Solar radiation estimation methods using ANN and empirical models

doi:10.1016/j.compag.2019.03.022

Computers and Electronics in Agriculture

Volume 160, May 2019, Pages 160-167

https://doi.org/10.1016/j.compag.2019.03.022 Get rights and content

Highlights

•
Hargreaves method to estimate solar radiation in two station of Greece.
•
Estimation of ANNs model to describe daily solar radiation.
•
Multi-linear regression models with different combination of input variables.
•
Comparison of Hargreaves, ANN and MLR models with data from two station.

Abstract

Many empirical equations and methods have been used and proposed in order to estimate the solar radiation (R_s). In this work, empirical equations, such as Hargreaves method, Artificial Neural Networks (ANN) technology and multi-linear regression methods (MLR) were used to estimate solar radiation. The daily meteorological measurements of air temperature, radiation, humidity and wind velocity from the stations of Aristotle University Farm and Amintaio in Northern Greece were used to derive the solar radiation models. The measurements of R_s were used to derive new and to evaluate existing models. Different combinations of input variables were examined in the ANN models, and in the MLR models different variables were used. The results of RMSE criteria of the examined models showed that they are in the same range with many other models describing R_s as summarized in many review articles. The use of extraterrestrial radiation and the square root of daily difference in temperature in the ANN and MLR models improve the accuracy of the results. The results of ANN models in comparison to MLR models using the same input variables are consistent between them.

Introduction

Solar radiation is critical in the design and operation of solar energy utilization systems, and it is one of the most significant parameters in reference evapotranspiration (ET_o) estimation. The estimation of ET_o is very interesting in agriculture, hydrology and ecology. The accuracy of solar radiation measurements and its respective estimations through models is then extremely important.

Numerous methods or equations have been developed and are being used to estimate the average monthly and daily solar radiation (R_s) in the last decades. These depend mainly upon the availability of meteorological data. For the designers and manufactures of solar equipment but also for other professionals, such as architects, agronomists and farmers, it is very important to have methods to indirectly estimate the solar radiation based on other readily available meteorological data (e.g. the actual and maximum sunshine duration, the average, maximum, minimum ambient temperature, the extraterrestrial radiation on horizontal surface (R_a), the longitude, altitude and latitude etc). The empirical methods of solar radiation and temperature based methods are considered as the most accurate methods to estimate ET_o (Priestley and Taylor, 1972, Aschonitis et al., 2015, Antonopoulos and Antonopoulos, 2017).

The solar radiation is a meteorological variable which is not measured or is of low accuracy in many cases. Particularly in developing countries, solar radiation measurements are not easily available due to the cost, maintenance and calibration requirements of the respective equipment (Despotovic et al., 2015). In the Greek territory, many meteorological stations have been established in the last decade. Most of them do not measure solar radiation and thus the indirect estimation of solar radiation by other easily measured meteorological parameters is one alternative solution. Many models have been proposed and developed based on mathematical formulas, such as empirical models, artificial intelligence techniques (Zhang et al., 2017) and satellite-based methods. The empirical methods are divided into three groups (Zhang et al., 2017), the sunshine duration fraction models, the modified sunshine duration fraction models and the non-sunshine duration methods. The widely adopted R_s estimation models are based on sunshine duration and temperature datasets (Besharat et al., 2013). The most common R_s models are sunshine based (Allen et al., 1998), but the use of these models is often limited by the lack of available sunshine records. The temperature based models such as the Hargreaves model, are preferred in the absence of sunshine data.

The artificial intelligence techniques have being developed in recent decades to predict or forecast solar radiation. Artificial neural networks (ANN) have been used to accurately predict solar radiation by Yadav and Chandel, 2014, Kisi et al., 2015, Premalatha and Valan Arasu, 2016, Zhang et al., 2017 among many others. Zhang et al. (2017) reviewed and compared models, amongst of them the ANN models, from the points of view of time scale and estimation type.

Significant efforts for the accurate estimation of ET_o under the Greek environmental-meteorological conditions have been carried out by several researchers. In many of them, the R_s estimation as the most significant input variable, has received special attention and was analyzed and examined (Alexandris et al., 2006, Valiantzas, 2006, Mavromatis, 2008, Ampas et al., 2007, Aschonitis et al., 2012, Paraskevas et al., 2013, Efthimiou et al., 2013, Valiantzas, 2018).

The aim of this study is to evaluate empirical equations as Hargreaves, Artificial Neural networks technology and multi-linear regression methods to estimate the solar radiation. The use of different input variables in the ANN and MLR models was evaluated based on measured R_s values. Daily data sets from two meteorological stations, in northern Greece were used.

