Research Article
Research Article
Research Article
Estimating Solar Insolation and Power Generation of Photovoltaic
Systems Using Previous Day Weather Data
Received 22 October 2019; Revised 2 January 2020; Accepted 16 January 2020; Published 18 February 2020
Copyright © 2020 Min Hee Chung. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Day-ahead predictions of solar insolation are useful for forecasting the energy production of photovoltaic (PV) systems attached
to buildings, and accurate forecasts are essential for operational efficiency and trading markets. In this study, a multilayer feed-
forward neural network-based model that predicts the next day’s solar insolation by taking into consideration the weather
conditions of the present day was proposed. The proposed insolation model was employed to estimate the energy production of a
real PV system located in South Korea. Validation research was performed by comparing the model’s estimated energy production
with the measured energy production data collected during the PV system operation. The accuracy indices for the optimal model,
which included the root mean squared error, mean bias error, and mean absolute error, were 1.43 kWh/m2/day, − 0.09 kWh/m2/
day, and 1.15 kWh/m2/day, respectively. These values indicate that the proposed model is capable of producing reasonable
insolation predictions; however, additional work is needed to achieve accurate estimates for energy trading.
1. Introduction PV systems in the city rather than in the rural or forested areas
because of land conservation concerns. Indeed, PV systems
Electricity consumption has been rapidly increasing around are more likely than other types of renewable energy systems,
the world despite efforts to improve energy conservation. In to be constructed in urban built environments.
2013, in Korea, the electricity consumption in the commercial The Korean government intends to introduce a smart
and public sectors increased by 10% to 65.8% compared to grid to promote the embracement of renewable energy [3].
values in 2001. Moreover, the electrification of energy con- As part of these efforts, the government announced a second
sumption has further intensified [1]. Since natural gas and basic plan for an intelligent power grid in 2018. Accordingly,
electricity have emerged as primary energy sources in the the government has set up policies on smart grid complex
household sector, this share of electricity consumption has construction, infrastructure, and facility expansion. It was
been steadily increasing. This is because households prefer reported that, in 2019, a distributed energy resource market
clean energy sources to maintain comfort and the convenience would be allowed in Korea. As a result, significant changes
of living. The Korean government is promoting buildings in are expected in the electric trading market, which has been
which renewable energy systems are used, to save on primary focused on centralized power generation companies. The
energy consumption and to realize energy independence. expansion of distributed generation will accelerate direct
Solar energy is known to be a good substitute for fossil energy trading among individuals and groups that can
fuels, which currently account for more than 80% of the produce and consume energy simultaneously [3, 4]. Such
primary energy supply [2]. Photovoltaic (PV) systems are the peer-to-peer (P2P) energy trading has been demonstrated in
most suitable replacements for fossil fuels because they do not various forms in the United States, United Kingdom,
produce CO2 emissions and do not pose the same risks as Netherlands, and Germany [5–7]. P2P energy trading is
those associated with other alternative energy supplies such as beneficial because it reduces the need for expensive and
nuclear power generation. It may be more desirable to install inefficient energy transportation with substantial losses [8],
2 Advances in Civil Engineering
thus improving the reliability of energy supply [9] and al- solar irradiance prediction method for hourly day-ahead
leviating transmission and distribution congestion problems forecasts by using hourly weather forecast data. The pro-
[10]. Forecasting solar insolation in advance is essential in posed model contained structured long- and short-term
the management of the PV system’s generated output in a memory networks, and it was compared with other algo-
distributed grid. It is necessary to provide information on rithms. By using weather forecasting data as input variables,
the energy production through the system to the prosumer, it is possible to reflect the change in solar radiation in ac-
who conducts the small-scale electricity trading as a weather cordance with the weather changes. However, the accuracy
forecast in advance. Time-based predictions may need to be of the solar radiation prediction changes based on the
provided according to the individual’s characteristics. prediction accuracy of the weather forecast data. Amrouche
However, it is necessary to provide daily production forecast and Le Pivert [19] proposed a solar radiation forecasting
information to a large number of prosumers. In small-scale model using artificial neural networks (ANNs) and special
energy trading, the provision of daily energy production modelling. In cases where there were no meteorological data
forecast information is more effective than providing de- from the predicted area, the solar radiation was estimated
tailed hourly energy generation forecast information. Daily using meteorological data from the nearby areas. This model
energy production by PV systems can be predicted simply predicted daily values, which were used to estimate power
based on daily irradiation. generation by PV systems. Long et al. [20] proposed a
Many researchers are presently directing their efforts on prediction model for the daily PV energy production by
solar insolation forecasting techniques. Soares et al. [11] meteorological parameters. The efficiency of the prediction
examined the intrahour solar insolation predictions based algorithm was improved by classifying the weather data
on hourly weather data for a period of 4 years. They esti- based on importance and with reduced input variables used
mated the hourly values of the diffuse solar radiation by as input data.
