2023 International Conference on Power and Renewable Energy Engineering

Crest Factor Oriented Finite Set Model Predictive

Control of the Mesh Microgrid
Miłosz Krysik †∗ , Krzysztof Piotrowski ‡

IHP—Leibniz Institute
for High Performance Microelectronics
Frankfurt (Oder), Germany
† [email protected], ‡ [email protected]
2023 International Conference on Power and Renewable Energy Engineering (PREE) | 979-8-3503-2190-6/23/$31.00 ©2023 IEEE | DOI: 10.1109/PREE57903.2023.10370442

∗ Corresponding author

Abstract—Single-phase voltage source inverters with an LC fil- dead-beat controller are used [3]. Nevertheless, the cascaded
ter are widely used in islanded Microgrids (MGs) for Renewable loops tend to limit bandwidth and complicate the tuning [4].
Energy Sources (RES) or Distributed Energy Resources (DER). In recent years, the rapid development and falling cost of
It is crucial for MGs that are highly penetrated by these energy
sources to maximize the power quality in the presence of the powerful processors have allowed high computational burden
higher order harmonics. This work presents the Finite Control methods to become more popular. One of these methods is
Set Model Predictive Control (FCS-MPC) for a single-phase Model Predictive Control (MPC). In principle, this method
voltage source inverter with an LC filter, applicable in islanded calculates the optimization problem with cost function at each
MGs. It is a strategy that determines the optimal switching sampling interval, across the prediction horizon. Nevertheless,
states within the inverters, operating over a finite set of possible
voltage levels to regulate the power flow and voltage in the whole only the first element in the sequence is executed. In the
microgrid, while considering operational constraints. We propose next iteration, the process is repeated with taking into account
to extend the cost function of this method to include not only new inputs. It can be easily adapted to a variety of systems
the voltage and current factors, but also the Crest Factor (CF), and is easy to understand [5]. MPC is widely used in power
which is calculated as the ratio of the peak value and the root electronics due to its fast dynamic response [6].
mean square of the signal. This factor shows how extreme the
peaks are and therefore indicates the power quality. The model There are plenty of different types of MPC. The most
performance was simulated using MATLAB/Simulink tools in commonly used in inverter control are Finite Control Set
a mesh microgrid with consideration of line impedance effects, MPC (FCS-MPC) and Continuous Control Set MPC (CCS-
where the power and load sharing were successfully validated MPC). The latter requires a modulator to calculate the optimal
with the use of Droop Control. Our approach improves the switching states to the controller [7]. This is not necessary
Total Harmonic Distortion (THD) of both voltage and current,
comparing with the typically used cost functions. when using the FCS-MPC [5]. Another difference is that CCS-
Index Terms—crest factor, mesh microgrid, finite control set MPC works with a fixed frequency, on the contrary to FCS-
model predictive control, primary control, inverter MPC. CCS-MPC may lead to complex algorithms, whereas
FCS-MPC is well-known as an easy and intuitive method [8].
I. I NTRODUCTION Therefore, for our approach, the FCS-MPC was chosen as
The use of Distributed Energy Resources (DER) rises dy- inverter control.
namically due to the energy grid transition triggered by the In principle, FCS-MPC is a finite horizon optimization,
climate change [1]. Single-phase inverters are widely used where the optimal solution is chosen from a set of possible
interfaces between DER and loads, especially in residential switching states. At each sampling time, the cost function is
loads. When DERs are linked to local loads and have the calculated for each possible state and then compared with
potential to function as an island, they form a Microgrid (MG). the reference value [9]. The cost function normally consists
Providing energy from non-dispatchable renewable sources of predicted either voltage or current, depending upon the
might generate lots of power quality issues due to their function of the inverter. Some studies use both variables in the
intermittency [2]. cost function to eliminate energy resonance between passive
The selection of the control method is critical to achieving components [4]. Authors in [4] and [8] used switching factors
appropriate power quality with the use of multiple inverters. to minimize energy losses during switching transitions. The
The most basic approach is to divide the control into two author in [10] used the current limit penalty to safeguard
cascaded loops. The inner loop is the current control loop, the inverter in the face of short circuit faults or further
where the signal is directly delivered to the gate drivers of disruptive current circumstances. Nevertheless, none of the
the transistors based on current measurements and references. above studies performed additional direct efforts to improve
The outer loop is the voltage control loop, where references power quality. Only in [11] the author include directly THD in
for the inner loop are set based on voltage measurements. the cost function but the implementation of THD prediction is
To perform this control, modulator-required methods such complex. However, to the best of the author’s knowledge, Crest
as Proportional–Integral (PI), Proportional-Resonant (PR), or Factor (CF) has not been yet included in the cost function. The

