Approximate explicit model predictive control approach for the unit commitment production of a district heating network

Unit Commitment (UC) production for District Heating Networks (DHNs) is a crucial operational task to meet heat demand under fluctuating parameters. The major concern for DHNs is to efficiently manage network operations, possibly requiring advanced control strategies, while maintaining low computational complexity that enables online deployment with minimal resources. To address this issue, we propose an Approximate Explicit MPC controller that surrogates a baseline UC-MPC controller with offline learning approach. It is shown that AEMPC can conveniently replicates nonlinear relations between input parameters and optimal outputs, leading to very similar performances compared to baseline MPC on simulations.