Shared method core before the three application pillars. Use a classical SMIB (single-machine infinite-bus) system repeatedly so the same equations appear as forward simulation, surrogate learning, stability labels, and parameter estimation.
Coming from another field¶
Chapters in this part can be read out of order when the Before you start box on the page is already satisfied.
| If this is you | Do this first |
|---|---|
| Machine learning, new to power systems | Power systems for ML people, then Problem framing |
| Power systems, new to machine learning | Machine learning for power-systems people, then Problem framing |
| New to both | Programming from zero and Your first lab |
| Comfortable with both | Use the outline and open the chapter you need |
Word-by-word translations: Glossary. Routes by role: Learning paths.
Foundations teach the shared plant and an early physics-informed path (supervised baseline, then forward and inverse PINNs). Dynamics, Stability, and Identification deepen the three engineering questions. Other SciML families (Neural ODEs, operators) arrive when needed — see SciML method families and Lab 4.
Learning outcomes¶
After this part you should be able to:
State forward, inverse, and stability-assessment tasks with clear inputs and outputs
Distinguish states (time-varying) from parameters (usually fixed unknowns)
Build or use a transparent numerical SMIB reference with documented events
Explain data loss vs physics residual loss in a PINN
Contrast classical least squares with inverse PINNs for named parameters
Know when Neural ODEs / operator learning are a different SciML branch
Suggested reading order¶
Problem framing — tasks, state vs parameter, forward vs inverse
Inverse PINNs then Lab 3
Optional landscape: SciML method families.
Pillars after this part¶
| Pillar | Engineering question |
|---|---|
| Dynamics | How will the system evolve? |
| Stability | Stable or not, and how close to the boundary? |
| Identification | Which parameters explain the measurements? |