Ibrahim H. A. Alsaleh

Real distribution feeders are multi-phase, unbalanced and full of discrete devices such as voltage regulators and capacitor banks, which makes the AC optimal power flow problem nonconvex and mixed-integer. This line of work builds convex formulations that still respect that physics: semidefinite and mixed-integer SDP relaxations for volt/var optimization, multi-period co-optimization of regulators and PV inverters, and a convex-iteration method that recovers exact solutions from an extended radial ACOPF model.

Once a distribution network can be dispatched, the next questions are how to price it and how to make demand respond. The papers here develop distribution locational marginal prices for multiphase systems, convex bilevel models of day-ahead distribution markets, a bi-level primal-dual model that couples price-based demand response with DLMP, and rank-penalized SDP for flattening net load.

Where a full grid model is unavailable or too slow for real-time use, the work turns to model-free deep reinforcement learning: first a single agent for demand response in a PV-rich feeder, then decentralized multi-agent control in NeomFlex, and curriculum-trained multi-agent RL for energy adequacy in active distribution networks.

As converters replace synchronous machines, stability depends on control loops: phase-locked loops, current limiters, droop and inner current controllers. Recent work, largely with Abdullah Alassaf, builds and analyzes dynamic models of these devices: numerical state-space models of grid-following resources, a 26-state nonlinear DFIG model and its equilibrium structure, DFIG low-voltage ride-through in weak grids with PLL effects and crowbar limits, blade-pitch dynamics, and wind-farm parameter identification with dynamic mode decomposition.

For grid-forming converters, the focus is current limiting: modal analysis of converters with projection-based limiters, and a constraint-activated, projection-free control law for power-limited droop-controlled networks.