Energy and Buildings 129 (2016) 563–573

Contents lists available at ScienceDirect

Energy and Buildings

journal homepage: www.elsevier.com/locate/enbuild

A pixel-based approach to estimation of solar energy potential on

building roofs
Yan Li a , Ding Ding b , Chunlu Liu a,∗ , Chongjie Wang b
a
School of Architecture and Built Environment, Deakin University, Geelong, Australia
b
School of Architecture and Urban Planning, Shandong Jianzhu University, Jinan, China

a r t i c l e i n f o a b s t r a c t

Article history: Precise estimation of solar energy on building roofs plays a critical role in sustainable development and
Received 14 March 2016 renewable energy consumption of high-density human habitats. Conventional solar radiation models
Received in revised form 21 June 2016 based on costly Light Detection and Ranging (LiDAR) data are only adequate for existing buildings, not
Accepted 5 August 2016
for future construction areas. In this paper, a pixel-based methodology is constructed for estimating solar
Available online 10 August 2016
energy potential over roofs. Buildings with flat roofs in a newly planned construction area are chosen as a
case study. The solar radiation at a certain cell is mathematically formulated in the pixel unit, and its yields
Keywords:
over a certain time period are calculated by considering multiple instantaneous solar irradiances and are
Building roofs
GIS
visually presented by image processing. Significant spatial and temporal variations in solar radiation are
Modelling measured. Within the study area, the maximum and minimum annual radiation yields are estimated at
Pixel 4717.72 MJ/m2 /year and 342.58 MJ/m2 /year respectively. Radiation contour lines are then mapped for
Solar radiation outlining installation ranges of various solar devices. For each apartment building, around 20% of roof
Photovoltaics areas can obtain 4500 MJ/m2 /year or more solar radiation yields. This study will benefit energy investors
Urban energy planning and urban planners in accurately predicting solar radiation potential and identifying regions with high
radiation over building roofs. The results can be utilised in government policies and urban planning to
raise awareness of the use of renewable energy sources.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction These devices are also easy to install with little need to dig, and can
achieve energy conversion without noise and greenhouse gases.
As fossil fuels are projected to decline and the global demand Furthermore, the direct utilisation of solar energy in buildings is
for energy is increasing rapidly, many countries have to seek mainly through solar photovoltaic technology and thermal tech-
sustainable development by generating and acquiring energy in nologies [5]. The former technology aims to absorb solar radiation
a sustainable way [1]. Renewable energy is sustainable energy and generate electricity directly using photovoltaic cells. In addi-
derived from the natural environment. Since it significantly con- tion, solar energy sources are also very cost effective. Compared to
tributes to low emissions and national energy security, renewable conventional energies, an investment in solar equipment provides
energy is of great importance for future environmental protection long lasting energy for future generations [6]. Therefore, utilisa-
and economic development [2]. In addition, the comfort of mod- tion of solar energy in buildings can be used to support sustainable
ern living conditions has been achieved at the cost of vast energy buildings, such as rating systems, namely Leadership in Energy and
resources. Thus, governments and engineers often devote extensive Environmental Design and Green Star [3,7].
effort to exploiting renewable energy sources in order to minimise However, great efforts are usually made in search of low
conventional energy expenditure and improve living standards [3]. manufacturing costs by technology research [8]. To promote the
Among renewable energy resources, solar energy is considered utilisation efficiency of solar energy and satisfy as much urban
a free, clean, unlimited, and environmentally and economically energy demand as possible, it is necessary to evaluate solar energy
friendly energy source [4]. In urban areas, a building rooftop is often potential and analyse its distribution over urban areas. Yet eval-
recognised as a suitable location for installing solar energy devices. uation of the solar energy potential for urban use is a significant
challenge as there are significant spatial and temporal variations in
solar radiation that are greatly impacted by many factors such as
∗ Corresponding author.
surface orientation, shadowing effects and terrain features [9]. To
E-mail address: [email protected] (C. Liu).
address these issues, researchers explored and developed various

http://dx.doi.org/10.1016/j.enbuild.2016.08.025
0378-7788/© 2016 Elsevier B.V. All rights reserved.
564 Y. Li et al. / Energy and Buildings 129 (2016) 563–573

