IOP Conference Series: Earth and Environmental Science

PAPER • OPEN ACCESS

Modelling and simulation for installation feasibility of standalone

photovoltaic system for Quetta, Pakistan
To cite this article: T Z Muhammad et al 2018 IOP Conf. Ser.: Earth Environ. Sci. 188 012008

View the article online for updates and enhancements.

This content was downloaded from IP address 181.214.46.149 on 31/10/2018 at 00:47

NEFES 2018 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 188 (2018)
1234567890 ‘’“” 012008 doi:10.1088/1755-1315/188/1/012008

Modelling and simulation for installation feasibility of

standalone photovoltaic system for Quetta, Pakistan

T Z Muhammad1, 3, K Cheng2, 3 and X Y Ma2

1
Solar Energy Research Institute, Northwestern Polytechnical University, Xi'an
710072, China
2
School of Power and Energy, Northwestern Polytechnical University, Xi'an 710072,
China

Email: muhammadtouseefzafar195@mail.nwpu.edu.cn /cksolar@163.com

Abstract. The impact and significance of renewable energy in developing countries is

increasing rapidly in latest years. There is much need of energy dependence to be based on off
grid photovoltaic systems in developing countries like Pakistan which are suffering from
severe power shortfalls annually and load shedding concern. In this paper, a concept
assessment and estimation of standalone photovoltaic system is done for the site of Quetta,
Pakistan in comparison to the location of the Copiapo-Chamonate. Copiapo-Chamonate site is
situated in Atacama Desert which is often labelled as the best solar irradiation receiving
location in the world with maximum solar insolation reaching to 9.28 kWh/m2/day. The
systems for both locations is modelled and simulated in PVSYST for a constant load of 11.9
kWh/day with seasonal tilt adjustment. According to the simulation, results of the parameters
including performance ratio, solar fraction, energy supplied to the load etc. show that Quetta
meet its individual household demand quite efficiently.

1. Introduction
The key issue in the world lately has been the damage of the climate conditions by greenhouse gases.
The emission of greenhouse gases is related to various factors around the world. One of the main
reasons is the burning of the conventional fuels like coal, oil and natural gas [1]. In order to reduce the
pollution concerns and to meet the swiftly increasing energy demand, government is paying huge
attention towards the non-conventional sources such as solar, wind and hydro [2]. Pakistan is a
developing country which is unable to meet its energy demand with acute energy shortfalls reaching
5500 MW in 2015. The major reason of shortfall is heavy reliance on the thermal energy rather
renewable energy [3]. The oil and gas are the major fuels in Pakistan to produce electricity with 80%
and 6% of the whole generation respectively. However, the renewable energy has not been exploited
fully until now. Pakistan is a favorable country for the solar energy exploitation which receives 15.5
×1014 kWh/year. The solar energy potential can be used to produce 1600 GW per year [4]. Quetta is
one of the major cities of Pakistan with considerably high level of irradiation up to 5.54 kWh/m2/day
as compared to other main populated cities of Pakistan [5]. The similar country which is undergoing
an enormous surge in exploiting the solar energy is Chile. In March 2016, the installed capacity of PV
was 1.1 GW but many new projects have been carried out. Most of these projects are based in
Atacama Desert. This place has been known as one of the places with the highest surface solar
irradiation. The main characteristics of Atacama Desert are high altitude and clear sky days [6]. The
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
NEFES 2018 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 188 (2018)
1234567890 ‘’“” 012008 doi:10.1088/1755-1315/188/1/012008

mean clearness index of Atacama Desert is 3% with global horizontal irradiation reaching up to 2500
kWh/m2 and maximum temperature of 38oC [7]. On the other side, more than 40,000 villages have no
access to the nationalpower grid in Pakistan. The southern province of Pakistan i.e. Sindh and
Baluchistan are suitable for solar energy with maximum daily sun hours of 7-8 [8]. Chile, similarly, is
in its way to maximize the proportion of its solar energy percentage in national energy market. The
20% energy is intended to be produced by the renewable energy resources until 2020. However, there
are lot of hindrances in the way of achieving this target such as lack of solar database and utilization
mechanism [9]. The 50% Chile’s economy is based on mining industry (mainly coal industry) and
80% of energy demand is needed in the Atacama Desert [10]. Many countries are now adopting the off
grid photovoltaic systems in public policies to ensure the easy access of energy by installation on
remote places [11]. The standalone photovoltaic systems can be regarded as the best choice for rural
areas because of simple installation system topology and control with a load demand of 1-100 KW
[12]. Two geographical location are selected to be used in this study for power output potential
production of off-grid photovoltaic system; Quetta, Pakistan and Copiapo-Chamonate (CC), Chile.
The simulation software used here is PVSYST which is developed by EPFL for the reliable and
efficient evaluation of various configurations of both off grid and on grid photovoltaic systems [13].
The fundamental purpose of this paper is to determine the potential of Quetta in comparison to CC
for generating power from off grid photovoltaic system for a standard household consumer. This paper
is structured in five sections. The first section is for the general description and introduction of the
problem considered. The second section is for determining the geographical parameters including
albedo and tilt angle. The third section is for designing the electrical parameters including electrical
components and loads. In the fourth section the discussion of simulation results is presented and, in the
section 5, the conclusions are drawn.

2. Geographical site parameters

2.1. Geographical location

Figure 1. Global Irradiation (kWh/m2) for Quetta Figure 2. Monthly Average Temperature (oC) for
and CC. Quetta and CC.

The sites chosen are Quetta and Copiapo-Chamonate (CC). The geographical coordinates of
CapiapoChamonate are 27.30 S latitude and 70.42 W longitude. The elevation of the site is 301 m.
The time zone is UT -4. The latitude and longitude of Quetta site is 30.1 N and 66.95 E respectively
with an elevation of 1702 m with time zone of UT +5. Since, CC is located in the Southern
Hemisphere and Quetta is situated in the Northern Hemisphere, the winter and summer months are
different for these two places. In CC, the winters are in the months from May to August and the
summer lasts from September to December. July is the month with lowest average temperature of
11.8oC and January with the highest temperature of 22.4oC. The temperature variance between
summer and winter at CC is not significant. Quetta has a significant trend of variation in climate

2
NEFES 2018 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 188 (2018)
1234567890 ‘’“” 012008 doi:10.1088/1755-1315/188/1/012008

between summers and winters. The summer season is from March to October with 30.5 oC as the
highest average temperature and the winter season is from November to February with average lowest
temperature of 7.3oC.
The global irradiation GI (kWh/m2) and average month temperature of the two locations is shown
in figures 1 and 2, respectively.

2.2. Tilt angle

The seasonal based tilt adjustment is used in this system. The average tilt angles for summer and
winter months can be calculated from the equations (1) and (2).