979-8-3503-2190-6/23/$31.00 ©2023 IEEE 174

Authorized licensed use limited to: Hochschule Bochum (FH). Downloaded on February 20,2025 at 12:33:28 UTC from IEEE Xplore. Restrictions apply.
use of CF can simplify the calculation of the cost function in
comparison with [11] while improving the power quality in
the system.
The performance of the method is validated in a mesh
microgrid. The mesh microgrid architecture enables multiple
pathways for energy transfer, allowing power to flow from
any available source to any load within the network. This
redundancy ensures system reliability and enhances resilience, Fig. 1. Single-phase inverter with output LC filter
as power can be rerouted dynamically in case of failures
or disruptions. Traditional radial microgrids, while effective
in certain scenarios, often encounter limitations in terms of The voltage Vab can be calculated according to [13] as:
scalability, reliability, and overall system efficiency [12].
The objectives of the paper are twofold. First is the vali- Vab = VDC S (5)
dation of the system with CF included in the cost function
where VDC is the DC-link voltage and S is the switching
of FCS-MPC. This is validated by analysis dynamic behavior
function that is summarized in Table 1. Switching function S
of the system through load changes and droop control. The
is calculated from:
second is to present the improvement of THD in a steady state
of functioning. The simulations were conducted in a meshed
microgrid with the use of two parallel connected inverters S = S1 S4 − S2 S3 (6)
supplying one load. Discrete-time model of the inverter is obtained from (3) and
This paper is organized as follows. Section II describes the for sampling time Ts can be expressed as:
model of the system, in particular the inverter model, CF, and
mesh MG. Section III presents the algorithm of FCS-MPC x(k + 1) = Aq x(k) + Bq vin (k) + Bdq iout (k) (7)
with the use of CF in the cost function. Section IV presents
where:
the results of the simulation in MATLAB/Simulink software.
The results are the validation of the system work and the Aq = eAT s (8)
improvement in THD decrease is shown. Finally, Section V Z Ts
provides the conclusion and future work. Bq = eAτ Bdτ (9)
0
II. M ODEL OF THE S YSTEM Z Ts
A. Inverter Bdq = eAτ Bd dτ (10)
0
The circuit of the full-bridge single-phase inverter with
These equations are adopted to create the FCS-MPC algorithm.
output LC filter is presented in Fig. 1. The system dynamics
model is well known [13] and can be expressed with the use
of Kirchoff’s law: B. Crest Factor
diL CF shows the ratio between peak (Vpeak ) and effective value
L = vab − vout (1) (VRM S ) and indicates how extreme the peaks are [14]. The
dt
dvout greater the CF, the greater the energy contained in higher
C = iL − iout (2) frequency components and therefore it indicates the purity of
dt
the wave. The CF is the indicator that is used very broadly
where L and C are the inductance and capacitance of the by engineers to analyze the quality of the sine wave. It is,
output filter. Variables iL , vout , and iout are measured and Vab however, a bit forgotten and underestimated in the scientific
is calculated. The equations (1) and (2) can be rewritten as a literature. CF is determined by the following formula:
state space model:
Vpeak
dx CF = (11)
= Ax + Bvin + Bd iout (3) VRM S
dt
where:
   −1
 TABLE I
iL 0 L S WITCHING STATES OF THE INVERTER
x= ,A = 1 ,
vout C 0
S1 S2 S3 S4 S
(4) 1 0 0 1 1
1  
L 0 1 1 0 0 0
B= , and Bd = −1
0 C 0 1 1 0 -1
0 0 1 1 0

175
Authorized licensed use limited to: Hochschule Bochum (FH). Downloaded on February 20,2025 at 12:33:28 UTC from IEEE Xplore. Restrictions apply.
 Start

Measure iL , iout , vout

for k=1:3

Calculate predicted values

iL (k + 1) and vout (k + 1) by (7)

Fig. 2. Mesh grid example

Calculate predicted
The CF value for the pure sinusoidal wave is 1.414 and it can CF(k+1) by using (11)
serve as a reference for the calculations. Any fluctuations from
this value can cause voltage stress and higher THD.