methods and technologies that are used for radiation estimation are used for terrain description and shadowing effects respectively.
over given building surfaces [10]. Hachem, Athienitis and Fazio Statistical analysis and potential distribution with centimetre-level
[11] chose two-story single housing units as case studies to inves- accuracy aim to promote assessment measures and indicate suit-
tigate the effects of roof shapes on solar energy potential. Yet able regions for installing solar devices over roofs. The proposed
these initial studies are often limited to self-shadowing in the built method is detailed in the next section.
environment. To address the spatial factors, geographic informa-
tion system (GIS) techniques are also frequently incorporated in
2. A pixel-based method for solar radiation estimation
the estimation. Based on digital maps produced by Google Earth,
Ordóñez, Jadraque, Alegre and Martínez [12] successfully deter-
2.1. Solar radiation estimation algorithm
mined the solar energy potential for installation of grid-connected
photovoltaic systems on rooftops in Andalusia. In addition, some
In this paper, the basic algorithm of solar radiation estimation is
GIS-based tools have been widely used for the solar radiation esti-
implemented in terms of the approach presented by Kumar, Skid-
mation. For example, PVGIS, a web-based database, that integrates
more and Knowles [24]. Corresponding solar declination ı, hour
many relevant data sources, such as USGS GTOPO30 digital eleva-
angle h and altitude angle ␣ can be then obtained and used to
tion models (DEM), is capable of calculating photovoltaic potential
describe the Sun’s position in the sky. The solar flux Io (W/m2 ) can
for many countries and regions [13]. Nevertheless, these solar radi-
be expressed as:
ation tools have not yet taken complex shadowing effects into
account in GIS environment [14].   ◦ 
Io = So 1 + 0.0344 cos 360 N/365 (1)
Since height grids in urban environments became easily acces-
sible data through airborne Light Detection and Ranging (LiDAR)
in which N is the number of days of the actual date since 01 January
surveys, building detection based on remote sensing data has been
and the solar constant So is set as 1367 W/m2 . The optical air mass
popular in the solar energy estimation on the extracted building
M, the atmospheric transmittance for direct radiation b and diffuse
footprints. This is because the LiDAR-based data acquisition tech-
radiation d can be written in Eqs. (2), (3) and (4), respectively.
nology allows the rapid reconstruction of terrain surfaces, which is
suitable for existing 2D models to assess building insolation [15].  0.5
In 2009, Tereci, Schneider, Kesten, Strzalka and Eicker [16] applied M = 1229 + (614 sin ˛)2 − 614 sin ˛ (2)
LiDAR data to build a Digital Surface Model (DSM), determining  
the annual solar potential for identified building roofs in combi- b = 0.56 e−0.65M + e−0.095M (3)
nation with ALK map data and GIS software. Lee and Zlatanova
[17] constructed a 2.5-D DSM on the basis of the use of LiDAR d = 0.271 − 0.294 × b (4)
and 2D/3D vector data to compute the morphological properties
of buildings. The implemented tool for solar radiation calculation Direct solar radiation Ib can be formulated as Eq. (5) when the
is not only suitable to roof areas but also useful to analyse building Sun’s rays are striking the surface whose normal makes an angle 
facades. In 2013, in order to estimate solar potential over build- with the direction to the Sun.  relies mainly on spatial factors, such
ing surfaces, Redweik, Catita and Brito [18] also proposed a solar as the slope and aspect of the surface, and temporal factors, such as
radiation method based on the r.sun radiation model developed by the solar azimuth and altitude.  denotes the location-based sun-
Šúri, Huld, Dunlop and Ossenbrink [19] and incorporated this in the shine percentage that aims to consider the probability of a cloudy
open source GRASS GIS [20]. The results revealed that the poten- sky. Diffuse solar radiation Id can be expressed as Eq. (6) for the
tial of building facades is lower than that of roofs although they surface with the tilt angle ˇ.
normally have large areas. By taking advantage of ESRI’s Solar Ana-
lyst Toolbox and LiDAR data, Kodysh, Omitaomu, Bhaduri and Neish Ib =  × Io × b × cos  (5)
[21] introduced a method for estimating solar potential on multiple
building rooftops through combining a DEM with an upward- Id = Io × d × cos2 ˇ/2 sin ˛ (6)
looking hemispherical viewshed algorithm. Lukač and Žalik [22]
presented a methodology through using graphics processing units When solar radiation strikes a particular site, the total radiation
with compute unified device architecture technology and LiDAR is equal to the sum of the direct and diffuse solar irradiances for an
data for solar potential estimation. In addition, some evolutionary unshaded spot. On the contrary, the total radiation excludes direct
approaches have also been applied in the optimal design of solar solar radiation if the spot is in shadow. This is because, unlike the
building models based on LiDAR data [10,23]. diffuse radiation, the direct radiation cannot reach a shaded spot.
Although previous research has paid extensive attention to Shadowing effects at any moment can be illustrated by an image,
evaluation of the solar energy potential over building surfaces, in which all regions are equally discretised by pixels. As shown
LiDAR-based procedures are more suited to existing buildings. in Fig. 1, the left picture demonstrates a 3D shading model for an
However, for investors, potential energy investments may bene- apartment building. The right picture depicts a shadow map of the
fit from assessment measures of solar potential according to future building from a top view, in which dark pixels mean shaded spots.
urban planning. Moreover, rapid development and wide use of solar For a shadow map at a time point, the pixel value of the greyscale
energies in urban areas more necessitate accurate measurements image Vpixel is a single number that represents the brightness of the
of solar resources, which means that higher resolution photogra- pixel. The most common pixel format is the byte image, in which
phy (or image) is an essential prerequisite in order to demonstrate this number is stored as an 8-bit integer that ranges from 0 to 255.
the distribution of solar potential over building surfaces and pro- In this research, the pixel value between 0 and 250 is taken to
vide accurate statistical analysis. Therefore, this paper establishes be a shaded cell in a shadow map, if 250 is selected as a threshold
a pixel-based methodology for estimating solar energy potential. value. From 250 to 255 is considered an unshaded cell. The area
Based on digital image processing using MATLAB, the method for each cell Acell is used to indicate the actual area in a map that a
allows for estimation of solar potential yields on rooftops (not only pixel represents in an image. The actual area represented by a pixel
limited to building roofs) over a specific period by formulating solar will be increased by 1/ cos ˇ times once a surface is inclined with
irradiances in pixel unit. The method is here applied to flat roofs of the tilt angle ˇ. Therefore, the Nth day daily solar radiation yield
N
for the pixel cell located in the ith row and jth column Ecell can be
buildings from future urban planning, in which GIS and SketchUp i,j
 Y. Li et al. / Energy and Buildings 129 (2016) 563–573 565

