1. Introduction
The technology development involved in the power system to process the generated and transmitted energy, especially with the large-scale use of static power converters, leads to a more significant concern with the power system stability. Among the problems that arise, rotor angle stability, or transient stability, appears with particular emphasis. The transient stability is defined by the ability of the power system, and the generators that comprise it, to remain in synchronism when subjected to a severe disturbance, such as a short-circuit in the transmission system. The loss of synchronism due to such instability is characterized by an aperiodic drift of rotor angle and speed [
1]. The transient stability is highly dependent on some system characteristics, such as the generator’s loading, fault location, fault clearing time, generator parameters such as inertia and reactances, transmission lines impedance, and system voltages magnitudes [
1]. Therefore, considering these characteristics on which transient stability relies, several control techniques and design methods for power systems have been developed in order to improve their stability.
The use of superconducting materials is one of the most promising technologies for this application. One of the fundamental properties of superconducting materials is to limit the electrical current flow that passes through it once it exceeds a critical threshold, at which point the material starts to present a considerable impedance value in series with the faulted power system. The impedance that appears may have an inductive or resistive characteristic depending on the type of material used, and the transition between states is done in a short time interval [
2]. Several studies propose the use of the superconducting fault current limiter (SFCL) in power systems to improve its transient stability [
3,
4,
5,
6,
7,
8,
9,
10]. In [
3], the use of SFCL for transient stability improvement is proposed, and validated by numerical analysis for symmetrical and asymmetrical faults. Further, reference [
4] shows a power quality improvement in the system with the SFCL by limiting voltage dips in the system during faults. In [
5], the design parameters of SFCLs for optimal contribution to system stability is discussed. References [
6,
7] show the contribution of SFCL to transient stability and its optimal location in a multi-machine system. The use of resistive SFCL to increase the system transient stability and its dynamic response is proposed and demonstrated, by numerical approach in [
9], and experimental results in [
8]. Furthermore, reference [
10] makes a comparative study between the resistive and inductive types of SFCL, showing its contribution to power system transient stability. The authors in references [
4,
10] also concluded that, for a system topology similar to the one proposed in this paper, the resistive device is the best option for the enhancement of transient stability.
In recent years, the exponential increase of photovoltaic plants connected to the system, some of which of large-scale capacity, presents a new challenge for the system stability [
11]. These plants do not have the natural ability to contribute to the system stabilization, like the rotational inertia of synchronous generators, due to its interface with the system through power converters. Hence, a power system with high participation of plants with this characteristic can lead to large-scale instability scenarios, such as blackouts, and potential risks to energy delivery [
12]. Some control strategies for these interface converters, such as the virtual synchronous generator (VSG), have been proposed to make these converters capable of acting in disturbance scenarios to avoid instability. In references [
13,
14,
15], different authors initially presented the basic concept and objectives for the emulation of synchronous machines. The authors in [
16] show a review and a comparison of different VSG implementations and their characteristics. In [
17], a study showing transient stability of VSG is performed using the Lyapunov direct method. The study conducted in [
17] demonstrate the ability of VSG to contribute to the system inertial stability, and also show the VSG parameters’ sensitivity to stability.
The VSG is designed to make the power converter capable of emulating the dynamic behavior of synchronous machines, thus making them contribute to the system stabilization through a feature usually called virtual inertia. Furthermore, due to the nature of its implementation through the use of a machine mathematical model, the VSG implementation is not limited to physical parameters like a real generator. This feature enables the possibility of developing adaptive controls for the VSG, that allow its machine parameters to be optimized during certain situations to instantly improve the converter’s dynamic response in the event of disturbances. Several papers present adaptive controls for the VSG parameters. In [
18,
19], a bang-bang control is implemented to change the VSG inertia constant, in order to damp frequency oscillations during disturbances. Meanwhile, the work in [
20,
21] proposes a damping coefficient and inertia constant adaptation control to further enhance the frequency stability. In [
22,
23] the adaptation is accomplished by a robust control approach. Reference [
22] presents a dual adaptivity of the inertia constant, with frequency and power regulation as objectives, while in [
23] a fuzzy logic is used to change inertia in order to improve frequency stability in a microgrid.