Tsa = T S
(1)
NS

TWa = T W
(2)
NW
Where Tsa,Twa, Ts, Tw, Ns and Nw are the summer average tilt angle, winter average tilt angle,
individual summer months tilt angle, individual winter months tilt angle, number of summer months
and number of winter months respectively. The individual tilt angles for every month can be obtained
from [13]. The tilt angles for Quetta in summer and winter months are 16o and 53o respectively and
13o and 47o for Copiapo- Chamonate, respectively.

2.3. Albedo
The albedo is the characteristics of a surface linked with its proportion of reflectance or absorption.
The usual range of albedo is 0.1-0.5 [14]. The dark black surface is considered to be at 0 while the
pure white surface is termed to be 1. The albedo of the two sites considered can be obtained from [15].
The albedo of the two geographical sites is given in the table 1.

Table 1. Albedo of Quetta and CC.

Albedo Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec
Quetta 0.16 0.16 0.17 0.24 0.25 0.28 0.28 0.28 0.25 0.25 0.25 0.20
CC 0.13 0.13 0.11 0.11 0.11 0.10 0.10 0.10 0.10 0.11 0.12 0.13

3. Electrical system

3.1. Electrical loads

The total daily energy load demand is 11.9 kWh/day and 4352 kWh/year. The local demand is
designed according to the needs of an average domestic consumer. The consumption is mainly
distributed along the duration time from 9 am to 6 pm. The loads used for the calculation are given in
the table 2.
Table 2. Annual values of load (Wh/day).
Number Power Use Energy
Lamps(LED or fluo) 10 10W/Lamp 5h/day 500Wh/day
TV/PC/Mobile 2 100W/Lamp 5h/day 1000Wh/day
Domestic appliances 1 500W/Lamp 4h/day 2000Wh/day
Fridge/Deep-freeze 2 24Wh/day 1000Wh/day
Dish-&Cloth-washers 1 2Wh/day 2000Wh/day
Ventilation 1 100W tot 24h/day 2400Wh/day

3
NEFES 2018 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 188 (2018)
1234567890 ‘’“” 012008 doi:10.1088/1755-1315/188/1/012008

Air condition 1 1000W tot 3h/day 3000Wh/day

Stand-by consumers 24h/day 24Wh/day

3.2. Electrical components

The values of electrical components used in the simulation are provided built-in in PVSYST software.
[15] The photovoltaic module used here is Si-poly Model named AC- 210 P/ 156-54S from Axitec
USA. The nominal power of the module is 210 W. The PV array has 16 modules; 4 in series and 4 in
strings. The total PV array power is 3360 W at STC conditions and 3058 W at 50oC. The total module
area is 23.5 m2. The maximum power point voltage and current of PV array are 102 V and 30 A
respectively. The Dural SC battery of Electrona is used in this system. The voltage is 72 V and
nominal capacity is 600 Ah. The high voltage is chosen for the battery system to reduce the cable
wiring losses. There are 12 batteries connected in series and 2 in parallel. The temperature 25oC is
considered as the temperature for keeping the batteries. The Universal controller is used for the direct
control of the sizing of the battery pack and PV array. These components will automatically adjust to
the whole system with minimum losses and maximum power production. The charging and
discharging limits of universal controller are calculated according to the limits of state of charge
(SOC) as shown in equations (3) and (4).
0.92  SOCC  0.75 (3)

0.20  SOCd  0.45 (4)

Where SOCc and SOCd are the charging and discharging percentages of the battery by the universal
controller, respectively. The approximate limit of voltage of the battery in charging mode is from 75.2
V to 82.3 V and for the discharging mode is from 70.8 V to 73.3 V. The both systems are designed for
LOL of 3% and the 3 days of autonomy. The general electrical schematic diagram is given in figure 3.

Figure 3. Schematic diagram of standalone photovoltaic system.

4. Model simulation and results

The available energy at CC is 4839 kWh/year and the specified production is 1440 kWh/kWp/year.
Whereas, at the Quetta, the specified production is 1448 kWh/kWp/year and the available energy is
5001 kWh/year. The monthly available energy at Quetta and CC is shown in figure 4. The maximum
available energy at CC is in the summer with maximum of 511.5 kWh in December and minimum in
June in 305.3 kWh. The maximum available energy at Quetta is in June i.e.481.7 kWh and minimum
in March i.e. 350.2 kWh. The available energy at both sites in summers is almost same but Quetta has
42.5 kWh more available energy than CC in winters. The performance ratio of Quetta and CC is
56.9% and 56.2% respectively. The performance ratio is independent of the orientation and location
and it is the ratio of actual obtained power versus the theoretical predicted power. The low PR at
Quetta can be obtained to the output variation caused by operation of the system on STC and actual
operating condition. While, the solar fraction of Quetta is 99.6% and for CC is 95.8%. The CC failed

4
NEFES 2018 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 188 (2018)
1234567890 ‘’“” 012008 doi:10.1088/1755-1315/188/1/012008

to meet the actual user demand and provided only 88.9% of the total load. The Quetta lacked only in
January with the provision of 95% of the load need. The solar fraction for both sides is given in figure
5. The performance ratio and solar fraction can be calculated from the equations (5) and (6)
respectively.
Yf
PR = (5)
Yr

Eu
SF = (6)
El
Where PR, SF, Yf, Yr, Eu, El are performance ratio, solar fraction, system yield, nominal system
yield, energy supplied to the user and energy need of the user respectively. The performance ratio PR
is the ratio of actual power produced by the PV array and the nominal power expected to be produced
by PV array at STC conditions. The Eu is the actual power supplied to the load and El is the actual
load demand. The solar fraction SF is the ratio of energy provided from the array to the load versus the
actual demand of the load.

Figure 4 Solar Fraction (SF) at Quetta and CC. Figure 5 Comparison between Energy supplied to
the load Eu and Energy demanded by the load El.

The missing energy at CC has shown a downfall curve in the winters especially in June. The Euat
Quetta performed at capacity despite in January while, in winters, CC was unable to supply the load
demand with 57.6 kWh energy shortfall in June. The energy supplied to the load Euand energy
demand of the load El for both sites is given in figure 6.
The missed energy is 183.66 kWh at CC with maximum of 57.65 kWh in June. The Quetta site
performed well in terms of missing energy that is 17.93 kWh only in January. The total unused energy
for Quetta is 534 kWh with maximum in the month of June and considerable average of 21.625 kWh
in the winters. Whereas, the excess energy at CC is in the months of Summer with an average of
65kWh with the maximum value of 118.4 kWh in December. The various losses of the off grid
photovoltaic system at CC and Quetta is given in figure 7. The most considerable losses at both sides
is in the working of the system at MPP (maximum power point) and temperature loss. Temperature
loss at Quetta is 6.6% and 5.5% at CC. The Quetta site has more temperature loss because of high
temperatures as compared to CC. However, the performance of CC at MPP is very low and 1872 kWh
energy is lost.