Fig. 1. A building model and its shadow map.

obtained by integrating the instantaneous solar irradiances within

Terrain data Building models
the time interval [sunrise, sunset].
⎧
⎪
⎪
sunset GIS + SketchUp
⎪
⎪
⎪
⎪ Acell (Ib (t) + Id (t)) dt if 250 ≤ Vpixel ≤ 255
⎨
N sunrise Shadow maps A building roof image
Ecell = (7)
i,j ⎪
⎪
sunset
with IhJ pixels with IhJ pixels
⎪
⎪
⎪
⎪ Acell Id (t) dt if 0 ≤ Vpixel < 250
⎩ Solar radiation MATLAB MATLAB
sunrise estimation
For a given region , the solar radiation yield E from the N1 th For i=1:I Read the value of the
to N2 th days during a year can be expressed as For j=1:J pixel Ci,j in the image
J J N2
I I
N
 
E= Ecelli,j = Ecell i and j ∈ ˝ (8)
i,j
i=1 j=1 i=1 j=1 N=N1 Is the
pixel Ci,j in shadow
After calculating the solar radiation yield, the radiation distri-
bution within ˝ can be mapped as a new greyscale image on the in the maps?
’
basis of new pixel values. A new pixel value Vpixel can be defined, Yes No
Is the
through a linear interpolation method, as pixel over roofs in
  Diffuse Direction and
’
Vpixel = 255 × Ecelli,j − Ecell,min / the image?
radiation diffuse radiation
  
Ecell,max − Ecell,min i and j ∈ ˝ (9) calculation calculation
where Ecell,max and Ecell,min are the maximum and minimum cell
solar radiation yields within ˝, respectively. Global irradiances
for cells
2.2. Implementation procedures of the pixel-based estimation
method Solar radiation
Yes
yields over roofs
The LiDAR-based procedure starts with data collection through
aerial detection and constructs a continuous DSM, in which the Fig. 2. Flowchart to calculate solar radiation yields over building roofs.
raster size usually ranges from 0.3 m to 1 m [18,25,26]. Unlike the
previous methods, the pixel-based procedure begins with build-
ing models of future urban planning. A series of high-resolution on the basis of urban terrain features so that SketchUp can cast vis-
shadow maps are produced for preserving the topographic features ible shadows of buildings with solar conditions in a complex built
of building surfaces. Based on digital image processing of MATLAB, environment. Secondly, a series of instantaneous shadow maps are
the data from each pixel can be independently processed within produced after casting visible shadows for building models. Black
the uniform square grid data structure. The whole workflow of the and white pixels are used to indicate shaded and unshaded cells
pixel-based method is illustrated in Fig. 2 by taking building roofs respectively. All maps have the same image dimensions and are
as an example. from the same top view. A binary image of the rooftop is also pro-
Firstly, future urban planning is used to perform the solar duced by SketchUp from the same view. Pixels that represent roofs
potential estimation through the 3D modelling platform SketchUp. could be displayed as black, and the rest is referred to as the back-
Terrain data can be easily imported into SketchUp from Google ground colour, white. Thirdly, the algorithm implements radiation
Earth. More accurate terrain data can also be utilised by SketchUp estimation calculation for pixels. Each pixel value in the shadow
in the form of elevation contour lines to describe the continuously maps is read by MATLAB so that the pixel can be identified as either
varying topographic surfaces. Building models are then constructed shaded or unshaded. Diffuse radiation calculation will be imple-
566 Y. Li et al. / Energy and Buildings 129 (2016) 563–573

Fig. 3. Location and urban planning of the study area.

