Energy management in a multi-building setup via distributed stochastic optimization

This master thesis work concerns distributed and stochastic optimization with application to an optimal energy management in a multi-building network setting. More precisely, an iterative distributed algorithm based on proximal minimization and randomization of stochastic uncertainty is studied. The algorithm is applied to a microgrid set-up where multiple buildings with their own chiller plants are sharing a thermal storage unit. The goal is to minimize the overall electricity cost over a one-day horizon by optimally setting the temperature set-points and storage usage while accounting for uncertainty (outside temperature, people occupancy, and solar radiation). The methodology used to solve the resulting chance-constrained optimization problem is based on the scenario approach that is extended to a distributed setting while retaining its probabilistic guarantees regarding feasibility. The obtained nite horizon solution can be applied in a receding horizon fashion, leading to a stochastic model predictive control (MPC) strategy.