using multilayer perceptron neural networks. However, The prediction accuracy varied in accordance with the
their model could not reflect the present-day weather accuracy of the weather forecasts. In addition, there have
change. Mathiesen and Kleissl [12] compared five fore- been many studies on various solar insolation prediction
casting models for the intraday solar insolation by using models with different prediction horizons and different
ground-based weather measurement data, and these efforts prediction methods. The previous solar forecasting models
resulted in the production of an accurate database to validate made predictions based on the time or day. However, a large
numerical weather predictions. The accuracy of each model amount of input data was needed for accuracy within the
in predicting insolation varied in accordance with the predicted time scale. The limitation of the prediction model
weather conditions. However, it was difficult to consider is that the training and prediction time increases as the
multiple models simultaneously. Sun et al. [13] proposed a number of input variables increases. This study aims to
random forest algorithm for estimating the daily solar ra- develop a simple solar insolation prediction model by using
diation based on meteorological data, solar radiation, and fewer weather data variables, which make it easy to acquire
three air pollution indexes for SO2, NO2, and PM10. They information. Typically, weather forecast data may be used to
optimized the proposed estimation model depending on the predict the next day’s solar insolation. However, predicting
input variables. In a city with a high degree of air pollution, the next day’s solar insolation using weather forecast data is
this model can be used to estimate the exact amount of solar characterized by a lot of uncertainty due to the uncertainty
radiation; however, the number of input variables increases. in the weather forecast data.
Sharma and Kakkar [14] also proposed an hourly global solar The objective of this paper is to propose a simple pre-
irradiance model for different forecasting horizons ranging diction model for day-ahead solar insolation using weather
from a few hours ahead to 48 hours ahead by using machine observation data. The predicted solar insolation would be
learning. The model with 1 hour ahead predictions was the used for estimating energy production in advance. The es-
most accurate; however, there was not enough time to timated energy production would give information for the
provide information to users. Gutierrez-Corea et al. [15] determination of optimal operating conditions, such as self-
modeled the spatial-temporal short-term global solar in- consumption or feed-in into the grid mode for PV systems
solation by using artificial neural networks. This model gave located in a distributed grid. The day-ahead solar insolation
accurate data predictions over time but increased the prediction model forecasts the solar radiation of the next day
complexity of the prediction model with more than 900 based on the weather data of the current day. It is assumed
input parameters. Huang and Davy [16] proposed a linear that the predicted model cannot be applied to systems in-
regression model for the intrahour solar irradiance that uses stalled in a specific condition; however, it can predict in-
the hourly clear sky index and geopotential thickness. The formation applicable to a wide range of conditions.
model predicted hourly solar irradiance only for summer. Therefore, the weather difference due to the differences in
Therefore, there were limitations in applying this model to the meteorological observation site and the installation lo-
other seasons. Vakili et al. [17] designed a prediction model cation of the PV system was ignored. The input variables use
for the total daily solar insolation for Iran by using a meteorological data that can be easily obtained from the
multilayer perceptron artificial neural network. The limi- meteorological administration. The prediction model could
tation of this model was that the solar insolation was esti- provide building users with useful information to plan their
mated based on the weather observation information, rather energy use in advance. This paper is organized as follows. In
than the future value. Qing and Niu [18] presented a novel Section 2, the data used for training and checking the
Advances in Civil Engineering 3
prediction model are described. Section 3 explains the Table 1: Station specifications.
prediction model and the verification method. The measured Item Specification
energy production for a PV system located in South Korea
Location Seoul, South Korea
was compared to the energy production prediction obtained Latitude 37.5714°N
through the solar insolation prediction model. Section 4 Longitude 126.9658°E
presents the results, and Section 5 presents the conclusions. Elevation 85.67 m
Climate type Humid continental
2. Meteorological Data Observed period 2014–2017
Table 5: Accuracies of predicted value ranges according to the input variables, NHN, and NHL.