mented if the pixel is identified as a shaded pixel. Otherwise the city in the social, economic and ecological aspects. In addition,
program will carry out the direct and diffuse radiation. Based on investors may consider potential investments in solar energy on the
the actual cell area, daily global irradiances for each pixel can be basis of assessment and prediction. As the high computational cost
estimated in terms of Eq. (7). Fourthly, each pixel value in the roof could be avoided and specifications are identical for some build-
image is read by MATLAB in order to identify whether it belongs ings, this paper selects the study area that consists of two adjacent
to building roofs. It is assumed that the pixel is displayed as a blocks to evaluate solar potential over flat roofs, as shown in the
colour of building roofs. The algorithm will implement solar radia- lower left picture of Fig. 3. There would be about 5500 inhabitants
tion estimation for building roofs by taking into account all pixels living in these two blocks.
Ci,j of shadow maps, whose positions in the shadow maps have the 3D building models are first constructed in order to demonstrate
same position in the building roof image. The fifth step is to calcu- the applicability of the proposed method, as shown in the upper
late the total solar radiation yield over building roofs for a period picture of Fig. 4. The primary GIS database consists of a 2D roof
through considering multiple instantaneous solar irradiances using image (or roofprints) for buildings and a digitised contour file. In
a specific time step. A new greyscale image can be obtained by this study, the elevation contours, as shown in this picture, are pro-
re-computing pixel values according to Eq. (9). duced with ArcGIS software based on dense spot elevations that are
evenly distributed on this study area. The lower left picture depicts
a shadow map at 10:00 on 30 January, in which shaded pixels are
3. Study site and data process
indicated by black. In this study, every shadow map is produced
for each hour between sunrise and sunset so that the daily energy
Based on the above-mentioned method, this research considers
yield for each pixel can be calculated by summing solar potential
several buildings with flat roofs from urban planning of Xiuyuan
values from every hour during the same day. A representative daily
eco-city as a study area to measure solar potential over build-
energy yield is calculated every five days so that the annual energy
ing roofs. The daily sunshine percentage  for the case study can
yield can be obtained with a five-day time step.
be obtained using the rate of the mean sunshine duration [27] to
As the buildings as well as their corresponding roofs are labelled
the ideal daily daylight duration between sunset and sunrise [28].
in the lower right picture of Fig. 4, there are two office buildings,
According to urban planning [29], the Xiuyuan eco-city (36.7◦ N,
namely Buildings A and B, and twenty-four apartment buildings,
117.4◦ E) will be situated in the east of the City of Jinan that is the
namely Buildings C–Z in these two adjacent blocks. The three-story
capital of Shandong province in Eastern China, as shown in Fig. 3.
office buildings will be located close to a main street facing west.
The construction of Xiuyuan eco-city was granted by the provincial
Heights of each story and parapet wall are designed as 4 m and
government in March 2015 and will be fully completed by the end
1.5 m respectively. The residential buildings will face south and
of 2020.
have three types, namely 4-unit, 6-unit and 8-unit apartments. The
To build an eco-city, various requirements involved in its criteria
buildings are designed as 11 stories with 2.8 m of story height.
must be satisfied [30]. Therefore, measuring solar energy poten-
Parapet walls are also 1.5 m in height. Moreover, the roofprints
tial is essential for further investigating how specific solar energy
excluding stair wells, elevator shafts and parapet walls are dis-
devices could contribute to fulfil the various objectives of an eco-
 Y. Li et al. / Energy and Buildings 129 (2016) 563–573 567

Fig. 4. Roofprints and shadow maps of the study area at 10:00 on 02 June.

played as black in this picture. These shadow and roofprint images 4694.86 MJ/m2 when considering the possibility of experiencing
have the same pixel dimensions of 1916 × 912 and each pixel rep- cloudy sky during the whole year. The amount is quite close to the
resents a cell with an actual area of 0.3 m × 0.3 m. maximum annual radiation of 4717.72 MJ/m2 within this study
area. The value is also in the range of 4540–5120 MJ/m2 /year
4. Radiation results in terms of the data provided by Jinan’s observation station of
solar radiation [31]. Moreover, the daily solar energy significantly
4.1. The annual solar radiation potential over roofs decreases when compared with clear-sky conditions, especially
daily values in July and August. This is because the ratio of the
The proposed method is applied to the study area. Fig. 5 shows actual mean sunshine duration to the theoretical duration only
the annual solar radiation yields over this study area. As expected, reaches 51.1% and 59.2% in these two months. Fig. 6 also shows
this figure clearly illustrates that irradiation levels of building roofs estimated monthly solar radiation yields at Site O and observed
are usually much higher when compared with ground surfaces. values from the solar radiation station [31]. They have a same
In the two-block study area with dimensions of 632 m × 301 m, trend. Although the differences between estimated and observed
the annual total solar energy reaches 7.082E + 8 MJ/year. Max- values are slightly more than 10% for the first three months, the
imum and minimum annual radiation yields are estimated at differences for other months are relatively low and the average
4717.72 MJ/m2 /year and 342.58 MJ/m2 /year within the study area value is less than 4%. It is possible that the observed radiation
respectively. To map the radiation distribution, 255 and 0 are then yields are a little higher than the estimated, because observation
defined as their corresponding pixel values respectively points are never shaded. In addition, the average daily radiation
Within this study area, this research selects several sites of pos- over Roof I is slightly lower than the yield at Site O, but they also
sibly installing solar collectors and average solar radiation yields have a same trend. As a result, Site O is a relatively suitable place
over two roofs to demonstrate their changes. As labelled in Fig. 5, for installing solar collects within this roof.
Site O is at the geometric centre of the roof of Building I (or Roof Unlike the 11-story apartment buildings, Building A is a 3-story
I) and is rarely shaded on the rooftop. Its corresponding pixel is in commercial building. The height of Site P is only 12.06 m above
Row 152 and Column 755 in the roof image. Site P is on the rooftop ground according to the design. The daily radiation yield that Site P
of Building A, located at centre of the roof section between Build- receives is generally smaller than the yield at Site O by comparison.
ings C and D. Its corresponding pixel is in Row 208 and Column However, daily radiation yields are almost identical between May
319. Site Q is at the centre of the ground between Buildings I and J, and August. This is because the solar altitude becomes great so that
and the corresponding pixel is in Row 207 and Column 835. Fig. 6 Site P is also little shaded by Building D during the period. Similarly,
illustrates the variation for the daily solar energy yields during the the radiation yield at Ste P is higher than the average radiation over
whole year, including the average radiation yields over the roofs of the roof of Building A from April to August, but lower in February,
the apartment and commercial buildings. March, September, October and November. Consequently, Site P is
For Site O, its annual total solar energy reaches seasonally suitable to solar energy devices over the roof.
7215.91 MJ/m2 /year under clear-sky conditions. The full line Moreover, the daily radiation yield at Site Q is usually below
gives the change of the solar energy at Site O in Fig. 6. Theoreti- 1 MJ/m2 /day in January, February, November and December. How-
cally, the daily solar energy at Site O can be above 25 MJ/m2 /day ever, the daily radiation is almost identical with the radiation level
between May and August, while it could be below 10 MJ/m2 /day at Site O from May to July because of less shadowing effects from
in January and December. The maximum daily energy of the Building J. Overall, Site Q is the least suitable location of installing
site receives 3.4 times as many radiation yields as the minimum
value. However, Site O can receive an annual total solar energy of
568 Y. Li et al. / Energy and Buildings 129 (2016) 563–573

