Optimal Coordination of Protection Devices in Distribution Networks With Distributed Energy Resources and Microgrids

Microgrids are promising to enhance power distribution systems’ efficiency, quality, sustainability, and reliability. However, microgrids operation can impose several challenges to traditional protection schemes, like changes in the power flow direction and an increase in short-circuit currents. Microgrids can include several distributed generation technologies with different behaviours during short-circuit conditions, requiring additional protection schemes and devices. In this way, the optimized coordination of reclosers and fuses in distribution networks with directional overcurrent relays, which operate as interconnection devices, might overcome many imposed protection challenges. Regarding different generation technologies, voltage-restrained overcurrent relays and frequency relays are presented as local microgrid protection for rotative and inverter-based distributed generators, respectively. The optimized coordination of these protection devices maximizes microgrid benefits and minimizes operation drawbacks by reducing interruptions impacts and energy not supplied to consumers. This work proposes, thus, a mathematical model for the optimal coordination of protection devices in distribution networks with distributed energy resources operating in grid-connected and islanded modes. The minimization technique of operating times using an elitist genetic algorithm with variable crossover and mutation processes is proposed, as well. The results show adequate coordination using passive and low-cost protection devices.

short-circuit conditions within microgrids. Protection devices 69 from distribution networks and points of common coupling 70 (PCC) must coordinate simultaneously with microgrids' local 71 protection. Although the practical application of frequency 72 relays is common in inverter-based DGs (IBDGs), this tech-73 nology is rarely present in literature papers on the optimal 74 coordination problem, especially in obtaining the coordi-75 nation of frequency relays with OCRs. In DGs protection 76 schemes, most energy utilities recommend frequency relays 77 and voltage-restrained OCRs, ANSI 81 and 51V. Therefore, 78 51V relays can be an attractive solution to protect rotating 79 DGs, while frequency relays protect IBDGs to avoid mis-80 coordination regarding their low contribution to short-circuit 81 currents. 82 This work proposes, hence, a mathematical model to solve 83 the optimal coordination problem of protection devices in 84 distribution networks with microgrids, including renewable 85 distributed generation and energy storage systems. It consid-86 ers the minimization of operating times of reclosers and fuses 87 present in the distribution network, as well as the microgrid 88 protection devices. These devices comprise DOCRs present 89 in the PCC for each microgrid and the frequency relays 90 or voltage-restrained OCRs installed at the point of DER 91 connection (PoC). 92 The coordination of protection devices is a mixed inte-93 ger nonlinear combinatorial optimization problem. Most 94 approaches in the specialized literature solve this problem 95 using metaheuristics because such a technique generally has 96 better computational tractability than mixed integer nonlin-97 ear programming models, which also have no guarantee of 98 finding the optimal solution. Some works have proposed a 99 linearization, but this strategy simplifies the problem and 100 frequently does not include different time-current curves 101 or intervals for the pickup current. Among the approaches 102 using metaheuristics, particle swarm optimization (PSO) and 103 genetic algorithms (GA) are the most common. However, the 104 classic PSO is inadequate for problems including discrete and 105 continuous values, requiring additional procedures to yield 106 better performance. Thus, this work proposes a specialized 107 GA with elitism, in addition to variable crossover and muta-108 tion processes. 109 In contrast to other works, this paper addresses the coordi-110 nation of traditional protection devices simultaneously with 111 the protection devices from the PCC and local protections 112 from the PoC. The proposed technique allows the protection 113 system coordination for distribution networks with DGs from 114 different technologies and also considers the islanded opera-115 tion of these sources with part of distribution network loads 116 operating as a microgrid.

117
The main contribution of this work includes the develop-118 ment of a methodology for simultaneous coordination of the 119 protection system in distribution networks with microgrids' 120 protection through passive and low-cost protection devices. 121 The recommendation of distribution companies is followed 122 by adding the ANSI 81 and 51V devices into the mathemati-123 cal formulation of the protection system coordination. 124 VOLUME 10, 2022

125
Traditional protection schemes in distribution networks,  Fuses' behaviour is calculated using linear interpolation 152 with samples provided by the manufacturer [22].