5
NEFES 2018 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 188 (2018)
1234567890 ‘’“” 012008 doi:10.1088/1755-1315/188/1/012008

Figure 6. Available Energy Eav (kWh) at CC Figure 7. Major Losses at Quetta and CC.
and Quetta.

5. Conclusion
This paper simulates and compares the two geographical sites of Quetta and CC for an off grid
photovoltaic system. CC, Atacama is considered to be one of the best receiving solar irradiation place.
PVSYST model results show that Quetta performs well in comparison to CC. The following results
are obtained from the studies done in this paper.
⚫ Solar Fraction at Quetta site was 100% the whole year, however, lacking only in January with
missing energy of 17.93kWh
⚫ The energy used by the load at Quetta is 3.807% more as compared to CC
⚫ Performance ratio PR at Quetta is 0.7% more than at CC
⚫ The Quetta site has produced 3.96% excess annual energy as compared to CC
This paper can provide the further research on modeling the off grid photovoltaic systems in
Pakistan for domestic consumers. According to the author, special attention should be paid in reducing
the temperature losses, tilt angle adjustment and MPP running losses. The photovoltaic roofs with low
reflectance and less albedo are recommended. The tilt angles should be modified according to the
location temperature to increase performance ratio PR.

References
[1] Danish, Wang ZH 2017 Role of renewable energy and non-renewable energy consumption on
EKC: Evidence from Pakistan J. Clean. Prod. 156 855-64.
[2] Kamran M 2018 Current status and future success of renewable energy in Pakistan Renew.
Sustain. Energ. Rev. 82 609-17.
[3] Qureshi T M, Ullah K and Arentsen M J 2017 Factors responsible for solar PV adoption at
household level: A case of Lahore Pakistan Renew. Sustain. Energ. Rev.78 754-63.
[4] Zafar U, Rashid T U, Khosa A A, Khalil M S and Rahid M 2013 An overview of implemented r
enewable energy policy of Pakistan Renew. Sustain. Energ. Rev. 82 654-65.
[5] Khalid A and Junaidi H 2013 Study of economic viability of photovoltaic electric power for
Quetta-Pakistan Renew. Energ. 50 253-8.
[6] Ferrada P, et al 2017.Potential for photogenerated current for silicon based photovoltaic
modules in the Atacama Desert Sol. Energ.144 580-93.
[7] Olivares D, Ferrada P, de Matos C, Marzo A, Cabrera E, Portillo C and Llanos J
2017.Characterization of soiling on PV modules in the Atacama Desert Energy Procedia 124
547-53.
[8] Sheikh M A 2010 Energy and renewable energy scenario of Pakistan Renew. Sustain. Energ.
Rev. 14 354-63.
[9] Escobar R A, Ortega A, Cortés C, Pinot A, Pereira E B, Martins F R and Boland J 2014. Solar
energy resource assessment in Chile: satellite estimation and ground station measurement
Energy Procedia 57 1257-65.

6
NEFES 2018 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 188 (2018)
1234567890 ‘’“” 012008 doi:10.1088/1755-1315/188/1/012008

[10] Moreno-Leiva S, Diaz-Ferran G, Haas J, Telsnig T, Diaz-Alvarado F A, Palma-Behnke R,

Kracht W, Roman R, Chundinzow D and Eltrop L 2017 Towards solar power supply for
copper production in Chile: Assessment of global warming potential using a life-cycle
approach J. Clean. Prod. 164 242-9
[11] Bellinaso L V, Schwertner C D and Michels L 2016 Price-based power management of off-grid
photovoltaic systems with centralised dc bus IET Renew. Power Gener. 10 1132-9.
[12] Salas V, Olias E, Barrado A and Lazaro A.Review of the maximum power point tracking
algorithms for stand-alone photovoltaic systems Sol. Energ. Mater. Sol. Cells 90 1555-78.
[13] NASA’s surface meteorology and Solar Energy https://eosweb.larc.nasa.gov/cgibin/sse/sse.cgna
[14] Bonafoni S, Baldinelli G, Rotili A and Verducci P 2017 Albedo and surface temperature
relation in urban areas: Analysis with different sensors Urban Remote Sensing Event (Dubai,
United Arab Emirates, 6-8 March 2017) pp 1-4
[15] www.pvsyst.com/en/about-us/our-objectives

7
IOP Conference Series: Earth and Environmental Science

PAPER • OPEN ACCESS Related content

- Research on the Application of Cluster
Application of clustering analysis in the prediction Analysis in Criminal Community Detection
Tian Junjing
of photovoltaic power generation based on neural - Research of PV Power Generation MPPT
based on GABP Neural Network
network Yu Su and Xianfu Lin

- Optimal Output of Distributed Generation

To cite this article: K Cheng et al 2017 IOP Conf. Ser.: Earth Environ. Sci. 93 012024 Based On Complex Power Increment
D Wu and H Bao

View the article online for updates and enhancements.

This content was downloaded from IP address 222.90.67.226 on 04/03/2019 at 15:43

NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

Application of clustering analysis in the prediction of

photovoltaic power generation based on neural network

K Cheng1,2, L M Guo1, Y K Wang1 and M T Zafar1

1
School of Power and Energy, North-western Polytechnical University, Xi’an 710072,
Shaanxi province, China

E-mail: cksolar@163.com

Abstract. In order to select effective samples in the large number of data of PV power
generation years and improve the accuracy of PV power generation forecasting model, this
paper studies the application of clustering analysis in this field and establishes forecasting
model based on neural network. Based on three different types of weather on sunny, cloudy
and rainy days, this research screens samples of historical data by the clustering analysis
method. After screening, it establishes BP neural network prediction models using screened
data as training data. Then, compare the six types of photovoltaic power generation
prediction models before and after the data screening. Results show that the prediction
model combining with clustering analysis and BP neural networks is an effective method to
improve the precision of photovoltaic power generation.

1. Introduction
With the expansion of installed capacity of photovoltaic power generation, its proportion in the power
grid is also increasing these days. The random fluctuation characteristics of its output will cause the
considerable impact on the power grid system, which indirectly affects the operation of the security
and stability of the power system. Therefore, it is extremely important to predict the photovoltaic
power generation effectively to ensure the security and stability of the power system. At present, some
forecasting models for photovoltaic power generation are SVM model [1,2], ARMA model [3], gray
theoretical model [4] and neural network model, etc. [5,6]. However, the neural network model
generally has better performance among them [7]. Therefore, this study uses this method to predict the
photovoltaic power generation.
In the process of BP neural network prediction, the quality of the input samples will directly affect
the accuracy of the prediction model. Practically, when the photovoltaic power plant collects and
stores the running data, the errors occur which form the abnormal samples. These anomaly samples
deviate from the factual basis and the corresponding weather patterns are not correctly described
according to its established model. Therefore, it is necessary to screen out these anomalous samples.
Manually, it is very difficult to screen out anomalous samples because of a large amount of data
recorded in the past years. Therefore, most of the research papers use raw data collected from the
untreated power plants [4-8], and only a few papers deal with filtered data [9]. However, these papers
do not focus on the specific screening method. There are some papers, which use the wavelet analysis
method, the mean fill method and other methods to deal with the original data [10,11]. However; there
is no specified method to verify the accuracy and effectiveness of the power generation forecasting
model.

