Active Power Management of Virtual Power Plant under Penetration of Central Receiver Solar Thermal-Wind Using Butterﬂy Optimization Technique

: Striving for the suppression of greenhouse emissions, the modern power network is facing fundamental changes with the utilization of renewable energies (REs) for the future carbon-free society. The utilization of intermittent renewable-green power needs a better power management system and virtual power plant (VPP) can be a vital candidate that meets this demand. This study investigates a coordinated control grid integrated virtual power plant (VPP) in the presence of Central Receiver Solar Thermal System (CRSTS), Wind Turbine Generator (WTG), and Electric Vehicle (EV). To this end, CRSTS employed with thermal storage acts as a dispatchable renewable generating unit and coordinated control of the system units are achieved using the available control strategy on interconnected microgrids in the modiﬁed form, employing communication time delay. The proposed control strategy employs the proportional-integral (PI) and PI-derivative (PID) controller. Coordinated power control with real-time communication delay in grid integrated VPP in presence of CRSTS, WTG, and EV is a novel approach. Genetic algorithm (GA), Particle Swarm Optimization (PSO), Slap Swarm Algorithm (SSA), and recent Butterﬂy Optimization Algorithm (BOA) are used for tuning the necessary control parameters. The results establish the superiority of the BOA over SSA and PSO in suppressing system frequency deviations and tie line power deviation. The analysis of the dynamic response reveals that the consideration of the communication delay in the system expressively impedes the stable operation of the power system.


Introduction
Recent developments in power systems have led to more complex dynamic networks and have radically revolutionized the very concept of the conventional power system. In particular, the increased deployment of renewable energy (RE) sources and higher propagation of prosumers [1] into electricity networks are encouraged for curbing global warming and reducing fossil fuel use. However, due to certain socio-technical drawbacks and intermittent nature of RE sources, their integration creates issues in power system operation. Maintaining power quality, ensuring reliability and stability become a challenging task. As a solution, several RE units aggregated as a single entity have emerged as a heat and power (CHP), battery, wind, and solar photovoltaic (PV). Although certain literature has reported RE-based VPP control strategies, there is still ongoing research to develop a mature control structure similar to those in conventional power system. It has been stated that VPPs are usually operated in grid-connected mode, as the isolated operation may not be feasible. Controlling DERs in VPP is akin to the secondary control approach of microgrids that takes comparatively higher time for control action as compared to primary control strategy [18]. Therefore, VPP can make use of ICT that causes certain time delay. These communication delays play a significant role in VPP design, operation, and analysis, as longer delays could be detrimental to the power system's health [22]. Therefore, in the context of VPPs, communication delay needs to be set appropriately by finding the time-lag required before regulatory action is introduced to contain the power imbalance and oscillations.
Meanwhile, optimization techniques and control strategies for microgrid/interconnected have also attracted research attention. Several optimization techniques, such as genetic algorithmic technique (GA) [18,23], Particle Swarm Algorithmic technique (PSO) [24], and social spider algorithmic technique (SSA) [25] have been researched to this end. Recently, butterfly optimization algorithm (BOA) was reported to provide a better solution in solving several complex engineering problems compared to other metaheuristic algorithms [26]. Therefore, the performance of the BOA may be explored in providing better tuning of controllers' parameters and hence better frequency regulation of the VPP model.
Recent research on control strategies for interconnected microgrids has been reported in [27][28][29]. In [27], the authors considered the fmincon solver method, a Matlab function and compared the performance with the proportional-integral-derivative (PID) based control strategy. The grasshopper optimization technique-fuzzy PID-based control strategy [28] and comparative performance of social-spider tuned PID control strategy vis-a-vis GA tuned PID control strategy were investigated in [29]. GA, PSO, sine-cosine algorithm, and Grasshopper optimization algorithm tuned PID control strategy were researched in [30]. The communication delay is significant for the realization of such interconnected microgrid system operation, because when this delay exceeds a threshold value, the microgrid may oscillate and may lose stability [31]. Therefore, for better realization of interconnected microgrids, communication delay (time required for receiving and sending information between the control center and the devices) should have been incorporated.
In light of the above, the present work develops a control strategy for VPPs based on control strategy of an interconnected microgrid [31]. The major contributions of this research work are: (a) The development of grid-integrated VPP in the presence of CRSTS, WTG, and EV and implementing it in MATLAB Simulation platform; (b) The application of available control strategy on interconnected microgrid in modified form for active power regulation of VPP model; (c) Considering communication delay in control system design and evaluations; (d) Use of a recent meta-heuristic optimization tool, namely BOA, for optimizing the control parameters and comparing the performance with GA, PSO, and SSA.

