An Adaptive Emergency Approach for Hybrid Networked Microgrids Resilience

The low inertia of renewable-based distributed energy resources (DERs) renders hybrid networked microgrids (<inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs) dynamically susceptible to transients. Such fragility makes it very difficult for <inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs operators to maintain a reasonable margin for the resilient operation during extreme condition contingencies. This paper presents a three-stage emergency approach to improve resilience of <inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs through maintaining dynamic security. The proposed approach targets preserving the resilient operation of <inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs by preventing unnecessary tripping of the DERs after unintentional islanding incident. To do so, a resilient operation zone (ROZ) is introduced which determines the secure operating zone for <inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs and the limits for implementing the corrective countermeasures for resilience augmentation. The proposed approach is outlined in three stages: First, offline analysis is carried out to model and calculate the ROZ. Second, hybrid <inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs operating point is monitored at the pre-event stage and the calculated ROZ at offline stage is adapted to the operating conditions. The third stage is responsible for real-time evaluation of hybrid <inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs security using the ROZ and implementation of the countermeasures. Comprehensive simulation studies presented in this paper demonstrate effectiveness of the proposed scheme for enhancing resilience of hybrid <inline-formula> <tex-math notation="LaTeX">$\text{N}\mu $ </tex-math></inline-formula>Gs.

entire power system, the power system resilience could be 31 further improved by networking the adjacent individual µGs 32 and establishing networked µGs (NµGs) [5]. A NµG might 33 include only AC or DC µGs; however, AC and DC µGs can 34 be practically networked to configure hybrid AC/DC NµGs 35 which can contribute more to the power system resilience. During extreme condition contingencies, when the upstream 38 power supply is interrupted [6], the NµGs are exposed to 39 an unintentional islanding incident which can challenge the 40 resilient operation of the NµGs. The main reason is the low 41 physical inertia of power-electronic converters which render 42 hybrid NµGs dynamically fragile against incipient transients 43 and makes it very difficult for NµGs operators to maintain 44 a reasonable margin for the resilient operation [7]. Due to 45 such a vulnerability, certain procedures and grid codes have 46 recommended fast DER tripping through the loss of mains 47 Based on the literature review study presented in Section I.A, 93 it can be observed that resilience enhancement through pre-94 serving dynamic security is worthy of study which is not cov-95 ered in the available literature, yet. Such an approach deems 96 vital since in case of dynamic insecurity and losing DERs, 97 black starting under severe weather conditions would be very 98 difficult. To fill this gap, an adaptive emergency approach is 99 presented in this paper which aims at expediting the resilient 100 operation of hybrid NµGs. The main contributions of this 101 paper are: 102 • Developing an analytical method, which is based on 103 region of attraction concept in non-linear control theory, 104 to identify the dynamic security boundaries of a hybrid 105 NµGs.

106
• Proposing a suite of emergency approaches to maintain 107 the DERs in-service before reaching the boundaries of 108 dynamic insecurity. The proposed emergency counter-109 measures adapt to different operation conditions of the 110 hybrid NµGs.

113
• Exploiting the potential of NµGs facilities to prevent 114 the need for black-start and mitigate impacts of extreme 115 events.

117
A. OVERVIEW 118 The chronological outline of the proposed adaptive emer-119 gency approach is depicted in Fig. 2. The main objective 120 is to preserve the resilient operation of NµGs by prevent-121 ing unnecessary tripping of the DERs after an unintentional 122 islanding incident. To do so, a resilient operation zone (ROZ) 123 is introduced and the operation trajectories of the NµGs 124 are preserved within the ROZ through suite of emergency 125 approaches. The ROZ is the locus of all points within state 126 variables plan at which, the NµGs resilient operation is 127 retrieved subsequent to an unintentional islanding incident. 128 In the proposed method, unlike the conventional loss of mains 129 protection schemes, DERs can remain in-service and enhance 130 resilience of the hybrid NµGs as long as the DERs state vari-131 ables lie within the boundaries of the ROZ. Only in case that 132 the boundaries of the ROZ are violated, the corresponding 133 DER is tripped to avoid likely damages.