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

Therefore, based on the statistical data of different kinds of weather, this paper cleans the historical
data using the clustering analysis. Then, based on the BP neural network, 6 different kinds of
photovoltaic power generation forecasting models are established and verified in the sunny, cloudy
and rainy days in MATLAB software [12,13]. The results show the validity and accuracy of the
clustering analysis of historical data.

2. Prediction model of photovoltaic power generation

2.1. Prediction model of photovoltaic power generation

In this paper, the clustering analysis and BP neural network are combined to process the raw data
collected by the power station. The processed data is used as the input of the neural network to
establish a reliable photovoltaic power generation forecasting model. The data in the research process
comes from the actual record of the data collection device of Xi'an Jiayang New Energy Co., Ltd. PV
power plant. The model is designed based on the photovoltaic system installed on the roof of the
building. The total installed capacity of the photovoltaic power station is 1.2 MW. The measured data
of the power station in the time period of 2012.01-2015.01 is selected as the original sample. The
selected samples consist of total 868 data samples.
The main process of the model is shown in figure 1.

Collec t historic al data

(meteorological data, PV power
generation, etc.)

Weat her type

Sunny Cloudy Rainy

Cluster
Forecasting analysis (data
model based on cleaning)
BP neural
network Forecasting
model based on
BP neural
network
Input simulation data Input simulation data

Forecast power Forecast power

generation generation

Calculate MAPE Calculate MAPE

Figure 1. The framework of photovoltaic power generation forecasting model.

 Firstly, 3 typical weather types of sunny, cloudy and rainy days are taken as examples to
analyze the theoretical radiation, actual radiation and power generation. The outliers are
filtered out according to the cluster pedigree.
 Secondly, the historical data of 2012.01~2014.01 is used as the training sample. The
theoretical radiation and the highest and the lowest temperatures are used as the input of BP
neural network. The historical data collected by the data collector of PV power plant are used
as the output of BP neural network. The neural network model is trained to establish the
prediction models of 6 kinds of photovoltaic power generation
 Thirdly, the simulation data of 2014.02 ~ 2015.01 is put into the prediction model of
photovoltaic power generation. The purpose is to make use of the BP neural network's

2
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

nonlinear learning ability to predict the daily power generation under different weather types
 Finally, in order to verify the need of data screening, the MAPE (the mean absolute percentage
error) value is calculated for the predicted power generation and the actual generation and the
performance is evaluated.

2.2. Clustering analysis model

Clustering analysis is a process of dividing an object of observation into several groups or classes
based on certain quantitative features. The division is done in such a way that the same class within
the data object has a high degree of similarity and low degree of similarity between various types of
data object. The main purpose of clustering is to bring similar things together.

2.2.1. Clustering method. According to the different clustering methods, clustering analysis can be
divided into system clustering method, adding method, decomposition method and dynamic
classification method, etc. [14]. In this paper, the purpose of clustering analysis is to deal with the
outlier data (abnormal samples) from the original data to improve the quality of the original data.
Therefore, this paper uses the system clustering method. The main algorithm steps are as follow:
 To convert the objects in the sample set;
 The processed objects are divided into n classifications and every class contains an object;
 Calculating the distance between each of the two classes;
 Merging the nearest two classes as a new class;
 Calculating the distance between the new class and the current class. If the distance is close
enough, then merge into a class. This process continues until all classes (or samples) are
combined. If this step works well then we will move further to step 6, otherwise, to step 4.
 Drawing the clustering pedigree of the sample set, as shown in figure 2;
 According to the number of classes to be classified, the corresponding classification results are
obtained.

Tree with average connectivity（Between groups）

Cluster after adjusting the distance
0 5 10 15 20 25
6

2
class a

4
Y

class b
1

class c
5

Figure 2. Clustering pedigree chart.

Figure 2 shows the data clustering results for randomly selected eight groups on rainy day. The
sample can be clustered into two, three, four or five categories. As shown in the figure, when the class

3
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

distance is 5, the samples can be clustered into three classes. The samples 6, 7, 2, 3 and 4 are class a,
sample 1 and sample 8 are b, and sample 5 is c.

2.2.2. Select the metric. The metric directly affects the quality of the cluster, so the statistical analysis
of the historical data of the database is carried out to determine the metric. In the figures 3-5 of cloudy
weather, some samples are found away from the main data sets which are outliers. While, the data
distribution exhibits a distinct oblique strip.

4000 30

Actual daily radiation/ MJ·m-2

Power generation/ kWh

3000
20

2000

10
1000

0 0
0 5 10 15 20 25 30 5 10 15 20 25 30
Actual daily radiation/ MJ·m-2 Theoretical daily radiation/ MJ·m-2

Figure 3. Scatter plots of actual radiation and Figure 4. Scatter plots of theoretical and actual
power generation. radiation.

3000
io n /k W h

2000
P o w er g en erat

1000

30
0
/㎡
25
The 0
oret 20
on/MJ
ical 1 0 iati
radi
atio 15
al rad
u
n/M 20
J/㎡ 10 Act
30

Figure 5. Spatial scatter plots of actual radiation, theoretical radiation and power generation.

According to the previous study results [15], the data covered by the oblique range belong to a
class. Selecting the cosine coefficient as the criteria, the data on the oblique bar is classified as valid
data and other data is classified as the outlier which is far from the oblique bar. After the outlier
deletion, the discrete degree of the sample is optimized. The cosine distance is measured according to

4
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

formula (1):
0

 X ik X jk
k 1
dij  cos(ij )  (1)
n n

 X ik2  X 2jk
k 1 k 1

In the above formula, d ij -the distance between simples X ik and X jk ; X ik - the k-the variable of
the i sample; X jk - the k-the variable of the j sample; n - number of samples.

2.3. Neural network model

Fundamentals. The neural network is a network system that performs the parallel processing and non-
linear transformation of information same as human brain. The BP neural network chosen in this paper
is a forward process using error back propagation algorithm of the network. Its structure is shown in
figure 6.

Error back propagation (learning algorithm)

i j k
X1

Input X2
mode
…

Xm
wij w jk
m q L
input layer hidden layer output layer
Figure 6. Structure of BP neural network.