RMSE MBE MAE
Input variables Number of hidden layers∗ NSE
(kWh/m2/day) (kWh/m2/day) (kWh/m2/day)
1 1.764–2.000 − 0.891 to − 0.787 1.368–1.529 − 0.503 to − 0.169
2 1.763–2.023 − 0.811 to − 0.649 1.355–1.552 − 0.920 to − 0.210
Thigh, Tlow, RH, DWS, PR,
3 1.760–2.015 − 0.840 to − 0.568 1.373–1.506 − 1.092 to − 0.164
PD, SD, CSD, CC
4 1.799–2.015 − 0.903 to − 0.711 1.382–1.559 − 0.937 to − 0.216
5 1.793–2.102 − 0.853 to − 0.589 1.366–1.631 − 1.068 to − 0.207
1 1.975–2.227 − 1.301 to − 0.969 1.515–1.692 − 0.292 to − 0.016
2 1.975–2.138 − 1.076 to − 0.714 1.550–1.635 − 0.162 to − 0.016
Thigh, Tlow, RH, DWS,
3 1.792–2.318 − 1.221 to − 0.567 1.432–1.773 − 0.391 to 0.164
SD, CSD, CC
4 1.889–2.187 − 1.005 to − 0.547 1.436–1.632 − 0.245 to 0.071
5 1.807–2.338 − 1.227 to − 0.419 1.444–1.773 − 0.423 to 0.150
1 1.425–1.506 − 0.272 to − 0.129 1.157–1.204 0.411–0.473
2 1.423–1.597 − 0.400 to − 0.090 1.131–1.297 0.337–0.474
Thigh, Tlow, RH, DWS, SD, CSD 3 1.421–1.660 − 0.376 to − 0.114 1.131–1.342 0.284–0.482
4 1.444–1.692 − 0.464 to − 0.121 1.148–1.372 0.256–0.458
5 1.450–1.720 − 0.291 to − 0.118 1.153–1.405 − 0.338 to 0.454
∗
The NHN for each layer was conducted by changing both to 10–20.a
found to be largely as a result of rainfall, humidity variations, production by using equation (11). To check the validity of
and changes in suspended matter concentrations. In 24 out the approach, the energy production was calculated from the
of the 29 days, the previous day’s rain came to a halt or it measured insolation and compared with the actual energy
rained on the estimation day. produced by the PV system. Table 10 presents the accuracy
The RMSE and MAE patterns exhibited similar ten- results for the energy production estimations. The predicted
dencies to the patterns of the monthly solar insolation power generation calculated from the measured insolation
according to the Sun’s trajectory. The RMSE and MAE was similar to the actual power generation, as shown in
values were smallest in December during the winter solstice. Figure 5. There was a difference between the power gen-
The RMSE and MAE values then increased with the increase eration calculated by the measured insolation and the actual
in solar radiation until June. The solar insolation, RMSE, and production values. Detailed meteorological observations
MAE values decreased after June. The predicted values for cannot be performed at all PV installation sites. As a result, it
the period between October and December, where the MBE is determined that an error occurred due to the positional
values showed positive values, were overestimated. On the difference. It is also possible that a difference occurred due to
contrary, the predicted values for January to March were the limitation of the numerical calculations. However, the
underestimated compared to the measured values. The RMSE, MAE, and MBE of the energy generation calculated
predicted values between October and December were from the measured insolation and the actual production
overestimated because of the learning effect of the summer were 0.85 kWh/day, 0.38 kWh/day, and 0.66 kWh/day, re-
data. Conversely, the predicted values between January and spectively, indicating that the calculation procedure could be
March did not reflect the increase in solar insolation because used for estimating the energy production of a PV system.