DER Modeling in VPP
The schematic structure of CRSTS is displayed in Figure 1. VPP design is illustrated in Figure 2, where DERs such as CRSTS, EV, and WTG are considered as Area-1 and a large conventional non-reheat turbine is considered as Area-2. The system components with respective mathematical models are illustrated below. The nomenclatures with values of some design parameters considered in this work are listed in Table 1.

Central Receiver Solar Thermal System (CRSTS)
CRSTS is one of the most appropriate DER types for VPP due to its potential and thermal storage capability. In particular, thermal energy storage capability responds to power demand fluctuations Sustainability 2020, 12, 6979 4 of 16 while leveling the output variation due to stochastic nature of solar energy. CRSTS's thermal electricity conversion efficiency is about 28% [32]. Worldwide, there are around 35 CRSTS projects in operation [3], with capacities in the order of MWs. Figure 1 shows the schematic diagram of a CRSTS. CRSTS converts sunshine to electricity. It operates by concentrating the solar rays to a central receiver using heliostats. A heat transfer fluid is heated and stored in storage tanks. As per the demand, the stored thermal energy is used to produce steam, which drives the turbine generator to produce clean energy. In the proposed work, the CRSTS is considered as a dispatchable power generator. The transfer function model of CRSTS is [34]: where KRF, KRV, KG, KT represent the gain of the refocus, receiver, governor, and turbine of CRSTS, respectively. The TRF, TRV, TG, TT are the time constants of refocus, receiver, governor, and turbine of CRSTS. CRSTS converts sunshine to electricity. It operates by concentrating the solar rays to a central receiver using heliostats. A heat transfer fluid is heated and stored in storage tanks. As per the demand, the stored thermal energy is used to produce steam, which drives the turbine generator to produce clean energy. In the proposed work, the CRSTS is considered as a dispatchable power generator. The transfer function model of CRSTS is [34]: where KRF, KRV, KG, KT represent the gain of the refocus, receiver, governor, and turbine of CRSTS, respectively. The TRF, TRV, TG, TT are the time constants of refocus, receiver, governor, and turbine of CRSTS. CRSTS converts sunshine to electricity. It operates by concentrating the solar rays to a central receiver using heliostats. A heat transfer fluid is heated and stored in storage tanks. As per the demand, the stored thermal energy is used to produce steam, which drives the turbine generator to produce clean energy. In the proposed work, the CRSTS is considered as a dispatchable power generator. The transfer function model of CRSTS is [34]: where K RF , K RV , K G , K T represent the gain of the refocus, receiver, governor, and turbine of CRSTS, respectively. The T RF , T RV , T G , T T are the time constants of refocus, receiver, governor, and turbine of CRSTS. Synchronizing tie-line co-efficient between area-1 and 2 4.397

Wind Turbine Generator (WTG)
The wind turbine unit is one of the fasted growing renewable energy technologies. Nowadays, wind power has become the most important part of modern energy supply because of its capacity and availability. Energy conversion from wind to electricity is done by a wind turbine system, which includes a rotor with blades, gearbox, electrical generator, and power electronics component. The power output of the wind turbine depends on the speed of wind at that instant. Wind turbines contain several nonlinearities, so while modeling this must be considered. The wind turbine model is simplified into a first order system. The transfer function of WTG based on modeling in [23] is expressed by following equation: where K WTG , T WTG are the gain and time constant of WTG.

Electric Vehicle (EV)
The electric vehicle is the most promising technology that provides s solution to the environmental issues by reducing CO 2 emissions [35,36], and at the same time minimizes the use of oil on a per kilometer basis. EVs can work as an energy storing device to absorb renewable energy and provide support to grid service when it is plugged-in. The bidirectional connection capability of converters in EVs' structure has made them a secure backup for frequency support in power systems. The model which is used in this paper is the first order system. The transfer function model of EV based on modeling in [37] is expressed by the following equation: where T Td is the time constant of EV.

