Abstract
South Africa first implemented rolling blackouts known as load shedding to protect the integrity and maintain stability of their ailing electrical grid over a decade ago. Since then, load shedding has become a daily occurrence, which South Africa plans to alleviate through the addition of grid-connected renewable generation. Significant renewable energy (RE) additions can however negatively affect a grid, prompting South Africa to develop RE-specific grid code requirements as mitigation. Given this strategy to address the country’s supply challenges using RE, and the still unknown effects of the country’s aggressive RE integration planned, South Africa must perform the necessary simulations tailored to the country’s unique grid conditions, and local RE-specific grid code requirements to ensure a RE-driven solution are indeed viable. Prompted by the lack of a suitably tailorable simulation platform, this paper proceeds to discuss the development of a tailorable grid code-guided renewable power plant (RPP) behavioral studies testbed. The testing methodology involves feeding grid disturbance event data to an RPP modeled after Eskom’s Sere wind farm in South Africa, which is operated in line with local RPP-specific grid code voltage and frequency requirements to assess the testbed’s grid code-guided RPP operating approach. Results show the testbed to effectively distinguish between grid code-specified no-fault, fault ride-through, and trip conditions, operating the simulated RPP accordingly. Consequently, as compared to comparable RE integration simulation platforms, the reviewed testbed has the potential of producing individualized results, owing to its tailorable grid, RPP, and incorporated grid code guiding specifications.
1. Introduction
Since late 2007 Eskom, South Africa’s primary electricity provider faced a challenge where demand started to outweigh supply, forcing Eskom to introduce planned blackouts now known as load shedding to maintain the stability of the country’s electrical grid [1]. Since then, Eskom struggled to keep breakdowns of their aging poorly maintained coal-dominated generation fleet below 9 500 MW, at which point load shedding is initiated. Statistics [2–7] however show conditions to have since deteriorated to the point where load shedding of up to 6 000 MW has become a daily occurrence, as the country’s average energy availability factor reaches an all-time low of 57.8% for the 2022 calendar year. As part of the solution, South Africa’s 2030 Integrated Resource Plan (IRP) [8, 9] reviled a strategy to significantly increase grid-connected renewable generation by 14 725 MW leading up to 2030, representing a 33% increase in the country’s generation capacity [10, 11].
Since the IRP’s original release in 2010, 6 422 MW of the planned renewable generation additions have become operational, of which around 4 100 MW are already grid-connected.
Such renewable generation additions, however, bring with them added challenges to existing electrical grids, which become more prominent as grid-connected renewable energy (RE) increases. The most significant of these are noted by Li [12] to be increased voltage levels and short-circuit currents, as well as a deterioration in reliability of supply and power quality, with the loss of system inertia being a contributing factor. System inertia refers to the spinning reserves often associated with conventional generation and plays a vital role in maintaining system stability. Saha et al. [13] investigate the implications of system inertia loss, noting grid frequency stability to be the most affected, posing significant added challenges to system operators in terms of maintaining system frequency within prescribed limits. The consequences of system inertia loss are corroborated by Mararakanya and Bekker [14], who furthermore note that large-scale RE integration implications can greatly vary between grids, highlighting the importance of considering specific conditions, as well as renewable power plant (RPP)-related requirements of a country or region.
In the case of South Africa, such requirements include RE-specific grid codes developed as mitigation to unfavorable effects associated with high penetration RE integration. To perform effective and accurate RE grid integration simulation studies concerning grid-connected RE’s behavior in South Africa, the simulation testbed will need to incorporate both South Africa’s unique grid characteristics, and grid codes RPPs need to adhere to. To perform such studies, simulation platforms developed by existing literature were considered. Paquin et al. [15] implemented a MATLAB-integrated OPAL-RT real-time simulation model of a 24-bus network connected to a doubly fed induction generator wind farm to perform detailed RE performance studies. Merabet et al. [16] developed a MATLAB-integrated OPAL-RT wind turbine emulating system to simplify generator and grid side control scheme development. Li [12] developed a MATLAB-integrated OPAL-RT digital-analog hybrid real-time simulation model for grid-connected behavioral studies of single- or multigeneration units. Kemal et al. [17] developed a MATLAB-integrated OPAL-RT real-time simulation model to demonstrate the potential of hardware-in-the-loop simulations when studying various aspects of RE integration.
