Advanced topics overview

Start here¶

1. From assessment to decision — how Identification, Dynamics, and Stability connect toward operating-point decisions

Planned directions (stubs later)¶

• Operator learning (DeepONet / FNO) as a forward function→function branch — not a PINN variant

• Uncertainty quantification for estimated parameters

• Multimachine / network scaling after SMIB mastery (ANDES / Kundur path under Stability)

• Selected converter synchronization case studies (after Labs 1–5; VSM–swing bridge first)

Fast-forward topics¶

Short pointers to directions this book does not teach yet. Each needs Labs 1–5 first. Method background is Thuerey et al. (2021); linked external code carries its own license.

• Probabilistic learning for uncertainty. Generative models (diffusion, flow matching) learn a distribution of trajectories or parameters instead of one answer. In this book’s terms: a spread of plausible $(H, D)$ or a probability of instability, instead of a point estimate. Background: probabilistic learning introduction.

• Graph architectures for networks. Multimachine systems are graphs (buses and lines). Graph neural networks and graph-based generative models respect that structure and are a natural next step after the Kundur path. Background: network architectures, graph-based diffusion models.

• Reinforcement learning for control. Learning control actions (e.g. damping) from interaction. It is outside the three pillars of this book; treat it as curiosity. Background: reinforcement learning introduction; applied example: the ANDES RL gallery entry below.

Linked further practice (ANDES gallery)¶

Summarize-then-link only — not core TOC spines:

• Forced oscillation (WECC 179) — large-system elective

• RL oscillation damping (SMIB) — outside the three pillars; curiosity

• Gallery index: https://docs.andes.app/en/stable/gallery/index.html

Status¶

Synthesis page filled; other Advanced topics remain stubs until Labs 1–5 deepen.