Conventional Synchronous Generator
The non-reheat turbine is one of the proficient and flexible machines which are capable of fulfilling a wide range of steam cycle applications. In this paper, the conventional plant used is not a reheat turbine type and the non-reheat turbine consists of three parts and their transfer function models are: Governor with Dynamics: Turbines with Dynamics: Load and Machine Dynamics: where T RF , T RV represent the time constant of conventional generator and turbine, and the overall conventional system's gain and time constant are depicted by K P and T P .

Information and Communication Technology (ICT)
Future VPPs would connect millions of prosumers, DERs including RE S , energy-storing devices, and a plethora of energy-consuming appliances. It would essentially allow two-way communication for receiving and sending information regarding running status, energy management, analysis, conveying control signals to all the VPP system components. This could only be achieved by profound application of information and communication technologies (ICT). Hence, it is highly likely that ICT would become one of the most essential VPP research topics. ICT such as internet of things or cloud-based infrastructure are the emerging technologies. The ICT network takes certain time for communicating information, e.g., for receiving/sending data from/to the control center and the VPP system components, called communication delay. In this work, the communication delay required for receiving as well as sending information is approximated to a single delay and implemented in the controller forward loop.
The transfer function model of communication delay e −sT D using Pede Ist order approximation [22,38] is represented in the following form: where T D is the delay time.

Formation of the Objective Function
The main objective of the VPP is to maintain a demand-supply balance, irrespective of random variations in RE generation and load. To this end, in association with ICT and appropriate control strategy, VPP minimizes the power imbalance. Therefore, an integral of the squared error (ISE) of the frequency deviation of VPP, tie-line, and main grid frequency deviation is taken as the objective function (J), which is expressed as follows: It is subject to: where, K P , K I , and K D in are the parameters of the PID controller. By using popular optimization techniques GA, PSO, SSA, and recently BOA, the gains of PI and PID controllers are optimized to study the proposed system responses.

Adopted Control Strategy
The VPP must provide control signal to the DERs so as to keep supply-demand balance. To this end, researchers have favored a centralized control concept amongst the available architectures, such as Sustainability 2020, 12, 6979 7 of 16 distributed and hybrid control [18]. In centralized control, the VPP possesses the status of the DERs; accordingly, the central controller dispenses signal corresponding to target power demand from the DERs. The EMS in VPP manages the prosumers' resources such as controllable load, for instance, EV charging stations and DERs. The vehicle to grid mode of EV, WTG, and CTSTS is the DERs to be monitored by the EMS. Controllers employed with the CTSTS and EV do the necessary supervisory control action to virtualize the VPP. EMS interacts and sends signals to the controllers so as to adjust the power generation and maintain a power balance. PI and PID controllers have been employed for the control system. The controller input is the frequency deviation of the system. The GA, PSO, SSA, and BOA are used to design the controller for the model. The results of the algorithms have been compared and analyzed separately after simulation. BOA is discussed in the following section, while PSO and SSA are well documented in the literature [24,25].

Overview of Butterfly Optimization Algorithm (BOA)
In the VPP model, generating units are employed with the controllers (PI/PID). The parameters of these controllers are optimized using GA, PSO, SSA, and the recently developed BOA. GA, PSO, and SSA are detailed in references [23][24][25], respectively, so the authors present only BOA in the present paper.
Recently, in 2018, Arora and Singh developed a Butterfly Optimization algorithm (BOA) by utilizing the food searching strategy and mating behavior of butterflies [26]. The optimization function considers butterflies as search agents [39]. The entire conception of sensing and processing the modality depends on three important factors viz. sensory modality 'c', stimulus intensity 'I', and power exponent 'a'. The sensory modality is fragrance, 'I' is the magnitude of the physical stimulus and the parameter 'a' allows for regular expression, linear response, and response compression in BOA. Using these concepts, in BOA, the fragrance is formulated as a function of the physical intensity of stimulus as follows: where 'f ' is the perceived magnitude of the fragrance, i.e., how much stronger the fragrance is perceived by other butterflies. We can consider 'a' and 'c' in the range [0, 1] for our study. The algorithm has two key steps, viz., global search phase and local search phase. In global search phase, the butterfly takes a step toward the fittest butterfly/solution 'g * ', which is represented by the following equation: where x t i is the solution vector x i for ith butterfly in iteration number 't'. Here, g * represents the current best solution found among all the solutions in current iteration. Fragrance of ith butterfly is represented by f i and r is a random number in [0, 1].
Local search phase can be represented as: where x t j and x t k are jth and kth butterflies from the solution space. If x t j and x t k belong to the same swarm and r is a random number in [0, 1], then Equation (12) becomes a local random walk. The corresponding flow diagram of BOA technique is displayed in Figure 3. The considered system parameters are tabulated in the Appendix A.

