VictorMylle/Thesis
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
b3f05f386f0d4c1b44bc96d4f82793bfe4d104a4
Thesis/Result-Reports/Policies.md

20 KiB
Raw Blame History

Policies for battery charging

Battery

Capacity: 2MWh
Charging Power: 1MW
Discharging Power: 1MW

Baseline policy

Determining 2 thresholds for battery charging and discharging. Charging when the imbalance price is below the lower threshold and discharging when the imbalance price is above the upper threshold.

Training data: 01-01-2020 - 31-12-2022 Test data: 01-01-2023 - 12-12-2023

Charge threshold Discharge threshold Charging Cost Discharging Profit Total Profit Charge cycles Mean charging price Mean discharging price
0 150 175 165982 989322 823340 1701.62 32.4889 348.51
0 175 200 190060 1.00145e+06 811389 1594 37.0705 367.44
0 125 150 132766 939145 806378 1729.88 27.7746 333.447
0 200 225 219726 1.01789e+06 798159 1493.12 42.9038 390.49
0 150 200 151145 937599 786454 1514.88 32.4889 367.44
0 125 175 122067 904885 782818 1584.62 27.7746 348.51
0 175 225 173050 950270 777219 1415 37.0705 390.49
0 100 125 69876.4 829454 759577 1662.88 19.9739 315.1
0 150 225 136563 886279 749716 1339.12 32.4889 390.49
0 125 200 109873 853728 743855 1402.5 27.7746 367.44
0 100 150 63807.1 802437 738630 1502.5 19.9739 333.447
0 100 175 56405.8 770173 713768 1366.75 19.9739 348.51
0 75 100 14153.7 724660 710506 1684.5 11.3023 284.61
0 125 225 97524.1 804468 706944 1231.88 27.7746 390.49
0 25 50 -56542.1 637628 694171 2696.12 -16.9846 204.943
0 75 125 8003.32 688780 680776 1407.38 11.3023 315.1
0 100 200 49469.3 721380 671911 1196.12 19.9739 367.44
0 50 75 -47575.2 614964 662539 1752 -9.65396 254.211
0 75 150 5147.22 661439 656292 1254.62 11.3023 333.447
0 50 100 -46110.5 592694 638805 1417.5 -9.65396 284.61
0 100 225 42046.6 677560 635514 1042.88 19.9739 390.49
0 75 175 850.49 631552 630702 1129 11.3023 348.51
0 25 75 -63342.8 561356 624699 1637 -16.9846 254.211
0 50 125 -46686.3 556263 602950 1150.5 -9.65396 315.1
0 25 100 -61448.6 537683 599132 1306.5 -16.9846 284.61
0 75 200 -1767.94 590094 591862 981.5 11.3023 367.44
0 50 150 -45858.6 531472 577330 1007.5 -9.65396 333.447
0 25 125 -61228.7 500861 562089 1045.62 -16.9846 315.1
0 75 225 -3475.71 555727 559203 855.875 11.3023 390.49
0 50 175 -46084.1 506348 552432 900.875 -9.65396 348.51
0 25 150 -60177.2 476823 537000 906.75 -16.9846 333.447
0 50 200 -43738 473478 517216 779.25 -9.65396 367.44
0 25 175 -59466.7 453988 513455 808.25 -16.9846 348.51
0 50 225 -40529.8 452651 493181 693.875 -9.65396 390.49
0 25 200 -55766.8 426055 481822 699.875 -16.9846 367.44
0 25 225 -51953 407484 459437 623.875 -16.9846 390.49
0 0 75 -107853 332344 440197 851.375 -73.2011 254.211
0 0 100 -97804.3 341072 438876 761.5 -73.2011 284.61
0 0 50 -121468 314671 436139 993.625 -73.2011 204.943
0 0 125 -91087.2 342646 433734 686.125 -73.2011 315.1