of the influence of the winter data. The energy production was calculated using the estimated
solar insolation using the proposed procedure. The power
generation obtained by the predicted insolation fluctuated
4.2. Application of the Optimal Prediction Model to less than the actual energy production. Some errors were due
PV Output Estimations. In order to compare the PV to the insufficient predictions of the daily weather condition
production forecast, the energy production of the 1.75 kW changes. Although the forecast model reflected the daily
PV system was monitored for 364 days from August 2018 to variation, it did not predict the actual magnitude of the
July 2019. The weather conditions during the monitoring change. In the spring and summer observation, there was a
period are listed in Table 9. Conditions were generally clear; wide variation in the day to day solar radiation. This was
however, there were some rainy days due to the effects of because, unlike previous studies in which the prediction was
typhoons. There was a precipitation period of 97 days in total by hours, the forecast horizon was long and did not reflect all
during the entire monitoring period. During this period, the of the variability during that forecast periods. The prediction
average daily cloud cover fluctuated greatly. As a result, the model proposed in this study predicts the next day’s solar
variation in solar insolation was larger than that for the other insolation by using the weather conditions of the previous
periods. The power generation range for the PV system was day. However, the weather conditions of the monitoring
0.1–10.8 kWh/day depending on the weather conditions. period changed extensively each day, which made it difficult
The predicted solar insolation was estimated by using the to predict the insolation. The energy production error was
optimal solar insolation prediction model. The predicted especially significant on days when the solar radiation
solar insolation was then used to calculate the PV suddenly changed due to rain or snow. In order to provide
8 Advances in Civil Engineering
9 9
8 8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
0 0
1-Jan
1-Feb
1-Mar
1-Apr
1-Jun
1-Jul
1-Aug
1-Sep
1-Oct
1-Dec
1-Mar
8-Mar
15-Mar
22-Mar
29-Mar
5-Apr
12-Apr
19-Apr
26-Apr
1-Nov
1-May
3-May
10-May
17-May
24-May
31-May
Date Date
Observed value
Observed value
Predicted value
Predicted value
(a) (b)
9 7
Solar insolation (kWh/m2/day)
Solar insolation (kWh/m2/day)
8 6
7
5
6
5 4
4 3
3
2
2
1 1
0 0
1-Sep
8-Sep
15-Sep
22-Sep
29-Sep
6-Oct
13-Oct
20-Oct
27-Oct
13-Jul
20-Jul
27-Jul
3-Nov
10-Nov
17-Nov
24-Nov
1-Jun
8-Jun
15-Jun
22-Jun
29-Jun
6-Jul
3-Aug
10-Aug
17-Aug
24-Aug
31-Aug
Date Date
Observed value Observed value
Predicted value Predicted value
(c) (d)
5
Solar insolation (kWh/m2/day)
4.5
4
3.5
3
2.5
2
1.5
1
0.5
19-Dec
25-Dec
31-Dec
1-Jan
7-Jan
13-Jan
19-Jan
25-Jan
31-Jan
1-Dec
7-Dec
6-Feb
12-Feb
18-Feb
24-Feb
13-Dec
Date
Observed value
Predicted value
(e)
Figure 3: Comparison between the predicted and observed values of daily solar insolation: (a) over one year; (b) spring; (c) summer; (d) fall;
(e) winter.
Advances in Civil Engineering 9
Table 7: Accuracies of the optimal predicted values as compared to the measured values.
Month Number of valid days RMSE (kWh/m2/day) MBE (kWh/m2/day) MAE (kWh/m2/day)
January 31 0.717 − 0.376 0.580
February 28 1.133 − 0.246 1.003
March 31 1.112 − 0.356 0.917
April 28 1.742 − 0.595 1.581
May 31 1.753 − 0.797 1.536
June 30 2.034 − 0.901 1.744
July 31 1.760 0.029 1.485
August 31 1.753 − 0.439 1.491
September 29 1.650 − 0.432 1.455
October 30 0.925 0.559 0.643
November 26 0.668 0.609 0.481
December 31 0.767 0.521 0.595
8
7 Weather elements Value
6 2 August 2018 to 31 July 2019 (364
Monitoring period
5 days)
4 The day’s highest
3 − 6.6–39.6°C
temperature
2 Minimum humidity 10–91%
1
Daily wind speed 0.8–3.8 m/s
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Number of rainy days 97 days
Month Sunshine duration 9.6–14.8 hours
Average
Figure 4: Range of monthly solar insolation by measured data from according to weather conditions and operating schedules. In
the meteorological administration. general, the summer and winter energy consumption in
buildings is related to the external weather conditions. On
accurate forecasts of the energy production according to days where energy consumption is expected to be high in
weather condition changes as basic information for energy buildings, the energy produced by PV systems will be
trading, it will be necessary to compensate for dramatic consumed preferentially in buildings, even if less or more
fluctuations in weather conditions. The error magnitude was energy is produced through PV systems. On the contrary, if
