Data and code supporting the PhD thesis: Bridging Machine Learning and Optimization for Large-Scale Transmission System Operation

This repository contains datasets and code developed during a PhD project on optimization and machine learning methods for power system operation under uncertainty. The research focuses on problems such as security-constrained substation reconfiguration, transmission congestion management, and stochastic multi-period optimal power flow (OPF). The datasets are generated using a combination of benchmark power system test cases (e.g., IEEE and PEGASE networks), simulated load and renewable generation scenarios, and optimization outputs obtained from mathematical programming models (e.g., MILP, AC/DC OPF) and decomposition methods such as ADMM. In addition, supervised machine learning models (e.g., graph neural networks and recurrent neural networks) are trained on these datasets to approximate or accelerate optimization tasks. The data primarily consist of numerical time-series and network-based variables, including load profiles, generation dispatch, power flows, topology configurations, and scenario-based uncertainty realizations.