Automatic Process Monitoring in a District Heating Substation Utilizing a Contextual Shewhart Chart

Fault detection methods play a key role in enabling proactive maintenance in district heating systems. Faults are estimated to cause around 40 percent of energy consumption and it is therefore critical to employ methods to decrease this unnecessary waste of energy. For detection of these faults, a data-driven process monitoring methodology is presented which uses a modified version of the Shewhart chart, which is called contextual Shewhart chart. A process variable’s normal operating range often shifts, and this can be due to external factors (e.g., outdoor temperature), and this is not captured by a regular Shewhart chart. However, the proposed contextual Shewhart chart can capture these effects by using a so-called contextual variable to vary the acceptable operational ranges, based on the identified external factor. The methodology has been applied to real data from a district heating substation and has shown promising results.