Solar Energy 81 (2007) 102110 www.elsevier.

com/locate/solener

The London Heat Island and building cooling design

Maria Kolokotroni *, Yuepeng Zhang, Richard Watkins
Brunel University, Mechanical Engineering, School of Engineering and Design, Uxbridge, Middlesex, UB8 3PH, UK Received 3 October 2005; received in revised form 1 June 2006; accepted 7 June 2006 Available online 24 August 2006 Communicated by: Associate Editor Matheos Santamouris

Abstract Londons urban heat island increases the mean air temperature which aects the demand for heating and cooling buildings. Measured air temperature data have been used as input to a building energy simulation computer program to assess the heating and cooling load of a typical air-conditioned oce building positioned at 24 dierent locations within the London Heat Island. It is found that the urban cooling load is up to 25% higher than the rural load over the year, and the annual heating load is reduced by 22%. The eect of raised temperature and urban context are assessed separately, and the sensitivity of the net impact to the internal gains in a building is determined. For the estimation of peak cooling demand, we propose hourly temperature corrections based on radial distance from Londons centre to be applied to standard published temperatures for the region. For more detailed investigations over the cooling season a range of models is available. These are reviewed in this paper and we describe preliminary results of an Articial Neural Network (ANN) model that predicts location specic hourly temperatures for London, taking into account radial distance from central London, hourly air temperature measured at the meteorological station and associated synoptic weather data. 2006 Elsevier Ltd. All rights reserved.
Keywords: Heat island; London; Energy demand; Building design

1. Introduction Heat islands are well-established consequences of the urban environment. In general, urban centres are warmer than the surrounding area and this can be benecial, or not, in terms of the energy used in providing comfortable conditions in buildings. In hot or cold climates the annual balance of the impact of a heat island may be clear, but in temperate regions the reduced heating season loads may signicantly oset the higher cooling loads in the summer. This paper addresses this impact in the case of London. First and in order to demonstrate the eect, the heating and cooling loads for a typical oce building have been assessed using measured hourly air temperature data. The methodology of data collection and quantication of the
*

Corresponding author. Tel.: +44 1895 266 688; fax: +44 1895 256 392. E-mail address: maria.kolokotroni@brunel.ac.uk (M. Kolokotroni).

London Urban Heat Island has been recently reported previously in Watkins et al. (2002a) and Kolokotroni et al. (2006). Using these measured air temperature data as one of the input parameters together with additional weather data and building construction and operational characteristics, the energy performance of a typical air-conditioned oce building has been modelled using an energy simulation programme in a variety of urban contexts to determine the eect of the heat island over a year. It will be shown that location specic air temperature has a marked eect on energy consumption and therefore, designers should take this into account. The paper provides information of how location specic urban air temperatures in London can be considered for the calculation of peak cooling demand (Section 4) and proposes a model for hourly air temperature calculation based on available meteorological station data and location of site within the Urban Heat Island (Section 5).

0038-092X/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.solener.2006.06.005

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110

103

2. Modelling the energy demand of a typical oce building Much research has been carried out investigating the eect of increased air temperature in urban areas on energy demand by buildings. In London, Chandler, using data from 195160, found a reduction in annual heating degree days (base 15.6 C) of about 10% between central London and a rural area (Chandler, 1965). More recently, climate change has been linked to Urban Heat Island and its impact on building environmental design. Graves et al. (2001) have presented temperature data for designers and guidelines for reducing the eect of the heat island in London. GLA (2002) discusses the eect of climate change including Urban Heat Island on London and CIBSE (2005) examines in detail the eect of climate change on the indoor built environment in the UK. Elsewhere, in Munich, Brundl reported that annual heating degree days (base 15.0 C) were 14% lower in a central urban area compared to a suburban one (Brundl and Hoppe, 1984). Energy use for cooling was not considered because summer cooling was not prevalent. Energy requirements for air-conditioning are considered higher than for heating (Landsberg, 1981), and both Landsberg (1981) and Taha (1997) concluded that the elevation of urban temperatures imposes a net energy penalty in several American cities because of increased cooling requirements. Shading eects of surrounding buildings are often neglected in calculating the loads in a building. This can lead to over-sizing air-conditioning equipment, and in general, neglects a basic aspect of urbanization. In central Athens, the annual cooling load for an apartment block with windows shaded 50% of the time was found to be 15 50% higher than when modelled using weather data from an open site on a hill 2 km away (Hassid et al., 2000). Also in Athens, where the mean heat island intensity exceeds 10 C, it was found that the cooling load of urban buildings may be doubled and the peak electricity load for cooling purposes may be tripled especially for higher set point temperatures. During the winter, the heating load of central urban buildings is found to be reduced up to 30% (Santamouris et al., 2001). Akasaka et al. (2002) report that due to the heat island the cooling load of Tokyo has increased about 20% since 1900 with a corresponding decrease of about 40% for heating load. Strategies to reduce cooling energy use in buildings due to the heat island are proposed by Akbari and Konopacki (2005) classied to direct (reducing heat gain through the building shell) and indirect (reducing the ambient air temperature). They have developed summary tables for the US sorted by heating and cooling degree-days based on simulations to quantify the eect of these strategies. In this research project, a commercial building energy simulation computer program (TAS) has been used to predict the impact of the heat island and the urban environment on energy demand by buildings. Some preliminary results were reported in Watkins et al. (2002b) and Kolokotroni et al. (2004). In this paper, additional analysis is

