Powelectrics are a company providing sensor, telecommunications and computing solutions, thus empowering the industrial internet of things through accurate and affordable sensor data. www.powelectrics.co.uk
Anomaly detection is the process of identifying novel or unexpected observations within a dataset. This process has been well studied in various domains but there has been limited application of these techniques to the Internet of Things (IoT).
This project applies anomaly detection techniques to continuous data streams from IoT, which could automate detection of power outages and the early detection of faults for machinery to allow for scheduled maintenance. The results could also be used to detect anomalous states in the data from sensors in buildings
The project can use data from SEND to train the detection techniques based upon real fault data. As a demonstrator of an “at scale” integrated smart energy network there are numerous opportunities to integrate streaming anomaly detection into the data collection and enrichment process potentially moving towards a more online and automated system.
Project progress (June 2019):
- Literature review of data analytics, anomaly detection and IoT completed.
- Performed replication studies of two major anomaly detection techniques applied to streaming data.
- Using open source anomaly detection methods to design novel detection benchmark.
Business Contact(s): Dave Oakes
Keele Graduate Researcher: Andrew Cook