Of the simulation platforms reviewed, few offered the potential of being tailorable, especially in terms of representing the South African grid. Also, none incorporated grid code requirements as a way of governing RPP operation, which needed to study grid-connected RPP behavior when operated in line with local requirements RPPs need to adhere to. As no suitably tailorable simulation platform was available, this study sets out to develop a novel real-time RPP grid-integrated behavioral studies testbed, which is tailorable in terms of grid side, RPP, and grid codes guiding RPP operation.
To achieve real-time simulation, existing literature often implements MATLAB-integrated OPAL-RT models. Assessing the strengths and limitations of this combination through modeling of the WSCC 9 bus, and New England 39 bus system, is Singh et al. [18], note this combination to be ideal for real-time power generation modeling applications. Noureen et al. [19] researched OPAL-RT as a renowned real-time simulator developer, summarizing the strengths and advantages of OPAT-RT’s implementation with MATLAB as familiar front-end software for real-time power modeling. Given the success and accuracy achieved by comparable studies implementing this combination, this study’s testbed will be developed using a MATLAB circuit brought to real-time simulation implementing OPAL-RT’s RT-LAB and OP4510 real-time simulator. The testbed consists of three tailorable sections, namely the grid, RPP, and grid codes guiding RPP operation concerning point of connection (POC) conditions. The focus of this paper specifically falls on the testbed’s active grid code voltage and frequency validation abilities, and testing of these respective testbed sections. This is done by replaying grid data recorded during previous disturbances, allowing POC conditions to enter different regions of RPP operation, from where the RPP’s response in line with grid code requirements is assessed. Results show the grid code validation subsystem designs to effectively track and assess POC conditions, enabling them to operate simulated RPPs according to grid code requirements. This allows tailored results to be produced, as an RPP’s behavior and response are observed relative to replayed real-world conditions, and grid code operating specifications specific to the RPP simulated.
Section II follows, detailing the testbed’s design, focusing on the South African grid code voltage and frequency validation subsystems assessed in this paper. Section III reviews the real-time simulator equipment setup, followed by two datasets simulated in section IV, allowing the respective grid code validation sections’ operation to be assessed. Lastly, conclusions are drawn in section V.
2. Real-Time Simulation Testbed
The tailorable testbed design implements a MATLAB-integrated OPAL-RT circuit, allowing simulations to be performed in real time. Consequently, the design incorporates a Master (SM) and Console (SC) split subsystem design, allowing the main Master subsystem circuit to be simulated using OPAL-RT’s OP4510 real-time simulator, while data imports and circuit measurements are handled by the Console subsystem housed on a host computer. For this study, testbed parameters are limited to that of South Africa, which includes two grid event datasets, an Eskom wind farm representing RPP and the integration of relevant local grid code voltage and frequency requirements. The testbed design is nonetheless tailorable and can be adapted to replay any generic or recorded grid data, represent any type and size of RPP, and incorporate any set of grid code voltage and frequency requirements.
Figure 1 follows, depicting the testbed’s main SM subsystem-housed circuit, followed by an outline of the circuit inputs, sections, and outputs. Considering Figure 1, it should additionally be noted that the purpose of passing inputs through an OpComm block is to establish communication between the physically separated host computer and OP4510 real-time simulator during model execution.
2.1. Inputs
From Figure 1, the “Vnu”, “Vnl”, “Vpu spreadsheet”, and “f spreadsheet” inputs are obtained from the SC subsystem.
2.2. Circuit Sections
From Figure 1, the grid, POC, RPP, voltage validation, and frequency validation circuit sections forming the main circuit of the model can be considered, which is defined as follows.
2.2.1. Grid
Observing the “Grid” section labeled in Figure 1, the first subcomponent is seen as the grid representing source, using “Vpu spreadsheet” and “f spreadsheet” as inputs to replay previously recorded grid events in the simulation environment. The imported spreadsheet data are therefore converted to a three-phase output, concerning base values of 25 kV at 50 Hz selected for this study. Adjacent, connected through Bus 1 is a 25 kV/690 V wye-delta step-down transformer to match POC and RPP voltage. The final component of the grid section is a 100 MW grounded, resistive load connected adjacent to Bus 2.