included (Section 3) and new data on proposed temperature corrections suitable for summer cooling design assessment are presented (Section 4). 2.1. External conditions One years measured hourly air temperature data from 24 locations within London were combined with regional weather data to form 24 weather les for the building energy simulation model. Wind speed and direction data were obtained from Heathrow airport, 23 km WSW of central London, and humidity, cloud cover, and solar radiation data from the London Weather Centre, central London. These regional data were assumed to be applicable to all 24 sites an approximation. Wind speed can aect inltration and convective heat transfer. In this study of an air-conditioned building, both the inltration and ventilation rates are scheduled and are thus independent of wind-speed. The convective heat loss coecient is varied by the model according to the hourly regional wind speed, but this is not modied here for urban settings. 2.2. Typical Oce building The dominant type of air-conditioned building in London is the oce, and to represent this a typical design has been selected from a set of designs widely used in the UK in comparative energy studies. It is a standard, as distinct from prestige, air-conditioned oce termed the ECON 19/3 building; taken from the Energy Consumption Guide 19 (BRECSU, 1999). It is a three storey open plan building, 9 m high, 30 m long and 15 m wide orientated with the longer sides facing north:south, with 60% glazing on these facades. There is clear double glazing with no shading. The end walls are unglazed. Walls and roof are concrete with insulation. Intermediate oors are of concrete with false ceilings, and the ground oor is uninsulated. The walls have a solar absorptance of 40% and the roof 65%. The surrounding land was set to have 20% ground reectance to solar radiation. An air-conditioning system (vapour compression) and heating system (gasred) operates from 06:00 to 18:00 (to include pre-conditioning) maintaining the internal air temperature between 20 and 24 C. Fresh air is supplied during occupied hours with a total air change rate of 1.1/hour (including inltration of 0.5 ach/hour at all times). Internal gain from lights, occupants and plant is 43 W/m2. 2.3. Urban contexts In urban areas, buildings usually experience a degree of over-shadowing which reduces solar gain. This eect has been modelled by surrounding the test building with neighbouring blocks to the same height (9 m) at a varying distance depending on the appropriate site categorization (Table 1). The spaces formed between the buildings, the street gorges, were given a height to width ratio that varied

104 Table 1 Criteria used for categorizing sites Ct 1 2 3 4 5 6 7 8 H/W ratio of street, x x=0 x=0 x=0 0 6 x < 0.3 0.3 6 x < 0.5 0.5 6 x < 1 16x<2 P2 Description

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110

Rural elds, or large park, or trees Housing near park or eld Urban derelict or unbuilt area Low density residential area Medium density urban area High density urban area Very high density urban area Exceptionally high density urban area

London and warmer, but has heavy over-shadowing (category 7, Table 1). Both these sites have particularly low cooling loads for their mean temperatures. Site C is 3.2 km from London in the centre of the heat island, warm, but in a more open site (category 5, Table 1), and has a high cooling load for its temperature. Site D is at the centre, warm and has over-shadowing (category 6, Table 1). Thus at the three warmest sites (B, C and D) the order of decreasing cooling load follows the order of increasing urbanization, i.e., categories 5, 6 and 7 (Table 1). 3.1.2. Annual heating and cooling loads Fig. 2 shows how locations with higher cooling loads tend to have lower heating loads. A reason for scatter in the relationship is that some sites with higher mean temperatures (increasing cooling load), experience over-shadowing which reduces cooling load. 3.1.3. Annual total load Fig. 3 shows the total (heating + cooling) load at each location and how this varies with site category. The line shows the mean of the values for each category. The trend is for total load to increase with increasing urbanization, at least to site category 5 (where it is about 8.5% higher than the total rural load). There appears to be a reduction

from zero (no over-shadowing) to 2.0 (heavy over-shadowing for category 8 Table 1). 2.4. Internal gains Internal gains aect the ratio of heating to cooling load, and the relative signicance of a change in solar gain with over-shadowing. To determine the eect of this the building was modelled with four alternative levels of internal gain. Base 43 W/m2, +25% 54 W/m2, +50% 65 W/m2 and 25% 32 W/m2. These higher and lower gains are not articial and are quite likely to be found in practice (BRECSU, 1999). 3. Energy demand simulation results

