Machine Learning Algorithms for Abnormality Detection of Chilled Water Systems (OR-16-C079)

Many facility departments have installed energy meters seeking to pin point how, where and when energy is being inefficiently utilized. This has led to the recording of vast amount of data over extended periods of time, which makes it very difficult to manage and manually analyze. Luckily, techniques in machine learning have shown promising results in automated knowledge discovery making it more and more crucial when large data is at hand. This paper applies machine learning (ML) algorithms to detect abnormalities in chilled water systems (CWS) at building level. Two abnormal situations are pursued: chilled water sensor misreadings and low thermal efficiency in terms of delta T. The visualization of building chilled water historical data provides general trends and an initial identification of the building abnormalities; this visualization also helps to lay down the requirements for the abnormality detection algorithms, and eventually, their selection.