Research & projects

High-voltage AC–DC converters sit at a tradeoff between conversion efficiency, thermal stress, and transient stability. Manual tuning of control parameters tends to optimize for one operating point at the expense of others. This project trains a search procedure over the converter's control surface, using simulated grid disturbances to score candidate parameter sets on both steady-state efficiency and recovery behavior after a fault.

Static sensor configurations degrade as load and topology shift throughout the day. This project explores a small neural network that ingests recent measurement history and recommends which sensors to prioritize for sampling, trading measurement freshness against communication and power budget. Early results focus on distribution-network case studies, reusing the observability model built for the PMU placement work.

Home for coursework projects, algorithm implementations, and smaller experiments outside the two active research tracks — mostly C++ data structures work, Python tooling, and MATLAB scripts used to validate the power-systems research.

Distribution networks are far less observable than transmission grids, and instrumenting every node with a PMU is rarely economical. This project formulates placement as a constrained coverage problem over the network graph, searching for the smallest sensor set that keeps the whole feeder observable under contingency. The current implementation compares heuristic and metaheuristic search strategies against a baseline integer formulation, measuring observability redundancy and estimation error introduced by partial coverage.