42000

3.1. Main results base gain 3.1.1. Annual cooling load and mean temperature Fig. 1 shows the relationship between the measured mean 24 h temperature of each location over the year and the annual cooling load. Approximately two thirds of the variance in cooling load is associated with varying mean temperature (r2 = 0.67, Fig. 1). The cooling loads at three sites are at some distance from the regression line. Site A is 1.6 km from the centre, has over-shadowing (category 6, Table 1) and is 250 m from a large park. Site B is in the city of
42000

40000

Cooling load, kWh

y = -1.40x + 55600 r2 = 0.56 n = 24

38000 36000 34000 32000 30000 12000

13000

14000

15000

16000

Heating load, kWh

Fig. 2. The relationship between simulated annual heating and cooling load at each location.

C
Annual cooling load, kWh
40000

54000

Total annual load, kWh

13.0

38000 36000 34000 32000 30000 10.5

D B A
y = 3275x - 1873 r2 = 0.67 11.0 11.51 2.0 12.5

53000 52000 51000 50000 49000 48000 47000 0 1 2 3 4 5 6 7 8

Mean annual temperature, C

Site category
Fig. 3. Simulated annual total load at each location, separated by site category.

Fig. 1. The relationship between measured annual mean temperature and simulated cooling load.

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110

Total annual load, kWh
54000 53000 52000 51000 50000 49000 48000 47000 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
8.0 6.0 BASE -25% gain +25% gain +50% gain

105

% change wrt cat-1

4.0 2.0 0.0 1 -2.0 -4.0 -6.0 -8.0

Radial distance Xr, km

Fig. 4. Simulated annual total load at each location, separated by distance from the city centre.

Site category

Fig. 5. Simulated percentage change in total annual load with site category alone for dierent internal gains. Rural weather.

(down to 2% of total rural load) after this point, with increasing over-shadowing, but the small sample makes this speculative at this stage. Fig. 4 shows the same total loads, but how these vary with the distance of each site from the city centre. The line shows the mean of the values at each distance. Total load tends to increase closer to the city, but the range of scatter also increases. This may be associated with the increasing contrast in environments as one approaches a city centre: hard surfaces (open, or heavily over-shadowed) on the one hand, contrasted with large parks. At a radial distance of 3.2 km (Xr = 3.2), the highest total demand is from a hard surfaced site that has the highest annual cooling load of all sites. The lowest total demands at Xr = 3.2 are at a site next to a park and a site in a hard area of the City of London but heavily over-shadowed. At a radial distance of 1.6 km the site is near Regents Park and has a low total demand. This point is responsible for the sudden fall in the mean demand line on the graph and may be atypical of load at this distance. 3.2. Separation of the temperature and urban context eects variable gain There are two main factors aecting load: higher urban temperatures increasing cooling load, and higher overshadowing decreasing it, with the opposite eects for heating. The size of these eects when operating independently are evaluated here. The sensitivity of the results to variable internal gain is also given here; the higher the internal gain the more the annual loads shift from heating to cooling, aecting the balance of the eect of the heat island. 3.2.1. The eect of context alone The building energy simulation program has been run using rural weather acting on the typical air-conditioned oce (ECON 19/3) building in each of the seven contexts (site categories). This introduces the eects of over-shadowing, while controlling for external temperature. Fig. 5 shows the change in total annual load (heating + cooling) with site category. The change is plotted as a percentage of the load in a rural setting (category 1, Table 1).

The following observations can be made: The eect of site category depends on the level of internal gain. For a low internal gain building, increasing over-shadowing leads to an increase in total annual load, even at the densest urban category 7. For the BASE gain building, greater over-shadowing leads to an increase in total load, but starts to fall for the most urban site type. Higher gain buildings have a reduced total load with any level of over-shadowing. The BASE gain buildings total load increases by about 2% at the urban site (category 6), before falling to almost the same as in a rural context, at the most over-shadowed site (category 7). As noted above, for the BASE and low gain buildings, increasing context density leads initially to an increase in total annual load, and then a reduction. This is probably because of the angle of the sun being higher during the cooling season compared to the heating season. Applying a wide context aects gain during the heating season much more than in the cooling season. The trend of reducing total annual load with increasing site category is apparent as early as category 4 for the high gain building. This suggests that this eect, rst observed in Fig. 3 where temperature was uncontrolled, and for just one site and therefore speculative, is in fact a real eect and not an anomaly. 3.2.2. The eect of temperature alone The building energy simulation program has also been run using the measured hourly temperatures at each location, but keeping the typical air-conditioned oce (ECON 19/3) building in the open, as if in a rural context. This introduces the eects of temperature, while controlling for over-shadowing. Fig. 6 shows the change in total annual load (heating + cooling) with the annual mean temperature for each location. The change is plotted as a percentage of the load at the rural site temperature (10.7 C). Polynomial