134
In Fig. 2  where, E f is the voltage set point of IBDER. asymptotically stable equilibrium point is: 224 is a set of points, such that any trajectory originating from 225 x 0 ∈ at time 0 will be attracted to the stable equilibrium 226 point.

227
Theorem 1 (Lyapunov's indirect method) [27]: Let A be a 228 Jacobian matrix of (8) at x * : The x * is an asymptotically stable equilibrium point of 231 (8) if all the eigenvalues associated with A are located on 232 the left-half plane which is denoted as secure equilibrium 233 point (SEP), hereinafter. Likewise, x * is unstable equilibrium 234 point of (8) if A has eigenvalues on the right-half plane 235 which is referred to as unsecure equilibrium points (UEPs), 236 hereinafter. Given the hybrid AC/DC NµGs depicted in Fig. 4, this 239 section represents the dynamics of a hybrid NµGs as the 240 system denoted by (8).

241
The hybrid NµGs in Fig. 4 can be represented by the 242 dynamics of associated center of inertia as: In case the inverter of the battery is operated as the grid 269 forming DERs within an AC µG, the dynamics can also be 270 represented by (6). In case the batteries are operation in grid 271 following mode, they can contribute to the countermeasures 272 by rapidly charging and discharging which is discussed in 273 Table 2.  Hence, δ dc and ω dc in (14) and (15) follow the dynamics 279 represented in (6). In (14)-(16), H dc and D dc are: where, DC is set of DC µGs. In (17) In (25), the negative sign corresponds to x dc(1) * i and positive 307 sign stands for x dc(2) * i . In (26), the index j encompasses both 308 SGBDERs and IBDERs with in an AC µG. The discriminated 309 vectors of equilibrium points for SGBDERs and IBDERs are 310 as (27) and (28), shown at the bottom of the next page.

311
In (27)   In (30) CoI is a positive value; hence, 320 to have the left-half plane: The requirement in (31)  Ms is close to π/2 331 radians representing high X/R ratio of SGBDERs and step-332 up transformers. Hence, δ µ,ac(1) * s is less than and δ µ,ac(2) * s 333 is larger than π/2 radians. In (28), the condition is the same 334 as (25) and δ µ,ac(1) * i is less than π/2 radians. Therefore, based 335 on the requirements stated by at post-islanding condition. Here, the hurricanes are taken 370 into the account and the wind speed at pre-islanding stage is 371 the main monitored factor. The pre-islanding data monitoring 372 is repeated periodically with updating rate of T Update seconds 373 to adapt the emergency countermeasures to any change in 374 NµGs operation condition (blocks #3 and #4). Note that, 375 T Update is directly dependent on the polling and updating rate 376 of the NµGs data acquisition system and can be adjusted 377 based on the characteristic of NµGs data acquisition system. 378 In block #5, a fragility analysis is performed which cor-379 relates the measured wind speed with the fragility curves of 380 NµGs facilities using (32) [28]:  The calculated failure probability in (32) is used to update 394 the dynamic security model, (13), based on Table 1. The 395 updated model augments performance of the proposed emer-396 gency approach by representing more realistic mimic of post-397 islanding condition. In Table 1, the failure for an asset is 398 concluded in case the failure probability calculated in (32) 399 is greater than a pre-defined value, say 70%. This value 400 should be set by the NµGs operator through stablishing a 401  to NµGs islanding incident. In Fig. 1, the P e,N µ is roughly 423 constant during the emergency dynamic security preservation 424 stage; on the contrary, P e,N µ follows the system dynamics.

425
For emergency dynamic security preservation stage in Fig. 1 If the NµGs is importing power from the main grid at pre-440 event condition, r1 is a negative value. Referring to (36) and 441 (37), for r 1 ≤ 0, both ω CoI . 446 Therefore, the two portions of the limit cycle are sufficient 447 to evaluate the hybrid NµGs security which are depicted in 448 Fig. 7.