Where wij is the connection weight between the input layer node and the hidden layer node, and w jk is
the connection weight between the hidden layer node and the output layer node. The input of the
hidden layer and the output layer node is the weighted sum of the output of the previous node [16]. In
this paper, a three-layer neural network is used which contains only one hidden layer.
Evaluation of photovoltaic power generation forecasting model. In this paper, a two-layer forward
neural network based on Levenberg-Marquardt (LM) algorithm is adopted. The neural network
toolbox of Matlab 2010b is used to build and train the network and obtain the simulation results. The
training data and test data are randomly selected by neural network as 90% and 10% respectively of
the total data. In order to ensure the credibility of the results, test data is not used in the training part.
The simulation results use the average absolute percentage error (MAPE) to evaluate the entire
system, which is calculated as (2):
N P i Pi
 100 
MAPE   
 N  i 1
f

Pa i
a
% (2)

In the above formula, N - the total number of data; Pf - the actual power generation; Pa - predicted
power generation; i - the data serial number. The lower MAPE value ensures the better predictive
effects.

5
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

3. Clustering process and results

The original data of three kinds of weather types such as sunny, cloudy and rainy days are extracted by
SPSS (statistical analysis software). The cosine coefficient is used as the criteria to cluster and analyze
the different weather types respectively. When the number of clusters is 4, for sunny days, according
to the clustering pedigree, as class has 119, bs class has 23, cs class has 5, and ds class has 3; for rainy
days, according to the cluster spectrum, ar class has 4, br class has 2, cr class has 19, and dr class has
36; for cloudy days, ac class has 10, bc class has 30, cc class has 61, and dc class has 67, comparing the
characteristics of the number of samples contained in each cluster and screening a few samples with
the least number of clusters. Figure 7 shows the filtered cloudy sample scatter plots.

30
Actual daily radiation/MJ·m-2

20

10

0
5 10 15 20 25 30
Theoretical daily radiation/ MJ·m-2
Figure 7. The theoretical and actual daily radiation scatter plots after screening in cloudy.

The clustering results of the various data samples are shown in table 1. In the sunny days, the
atmospheric state is relatively stable and the reverse radiation of the atmosphere is weak. Therefore,
the theoretical daily radiation is close to the actual amount of radiation in the whole day, and the
degree of data discrimination is small and the screening ratio is 10.7%. In the cloudy days will have
large random variation in terms of clouds and the similarity of the data is low, the screening ratio is
13.1%. In the rainy days, the atmospheric state fluctuates at a higher degree and the screening ratio is
also large which is 21.3%.

Table 1. The clustering result statistics for all kinds of sample data.
Weather type Sunny Cloudy Rainy Total
Before screening 150 168 61 379
After screening 134 146 48 328
Screening ratio 10.7% 13.1% 21.3% 13.5%

4. Predictive results and analysis

4.1. Forecast results

In order to verify the effectiveness of clustering analysis for the prediction of photovoltaic power
generation, the prediction model of PV generation is established with raw data as a comparison. Firstly,
cluster analysis is used in data cleaning for different kinds of weather. Then, corresponding type of
model is selected from forecasting models and simulation data is entered. Finally, the power
generation is predicted.
Figures 8-10 shows the predicted power generation, actual power generation, and the data for the

6
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

three kinds of weather before and after the clean-up process (the day number in the x axis is the data of
the forecast in a year). From the results, it can be seen that the difference between the predicted value
and the actual value of the model established by the weather type is relatively large before the data is
cleaned up. After data clearing, the difference between the predicted value and the actual value
become small. The predicted value of the model after the cluster analysis is closer to the actual value.

4000 Actual power generation Actual power generation

Before screening Before screening
Power generation/ kWh

3000 After screening

Power generation/ kWh

3000 After screening

2000 2000

1000 1000

0
20 80 146 206 263 335 0
2 45 112 148 272 329
Day number Day number

Figure 8. The relationship between the actual Figure 9. The relationship between the actual
power generation and the prediction of power power generation and the prediction of power
generation in sunny weather. generation in cloudy weather.

1500
Actual power generation
Before screening
Power generation / kWh

After screening
1000

500

0
62 218 250 256 281
Day number

Figure 10. The relationship between the actual power generation and the prediction of power
generation in rainy weather.

4.2. Evaluation of the results

Table 2. MAPE values of various weather neural network models.

Weather sunny cloudy rainy average
type value
Prediction Before 32.4% 45.6% 85.8% 46.8%
error screening
After 20.7% 22.6% 34.5% 23.6%
screening

Table 2 shows the difference between the predicted results of the six neural network models.
Compared with the MAPE mean value, it can be seen that the predicted value of the model after the
cluster analysis is closer to the actual value in each case. The average MAPE value of the model after
clustering analysis is 23.2% which is much smaller than the conventional models. Indicating that the

7
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

clustering analysis is effective, and the prediction model of photovoltaic generation is more accurate
base on the data after clustering analysis. Therefore, it is necessary to screen data.
In the case of sunny days, the prediction results of two models are closer, because the sunny
weather atmospheric state is relatively stable and the degree of data discrete is small, the MAPE value
of the prediction model that after screening is 11.7% lower than that before screening, the prediction
effect is better. In the case of cloudy days, the thickness and location of the clouds is difficult to
predict, the predicted results of the two models are relatively different, the MAPE value of the
prediction model that after screening is 23% lower than that before screening. In the case of rainy days,
the MAPE value of the prediction model that after screening is 51.3% lower than that before screening,
which difference is the biggest. It can be seen from the table 2, the prediction of rainy days is far less
accurate than sunny and cloudy days. The reason is the rain weather atmospheric changes are more
complex, in different rainy days, clouds, rain and vapor will be very different, and the absorption of
solar radiation will be very different, this will produce a large relative error, so it is difficult to achieve
the desired prediction effect in rainy days.
Based on overall assessment, the MAPE value of the predicted model is 23.6% which is less than
46.8% of the pre-screening prediction model. The comparison results show that photovoltaic power
generation models are more accurate when using the clustering analysis of the original data. Therefore,
it is necessary to carry out data cleanup.

5. Conclusion
In this paper, the prediction model is established by combining the clustering analysis and the BP
neural network. The abnormal samples in the historical monitoring data of the PV power plant are
screened out, and the prediction results are evaluated and analyzed. Research indicates:
 Using the cosine coefficient as the evaluation criteria, the abnormal samples are removed
using clustering analysis and screening of historical data, which can effectively filter the
discrete operating data.
 After the data is cleaned up as the training data, the BP neural network prediction models can
be established which can improve the accuracy of the prediction results.
 The data cleaning is necessary for establishing the prediction model using the historical data.
The cluster analysis can be effectively applied to the photovoltaic power generation
forecasting system.
The above research results can provide an effective method for the forecasting technology of PV
power plants, which is beneficial to the popularization and application of photovoltaic power
generation technology.

