In the context of sensor data generated by Building Management Systems (BMS), early warning signals are still an unexplored topic. The early detection of anomalies can help preventing malfunctions of key parts of a heating, cooling and air conditioning (HVAC) system that may lead to a range of BMS problems, from important energy waste to fatal errors in the worst case. We analyse early warning signals in BMS sensor data for early failure detection. In this paper, the studied failure is a malfunction of one specific Air Handling Unit (AHU) control system that causes temperature spikes of up to 30 degrees Celsius due to overreaction of the heating and cooling valves in response to an anomalous temperature change caused by the pre-heat coil in winter period in a specific area of a manufacturing facility. For such purpose, variance, lag-1 autocorrelation function (ACF1), power spectrum (PS) and variational autoencoder (VAE) techniques are applied to both univariate and multivariate scenarios. The univariate scenario considers the application of these techniques to the control variable only (the one that displays the failure), whereas the multivariate analysis considers the variables affecting the control variable for the same purpose. Results show that anomalies can be detected up to 32 hours prior to failure, which gives sufficient time to BMS engineers to prevent a failure and therefore, an proactive approach to BMS failures is adopted instead of a reactive one.

In large buildings, linking heating, cooling or ventilation systems between themselves and to physical spaces is a very time-consuming task that requires highly skilled engineering knowledge, as all these systems are interconnected and they have a certain influence to each other (ventilation systems are often connected to heating and cooling), which often makes task of locating the sources of error or anomalies very time consuming and difficult as they are performed manually. A different approach would be to work out relationships and equipment linkage from time series data provided by the sensors, thus inferring equipment links from which anomalies can be traced back to the source more easily. This paper proposes a data-based solution to obtain equipment relationships based on cross-correlations to relate Air Handling Units (AHUs) to their respective areas of operation. We also propose a methodology, in particular for AHUs, to identify whether or not to trust correlations based on the difference between supply and return temperature. A case study is presented based a large building with 16 AHU systems.

In building management systems (BMS), a medium building may have between 200 and 1000 sensor points. Their labels need to be translated into a naming standard so they can be automatically recognised by the BMS platform. The current industrial practices often manually translate these points into labels (this is known as the tagging process), which takes around 8 hours for every 100 points. We introduce an AI-based multi-stage text classification that translates BMS points into formatted BMS labels. After comparing five different techniques for text classification (logistic regression, random forests, XGBoost, multinomial Naive Bayes and linear support vector classification), we demonstrate that XGBoost is the top performer with 90.29% of true positives, and use the prediction confidence to filter out false positives. This approach can be applied in sensors networks in various applications, where manual free-text data pre-processing remains cumbersome.