106
14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 -2.010.0 -4.0 -6.0 -8.0 10.5 11.0 11.5 BASE -25% gain +25% gain +50% gain

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110

12.0

12.5

13.0

Annual mean temperature, C

Fig. 6. Simulated percentage change in total annual load with mean temperature alone for dierent internal gains. Rural site category.

regression lines have been tted to help visually in following the trend of the dierent sets of points. The following observations can be made: The eect of site temperature depends on the level of internal gain. In all but the lowest gain building, higher mean temperatures lead to increased total load. In the low gain building, higher temperatures have only a small eect on total load and at the highest temperatures total load decreases. The BASE gain buildings total load increases by about 7% at the warmest site in the heat island compared to the load at rural temperatures. With both increasing site category and mean temperature (both general concomitants of increasing urbanization), it can be concluded that the size and sign of the eect on total annual load depend on the level of internal gains. Note that the level of solar gain can also vary, given a xed degree of over-shadowing, if the facade design is dif ferent: higher or lower percentage glazing, shading features, etc. The percentage glazing used here is as given in the specication for the ECON 19/3 oce. The eect of varying solar gain is likely to be similar to that of varying internal gain. 4. Summer design temperature The results presented above suggest that air temperature contributes the highest percentage of change in energy demand and indicate that temperature variations within the London Heat Island should be considered in the design of buildings. In particular there are specic implications for passive cooling techniques due to increased temperatures, (Geros et al., 2005). In order to estimate peak cooling loads, it is desirable to modify published temperature values (CIBSE, 2006) to account for the position of a building within the London Heat Island. Such a method was rst proposed in Graves et al. (2001). CIBSE (2006) includes hourly design data

for typical days for each month. These days are derived from the 97.5 percentile daily global irradiation exceedence data. From available data (July and August 1999 and 2000) Graves et al. (2001) selected 6 days (5% of data) with the highest solar radiation and calculated the temperature difference (site air temperature minus Bracknell air temperature). From these, tables were proposed containing air temperature correction data for locations within the Great London Area. In order to simplify the procedure, the authors have repeated this study with one adjustment. Heathrow air temperatures have been used instead of Bracknells as Heathrow is the met station used for climatic data for London (CIBSE, 2001). The London region has been divided into three concentric annular zones, and the mean heat island intensities with respect to Heathrow computed for each zone and hour. Dierent radii were examined for signicant changes in the zone means before selecting: 03 km (core temperature), 310 km (semi-urban), and 1023 km (suburban). Beyond this lie rural areas. We propose these hourly temperature corrections to Heathrow data for summer peak cooling estimation based on radial distance (CIBSE, 2006). Table 2 presents the proposed adjustments to be added to Heathrow design data. Heathrow is towards the edge of the London Heat Island but signicantly warmer than a rural site. The heat island air temperature corrections in Table 2 are signicantly lower than if they were given relative to a true rural site. The real Urban Heat
Table 2 Proposed air temperature corrections based on radial distance from city centre Hour Distance from city centre 03 km 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Average 1.9 1.9 1.9 1.7 1.5 1.4 1.7 1.5 1.7 1.8 1.6 1.4 1.7 1.6 0.6 0.9 0.9 0.7 0.4 0.4 0.6 1.0 1.5 1.5 1.3 310 km 1.0 1.1 1.1 0.9 0.8 0.7 1.3 1.9 2.0 2.2 2.0 1.8 2.1 2.2 1.1 1.4 1.4 1.0 0.8 0.4 0.3 0.3 0.6 0.6 1.2 1023 km 0.9 0.9 0.7 0.7 0.7 0.6 0.4 0.8 1.3 1.5 1.4 1.3 1.3 1.4 0.4 0.5 0.4 0.2 0.0 0.4 0.9 1.4 1.3 1.3 0.0 2.3 2.5 2.0 2.3 2.3 2.3 2.5 1.3 0.7 0.4 0.7 0.0 0.2 0.2 0.9 0.0 0.1 0.2 0.8 0.5 1.9 2.7 2.7 2.8 1.1 Rural reference