0 0 25 -126590 305984 432574 1041.75 -73.2011 193.87
0 0 150 -85919 341046 426964 635.75 -73.2011 333.447
0 0 175 -82378.2 338310 420688 600.625 -73.2011 348.51
0 0 200 -75822.2 328188 404010 541.25 -73.2011 367.44
0 0 225 -69488.7 320776 390265 494.75 -73.2011 390.49
0 -25 125 -102495 269615 372110 534.062 -105.73 315.1
0 -25 100 -108934 262527 371462 574.938 -105.73 284.61
0 -25 75 -118553 252453 371006 626.062 -105.73 254.211
0 -25 150 -97080 271209 368289 503.438 -105.73 333.447
0 -25 50 -129918 236112 366030 683.562 -105.73 204.943
0 -25 175 -92889.5 272520 365409 480.438 -105.73 348.51
0 -25 25 -132518 227910 360428 695.562 -105.73 193.87
0 -25 200 -85482.7 268638 354120 438.5 -105.73 367.44
0 -25 225 -79253.2 264052 343305 403.75 -105.73 390.49
0 -25 0 -139125 199831 338956 724.312 -105.73 146.436
0 -50 125 -102665 216740 319405 431.562 -127.845 315.1
0 -50 175 -94272.2 223434 317706 393.188 -127.845 348.51
0 -50 150 -97889.1 219525 317414 409.312 -127.845 333.447
0 -50 100 -108274 208273 316548 458.312 -127.845 284.61
0 -50 75 -115707 198325 314032 488.438 -127.845 254.211
0 -50 200 -87698.7 223235 310933 363.625 -127.845 367.44
0 -50 50 -124664 184027 308691 525.688 -127.845 204.943
0 -50 25 -126605 177834 304440 533.062 -127.845 193.87
0 -50 225 -81706.8 221510 303217 337.75 -127.845 390.49
0 -50 0 -131679 158118 289797 551.188 -127.845 146.436
0 -50 -25 -135234 140945 276180 565 -127.845 139.469
0 -75 175 -90764.9 184846 275611 325.812 -146.153 348.51
0 -75 125 -98076.6 176722 274799 355.062 -146.153 315.1
0 -75 150 -93979.3 179914 273893 338.188 -146.153 333.447
0 -75 100 -103131 169294 272425 375.062 -146.153 284.61
0 -75 200 -85036.8 186067 271104 302.375 -146.153 367.44
0 -75 75 -109431 159779 269210 395.938 -146.153 254.211
0 -75 225 -79978.8 185752 265731 282.75 -146.153 390.49
0 -75 50 -116483 147147 263630 421.062 -146.153 204.943
0 -75 25 -117886 141847 259734 424.812 -146.153 193.87
0 -75 0 -120983 126071 247054 435.188 -146.153 146.436
0 -75 -25 -123261 111930 235190 443.625 -146.153 139.469
0 -75 -50 -124708 101672 226380 449.25 -146.153 136.356
0 -100 200 -76499.7 137174 213674 219.625 -179.829 367.44
0 -100 175 -79830.5 132966 212797 230.125 -179.829 348.51
0 -100 225 -72449.9 138631 211081 207.625 -179.829 390.49
0 -100 150 -82202.7 128215 210418 237.125 -179.829 333.447
0 -100 125 -84192.7 123976 208169 243.625 -179.829 315.1
0 -100 100 -86319.5 117325 203645 250.25 -179.829 284.61
0 -100 75 -90219.7 109153 199373 259.25 -179.829 254.211
0 -100 50 -94055.9 99443.5 193499 269.875 -179.829 204.943
0 -100 25 -94464.1 96073.2 190537 270.875 -179.829 193.87
0 -100 0 -95713.7 87279.5 182993 274.625 -179.829 146.436
0 -100 -25 -97020.3 78656.5 175677 278.25 -179.829 139.469
0 -100 -50 -97722.9 71560.6 169283 280.375 -179.829 136.356
0 -100 -75 -98604.6 63373.8 161978 282.625 -179.829 134.012