smaller in the fall and winter periods with less precipitation, the prosumer is planning to sell the produced energy to the
but these seasons were predicted to be less than the actual grid because the projected energy consumption is low in the
energy production. In regions with distinct characteristics of building, the energy trade profits may be less than expected if
seasons, it is necessary to present the solar insolation pre- the system actually produces less power than expected. In
diction model by season. order to propose the optimal operations for P2P energy
P2P energy trading should also take into account the trading, it is necessary to predict the energy consumption of
energy price and the energy consumption of buildings the building, the price of energy, and the amount of energy
10 Advances in Civil Engineering
12 12
10 10
8 8
6 6
4 4
2 2
0 0
1-Aug-18
1-Sep-18
1-Oct-18
1-Nov-18
1-Dec-18
1-Jan-19
1-Feb-19
1-Mar-19
1-Apr-19
1-May-19
1-Jul-19
1-Jun-19
1-Mar-19
11-Mar-19
21-Mar-19
31-Mar-19
10-Apr-19
20-Apr-19
30-Apr-19
10-May-19
20-May-19
30-May-19
Month Month
Monitored power generation by the PV system Monitored power generation by the PV system
Estimated power generation by predicted irradiance Estimated power generation by predicted irradiance
Estimated power generation by measured irradiance Estimated power generation by measured irradiance
(a) (b)
The produced energy (kWh/day)
12 12
10 10
8 8
6 6
4 4
2 2
0 0
1-Sep-18
11-Sep-18
21-Sep-18
1-Oct-18
11-Oct-18
21-Oct-18
31-Oct-18
10-Nov-18
20-Nov-18
30-Nov-18
1-Aug-18
11-Aug-18
21-Aug-18
31-Aug-18
9-Jun-19
19-Jun-19
29-Jun-19
9-Jul-19
19-Jul-19
29-Jul-19
Month
Month
Monitored power generation by the PV system
Monitored power generation by the PV system
Estimated power generation by predicted irradiance
Estimated power generation by predicted irradiance
Estimated power generation by measured irradiance
Estimated power generation by measured irradiance
(c) (d)
Figure 5: Continued.
Advances in Civil Engineering 11
1-Dec-18
11-Dec-18
21-Dec-18
31-Dec-18
10-Jan-19
20-Jan-19
30-Jan-19
9-Feb-19
19-Feb-19
Month
Figure 5: Comparison of the predicted and measured energy production insolations: (a) over one year; (b) spring; (c) summer; (d) fall; (e)
winter.
to be produced by PV systems together to suggest the op- to the actual PV system’s output. The accuracy in-
timal use method. dexes for the PV system’s energy output using the
optimal model, which included the root mean
5. Conclusions squared error, mean bias error, and mean absolute
error, were 2.70 kWh/m2/day, − 0.36 kWh/m2/day,
The recent energy-saving building policies enacted in South and 2.28 kWh/m2/day, respectively.
Korea are expected to lead to the realization of distributed
The error of the predicted values for solar insolation
energy generation in a smart grid. P2P trading of the energy
varied from season to season. For instance, in the summer
produced in buildings will soon take place. Predicting energy
when the difference between the maximum and minimum
production is essential for P2P energy trading planning, and
insolation was large, results did not fully reflect insolation
the energy produced through PV systems will be mainly
changes in response to weather changes. The predicted values
traded. In this study, the potential applicability of the energy
tended to represent the median value. Although this provides
production prediction of PV systems was investigated by
stable values, data were less predictable in the summer when
using the predicted solar insolation derived from a simple
energy production was high. On the contrary, in winter, the
model to provide information to prosumers engaging in
accuracy of the forecasted insolation was high for high solar
small-scale electricity trading. The predictive model pro-
insolation days. However, the predictions were higher than
posed by the existing researchers attempted to accurately
the actual solar insolation for the low solar insolation days.
predict the hourly or daily radiation by using a large amount
The proposed model did not reflect sudden weather changes
of information. However, in this study, the amount of solar
on any day because it used the daily weather conditions of the
radiation the next day was predicted by using a small
day before the forecast. To further improve the prediction
amount of weather information. The results are summarized
model, it is necessary to execute the prediction model in the
as follows:
morning or on an hourly basis rather than the day before the
(1) The prediction model for solar insolation was built forecast and to further optimize the prediction model in
by using an MLF, and the proposed model uses the accordance with the variations in seasonal characteristics.
