Why is the gap so big and falling so slowly?

wangyu hu

February 18, 2023 10:53

Hello every one,
I used matlab yalmip gurobi to solve a MINLP problem, but in the process of running, I got such a result.
It really took a long time. I want to hurry as soon as possible. How can I modify it?Thank you.

0

wangyu hu

February 18, 2023 10:58

Set parameter MIPGap to value 0.15
Set parameter CrossoverBasis to value 0
Set parameter NodefileDir to value ""
Set parameter NonConvex to value 2
Set parameter PreSOS2BigM to value 0
Set parameter TuneTrials to value 3
Gurobi Optimizer version 9.5.1 build v9.5.1rc2 (win64)
Thread count: 2 physical cores, 4 logical processors, using up to 4 threads
Optimize a model with 8688 rows, 2939 columns and 16170 nonzeros
Model fingerprint: 0x510e4880
Model has 889 quadratic constraints
Variable types: 1470 continuous, 1469 integer (1469 binary)
Coefficient statistics:
Matrix range [1e-05, 2e+06]
QMatrix range [8e-01, 2e+02]
QLMatrix range [1e+00, 5e+06]
Objective range [1e+00, 1e+00]
Bounds range [1e+00, 1e+00]
RHS range [1e-05, 2e+06]
QRHS range [4e+02, 5e+06]
Presolve removed 5717 rows and 1466 columns
Presolve time: 0.24s
Presolved: 9999 rows, 3089 columns, 27942 nonzeros
Presolved model has 16 quadratic constraint(s)
Presolved model has 1443 bilinear constraint(s)
Variable types: 2436 continuous, 653 integer (645 binary)
Root relaxation: objective -7.713198e+06, 3419 iterations, 0.12 seconds (0.07 work units)

