Interconnection Study and Optimization of Grid

Connected Photovoltaic System Using

Electromagnetic Transient Program

Bikrant Poudel Bikram Bhandari Ebrahim Amiri Parviz Rastgoufard

University of New Orleans University of New Orleans University of New Orleans University of New Orleans
New Orleans, USA New Orleans, USA New Orleans, USA New Orleans, USA

Thomas E. Field Roy A. McCanne

Entergy Corporation University of Arkansas
Jackson, USA Fayetteville, USA
2021 IEEE Kansas Power and Energy Conference (KPEC) | 978-1-6654-4119-3/21/$31.00 ©2021 IEEE | DOI: 10.1109/KPEC51835.2021.9446233

Abstract—This paper presents the interconnection study of a [5] that have occurred in the US power grid, the North Amer-
photovoltaic (PV) plant model with a transmission test system ican Electric Reliability Corporation (NERC) has developed a
using an electromagnetic transient program (EMTP). The effects guideline for the interconnection of inverter-based resources
of PV controller gain parameters on the system’s response are in-
vestigated. Then, the PV system control parameters are optimized with the bulk power system [6]. The connection of solar
using Particle Swarm Optimization (PSO). It is observed that the photovoltaic (PV) plants to the existing grid is asynchronous
controller gain parameters of the PV model have a significant through the power electronics interface, and, therefore, the
impact on the PV plant’s response and its interaction with the PV plants are often referred to as Inverter- Based Resources
system, which can be tuned to reduce the negative effects by (IBR) [6]. With the use of power electronics in IBRs, new
implementing optimization techniques.
Index Terms—Photovoltaic, EMTP, Interaction study, Opti-
opportunities for grid control can be explored to help the
mization, Particle swarm optimization. grid during abnormal grid conditions [6]. NERC reliabil-
ity guideline for BPS connected IBR performance provides
I. I NTRODUCTION recommendations for steady-state and dynamic performance
characteristics of IBRs. This guideline also explains different
With the ongoing increase of renewable energy penetration aspects of protective functions to the monitoring capability of
into power grids, the integration study of photovoltaic (PV) IBRs.
plants into the power grid, and the related control strategies, The dynamic instability of the PV system during the fault
has become an important topic [1]. Interconnection studies can can be mitigated by tuning PV plant control parameters [6].
help utilities not only with the optimal design and operation However, manual tuning of PV system control parameters
of the PV, but it also prevents instability in the system. considering multiple fault scenarios is time consuming and
Although the manufacturer-specific models of PVs are de- maybe even impossible. Hence, it is recommended to adjust
sired for accurate interconnection studies, they are typically the control parameters by implementing systematic optimiza-
difficult to procure [1], [2]. Hence, a generic model of the tion algorithms [7].
PV plant is often used in interconnection studies to identify In this paper a generic model for a grid-connected PV plant
potential issues and propose proper countermeasures. Under is integrated with a part of a 115/230kV transmission system
certain system conditions, (e.g., low short circuit current), it (Fig. 1) in EMTP-RV. The protection and control functionality
is suspected that the positive sequence stability programs such of the PV system is designed based on the NERC IBR
as PSS/E and DSA Tools may not provide an accurate response reliability guideline [6]. The analysis includes the dynamics
(e.g., voltage swings, oscillations, etc.) [1], [2]. Therefore, it of the PV system and other components such as generator
is recommended to use 3-phase time domain programs such automatic voltage regulator (AVR). The control parameters
as Electromagnetic Transient Program (EMTP) to investigate of PV system are optimized via Particle Swarm Optimization
the response of the PV plants during system disturbances and (PSO) to get a stable performance during the steady state and
contingencies [1]. fault conditions. This paper is organized such that section II
Over the years, learning from the large scale solar PV describes the test system and the PV plant model,and section
response to disturbances such as the “Blue Cut Fire” [3], the III presents the performance analysis of the PV system and
“Canyon 2 Fire” [4], and the “Angeles Forest Disturbance” its sensitivity to the control parameters. In section III, the

978-1-6654-4119-3/21/$31.00 ©2021 IEEE

Authorized licensed use limited to: The University of Texas Rio Grande Valley. Downloaded on February 03,2025 at 20:24:59 UTC from IEEE Xplore. Restrictions apply.
 





 

%& 




 


  !"  