2.2.2. Point of Connection
Bus 3 labeled as the POC in Figure 1 is the connection point of the RPP to the rest of the grid representing network. This is also the point where complex voltage and current measurements are obtained and used to perform the active grid code voltage and frequency validation assessed in this paper.
2.2.3. Renewable Power Plant
The RPP labeled in Figure 1 may represent any renewable power plant for which grid integration behavioral studies need to be performed. For this study, the RPP is modeled after Eskom’s Sere wind farm in South Africa. The simulated RPP will consequently consist of 46 × 2.3 MW asynchronous wind turbines with an output voltage of 690 V [20, 21]. Wind speed is furthermore assumed constant at 8 m/s, as a means of limiting factors that may affect parameters unrelated to the purpose of this study.
2.2.4. Voltage Validation Subsystem
The voltage validation subsystem was designed to operate simulated RPPs in line with South African RPP grid-code voltage requirements. For this study, based on the simulated Sere representing RPP’s specifications, grid-code voltage ride-through requirements of category C nonsynchronous RPPs are considered in Figure 2.
To relay Figure 2 operating boundaries to a circuit that actively monitors and operates RPPs accordingly, the voltage validation subsystem was split into three sections. The first of these identifies the presence of voltage events, using the process illustrated in Figure 3.
As inputs, POC voltage and Vnu/Vnl boundaries given by the shaded “continuous operating range” of Figure 2 are used. Relational operators then compare the measured POC voltage to Vnu/Vnl boundaries, allowing boundary violations to be identified. If this is the case, a logical operator block passes a “1,” indicating an active voltage event, which is passed on to the adjacent section, for which the operation is given in Figure 4.
The purpose of the operation illustrated in Figure 4 is to measure a voltage event’s event time. This involves inverting the signal received from the previous section which, for an active voltage event, will trigger both a circuit timer and switch. The timer determines the event time by subtracting the time of initiation of the event from the total simulation run time, passing it on to the switch. The triggered switch normally passes a “0” and then outputs the calculated event time which is passed on to the final section for which the operation is illustrated in Figure 5.
The process illustrated in Figure 5, in addition to the event time, also uses the POC voltage and applicable grid-code voltage ride-through boundaries to identify voltage ride-through violations. This is achieved using lookup tables to represent Figure 2 grid code voltage ride-through boundaries. Using the event time obtained from the previous section, the lookup tables then find and return the corresponding instantanious upper and lower voltage ride-through values. These instantaneous upper and lower voltage values are compared to the instantaneous POC voltage, allowing a violation to be identified. Thus, if the POC voltage violates either the instantaneous upper or lower voltage value, a “1” is passed, disconnecting the simulated RPP.
2.2.5. Frequency Validation Subsystem
The frequency validation subsystem was designed to operate simulated RPPs in line with South African RPP grid-code frequency ride-through requirements. This is given in Figure 6 and remains the same for all types and sizes of RPPs in South Africa.
To relay the grid code frequency ride-through requirements of Figure 6 to a circuit that can actively monitor and operate RPPs accordingly, the operation depicted in Figure 7 was applied.
The first section represented in Figure 7 is dedicated to calculating POC frequency using complex POC voltage, as POC frequency could not be measured directly for phasor simulations. A sequence analyzer block converts the measured complex POC voltage in the format given by (1) into a more appropriate magnitude (|u|) angle (θ) format, respectively, implementing (2) and (3), as in [23].
The angle output in degrees is then converted to radians (rad) using a “Gain” block implementing (4) as in [23], before being passed on to a “calculate frequency” section.
This section works by calculating the derivative of the phase angle of the input voltage, concerning a phasor rotating at 50 Hz, determining the desired POC frequency as output [24]. The final Figure 7 section then uses comparetors and time delayes to represent the respective time-sensitive Figure 6 ride-through boundary limits. A passed signal consequently indicates a grid-code frequency ride-through violation, disconnecting the RPP.
2.3. Measurements
To demonstrate and assess the operation of the testbed’s active grid code voltage and frequency validation subsystems, POC voltage and frequency measurements will be obtained for this study, for which the details follow.
