observer
minimizes the cost vector. For the lowest value of the cost current, with THD of 2.64%. The harmonic analysis of the
function, the corresponding switching states are selected by the observed inductor current shows that the quality of the current
controller to be applied to the inverter. waveform is acceptable by using this control scheme. In this
IV. SIMULATION RESULTS AND DISCUSSION case study, the real power and reactive power references are set
to 1 kW and 0 VAR respectively.
Several case studies are investigated for evaluating the
performance of the proposed control scheme and designed B. Decoupled Control of Real and Reactive Power
observer. Real-time simulation of the grid-interactive current In this section, the real and reactive power output of the
sensor-less MPC inverter with decoupled control of real and inverter are each controlled independently. In order to evaluate
reactive power is performed on Typhoon HIL402 control the performance of the controller, a step change is applied to the
platform. The system specifications are provided in Table II. It
is shown that the control scheme is able to track real and reactive TABLE II: SYSTEM SPECIFICATIONS
Parameter Value
15 15 Inverter DC Side Voltage 300 V
iL iL
10 10 Sampling Time 20 μs
Current (A)

Current (A)

5 5 Filter Inductance 2.4 mH

0 0
Filter Capacitance 15 μF
-5 -5
Grid Voltage 120 VRMS
-10 -10
-15 -15
0.5 0.51 0.52 0.53 0.54 0.55 0.5 0.51 0.52 0.53 0.54 0.55
Time (s) Time (s) 100

(a) (b)
Magnitude (% of Fundamental)

15
15 iL iL iref §
10 10
0.3
Current (A)

5
Current (A)

5
0 0
0.2
-5 -5

-10 -10 0.1

-15 -15
0.5 0.51 0.52 0.53 0.54 0.55 0.5 0.51 0.52 0.53 0.54 0.55
0
Time (s) Time (s) 0 100 200 300 400 500 600 700 800 900 1000

(d) Frequency (Hz)

(c)
Fig. 3. a) Observed inductor current, b) measured inductor current, and c) overlap Fig. 4. FFT analysis of observer current during steady-state condition; real
comparison of observed and measured currents while the power output of the power is set to 1 kW, reactive power is set to 0 VAR, THD is 2.64%
inverter are 1 kW and zero VAR using proposed MPC. d) The generated reference
current.

Authorized licensed use limited to: Carleton University. Downloaded on August 02,2020 at 21:50:54 UTC from IEEE Xplore. Restrictions apply.
 Reactive Power (VAR)
Qref MPC reference reactive power from 0 kVAR to 1 kVAR at instant
1000 0.54 s. As shown by Fig. 5a, the proposed control scheme tracks
the reference reactive power effectively. The same scenario is
500 repeated for changing the real power reference. The real power
reference changes from 1 kW to 3 kW at instant 0.54 s. The
0 controller continues to follow the step change in the real power
0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 accurately as shown in Fig. 5b. Hence, the results show that the
Time (s) control is independent and has fast dynamic behavior.
(a)
C. Comparison with PR-Based Controller
Pref MPC The performance of the proposed controller is compared
Real Power (W)

3000
with conventional PR controller as a benchmark. The PR-based
2000 controller is tuned through numerous trial-and-error attempts for
1000
finding the optimum controller gains. The proportional gain KP
is set to 1 and the resonance gain KR is set to 6000. The
0 performance of the controller is critically dependent on these
0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 gain values. Conversely, the proposed controller in this paper
Time (s) leverages MPC technique, which does not require substantial
(b) tuning efforts and avoids this obstacle. In this comparison
between PR and MPC based control schemes, a step change
Fig. 5. Dynamic behavior of the controller when step changes are applied in
reactive and real power reference from a) 0 kVAR to 1 kVAR at instant 0.54s
from 0 kVAR to 1 kVAR is applied to the reference reactive
b) 1 kW to 3 kW at instant 0.54s. power at instant 0.54 s, which consequently causes a step change

iref PR MPC 50 iref PR MPC

Current (A)
20
Current (A)

0 0

-20 -50
0.5 0.52 0.54 0.56 0.58 0.6 0.5 0.52 0.54 0.56 0.58 0.6
Time (s) Time (s)
(a) (a)
25 30
iref PR
iref PR
20
20
Current (A)
Current (A)

15

10
10
5 ǻt < 0.4 ms ǻt < 0.49 ms

0 0
0.5398 0.5399 0.54 0.5401 0.5402 0.5403 0.5404 0.5405 0.5406 0.5398 0.5399 0.54 0.5401 0.5402 0.5403 0.5404 0.5405 0.5406
Time (s) Time (s)
(b) (b)
25 30
iref MPC
20 iref MPC
20
Current (A)
Current (A)