Simulated Results and Analysis
In order to investigate the system response, the proposed system shown in Figure 2 was modeled and simulated in MATLAB. System conditions detailed in Table 2 were enforced and three different optimization techniques were utilized to benchmark their performances.

Time Domain Analysis of the VPP Model Employed with PI and PID
The time domain analysis obtained by simulation of the MATLAB Simulink model (Figure 2) is shown in Figures 4-6. Controller parameters in this case were optimized using particle swarm optimization algorithms (PSO). From the responses Δf1, Δf2, and ΔPtie, as presented in Table 3, it was

Simulated Results and Analysis
In order to investigate the system response, the proposed system shown in Figure 2 was modeled and simulated in MATLAB. System conditions detailed in Table 2 were enforced and three different optimization techniques were utilized to benchmark their performances.

Time Domain Analysis of the VPP Model Employed with PI and PID
The time domain analysis obtained by simulation of the MATLAB Simulink model (Figure 2) is shown in Figures 4-6. Controller parameters in this case were optimized using particle swarm optimization algorithms (PSO). From the responses ∆f 1 , ∆f 2 , and ∆P tie , as presented in Table 3, it was observed that the PI controlled system has higher overshoot, under-shoot, and settling time as compared to the PID controlled system. Because of the derivative term in PID controller, it acts faster to the system disturbance in the VPP and the EMS system sets the appropriate coordination between the DERS to cater the demand, which in turn contains frequency fluctuation. observed that the PI controlled system has higher overshoot, under-shoot, and settling time as compared to the PID controlled system. Because of the derivative term in PID controller, it acts faster to the system disturbance in the VPP and the EMS system sets the appropriate coordination between the DERS to cater the demand, which in turn contains frequency fluctuation.     observed that the PI controlled system has higher overshoot, under-shoot, and settling time as compared to the PID controlled system. Because of the derivative term in PID controller, it acts faster to the system disturbance in the VPP and the EMS system sets the appropriate coordination between the DERS to cater the demand, which in turn contains frequency fluctuation.

Time Domain Analysis for Optimization Algorithms GA, PSO, SSA, and BOA
The time domain analyses in Figures 7-9 were obtained by PID controller-based simulation of the proposed MATLAB Simulink VPP model using optimization algorithms PSO, SSA, and BOA. The responses of Δf1, Δf2, and ΔPtie using GA, PSO, SSA, and BOA are illustrated, which show that the response in the case of BAO was better than GA, PSO, and SSA at all the times when changes occurred frequently in the system. It can be clearly seen from Table 4 that in the case of GA, PSO, and SSA, the overshoot, under-shoot, and settling time were more when the system was getting disturbed, but in the case of BOA, it showed better control over system controller parameters.   The responses of ∆f 1 , ∆f 2 , and ∆P tie using GA, PSO, SSA, and BOA are illustrated, which show that the response in the case of BAO was better than GA, PSO, and SSA at all the times when changes occurred frequently in the system. It can be clearly seen from Table 4 that in the case of GA, PSO, and SSA, the overshoot, under-shoot, and settling time were more when the system was getting disturbed, but in the case of BOA, it showed better control over system controller parameters.