% change wrt Trural

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110

107

Island Intensity can be obtained by subtracting the rural reference column from an annular column in Table 2. For example, for a site located in the core temperature area (03 km) at 03.00 h, the real Urban Heat Island Intensity is 4 K (1.9 (2.3)). 5. Model for predicting location specic hourly temperatures However, as discussed in Section 3 site category (microclimate) has been shown to have a marked eect on the heat island intensity and additional work is required to enable us to include the eect of physical site characteristics into urban heat island prediction algorithms. We rst review available techniques for developing models and their suitability for use by building designers. Based on this review, an Articial Neural Network (ANN) model is proposed as the most suitable and some preliminary results are presented. 5.1. Review of models The Urban Heat Island Intensity (UHII) has been primarily studied by urban climatologists. Their research focuses on microclimatology and boundary-layer climatology and a comprehensive review on the advances over the last twenty years has been recently published (Arneld, 2003). It includes ten recommendations for further research, one of which is that simple models are needed to estimate UHII within urban areas, as a function of time, weather conditions and structural attributes, for practical applications such as road climatology, energy conservation and weather forecasting. The following categories of models have been developed to predict the UHII: 1. Climatology models: These are detailed but are complex and require specic expertise for accurate predictions. For example the Colorado State University mesoscale model (CSUMM) has been developed to predict UHII (Taha, 1999; Leuzzi and Monti, 2004; Taha et al., 2000). The model requires the knowledge of soil parameters such as density, specic heat, thermal diusivity, moisture availability, roughness length and albedo in addition to knowledge of the large scale geostrophic wind and the initial proles of temperature and specic humidity. Such models might be too complicated for the purpose of estimating air temperatures to assess specic building cooling requirements. Methods of coupling urban models and building thermal performance models have been proposed (Flor and Dominguez, 2004; Tanimoto et al., 2004) which could be useful in estimating building cooling loads accurately but still the detail of required inputs is quite complex. Climatologic models are necessary in cases of complex terrain, where large temperature dierences are formed due to changes in elevation or in land use or both.

2. The UHII has also been studied by building physicists who have developed empirical models. For example, the Cluster Thermal Time Constant (CTTC) (Swaid and Homan, 1990) model has been developed for predicting the UHII in hot-arid climates and has also been extended to semi-arid climates (Elnahas and Williamson, 1997). Runnals and Oke (2000) describe an empirical model to disentangle the multiple controls on Vancouver UHII, including time of the day, wind speed, cloud cover and rural thermal admittance. Such models depend on measured local physical parameters and are very sensitive to empirically measured coecients. Developed empirical models use experimentally measured coecients which are site-specic and therefore only applicable to certain urban environments and climatic conditions. Interpolation and application of these models to dierent urban environments would not be possible without large errors. Sensitivity of results to small variations of coecients would also prohibit interpolation. 3. Computational Fluid Dynamics (CFD) models have also been used, mainly to predict air ow within the urban environment but also temperature elds (for example Tabahashi et al., 2004; Assimakopoulos et al., 2006). Such models, although reasonably accurate in many cases, require a large amount of input data with the corresponding uncertainties. Models are usually based on the wind and temperature and can not take many other important factors as inputs, although improvements to basic CFD models have been proposed (Tabahashi et al., 2004). 4. The UHII has also been studied using analysis of variance and regression, primarily to examine the eect of weather conditions on the UHII (as reported by Arneld (2003)), including recent studies (for example Morris and Simmonds, 2000). Many of these studies conrmed the rst empirical generalisation oered by Runnals and Oke (2000). UHII decreases with increasing wind speed and cloud cover, UHII is greatest during anticyclonic conditions and is best developed in the summer or warm half of the year. The authors have used statistical techniques in the case of London to study the eect of weather conditions and distance from the thermal centre with success (Watkins, 2002). However, it was also found that the UHII is dependent on site category and its physical characteristics. Probability methods such as the Monte Carlo method has been also used to predict UHII (for example Montavez et al., 2000; Sailor and Fan, 2002). 5. The UHII as aected by synoptic climatic parameters has also been studied using Articial Neural Networks (ANNs) (Mihalakakou et al., 2002; Kim and Baik, 2002). ANN models are suited to the proposed research because they are particularly good at representing any complex non linear functions whose analytical forms are dicult or impossible to obtain. The UHII problem is complicated and some relationships between dierent

108

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110

factors are still not fully understood. Kim and Baik (2002) have compared the ANN model results to a multiple linear regression model and concluded that the ANN model gave improved predictions by up to 6.5%.

