Skip to content

Master Thesis (TFM)

Master's Thesis

Moj master rad na ETSEIB-u (2026): estimacija stanja u distributivnim mrežama grafovskim i fizički informisanim neuronskim mrežama, kao ceo postupak na Baran-Wu mreži od 33 čvora, uporedo sa GridCal-ovim estimatorom najmanjih težinskih kvadrata.

My master’s thesis at ETSEIB (2026): state estimation in distribution networks with graph neural networks and physics-informed neural networks, as an end-to-end pipeline on the 33-bus Baran-Wu feeder, set beside GridCal’s weighted least squares estimator.

Jezik
engleski
Fajlova
30
Veličina
34 MB
Otvori folder na Google Drive-u (otvara se u novom tabu)

Besplatno, za gledanje i preuzimanje

Odakle da počneš

Prvo pročitaj rad. Za kod otvori glavnu svesku u 03 - Code: ona učitava sačuvane modele, pa radi bez treniranja.

Where to start

Read the thesis first. For the code, open the main notebook in 03 - Code: it loads the saved models, so it runs without training.

Šta je u folderu

4 dela, pod imenima koja imaju na Drive-u.

  1. Master rad

    01 - Thesis - Advanced State Estimation in Distribution Networks Using GNN and PINN.pdf

    Rad (123 strane): estimacija stanja u elektroenergetskim sistemima, neuronske, grafovske i fizički informisane mreže, i ceo postupak na Baran-Wu mreži od 33 čvora sa simuliranim PMU, smart-meter, SCADA i pseudo-merenjima.

    Thesis. The thesis (123 pages): state estimation in power systems, neural, graph and physics-informed networks, and the pipeline on the 33-bus Baran-Wu feeder with simulated PMU, smart-meter, SCADA and pseudo-measurements.

  2. Slajdovi odbrane

    02 - Defence slides.pdf

    25 slajdova sa odbrane rada.

    Defence slides. The 25 slides of the defence.

  3. Kod

    03 - Code

    Sve što je potrebno da sveska iz rada radi: glavna sveska, paket gridcal_utils, MATPOWER mreže od 5 do 118 čvorova, skup od 2.000 primera i istrenirani GNN i PINN modeli. Python 3.11 sa GridCalEngine-om i PyTorch Geometricom.

    Code. Everything the thesis notebook needs to run: the main notebook, the gridcal_utils package, MATPOWER cases from 5 to 118 buses, the 2,000-sample dataset and the trained GNN and PINN models. Python 3.11 with GridCalEngine and PyTorch Geometric.

  4. Od mentora i fakulteta

    04 - From the supervisor and the school

    Zadatak rada, GridCal tutorijali (tokovi snaga, optimalni tokovi snaga, estimacija stanja, analiza ispada i početni GNN estimator stanja) i ETSEIB-ovo uputstvo za izgled master rada.

    From the supervisor and the school. The thesis brief, GridCal tutorials on power flow, optimal power flow, state estimation, contingency analysis and a GNN state estimator, and ETSEIB’s thesis style guide.

Dobro je znati

  • Rad i kod su primer, a ne za prepisivanje.The thesis and the code are an example, not ones to copy.
  • Program i pravila polaganja se menjaju iz godine u godinu, pa važeća proveri kod predmetnog nastavnika.Programmes and exam rules change from year to year, so check the current ones with the teacher.

Ako je nešto u ovom folderu tvoje i želiš da se ukloni, piši mi na milossaric@outlook.com i uklanjam ga.Wrote something in this folder and want it taken down? Email me at the same address and I’ll remove it.

Gde sam ovo primenio

Rad ima svoju stranicu na ovom sajtu, sa igrom na mreži.

The thesis has its own page on this site, with a game to play on the grid.

State estimation on a power grid

Još predmeta

Svi predmeti