0 0 -7340583.0 0 1061 - -7340583.0 - - 1s
0 0 -5456100.0 0 1037 - -5456100.0 - - 2s
H 0 0 -362049.4931 -5456100.0 1407% - 2s
H 0 0 -383732.1446 -5456100.0 1322% - 2s
0 0 -5383226.9 0 839 -383732.14 -5383226.9 1303% - 2s
0 0 -4753175.2 0 1201 -383732.14 -4753175.2 1139% - 3s
0 0 -4751388.3 0 1191 -383732.14 -4751388.3 1138% - 3s
0 0 -4514750.7 0 1186 -383732.14 -4514750.7 1077% - 3s
0 0 -4513561.5 0 1246 -383732.14 -4513561.5 1076% - 4s
0 0 -4432402.9 0 1154 -383732.14 -4432402.9 1055% - 5s
0 0 -4415286.3 0 1221 -383732.14 -4415286.3 1051% - 5s
0 0 -4400268.0 0 1300 -383732.14 -4400268.0 1047% - 6s
0 0 -4398325.0 0 1182 -383732.14 -4398325.0 1046% - 6s
0 0 -4369984.6 0 1220 -383732.14 -4369984.6 1039% - 7s
0 0 -4366191.3 0 1251 -383732.14 -4366191.3 1038% - 7s
0 0 -4355759.8 0 1292 -383732.14 -4355759.8 1035% - 8s
0 0 -4354081.3 0 1185 -383732.14 -4354081.3 1035% - 8s
0 0 -4347553.3 0 1286 -383732.14 -4347553.3 1033% - 9s
0 0 -4345474.9 0 1269 -383732.14 -4345474.9 1032% - 9s
0 0 -4342547.0 0 1264 -383732.14 -4342547.0 1032% - 9s
0 0 -4341583.0 0 1307 -383732.14 -4341583.0 1031% - 10s
0 0 -4340819.5 0 1279 -383732.14 -4340819.5 1031% - 10s
0 0 -4336649.3 0 1135 -383732.14 -4336649.3 1030% - 15s
0 2 -4336649.3 0 1129 -383732.14 -4336649.3 1030% - 15s
20 23 -3586894.9 8 1132 -383732.14 -3895079.4 915% 949 20s
60 61 -3191901.1 20 987 -383732.14 -3895079.4 915% 972 25s
152 159 -2921582.2 54 455 -383732.14 -3895079.4 915% 656 30s
305 278 -3471022.5 4 1334 -383732.14 -3891830.7 914% 464 35s
364 334 -3363563.8 17 1093 -383732.14 -3891830.7 914% 447 42s
387 367 -3163760.1 22 1078 -383732.14 -3891830.7 914% 502 52s
463 426 -3027670.4 30 1055 -383732.14 -3891830.7 914% 465 56s
486 427 -2937899.1 43 235 -383732.14 -3891830.7 914% 468 62s
H 487 406 -435465.6558 -3891830.7 794% 467 65s
489 407 -3022331.1 48 271 -435465.66 -3876612.8 790% 465 70s
491 408 -922522.93 105 337 -435465.66 -3490080.9 701% 463 77s
493 410 -2342155.3 72 275 -435465.66 -3266108.5 650% 461 82s
496 412 -2891737.0 39 314 -435465.66 -2891737.0 564% 458 85s
502 416 -689802.26 109 314 -435465.66 -2769523.4 536% 453 90s
511 422 -2749007.0 41 283 -435465.66 -2749007.0 531% 445 95s
517 426 -2746019.7 47 310 -435465.66 -2746019.7 531% 440 104s
518 426 -518114.39 140 310 -435465.66 -2746019.7 531% 439 105s
520 431 -2740257.6 12 1257 -435465.66 -2740257.6 529% 532 118s
522 432 -2694582.4 13 1263 -435465.66 -2724339.3 526% 539 122s
528 436 -2691740.2 14 1215 -435465.66 -2718827.9 524% 545 125s
535 442 -2538093.2 16 1123 -435465.66 -2694004.5 519% 561 130s
542 445 -2477096.7 17 1171 -435465.66 -2694004.5 519% 569 135s
553 454 -2437827.5 20 1025 -435465.66 -2694004.5 519% 604 140s
563 462 -2369429.7 22 972 -435465.66 -2694004.5 519% 619 145s
573 469 -2375749.5 25 1025 -435465.66 -2694004.5 519% 631 150s
587 477 -2300924.5 28 1056 -435465.66 -2694004.5 519% 654 155s
607 496 -2009059.8 34 894 -435465.66 -2694004.5 519% 666 162s
616 497 -1983229.1 36 914 -435465.66 -2694004.5 519% 669 165s
628 509 -1942341.6 39 892 -435465.66 -2694004.5 519% 680 170s
643 521 -1860475.1 45 914 -435465.66 -2694004.5 519% 684 175s
662 527 -1773302.5 51 890 -435465.66 -2694004.5 519% 681 182s
668 531 -1782337.8 53 1113 -435465.66 -2694004.5 519% 686 186s
687 544 -1715181.1 56 961 -435465.66 -2694004.5 519% 705 194s
699 546 -1699214.5 60 1046 -435465.66 -2694004.5 519% 710 196s
713 551 -1624891.6 65 918 -435465.66 -2694004.5 519% 725 202s
726 571 -1484962.5 71 774 -435465.66 -2694004.5 519% 726 205s
762 573 -1399661.8 81 744 -435465.66 -2694004.5 519% 705 210s
816 594 -574869.28 103 627 -435465.66 -2694004.5 519% 696 217s
851 587 -447845.28 113 616 -435465.66 -2692339.0 518% 687 221s
873 599 -2036687.8 21 1171 -435465.66 -2692339.0 518% 706 229s
883 602 -2224034.8 24 1150 -435465.66 -2692339.0 518% 711 234s
889 611 -2193547.0 28 1159 -435465.66 -2692339.0 518% 710 243s
900 618 infeasible 36 -435465.66 -2692339.0 518% 727 249s
913 636 -1763919.4 43 1042 -435465.66 -2692339.0 518% 739 256s
939 641 -1619716.7 52 1029 -435465.66 -2692339.0 518% 737 260s
953 681 -1540078.6 58 877 -435465.66 -2692339.0 518% 735 266s
998 725 -1432257.8 72 805 -435465.66 -2692339.0 518% 731 270s
1057 764 -1123128.5 84 715 -435465.66 -2692339.0 518% 713 275s
1216 874 -557441.59 105 668 -435465.66 -2689779.6 518% 670 284s
1323 855 -2659028.3 17 1035 -435465.66 -2689602.2 518% 637 290s
1352 853 -2479827.7 22 1035 -435465.66 -2689602.2 518% 648 298s
1360 866 -2385124.6 27 1056 -435465.66 -2689602.2 518% 664 304s
1375 872 -2258562.3 35 996 -435465.66 -2689602.2 518% 675 310s
1386 895 -2090806.4 37 1016 -435465.66 -2689602.2 518% 685 316s
1413 918 -1823789.2 43 855 -435465.66 -2689602.2 518% 694 323s