Section snippets

Empirical methods

Hargreaves and Samani, 1982, Hargreaves and Samani, 1985 developed a simple model for estimating the incoming shortwave solar radiation R_s (MJ m⁻² day⁻¹), which requires only temperature data, as follows: $R_{s} = K_{RS} \cdot R_{a} \cdot {(TD)}^{0.5}$

where K_RS is the adjustment coefficient of the radiation formula (°C^−0.5), R_a is the extraterrestrial radiation (MJ m⁻² day⁻¹) and TD (equal to T_max-T_min) is the temperature difference between maximum (T_max) and minimum (T_min) daily temperature (°C). Hargreaves (1994) recommended

Results

The results of this work concern the use of the empirical equation of Hargreaves, the development of ANN models and the development of multiple-linear regression models for estimating solar radiation. Two sets of meteorological data from different regions in Northern Greece were used to develop and compare the models.

Discussion

The empirical equation of Hargreaves and models of Artificial Neural networks and multi-linear regression methods for estimating solar radiation were evaluated in this study based on the available datasets from two regions of Northern Greece. The results showed that these models have the ability to describe R_s with relatively high accuracy.

The comparison of the proposed ANN and MLR models showed that the MLR models performed better for the AUTH station, while the opposite was observed for the

Conclusions

The suitability of Hargreaves method, Artificial Neural networks (ANN) and multi-linear regression methods (MLR) to estimate solar radiation was evaluated using daily meteorological data from two stations in Northern Greece.

The results of these models were compared with measured R_s values. The results of RMSE criteria showed that they are in the same range with many other models describing R_s as summarized by Zang et al. in a review article of R_s models. The use of extraterrestrial radiation

Acknowledgements

The authors acknowledge Municipality of Amintaio and School of Agriculture of Aristotle University of Thessaloniki in Greece for providing the data sets of meteorological stations.

References (45)

S. Alexandris et al.
Daily reference evapotranspiration estimates by the Copais approach
Agric. Water Manage.
(2006)
V.Z. Antonopoulos et al.
Simulation of water temperature and dissolved oxygen distribution in Lake Vegoritis
Greece. Ecol. Model.
(2003)
V.Z. Antonopoulos et al.
Daily reference evapotranspiration estimates by artificial neural networks technique and empirical equations using limited input climate variables
Comput. Electron. Agric.
(2017)
F. Besharat et al.
Empirical models for estimating global solar radiation: a review and case study
I. Daut et al.
Combination of Hargreaves method and linear regression as a new method to estimate solar radiation in Perlis
Northern Malaysia. Solar Energy
(2011)
M. Despotovic et al.
Review and statistical analysis of different global solar radiation sunshine models
Renew. Sustain. Energy Rev.
(2015)
A.A. El-Sebaii et al.
Estimation of global solar radiation on horizontal surfaces in Jeddah
Saudi Arabia. Energy Policy
(2009)
A.A. El-Sebaii et al.
Global, direct and diffuse solar radiation on horizontal and tilted surfaces in Jeddah
Saudi Arabia. Applied Energy
(2010)
S.K. Gianniou et al.
Evaporation and energy budget in Lake Vegoritis
Greece. J. Hydrol.
(2007)
Z. Jin et al.
General formula for estimation of monthly average daily global solar radiation in China
Energy Convers. Manage.
(2005)

A.K. Katiyar et al.

Simple correlation for estimating the global solar radiation on horizontal surfaces in India

Energy

(2010)

O. Kisi et al.

Long-term monthly evapotranspiration modeling by several data-driven methods without climatic data

Comput. Electron. Agric.

(2015)

H. Li et al.

Global solar radiation estimation with sunshine duration in Tibet

China. Renew. Energy

(2011)

K. Loague et al.

Statistical and graphical methods for evaluating solute transport models: overview and application

J. Contam. Hydrol.

(1991)

H.R. Maier et al.

Methods used for the development of neural networks for the prediction of water resource variables in river systems: current status and future directions

Environ. Model. Softw.

(2010)

N. Premalatha et al.

Prediction of solar radiation for solar systems by using ANN models with different bach propagation algorithms

J. Appl. Res. Technol.

(2016)

T. Raziei et al.

Estimation of ETo with Hargreaves-Samani and FAO-PM temperature methods for a wide range of climate in Iran

Agric. Water Manage.

(2013)

M. Rivero et al.

A new methodology to extend the validity of the Hargreaves-Samani model to estimate global solar radiation in different climates: case study Mexico

Renew. Energy

(2017)

J.D. Valiantzas

Simplified versions for the Penman evaporation equation using routine weather data

J. Hydrol.

(2006)

W. Wu et al.

Protocol for developing ANN models and its application to the assessment of the quality of the ANN model development process in drinking water quality modelling

Environ. Model. Softw.