C. Mesh Microgrid Calculate cost function by (12)

The system is validated with different lengths of transmis-
sion lines, and thus with different impedance of these lines. In
Fig. 2 an example of a chunk of the mesh MG is presented. The
considered example shows supplying the load with Inverter 1 N
and Inverter 3. Inverter 2 is inactive in the scenario. The load is n=3? n = n+1
located by the Inverter 1, thus the length of the transmission
line between Inverter 1 and the load can be neglected. The Y
power from Inverter 3, as a result of the redundancy in mesh Apply the most optimal solution
MG, could be transferred in two ways. Either through the line
of impedance Zline13 or through the line of impedance Zline23
and then through Zline12 . The length of these transmission
lines could significantly vary and there can be many routes for Stop
delivering the energy, therefore the system should be resisted
to the length of the transmission line. Fig. 3. Flowchart of the FCS-MPC algorithm
The transmission line was simulated as series inductance
and resistance. The values are given in Table II in mH/km
and Ω/km. The above values used in the simulation were Ji = (i∗L − iL (k + 1))2 (14)
multiplied according to the length of the transmission line.
JCF = (CF ∗ − CF (k + 1))2 (15)
III. P ROPOSED FCS-MPC A LGORITHM
where i∗L , vo∗ , CF ∗ are the reference values for current,
The diagram of the proposed FCS-MPC is presented in voltage, and CF. The predicted values of the filter current
Fig. 4. The behavior of the system is designated by the cost iL (k+1) and the voltage vo (k+1) are calculated from (7). The
function, which is calculated at each sampling time. Possible predicted voltage CF is calculated using (11). CF value of the
predictive values are calculated for all possible states with the voltage is calculated with the use of peak2rms function from
use of (7). There are four possible states, as demonstrated in MATLAB, and it is computed over one period and updated on
Table I, however, two of these sets produce the same output, each sampling time.
so it is possible to decrease the number of states to three. The The reference values are computed with the use of well-
cost function is expressed as: documented in the literature Virtual Impedance-Based Droop
J = λv Jv + λi Ji + λCF JCF (12) Control [10]. Droop control specifies the set points for every
inverter. It simulates the behavior of a synchronous generator
where: [15], namely if active power increase then the frequency
Jv = (vo∗ − vo (k + 1))2 (13) decrease. Similar relations occur between the reactive power

176
Authorized licensed use limited to: Hochschule Bochum (FH). Downloaded on February 20,2025 at 12:33:28 UTC from IEEE Xplore. Restrictions apply.
 two rows present frequency and RMS voltage respectively.
Both quantities are presented in per-unit values. The change
in power sharing did not influence the frequency and only

TABLE II
PARAMETERS OF THE SYSTEM

Grid voltage 24 V / 50 Hz
DC Bus voltage 35 V
Filter inductance 10 mH
Filter capacitance 20 µF
Transmission line inductance 2.1 mH/km
Transmission line resistance 0.5 Ω/km
Load active power 2.88 W
Voltage weighting factor λv 1
Current weighting factor λi 0.3
CF weighting factor λCF 1.2
Fig. 4. Diagram of the system with MPC controller

and the voltage. The values transferred to the droop control

loop are transformed into stationary (alpha-beta) reference
frame with the use of a Second-Order Generalized Integrator
[13]. The virtual impedance loop is developed to improve
the stability of the system. The summarized algorithm after
modelling is presented in Fig. 3.

IV. R ESULTS
MATLAB/Simulink was used for investigating and model-
ing the proposed approach. The system was modeled according
to the smart grid emulator presented in [16]. The emulator was
created to imitate, as much as possible, the real behavior of the
electricity grid. It helps to define conditions and scenarios for
testing, which is hard to produce within a real energy network.
In the emulator, voltages and currents were scaled-down to
allow testing to be carried out safely.
The specification of the system is summarized in Table II. Fig. 5. Power sharing between two inverters
The simulations are presented in three parts. The aim of the
first two is the validation of the basic function of the MG
like the droop control and the load changes, when the CF
is included in the cost function. The second one presents the
comparison of the THD for steady-state work for systems with
and without CF in cost function. The simulations were made
for different lengths of the transmission lines, namely 1, 10,
and 20 km.