References
[1] Li R and Li G M 2008 Photovoltaic power generation output forecasting based on support
vector machine regression technique Electr Pow 41(2) 74-8
[2] Shi J, Lee W J, Liu Y Q, et al 2012 Forecasting power output of photovoltaic systems based on
weather classification and support vector machines IEEE Trans Ind Appl 48(3) 1064-9
[3] Lan H, Liao Z M and Zhao Y 2011 ARMA model of the solar power station based on output
prediction Electr Meas Instrum 48(54) 31-4
[4] Wang S X and Zhang N 2012 Short-term output power forecast of photovoltaic based on a grey
and neral network hybrid model Autom Electr Pow Syst 36(19) 1-5
[5] Christophe P, Cyril V, Marc M et al 2010 Forecasting of preprocessed daily solar radiation time
series using neural networks Sol Energy 84(12) 2146-60
[6] Zhang Y X and Zhao J 2011 Application of recurrent neural networks to generated power
forecasting for photovoltaic system Pow Syst Prot Control 39(15) 96-101
[7] Şenkal O and Kuleli T 2009 Estimation of solar radiation over Turkey using artificial neural
network and satellite data Appl Energ 86(7-8) 1222-8
[8] Wang F, Mi Z Q, Zhen Z et al 2013 A classified forecasting approach of power generation for

8
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012024 doi:10.1088/1755-1315/93/1/012024

photovoltaic plants based on weather condition pattern recognition Proc CSEE 33(34) 75-82
[9] Lu J, Zhai H Q and Liu C 2010 Study on statistical method for predicting photovoltaic
generation power East China Electr Pow 38(4) 563-7
[10] Lang Y and Zhang W T 2016 Research on the application of MATLAB in the data processing
of photovoltaic power generation Electr Technol Softw Eng 2016 87
[11] Zhang H N, Zhang J T, Yang L B et al 2016 Date processing method for PV power station data
acquisition system Electr Energ ManageTechnol 2016 9-13
[12] Chen C S, Duan S X, Cai T et al 2011 Online 24-h solar power forecasting based on weather
type classification using artificial neural network Sol Energy 85(11) 2856-70
[13] Dai Q, Duan S X, Cai T et al 2011 Short-term PV generation system forecasting model without
irradiation based on weather type clusting Proc CSEE 31(34) 28-35
[14] Wu S, Pan F M et al 2014 SPSS Statistical Analysis (Tsinghua University Press) pp 315-30
[15] Mohamed A and Badrul C 2015 Solar power forecasting using artificial neural networks Proc of
The 47th Annual North American Power Symp (USA: Charlotte, North Carolina)
[16] Chen C S, Duan S X and Yin J J 2009 Design of photovoltaic array power forecasting model
based on neutral network Trans China Electrotech Soc 24(9) 153-8

9
IOP Conference Series: Earth and Environmental Science

PAPER • OPEN ACCESS

Prediction of energy balance and utilization for solar electric cars

To cite this article: K Cheng et al 2017 IOP Conf. Ser.: Earth Environ. Sci. 93 012025

View the article online for updates and enhancements.

This content was downloaded from IP address 191.101.76.134 on 09/11/2017 at 12:11

NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

Prediction of energy balance and utilization for solar electric

cars

K Cheng1,2, L M Guo1, Y K Wang1 and M T Zafar1

1
School of Power and Energy, Northwestern Polytechnical University, Xi’an 710072,
Shaanxi province, China

E-mail: cksolar@163.com

Abstract. Solar irradiation and ambient temperature are characterized by region, season and
time-domain, which directly affects the performance of solar energy based car system. In this
paper, the model of solar electric cars used was based in Xi’an. Firstly, the meteorological data
are modelled to simulate the change of solar irradiation and ambient temperature, and then the
temperature change of solar cell is calculated using the thermal equilibrium relation. The above
work is based on the driving resistance and solar cell power generation model, which is
simulated under the varying radiation conditions in a day. The daily power generation and solar
electric car cruise mileage can be predicted by calculating solar cell efficiency and power. The
above theoretical approach and research results can be used in the future for solar electric car
program design and optimization for the future developments.

1. Introduction
Solar electric cars rely on the photovoltaic effect to produce electricity. In these types of cars, the
controller and battery are used to drive the motor by overcoming the driving resistance. The solar
energy characteristics are clean, pollution-free and inexhaustible. The application of solar energy in
the renewable based cars has broad prospects and has emerged to be a hot topic [1,2]. In 1984, the
solar electric car of Japan’s Tokai University won the competition named as World Solar Challenge by
using an average speed of 100 km/h [3]. Domestically, the research institutions of the universities
including Tsinghua University, Shanghai Jiao Tong University and Jilin University have also carried
out relevant research, and successfully produced a number of solar electric cars [4].
The core components of the solar electric car energy system are solar cells. The performance of the
battery is affected by solar radiation and temperature, and the power generation capacity changes with
time. Domestic scholars have done some research in proposing the corresponding solution to improve
the matching of the power system and the energy efficiency [5-7]. The prediction of energy balance
and utilization is still very complex because of many contributing factors such as change in solar
irradiation, temperature and environmental factors. These above-mentioned factors change in the form
of season, time domain, regional and domestic variations. Therefore, the theoretical based
performance prediction of the solar electric car energy system is carried out by using driving
resistance, solar cell power generation model, metrological data and thermal variations.

2. Running resistance and drive power

Solar electric cars rely on driving force to overcome the driving resistance to be able to drive. Driving
resistance is the most important energy load of solar cars. Driving resistance mainly includes rolling

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

resistance Rr (N), air resistance Ra (N), acceleration resistance Rac (N) and slope resistance Rc (N),
the formulae are given in preceding sections [8-10]:

2.1. Rolling resistance

The wheel in the car driving process continues to rotate and the resistance generated in the wheel is
rolling resistance which is expressed as:
Rr   r  W (1)

In the above formula, r - rolling resistance coefficient and W - total car weight (kg).

2.2. Air resistance

The car bears the resistance against the air while moving and it is named as the air resistance written
as:
1
Ra  Cd  Av 2 (2)
2
In the above formula, C d - air resistance coefficient;  - air density (kg/m3); A - windward area
(m2); v - travel speed (m/s).

2.3. Slope resistance

When the car moves uphill, the car's gravity component along the slope is expressed as the slope of the
car resistance, expressed as:
Rc  m  g  sin  (3)
In the above formula, m - car mass (kg), g - acceleration due to gravity (m/s2),  - slope angle (°).

2.4. Accelerate resistance

When the car accelerates, it is necessary to overcome the inertia force which is the same as to
accelerate the resistance. The mass of the car is divided into two parts: translation mass and
accelerated mass. Acceleration is not only the translation of the mass of inertial force, rotation mass
also has inertial force. The acceleration resistance of the car can be expressed as:
Rac    m  a (4)
In the above formula,  - car rotation mass conversion factor, m- car mass (kg), a -car acceleration
(m/s2).