1445 1025 -1723390.5 52 710 -435465.66 -2689602.2 518% 704 332s
1563 1095 -1188552.0 86 436 -435465.66 -2689602.2 518% 669 337s
1678 1152 -543682.53 109 482 -435465.66 -2683694.6 516% 644 344s
1797 1183 -2218841.4 17 1147 -435465.66 -2683694.6 516% 622 351s
1874 1197 -1887292.1 23 1041 -435465.66 -2683694.6 516% 615 357s
1908 1223 -1488702.9 31 779 -435465.66 -2683694.6 516% 620 364s
1934 1295 -1147906.7 48 870 -435465.66 -2675683.4 514% 625 372s
* 2047 1305 113 -473835.4654 -2673366.5 464% 609 379s
* 2051 1305 112 -473844.3990 -2673366.5 464% 608 379s
* 2053 1305 111 -473854.3268 -2673366.5 464% 607 379s
2091 1331 -2410551.9 22 1110 -473854.33 -2672935.2 464% 612 386s
2117 1359 -2054439.9 39 997 -473854.33 -2672935.2 464% 622 393s
2145 1417 -1673289.7 47 904 -473854.33 -2666838.5 463% 629 402s
2221 1439 -2373255.6 20 1080 -473854.33 -2666838.5 463% 628 411s
2261 1450 -1896099.2 35 964 -473854.33 -2666838.5 463% 634 418s
2272 1487 -1656375.8 45 1012 -473854.33 -2653365.5 460% 639 426s
2317 1519 -2352238.8 24 1127 -473854.33 -2653365.5 460% 648 436s
2349 1557 -1579656.7 39 1164 -473854.33 -2653365.5 460% 659 445s
2387 1645 -1141367.6 55 857 -473854.33 -2653331.6 460% 669 455s
2495 1707 -2401349.8 17 1103 -473854.33 -2653331.6 460% 656 463s
2593 1744 -1893451.9 26 1073 -473854.33 -2653331.6 460% 648 472s
2630 1768 -1428861.7 41 1153 -473854.33 -2653331.6 460% 657 480s
2670 1813 -1154329.1 49 988 -473854.33 -2640210.9 457% 664 491s
2728 1853 -2299572.3 17 1126 -473854.33 -2640210.9 457% 673 503s
2778 1879 -2054988.8 25 1104 -473854.33 -2640210.9 457% 680 515s
2804 1914 -1604543.8 45 983 -473854.33 -2617718.1 452% 692 529s
2851 1953 -2098615.0 28 1065 -473854.33 -2617718.1 452% 702 540s
2890 2000 -1375219.4 52 860 -473854.33 -2617170.6 452% 709 561s
2953 2047 -1872041.8 23 1041 -473854.33 -2617170.6 452% 722 574s
3002 2206 -1203131.3 43 781 -473854.33 -2617151.9 452% 729 586s
3205 2247 -2345515.0 18 1159 -473854.33 -2617151.9 452% 705 597s
3249 2316 -1697442.0 27 958 -473854.33 -2617151.9 452% 717 611s
3330 2439 -1385836.3 58 940 -473854.33 -2617151.9 452% 719 631s
3481 2496 -880033.18 94 823 -473854.33 -2602001.1 449% 710 647s
3585 2539 -2580304.0 18 1217 -473854.33 -2602001.1 449% 711 662s
3638 2566 -2197551.0 23 1103 -473854.33 -2602001.1 449% 723 677s
3669 2648 -2089203.4 26 1077 -473854.33 -2602001.1 449% 726 692s
3775 2752 -1747518.9 37 949 -473854.33 -2602001.1 449% 730 709s
3929 2841 -1029785.6 48 535 -473854.33 -2591900.9 447% 726 730s
4038 2941 -1723067.6 37 1031 -473854.33 -2585164.1 446% 730 758s
4168 2986 -2531234.8 19 1239 -473854.33 -2583685.5 445% 724 776s
4213 3033 -1489604.1 41 965 -473854.33 -2583685.5 445% 736 798s
4270 3079 -930922.87 59 699 -473854.33 -2574045.3 443% 743 814s
4337 3153 -1862574.1 29 1037 -473854.33 -2568933.9 442% 750 830s
4425 3204 -1106333.8 72 547 -473854.33 -2562510.5 441% 757 845s
4492 3257 -1946664.4 21 1005 -473854.33 -2562510.5 441% 764 866s
4557 3319 -1199314.2 36 864 -473854.33 -2561338.4 441% 779 884s
4631 3436 -1953541.0 29 1085 -473854.33 -2561183.4 441% 787 903s
4787 3584 -1600599.2 25 886 -473854.33 -2554296.7 439% 789 923s
4972 3796 -1959328.9 22 1120 -473854.33 -2553424.4 439% 783 946s
5189 4048 -1607586.1 45 841 -473854.33 -2553294.7 439% 773 970s
5515 4161 -2014126.3 26 1158 -473854.33 -2550025.4 438% 755 998s
5658 4248 -2066086.9 25 1029 -473854.33 -2548517.1 438% 760 1022s
5780 4351 -2239696.0 19 1118 -473854.33 -2544425.2 437% 762 1051s
5913 4406 -2330889.8 20 1218 -473854.33 -2543435.5 437% 768 1080s

0

Jaromił Najman

Gurobi Staff

February 20, 2023 09:32

Hi,

There can be many reasons for a slowly solving non-convex MINLP.

The coefficient ranges are not great in your case. Please refer to our Guidelines for Numerical Issues and try to improve the model scaling. This may already drastically improve the solver's performance.
It looks like you did not provide finite bounds for all variables. In particular, it is strongly recommended to provide tight finite bounds for all variables participating in nonlinear terms when solving a non-convex model.
Your model is quite big given that it's a non-convex MINLP. Thus, it it somewhat expected to take quite some time to solve.
If possible, you should upgrade to Gurobi's latest version 10.

You could also write the model to an LP file and share it with the Community. Maybe someone has an idea of how to solve the model in a better way. Note that uploading files in the Community Forum is not possible but we discuss an alternative in Posting to the Community Forum.

Best regards,
Jaromił

0

Why is the gap so big and falling so slowly?

Comments

Didn't find what you were looking for?