With the increasing number of large-scale non inertial power sources being connected to the grid, and the VSG proving to be a viable alternative to stability, this work’s motivation is to propose an adaptation of the VSG virtual resistance parameter, to improve the transient stability of these non-inertial VSG interface converters. As seen in the papers discussed earlier, most adaptive control strategies proposed in the literature are focused only on the systems’ frequency stability in the presence of small-scale VSG, as in microgrids. In these propositions, only the parameters related to the synchronous generator electromechanical modeling are manipulated, in this case, inertia and damping. The application of adaptive controls that modify the VSG inertia and damping coefficient to improve transient stability should be a viable approach. However, it is a known fact in the power system stability literature that a growth in system inertia and damping eventually leads to diminishing returns in the system critical clearing time (CCT). A saturation effect appears from a certain value of inertia onwards, and a further increase in CCT with the increase of generator inertia becomes impractical [
24]. The novelty of this work is to address the transient stability using an adaptive control for armature resistance of VSG. At this moment, there is no VSG adaptive control in the literature concerning the transient stability of power systems for large-scale non-inertial renewable energy plants. Moreover, no adaptive control that involves the adaptation of the machine electrical windings modelling parameters has been proposed to this day.
In this paper, an adaptive control for VSG-based power converters is proposed based on the performance of resistive SFCLs to improve the power system transient stability. The control is based on the alteration of the VSG armature resistance in the adopted synchronous machine model. The effectiveness of the proposed control on the transient stability of a single machine infinite bus (SMIB) system is verified at first through a theoretical analysis, to determine the evolution of the fault critical clearing angle (CCA) and CCT of the system. The equal-area criterion (EAC) is applied to evaluate the CCA while a numerical solution of the system swing equation is used to determine the CCT. Additionally, a Hardware-in-the-loop (HIL) test bench and a PSCAD simulation are proposed, to determine the consistency of the theoretical results and present additional results for the proposed SMIB system. The results show an increase in the power system CCT that is consistent with an increase in the VSG virtual resistance, which confirms the improvement in power system transient stability with the adoption of the adaptive control proposed.
The structure of this paper is as follows.
Section 2 presents a theoretical approach to the problem.
Section 3 presents the proposed adaptive control and HIL test-bench.
Section 4 presents a discussion about the results obtained and, finally,
Section 5 presents the conclusions.
2. Transient Stability in SMIB System with Proposed Control—Theoretical Analysis
The SMIB system illustrated in
Figure 1 is used to evaluate the effect of the proposed control on a power system in terms of transient stability.
The system parameters are given in
Table 1. In this system, the synchronous generator power is delivered to the infinite bus via two transmission lines whose impedance is
ZL. The generator is connected to the transmission system via a transformer whose impedance is
ZT. The impedance (
ZG) represents the generator transient impedance used in the transient stability analysis.
E′ and
V are the generator internal voltage and the infinite bus voltage, respectively.
The system is subjected to a three-phase ground fault on the transformer bus. The system transient stability without the extra resistance is evaluated using the EAC. For this, the swing equation is given in Equation (1), and the system power-angle relation in Equation (2) [
1]:
where
is the electrical power transmitted between bus 1 and 2,
is the mechanical power in the generator shaft,
is the generator rotor angle,
is the angular speed,
is the inertia constant,
is the damping coefficient,
is the system admittance matrix,
is the first-row and first-column complex term of matrix
,
is the first-row and second-column complex term of matrix
, and
is the angle of the complex term
.