15

10
10
5 ǻt < 0.1 ms
ǻt < 0.15 ms

0 0
0.5398 0.5399 0.54 0.5401 0.5402 0.5403 0.5404 0.5405 0.5406 0.5398 0.5399 0.54 0.5401 0.5402 0.5403 0.5404 0.5405 0.5406
Time (s) Time (s)
(c) (c)
Fig. 6. Comparison study of PR-controlled current and MPC-controlled Fig. 7. Comparison study of PR-controlled current and MPC-controlled
current, where reactive power is step changed from 0 kVAR to 1 kVAR. The current, where real power is step changed from 1 kW to 3 kW. The current in
current in a) depicts the steady state behavior before and after change, and b) a) depicts the steady state behavior before and after change, and b) depicts the
depicts the enlarged zoom of dynamic response for PR control. c) shows the enlarged zoom of dynamic response. c) shows the enlarged zoom of dynamic
enlarged zoom of dynamic response for MPC. response for MPC.

Authorized licensed use limited to: Carleton University. Downloaded on August 02,2020 at 21:50:54 UTC from IEEE Xplore. Restrictions apply.
in the reference current. The steady state behavior before and Impedance Source Inverter With LVRT Capability," IEEE Access, vol. 7,
after the change is shown in Fig. 6a. The output current in PR pp. 35731-35742, 2019.
[6] S. Jain, M. B. Shadmand, and R. S. Balog, "Decoupled Active and Reactive
and MPC control methods are shown in Fig. 6b and 6c Power Predictive Control for PV Applications Using a Grid-Tied Quasi-Z-
respectively. The results show that the settling time of the PR- Source Inverter," IEEE Journal of Emerging and Selected Topics in Power
controlled current is about 0.4 ms, while the settling time of the Electronics, vol. 6, no. 4, pp. 1769-1782, 2018.
MPC-controlled current is less than 0.1 ms. Furthermore, the [7] S. R. Mohapatra and V. Agarwal, "Model Predictive Controller With
PR controlled current exhibits underdamping with an overshoot Reduced Complexity for Grid-Tied Multilevel Inverters," IEEE
Transactions on Industrial Electronics, vol. 66, no. 11, pp. 8851-8855, 2019.
of 48%, and MPC controlled current shows overdamping [8] S. R. Mohapatra and V. Agarwal, "Model Predictive Control for Flexible
behavior. The same scenario is repeated to step change the real Reduction of Active Power Oscillation in Grid-tied Multilevel Inverters
power reference from 1 kW to 3 kW at instant 0.54s. The steady under Unbalanced and Distorted Microgrid Conditions," in 2019 IEEE
state behavior before and after the change is shown in Fig. 7a. Industry Applications Society Annual Meeting, 2019.
The output current in PR and MPC control methods are shown [9] H. Bai, X. Wang, and F. Blaabjerg, "A Grid-Voltage-Sensorless Resistive-
in Fig. 7b and 7c respectively. The settling time is shown to be Active Power Filter With Series LC-Filter," IEEE Transactions on Power
Electronics, vol. 33, no. 5, pp. 4429-4440, 2018.
about 0.15 ms for the MPC-controlled current, while for the PR- [10] B. Wang, Y. Xu, Z. Shen, J. Zou, C. Li, and H. Liu, "Current Control of
controlled current, it is about 0.49 ms. The PR controlled current Grid-Connected Inverter With LCL Filter Based on Extended-State
exhibits underdamping with an overshoot of 50%. Hence, the Observer Estimations Using Single Sensor and Achieving Improved Robust
PR-based controller presents lower performance, along with the Observation Dynamics," IEEE Transactions on Industrial Electronics, vol.
need for precise tuning of controller gains. However, the 64, no. 7, pp. 5428-5439, 2017.
[11] C. Zheng, T. Dragiþeviü, and F. Blaabjerg, "Current-Sensorless Finite-Set
proposed MPC-based controller does not suffer from gain tuning Model Predictive Control for LC-Filtered Voltage Source Inverters," IEEE
process, and has faster dynamic response to reference changes. Transactions on Power Electronics, vol. 35, no. 1, pp. 1086-1095, 2020.
[12] A. Khan and F. Blaabjerg, "Novel shunt-less filters for grid-connected
V. CONCLUSIONS transformerless photovoltaic applications," in CPE-POWERENG 2018,
2018.
A current observer-based predictive control for single-phase
[13] A. Khan, A. Iqbal, and M. B. Shadmand, "Novel LCL Filter for Non-
grid-interactive inverter has been developed. The proposed Isolated Photovoltaic Inverters with CM Current Trapping Capability for
control method features decoupled power control with fast Weak Grids," in IECON 2018 - 44th Annual Conference of the IEEE
