Electricity consumption forecasting models for administration buildings of the UK higher education sector
Introduction
This research attempts to identify the relationship among the daily electricity consumption of an administration building and weather parameters such as solar radiation, ambient temperature, wind speed and relative humidity and also with the type of week day (i.e. either it is a working day or non-working day) for the purpose of predicting the building's future electricity consumption. Electricity is a key energy source and plays an important role in facilitating a country's economic development [1]. In the UK, electricity is generated typically in conventional power plants and contributes to 30% of the UK's total carbon emissions [2]. As a result of the UK government's initiatives, mild winter, energy efficiency and tight economic conditions, electricity consumption in the UK in 2011 decreased by 6.9% compared to the 2007 consumption level [3]. Energy consumption in buildings makes up over 40% of all UK energy use [4]. The higher education (HE) sector in the UK spent £400 m on its energy bills in 2012 [5]. Fig. 1 shows the sector's annual energy consumption for 2009, 2010 and 2011. In 2011, electricity was directly responsible for 63.4% of the HE sector's total carbon emissions [6].
In a typical HE campus, different building types have different energy consumption profiles. e.g. academic or administration buildings will require energy during the day time only whereas a student residential hall or a chemical laboratory will require energy throughout the day. Therefore, for the evaluation of potential energy usage, it is necessary to understand the types of building and categories of space in a university campus. Based on five years data (from year 2003–2004 to 2007–2008) collected from 70 UK universities, the academic space (teaching and research) has seen a 2% reduction from 44.2% in 2003–2004 to 42.2% in 2007–2008. During the same period, the second largest space category, i.e. administration space has risen from 24.8% to 26% [7]. Fig. 2 presents a breakdown of different space types in a typical HE campus in the UK for the period 2007–2008. It is apparent that after academic facilities which occupy 42% of a campus space, administration facilities are the second largest space category with a share of 26% in the UK [7].
Electricity consumption in administration buildings is mainly linked with heating, ventilation and air conditioning (HVAC) components, lighting, IT equipment, lifts and other equipment. A small proportion of daily electricity consumption occurs in the appliances used for cleaning, and charging of mobile devices. Major components of a HVAC system include air conditioners, air handling units, fans, pumps and boilers. In the UK, HVAC components are the major electricity users (41%) in administration buildings, followed by lighting (33%), IT equipment (15%) and other equipment (11%) [8].
The financial year in the UK universities runs from 1 August to 31 July. Energy managers are responsible for preparing a budget forecast for their university buildings. To calculate the annual budget for electricity purchase, a reliable forecast of electricity consumption is desirable. Other benefits of reliable electricity forecasting include the followings:
- (a)
identification of variables having a significant effect on electricity consumption;
- (b)
identification of electricity saving potential;
- (c)
estimation of electricity consumption for similar types of buildings;
- (d)
policy development and improvement of production and distribution facilities for electricity;
- (e)
for the utility companies, a reliable model helps in understanding the peak and base load demands of different consumers during the different periods of a year.
An administration type building “Technopark” located at London South Bank University (LSBU) has been considered in this study and its daily electricity consumption data was obtained from the office of the energy manager at LSBU for the period 1 January 2007 to 31 December 2013. Two kinds of models have been developed and tested in this study. One model is based on multiple regression (MR) method and the other model is based on genetic programming (GP) method. Both modeling techniques have been comprehensively discussed with their background and with their advantages and disadvantages in Section 2. Their use in the forecasting of energy consumption has also been discussed in detail with relevant literature review.
Section snippets
Background of modeling techniques used
As described in Section 1, a reliable forecast of energy consumption of a building helps the energy managers in a number of ways. Now a day there are a number of building simulation software (e.g. Energy Pro, DOE2) available in the market for the forecasting of a building's energy consumption. However, their use requires a high level of training for the users and simulation requires a comprehensive input data required such as; description of the building layout, constructions, usage,
Description of the Technopark building
Technopark is an administration type building located at the Southwark campus of London South Bank University in London, UK. It comprises of three floors with a gross internal area (GIA) of 7811 m2. Ground and first floors are mainly occupied by the university administration and finance offices. The second floor is mainly occupied by external private companies which lease the space from the university. Fig. 4 shows the front elevation of the Technopark building.
The building's daily operating
Development of models
The details about the development of both MR and GP models are given in this section.
Results and comparison
This section presents the test results for both the models. The predicted test results are analyzed and compared with each other and also with real electricity consumption datasets.
Conclusions
This paper presents two models, i.e. multiple regression (MR) and genetic programming (GP) models for the forecasting of daily electricity consumption of an administration building. Five independent variables have been used for development of both the models. The developed MR and GP models were trained for the electricity consumption data from January 2007 to December 2012 and were tested against actual daily electricity consumption dataset for the period January 2013 to December 2013. Results
Acknowledgments
K.P. Amber and M.W. Aslam would like to thank Mirpur University of Science and Technology, Mirpur AJK, Pakistan for providing PhD sponsorship. Thanks to the Energy Manager at London South Bank University for providing the energy consumption data.
References (34)
- et al.
Carbon and environmental foot printing of low carbon UK electricity futures to 2050
Energy
(2012) - et al.
Determinants of energy use in UK higher education buildings using statistical and artificial neural network methods
Int. J. Sustainable Built Environ.
(2012) - et al.
Multiple regression models to predict the annual energy consumption in the Spanish banking sector
Energy Build.
(2012) - et al.
Regression models for predicting UK office building energy consumption from heating and cooling demands
Energy Build.
(2013) - et al.
Enernet: studying the dynamic relationship between building occupancy and energy consumption
Energy Build.
(2012) Multiple regression models for energy use in air-conditioned office buildings in different climates
Energy Convers. Manage.
(2010)Electricity consumption forecasting in Italy using linear regression models
Energy
(2009)Development of regression equations for predicting energy and hygrothermal performance of buildings
Energy Build.
(2008)- et al.
Predicting electricity energy consumption: a comparison of regression analysis, decision tree and neural networks
Energy
(2007) - et al.
Forecasting energy consumption using a grey model improved by incorporating genetic programming
Energy Conserv. Manage.
(2011)