Outlier detection and data filtering for wireless sensor and actuator networks in building environment

The management systems of smart buildings aim to provide a comfort environment while reducing energy consumption. Efficient energy management requires changes not only in the way the energy is supplied but also in the way the devices are controlled. Such system requires various sensing data, data analysis and optimization algorithms in decision making. Wireless sensor networks have been employed widely for data collection. However, these sensors due to different factors can cause variations in raw sensor measurements, inaccurate information transferred to the base station may result in misguidance to the control system. The main objective in outlier detection is to find the data that is deviating from the other data based on the algorithm techniques. In this paper, Hodrick Prescott filters are adopted for reducing the noise and errors from real sensor data. Experiments and analysis conducted show that output data is more stable when compared with the results obtained by other methods.