daily weather conditions. The predicted solar inso- The prediction of the PV system energy production
lation was based on the estimated weather conditions through solar insolation predictions showed a pattern similar
on the day before the prediction. To find the optimal to the solar insolation prediction results. The energy pro-
architecture, the number of input variables, the duction estimation by the predicted insolation tended to yield
number of hidden neurons, and the number of underestimates compared to the real produced energy. The
hidden layers were changed. monitoring period was characterized by severe weather
fluctuations. To accommodate such weather fluctuations in
(2) The input variables of the optimal model included
the future, it may be necessary to compensate for climate
the highest and lowest temperature, minimum hu-
change. The accuracy of the prediction model presented in
midity, wind speed, sunshine duration, and con-
this study is much lower on sunny days or cloudy days with
tinued sunshine duration. The optimal model
precipitation. There was a limit to predicting the exact amount
consisted of 3 hidden layers and 11 hidden neurons.
of solar radiation, including various weather variables, in the
(3) The energy output of the PV system was calculated relationship between the weather conditions of the previous
from the solar insolation estimated by the optimal day and the weather conditions of the next day. There was no
model. The calculated energy output was compared uncertainty about this change in weather. In this research, the
12 Advances in Civil Engineering
solar insolation of the horizontal plane was predicted, while [9] K. Alanne and A. Saari, “Distributed energy generation and
that of the inclined insolation was calculated by the equations sustainable development,” Renewable and Sustainable Energy
and added to the solar power generation equation. In this Reviews, vol. 10, no. 6, pp. 539–558, 2006.
process, there was a difference between the actual value and [10] J. Thomsen, N. S. Hussein, C. Senkpiel, N. Hartmann, and
the predicted value obtained using the equation. In the future, T. Schlegl, “An optimized energy system planning and op-
it may be necessary to distinguish the prediction model results eration on distribution grid level—the decentralized market
agent as a novel approach,” Sustainable Energy, Grids and
according to weather conditions or to increase the accuracy by
Networks, vol. 12, pp. 40–56, 2017.
shortening the prediction horizon.
[11] J. Soares, A. P. Oliveira, M. Z. Božnar, P. Mlakar,
If P2P energy trading becomes active in the future, it will J. F. Escobedo, and A. J. Machado, “Modeling hourly diffuse
be necessary to predict the energy production based on the solar-radiation in the city of São Paulo using a neural-network
weather conditions of the forecast day. Therefore, an ac- technique,” Applied Energy, vol. 79, no. 2, pp. 201–214, 2004.
curate insolation prediction model will be needed to predict [12] P. Mathiesen and J. Kleissl, “Evaluation of numerical weather
the energy production of the PV systems involved in P2P prediction for intra-day solar forecasting in the continental
energy trading. The model described here represents a good United States,” Solar Energy, vol. 85, no. 5, pp. 967–977, 2011.
potential model for such purposes, and future optimization [13] H. Sun, D. Gui, B. Yan et al., “Assessing the potential of
work will be forthcoming. random forest method for estimating solar radiation using air
pollution index,” Energy Conversion and Management,
Data Availability vol. 119, pp. 121–129, 2016.
[14] A. Sharma and A. Kakkar, “Forecasting daily global solar
The data used to support the findings of this study are irradiance generation using machine learning,” Renewable
available from the corresponding author upon request. and Sustainable Energy Reviews, vol. 82, no. 3, pp. 2254–2269,
2018.
[15] F.-V. Gutierrez-Corea, M.-A. Manso-Callejo, M.-P. Moreno-
Conflicts of Interest Regidor, and M.-T. Manrique-Sancho, “Forecasting short-
The author declares that there are no conflicts of interest. term solar irradiance based on artificial neural networks and
data from neighboring meteorological stations,” Solar Energy,
vol. 134, pp. 119–131, 2016.
Acknowledgments [16] J. Huang and R. J. Davy, “Predicting intra-hour variability of
This research was supported by the Basic Science Research solar irradiance using hourly local weather forecasts,” Solar
Energy, vol. 139, pp. 633–639, 2016.
Program through the National Research Foundation of
[17] M. Vakili, S. R. Sabbagh-Yazdi, S. Khosrojerdi, and K. Kalhor,
Korea (NRF), funded by the Ministry of Science, ICT & “Evaluating the effect of particulate matter pollution on es-
Future Planning (NRF-2017R1C1B20117911). timation of daily global solar radiation using artificial neural
network modeling based on meteorological data,” Journal of
References Cleaner Production, vol. 141, pp. 1275–1285, 2017.
[18] X. Qing and Y. Niu, “Hourly day-ahead solar irradiance
[1] Korea Energy Economic Institute, Energy Consumption prediction using weather forecasts by LSTM,” Energy, vol. 148,
Survey, Ministry of Trade, Industry and Energy, Sejong City, pp. 461–468, 2018.
