Reducing the Operation Performance Gap – improving building simulation tools through data-driven and real-time approaches
George Papachristou, Loughborough University
The existing building stock continues to have high greenhouse gas emissions. Meeting the challenges of decarbonisation will require an improved understanding of how real buildings work in real life settings and how they can be successfully adapted to a low-carbon future. Digital innovations and technologies are growing and becoming integral to many sectors, with much more proliferation and integration likely in the future. For existing buildings this means that multiple sensors and controls will be recording a wealth of real-time time series measurements on all aspects of building performance including meter readings, temperatures, occupancy etc. However the current range of building performance models are not easily able to utilise and react to these measurements, as they were primary developed for early stage design work. This project seeks to address this gap, by contributing new knowledge to the theory of building simulation focussing on real world and real time applications.
Detailed sensor measurements of building performance and occupancy behaviour in 20 homes have been measured for over 12 months as part of the EPSRC-funded REFIT project. The data arises from dedicated sensors, automated meter reading and Smart Home equipment installed in the homes. This dataset includes individual room temperatures, half hourly gas readings, individual radiator temperatures and appliance measurements, and represents one of the most finely-grained building performance datasets of homes collected in recent years. The data collection is ongoing, with plans to collect a further 12 months of data over the period 2014-2015. This dataset will form the basis of this study, incorporating building simulation knowledge and theory with real world measurements to develop new modelling approaches to understanding building performance.
The aim of the project is to develop new approaches to modelling the thermal characteristics of buildings, based on models which can integrate and react to real-time measurements arising from in-situ sensors. The overall goal is to develop new approaches to reducing the Operation Performance Gap. Prototype models will be developed and tested, which can then be used for improved and more extensive prediction of changes to the building construction, the energy systems, the current control strategies and the behaviour of the occupants.
A series of different models will be developed based on a variety of techniques (steady-state, lumped parameter, finite difference, response factor etc.) and complexity (singe/multiple zone, choice of time steps, steady-state or dynamic HVAC models etc.) for the 20 homes. The models will be tested against the measured performance. Using the results new knowledge of the strengths and weakness of different modelling approaches for existing building modelling will be drawn. Additionally, new modelling methods and techniques which have the capability to utilise multiple real-time measurements to reduce the Operation Performance Gap will be developed. The impact using time series measured data has on choices and assumptions of model inputs such as construction properties, ventilation rates and occupancy schedules will be assessed. Lastly, a prototype data-driven real-time model will be developed and tested out on a previously unused portion of the REFIT data. The opportunities of model calibration and tuning in real-time, and the potential to scale-up the approach or a simplified version to larger samples of buildings will be explored. A comparison of non data-driven vs data driven modelling will be carried out in order to highlight the strengths of the new approach.
The expected outcomes include a new model and/or a set of model techniques that can include real-time performance data as part of their prediction algorithms, and which can update and calibrate in real time, to improve the thermal modelling of existing buildings.