 

 
  

 
   
   

 
 


'

  %& 
# $ 
# $

#$ #$   



 
%&  

#*#&"+ #,-

) 


 (  (
  

Fig. 1. One line diagram of test system with integrated PV park model.

PSO is adopted to optimize the control parameters of the PV curves based on PRC-024-2 [6], deep a voltage sag detector,
system considering the dynamics of the PV system and the overcurrent protection and a dc resistive chopper control. In
generators, a series of optimization results are presented, and this study, the PV plant is set to operate on Q- control mode
the conclusions are discussed. of operation. Real and reactive power output of the PV plant
is set to 49 MW and -19 Mvar respectively.
II. D ESCRIPTION OF THE T EST S YSTEM
III. PV S YSTEM C ONTROL U NIT
Fig. 1 presents the one-line diagram of the integrated PV
A. Effect of inverter controller gain Kv and PVPC gains Kp
park test system. It consists of 9 buses labeled BUS 1 through
and Ki
BUS 9. The PV plant is connected at BUS 1. There are three
115 kV generators connected to BUS 3, BUS 4, and BUS 7 The response of the the PV park (i.e., real and reactive
respectively. A 230 kV generator is connected to the BUS 5. power) is highly sensitive to the value of the inverter controller
One autotransformer is connected between BUS 5 and BUS gain Kv , and the PVPC gains Kp and Ki . To determine the
6. impact of the controller gains on the system’s response, a
Fig. 2 illustrates the block diagram of the EMTP-RV imple- sensitivity analysis is carried out for different values of Kv,
mentation of the PV plant model. The EMTP-RV PV plant Kp and Ki using using parametric analysis in EMTP-RV. The
model uses the Maximum Power Point Tracking (MPPT) analyses are conducted on the test system of Fig. 1 when a
algorithm to get equivalent circuit parameters for the PV cells 15 cycle 3-phase line to ground fault (with Line 1-6 out) is
[8]. The PV Park Controller (PVPC) block includes a mea- applied at the PV bus (BUS 1).
suring block, an outer voltage (or power factor) control and The range of the controller gain variables are presented in
a slow proportional-integral reactive power control [8]. The Table I.
PVPC adjusts the PV inverter controller’s voltage reference
TABLE I
to deliver the desired reactive power output at the POI [8]. R ANGE OF Kp , Ki , AND Kv .
A measurement block measures the voltages and currents at
Variable Tested Range Stable Range
the POI and calculates the voltage magnitude, active power, Kp 0.1 - 10 1-5
and reactive power [8]. The PVPC can be operated in V- Ki 0.1 - 200 1 - 10
control, Q-control, PF-Control or Q(V) curve control mode. Kv 0.25 - 2 0.3 - 0.5
Shunt AC harmonic filters, including two band-pass filters,
are implemented to filter the AC harmonics. The protection Figs. 3, 4 and 5 illustrate the real and reactive power
unit consists of low voltage and over voltage ride-through profile of the PV plant for different controller gain values.

Authorized licensed use limited to: The University of Texas Rio Grande Valley. Downloaded on February 03,2025 at 20:24:59 UTC from IEEE Xplore. Restrictions apply.
     
 
& "      
     
  "      
  &" "  
! 

 
    



 

 

 ) 


   "
'(
   &

  
   
" 

 * #  
  
(.(.
 , ! "  "
 / ,
  

  " 
 
 ,
"   
    
 
  

   '+  , 

  * #- 

  
 
 



  



 



$%  


 
 


Fig. 2. Block diagram of PV park model.

 #  

As observed, in Fig. 3, and Fig. 4, the response is oscillatory
   !"#


    !"

when Kp = 1, Ki = 5 and Kv = 2 and when Kp = 10, 
$% &' ( $
)  *+,-
* + -* +.-


#$ %& ' #
()  *+,-
* + -* +-

Ki = 1 and Kv = 0.5 respectively. 

     


    

 
 

   !"#



$% &' ( $
)  *+,-
* + -* +.-
    !"

#$ %& ' #
()  *+,-
* + -* +-




     

 
    
















    



(a) (b)

Fig. 5. Simulation with Kp = 1, Ki = 5 and Kv = 0.3 (a) Real Power,


and (b) Reactive Power















   

(a) (b)
   ! "#$

    ! "#$

%& '( ) %
*  +,-.
+ , .+ ,/. %& '( ) %
*+   ,-./
, - /, -/

Fig. 3. Simulation with Kp = 1, Ki = 5 and Kv = 2 (a) Real Power, and

     

 
    

(b) Reactive Power 






   ! "#$

    !"