2.5. Drive power

The air resistance is mainly determined by the design of the car. However, the acceleration resistance
is determined by the way of running and slope resistance is determined by the driving the car on the
road. Therefore, in the overall design, the car is considered runningon a flat road with the uniform
speed but the impact of acceleration and slope resistance is not considered.
Knowing the rolling resistance and air resistance, the calculation of the drive power PD (W) of the
solar electric car power system can be done as follows:
PD  ( Rr  Ra )  v (5)

3. Solar cell power generation model

2
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

The principle of solar cell power generation is based on the semiconductor photovoltaic effect of solar
radiation directly converted to electricity. The maximum power generation Pmpp of the solar cell is
obtained under standard test conditions where the atmospheric thickness is AM 1.5, the temperature is
25°C, and the solar irradiance is 1000W/m2. Since, the actual performance of the solar cell is affected
by solar radiation and temperature, it is necessary to use the maximum power temperature coefficient
to correct the power at the actual temperature and obtain the final actual power according to the
approximate relationship with the solar irradiance [11].

3.1. Temperature correction

When the battery temperature is T (°C), the maximum power PT (W) of the solar cell is:
PT  [1  (T  25)   ' ]  Pmpp (6)

In the above formula,  ' - the maximum power temperature coefficient is generally around - 0.5%
/°C.

3.2. Irradiation correction

In order to simplify the calculation, the solar cell power generation P is estimated to be proportional to
the incident solar irradiance,
P  Gon  PT 1000-1 (7)

In the above formula, G on - the solar irradiance received by the solar cell (kWh/m2), its value can
be obtained according to the hourly meteorological data.

4. Hourly weather data

The location of the operation is located in Xi'an at 34.5°north latitude, 108.5°east longitude and 693
m above sea level. Since so far, China has not yet established a unified standard and national hourly
weather data, especially solar radiation data. Therefore, the use of already known method of time
based meteorological data generation is more appropriate at this stage. The weather data for solar and
ambient temperatures is derived from NASA's historical data [12], as detailed in table 1:

Table 1. Solar energy and temperature.

month daily total solar sunny total solar Daily average daily range of
irradiation irradiation temperature temperature
kWh/m2/d kWh/m2/d °C °C
1 2.84 3.83 -5.49 7.44
2 3.34 4.86 -2.15 7.26
3 3.85 6.21 3.20 8.36
4 4.80 7.17 11.0 9.41
5 5.29 7.80 16.6 9.61
6 5.27 7.76 20.4 9.06
7 5.11 7.53 22.1 8.47
8 4.57 5.98 20.7 7.90
9 3.69 5.98 16.2 7.33
10 3.06 4.87 9.80 6.63
11 2.71 3.71 2.98 6.43
12 2.61 3.44 -3.18 5.80

The average daily sun irradiation is the daily average of the total radiation for the month, and the

3
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

sunny day total solar irradiation is the amount of radiation in the sunny day, and the following formula
is used [13]:
rt   /24  (a  b cos )(cos   cos s )(sin s    cos s )1 (8)

In the above formula, rt - the ratio of total radiation in hour to total radiation, s - sunset angle,  –
hour angle; the coefficient a, b are calculated as follows:
a  0.409  0.5016  sin(s   / 3) (9)

b  0.6609  0.4767 sin(s   / 3) (10)

The solar irradiance at different time can be obtained by calculating rt of different time and the
daily total radiation.
In June and December, for example, the daily total solar irradiation and sunny total solar irradiation
distribution with time is shown in figure 1:

1000 30
Jun.daily Jun.
900 25 Dec.
Jun.sunny
800 Dec.daily 20
700
Irradiance /w/m2

Temperature /℃

Dec.sunny
600 15
500 10
400
5
300
200 0
100 -5
0 -10
0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Time /h Time / h
Figure 1. Hourly solar irradiance. Figure 2. Hourly ambient temperature.

The data show that [14,15], the ambient temperature in the day can be approximated with a cosine
function. Moreover, the maximum temperature generally occurs at 14:00, and the lowest temperature
appears in the 1 hour before the sunrise. According to the daily average temperature, temperature
differences and the change in sunrise time in the months of June and December are shown in figure 2.

5. Prediction of energy balance and utilization

5.1. Solar electric car conceptual model

The solar electric car model is shown in figure 3, which has length of 4 m, width of 1.7 m, height of 1
m, and weight of 350 kg. Solar electric cars rely on the photovoltaic effect to produce electricity. In
these types of cars, the controller and battery are used to drive the motor by overcoming the driving
resistance. For maximum power output, the system uses MPPT solar controller. When the intensity of
light is large, the electric energy can drive the motor directly through the control system and drive the
wheel. When the vehicle is at low speed or stagnant state, excess electricity can be stored in the battery.
Otherwise, the battery through the control system discharge drives motor and then drives the car. The
speed of the car is dependent on the control system which adjusts the current of the motor.

4
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

solar cell

battery controller motor

Figure 3. Solar electric car model.

The surface of solar electric car is equipped with 432 pieces of HH125M165-S2 monocrystalline
silicon solar cells. In addition, 144 pieces of these cells constitute a PV module and a total of three PV
modules are used. Under standard conditions, the power comes out to be 1223 Wp. Under the standard
condition, the efficiency  of HD125M165-S2 monocrystalline silicon solar cell is 18.2%, peak
power Pmpp is 2.833Wp, and the maximum power temperature coefficient  is -0.46%/°C.
The four NCR18650A lithium ion battery stacks are used in series to form a power storage unit. In
addition, sub string is made by joining 10 groups of these power storage units and 45 groups of sub-
string are then connected in parallel. The total capacity of the battery pack is 2.09 kWh, the maximum
discharge current is 1395 A and the total weight is 81.9 kg.
The information relevant to the solar electric model car include temporary cruise speed of 30 km/h
(8.33 m/s), rolling resistance coefficient of 0.015, the wind area of 0.5 m2, air resistance coefficient
0.35, and air density is 1.205 kg/m3.
According to the formulas (1)-(6), the rolling resistance is 51.5 N and the air resistance is 7.3 N.
Therefore, the total resistance is 58.8N and the required power is 489.8 W. The different efficiencies
come out to be 95%, 90%, 95% and 75% for motor, mechanical transmission, MPPT solar controller
and battery respectively. In addition, the required solar power supply is 804 W.