2.3.1. Voltage
To represent POC voltage conditions, per unit (pu) POC voltage values are used for analytical purposes. The pu voltage measured at the POC (VPOC(pu)) relates to actual voltage values as given by (5) in [25].where “actual voltage” refers to the voltage at the POC, while the “base voltage” is represented in the three-phase system as per (6), given by [25].
Considering (6), VL represents the line voltage, and Vp represents the phase voltage of the three-phase system. POC voltage measurements are considered significant, as they are used by both voltage and frequency validation subsystems to determine when grid-code violations occur, allowing them to operate simulated RPPs actively in line with grid-code voltage and frequency requirements.
2.3.2. Frequency
As discussed, POC frequency is determined as part of the frequency validation circuit, and not measured at the POC. POC frequency is therefore obtained at that point in the circuit.
3. Equipment Setup
The MATLAB-designed testbed’s integration with OPAL-RT technologies requires it to consist of at least two subsystems, of which the SC subsystem will run on the host computer, while the SM subsystem is executed using the real-time simulator. As these are two physically separated machines, some configuration is required. Following the integration of the MATLAB/Simulink model with OPAL-RT’s RT-LAB software, the equipment was set up in the lab as in Figure 8.
From Figure 8, a TCP/IP cable is seen used to establish the physical, wired connection using the local network between equipment, while communication is setup and configured using the host computer and RT-LAB software. This connection is confirmed using the Windows command prompt to ping the real-time simulator from the host computer, after which the MATLAB model can then be loaded, compiled, and executed in real time.
4. Real-Time Simulation Case Study Results
This section simulates two datasets obtained from previously recorded grid events to demonstrate and assess the active monitoring abilities of the voltage and frequency validation subsystems. Simulation data are stored in a spreadsheet, containing pu voltage (V_feed_in) and frequency (F_feed_in) measurements taken every 1/3 second, and fed to the simulation to recreate conditions using the “grid representing source” discussed in section II.
4.1. Dataset One
Figures 9 and 10 follow, representing the voltage and frequency results obtained during the first 100-second simulation dataset, followed by Table 1 containing the logged events noted during simulation one.
In Table 1, event one records V_POC conditions violating the Vnl = 0.85 boundary of the RPP, thereby entering an LVRT (low voltage ride-through) state. At event two, V_POC conditions remain in the LVRT state, causing the voltage ride-through limits (Vmax/Vmin) to adjust in line with the RPP’s Figure 2 grid code voltage ride-through requirement graph. Event two furthermore shows F_POC conditions to violate the Fnl = 49 Hz boundary, causing the RPP to also enter an LFRT (low-frequency ride-through) state. Event three sees V_POC conditions recover to within normal limits, resetting the Vmax/Vmin response of the voltage validation subsystem. F_POC however remains within the LFRT region, causing Fmin to have adjusted to 47 Hz. By event four, F_POC remains within the LFRT region, while V_POC conditions re-enter an LVRT state. Event five then shows V_POC conditions to violate the LVRT state’s Vmin boundary value, causing the V Violation voltage trip recorded. This, in turn, disconnects the RPP, which remained disconnected for the remainder of the simulation.
4.2. Dataset Two
Figures 11 and 12 follow, representing the voltage and frequency results obtained during the second 134-second dataset simulation, followed by Table 2 containing the logged events noted during simulation two.
In Table 2, event one records F_POC conditions violating the Fnl = 49 Hz boundary, causing the RPP to enter an LFRT state. By event two, the Fmin limit has been adjusted to 47 Hz in line with the RPP’s Figure 6 graph, which F_POC is then in violation of, causing a F Violation trip to be generated. At this stage, the RPP is disconnected and remains disconnected for the remainder of the simulation. A third event then shows a sudden drop in V_POC, causing conditions to enter an LVRT state, while F_POC conditions remain such as to maintain the active trip. By event four, F_POC conditions recover to within LFRT boundaries, releasing the active trip generated by the frequency validation subsystem, although the RPP remains disconnected. Event five shows that V_POC conditions later deteriorate to below the LVRT condition’s Vmin limit, generating a V Violation trip from the voltage validation subsystem, although again not affecting the already disconnected RPP.
