Transcript Poster given at annual Energy Institute Colloquium

ANNUAL COLLOQUIUM 2011
The Danger of Data. Assessing the availability and quality of data for tertiary
sector energy demand forecast models.
Ed Sharp PhD candidate - LoLo CDT – UCL energy institute – [email protected]
Introduction
Energy demand forecasting has been carried out since the oil crisis in the 1970’s and the subsequent realisation of the need to match supply to demand. The
methodologies employed in these models have developed iteratively alongside the associated applications. Despite the resultant incorporation of a complex and varied
number of methods models still remain data intensive and dependent on the availability and quality of input data to some extent. This poster represents the result of a
request by EDF to identify and analyse datasets that would be beneficial to their model forecasting demand in the tertiary sector across Europe. Key drivers of demand
which are used as inputs into these models were identified through EDF’s models and the literature, common causes of variation between sources of these variables
were then explored. The results below represent the findings in the context of the UK, data availability is shown as a time series with the maximum range shown as the
percentage of the mean value for selected years.
Variable Data
Causes of Variation Between Data Sources
Variable Data
Sectoral classification: The predominant cause of
variation between data sources is the lack of
standard classification of the sector. Most widely
used are The United Nation’s International Standard
Industrial Classifications (ISIC) and Eurostat’s
Nomenclature statistique des Activités économiques
dans la Communité Européenne (NACE). However
the former includes significant sub sectors omitted
from the latter (agriculture, forestry and fishery).
This results in the substantial divergence seen in the
Energy Consumption, Employee numbers and GVA
variables where those sources using the ISIC
scheme provide significantly higher values than
those using NACE.
Other causes of variation between data sources
include: differing methods of calculation in
particular GDP where complicated techniques
require
expert
knowledge
to
understand.
Inconsistent methods of harmonisation for
example changing population values used to create
per capita GDP values. Variations in the precision
with which data is stored, for example where
population statistics are stored to the nearest person
created an unjustified perception on the accuracy of
the data. Semantic inconsistencies for example
where energy consumption is not explicitly referred
to as either primary or delivered. All of these issues
would not be a problem if they were clearly stated
which highlights the lack of quality metadata.
Conclusions
The above sources of variation are exacerbated by a lack of data for certain variables, significantly floor space which is the key driver of energy demand in the nondomestic sector. All of these issues could be mitigated by some simple measures. These include the creation of a centralised repository for the data which could make
the reason for divergence clear and transparent through the creation of high standards for metadata. Beyond this many variables would benefit from the introduction
and use of mandatory classification schemes, a measure that would be significantly more costly and complicated to implement. The most important lesson of the
research however is to utilise online sources critically and beware of the danger of data.
PhD Research - The spatiotemporal patterns of energy demand and supply in the UK
Aims and Objectives:
1. Create a simulation model depicting UK energy demand into the future
using a fine spatial and temporal scale.
2. Creation of demand and supply profiles at approximately 0.5 degrees
spatial and 1 hour temporal resolution.
3. Model the influence of climate on these profiles using measured
meteorology and widely used scenarios.
4. Analyse the difference between supply and demand at alterable spatial
scales.
First steps:
1. Assess and summarise current state of related research
2. Download, harmonise and consolidate existing data on variables affecting demand
including but not limited to meteorological, socio-demographic, technological and
physical.
3. Characterise demand over space and time in a recent historical year.
4. Develop simple models to depict demand based on the prior steps.
5. Iteratively advance these models in sophistication and robustness.
6. Repeat process for supply using existing models and structures.
7. Analyse difference between supply and demand at alterable spatial scales.
London Loughborough Centre
for doctoral research
in energy demand
Central House
14 Upper Woburn Place
London, WC1H 0NN
www.lolo.ac.uk