%& '( ) %
*  +,-.
+ , .
+ ,/. #$ %& ' #
()  *+,-
* + -
* +- 



     


    



     

     
   
 


(a) (b)

 Fig. 6. Simulation with Kp = 1, Ki = 2.5 and Kv = 2 (a) Real Power,
 
and (b) Reactive Power



















   

(a) (b)
to a value between 0.3 and 0.5, the real and reactive power
Fig. 4. Simulation with Kp = 10, Ki = 1 and Kv = 0.5 (a) Real Power, responses do not oscillate after the fault is cleared. Setting the
and (b) Reactive Power
value of Kp and Ki less than 1, and the value of Kv less than
The oscillatory response mainly occurs for the values of Kp 0.3 prolongs the transient period and takes longer duration for
and Ki higher than 5 and 10 respectively. If the value of Kv the reactive power response to settle into the steady state. Fig
is set higher than 0.5, some negative real power is noticed 5 presents a case with Kp , Ki and Kv set to 1, 5, and 0.3
post fault recovery (Fig. 6). If the controller gain of Ki is set respectively, where the PV system becomes stable after the
between 1 and 10, Kp is set between 1 and 5, and Kv is set fault is cleared.

Authorized licensed use limited to: The University of Texas Rio Grande Valley. Downloaded on February 03,2025 at 20:24:59 UTC from IEEE Xplore. Restrictions apply.
 B. Control parameter optimization • To reduce the post fault oscillation.
Although the parametric analysis of Section II.A quantifies • To remove the negative real power output of the PV plant.
• To maintain the reactive power output of the PV plant at
the impact of the controller parameters on the system’s re-
sponse, the manual effort to tune the controller parameters is the desired level.
time consuming. Hence, it is preferred to use a systematic To calculate the fitness function before the fault is applied,
optimization algorithms to tune the value of the controller the reactive power output of the PV system is compared
parameters in a timely manner. The optimization algorithm with the rated value. If the reactive power output is not
can be applied on a single PV plant or multiple interconnected approximately equal to the rated value, a flag (f lagQ ) is raised
PV plants connected to the grid. For this purpose, the PSO (set to 1).After the fault is cleared, the PV system real power
technique is adopted and applied to optimize the PV plant is monitored and, if negative, another flag (f lagP ) is raised.
performance. The system presented in Fig. 1 is reduced to Similarly, after the fault the PV system is expected to start
the system presented in Fig. 7 in which one of the boundary stabilizing. If oscillation is detected at any 0.5 s interval, a flag
equivalence is replaced with a model withgenerator dynamics. (f lagOn ) is raised for each time interval. The real and reactive
This is done in order to study the dynamic response of power oscillations is detected using a root mean square (RMS)
generators and a PV system. energy filter [12] (Fig. 8). The fitness is calculated based on
The optimization process searches for the best solution relation (2).
among the set of available parameters. The optimization
problem can be defined to either minimize or maximize the f itness = f lagQ × 2N + f lagP × 2N −1 + f lagOn × 2N −2 (2)
objective function. Generic form of minimization problem is + f lagO(n−1) × 2N −3 + ...... + f lagO0 × 2N −n
presented in (1) [11].
The flowchart of the optimization approach is presented in
minimize f (x) (1) Fig. 9. In this setup, simulation is initialized in JavaScript. The
script runs a 3-phase time domain transient EMTP-RV simula-
subject to x ∈ Ω
tion, where the oscillation detection, negative power detection
f : n →  is a minimization real-valued function also and reactive power generation detection blocks compute the
known as objective/cost function. x = [x1 , x2 , ..., xn ]T is an flags during the simulation run. Next, detected flags are sent
n-vector of independent variables. Ω is a subset of n known back to the script which calculates the fitness based on (2).
as constraints or feasible set. The PSO then calculates the next set of particles for the next
PSO is a heuristic optimization method based on social simulation. This process continues until the stopping criterion
interaction (behavior of flocks of birds and schools of fish) [9], is fulfilled. It should be noted that for the PV test system
[10]. This population-based heuristic optimization method has operating in Q-control mode, the above mentioned objective
been applied in a number of practical applications, including can be fulfilled by optimizing the PVPC gains Kp , Ki and
electric power systems [10]. converter controller gain Kv . Similarly, for the PV test system
In PSO, the population set (solution set) is called a swarm operating in V-control mode, the PVPC gains Kp , Ki , Kv
and in each iteration of PSO the algorithm value of the entire and converter controller gain Kv , and for the PV test system
swarm is updated. For the application of PSO in this paper, operating in PF-control mode, the PVPC gains Kp , Ki , Kv
the problem dimension (number of particles in a swarm d) and converter controller gain Kv are optimized. The gains’
depends on the number of parameters to be optimized for the lower and upper limit (constraints) are listed in Table II. The
PV plant/s in the system. In PSO, a function is minimized values of velocities c1 and c2 are initialized to 1.5.
over the function n initializing with a randomly generated
n population sets each with d particles. A velocity vector TABLE II
C ONSTRAINTS FOR CONTROL PARAMETERS OPTIMIZATION .
is associated with each set of points. The objective function
(fitness) is calculated at each particle based on the fitness value Controler Parameter Range
the of the new population andthe new velocity of each particle Kp 1-5
Plant controller Ki 1-2
is determined. The best fitness for each particle is tracked Kv 1-5
as and referred to as “pbest”. The overall best run is called Converter control Kv 0.1-1
“gbest”. In PSO, particle interaction is performed by updating
the velocities of each particle based on each particle’s pbest as
well as a global gbest. By doing this, the particles are snapped D. Optimization results
towards their personal best as well as the global best of the The algorithm is set to terminate the simulation if the fitness
entire swarm [9]. The completion basis for PSO is a certain is less than 8 or a maximum number of 100 iterations is
number of iterations or a certain fitness value. reached. Fig. 10 illustrates the working of PSO across multiple
simulations for PV system operating in Q-control mode. The
C. PSO Implementation on the PV system color bar shows the fitness of each simulation at different
PSO implementation to optimize the PV system control values of Kp , Ki and Kv . The optimized parameters are
parameters has the following objectives: reached after 30 iterations, and the best values are listed in