(2014)

A.K. Yadav et al.

Solar radiation prediction using artificial neural network techniques: a review

Renew. Sustain. Energy Rev.

(2014)

J. Zhang et al.

A critical review of the models used to estimate solar radiation

Renew. Sustain. Energy Rev.

(2017)

Cited by (106)

Solar energy and wind power supply supported by battery storage and Vehicle to Grid operations
2024, Electric Power Systems Research
Worldwide activity in renewable energy is a motive power to introduce technological innovations. Integrating intermittent energy sources such as solar energy and wind power with battery storage and Vehicle to Grid operations has several advantages for the power grid. The first advantage is that energy storage supports the power grid during the periods that the power grid is facing challenges from high peak demand. The second advantage is that using battery storage and Vehicle to Grid operations would shift the power grid load from the peak and busy time to less demand time. And the third advantage uses energy storage and Vehicle to Grid operations to smooth the fluctuating power supply fed into the power grid by intermittent renewable energy resources. This energy storage idea is of particular importance because, in the future, more renewable energy sources are integrated into the power grid worldwide. The research objective includes the results and examines the role and advantages of battery storage and Vehicle to Grid operations integrated into intermittent sources. The battery storage and Vehicle to Grid operations will create a renewable power supply and enhance the power grid reliability, including a large proportion of intermitted renewable energy sources.
Predicting daily solar radiation using a novel hybrid long short-term memory network across four climate regions of China
2023, Computers and Electronics in Agriculture
Accurate daily solar radiation (R_s) data is crucial for the growth and development of crops, but its direct observation is relatively scarce in most regions around the world. Therefore, accurately estimating daily R_s is of great significance. Regarding the above practical requirement, this study developed a novel hybrid deep learning model (MEA-DBN-LSTM), which combined the Long Short-Term Memory model (LSTM), Deep Belief Network (DBN), and Mind Evolutionary Algorithm (MEA), to estimate the R_s across four different climate regions of China using five easily accessible meteorological input combination (sunshine-based, temperature-based and three complex-based). The results showed that MEA-DBN-LSTM was proved to be the most recommended R_s estimated model across four different climate regions of China, and the corresponding median R², NSE, RMSE, MAE, and MAPE values ranged 0.805–0.999, 0.656–0.954, 1.069–4.289 MJ m⁻² d^–1, 0.889–3.532 MJ m⁻² d^–1, and 0.055–0.296 respectively. Correspondingly, compared to the empirical R_s model, the improvement values of MEA-DBN-LSTM were 12.33–52.13 % for R², 14.80–69.38 % for NSE, 33.75–69.77 % for RMSE, 36.03–62.70 % for MAE, and 42.40–66.67 % for MAPE, respectively. Moreover, the MEA-DBN-LSTM exhibits improvements in R², NSE, RMSE, MAE, and MAPE, ranging 1.40–13.50 %, 1.27–24.79 %, 1.81–61.18 %, 1.90–61.30 %, and 2.31–57.55 %, respectively, compared to other forms of LSTM-type models. Specifically, this study found the DBN modules significantly improved the complex-based single LSTM model, but without impact on sunshine-based and temperature-based single LSTM models. Regarding the MEA modules, it was effective in the hyperparameter optimization compared to DBN-LSTM. Furthermore, significant accuracy discrepancies exist among R_s models using different input combinations across four different climate regions, and the complex-based was generally superior to other input combinations. Overall, our findings can provide accurate daily R_s across four different climate regions of China using easily accessible meteorological data, which is of great significance for crop water consumption estimation, agricultural water resources management, crop growth conditions optimization, and sustainable agricultural development.
Estimation of one-minute direct normal irradiance using a deep neural network for five climate zones
2023, Renewable and Sustainable Energy Reviews
Due to its high cost and significance for solar thermal technologies, the calculation of direct normal irradiance has dominated the modelling of solar radiation components. There are three methods described in the literature to calculate direct normal irradiance: regression analysis, probabilistic approach, and application of machine learning models. To demonstrate that machine learning models can achieve sufficient generality for solar radiation data with good statistical results, a new methodology is proposed with three main goals: that required only one measure data as input, trained a deep neural network model using only one station referred as the reference station, and evaluate the direct normal irradiance estimated by the model with the measured values from others stations of the same climate zone. The results then were compared against the most prominent separation models in the literature. The proposed methodology shows that the model learn the radiative characteristics of the reference station and is a feasible method to estimate the component of solar radiation in others location. Therefore, the proposed approach is a viable methodology that has the ability to reach the sought-after quasi-universal model by researchers, as demonstrated by the neural network’s similar or superior performance compared to the separation models.
