Verfahren und System zur Erkennung von Unregelmässigkeiten in einem Wärmeenergieaustauschsystem unter Verwendung eines Modells und Verfahren zum Trainieren des Modells

Details

Veröffentlichungs-Nr.

EP4814390

Anmeldetag

27. März 2025

Veröffentlichung

30. September 2026

Rechtsraum

EP

IPC

F24D10/00, G05B23/02

Abstract

A computer implemented method and system is provided for detecting operational irregularities in a thermal energy exchange system. The system comprises a first side that generates heating/cooling, a second side that consumes heating/cooling, and an interface formed by one or more substations that facilitate thermal energy exchange between the two sides. The method includes obtaining operational data from one or more sensors monitoring a district heating substation over a plurality of time intervals, and utilizing a machine learning model that is pre-trained on mechanical fault signature data to extract features indicative of fault conditions. The model is then fine-tuned with fault-indicative data obtained from the district heating substation to adapt it to the specific operating domain. Subsequently, the fine-tuned machine learning model is applied to feature vectors extracted from the operational data to determine whether the substation and/or any coupled downstream heating/cooling systems are in a faulty or normal operational state, and an indication of a fault is output when a faulty state is detected.

Anmelder

Firmen

VITO NV
Cheddad, Abbas
Alawadi, Sadi
Ghazi, Ahmad Nauman

Land

🇧🇪 Belgien

Fachgebiete