Fig. 5. Annual radiation map.

Fig. 6. Solar energy change during the whole year.

solar energy collectors because of poor solar resources during most are located to the north of buildings. In summer, areas of receiving
of the year. direct solar radiation increase due to greater altitudes so that most
pixel values rise and the map turns white. The areas greater than
4.2. The seasonal solar radiation potential over roofs 500 MJ/m2 slightly increase to 10,550 m2 over roofs and 76,187 m2
over the ground. As shown in the autumn and winter radiation
The proposed method can also be applied to the analysis of solar maps, most pixels considerably increase in the grey level. In par-
radiation over specific periods. In order to demonstrate the vari- ticular, the winter radiation map illustrates that the most regions
ation of solar radiation with the seasons, this research presents over these two commercial building roofs are not suitable to install
greyscale maps in Fig. 7 in terms of Eq. (9), in which Ecell,max and solar energy devices.
Ecell,min still adopt annual maximum and minimum radiation yields. However, each commercial building roof covers around
To take the left block for example, the seasonal maximum solar 1400 m2 , which is nearly 1.4 times more areas than Roof I excluding
radiation reaches 1527.92 MJ/m2 in spring, 1581.90 MJ/m2 in sum- elevator shafts, 2.3 times for Roof C and 3.4 times for Roof D. the
mer, 982.55 MJ/m2 in autumn and 626.36 MJ/m2 in winter within total solar energy yield over each commercial building roof should
this area of 89,782 m2 . not be overlooked. Table 1 lists seasonal solar energy yields of these
Within the left block, areas, in which the radiation yield is 14 buildings in detail.
greater than 500 MJ/m2 , covers around 10,253 m2 over roofs and It is estimated that Roof A or B can receive 1.62 × 106 MJ solar
73,030 m2 over the ground in spring. Most low radiation regions energy in spring. The energy yield can slightly rise to around
 Y. Li et al. / Energy and Buildings 129 (2016) 563–573 569

Fig. 7. Seasonal radiation maps.

Table 1 solar energy yields of rooftops are significantly affected by roofprint

Solar energy over roofs (MJ).
areas. However, for residential buildings, the per unit area energy
Roof Spring Summer Autumn Winter yield is decreased by almost 45% in autumn when compared with
A 1.62 × 10 6
1.73 × 106
9.89 × 105
5.60 × 105 summer. The commercial building is reduced by 42%, which means
B 1.62 × 106 1.74 × 106 9.90 × 105 5.63 × 105 its seasonal solar energy has a relatively slow decrease.
C 7.54 × 105 8.09 × 105 4.50 × 105 2.65 × 105
D 5.03 × 105 5.42 × 105 2.99 × 105 1.76 × 105
E 7.54 × 105 8.06 × 105 4.51 × 105 2.65 × 105 5. High-resolution mapping of solar radiation distribution
F 7.68 × 105 8.25 × 105 4.57 × 105 2.69 × 105 over roofs
G 4.92 × 105 5.27 × 105 2.94 × 105 1.72 × 105
H 7.58 × 105 8.17 × 105 4.25 × 105 2.67 × 105
I 1.26 × 106 1.36 × 106 7.52 × 105 4.43 × 105
The high-accuracy measurement is a key step in accurately eval-
J 4.90 × 105 5.25 × 105 2.93 × 105 1.72 × 105 uating topographical features of solar radiation over roofs. In this
K 1.25 × 106 1.35 × 106 7.50 × 105 4.42 × 105 section, high-resolution distribution of solar radiation over roofs is
L 1.27 × 106 1.36 × 106 7.56 × 105 4.45 × 105 mapped for accurately identify roof portions with high radiation
M 4.90 × 105 5.25 × 105 2.93 × 105 1.72 × 105
intensities and outline installation ranges of various solar devices.
N 1.24 × 106 1.33 × 106 6.73 × 105 4.36 × 105

