Optimization through Automated Decision Support in District Cooling - Minimizing Electricity Use through Software-Based Start/Stop Control and Chiller Dispatch

Abstract

This thesis investigates how software-based decision support can improve operational efficiency in district cooling through automated start/stop control and chiller dispatch. The study was conducted as a design science case study at Göteborg Energi’s district cooling network in Gothenburg, Sweden. The work addresses three main questions: identifying the main operational limitations in the current control strategy, designing a software-based artifact for improved dispatch decisions, and assessing which parts of the artifact are case-specific and which may be transferable to similar district cooling systems. The current operation was analyzed using historical operational data from 2025 together with input from domain experts. This analysis showed that the existing control strategy is only partly standardized in practice, since actual chiller dispatch varies under similar operating conditions and remains partly dependent on operator judgement. To address these limitations, a software-based artifact was developed consisting of empirical chiller performance models, a Mixed Integer Non-Linear Programming (MINLP) optimization engine, and an interactive dashboard for visualization and fleet management. The artifact was evaluated through simulation using historical operating conditions from 2025 as a baseline. The results indicate that the proposed control strategy can improve operational efficiency while satisfying cooling demand and limiting daily start/stop events in accordance with the defined success criteria. The largest improvements were observed during the summer period when the electricity use was reduced by as much as 41.69%. The thesis thus contributes both a practical prototype and a design-oriented approach that may be adaptable to similar district cooling systems after local recalibration of the chiller models.