153
IIDs are also based on (1) for permanent faults and 50TD 154 for temporary ones. Moreover, these devices include a direc-155 tional unit to allow their operation in only one direction, that 156 is, for faults external to the microgrid.

157
The operating times of 51V relays are also based on (1), but  the PoC in pu, as in (2).
Frequency relays from microgrids' PoCs with IBDGs con-162 sider the frequency limits defined by IEEE 1547 [23]. The fre-163 quency estimation is based on the DG power change between 164 normal and short-circuit conditions and the system's inertia. 165 During a short-circuit event, the power demand suddenly 166 increases, leading to an increment in the mechanic power due 167 to the machines' control response. Since DGs controller tends 168 to reduce the difference between primary and electric power, 169 the greatest unbalance occurs at the beginning of the short-170 circuit event, i.e., the difference between pre-and post-fault 171 generated power. In (3), P is the power unbalance, H sys 172 is the system's inertia, and f is the frequency deviation 173 in the time domain, [24], [25]. The frequency deviation rate 174 performs a similar behaviour, starting with the highest value 175 and decreasing over time.
Among DERs, wind turbines have a significant amount of 178 kinetic energy in their blades, which is essential to consider 179 in the system's inertia estimation in addition to conventional 180 machines [25]. Biomass generators can be represented as 181 dispatchable DG, being considered a conventional machine. 182 PV units have almost no contribution to the system's inertia. 183 The sum of total inertia constants due to conventional and 184 wind turbines, H C and H W , respectively, is shown in (4), 185 where Nc and Nw are the amounts of each turbine.

186
Thus, the aforementioned features allow the frequency 188 estimation and, consequently, the necessary information for 189 coordinating frequency relays with microgrids' IIDs. Oper-190 ating times of ANSI 81 relays are time defined and depend 191 on the frequency deviation level, presenting four conditions, 192 as described in Table 2.

193
The proposed protection and control scheme can also 194 be extended to the coordination and selectivity of protec-195 tion devices inside microgrids during the islanded operation 196 mode. However, this work focuses on the grid-connected 197 mode assuming the entire distribution network topology. The The minimum pickup current of reclosers is set at dif- The coordination sensibility between distribution network 244 reclosers is ensured using faults I 2∅ and I 1∅ 40 for the phase units 245 and I 1∅ 0z and I 1∅ 40z for neutral units. In Fig. 1 (c), there is an DOCRs have a smaller interval for pickup currents than 252 distribution network reclosers because lower magnitude cur-253 rents flow through the PCC branch, depending on the micro-254 grid capacity. At the same time, the upper limit must be 255 small since IBDGs also have a small short-circuit current 256 contribution during a fault condition [21].

257
Coordination and selectivity between DOCRs and 258 reclosers consider fault currents I 2∅ and single-phase fault 259 currents with 5 impedance, I 1∅ 5 , for phase units. Neu-260 tral units use the zero sequence short-circuit currents I 1∅ 0z 261 and I 1∅ 5z . Users can set longer ranges, but the ride-through 262 requirements must be taken into account. The same faults are 263 considered for selectivity between 51V relays and DOCRs. 264 Fig. 1 (d) shows an example system with a microgrid, 265 MG1, between reclosers Rj and Rj + 1. For a fault in Rj 266 protected feeder section, we assume that Dk must trip faster 267 than Rj for permanent (10) and temporary faults (11). In Fig. 1 268 (e), the example system is a microgrid with a fault in the pro-269 tection zone of the downstream recloser Rk. During perma-270 nent or temporary faults in Rk protected section, Dj must trip 271 slower than Rk, avoiding unnecessary MG1 disconnection, 272 as respectively given in (12) and (13). Coordination times 273 t 51−D coord , t D−51 coord , t 50−D coord , and t D−50 coord ensure the coordination and 274 selectivity between such devices.
Frequency relays must trip if the frequency drift surpasses 286 some intervals established by IEEE 1547, as presented in 287 Table 2. Each frequency interval has a clearing time limit, 288 depending on the variation level. For a fault outside the 289 microgrid, as shown in Fig. 1 (f), the IID should trip first than 290 the frequency relay (15). Some particular cases require the 291 instantaneous operation of the frequency relay, making coor-292 dination impossible. The proposed method does not perform 293 the coordination in such cases.