dynamic response. This controller increases reliability by Industrial Electronics Society, 2018.
removing current sensors from the system. The model equations [14] A. Khan and F. Blaabjerg, "Modified transformerless dual buck inverter
for the full state observer were formulated then simulated in real- with improved lifetime for PV applications," in (IRPS 2018), 2018.
[15] A. Khan, S. D’silva, M. Hosseinzadehtaher, M. B. Shadmand, and H. Abu-
time using Typhoon HIL 402 control platform. The measured Rub, "Differential and Common Mode Active Resonance Damping Control
and observed currents were directly compared, and metrics were for Shunt-less LCL Filter Based Grid-Connected PV Inverters," in PECI
captured to show the quality of the observer. The accuracy and 2019, 2019.
quality have not been compromised compared to sensor-based [16] A. Khan, M. B. Shadmand, and H. Abu-Rub, "Active Power Decoupling
methods. The controller does not suffer from substantial gain Control Scheme for Distorted Grids," in PECI 2019, 2019.
[17] M. Mirhosseini, J. Pou, B. Karanayil, and V. G. Agelidis, "Resonant Versus
tuning efforts due to applying MPC technique on the proposed Conventional Controllers in Grid-Connected Photovoltaic Power Plants
controller. Case studies have shown the superiority of the Under Unbalanced Grid Voltages," IEEE Transactions on Sustainable
proposed controller over similar observer-based PR control Energy, vol. 7, no. 3, pp. 1124-1132, 2016.
schemes. [18] R. Teodorescu, F. Blaabjerg, M. Liserre, and P. C. Loh, "Proportional-
resonant controllers and filters for grid-connected voltage-source
ACKNOWLEDGMENT converters," IEE Proceedings - Electric Power Applications, vol. 153, no. 5,
pp. 750-762, 2006.
This material is based upon work supported by the [19] S. Vazquez et al., "Model Predictive Control: A Review of Its Applications
Department of Energy, Office of Energy Efficiency and in Power Electronics," IEEE Industrial Electronics Magazine, vol. 8, no. 1,
Renewable Energy (EERE), Solar Energy Technologies Office, pp. 16-31, 2014.
under Award Number DE-EE0008767. [20] M. Hosseinzadehtaher, A. Khan, M. W. Baker, and M. B. Shadmand,
"Model Predictive Self-healing Control Scheme for Dual Active Bridge
Converter," presented at the SGRE 2019, Qatar, 2019.
REFERENCES [21] Y. Yang, F. Blaabjerg, H. Wang, and M. G. Simões, "Power control
[1] M. Rasheduzzaman, J. A. Mueller, and J. W. Kimball, "An Accurate Small- flexibilities for grid-connected multi-functional photovoltaic inverters," IET
Signal Model of Inverter- Dominated Islanded Microgrids Using $dq$ Renewable Power Generation, vol. 10, no. 4, pp. 504-513, 2016.
Reference Frame," IEEE Journal of Emerging and Selected Topics in Power [22] M. Ciobotaru, R. Teodorescu, and F. Blaabjerg, "A new single-phase PLL
Electronics, vol. 2, no. 4, pp. 1070-1080, 2014. structure based on second order generalized integrator," in 2006 37th IEEE
[2] Y. A. I. Mohamed, M. A. Rahman, and R. Seethapathy, "Robust Line- Power Electronics Specialists Conference, 2006.
Voltage Sensorless Control and Synchronization of LCL -Filtered [23] D. G. Luenberger, "Observing the state of a linear system," IEEE
Distributed Generation Inverters for High Power Quality Grid Connection," transactions on military electronics, vol. 8, no. 2, pp. 74-80, 1964.
IEEE Transactions on Power Electronics, vol. 27, no. 1, pp. 87-98, 2012. [24] L. Wang, Model predictive control system design and implementation using
[3] S. Ziegler, R. C. Woodward, H. H. Iu, and L. J. Borle, "Current Sensing MATLAB®. Springer Science & Business Media, 2009.
Techniques: A Review," IEEE Sensors Journal, vol. 9, no. 4, pp. 354-376, [25] S. Huang, Y. Konishi, Z. Yang, and M. Hsieh, "Observer-Based Capacitor
2009. Current Sensorless Control Applied to a Single-Phase Inverter System With
[4] A. Patel and M. Ferdowsi, "Current Sensing for Automotive Electronics— Seamless Transfer," IEEE Transactions on Power Electronics, vol. 34, no.
A Survey," IEEE Transactions on Vehicular Technology, vol. 58, no. 8, pp. 3, pp. 2819-2828, 2019.
4108-4119, 2009.
[5] M. Easley, S. Jain, M. B. Shadmand, and H. Abu-Rub, "Computationally
Efficient Distributed Predictive Controller for Cascaded Multilevel

Authorized licensed use limited to: Carleton University. Downloaded on August 02,2020 at 21:50:54 UTC from IEEE Xplore. Restrictions apply.