5.1. Annual solar radiation and radiation contour maps

1.73 × 106 MJ in summer. Roof D, G, J or M can only obtain around
5 × 105 MJ in spring and 5.3 × 105 MJ in summer due to small build- Fig. 8 shows a high-resolution radiation map with
ing roofprints. However, for Roof I, K, L or N, the rooftop can obtain 0.035 m × 0.035 m cell size, which means that the demonstrated
around 1.2 × 106 MJ in spring and 1.3 × 106 MJ in summer. Clearly, results can achieve centimetre-level accuracy. Only Buildings A, C,
570 Y. Li et al. / Energy and Buildings 129 (2016) 563–573

Fig. 8. An annual radiation map with 0.035 m × 0.035 m cell size. (a) Solar radiation contour lines (b) Radiation areas (bar chart) and radiation yields (line chart)

Fig. 9. Topographical features of annual solar radiation over building roofs.

(a) Solar radiation contour lines
(b) Radiation areas (bar chart) and radiation yields (line chart)

D and E are selected as a case study in this section to avoid a higher 1916 × 912 pixels and 3705 MJ/m2 /year based on 4768 × 2599 pix-
computational cost induced by a large pixel matrix. The annual els in terms of different mesh sizes, and these two measurements
radiation over the roofs is estimated at 3688 MJ/m2 /year based on have a margin of error of 0.45%.
 Y. Li et al. / Energy and Buildings 129 (2016) 563–573 571

Fig. 10. Topographical features of summer solar radiation over building roofs.
(a) Solar radiation contour lines
(b) Radiation areas (bar chart) and radiation yields (line chart)

Based on the annual solar radiation yields per unit area, Fig. 9 maximum solar radiation of Roof A is lower than 4500 MJ/m2 /year
depicts a radiation contour map illustrated with contour lines, unlike the other roofs.
which shows “hills” and the steepness, and the statistical analysis As shown in Fig. 9b, solar radiation areas and radiation yields
for different radiation levels. The contour map of annual radiation with different radiation levels for each building have been depicted
visualises the regions where common solar water collectors or pho- in the bar and line charts, respectively. Specifically, the 2500 MJ
tovoltaic panels are advisable to install as the devices are able to radiation ranges have an area of 1269.72 m2 over Roof A, 537.93 m2
receive more solar radiation yields during the whole year. over Roof C, 354.83 m2 over Roof D, 538.12 m2 over Roof E, as shown
As shown in Fig. 9a, this contour map uses a 250 MJ contour in the bar chart. However, the radiation area of Roof A drops dra-
interval. The red and outermost lines outline the range of 2500 MJ matically with the annual radiation yield rises. Its 4250 MJ radiation
radiation yields, within which the annual solar radiation yields area becomes smaller than other buildings, and there is no 4500 MJ
per unit area can reach at least 2500 MJ/m2 /year. The 2000 MJ radiation region over Roof A. The line chart also illustrates the same
and 2250 MJ contour lines are neglected in the map, because they trend in the annual solar radiation. The lines of Buildings C and E
are almost identical with the 2500 MJ contour line so that it is are almost identical and their trends are similar with the line of
not easy to identify them. Within the range outlined by yellow Building D. However, the annual solar radiation yield of Building A
lines, the solar radiation over the roofs is equal to or greater than is not lower than other buildings until the range is reduced to the
2750 MJ/m2 /year. It can be clearly observed from the 3000 MJ con- contour lines of 4250 MJ/m2 /year.
tour lines that there is a significant decrease in the annual solar
radiation within Roof A’s regions that connects with other three 5.2. Seasonal solar radiation contour maps
buildings, especially the regions around northwest corners of Roofs
C, D and E. The 4000 MJ and 4250 MJ contour lines over Roof A illus- Fig. 10 illustrates the solar radiation evaluation in terms of solar
trate the most suitable places for installing solar devices. However, radiation yields per unit area in summer. Investors may benefit
from the summer radiation contour map of Fig. 10a in installing
572 Y. Li et al. / Energy and Buildings 129 (2016) 563–573

Fig. 11. Topographical features of winter solar radiation over building roofs.