5.2. Proposed model A preliminary study was carried out using similar techniques (ANNs) to investigate the eect of synoptic climatic conditions experienced over London. In addition, distance of location from the centre of the heat island was included as one of the parameters. To start with, standard neural network architecture procedures were followed, which involve the following: 1. 2. 3. 4. 5. 6. prepare the network database; design the neural network; train the network; test and diagnostic check; adjust the parameters of the neural network; repeat steps 35, until the performance of network is satisfactory.

North Road: 3.2 km Fitzjohns Avenue: 4.8 km St.Marks Road: 6.4 km Ivy Road: 9.6 km Hallywell Crescent:12.8 km River Road: 16.0 km Kent Avenue: 19.2 km Sevenoaks Road: 22.4 km

5.2.1. Input data As mentioned before the aim is to predict hourly ambient air temperature at specic locations. Therefore, hourly ambient air temperature of the meteorological (reference) station and hourly synoptic data of London are two sets of the input data. As indicated before, radial distance from central London is one of the parameters to inuence location specic air temperature. Therefore, radial distance is the third input for the network. Ten stations were selected from the available database for which full historical hourly data for one year (01/01/ 2000 to 30/09/2000) is available. Initially, one day was considered (from 00:00 to 23:00) as a whole cycle. As data from the period 01/01/2000 to 30/09/2000 were used, 274 sets of data for each station were produced which include the following: Synoptic weather data: These include cloud cover, humidity, wind speed, radiation diuse and radiation global; London Weather Centre hourly data were used. Ambient air temperature of meteorological station: London Heathrow is used as the reference station this is because Heathrow weather data are typically used for thermal modelling for London locations. Radial distance from Central London: The British Museum has been used as the centre point of London in this project this is the centre of the monitored data. Therefore the radial distances for each of the ten investigated locations are as follows to cover the range of monitored data distances: Great Russell street: 0.24 km West Smitheld: 1.6 km

5.2.2. Design of the network The choice of the type of network and learning rule is very important for successful results and its choice is dependent on the type of available data for input and the form of the output required. In this case, a denite function between the inputs and outputs could not be identied and therefore the feed-forward network seems the most suitable type of network. There are a number of learning rules available, of which back-propagation has been used in similar work before (Mihalakakou et al., 2002). Back-propagation was created by generalising the WidrowHo learning rule to multiple-layer networks and nonlinear dierentiable transfer functions. Input vectors (data) and the corresponding output vectors (data) are used to train a network until it can approximate a function, associate input vectors with specic output vectors, or classify input vectors in an appropriate method as dened by the trainer. Trials with the available data indicated that this is a suitable network and learning rule. Three layers were designed; input layer, hidden layer and output layer. The next step was to choose a suitable training algorithm; nine promising algorithms were tested and the scaled conjugate gradient back-propagation (trainscg) was selected for further development; this proved to have best performance in function approximation for this particular network design. All parameters were determined by following the repeating procedure of train test adjust parameters, until the performance was satisfying; the nal learning rate was 0.2, error goal is 0.5, nerves in input layer 7, nerves in hidden layer 19, nerves in output layer 1. As the inputs are hourly data, building the model for 24 h results to 145 elements in input vectors, and 24 elements in output vectors. This treatment produced a relatively large network compared with the number of elements in input vectors; therefore the total number of samples (data) becomes too small to train such a network. The solution was to divide it into 24 small networks, one for each hour of the day. In the following sections, we present the results of one example (the network for 15:00 h) to explain the process. 6. Results Two important factors, which could be used to evaluate the trained network, are training error and prediction error. 100 sets of trained data were selected as the input for the network and simulative results were produced.

