Optimising the operation of district heating networks by combining forecasting and decision-making tools

Abstract

District heating networks (DHNs) are crucial for sustainable urban heating. Optimising their management enhances energy efficiency and reduces environmental impact by utilising advanced monitoring technologies for real-time data collection and machine learning algorithms to predict heat demand and fine-tune system parameters. This paper presents the design and initial results of an optimisation algorithm aimed at improving district heating operations. This study proposes a hybrid approach that combines simulation, machine learning and decision-making tools to optimise DHN operations. A grey-box methodology, which integrates both physics-based and data-driven models, is employed. The approach leverages monitoring data and contextual information to enrich decision-making processes and improve operational efficiency. Initial results suggest that supply temperatures in the generation systems can be reduced by 3°C–5°C while maintaining a 21°C comfort set-point in the buildings. This temperature reduction translates into energy savings of approximately 5% for the entire DHN. The primary limitation is the availability of data, particularly regarding the characterisation of the DHN. Accurate thermal properties of the distribution system are essential for modeling its behavior, but these characteristics are not always well-documented by operators. The proposed design demonstrates improved accuracy over current district heating optimisation trends due to the integration of contextual data. This allows for more precise fine-tuning of operational parameters, resulting in enhanced energy generation and distribution efficiency.