solar devices such as solar heating systems with seasonal storage. as slab or underfloor heating systems, which are used to provide
The ranges outlined by contour lines may also be useful for com- more hot water produced by solar energy in winter. The figure illus-
mercial and residential zones with urban central cooling systems as trates that there is a total 400 MJ radiation area of 1754.82 m2 over
the most electricity generated from photovoltaic systems are con- roofs. Roof A accounts for 41.78%, Roof C for 21.94%, Roof D for
sumed by air conditioners in summer. As shown in Fig. 10b, the 14.33% and Roof E for 21.95%. The proportion of Roof A is greatly
range areas of the 200 MJ, 400 MJ and 600 MJ radiation can reach reduced to 8.52% when considering the 600 MJ radiation range.
2881.30 m2 , 2859.15 m2 , 2841.73 m2 , which are very close to the The corresponding radiation yield also has a dramatic decrease
total roof area of around 2882 m2 . The radiation area and solar over Roof A. This is because Roof A is usually shaded by the apart-
radiation yield of the commercial building significantly decrease ment buildings in winter so that the seasonal solar radiation yields
after the range is reduced to 800 MJ/m2 /year. Roof A’s regions per unit area cannot reach 600 MJ/m2 /year within most regions
adjacent to the three apartment buildings are not able to meet of the roof, except for the region near the southwestern corner.
1400 MJ/m2 /year. Thus, the range outlined by the 1400 MJ contour The radiation contour map also illustrates that most regions of
lines is first considered for installing the above-mentioned solar 600 MJ/m2 /year are located between elevator shafts for the roofs
devices in order to receive more solar radiation in summer. More of the apartment buildings.
solar devices could be extended to the 1200 MJ radiation range.
These areas mainly concentrated in Roof A’s central regions that
do not adjoin other buildings, and in regions to the south of and 6. Conclusions
between elevator shafts over Roofs C–E. In addition, it is clearly
observed from Fig. 10b that the radiation area over the commercial This paper selects solar radiation evaluation over roofs as
building is not smaller than ones of the apartment buildings for all building energy research due to a rapid growth in solar energy
solar radiation ranges. applications and interest in sustainable development of cities.
The winter radiation contour map and its statistical analysis are However, previous research paid extensive attention to the LiDAR-
presented in Fig. 11. The evaluation may be beneficial for investors based procedures that are more suited to existing buildings. Taking
to consider the installation of solar hydronic heating devices, such advantage of future urban planning, this research establishes a
pixel-based methodology for estimating solar energy potential over
 Y. Li et al. / Energy and Buildings 129 (2016) 563–573 573

