Visual Pattern Recognition as a Means to Optimising Building Performance?

摘要

The move towards the use of smart systems to record data about the performance of building presents an opportunity to examine patterns of behaviour, which shed light on its performance peaks and troughs. Answers often lie within these peaks and troughs. For example, smart systems are used to record in-use data such as room temperatures, thermal comfort, and lighting usage. Each system is designed for an expected 'behaviour pattern' that is bounded by a 'threshold' boundary. This is the expected performance the system is designed for. Any deviations in this performance may indicate of a system malfunction, its overuse or underuse due to unexpected usage of building space/s (e.g. large number of visitors, doors/windows being left open). Performance when bounded within the threshold limits would be considered to be 'normal' or 'expected'. Thus, answers fall within four categories of performance behaviours. These include system malfunction, system overuse, system underuse, and normal performance. As data and information are accumulated over a period of time they present opportunities to observe system behaviour patterns and present opportunities to map these patterns and classify within learning clusters of 'expected' and 'unexpected' thresholds. Doing so would enable building owners to truly understand the building, so performance can be firstly understood and then optimised. Thus, an unusual activity that is significantly beyond the expected 'norm' would present an opportunity to learn about the building so a healthy function can be determined and maintained. The generation of large datasets through extensive monitoring has created a potential environment in which big-data style analytics could be applied for holistic performance assessment and pattern recognition. This research builds on the work previously completed by Gerrish et al. [1-4, 21] and utilises techniques of visualisation to demonstrate such behaviour patterns and presents learning opportunities for optimal performance. This is demonstrated through visualisation of energy performance data for a case-study building in the UK.