5.2. Solar cell temperature

Solar cell temperature is generally higher than the ambient temperature. However, sunshine makes the
temperature rise but the wind blows the temperature down. The solar cell temperature T is simplified
according to the thermal equilibrium relation [16].
  Gon  (1   )      (T 4  Tsky
4
)  h(T  Ta ) (11)

The formula on the left side is the heat energy absorbed by the solar cell and the right side shows
the release of heat which includes the radiant heat to the sky and the convective heat transferring with
the environment. In the above formula,  - solar cell heat radiation emissivity is set to 0.95; G on - solar
irradiance obtained by solar cells. According to meteorological data obtained by time;  - solar cell
photoelectric conversion efficiency and  - Stephen-Boltzmann Constant, 5.76×10-8W/m2∙K4. Tsky -
Sky temperature, calculated as follows:
Tsky  0.0552Ta1.5 (12)

Ta is the ambient temperature according to the meteorological data obtained by time. While, h - air
convection heat transfer coefficient is calculated by the following formula:
h  2.8  3.0vw (13)

Where vw is the surface wind speed which is approximately equal to the cruising speed of 8.33 m/s.
In June and December, for example, by using the formula (11) the changes in the solar cell

5
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

temperature T with the time are shown in the results in figure 4.

50
Jun.daily
45
Jun.sunny
40
Dec.daily
35
Dec.sunny
Temperature /℃ 30
25
20
15
10
5
0
-5
-10
0 2 4 6 8 10 12 14 16 18 20 22 24
Time /h
Figure 4. Hourly solar cell temperature.

5.3. Prediction results

The change in the solar cell temperature with the time will result in the change of efficiency which is
shown in figure 5. Further, with the change of solar irradiance with time, the peak power changes are
shown in figure 6.

22 Jun.daily
3.0 Jun.daily
Jun.sunny
21 Jun.sunny
Dec.daily 2.5 Dec.daily
Dec.sunny
Peak Power /Wp

20 Dec.sunny
Efficiency /%

2.0

19
1.5

18 1.0

17 0.5

16 0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Time /h Time /h

Figure 5. Hourly solar cell efficiency. Figure 6. Hourly peak power of solar cell.

The time of 12 o’clock can be taken as symmetry for the solar power produced all around the day.
By the impact of ambient temperature hysteresis, the power of the morning session is slightly higher
than the afternoon. The average efficiency in June was 18.0% and 17.7% with average power of 0.960
W and 1.378 W respectively. On the other hand, the average efficiency in December was 20.1% and
19.9%, with the average power of 0.893 W and 1.164 W respectively.
The solar cell efficiency and power change with time from January to December. The total
theoretical solar power generation from January to December is shown in figure 7. Further, according
to the needs of the solar electric car photovoltaic power supply, from January to December, the solar
electric car daily cruise stroke is shown in figure 8.
In the whole year, the maximum daily generation and cruise travel took place in May and the
minimum occurred in December. In June, for example, the performance prediction results show that
the average daily generating capacity of 432 solar modules is 6.33 KWh and 9.10 KWh under normal
daily radiation and sunny day radiation respectively. According to the cruise speed of 30 km/h and the
required solar power supply capacity of 804.0 W, the maximum theoretical travel is 232.0 km and

6
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

333.15 km with cruising time of 5.5 hours and 7.9 hours respectively. It should be noted that the actual
cruise travel and time will be less than the above forecasted results due to acceleration and slope
resistance.

10 250
Daily Daily
9 Sunny Sunny

Driving Distance /km

8 200
Power /kW·h

7
150
6

5
100
4

3 50
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Month Month

Figure 7. Theory daily power generation. Figure 8. Daily driving distance.

6. Conclusion
Based on the concept model of solar electric car, this paper studies the performance prediction of
energy system, and can get the following conclusions:
 The model and method established in this paper can be used to predict the performance of
solar energy car energy system in different months. It can be used to verify and analyze the
solar electric car in the developing stage. It can provide base for the selection and
improvement of the scheme.
 The performance of the solar electric car energy system changes in a day with the time of 12
o’clock as the center. The 12 o’clock can be considered as symmetrical point. However, the
lag of environmental temperature results into the morning value slightly greater than afternoon.
 Affected by changes in solar energy and ambient temperature in different months, the working
capacity of solar energy vehicle energy system has nearly doubled the scope of the design.
And the applications of the solar electric car system will have greater impact in the future.

References
[1] Kelly 2002 Raycing with the sun: Solar cars thrive in summer heat JOM Journal of the
Minerals, Metals and Materials Society 54 14-7
[2] Peter P and Phil H 2002 Critical speed control of a solar car Optimization and Engineering 3
97-107
[3] Surhone L M, Tennoe M T, Henssonow S F, Racing S C and Highway S 2010 World Solar
Challenge (Saarbrücken: Betascript Publishing)
[4] Zhao Y F, Wan J, Zhu Z P, Chen J, Le Z G, Ma F W, Ding Y and Zhao F Q 2011 The
application of solar cells on vehicle Agricultural Equipment and Vehicle Engineering 237
39-42
[5] Shi W K, Lv D H and Liang T Y 2008 A foundation investigation of parameters matching for
power train smission of solar electric car Acta Energlae Solaris Sinica 29 959-64
[6] Wang X C, Zhang P S, Zhang L H and Jia W 2009 Research on power control system of solar
electric vehicle Machinery and Electronics 12 36-9
[7] Wang X Y and Kang L Y 2010 The research and development of maximum power point tracker
for solar vehicle Power Electronics 44 12-3
[8] Guerra-zubiaga D A, Ramon-Raygoza E D, Rios-Soltero E F, Tomovic M and Molina A A

7
NEFES 2017 IOP Publishing
IOP Conf. Series: Earth and Environmental Science1234567890
93 (2017) 012025 doi:10.1088/1755-1315/93/1/012025

2009 PLM tools taxonomy to support product realization process: A solar racing car case
study Product Realization 1 1-26
[9] Japan Solar Energy Association 2010 Mechanism and Production of Eco-electric Vehicle (Xi'an:
Xi'an Jiaotong University Press)
[10] Hiroyuki O, Sumio N and Dai H 1998 Development of aerodynamics for a solar race car JSAE
Review 19 343-9
[11] Luo Y J, He Z N and Wang C G 2005 Solar Energy Technology (Beijing: Chemical Industry
Press)
[12] NASA: Surface meteorology and Solar Energy [EB/OL] http://eosweb.larc.nasa.gov/cgi-
bin/sse/sse.cgi?na
[13] Zhang H F 2004 Solar Energy Heat Utilization Principle and Computer Simulatio (Xi'an:
North-western Polytechnical University Press)
[14] Akasaka H et al 2000 Development of expanded AMeDAS weather data for building energy
calculation in Japan ASHRAE Transactions 206 455-6
[15] Tsinghua University, Tongji University, Xi'an Institute of metallurgy and architecture and
Chongqing Institute of civil engineering and architecture 1986 Air Conditioning (Beijing:
China Architecture and Building Press)
[16] Yang S M and Tao W Q 2006 Heat Transfer (Beijing: Higher Education Press)