The admittance matrix for the pre-fault and post-fault system
is calculated using Equation (3), and is used in conjunction with Equation (2) to plot the power-angle curve for the pre-fault system in the EAC analysis:
During the fault, the generator electrical power becomes zero due to fault position and characteristic. After the fault clearing, the system topology is not altered and, therefore, the post-fault power-angle curve is the same for the pre-fault system. For this system configuration, the initial δ angle is equal to 0.6467 rad, while the CCA found using EAC is equal to 1.255 rad.
The proposed adaptation for the armature resistance works by adding an extra resistance in the virtual armature circuit of the VSG, as indicated in
Figure 1b. In this scenario, for the pre-fault and post-fault systems, there is no extra resistance added. Therefore, the system admittance matrix is the same as in Equation (3). In contrast, during the fault occurrence, the extra resistance cannot be neglected and is considered in the calculation of the admittance matrix. The power-angle equation for the system during the fault is given by Equation (4) when considering this extra resistance [
4]:
where
e
are the imaginary parts of the generator and transformer impedances.
Due to the presence of a three-phase fault in the transformer bus, the admittance between buses 1 and 2 (
Y12) is nulled, and the second term of Equation (2) is zero. Consequently, the power-angle equation is dependent only on the real part of
Y11, which in turn can no longer be neglected due to the presence of the SFCL resistance.
Figure 2 shows the modification of the power-angle curves for different values of SFCL resistance.
The acceleration areas for CCT in
Figure 2 are defined by the areas between points ADGF (for
= 0.03 p.u.), ACJI (for
= 0.06 p.u.), and ABML (for
= 0.10 p.u.). The deceleration areas are defined by the areas between points FEN (for
= 0.03 p.u.), IHN (for
= 0.06 p.u.), and LKN (for
= 0.10 p.u.). In
Figure 2, it is possible to observe that, with an increase in the resistance added during the fault, there is a gradual decrease in the acceleration area as the machine shaft acceleration is dampened. The added resistance is responsible for partially absorbing the extra energy injected in the generator shaft during the fault. Consequently, the machine rotor acceleration is dampened, and an increase in the CCA occurs. With the increase in CCA, the system remains in synchronism for a longer time, increasing its stability margin.
Figure 3 illustrates this behavior, showing the evolution of CCA for different resistance values (from 0.0 p.u. to 0.16 p.u.).
The value of 0.16 p.u. is adopted as a limit because, at this point, the acceleration area A
1 in the equal-area criterion is null, and CCA value equals the maximum angle for stability at 2.538 rad. The CCT is determined using the numerical solution of Equation (1), applying the fourth-order Runge–Kutta method.
Figure 3 also shows the growth of CCT with the increasing SFCL resistance. It is possible to observe an improvement in system stability margin with the increase in
, as observed for the CCA. For example, through the graph shown in
Figure 3, the potential for increasing the CCT can be observed, since for an SFCL impedance of 0.1 p.u., a 100% increase in CCT is verified.
4. Results and Discussion
The system proposed in the previous section was simulated in a HIL environment, in order to verify the results found in the theoretical analysis made in
Section 2. Thus, the HIL test-bench goal is to check if an increase in simulated CCT follows the adoption of the proposed adaptive control, thus proving its contribution to enhancing the system transient stability. Ergo, as proposed in the theoretical analysis, the CCT of the HIL test bench is checked for different
values. The system is simulated for
values ranging from 0.0 p.u. to 0.16 p.u., with a step of 0.01 p.u. For each
value simulated, the CCT is found through repeated simulations in HIL, adjusting the fault time until the maximum fault time in which the system maintains synchronism. In order to shorten the process of finding the CCT, the theoretical CCT found is used as a starting point, and the bisection method is adopted with an accuracy of 1 ms. Altogether, 10 real-time simulations were performed on average to find the CCT in each scenario, which represents an average of 170 simulations performed for all scenarios. Simulation results were obtained using PSCAD software, to validate the theoretical analysis and to compare with the C-HIL results. The PSCAD simulates both power electronics and the proposed control strategy, while C-HIL has the control embedded in a microcontroller platform. By using an external digital controller, the C-HIL presents a gain in fidelity in relation to the purely simulated approach. The results for the system CCT with the adaptive VSG control is shown in
Figure 9.