296
The proposed penalized objective function (POF) in (16)     as selection, crossover, and mutation. The selection stage 345 randomly chooses two pairs of individuals from a population 346 and compares their quality (POF value). The best individual 347 from each pair goes through the crossover process.

348
In the crossover stage, genetic material is exchanged 349 between selected individuals. Genes are randomly mixed, 350 creating a new pair of individuals that will compose the 351 new population. The crossover process is variable, randomly 352 modifying the individual genes [28]. The crossover rate, ρ c , 353 can vary according to (27). Such an approach prevents explo-354 ration restricted only to local solutions. The number of similar 355 solutions is represented by SI i in the i-th generation, i.e., 356 similar solutions concerning other population individuals, η p . 357 The adjustment factor k c min defines a minimum crossover rate 358 to the GA process, while k c max is the maximum. Thus, the 359 crossover process starts at high rates and decreases as the pop-360 ulation loses its diversity. Before including the crossover indi-361 viduals in the new population, the mutation process begins. 362 The population mutation rate, ρ m , can also vary according 364 to the same concept, as is given in (27). Unlike the crossover 365 process, the mutation rate increases as similar individuals in 366 the population increase. In this case, the factors k m min and k m max 367 exchange their positions and the superscript m is employed, 368 representing the mutation parameters (28).
The elitism technique allows a more efficient exchange of 371 genetic material between population individuals and is fre-372 quently applied in the specialized literature [28]. Therefore, 373 such a technique is also applied in the proposed methodology. 374 Fig. 2 shows, from (a) to (j), the genes considered on each 375 chromosome as a row of numeric values that represent adjust-376 ment parameters of protection devices. Constraints of protec-377 tion devices' parameters involve:   The short-circuit current is calculated by multiplying the  Fig. 3 shows a 135-bus unbalanced distribution system 415 employed to evaluate the proposed methodology. This net-416 work has 13.8kV and 7.065 MVA [29]. The protection sys-417 tem without microgrids consists of four reclosers and eight 418 fuses. Five microgrids (MG) are then installed in the 135-419 bus system, where MG1 and MG2 are supplied by rotating 420 DGs. MG3 and MG4 are supplied by photovoltaic panels 421 with energy storage systems, while a full-converter wind DG 422 supplies MG5 also with energy storage. Microgrids have the 423 same nominal capacity, and their total power represents 30% 424 of the total system's load demand, with a power factor of 0.92. 425 The proposed method is implemented in C++ general 426 programming language due to its speed and computational 427 efficiency. Tests were performed on a personal computer with 428 Intel(R) Core(TM) i7-7700, 3.60 GHz, and 16GB of RAM.    when β is equal to or higher than γ . Fig. 4 (e) shows that 463 using β and γ equal to 60 and 90 provides good results for all 464 values of α. Therefore, the most suitable values are 3, 60, and 465 90 for factors α, β, and γ , respectively. The best OF result 466 in the sensibility analysis was achieved using these values. 467 Similar tests were performed for tunning variable crossover 468 and mutation rates, generation number, and population size. 469 The number of generations and the population size are 470 equal to 500 and 1500, respectively, while maximum and 471 minimum crossover and mutation rates are 0.9, 0.5, 0.15, and 472 0.01. Elite solutions represent 1% of the current population. 473 Such settings are updated every generation.