A. Evaluation of the power sharing

Two identical inverters are connected as presented in Fig. 2.
The results are presented in Fig. 5. The inverters are controlled
with the use of FCS-MPC with CF. At the beginning, the
load is shared equally between two inverters. Every column
represents a different length of the transmission line. The entire
run takes 2 seconds. At the time t = 1.05s, the power sharing
ratio between inverters is set to 3:1. The time to change the
ratio was chosen so that the transition did not take place in
the favorable case, like when crossing zero axes. The first Fig. 6. Detailed view on voltage and current during power sharing

177
Authorized licensed use limited to: Hochschule Bochum (FH). Downloaded on February 20,2025 at 12:33:28 UTC from IEEE Xplore. Restrictions apply.
the small disturbances in voltage exist. These deviations from C. Waveform quality
reference value can be easily restored by secondary control.
The bottom row presents the active and reactive power. The
To compare the quality of the output waveforms of the
blue line is Inverter 1 and the red is Inverter 2. The system
inverter, the THD was used. The system worked in a steady-
did not consume any reactive power and the step of the active
state condition for 2 seconds. Both voltage and current were
power happens quick and without overshoots. Fig. 6 presents
measured at the point of connection of the load. The results for
the closer look at the moment of the transition. The first
different lengths of transmission lines are presented in Fig. 8.
row presents the voltage in per-unit values and the second
The blue bars represent the system without CF in the cost
row present the current in per-unit values. The green line
function, and the red bars represent the system with CF in
is the output current of Inverter 1 and the red line is the
the cost function. The THD of the system with CF was better
output current of Inverter 2. The blue line is the load current.
in both current and voltage measurements. Independent of the
With longer transmission lines, there is a small difference
length of the transmission line, the THD without CF reaches
between the shared current of the inverters in the period of
around 1.2%, where the cost function with CF has the result of
1:1 ratio. The longer the line, the bigger the difference, but the
an average of 0.8%. The Fig. 9 presented the detailed harmonic
value of the difference is not of concern. During power ratio
distribution after FFT analysis. The frequency plot starts with
changes, only small disturbances in both voltage and current
second harmonics.
are presented, and the transition overall is happening without
notable overshoot in voltage or current.

B. Response to the load changes

The response to load changes is presented in Fig 7. The
setup for this simulation run and the presentation of the results
on a graph is the same as in the previous subsection. At the
time t = 0.3s, another load in parallel is connected and at
the time t = 0.7s is disconnected. Both of the loads have
the same impedance. There are negligibly small fluctuations
in frequency during load changes. Nevertheless, changes in
voltage value can be observed. At the time of connection
with another load in parallel, the undershoot is presented,
and at the time of disconnection the load, the overshoot is
presented. These distortions never exceed allowable magnitude
and duration limits set by the standards [17]. There is a
small mismatch in power-sharing after load changes, especially
with 20-kilometer transmission lines. This, as well as the
disturbances in reactive power are caused by the inductive
component of transmission lines.
Fig. 8. Comparison of THD