A detailed comparison between C-HIL and theoretical CCT is presented in
Table 3, showing the CCT relative error for individual
scenarios and the root mean square error (RMSE) considering all scenarios.
The results revealed an increase in CCT when increasing the virtual resistance added. In particular, it is possible to observe an increase of 470 ms between the CCT for the highest value of
, in real time C-HIL and the VSG, without changing the armature resistance in its model. This event represents a 346% increase in CCT and, therefore, in the power system stability margin. So, the results obtained through the real-time C-HIL corroborate the theoretical and PSCAD simulations analysis carried out, showing a consistent increase in the CCT with the increase in
. The curves shown in
Figure 9 reveal the same shape across the entire range of
. A deviation between theoretical and simulated results (C-HIL and PSCAD) is observed, beginning at
= 0.10 p.u. This could be explained mainly by the fact that, in the theoretical analysis, a simplified model of the machine is used to numerically determine the critical clearing angle and time. The machine model used in the VSG implementation is a high order model that includes both the swing equation and the machine winding modelling. The lack of fidelity in the model used in the theoretical analysis becomes noticeable when near the system stability boundary. However, the relative errors shown in
Table 3 indicate a bounded error along the entire range of
values, reaching a maximum of 16% at the
limit. Further, the small RMSE calculated confirms the proximity between theoretical, PSCAD simulation and HIL CCT curves. Hence, the present results demonstrate that, although a real SFCL is not present in the system, its virtualized behavior in the VSG causes the contribution to the transient stability enhancement to be included in the system operation.
Similar results concerning the increase of CCT could be obtained for an adaptation of the inertia constant. Using the CCT as a figure of merit (FOM), a comparison between both parameter adaptation methods can be presented.
Figure 10 shows the theoretical CCT for an inertia constant
variation between 1.0 p.u. and 20.0 p.u., while not adding any extra resistance in the VSG armature circuit. The CCA does not change, because the inertia variation does not affect the system power-angle curves. Nonetheless, the change in inertia implicates in the swing equation resolution, leading to slower dynamics and a reduced machine acceleration during a fault, therefore increasing CCT. However, as discussed in the Introduction
Section 1, the relation of CCT and inertia has a saturation characteristic, which eventually leads to diminishing returns in CCT with the rise in the inertia constant
value. This saturation characteristics is shown in
Figure 10, beginning at around
= 10.0 p.u. From this point, an increase in inertia leads to diminishing returns in CCT, which leads to a maximum practical CCT between 300 ms and 400 ms. Meanwhile, the maximum theoretical CCT achieved by the proposed resistance adaptive control is 787 ms. Therefore, this result shows an advantage of the control strategy proposed in this paper due to its wider range of CCT.
The VSG behavior for angular speed and rotor angle can be analyzed to confirm the improvement in stability. For this analysis, the fault clearing time is kept constant at 250 ms, and the system is simulated in real-time for different
values.
Figure 11 shows the rotor angular speed for the VSG in this situation, while
Figure 12 shows the behavior of the rotor angle. The fault is applied in 0.5 s and extinguished at 0.75 s.
From the curve in
Figure 11, it is possible to observe that the higher the resistance added into the VSG model during the fault, the lower the maximum speed reached by the machine at the fault clearing time. The acceleration of the VSG virtual rotor is not as steep in scenarios where adaptive control acts by increasing the resistance parameter. In other words, the extra energy injected into the machine’s rotor in the short-circuit is reduced. This result serves to corroborate the increase in CCT shown in
Figure 9, proving the contribution of the proposed control to the system transient stability enhancement.
Figure 11 demonstrates the increment of CCA with the increase of
found through the EAC. The rotor acceleration is reduced, and the angular displacement is lower at the end of the fault. As a result, the system can reach higher levels of angular separation without hitting the critical energy absorption point.