Optimal policy

Charge threshold: 150 €/MWh
Discharge threshold: 175 €/MWh

Profit on test data: € 359548
Charge cycles: 855.938

Smarter Policy using predicted NRV

Test data: 01-01-2023 until 08-10–2023

 Policy Threshold Step Size (€)  Total Profit (€)  Charge Cycles
 Baseline (charge: 150, discharge: 175)  25 251202.59  725
 Baseline (yesterday imbalance price)  25 342980.0938 903
 GRU Predicted NRV (mean reconstructed imbalance price) 50 325362.81  856
 GRU Predicted NRV (mean reconstructed imbalance price) 25 334058.65 862
GRU Predicted NRV (mean thresholds) 25 339846.9062 842
Diffusion Predicted NRV (mean thresholds) 25 338168.0312 886

Penalty parameter for charge cycles

 Policy  Penalty parameter  Total Profit (€) Charge Cycles
Baseline (yesterday imbalance price) 10 335968.9062 895.4375
Baseline (yesterday imbalance price) 20 328577.2812 886.5625
Baseline (yesterday imbalance price) 50 306301.9688 850.9375
Baseline (yesterday imbalance price) 150 251367.0469 749.7500

| Diffusion Predicted NRV | 10 | 331288.7812 | 880.3750 | | Diffusion Predicted NRV | 20 | 324568.7812 | 871.5000 | | Diffusion Predicted NRV | 50 | 304761.1875 | 837.3125 | | Diffusion Predicted NRV | 150 | 238441.5625 | 670.1875 |

TODO

  • Baseline

  • Profit penalty parameter als over charge cycles voor een dag -> parameter bepalen op training data (convex probleem) (< 400 charge cycles per jaar) (over een dag kijken hoeveel charge cycles -> profit - penalty * charge cycles erover, (misschien belonen als eronder charge cycles))

  • Meer verschil bekijken tussen GRU en diffusion

  • (In Progress) Andere lagen voor diffusion model (GRU, kijken naar TSDiff)

  • Policies met andere modellen (Linear, Non Linear)

  • Visualize the policies over the whole testr set -> thresholds plotten voor elke dag (elke policy) -> mss distribution om overzichtelijk te houden (mean and std)

  • Probleem met diffusion model (activation function? waarom direct grote waardes?)

  • Autoregressive confidence problem -> Quantiles zelf uit elkaar halen (helpt dit?)

  • time steps reducing for diffusion model (UNet activation functions?)

  • (State space model? S4)

TODO:

  • diffusion model oefening generative models vragen

  • [ ] Non autoregressive models policy testen (Non Linear eerst) -> als dit al slect, niet verder kijken, wel vermelden

  • Policy in test set -> over charge cycles (stop trading electricity)

  • penalty bepalen op training data

  • cycles en profit herschalen naar per jaar

baseline -> NRV van gisteren gebruiken om thresholds te bepalen voor vandaag andere policies -> NRV van vandaag voorspellen met model en thresholds bepalen voor vandaag

Eerste baseline -> thresholds bepalen op training data maar ook stoppen als 400 cycles (herschalen) per jaar bereikt zijn -> thresholds zouden anders moeten zijn (Ook met penalty parameter) -> deze toepassen op test set (ook stoppen als 400/jaar bereikt zijn)

Visualizatie van thresholds over test set voor baselines en complexere modellen -> zonder penalties tonen

1 a 2 Case studies (extreme gevallen, thresholds 150, -5, normale mss)

  • Generatie van NRV (echte NRV)
  • Thresholds die eruit komen
  • Profit en charge cycles

Policy volledig fixen en later training script met policy direct erachter (tijdens schrijven door laten runnen)

  1. Policy
  2. Finetuning van modellen (+ vergelijken met elkaar opbv profit en error)
  3. Ablation Study (input features weghalen en kijken wat er gebeurt) ( 4) Diffusion tussendoor )

Inleiding + Literatuurstudie + Tabellen die we gaan bespreken -> updaten met nieuwe data dan

Nog eens 3e meeting opbrengen voor 2e deel maart.

Reference in New Issue View Git Blame Copy Permalink