Understanding the Reliability of Localized Near

Future Weather Data for Building Performance

Prediction in the UK

Hu Du and Phil Jones Bobo Ng

Welsh School of Architecture Department of Mechanical and Construction Engineering
Cardiff University Northumbria University
Cardiff, United Kingdom Newcastle upon Tyne, United Kingdom
[email protected]

Abstract— Access to reliable site-specific near future weather integration of renewable energy. A recent review by Lazo et al.
data is crucial for forecasting temporally-dynamic building [2] shows that the optimization of energy management with
energy demand and consumption, and determining the state of weather forecasting can generate 15-30% savings in most
on-site renewable energy generation. Often there is a missing link cases. At the individual building level, the use of both thermal
between weather forecast providers and building energy and electric loads prediction technologies with weather forecast
management systems. This short paper discusses the potential to could optimize energy demand, storage and supply, and reduce
conduct building performance modelling using localized high the building response time, thus reducing costs and greenhouse
resolution weather forecast freely available from the United gas emissions. At city scale, high-accuracy multi-point weather
Kingdom Met Office DataPoint service. It creates a great
forecasts can reflect the real environment at urban condition
opportunity for building performance simulation professionals
and building energy managers to re-use site-specific high
such as considering urban heat island effects, therefore it
resolution weather forecast data to predict near future building increases the accuracy of performance prediction in real life
performance at both individual building and city scale. In this urban conditions.
paper, authors have developed a framework of forecasting near Access to reliable site-specific near future weather data is
future building performance and a Matlab script to crucial for all the works mentioned above. Often there is a
automatically gather observed weather data from 140 weather missing link between weather forecast provider and building
stations and weather forecasts for nearly 6,000 locations in the
energy management system. This short paper discusses the
UK. To understand the reliability of weather forecast, three-
hourly forecasts of temperature, relative humidity, wind speed
potential to conduct building performance prediction using
and wind direction are compared with observations from localized high resolution weather forecast.
weather stations. This provides evidences to use the next 24-hour
forecast to predict dynamic building energy demand and A. Weather data for building performance prediction
consumption, and determine the on-site renewable energy Appropriate meteorological data are needed for predicting
generation output. Because of the high accuracy of forecast, the annual performance of energy systems. This led to the
rolling forecast can be recorded on daily basis to construct development of so-called Typical Meteorological Year (TMY)
weather files for locations that do not have weather stations. This data in the USA, or a term of Test Reference Years (TRY)
will increase current 14 locations of the CIBSE weather data to mainly used in Europe [3]. TMY data provides building
nearly 6,000 locations covering population centers, sporting
simulation professionals with a reasonably sized annual data
venues and tourist attractions.
set that holds hourly meteorological values for a specific
Keywords—building performance prediction; near future location.
weather data; Met Office; DataPoint; wind speed; In the USA, the TMY data set is composed of 12 typical
meteorological months (January through December) that are
I. INTRODUCTION concatenated essentially without modification to form a single
The European Energy Efficiency Directive establishes a set year. The 12 selected typical months for each station were
of binding measures in 2012 to help the EU reach its 20% chosen using the Finkelstein-Schafer statistic method[4], which
energy efficiency target by 2020[1]. Under the Directive, all considering the combination of global horizontal solar
EU countries are required to use energy more efficiently at all radiation, dry bulb temperature, direct normal solar radiation,
stages of the energy chain from its production to its final wind speed and dew point temperature, which are the important
consumption. Smart building energy demand response variables for building performance simulation.
solutions considering weather forecasts enable end users to In the UK, the Chartered Institution of Building Services
reduce energy bills, improve their thermal comfort conditions, Engineers (CIBSE) developed TRY and Design Summer Year
increase the efficiency of energy grid and contribute to the (DSY) data sets for 14 cities in the UK. The selection of typical
This study is part of the Sêr Cymru (‘Stars Wales’ in English) program
funded by the Welsh government.