449
Block #11 in Fig. 6, determines the suite of countermea-450 sure to be used after the unintentional islanding scenario 451 is unfolded. The objective of these countermeasures is to 452 alleviate the consequences of the disturbance before the time 453 that trajectories pass the ROZ boundaries and tripping of all 454 DERs has happen. In case of exporting power to the main grid, the coun-457 termeasure with the priority is to charge the battery storage 458 systems. In case of inadequacy, the next priority is prompt 459 cutting down (not tripping) the outputs of IBDERs (includ-460 ing DERs at DC side) [29]. This can reduce the generation 461 excess and maintain resilient operation without DER tripping. 462 If the amount of IBDER curtailment is not sufficient, some 463 SGBDERs might also be tripped as the second priority. The 464 proposed method to determine sufficiency of a countermea-465 sure for resilient operation is presented in Section II.E. In case 466 of importing power from the grid, the priority is to rapidly 467 discharge the battery storages. In case of insufficiency, the 468 next priority is rapid load shedding is used where the load 469 shedding priority will be defined by NµGs operator.

507
This section examines the proposed scheme on a system 508 depicted in Fig. 10.

509
The system data are available in [4] where µGs1 and 2 are 510 considered as the AC and the rest are DC µGs. The DER 511 installed capacity and the peak load associated with each 512 are reported in Table 3. The studied cases are represented 513 in Table 4. Here, the α in (38) is considered 0.7 which is 514 determined based on the dynamics of the hybrid NµGs under 515 study and simulation studies. The updating rate of NµGs with 516 data acquisition system is assumed to be 1 second. Hence, 517 T Update in Fig. 6 is considered to be 1 second. In this study, 518 the simulations are conducted using the DIgSILENT Power 519 Factory software in a personal computer with Intel Core TM i7 520 CPU @3 GHz and 12 GB RAM.

521
The simulation results for Cases I and II in Table 4 are 522 presented in Figs. 11 and 12. At pre-islanding stage, the 523 NµGs is operation point on associated SEP; hence, the locus 524 of NµGs state variables is on the x N µ (1) * CoI in Fig. 11. Here, the 525 time required to form ROZ at pre-islanding stage is 600 ms 526 and 400 ms for Cases I and II, respectively. Following to 527 an islanding incident, the state variables move toward the 528 boundaries of ROZ within ω N µ CoI region for Case I (Fig. 11(a)) and within ω N µ CoI region for Case II (Fig. 11(b)). This 531 observation is in line with the discussion made in Fig. 7.

532
In Case I, the pre-islanding energy trade between NµGs 533 and the main grid is 5.2 MW (import). Referring to Fig. 11(a), 534 in case the 5.2 MW load is curtailed at T1, which the time 535 NµGs trajectories exceed the boundaries of countermeasures 536 actuation limit, the NµGs trajectories are steered towards 537 the SEP and resilient operation of NµGs can be retrieved. 538 This can also be observed from temporal characteristic of 539 rotor angular velocity at NµGs center of inertia in Fig. 12(a). 540 On the contrary, resilient operation of NµGs is forfeited when 541 the 5.2 MW load curtailment is occurred at T2 in Fig. 11(b), 542 i.e. beyond the ROZ. Here, the pole slip event in Fig. 12(b) 543 yields in insecurity which in turn, results in tripping of all 544 DERs and losing 10 MW load.