South Korea, 2015. [19] B. Amrouche and X. Le Pivert, “Artificial neural network
[2] International Energy Agency, Key World Energy Statistics,
based daily local forecasting for global solar radiation,” Ap-
International Energy Agency, Paris, France, 2017.
plied Energy, vol. 130, pp. 333–341, 2014.
[3] I. S. Bayram, M. Z. Shakir, M. Abdallah, and K. Qaraqe, “A
[20] H. Long, Z. Zhang, and Y. Su, “Analysis of daily solar power
survey on energy trading in smart grid,” in Proceedings of the
prediction with data-driven approaches,” Applied Energy,
2014 IEEE Global Conference on Signal and Information
vol. 126, pp. 29–37, 2014.
Processing (GlobalSIP), pp. 258–262, IEEE, Atlanta, GA, USA,
[21] M. H. Chung, “Modelling of solar irradiance forecasting using
December 2014.
[4] S.-V. Oprea, A. Bâra, A. Uţă, A. Pı̂rjan, and G. Căruţaşu, local meteorological data,” KIEAE Journal, vol. 17, no. 6,
“Analyses of distributed generation and storage effect on the pp. 273–278, 2017.
electricity consumption curve in the smart grid context,” [22] M. H. Chung, “The correlation of solar radiation and at-
Sustainability, vol. 10, no. 7, p. 2264, 2018. mospheric elements including air pollution,” KIEAE Journal,
[5] E. Mengelkamp, J. Gärttner, K. Rock, S. Kessler, L. Orsini, and vol. 19, no. 1, pp. 69–74, 2019.
C. Weinhardt, “Designing microgrid energy markets,” Ap- [23] J. Sola and J. Sevilla, “Importance of input data normalization
plied Energy, vol. 210, pp. 870–880, 2018. for the application of neural networks to complex industrial
[6] C. Park and T. Yong, “Comparative review and discussion on problems,” IEEE Transactions on Nuclear Science, vol. 44,
P2P electricity trading,” Energy Procedia, vol. 128, no. 3–9, no. 3, pp. 1464–1468, 1997.
2017. [24] C. F. M. Coimbra, H. T. C. Pedro, and J. Kleissl, “Stochastic-
[7] C. Zhang, J. Wu, C. Long, and M. Cheng, “Review of existing learning methods,” in Solar Energy Forecasting and Resource
peer-to-peer energy trading projects,” Energy Procedia, Assessment, pp. 383–406, Academic Press, Oxford, UK, 1st
vol. 105, pp. 2563–2568, 2017. edition, 2013.
[8] J. Jimeno, J. Anduaga, J. Oyarzabal, and A. G. de Muro, [25] E. W. Law, A. A. Prasad, M. Kay, and R. A. Taylor, “Direct
“Architecture of a microgrid energy management system,” normal irradiance forecasting and its application to con-
European Transactions on Electrical Power, vol. 21, no. 2, centrated solar thermal output forecasting—a review,” Solar
pp. 1142–1158, 2011. Energy, vol. 108, pp. 287–307, 2014.
Advances in Civil Engineering 13
[26] M. Daigne, M. David, P. Lauret, J. Boland, and N. Schmutz, [42] W. Omran, Performance Analysis of Grid-Connected Photo-
“Review of solar irradiance forecasting methods and a voltaic System, Ph.D. dissertation, Department of Electrical
proposition for small-scale insular grids,” Renewable and and Computer Engineering, University of Waterloo, Ontario,
Sustainable Energy Reviews, vol. 27, pp. 65–76, 2013. Canada, 2010.
[27] A. Mellit and A. M. Pavan, “A 24-h forecast of solar irradiance [43] C. J. Willmott and K. Matsuura, “Advantages of the mean
using artificial neural network: application for performance absolute error (MAE) over the root mean square error
prediction of a grid-connected PV plant at Trieste, Italy,” Solar (RMSE) in assessing average model performance,” Climate
Energy, vol. 84, no. 5, pp. 807–821, 2010. Research, vol. 30, no. 1, pp. 79–82, 2009.