978-1-5090-1846-8/16/$31.00 ©2016 IEEE

months applied by the CIBSE is similar to the selection to four days are as accurate as their one-day forecasts were 30
procedure of TMY data, however it only applies to dry bulb years ago[6], and the horizontal resolution of their forecast has
temperature, global horizontal solar radiation and wind speed. been increased to 1.5 km in 2010 from 150km in 1990.
The DSY data, which consists of an actual 1-year sequence of
hourly data, are selected based on dry bulb temperatures during D. Data sharing Application Programming Interface
the period April–September only. This enables designers to As part of the UK Government’s Open Access movement,
simulate building performance for a year with a hot, but not the Met Office is working towards making the data available to
extreme, summer. the public. Since 2011, the Met Office released 3-hourly site-
specific forecast data feeds for nearly 6,000 locations in the
B. Weather stations UK, which covers population centers, sporting venues and
Comparing with the 14 locations of CIBSE weather data, tourist attractions, through the Met Office DataPoint service
the Met office has access to 140 observation weather stations in [7]. The DataPoint service is an Application Programming
the UK (shown in Fig 1 left). If all data are available to Interface (API) that provides last 24-hour land and marine
building professionals in a right format, there is a potential to observations, lightning strikes frequency map, monthly
reduce the distance between the location of the building to be regional climate anomalies, historic regional climate data and
modeled and the location of weather data used for modeling, more importantly 3-hourly weather forecast for next 5 days.
from hundreds of kilometers to tens of kilometers. However, The files from the DataPoint are in a format of Extensible
currently the full set observation data from all 140 weather Markup Language (XML) which is suitable for web application
stations are only available to few research institutes and they developers to use. Forecasts are regularly updated on hourly
are not in a ready-to-use format that can be fed into building basis and make use of a range of weather models run on the
simulation programs directly. Met Office’s supercomputer.
Often the 140 weather stations are located outside the cities The high spatial resolution forecast has significantly
and they are still tens of kilometers away from the real location increased the density of weather data locations comparing with
of the buildings to be studied in most of cases. For example, for traditional 140 observation stations. For example, weather
buildings in the center of London, Heathrow or Gatwick airport forecasts are made available for around 300 locations near
weather data is used for simulation. In the real life London (shown in green dots in Fig 1 right) from the DataPoint
environment, the temperature difference between the two service, comparing with only 10 observation stations (in red
locations could be up to 10 oC [5] due to urban heat island dots). This provides a great opportunity for building
effect and micro-climate. Increasing the number and density of performance simulation professionals to re-use the Met Office
weather stations could potentially increase the accuracy of data for predicting near future building performance. It reduces
building simulation, however it is often limited by its cost and the distance between the location of building and the location
time for data collection. Typically it needs 20-30 years of weather data from tens of kilometers to several kilometers or
recorded data to produce a set of TRY/TMY data. Therefore meters, therefore it increases the accuracy of building
there is a need to consider the alternative source of weather performance prediction.
data for building simulation.
II. THE CHALLENGES AND RESEARCH METHOD
Based on the skill sets and availability of recourses, the
authors proposed the following framework (illustrated in Fig 2)
for predicating near future building thermal performance and
energy consumption using DataPoint, Matlab and EnergyPlus.

Fig. 1. 140 observation stations in the UK (red dots) and around 300 sites
near London having weather forecast data (green dots)

C. The UK Met Office Weather forecast

Forecasting weather involves predicting future variations in
a complex system. The UK Met Office started making
regularly use of numerical method to forecast weather in 1959. Fig. 2. The proposed framework for near future building performance
The advances in technology, such as the recent Cray XC40 predication
supercomputer, and better understanding of weather events Due to the paper limit, the alterative frameworks,
have helped them to deliver huge improvements in forecasting programming language and simulation engines are not
accuracy over recent decades. Today's Met Office forecasts out
discussed here. For early stage research, the combination of TABLE I. WEATHER PARAMETERS IN THE FORECAST AND
OBSERVATION XML FILES
tools above has been proved to be efficient at lab condition.
Symbol Description Forecast Observation Unit/Notes
At current stage of research, there are three main challenges i Location ID √ √ /
in making use of the Met Office DataPoint site-specific lat Latitude √ √ degree
forecast data for building performance prediction: lon Longitude √ √ degree
W Weather type √ √ 0-30
• Automatically collecting forecast and observation data T Temperature √ √ o
C
on regular basis. Relative
H √ √ %
humidity
• Converting the collected Extensible Markup Language S Wind speed √ √ mph
(.xml) weather files into a format suitable for building 16-point
simulation professionals. D Wind direction √ √ compass
direction
• Understanding the reliability of localized forecast for G Wind Gust √ √ mph
building performance modelling. This was not possible P
Sea level
√ hPa
until large sets of site-specific forecasts were made pressure
available to the public. It is a common practice to V Visibility √ √ m
Pressure F, Falling; R
understand the reliability when dealing with unknown Pt
Tendency
√
Rising, S Steady
large data sets. Dew point o
Dp √ C
temperature
In previous research [8], authors have tackled the first two
U Solar UV Index √ 1-11
challenges by developing a method of collecting weather Precipitation
forecast data for nearly 6000 locations and observation data Pp √ %
Probability
from 140 weather stations in the UK through the Met Office Feels Like o
F √ C
DataPoint service API. The method has been successfully Temperature
translated into a Matlab script to automatically gather data on Wet bulb
Missing
Linked to other
daily basis because of the rolling updates of weather forecast. temperature parameters
Solar radiation Missing Related to others
The Matlab script could capture the latest weather forecast and
stored on a local computer for future use. This short paper B. Temperature
particularly focuses on the third challenge – to understand the
Figure 3 shows the small difference between observed 3-
reliability of forecast for building performance modeling. The
hourly temperature and the forecast (general less than 1 degree
following research method has be employed:
• Develop a checklist of weather parameters available in
forecast and observation XML files and identify the
difference and additional parameters needed for
building performance prediction.
• Compare key common weather parameters existed in
both forecast and observation files and quantify the
difference.

