A method and system for detecting operational irregularities in a thermal energy exchange system, and a method of training a machine learning model

A system for detecting operational irregularities in a district thermal energy exchange system, wherein the thermal energy exchange system comprises a first side, a second side and an interface between the first side and the second side, wherein heating/cooling is generated at the first side upstream to the interface, wherein heating/cooling is consumed at the second side downstream to the interface, and wherein the interface is provided by means of a plurality of substations connected to the first side and the second side, each substation being configured to facilitate thermal energy exchange between the first side and the second side, wherein the system comprises a processing unit configured to:

• obtaining measurement data indicative of one or more operational parameters of the first side, wherein the measurement data is collected by means of a measurement system;
• and
• monitoring the measurement data so as to identify operational irregularities at the interface and/or the second side, wherein the processing unit comprises:
• a cluster based processing module configured to group the plurality of substations in clusters according to one or more predetermined criteria, to rank the plurality of substations in accordance with a set of key performance indicators, and to select a first set of substations corresponding to worst-performing substations in the ranking of substations;
• and
• an instance based processing module configured to model each substation by means of a trained machine learning model, wherein the trained machine learning model is configured to provide a predicted output based on one or more inputs indicative of measurements, wherein the instance based processing module is configured to compare the predicted output of the trained machine learning model to a measured output, and to select a second set of substations for which the difference between the predicted output and the measured output exceeds a predetermined threshold value;
• and
• wherein the processing unit is configured to identify substations present in both the first and second set of substations as substations being associated to at least one faulty subsystem, the at least one faulty subsystem being located in the substation itself and/or at the second side connected to the substation;
• and
• wherein cluster based processing is performed periodically, and wherein the instance based processing is performed continuously.