Considering Figures 9–12 response graphs of the respective datasets simulated, it can first be observed that the testbed succeeds in tracking POC conditions throughout, shown by the green dotted voltage ride-through boundaries adjusting in line with the respective voltage or frequency limits they represent, when POC conditions operate in the ride-through region between the blue dotted normal variation limits and green dotted trip limits. Noting the nature of these green ride-through boundary adjustments, it is seen to be in line with Figure 2 and 6’s grid code requirement graphs, showing the testbed to effectively represent integrated grid code requirements when POC conditions enter abnormal conditions. The real-time simulation result graphs are then also considered alongside Table 1 and II’s logged simulation data, confirming the adjustment of Vmax, Vmin, Fmax, and Fmin ride-through boundaries, concerning measured V_POC and F_POC values. The table data furthermore confirm that the testbed did not respond to fluctuations within normal boundaries, that the RPP was kept connected in support of the grid whilst within the ride-through regions, and that a trip disconnecting the RPP was generated for conditions outside Vmax, Vmin, Fmax, and Fmin boundaries.
5. Conclusion
In this paper, the active grid code voltage and frequency validation abilities of a tailorable RE grid integration behavioral studies testbed, tailored to South African conditions, were assessed. The testbed design saw the implementation of MATLAB as front-end software for the model and then brought to real-time simulation implementing OPAL-RT’s RT-LAB, and OP4510 real-time simulator. The testbed design spawns from the identified need for a tailorable simulation platform capable of incorporating country- or region-specific parameters relating to grid conditions and RPP-operating specifications when performing RE grid integration behavioral studies. To test the developed testbed’s respective grid code voltage and frequency validation abilities, previously recorded grid data were replayed, causing POC conditions to enter the simulated RPP’s normal, fault ride-through, and trip regions. The testbed’s operation of the simulated RPP concerning POC conditions was assessed, for which the results showed that the testbed effectively tracked POC conditions throughout, operating the RPP in line with South African grid code voltage and frequency requirements incorporated. This involved not responding to normal POC condition fluctuations, riding through minor disturbances in support of the grid, and disconnecting the RPP when POC conditions deteriorate outside ride-through requirement boundaries. The ability to simulate RPPs in line with grid-connected operation requirements then allows studies relating to RPP grid-support-imposed strains and RPP support limitations of a specific implementation to be studied. The testbed’s novel grid code guided approach to RPP operation and tailorable grid, RPP, and RPP operating grid code requirements then set it apart from other less- or nontailorable platforms, as it allows individualized case studies to be performed, producing results of greater relevance to a specific implementation.
Nomenclature
| F: | Frequency, Hz |
| HFRT: | High-frequency ride through |
| HVRT: | High-voltage ride through |
| LFRT: | Low-frequency ride through |
| LVRT: | Low-voltage ride through |
| MATLAB: | Programming platform used for designing and analyzing systems |
| OP4510: | OPAL-RT real-time simulator |
| OPAL-RT: | A leading developer of real-time simulation software and simulators |
| P: | Power, W |
| POC: | Point of connection |
| pu: | Per unit system, quantities are expressed as a ratio relative to a common base |
| rad: | Radians, a unit of angle |
| RE: | Renewable energy |
| RPP: | Renewable power plant |
| TCP/IP: | Connection between computer and simulator |
| V: | Voltage, V |
| : | Absolute value |
| : | Angle, degrees |
| : | pi, 3.14159 |
| abc: | Three-phase value |
| B3: | Bus 3/Point of Connection |
| base: | Base value chosen for pu system |
| in: | Input |
| L: | Line-to-line measurement |
| nu: | Upper grid code boundary value of an RPP’s continuous operating range |
| nl: | Lower grid code boundary value of an RPP’s continuous operating range |
| p: | Phase-to-ground measurement |
| spreadsheet: | Imported value |
Data Availability
The data supporting the current study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
The authors acknowledge the open access funding enabled and organized by SANLiC Gold. The research work of this paper was performed under a Cape Peninsula University of Technology (CPUT) postgraduate bursary, while APCs are covered under the CPUT/SANLIC publishing agreement with Wiley Journals.