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110

109

and distance from the centre of the heat island to predict site-specic air temperature in London. Work continues to improve the accuracy of predictions as described in Section 8. 7. Discussion The eect of the London Heat Island on energy used for heating and cooling depends on the degree of urbanization in a particular location, radial distance from the centre (depth within the heat island) and the relative contribution of solar gain to total gains in a building. Heating dominated buildings will tend to benet from a heat island whereas the opposite is the case for cooling dominated ones. Previous results from an analysis of variance of daytime temperature showed a fairly consistent ordering of mean temperature (or heat island intensity) with site category, when controlling for radial distance (Watkins et al., 2002a). It should be noted that the site categories and associated street gorge ratios (Table 1) reect the urban densities in London, and may not be appropriate for other cities where, for example, very much narrower streets are the norm. For a typical air-conditioned oce building (ECON 19/ 3) operating with internal gains of 43 W/m2, the annual heating load decreases by 22% from a rural site, with cooling load increasing 25%. Mean summer daytime temperatures for dierent sites are found to be associated with a site categorization based on the nature of surfaces and street gorge ratio. For a given radial distance, daytime temperatures tend to peak at category 5 sites. The annual primary energy use is also associated with site category, again tending to peak at category 5 sites, and reducing at more urbanized (over-shadowed) sites, despite them being warmer. When the ECON 19/3 building has 25% lower internal gain, total annual energy demand tends to be less sensitive to position within the heat island. This is because the heating and cooling loads are more balanced and a reduced heating load, deeper within the heat island, is matched by an increased cooling load. With higher gains, +25% and +50%, the pattern followed is similar to the base gain building. Therefore, estimation of cooling load due to external conditions will depend on the location of the building within the Urban Heat Island. Prediction of location dependent ambient temperature will improve accuracy in such calculations. A simplied method for adjusting temperatures based on Heathrow peak cooling calculations design data and radial distance from the centre of the London Heat Island is proposed (Section 4). However, for low energy design strategies, availability of location specic hourly air temperatures available for a year would contribute to more accurate predictions for their eectiveness. It was shown that an ANN model might be a suitable model to predict such values. Preliminary results from an ANN model based on meteorological air

Fig. 7. ANN training phase: correlation between simulative and measured results of the ANN model. Trained sets of data are used (measured results) as input vectors and output vectors (simulative results) are predicted. The training phase aims for the measured results to be the same as the simulative results or within a small training error (less than 0.3 in this case; mean square errors varied in the range of 0.10.3).

These results were compared with the corresponding measured data to nd the training error. Fig. 7 shows that there is good agreement between simulative and measured results. Most relative error values are less than 0.3. Mean square errors varied in the range of 0.10.3. From the 274 sets of data available from each station, 200 sets were then used for training and 74 for testing the predictions. The 200 sets of untrained data were randomly selected (20 from each station) for prediction testing. Fig. 8 shows that the network also performed well in prediction. 85% of the simulative result has a good agreement with the corresponding measured data, while relative error values are less than 0.6 and the mean square errors range between 0.2 and 0.5. These preliminary results demonstrate that the proposed network is successful in considering synoptic weather data

Fig. 8. ANN prediction phase using 200 sets of untrained data. 85% of the simulative results have good agreement with the corresponding measured data; relative error values are less than 0.6 and the mean square errors range between 0.2 and 0.5.

110

M. Kolokotroni et al. / Solar Energy 81 (2007) 102110 GLA , 2002. The impact of Climate Change on London: summary report, Greater London Authority, Available from: <http://www.london.gov.uk/gla/publications/environment/londons_warming02.pdf>. Graves, H., Watkins, R., Westbury, P., Littlefair, P., 2001. Cooling Buildings in London Overcoming the heat island, BR431, CRC. Elnahas, M.M., Williamson, T.J., 1997. An improvement of the CTTC model for predicting urban air temperatures. Energy and Buildings 25, 4149. Hassid, S., Santamouris, M., Papanikolaou, N., Linardi, A., Klitsikas, N., 2000. The eect of the Athens heat island on air conditioning load. Energy and Buildings 32, 131141. Kim, Y.H., Baik, J.J., 2002. Maximum urban heat island intensity in Seoul. Journal of Applied Meteorology 41 (6), 651659. Kolokotroni, M., Watkins, R., 2004. In: The impact of the London heat island on building heating and cooling loads, Contemporary Problems of Thermal Engineering, Gliwice-Ustron, Poland, pp. 233240. Kolokotroni, M., Giannitsaris, I., Watkins, R., 2006. The eect of the London urban heat island on building summer cooling demand and night ventilation strategies. Solar Energy 80 (4), 383392. Landsberg, H., 1981. Urban Climate. Academic Press. Leuzzi, G., Monti, P., 2004. Numerical modelling of ow and dispersion in Roma area. In: 9th Int. Conf. on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes. Mihalakakou, G., Flocas, H.A., Santamouris, M., Helmis, C.G., 2002. Application of neural networks to the simulation of the heat island over Athens, Greece, using synoptic types as a predictor. Journal of Applied Meteorology 41, 519527. Montavez, J.P., Jimenez, J.I., Sarsa, A., 2000. A Monte Carlo model of the nocturnal surface temperatures in urban canyons. Boundary-Layer Meteorology 96 (3), 433453. Morris, C.J.G., Simmonds, I., 2000. Associations between varying magnitudes of the urban heat island and the synoptic climatology in Melbourne, Australia. International Journal of Climatology 20 (15), 19311954. Runnals, K.E., Oke, T.R., 2000. Dynamics and controls of the nearsurface heat island of Vancouver, British Columbia. Physical Geography 21, 283304. Sailor, D.J., Fan, H.L., 2002. Modelling the diurnal variability of eective albedo for cities. Atmospheric Environment 36 (4), 713725. Santamouris, M., Papanikolaou, N., Livada, I., Koronakis, I., Georgakis, C., Assimakopoulos, D.N., 2001. On the impact of urban climate to the energy consumption of buildings. Solar Energy 70 (3), 201 216. Swaid, H., Homan, M.E., 1990. Prediction of Urban Air Temperature Variations using the Analytical CTTT Model 14, 313324. Tabahashi, K., Yoshida, H., Tanaka, Y., Aotake, N., Wang, F., 2004. Measurements of thermal environment in Kyoto city and its prediction by CFD simulation. Energy and Buildings 36 (8), 771779. Taha, H., 1997. Urban climates and heat islands: albedo, evapotranspiration, and anthropogenic heat. Energy and Buildings 25, 99103. Taha, H., 1999. Modifying a mesoscale meteorological model to better incorporate urban heat storage: a bulk-parameterization approach. Journal of Applied Meteorology 38 (4), 466473. Taha H., Chang, S., Akbari, H., Heat Island Group 2000. Meteorological and air quality impacts of heat island mitigation measures in three US cities. Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720. Tanimoto, J., Hagishima, A., Chimklai, P., 2004. An approach for coupled simulation of building thermal eect and urban climatology. Energy and Buildings 36 (8), 781793. Watkins, R., 2002. The impact of the urban environment on the energy used for cooling buildings. Ph.D. Thesis, Brunel University. Watkins, R., Palmer, J., Kolokotroni, M., Littlefair, P., 2002a. The London heat island results from summertime monitoring. BSER&T 23 (1). Watkins, R., Palmer, J., Kolokotroni, M., Littlefair, P., 2002b. The balance of the annual heating and cooling demand within the London heat island. BSER&T 23 (4), 207213.

