Advanced Analysis of Building Energy Performance using Computational Intelligence Approaches

Lead Research Organisation: Loughborough University
Department Name: Civil and Building Engineering


The UK government aims to achieve a 60% reduction in UK carbon emissions by 2050. Energy efficiency is a key component of the UK Government's new climate change strategy with 50% of current carbon dioxide emissions resulting from energy use in buildings. Central to this approach is the ability to monitor and control energy consumption accurately. Building performance is often measured through the metered energy use of the whole building, and/or through the more detailed monitoring of the individually controlled processes by the building management systems (BMS). Automatic meter reading (AMR) systems have existed for several decades which can provide accurate consumption data, typically at half-hour intervals. However, with the recent introduction of a legal framework for supplier-independent metering, many initiatives are in place to install such meters in large numbers of industrial and domestic sites. This has resulted in a broad increase in uptake of advanced AMR systems and their associated services and the emergenceof the smart metering paradigm. Unfortunately, it is not clear how this data is actually going to be used. The advanced metering pilot of the Carbon Trust has produced a breakdown of recommendations derived from introducing advanced metering into SMEs: 15% of recommendations are the result of analysis of the data alone; 25% were the result of analysis of the data combined with advice by phone and email, and 60% required personal contact with energy experts. In other words, without additional advice, only a small fraction of the potential savings can actually be achieved.This project aims at those 85% of recommendations, investigating how Computational Intelligence (CI) techniques can help in generating the analysis needed to gain the full benefit from the data.Computational Intelligence techniques, like artificial evolution and neural network, are perfectly suited to the analysis of metering data. In general, the techniques require very little domain knowledge, can automatically acquire domain knowledge, and are tolerant to noise. Through machine learning, they can adapt to individual sites, and to changes over time. This research will investigate the potential for using Computational Intelligence methods in the automatic analysis of metered building energy data and how these methods can be used to provide maximum benefit to a large range of energy users. The project will concentrate on identifying CI techniques that can be broadly applied to wide ranges of sites, that can be largely automated, require minimal training, system setup and manual data entry.The project will establish potential techniques, benefits, and requirements, as well as the extent to which building specific knowledge is required in the analysis together with the impact of time-varying boundary conditions (the weather and occupant driven heatloads). While this proposal concentrates on energy consumption by buildings in commercial and industrial sites, the research will also provide valuable insights for energy consumption of domestic buildings, energy consumption in industrial processes, and any other metered utility consumption.


10 25 50
Description The project investigated the use of calibrated models to identify the separate base, heating, and cooling energy use from metered total energy use. Both steady-state and simplified lumped capacitance models where evaluated, with both models being calibrated to the metered data using an Evolutionary Algorithm (EA). The calibration task is challenging in terms of the number of calibration variables, but the EA proved to be effective in calibrating the models. The (dynamic) lumped-capacitance model showed some improvement in model accuracy over the steady-state model. As well as identifying the separate load elements, the model results could also be interpreted to identify the different time-dependent operating modes of the energy systems. The research constituted a successful proof of concept, with further challenges being associated with the formulation of models that have a tangible physical meaning. Formulation of such models would allow the approach to be used directly in a comparison of performance enhancements resulting from the refurbishment of the building energy systems (by examining the changes in the model parameters from pre and post refurbishment calibration of the model).
Exploitation Route The approach investigated in the research has a potential application in the monitoring of building energy use, and in particular, the extent to which energy use is divided between a base-load, heating load, and cooling loads; the approach can also be used to identify the time-dependency of the energy system operation. With further development of the models, the approach may be used in evaluating the impact of building refurbishment through the examination of the model parameters calibrated with data from pre and post refurbishment periods. The approach might be exploited through its development by building energy control system providers and providers of energy facilities management software.
Sectors Construction,Digital/Communication/Information Technologies (including Software),Energy

Description The findings have been used in related research by the lead partner, Birmingham University (see EP/F062567/1)
First Year Of Impact 2009
Sector Digital/Communication/Information Technologies (including Software)
Description AEC Environmental Ltd 
Organisation Airborne Environmental Consultants (AEC)
Country United Kingdom 
Sector Private 
Start Year 2008
Description Birmingham 
Organisation University of Birmingham
Country United Kingdom 
Sector Academic/University 
PI Contribution The research grant EP/F062222/1 held by Loughborough University was subsidiary to the main grant held by Birmingham University EP/F062567/1. Birmingham University were project instigators and managers, and provided all data and computer science input to the project.
Collaborator Contribution Loughborough University provided knowledge on building energy and building energy systems, this being the focus of the modelling methods to be developed and evaluated by Birmingham University. The funding was to cover overheads for the Loughborough investigators time and for travel.
Impact The project resulted in an understanding of the applicability of computational intelligence methods in building energy monitoring and analysis, but no further tangible outcomes have resulted from the collaboration.
Start Year 2008
Description Optima Energy Management 
Organisation Optima Energy Management
Country United Kingdom 
Sector Private 
Start Year 2008