Time Domain Analysis for Optimization Algorithms GA, PSO, SSA, and BOA
The time domain analyses in Figures 7-9 were obtained by PID controller-based simulation of the proposed MATLAB Simulink VPP model using optimization algorithms PSO, SSA, and BOA. The responses of Δf1, Δf2, and ΔPtie using GA, PSO, SSA, and BOA are illustrated, which show that the response in the case of BAO was better than GA, PSO, and SSA at all the times when changes occurred frequently in the system. It can be clearly seen from Table 4 that in the case of GA, PSO, and SSA, the overshoot, under-shoot, and settling time were more when the system was getting disturbed, but in the case of BOA, it showed better control over system controller parameters.

Time Domain Analysis for Optimization Algorithms GA, PSO, SSA, and BOA
The time domain analyses in Figures 7-9 were obtained by PID controller-based simulation of the proposed MATLAB Simulink VPP model using optimization algorithms PSO, SSA, and BOA. The responses of Δf1, Δf2, and ΔPtie using GA, PSO, SSA, and BOA are illustrated, which show that the response in the case of BAO was better than GA, PSO, and SSA at all the times when changes occurred frequently in the system. It can be clearly seen from Table 4 that in the case of GA, PSO, and SSA, the overshoot, under-shoot, and settling time were more when the system was getting disturbed, but in the case of BOA, it showed better control over system controller parameters.   This analysis interpreted that BOA-optimized PID controllers give a better result than the GA, PSO, and SSA optimized counterparts and illustrated the superiority of BOA over the GA, PSO, and SSA algorithms. Furthermore, the convergence curves of GA, PSO, SSA, and BOA as shown in Figure  10 also confirm the superiority of BOA over GA, PSO, and SSA.   This analysis interpreted that BOA-optimized PID controllers give a better result than the GA, PSO, and SSA optimized counterparts and illustrated the superiority of BOA over the GA, PSO, and SSA algorithms. Furthermore, the convergence curves of GA, PSO, SSA, and BOA as shown in Figure 10 also confirm the superiority of BOA over GA, PSO, and SSA. This analysis interpreted that BOA-optimized PID controllers give a better result than the GA, PSO, and SSA optimized counterparts and illustrated the superiority of BOA over the GA, PSO, and SSA algorithms. Furthermore, the convergence curves of GA, PSO, SSA, and BOA as shown in Figure  10 also confirm the superiority of BOA over GA, PSO, and SSA.

Time Domain Analysis of the VPP Model with Increasing Communication Delay (T D )
The system stability is significantly influenced by communication delay, T D . If the communication delay in a VPP exceeds a threshold value, it causes oscillations, and in the worst case may lead to instability. This section presents the comparative performance of the proposed MATLAB Simulink VPP model with increasing T D1 . Figures 11-13 represent ∆f 1 , ∆f 2 , and ∆P tie , respectively, for a BOA-optimized PID controller for the proposed VPP model. The analysis of response for ∆f 1 , ∆f 2 , and ∆P tie , in terms of the peak overshoot, peak under-shoot, and settling time reveal that as the time delay was increased from T D1 = 1.2 s to T D1 = 1.4 s stepwise (with a step of 0.1 s) system oscillation was also increased. However, the proposed control scheme regulates the system frequency deviations. The moment T D1 increases to 1.5 s, the system oscillation increased significantly as compared to that of the oscillation occurred at 1.2 s and the system seems to be on the verge of losing stability. Hence, T D1 = 1.4 s could be considered as the threshold value of communication time delay and any value beyond this would impede the stable operation of the VPP. The Table 5 provides a quantitative representation of the increased system oscillations with the increase in the communication time delay in the system.

Time Domain Analysis of the VPP Model with Increasing Communication Delay (TD)
The system stability is significantly influenced by communication delay, TD. If the communication delay in a VPP exceeds a threshold value, it causes oscillations, and in the worst case may lead to instability. This section presents the comparative performance of the proposed MATLAB Simulink VPP model with increasing TD1. Figures 11-13 represent Δf1, Δf2, and ΔPtie, respectively, for a BOA-optimized PID controller for the proposed VPP model. The analysis of response for Δf1, Δf2, and ΔPtie, in terms of the peak overshoot, peak under-shoot, and settling time reveal that as the time delay was increased from TD1 = 1.2 s to TD1 = 1.4 s stepwise (with a step of 0.1 s) system oscillation was also increased. However, the proposed control scheme regulates the system frequency deviations. The moment TD1 increases to 1.5 s, the system oscillation increased significantly as compared to that of the oscillation occurred at 1.2 s and the system seems to be on the verge of losing stability. Hence, TD1 = 1.4 s could be considered as the threshold value of communication time delay and any value beyond this would impede the stable operation of the VPP. The Table 5 provides a quantitative representation of the increased system oscillations with the increase in the communication time delay in the system.