temperature, synoptic weather data and distance from the centre indicate a good agreement. Such a model can be used for more detailed prediction of location specic hourly air temperature. In this model, 85% of the simulative result agrees with the corresponding measured data and this is consistent with ndings by Kim and Baik (2002) who reported a 6.5% improvement in comparison to a multiple linear regression model. In our case, it has produced a marked improvement in comparison to linear regression analysis. This is because radial distance has been included as one of the inputs in addition to weather data. However, the match is not perfect yet indicating that there might be other parameters aecting the result. 8. Concluding remarks Site category (microclimate) has been shown to have an eect location specic air temperatures within the London Urban Heat Island. More work is needed to understand the main parameters for including such physical site characteristics into the prediction model. Therefore, future work will investigate how to classify and quantify physical urban characteristics with the aim to incorporate these into the ANN model. We believe that by incorporating location related weighting factors within the ANN model, the accuracy of predicting location specic air temperatures would be improved both for design applications using typical weather data and for forecasting applications using real weather data. References
Akasaka, H., Nimiya, H., Soga, K., 2002. Inuence of heat island phenomenon on building heat load. Building Services Engineering Research and Technology 23 (4), 269278. Akbari, H., Konopacki, S., 2005. Calculating energy-saving potentials of heat island reduction strategies. Energy Policy 33 (6), 721756. Arneld, A.J., 2003. Two decades of urban climate research: a review of turbulence, exchange of energy and water, and the urban heat island. International Journal of Climatology 23 (1), 126. Assimakopoulos, V.D., Georgakis, C., Santamouris, M., 2006. Experimental validation of a computational uid dynamics code to predict the wind speed in street canyons for passive cooling purposes. Solar Energy 80 (4), 423434. BRECSU, 1999. Energy use in oces (Energy Consumption Guide 19), BRECSU. Brundl, W., Hoppe, P., 1984. Advantages and disadvantages of the urban heat-island an evaluation according to the hygro-thermic eects. Theoretical and Applied Climatology 35 (12), 5566. Chandler, T.J., 1965. The Climate of London. Hutchinson&Co (Publishers) Ltd. CIBSE, 2006. Guide A, Environmental Design, CIBSE. CIBSE, 2001. Guide J, Weather, Solar and Illuminance Data, CIBSE. CIBSE TM36, 2005. Climate Change and the Indoor Environment: Impacts and Adaptation, CIBSE. Flor, F.S., Dominguez, S.A., 2004. Modeling microclimate in urban environments and assessing its inuence on the performance of surrounding buildings. Energy and buildings 36 (5), 403413. Geros, V., Santamouris, M., Karatasou, S., Tsangrassoulis, A., Papanikolaou, N., 2005. On the cooling potential of night ventilation techniques in the urban environment. Energy and Buildings 37 (3), 243257.