Figure 12 also shows that, for a higher resistance added by the adaptive control, more restrained are the angular oscillations, which is reflected in a dampened active power response after the fault clearance. This drives the system to lower settling times for the post-fault VSG dynamic response. It is worth noting that the base for the angular speed, represented in
Figure 11, is defined by the system nominal frequency in radians per second, in this case 376.99 rad/s. The initial value of rotor angle, shown in
Figure 12, is related to the initial machine angle
for steady state, calculated in
Section 2 by the EAC as 0.6467 rad.
Additionally, the phase plane shown in
Figure 13 can be used to condense the results exhibited in
Figure 11 and
Figure 12. A phase plane is a graphical tool that allows visualization of the system trajectory in fault situations. In cases where the system remains stable after the fault is cleared, it is possible to observe in
Figure 13 that the system trajectory remains bounded and moves towards a new equilibrium point. The trajectory is more restrained at both speed and angle axis for higher values of resistance. However, when the system loses synchronism after the fault (
= 0.05 p.u., for example), the trajectory of the system exits the stability region on an infinite movement. The phase plane is commonly compared to a sphere inside a bowl problem [
1]. An injection of energy in the ball starts its movement towards the bowl edge. When the energy injected is sufficient to make the sphere drop out of the bowl, the system is called unstable.
The parallel could be made to the power system stability by defining the sphere as the generator or the VSG, and the bowl is represented in a two-dimensional system by the rotor speed and angle, like the phase plane. When sufficient energy is added to the generator axis, the system is not able to absorb it, and the generator trajectory represented in the phase plane escapes from the bounded stability region. In the case regarding the adaptive control proposed, the extra virtual resistance is capable of absorbing part of the energy injected in the VSG virtual rotor, restricting its movement inside the stability region.
Figure 14 and
Figure 15 are used to show the system voltage behavior on the transformer low voltage (LV) side, in pre-fault, during-fault and post-fault instants.
Figure 14 shows the system line voltage for a 250 ms fault and
= 0.0 p.u.
Figure 15 shows two scenarios for non-zero
values.
Figure 15a presents a scenario in which the system remains stable for
= 0.06 p.u., while
Figure 15b shows the stable system for
= 0.1 p.u.
It can be observed in
Figure 14 that the voltage collapses shortly after the end of the fault, indicating the system instability for these conditions. For the C-HIL scenario shown in
Figure 14, where
= 0.0 p.u., the CCT defining the stability limit is 191 ms, as presented in
Table 3. In contrast,
Figure 15a,b show a disposition for the VSG voltage to return to nominal values. In particular, for the situation in
Figure 15b where
= 0.1 p.u., the VSG voltage after the fault has a faster settling time, and returns more quickly to a steady-state level. Moreover, the maximum voltage overshoot reached at the time of fault clearance for
= 0.1 p.u. is lower compared to the VSG with
= 0.06 p.u. In the first case, the voltage reaches 504 V when the fault is cleared, while for the second scenario the voltage reaches a maximum of 440 V. Both C-HIL scenarios in
Figure 15 presented post-fault stability, due to their higher CCTs provided by the adaptive control. For the scenario in
Figure 15a, the stability boundary is defined by a CCT of 261 ms. Meanwhile, the scenario shown by
Figure 15b has a stability boundary defined by a CCT of 355 ms. Both results are shown in
Table 3.
Eventually, if a control strategy such as the one proposed in this paper is implemented in a real power system, its interaction with the system protection devices must be carefully addressed. For example, the actuation time of overcurrent relays used to remove a faulted line, which usually consider the CCT for their coordination, must now be aware of the proposed control dynamics. The operation of out-of-step protection functions and distance relays are also of special importance when dealing with transient improving control strategies. On the other hand, an adaptive control that improves CCT in a power system could be used in a preventive scheme, to maintain the fault critical time in a fixed value for different generator loading levels, and system topology during the day.