roofs so that investors may benefit from assessment measures. [9] J. Hofierka, J. Kaňuk, Assessment of photovoltaic potential in urban areas using
Buildings from urban planning of Xiuyuan Eco-city are chosen as a open-source solar radiation tools, Renew. Energy 34 (10) (2009) 2206–2214.
[10] M. Bizjak, B. Žalik, N. Lukač, Evolutionary-driven search for solar building
case study. The shadow analysis of SketchUp software is employed models using LiDAR data, Energy Build. 92 (2015) 195–203.
in this research to produce a series of greyscale images. Each pixel [11] C. Hachem, A. Athienitis, P. Fazio, Parametric investigation of geometric form
value is then read by MATLAB to judge whether it is a shaded pixel or effects on solar potential of housing units, Sol. Energy 85 (9) (2011)
1864–1877.
not. The estimation of solar radiation yields is mathematically for- [12] J. Ordóñez, E. Jadraque, J. Alegre, G. Martínez, Analysis of the photovoltaic
mulated in each pixel. The solar radiation yields are calculated by solar energy capacity of residential rooftops in Andalusia (Spain), Renew.
considering multiple instantaneous solar irradiances using a spe- Sustain. Energy Rev. 14 (7) (2010) 2122–2130.
[13] M. Šúri, T.A. Huld, E.D. Dunlop, PV-GIS: a web-based solar radiation database
cific time step and visualised by image processing using MATLAB.
for the calculation of PV potential in Europe, Int. J. Sustain. Energy 24 (2)
Results reveal the spatial and temporal variation of solar (2005) 55–67.
radiation yields over roofs. The maximum and minimum [14] J. Liang, J. Gong, J. Zhou, A.N. Ibrahim, M. Li, An open-source 3D solar radiation
model integrated with a 3D Geographic Information System, Environ. Model.
annual radiation yields are estimated at 4717.72 MJ/m2 /year and
Softw. 64 (2015) 94–101.
342.58 MJ/m2 /year within the study area, respectively. The results [15] G. Agugiaro, F. Remondino, G. Stevanato, R.D. Filippi, C. Furlanello, Estimation
also demonstrate that the measurement can achieve centimetre- of solar radiation on building roofs in mountainous areas, Int. Arch.
level accuracy so that the modelling can be used to identify the Photogramm. Remote Sens. Spat. Inf. Sci. 38 (3/W22) (2011) 155–160.
[16] A. Tereci, D. Schneider, D. Kesten, A. Strzalka, U. Eicker, Energy saving
portions of roofs with high solar potential. For each apartment potential and economical analysis of solar systems in the urban quarter
building, 70% of roof areas can receive 3500 MJ/m2 /year or more Scharnhauser Park, in: Proceedings of the ISES Solar World Congress 2009:
solar radiation yields, and around 20% of roof areas can obtain Renewable Energy Shaping Our Future, Document Transformation
Technologies, Johannesburg, South Africa, 2009.
4500 MJ/m2 /year or more. Overall, the proposed radiation contour [17] J. Lee, S. Zlatanova, Solar radiation over the urban texture: LIDAR data and
maps illustrate the ranges for installing different types of solar col- image processing techniques for environmental analysis at city scale, in: J.
lectors. Particularly, the regions between elevator shafts are very Lee, S. Zlatanova (Eds.), 3D Geo Information Sciences, Springer Berlin,
Heidelberg, 2009, pp. 319–340.
suitable to install solar energy devices. [18] P. Redweik, C. Catita, M. Brito, Solar energy potential on roofs and facades in
The method presented in this paper can be applied to assess- an urban landscape, Sol. Energy 97 (2013) 332–341.
ment measures of solar radiation for energy investors and urban [19] M. Šúri, T.A. Huld, E.D. Dunlop, H.A. Ossenbrink, Potential of solar electricity
generation in the European Union member states and candidate countries,
planners, and the results can be utilised in government planning to
Sol. Energy 81 (10) (2007) 1295–1305.
raise awareness of the use of renewable energy sources. Further- [20] M. Neteler, H. Mitasova, Open Source GIS: A GRASS GIS Approach, Springer,
more, the proposed algorithm can be easily extended to evaluation New York, 2008.
[21] J.B. Kodysh, O.A. Omitaomu, B.L. Bhaduri, B.S. Neish, Methodology for
of solar radiation over pitched roofs and facades by producing
estimating solar potential on multiple building rooftops for photovoltaic
greyscale images from the direction perpendicular to the building systems, Sustain. Cities Soc. 8 (2013) 31–41.
surfaces. The specific application of solar devices, such as domestic [22] N. Lukač, B. Žalik, GPU-based roofs’ solar potential estimation using LiDAR
hot water or photovoltaic systems, are expected to be incorporated data, Comput. Geosci. 52 (2013) 34–41.
[23] J.H. Kämpf, D. Robinson, A hybrid CMA-ES and HDE optimisation algorithm
into the future research to provide a more complete picture. with application to solar energy potential, Appl. Soft Comput. 9 (2) (2009)
738–745.
References [24] L. Kumar, A.K. Skidmore, E. Knowles, Modelling topographic variation in solar
radiation in a GIS environment, Int. J. Geogr. Inf. Sci. 11 (5) (1997) 475–497.
[25] C. Catita, P. Redweik, J. Pereira, M.C. Brito, Extending solar potential analysis
[1] B.R. Gautam, F. Li, G. Ru, Assessment of urban roof top solar photovoltaic in buildings to vertical facades, Comput. Geosci. 66 (2014) 1–12.
potential to solve power shortage problem in Nepal, Energy Build. 86 (2015) [26] N. Lukač, S. Seme, D. Žlaus, G. Štumberger, B. Žalik, Buildings roofs
735–744. photovoltaic potential assessment based on LiDAR (Light Detection And
[2] World Energy Council, 2014 World Energy Issues Monitor, World Energy Ranging) data, Energy 66 (2014) 598–609.
Council, London, United Kingdom, 2014. [27] WeatherChina.org, Jinan Weather Forecast and Jinan Climate, in,
[3] X.Q. Zhai, R.Z. Wang, Y.J. Dai, J.Y. Wu, Y.X. Xu, Q. Ma, Solar integrated energy WeatherChina.org, 2015.
system for a green building, Energy Build. 39 (8) (2007) 985–993. [28] TimeBie, Daylight Hours of Jinan China, in, timebie.com, 2016.
[4] N. Lukač, D. Žlaus, S. Seme, B. Žalik, G. Štumberger, Rating of roofs’ surfaces [29] Shandong Jianzhu University, Specific urban planning of the Xiuyuan Eco-city,
regarding their solar potential and suitability for PV systems, based on LiDAR in, The Zhangqiu Urban Planning Bureau of Jinan City, Jinan, China, 2015.
data, Appl. Energy 102 (2013) 803–812. [30] O.N. Svensson, E. Svensson, Implementation of solar energy in eco-cities, in:
[5] X. Fang, D. Li, Solar photovoltaic and thermal technology and applications in School of Industrial Engineering and Management, KTH Royal Institute of
China, Renew. Sustain. Energy Rev. 23 (2013) 330–340. Technology, 2012.
[6] International Energy Agency, Solar Energy Perspectives, International Energy [31] J. Wang, J. Feng, A. Yuan, Calculation and distributive characteristics of solar
Agency, Paris, France, 2011. radiation in shandong province, Meteorol. Sci. Technol. 34 (1) (2006) 98–101.
[7] J. Zuo, Z.-Y. Zhao, Green building research–current status and future agenda: a
review, Renew. Sustain. Energy Rev. 30 (2014) 271–281.
[8] M. Socorro García-Cascales, M. Teresa Lamata, J. Miguel Sánchez-Lozano,
Evaluation of photovoltaic cells in a multi-criteria decision making process,
Ann. Oper. Res. 199 (1) (2012) 373–391.