545
In Case II, 4.45 MW was exporting at pre-islanding stage 546 where corresponding countermeasure to maintain NµGs 547 resilience is 4.45 MW DER curtailment. In Fig. 11(b), the 548 resilient operation of the NµGs is preserved by curtailing 549 VOLUME 10, 2022 FIGURE 9. Logic of DER curtailment (in case needed) subsequent to unintentional islanding event.     To validate the proposed method adaptability, the simu-574 lation results for different cases are reported in Table 5 are 575 presented in Table 6. Here, the performance of the proposed 576 scheme (PS) is compared with out-of-step (OOS) relay-based 577 [13] and undervoltage (UV) relay-based [14] approaches as 578 the common loss of mains practices. The settings of OOS 579 relay is derived from [30], i.e. DER tripping after one pole slip 580 incident. For UV relay, the settings are 0.8 p.u. with 200 ms 581 delay as recommended by [13]. Note that, the main objective 582 of the proposed method is maintaining the resilience through 583 supply continuity. Hence, the main index used in Table 6 for 584 comparison is the amount of load which curtailed/rescued 585 after unintentional islanding incident.

586
In Table 6, PS rescued considerable amount of load in all 587 listed cases. In cases where the NµGs is importing power 588 from the main grid, i.e. Cases I, III, IV, IX, and X, the imme-589 diate tripping of DERs through the local loss of mains relays 590 are avoided and DERs are maintained in service through the 591 logic depicted in Fig. 9. Doing so, the considerable amount of 592 load, which is equal to the in-service on-site DERs, is rescued. 593 In cases where the NµGs is exporting power to the main grid, 594 i.e. Cases II, V, and VI, 100% of the load is rescued upon 595 unintentional islanding. The main reason is availability of 596 sufficient on-site DERs which are kept in-service through the 597 proposed approach and contributed to resilient operation. The 598 results outlined in Table 4 demonstrate that PS is capable of 599 adapting to different operation condition of NµGs. Unlike the 600 PS, the deployment of OOS and UV schemes in most cases 601 resulted in substantial load curtailment.    In Case VII, where no active power and low reactive power,   Table 6 implies the undesirable impact of the available loss 614 of mains protection schemes on the resilience of the NµGs. 615 Case X in Table 5 expresses a scenario where unintentional 616 islanding is occurred at extreme wind condition. For this case, 617 the wind speed is given 50 m/s. The fragility curves for wind-618 based DERs and overhead lines connecting buses 6, 7 and 619 DC NµGs to the main AC bus are depicted in Fig. 14 [28]. 620 Referring to Fig. 14, 50 m/s wind speed results in failure of 621 DER1 in AC µG1 and also splitting of NµGs from each other 622 due to failure of main interconnecting links. Here unlike other 623 cases, the load curtailment by the PS is 5.5 MW which is 624 more than NµGs power transaction with the main grid at pre-625 islanding stage, 4.7 MW. The main reason is that according 626 to Table 1, the generation of DER1 (0.8 MW) is added on 627 top of the NµGs power transaction with the main grid at pre-628 islanding stage. 629 Fig. 15 depicts the rescued load after unintentional island-630 ing for different DER generation and exchange with main grid 631 conditions. In Fig. 15, negative values for exchange represent 632 the export to the main grid where, the resilient operation of 633 the NµGs is fully attained and all loads within the NµGs is 634 rescued.

635
The amount of rescued load decreases as the DER gener-636 ation decreases which confirm the direct impact of on-site 637 DERs on the resilience of power system. When the exchange 638 with the main grid turns into positive values, the amount 639 of rescued load decreases. The reason is that the positive 640  Table 6 for different network configurations. Here, five 647 configurations are considered for NµGs depicted in Fig. 10.

648
As can be seen, the PS can improve power system resilience 649 even if the µGs are not networked. However in Fig. 16,

698
Future works may consider the scalability issues regarding 699 the type and the number of components of the NµGs and 700 communication standpoints. In particular, the effect of elec-701 trical vehicle, for instance vehicle to grid model, could be 702 investigated. Furthermore, the performance of the proposed 703 approach in real-world applications could be studied through 704 experimental investigation. To this end, the authors aimed 705 at performing experimental validation tests in the reconfig-706 urable distribution grid laboratory of HEIG-VD, in Yverdon-707 les-Bains, Switzerland [31].