III. EARLY RESEARCH RESULTS

A. Parameters available from DataPoint API

The Met Office DataPoint API outputs (XML files) follow Fig. 3. Temperature Forecast vs. Observation, Dec2015 Heathrow
a defined rules for encoding documents in a format which is
both human-readable and machine-readable. As illustrated in Celsius). Note that the forecast is constructed with the most
Table 1, a range of weather related parameters are available in recent 24-hour forecast (a rolling forecast over that month).
these files. Note that sea level pressure, pressure tendency and The observations were taken from Met Office weather stations.
dew point temperature exist in the observation file, but not in
forecast file. Solar UV Index, Precipitation Probability, Feels C. Relative humidity
Like Temperature are only available from forecasts, not from The same format of comparison was made for relative
observations. A key weather parameter for building humidity in Figure 4. Forecasts generally follow observations
performance modeling - solar radiations (including global, with a difference of less than 10% in general.
direct and diffuse) are missing from both files.
The following results section compared the forecast and D. Wind speed
observation of temperature, relative humidity, wind speed and The forecast wind speeds also closely follow observations
wind direction which exists in both files (highlighted in grey in (shown in Figure 5). The differences are generally less than 5
Table 1). The comparisons were made using data of December miles per hour.
2015, London Heathrow which has a dedicated weather station.
 Fig. 7. Mean and standard deviations of the differences between
forecasts and observations, December 2015, Heathrow
Fig. 5. Relative Humidity Forecast vs. Observation, Dec2015 Heathrow
IV. CONCLUSION
Through the detailed comparison of 4 observed weather
parameters and their forecasts, the authors understood the
accuracy of near future weather forecast. This provided
confidence in creating ‘true’ weather data for locations that do
not have weather observation stations. The work also provided
evidences to use the next 24-hour forecast to predict dynamic
building energy demand and consumption, and determine the
state of on-site renewable energy generation.
Future work will focus on urban scale building energy
management using high spatial resolution forecast data, urban
heat land effect study and site-specific renewable energy
generation prediction.
Fig. 4. Wind Speed Forecast vs. Observation, Dec2015 Heathrow
ACKNOWLEDGMENT
E. Wind direction The authors thankfully acknowledge the DataPoint service
Both forecast and observation of wind direction use 16 provided by the UK Met Office. The authors also would like to
point compass as unit. In order to compare the data, the text thank W. Falkena at Delft University of Technology and others
descriptions of wind direction have been converted to values in for sharing the ‘xml2struct.m’ Matlab function.
range of 0-360 degree. The results (Figure 6) shows that
forecasts are very close to observations. REFERENCES

[1] European Parliament and Council, "Directive 2010/31/EU of the

European parliament and of the council on the energy performance of
buildings," Official Journal of the European Union, vol. L, pp. 13-35,
2010.
[2] D. Lazos, A. B. Sproul, and M. Kay, "Optimisation of energy
management in commercial buildings with weather forecasting inputs: A
review," Renewable and Sustainable Energy Reviews, vol. 39, pp. 587-
603, 11// 2014.
[3] A. Argiriou, S. Lykoudis, S. Kontoyiannidis, C. A. Balaras, D.
Asimakopoulos, M. Petrakis, et al., "Comparison of methodologies for
tmy generation using 20 years data for Athens, Greece," Solar Energy,
vol. 66, pp. 33-45, 1999.
[4] G. J. Levermore and J. B. Parkinson, "Analyses and algorithms for new
Test Reference Years and Design Summer Years for the UK," Building
Fig. 6. Wind Direction Forecast vs. Observation, Dec2015 Heathrow Service Engineering, vol. 27, pp. 311-325, November 1 2006.
[5] M. Santamouris, N. Papanikolaou, I. Livada, I. Koronakis, C. Georgakis,
A. Argiriou, et al., "On the impact of urban climate on the energy
Further analysis has been conducted to quantify the consumption of buildings," Solar Energy, vol. 70, pp. 201-216, // 2001.
distribution (Figure 7) of the differences between forecasts and [6] Met Office. (2016). Developing the science for long range weather
observations. The results are in line with previous findings that forecasts. Available: http://www.metoffice.gov.uk/news/in-depth/long-
1-day ahead forecasts of temperature, relative humidity, wind range-forecasting
speed and direction are very close to observations. Due to the [7] Met Office. (2016). Met Office DataPoint. Available:
http://www.metoffice.gov.uk/datapoint
page limit, study for longer period, other locations and [8] H. Du, P. Jones, and S. Lannon, "Creating localised near future weather
comparisons at urban/rural environment will be reported in a data for predicting the performance of buildings in the UK," in Clima
journal paper. 2016 -12th REHVA World Congress, Aalborg, Denmark, 2016.