Albert Poulose, Ph.D.

Research Notes

Short research framing for dynamic security, TSA, and CCT in my work. Teaching chapters, figures, and labs live in SciML for PS. Programme overviews are on the Research page.

Assessment concepts

In my research, dynamic security is the operational question after a disturbance: does the grid stay stable, and is a margin such as CCT still adequate? TSA and CCT are the assessment tools I use most for that question.

Dynamic security

In power systems, dynamic security is the ability of the grid, in its present operating condition, to remain stable after credible disturbances such as short circuits. It is assessed by checking whether generators and converters keep synchronism and whether a stability margin (for example CCT) remains adequate.

Based on IEEE/CIGRE definitions of power system stability and security: Kundur et al., IEEE Trans. Power Syst., 2004 and the converter-era update Hatziargyriou et al., IEEE Trans. Power Syst., 2021.

In the IEEE/CIGRE framework, the transient rotor-angle branch is where my TSA/CCT focus sits; converter-driven stability matters increasingly in renewables-rich grids. Teaching detail (classification tree, SMIB cases, CCT diagrams) is in SciML for PS, not duplicated here.

TSA asks whether synchronism is kept after a disturbance. CCT is one operational margin: how long can a fault remain before stability is lost.

Programme figures

Short visual summaries of the three research programmes. For problem statements, methods, and selected work, see Research.

Two converter problems, one security questionGrid-following (GFL)Current source into a weak or strong gridIGridOften limited in low-inertia regionsGrid-forming (GFM)Virtual synchronous machine (VSM)Voltage source behind reactanceE, δXGridEnergy-function CCT of the GFM unit
Conventional machines are assessed with synchronizing torque among generators. Converter-connected plants are not the same device: grid-following (GFL) and grid-forming (GFM) need different models. Direct energy functions give a CCT for the GFM unit when a GFL converter is in parallel, without waiting for a long time-domain sweep.
One trajectory model, two assessment tasksTime-domainsimulationANDES / EMT referenceAccurate, expensive to repeatPhysics-informedneural networkPredicts rotor-angle δ(t)Respects swing physicsTSA screenNominal clearingCCT searchBinary search on t_clSame trained checkpoint. Better trajectories help screening; they do not automatically give protection-grade CCT.
Scientific AI here is a surrogate for post-fault trajectories, not a classifier that only says stable or unstable. Time-domain simulation remains the teacher. The physics-informed neural network (PINN) is queried many times for planning-oriented TSA and CCT studies when repeating full simulation would be too slow.
A twin is a living model of the dynamics, not a 3D plantGrid nowPMU / SCADA laterPresent operating pointDigital twin01: what to monitor02: fast trajectoriesCCT and synchronismDecisionsContingency screeningOperational planningUpdate the model when the operating point changes, then ask the security question again.
The twin is the operating-point loop I am building toward. Measurements refresh the model; stability theory says which margins matter; the surrogate makes those margins cheap enough to ask often. It is not a visualisation of substations.