Combining contaminant event diagnosis with data validation/reconstruction: Application to smart buildings

In this work, a combined sensor data validation/reconstruction and contaminant event diagnosis approach is proposed for Smart Building systems, implemented as a two-stage approach. In the first stage, sensor communication faults are detected and missing data is estimated, in order to provide a reliable dataset to perform contaminant event diagnosis in the second stage. For the first stage, the sensor validation and reconstruction technique is based on the combined use of spatial and time series models. On the one hand, spatial models take advantage of the physical relation between different variables in the system, whilst on the other hand, time series models take advantage of the temporal redundancy of the measured variables, using Holt-Winters time series models. For the second stage, contaminant event diagnosis is based on contaminant detection and isolation estimator schemes, using adaptive thresholds by assuming certain bounds on the measurement noise and the model uncertainty. In order to apply these diagnosis schemes, state-space models have been considered in order to model the contaminant dispersion over the indoor building environment, where the contaminant event is modelled as a fault in the process which needs to be detected and isolated. Finally, the proposed approach is successfully demonstrated for the Holmes House smart building scenario.