Authorized licensed use limited to: The University of Texas Rio Grande Valley. Downloaded on February 03,2025 at 20:24:59 UTC from IEEE Xplore. Restrictions apply.

'


$
 !



 #$



 %& ""



'  (
  
! 
  
  
!
< 
$      45
! 8;/3   78

+%,- .  

  !
!
"" 

 /
1 6 # 6
/


)#*

( ""
$9:


   4,
!
 



 '$ 0
 %&  

 ' ''( 
   
  $
12&03
  


 
  
$

+%,- .

Fig. 7. Modified study area with PV park and dynamic generator.

TABLE III
B EST SET OF Kp , Ki AND Kv FOR PV SYSTEM OPERATING IN
Q- CONTROL MODE .
Parameter Value
Best converter control Kv 0.3253
Best plant control Kp 1.913
Fig. 8. Oscillation detection using “RMS Energy Filter” [12]. Best plant control Ki 4.475

      
       

 

     

    
 

 







    
    
   

(a) (b)
     
 

 
Fig. 9. PSO implementation for PV control parameter optimization using
 
 

 
EMTP-RV and JavaScript.
 

 

 

 

 

 

 

(c) (d)
 Fig. 11. Simulation result for optimal case with PV system operating in Q-

control mode (a) PV real power output, (a) PV reactive power output, (c)
PV bus (R BUS 1) voltage, and (d) Generator bus (R BUS 7, R BUS 8)

 voltage.





The best values derived from optimization with PV system

operating in V-control mode is listed in Table IV, and the
related PV plant response is presented in Fig. 12.


TABLE IV

  B EST SET OF PVPC Kp , Ki , Kv AND CONVERTER CONTROL Kv FOR PV

SYSTEM OPERATING IN V- CONTROL MODE
Fig. 10. PV system optimization using PSO.
Parameter Value
Best converter control Kv 0.42
Best plant control Kp 2.06
Table III. The profile of the PV real power, PV reactive power, Best plant control Ki 1.91
PV BUS voltage and the generator BUS voltage are captured Best plant control Kv 1.16
for the gbest particle and presented in Fig. 11. It is noted that
the reactive power output is different from the non-optimized
case of Fig. 5 since all ideal voltage sources were replaced The best values derived from optimization with PV system
with two real generators. The generator bus voltage transient operating in PF-control mode is listed in Table V, and and the
can be seen in Fig. 11d. related PV plant response is shown in Fig. 13.