Time Domain Analysis of the VPP Model with Increasing Communication Delay (TD)
The system stability is significantly influenced by communication delay, TD. If the communication delay in a VPP exceeds a threshold value, it causes oscillations, and in the worst case may lead to instability. This section presents the comparative performance of the proposed MATLAB Simulink VPP model with increasing TD1. Figures 11-13 represent Δf1, Δf2, and ΔPtie, respectively, for a BOA-optimized PID controller for the proposed VPP model. The analysis of response for Δf1, Δf2, and ΔPtie, in terms of the peak overshoot, peak under-shoot, and settling time reveal that as the time delay was increased from TD1 = 1.2 s to TD1 = 1.4 s stepwise (with a step of 0.1 s) system oscillation was also increased. However, the proposed control scheme regulates the system frequency deviations. The moment TD1 increases to 1.5 s, the system oscillation increased significantly as compared to that of the oscillation occurred at 1.2 s and the system seems to be on the verge of losing stability. Hence, TD1 = 1.4 s could be considered as the threshold value of communication time delay and any value beyond this would impede the stable operation of the VPP. The Table 5 provides a quantitative representation of the increased system oscillations with the increase in the communication time delay in the system.

Conclusions
Dynamic control that is based on the VPP concept can enhance system dynamics of power systems. This may lead to an increase in the share of renewable energy, while improving the transmission efficiency and reliability at the same time. The present work investigated the capability of using VPP to regulate the grid frequency with high penetration of CRSTS, WTG, and EV. Furthermore, the effect of real-time communication delays in VPP operation were also considered and incorporated in the system design.
The proposed grid-integrated VPP model was developed in the MATLAB Simulation platform and the impact with communication delays on the performance of the frequency regulation of VPP was evaluated. Simulation results show that increased communication delay can significantly degrade the response and even make the system unstable in some cases. Communication time delay of 1.4 s was found to be acceptable and hence considered as the threshold value for the proposed model. Any value beyond this would impede the stable operation of the VPP. On the control strategy, the analysis of the results endorses the superiority of the PID over PI controller and also establishes the dominance of the new optimization tool, i.e., BOA, over GA, PSO, and SSA. Finally, it has been observed that the available control strategy on interconnected microgrid could be employed for

Conclusions
Dynamic control that is based on the VPP concept can enhance system dynamics of power systems. This may lead to an increase in the share of renewable energy, while improving the transmission efficiency and reliability at the same time. The present work investigated the capability of using VPP to regulate the grid frequency with high penetration of CRSTS, WTG, and EV. Furthermore, the effect of real-time communication delays in VPP operation were also considered and incorporated in the system design.
The proposed grid-integrated VPP model was developed in the MATLAB Simulation platform and the impact with communication delays on the performance of the frequency regulation of VPP was evaluated. Simulation results show that increased communication delay can significantly degrade the response and even make the system unstable in some cases. Communication time delay of 1.4 s was found to be acceptable and hence considered as the threshold value for the proposed model. Any value beyond this would impede the stable operation of the VPP. On the control strategy, the analysis of the results endorses the superiority of the PID over PI controller and also establishes the dominance of the new optimization tool, i.e., BOA, over GA, PSO, and SSA. Finally, it has been observed that the available control strategy on interconnected microgrid could be employed for active power regulation of grid-integrated VPP in modified form, incorporating the real-time communication delays. In our future work we shall consider application of other controller, like fractional order PID controller in VPP control strategy. Acknowledgments: Authors would like to thank TEQIP-III NIT Silchar for providing technical support for this work.

Conflicts of Interest:
The authors declare no conflict of interest.

REs
Renewable