Fig. 7. Response to load changes Fig. 9. Detailed look on harmonics

178
Authorized licensed use limited to: Hochschule Bochum (FH). Downloaded on February 20,2025 at 12:33:28 UTC from IEEE Xplore. Restrictions apply.
 V. C ONCLUSIONS AND F UTURE W ORK [9] P. Li, Z. Song, H. Zhang and W. Jin, ”Adaptive Finite Control Set Model
Predictive Control Scheme for Single-phase Inverters with LCL Filter,”
In this paper, FCS-MPC as the inverter control algorithm in 39th Chinese Control Conference (CCC), Shenyang, China, 2020.
with a modified cost function was presented. The system under [10] A. M. Taher, H. M. Hasanien, A. R. Ginidi and A. T. Taha, ”Hierarchical
test presents a mesh MG, that consists of two inverters with Model Predictive Control for Performance Enhancement of Autonomous
Microgrids,” Ain Shams Engineering Journal, vol. 12, no. 2, pp. 1867-
LC filter, load, and transmission lines. CF, as one of the 1881, 2021.
widely used by engineers indicators of waveform quality, was [11] P. Li, R. Li and H. Feng, ”Total Harmonic Distortion Oriented Finite
included in the cost function of FCS-MPC. The simplicity of Control Set Model Predictive Control for Single-Phase Inverters,” Ener-
gies, vol. 11, no. 12, p. 3467, 2018.
calculation and correlation with THD leads to a decrease in [12] W. E. P. Sampath Ediriweera, N. W. A. Lidula, ”Design and protection
the amount of the harmonics in the output waveforms. The of microgrid clusters: A comprehensive review,” AIMS Energy, vol. 10,
system was validated in a mesh microgrid and shows that no. 3, pp. 375-411, 2022.
[13] M. Azab, ”A finite control set model predictive control scheme for
the system can handle basic primary control responsibilities single-phase grid-connected inverters,” Renewable and Sustainable En-
such as changes in load or power references. It is shown that ergy Reviews,, vol. 135, 2021.
the suggested method can meet the basic tasks of primary [14] W. G. Morsi and M. E. El-Hawary, ”Wavelet Packet Transform-Based
Power Quality Indices for Balanced and Unbalanced Three-Phase Sys-
control. The transitions in load changes as well as power tems Under Stationary or Nonstationary Operating Conditions,” IEEE
references were smooth and did not create spikes, nor ex- Transactions on Power Delivery, vol. 24, no. 4, pp. 2300-2310, 2009.
cessive generation of the reactive power. It can be concluded [15] S. Bayhan and H. Abu-Rub, ”Model Predictive Droop Control of
Distributed Generation Inverters in Islanded AC Microgrid,” in 11th
that the single-phase inverter with LC filter based on FCS- IEEE International Conference on Compatibility, Power Electronics and
MPC with CF included in the cost function can output a Power Engineering (CPE-POWERENG), Cadiz, Spain, 2017, 2017.
stable waveform that can track reference precisely and has [16] M. Krysik, K. Piotrowski and K. Turchan, ”Testing Smart Grid Scenarios
with Small Volume Testbed and Flexible Power Inverter,” Energies, vol.
good steady-state and dynamic performance in parallel work 15, no. 2, p. 428, 2022.
of the inverters. Independently of the transmission lines, the [17] ”IEEE Standard for Interconnection and Interoperability of Distributed
presented approach is superior to the traditional FCS-MPC. Energy Resources with Associated Electric Power Systems Interfaces,”
in IEEE Std 1547-2018 (Revision of IEEE Std 1547-2003), pp.1-138,
In the future, the algorithm will be implemented in the 2018,
DSP microcontroller. The system will be tested with nonlinear
loads, and more primary control features like Low Voltage
Ride Through or grid synchronization will be tested.
ACKNOWLEDGMENT
This work was supported by the European Union ebalance-
plus project under the H2020 grant no. 864283. The funding
institutions had no role in the design of the study, the collec-
tion, analyses, or interpretation of data, in the writing of the
manuscript, or in the decision to publish the results.
R EFERENCES
[1] C. Inês, P. L. Guilherme, M. G. Esther, G. Swantje, H. Stephen and H.
Lars, ”Regulatory challenges and opportunities for collective renewable
energy prosumers in the EU”, Energy Policy, Vol. 138, 2020.
[2] S. Parhizi, H. Lotfi, A. Khodaei and S. Bahramirad, ”State of the Art in
Research on Microgrids: A Review,” IEEE Access, vol. 3, pp. 890-925,
2015.
[3] Q. Liu, T. Caldognetto and S. Buso, ”Review and Comparison of Grid-
Tied Inverter Controllers in Microgrids,” IEEE Transactions on Power
Electronics, vol. 35, no. 7, pp. 7624-7639, 2020.
[4] Y. Li and Z. Zhang, ”FCS-MPC Based Primary Control With Improved
Performance for Islanded AC Microgrids,” in 11th Power Electronics,
Drive Systems, and Technologies Conference (PEDSTC), 2020.
[5] C. Liu, L. Lei, Q. Chen, L. Zhang and S. Quan, ”Model predictive
control of single phase grid-connected inverter with LC filter,” 2017
32nd Youth Academic Annual Conference of Chinese Association of
Automation (YAC), Hefei, China, 2017, pp. 115-119
[6] S. Vazquez, J. I. Leon, L. G. Franquelo, J. Rodriguez, H. A. Young,
A. Marquez and P. Zanchetta, ”Model Predictive Control: A Review
of Its Applications in Power Electronics,” IEEE Industrial Electronics
Magazine, vol. 8, no. 1, pp. 16-31, 2014.
[7] M. M. Aghdam, L. Li and J. Zhu, ”A Model Predictive Control of
Parallel Inverters for Distributed Generations in Microgrids,” in IEEE
International Conference on Power System Technology (POWERCON),
Wollongong, NSW, Australia, 2016.
[8] Z. Zhang, O. Babayomi, T. Dragicevic, R. Heydari, C. Garcia, J.
Rodriguez and R. Kennel, ”Advances and Opportunities in the Model
Predictive Control of Microgrids: Part I–Primary Layer,” International
Journal of Electrical Power & Energy Systems, vol. 34, 2022.

179
Authorized licensed use limited to: Hochschule Bochum (FH). Downloaded on February 20,2025 at 12:33:28 UTC from IEEE Xplore. Restrictions apply.