474
A graph of crossover and mutation is shown in Fig. 5 (a). 475 Rates show a mirrored behaviour since the same expression 476 is used in both variables, changing only the maximum and 477 minimum values. The crossover rate initially has the same 478 value as k c max . The first 150 generations refine several local 479 solutions in the search space, reducing the crossover rate 480 slightly. Thereafter, the value changes more often because the 481 method tries to increase population diversity whenever their 482 similarity increases. In parallel, the mutation rate starts at k m min 483 and increases, varying in the same proportion as the crossover 484 rate.

485
The proposed model is evaluated by performing a second 486 test with fixed rates of 0.7 and 0.075 for crossover and muta-487 tion. Fig. 5 (b) compares the OF progression in both cases, 488 variable and fixed rates. Before 100 generations, the case 489 using variable rates presents OF value lower than 100s, while 490 the case using fixed rates has OF value higher than 250s. 491 The sum of operating times for the coordination problem 492 is 69.03s, whereas with fixed rates, the sum of operating 493 times is 119.84s. OF results show the feasibility of the pro-494 posed methodology using variable crossover and mutation 495 rates.  tion problem are shown in Table 3 [3]. For reclosers' phase 499 units, the upper limit k 51P is set to 1.5, while for other relays,      Table 6 shows the main parameters of DOCRs. Operating 526 times of DOCRs D4 and D5 in 50TD have the minimum 527 value allowed by (32), except for D4 neutral unit. Hence, it is 528 not optimal since the OF can still be improved. Nevertheless, 529 it is a solution of excellent quality because most DOCRs 530 units have the best time allowed by constraints. Moreover, 531 the unit with the longest time has only a difference of 0.0001s 532 concerning the optimal setup. Other PCC relays have higher 533 times, mainly due to coordination restrictions with down-534 stream reclosers. 535 Table 7 shows the main coordination features of 51V local 536 protections. Coordination times of 51V relays are near the 537 limits imposed by constraint (21), indicating the high quality 538 of the obtained solution. Since the solution does not exceed 539 the model's restrictions, the parameters presented provide 540 selectivity with the distribution network protections, tripping 541 only in case of faults internal to the microgrid. 542 Fig. 6 shows the coordination and selectivity between D1 543 with relays R1(50) and R2(50/51). Magenta dashed areas 544 VOLUME 10, 2022   between R1 and F8 in 50TD characteristic. The coordination 559 range of characteristic 51 is wider, with a minimum limit 560 defined by a single-phase fault with impedances of 40 and 561 0 for phase and neutral units, respectively. A and B high-562 light the minimum and maximum current values measured by 563 the recloser for the same coordination range, in that order. 564 Fig. 8 shows the selectivity between D1 and V1 and 565 between D4 and f2. For selectivity between D1 and V1, 566 although the curves in the phase unit seem distant, the multi-567 plication of the DG voltage in the pickup current makes the 568 trip time between them very close, as shown in the coordina-569 tion times of Table 7. For selectivity between D4 and f2, the 570 maximum frequency variation for the coordination range is 571 3.42 Hz. Thus, relay f2 can trip in 5s for I 2∅ . The frequency 572 shift will be greater for a fault inside the microgrid, then f2 573 will trip in the instantaneous mode. 574 Fig. 8 shows the selectivity between D1 and V1 and 575 between D4 and f2. For selectivity between D1 and V1, Microgrids usually include many protection devices, 616 despite these protection devices being redundant. Both the 617 PCC and the PoC have a single protection device in the pro-618 posed methodology. A future proposal should include redun-619 dant protection devices to reinforce the microgrid reliability 620 in parallel and islanded operation modes.

621
The proposed methodology allows a proper operation 622 of microgrids in the distribution system, maximizing the 623 benefits provided by this technology. Moreover, microgrids 624 improve the quality and continuity of the electricity supply 625 service, encouraging the evolution of distribution systems due 626 to their influence on the expansion planning and distribution 627 network operation. The results obtained by the proposed 628 methodology raise expectations about this scenario.