Performance evaluation of seasonal solar irradiation models—case study: Karapınar town, Turkey
2023, Case Studies in Thermal Engineering
This paper presents the application of Particle Swarm Optimization (PSO) Algorithm, Artificial Neural Networks (ANNs) and Bagged Tree (BT) methods for forecasting seasonal solar irradiance in Karapınar town Turkey. These methods, namely ANN, PSO and BT methods are widely used for solar irradiance estimation. The study utilizes input data from 2007 to 2020, including air temperature, precipitation, snow mass, air density, and cloud cover fraction, to estimate solar irradiance as the output value. The performance of these methods is analysed according to error indicators such as mean absolute error (MAE), root mean square error (RMSE) and R². The results demonstrate that the BT method yields the lowest statistical error metrics. Additionally, the correlation matrix reveals a strong positive linear relationship (correlation coefficient of 0.9045) between air temperature and solar irradiance. The investigation of the impact of utilizing independent inputs on the forecast output has been conducted via the employment of multiple linear regression (MLR) subsets and diverse combinations.
Production of bio-oil from lychee-based biomass through pyrolysis and maximization of bio-oil yield with statistical and machine learning techniques
2023, Journal of Cleaner Production
This study's goal is to maximize the yield of pyrolysis oil utilizing lychee-based biomass for a quick pyrolysis process. Response surface methodology (RSM) and artificial neural network with generalized neural network (GNN) technique were used to optimize the maximum bio-oil yield while considering temperature, heating rate, retention time, and argon gas flow rate as independent variables and bio-oil production as the response. Typical neural networks have constraints such as a long training time, a large number of neurons, and a large number of hidden layers. A generalized neural network (GNN) has been constructed to address these limitations and build a non-linear controller for acquiring access to bio-oil production output. To anticipate bio-oil yield, a new prediction model is proposed that combines the benefits of an auto-adaptive management technique with the rapid reaction of a GNN. Three measures (MSE, RMSE, and R²) are explored in depth to assess the performance efficacy of the models. Based on the finding, 350 °C temperature, 125 min of retention time, 120 °C/min heating rate and 110 mL/min argon flow rate was used to produce the highest bio-oil production (38.43%). Based on techniques, both models used in the study produce equally acceptable outcomes for projecting bio-oil yield. The greatest R² value for the bio-oil yield ranges from 0.94 to 0.99, and all outcomes are classified as "good" regarding RMSE (all RMSE values are close to one). When all indicators are combined, GNN provides superior accuracy in predicting bio-oil yield in this investigation, followed by ANN.
Evaluation of artificial neural network models for predictive monitoring of biogas production from cassava wastewater: A training algorithms approach
2023, Biomass and Bioenergy
Anaerobic digestion (AD) is an established technology for resource recovery and renewable energy production. However, its performance is dictated by diverse variables including operating parameters and system design. Then, the development of modelling and optimization strategies, such as the application of machine learning (ML), to predict/maintain its performance have received significant attention. In this study, ML integrated based models for predicting cumulative biogas from cassava wastewater AD were developed using artificial neural networks (ANN). Biogas generation was investigated for 21 days with and without the addition of calcium particles from milled calcined chicken waste as a low-cost substitute for buffering. Digestion time, pH, and calcium eggshell-based concentration were selected as input features for biogas prediction through the implementation of the ANN approach. Moreover, seven different training algorithms were used to train the model, and a genetic algorithm (GA) was also implemented to optimize the ANN architecture, resulting in faster learning and higher prediction performance. Given the data set, the Levenberg-Marquardt algorithm with a hyperbolic tangent as a transfer function to the hidden and output layers was the most efficient model in predicting the biogas produced with an R-value of 0.9999.

View all citing articles on Scopus

View full text

Original papersSolar radiation estimation methods using ANN and empirical models

Highlights

Abstract

Introduction

Section snippets

Empirical methods

Results

Discussion

Conclusions

Acknowledgements

Agric. Water Manage.

Greece. Ecol. Model.

Comput. Electron. Agric.

Northern Malaysia. Solar Energy

Renew. Sustain. Energy Rev.

Saudi Arabia. Energy Policy

Saudi Arabia. Applied Energy

Greece. J. Hydrol.

Energy Convers. Manage.

Energy

Comput. Electron. Agric.

China. Renew. Energy

J. Contam. Hydrol.

Environ. Model. Softw.

J. Appl. Res. Technol.

Agric. Water Manage.

Renew. Energy

J. Hydrol.

Environ. Model. Softw.

Renew. Sustain. Energy Rev.

Renew. Sustain. Energy Rev.

Original papers
Solar radiation estimation methods using ANN and empirical models