Data-driven modelling and optimal management of district heating networks

This paper presents an optimization-based algorithm for the management of district heating networks, aiming to minimize their operational cost and maximize their efficiency. The method involves a mixed-integer formulation to properly control the operations of gas boilers and cogeneration systems at the central heating station. A novel modelling method for the district heating networks is also proposed, where temperature dynamics are directly learned from data through the identification of piece-wise affine models. The developed approach ensures that proper temperature levels are always provided to all users, meanwhile minimizing the network heat losses, and enabling cogeneration systems to optimally participate to the Day-Ahead Energy Market. The optimization system has been designed considering a real district heating network in the Milan area and first experiments on the real plant have been carried out. The economic evaluations are satisfactory, leading to significative energy saving and a consistent reduction of CO2 emissions and gas consumption.