Detecting central heating boiler malfunctions using smart-thermostat data

Abstract

With the advent of smart thermostats like Toon®, detailed information about operation and usage of central heating boilers has become easily available. However, this information is not used in a systematic way by most companies including Eneco, and the wealth of information within is not available to the users. This has a few drawbacks. For instance, when a mechanic is sent to repair a broken boiler, the mechanic has to rely on the data provided by the callcenter receiving the call from the customer. This data is often missing, incomplete or incorrect. This means that no reliable information about which parts to bring and how long the repair will take is available a priori. Secondly, some malfunctions could have been easily resolved by the end-user, for instance by refilling the system with water. This research aims to provide a starting point in a systematic and automated approach in analysing the behaviour of the boiler by detecting malfunctions as they occur. To do so, a mathematical model of a house is designed. On this model an Extended Kalman Filter is built which monitors important parameters of the system in real-time. The estimated parameters can in future research be used as features in a more complete fault detection and identification scheme. The filter has successfully been tested against simulated faults, and shows promising results when applied to real data.