Data-Driven Asset Management in District Heating: Barriers, Insights, and Future Pathways

District heating (DH) systems found in countries such as Denmark and Sweden are characterized by tens of thousands of kilometers of installed pipeline and generate large volumes of operational data. Nevertheless, renewal policy is to date largely dependent on traditional age and run-to-failure metrics and does not use the operational data effectively. Simultaneously, recent work on machine-learning (ML) techniques has opened new possibilities. Approaches such as survival analysis, anomaly detection on operational data (e.g., from smart meters), and digital twins reveals that failures can be forecast and ranked more accurately by ML models than by methods relying on traditional failure-rate assumptions. This paper asks: how should these data-driven and ML-based methods be structured and embedded so they actually change DH maintenance and renewal decisions? First, real DH case studies and recent literature are analyzed to identify four blockers for ML in asset management. Then, ML-oriented opportunities are synthesized into three clusters for DH. On this basis, three concrete development pathways are proposed: a DH-specific fault-registration infrastructure (DH-ELFAS) that provides high-quality labels for ML, reliability-driven and scenario-based portfolio planning that uses ML-derived failure probabilities and consequences as optimization inputs, and a compact set of standardized risk and reliability indicators to expose model outputs in regulatory and investment discussions.