[28] P. Lauret, E. Fock, R. N. Randrianarivony, and J.-F. Manicom- [44] T. Chai and R. R. Draxler, “Root mean square error (RMSE) or
Ramsamy, “Bayesian neural network approach to short time mean absolute error (MAE)?—arguments against avoiding
load forecasting,” Energy Conversion and Management, RMSE in the literature,” Geoscientific Model Development,
vol. 49, no. 5, pp. 1156–1166, 2008. vol. 7, no. 3, pp. 1247–1250, 2014.
[29] R. Marquez and C. F. M. Coimbra, “Forecasting of global and [45] J. E. Nash and J. V. Sutcliffe, “River flow forecasting through
direct solar irradiance using stochastic learning methods, conceptual models part I - A discussion of principles,” Journal
ground experiments and the NWS database,” Solar Energy, of Hydrology, vol. 10, no. 3, pp. 282–290, 1970.
vol. 85, no. 5, pp. 746–756, 2011. [46] Y. Chu, H. T. C. Pedro, M. Li, and C. F. M. Coimbra, “Real-
[30] F. S. Tymvios, C. P. Jacovides, S. C. Michaelides, and time forecasting of solar irradiance ramps with smart image
processing,” Solar Energy, vol. 114, pp. 91–104, 2015.
C. Scouteli, “Comparative study of Ångström’s and artificial
[47] C. Cornaro, M. Pierro, and F. Bucci, “Master optimization
neural networks’ methodologies in estimating global solar
process based on neural networks ensemble for 24-h solar
radiation,” Solar Energy, vol. 78, no. 6, pp. 752–762, 2005.
irradiance forecast,” Solar Energy, vol. 111, pp. 297–312, 2015.
[31] S. Haykin, Neural Networks—A Comprehensive Foundation,
[48] H. Sun, N. Zhao, X. Zeng, and D. Yan, “Study of solar ra-
Pearson Prentice Hall, New York, USA, 3rd edition, 2008.
diation prediction and modeling of relationships between
[32] J. Yang, H. Rivard, and R. Zmeureanu, “On-line building
solar radiation and meteorological variables,” Energy Con-
energy prediction using adaptive artificial neural networks,” version and Management, vol. 105, pp. 880–890, 2015.
Energy and Buildings, vol. 37, no. 12, pp. 1250–1259, 2005.
[33] J.-L. Zhang, “On the convergence properties of the Levenberg-
Marquardt method,” Optimization, vol. 52, no. 6, pp. 739–
756, 2003.
[34] L. S. H. Ngia and J. Sjoberg, “Efficient training of neural nets
for nonlinear adaptive filtering using a recursive Levenberg-
Marquardt algorithm,” IEEE Transactions on Signal Process-
ing, vol. 48, no. 7, pp. 1915–1927, 2000.
[35] F. M. Dias, A. Antunes, J. Vieira, and A. M. Mota, “On-line
training of neural networks: a sliding window approach for
the levenberg-marquardt algorithm,” in Proceedings of the
International Work-Conference on the Interplay between
Natural and Artificial Computation, pp. 577–585, Springer,
Las Palmas, Spain, June 2005.
[36] J. W. Moon, “Performance of ANN-based predictive and
adaptive thermal-control methods for disturbances in and
around residential buildings,” Building and Environment,
vol. 48, pp. 15–26, 2012.
[37] B. Liu and R. Jordan, “Daily insolation on surfaces tilted
towards equator,” ASHRAE Transactions, vol. 67, pp. 526–541,
1962.
[38] J. K. Page, “The estimation of monthly mean values of daily
total short wave radiation on vertical and inclined surface
from sunshine records for latitude 40°N–40°S,” in Proceedings
of the UN Conference on New Sources of Energy, vol. 4,
pp. 378–390, Rome, Italy, August 1961.
[39] J. V. Paatero and P. D. Lund, “Effects of large-scale photo-
voltaic power integration on electricity distribution net-
works,” Renewable Energy, vol. 32, no. 2, pp. 216–234, 2007.
[40] M. E. Ropp, M. Begovic, and A. Rohatgi, “Determination of
the curvature derating factor for the georgia tech aquatic
center photovoltaic array,” in Proceedings of the Conference
Record of the Twenty-Sixth IEEE Photovoltaic Specialists
Conference, pp. 1297–1300, Anaheim, CA, USA, September
1997.
[41] Ö. Ayvazoğluyüksel and Ü. B. Filik, “Estimation methods of
global solar radiation, cell temperature and solar power
forecasting: A review and case study in Eskişehir,” Renewable
and Sustainable Energy Reviews, vol. 91, pp. 639–653, 2018.