Authorized licensed use limited to: The University of Texas Rio Grande Valley. Downloaded on February 03,2025 at 20:24:59 UTC from IEEE Xplore. Restrictions apply.
              
best result successfully starts the PV system in a stable mode,


and is capable of removing post fault oscillations.

     




    

 


R EFERENCES




[1] S. Zhu, D. Piper, D. Ramasubramanian, R. Quint, A. Isaacs, and
 R. Bauer, “Modeling inverter-based resources in stability studies,” in



    
    
2018 IEEE Power Energy Society General Meeting (PESGM), Aug
    2018, pp. 1–5.
(a) (b) [2] Power and Energy Research Lab (PERL) University of New Orleans,
       
“Solar plant (PV) model in emtp for transmission interconnection
  studies,” Report submitted to Entergy Corporation, August 2020.
  [3] North Amirican Electric Reliability Corporation, “1,200 MW fault



 
induced solar photovoltaic resource interruption disturbance report,”
 
June 2017.



 

[4] Joint NERC and WECC Staff Report, “900 MW fault induced solar
 

 
photovoltaic resource interruption disturbance report,” February 2018.
 
[5] Joint NERC and WECC Staff Report , “April and May 2018 fault

    

    

    induced solar photovoltaic resource interruption disturbances report,”
(c) (d) January 2019.
[6] North Amirican Electric Reliability Corporation, “Reliability
Fig. 12. Simulation result for the optimal case with PV system operating in guideline bps-connected inverter-based resource performance,”
V-control mode (a) PV real power output, (a) PV reactive power output, (c) https://www.nerc.com/comm/OC Reliability Guidelines DL/Inverter-
PV bus (R BUS 1) voltage, and (d) Generator bus (R BUS 7, R BUS 8) Based Resource Performance Guideline, Accessed: 2019-05-20.
voltage. [7] M. A. Attia, “Optimized controllers for enhancing dynamic performance
TABLE V of pv interface system,” Journal of Electrical Systems and Information
B EST SET OF PVPC Kp , Ki AND CONVERTER CONTROL Kv FOR PV Technology, vol. 5, no. 1, pp. 1–10, 2018.
SYSTEM OPERATING IN V- CONTROL MODE [8] H. Gras, H. Ashourian, I. Kocar, U. Karaagac, and J. Mahseredjian,
Parameter Value “Simulation models of PV parks in EMTP-RV,” EMTPWorks 4.1.2 PV
Best converter control Kv 0.3079 park manual, 2019.
Best plant control Kp 1.7117 [9] J. Kennedy, “Swarm intelligence,” in Handbook of nature-inspired and
Best plant control Ki 3.975 innovative computing. Springer, 2006, pp. 187–219.
[10] A. Arzani, P. Arunagirinathan, and G. K. Venayagamoorthy, “Develop-
ment of optimal pi controllers for a grid-tied photovoltaic inverter,” in
2015 IEEE Symposium Series on Computational Intelligence, 2015, pp.

      


      1272–1279.
[11] E. K. Chong and S. H. Zak, An introduction to optimization. John
    


    




Wiley & Sons, 2004.


[12] M. Donnelly, D. Trudnowski, J. Colwell, J. Pierre, and L. Dosiek, “RMS-

energy filter design for real-time oscillation detection,” in 2015 IEEE

Power & Energy Society General Meeting. IEEE, 2015, pp. 1–5.




 



 
 



 
 

   

(a) (b)
       
 

 




 

 







 

 

 

    

    

   

(c) (d)
Fig. 13. Simulation result for the optimal case with PV system operating in
PF-control mode (a) PV real power output, (a) PV reactive power output, (c)
PV bus (R BUS 1) voltage, and (d) Generator bus (R BUS 7, R BUS 8)
voltage.

IV. C ONCLUSION
The PV plant model was integrated with a transmission
system, and the effects of the PV controller gains on the
system’s response were analyzed. It is concluded that the
controller gain parameters of the PV plant model have a
significant impact on the response of the PV plant and its
interaction with the system. The controller parameters were
tuned using a PSO technique for three different PV system
control modes to capture the best response. It is noted that the

Authorized licensed use limited to: The University of Texas Rio Grande Valley. Downloaded on February 03,2025 at 20:24:59 UTC from IEEE Xplore. Restrictions apply.