The potential of different artificial neural network (ANN) techniques in daily global solar radiation modeling based on meteorological data

The main objective of present study is to predict daily global solar radiation (GSR) on a horizontal surface, based on meteorological variables, using different artificial neural network (ANN) techniques. Daily mean air temperature, relative humidity, sunshine hours, evaporation, and wind speed values between 2002 and 2006 for Dezful city in Iran (32°16′N, 48°25′E), are used in this study. In order to consider the effect of each meteorological variable on daily GSR prediction, six following combinations of input variables are considered:

(I)

Day of the year, daily mean air temperature and relative humidity as inputs and daily GSR as output.

(II)

Day of the year, daily mean air temperature and sunshine hours as inputs and daily GSR as output.

(III)

Day of the year, daily mean air temperature, relative humidity and sunshine hours as inputs and daily GSR as output.

(IV)

Day of the year, daily mean air temperature, relative humidity, sunshine hours and evaporation as inputs and daily GSR as output.

(V)

Day of the year, daily mean air temperature, relative humidity, sunshine hours and wind speed as inputs and daily GSR as output.

(VI)

Day of the year, daily mean air temperature, relative humidity, sunshine hours, evaporation and wind speed as inputs and daily GSR as output.

Multi-layer perceptron (MLP) and radial basis function (RBF) neural networks are applied for daily GSR modeling based on six proposed combinations.The measured data between 2002 and 2005 are used to train the neural networks while the data for 214 days from 2006 are used as testing data.The comparison of obtained results from ANNs and different conventional GSR prediction (CGSRP) models shows very good improvements (i.e. the predicted values of best ANN model (MLP-V) has a mean absolute percentage error (MAPE) about 5.21% versus 10.02% for best CGSRP model (CGSRP 5)).     

Artificial neural network estimation of global solar radiation using air temperature and relative humidity

Measured air temperature and relative humidity values between 1998 and 2002 for Abha city in Saudi Arabia were used for the estimation of global solar radiation (GSR) in future time domain using artificial neural network method. The estimations of GSR were made using three combinations of data sets namely: (i) day of the year and daily maximum air temperature as inputs and GSR as output, (ii) day of the year and daily mean air temperature as inputs and GSR as output and (iii) time day of the year, daily mean air temperature and relative humidity as inputs and GSR as output. The measured data between 1998 and 2001 were used for training the neural networks while the remaining 240 days’ data from 2002 as testing data. The testing data were not used in training the neural networks. Obtained results show that neural networks are well capable of estimating GSR from temperature and relative humidity. This can be used for estimating GSR for locations where only temperature and humidity data are available.     

Simulation model of ANN based maximum power point tracking controller for solar PV system

In this paper the simulation model of an artificial neural network (ANN) based maximum power point tracking controller has been developed. The controller consists of an ANN tracker and the optimal control unit. The ANN tracker estimates the voltages and currents corresponding to a maximum power delivered by solar PV (photovoltaic) array for variable cell temperature and solar radiation. The cell temperature is considered as a function of ambient air temperature, wind speed and solar radiation. The tracker is trained employing a set of 124 patterns using the back propagation algorithm. The mean square error of tracker output and target values is set to be of the order of 10⁻⁵ and the successful convergent of learning process takes 1281 epochs. The accuracy of the ANN tracker has been validated by employing different test data sets. The control unit uses the estimates of the ANN tracker to adjust the duty cycle of the chopper to optimum value needed for maximum power transfer to the specified load.     

Use of artificial neural networks for mapping of solar potential in Turkey

Turkey has sufficient solar radiation intensities and radiation durations for solar thermal applications since Turkey lies in a sunny belt, between 36° and 42° N latitudes. The yearly average solar-radiation is 3.6 kWh/m²day, and the total yearly radiation period is ∼2610 h. The main focus of this study is to determine the solar-energy potential in Turkey using artificial neural-networks (ANNs). Scaled conjugate gradient (SCG), Pola-Ribiere conjugate gradient (CGP), and Levenberg-Marquardt (LM) learning algorithms and a logistic sigmoid transfer function were used in the network. In order to train the neural network, meteorological data for the last 3 years (2000–2002) from 17 stations (namely cities) spread over Turkey were used as training (11 stations) and testing (6 stations) data. Meteorological and geographical data (latitude, longitude, altitude, month, mean sunshine duration, and mean temperature) are used as inputs to the network. Solar radiation is in the output layer. The maximum mean absolute percentage error was found to be less than 6.7% and R² values to be about 99.8937% for the testing stations. However, the respective values were found to be 2.41 and 99.99658% for the training stations. The trained and tested ANN models show greater accuracies for evaluating solar resource posibilities in regions where a network of monitoring stations has not been established in Turkey. The predicted solar-potential values from the ANN were given in the form of monthly maps. These maps are of prime importance for different working disciplines, like those of scientists, architects, meteorologists, and solar engineers in Turkey. The predictions from ANN models could enable scientists to locate and design solar-energy systems in Turkey and determine the appropriate solar technology.     

Modeling of solar radiation using remote sensing and artificial neural network in Turkey

Artificial neural networks (ANNs) were used to estimate solar radiation in Turkey (26–45°E, 36–42°N) using geographical and satellite-estimated data. In order to train the Generalized regression neural network (GRNN) geographical and satellite-estimated data for the period from January 2002 to December 2002 from 19 stations spread over Turkey were used in training (ten stations) and testing (nine stations) data. Latitude, longitude, altitude, surface emissivity for ?_4, surface emissivity for ?_5, and land surface temperature are used in the input layer of the network. Solar radiation is the output. Root Mean Square Error (RMSE) and correlation coefficient (R²) between the estimated and measured values for monthly mean daily sum with ANN values have been found as 0.1630 MJ/m² and 95.34% (training stations), 0.3200 MJ/m² and 93.41% (testing stations), respectively. Since these results are good enough it was concluded that the developed GRNN tool can be used to predict the solar radiation in Turkey.     

A simple and efficient algorithm to estimate daily global solar radiation from geostationary satellite data

Surface global solar radiation (GSR) is the primary renewable energy in nature. Geostationary satellite data are used to map GSR in many inversion algorithms in which ground GSR measurements merely serve to validate the satellite retrievals. In this study, a simple algorithm with artificial neural network (ANN) modeling is proposed to explore the non-linear physical relationship between ground daily GSR measurements and Multi-functional Transport Satellite (MTSAT) all-channel observations in an effort to fully exploit information contained in both data sets. Singular value decomposition is implemented to extract the principal signals from satellite data and a novel method is applied to enhance ANN performance at high altitude. A three-layer feed-forward ANN model is trained with one year of daily GSR measurements at ten ground sites. This trained ANN is then used to map continuous daily GSR for two years, and its performance is validated at all 83 ground sites in China. The evaluation result demonstrates that this algorithm can quickly and efficiently build the ANN model that estimates daily GSR from geostationary satellite data with good accuracy in both space and time.     

Forecasting based on neural network approach of solar potential in Turkey

As Turkey lies near the sunny belt between 36 and 42°N latitudes, most of the locations in Turkey receive abundant solar energy. Average annual temperature is 18–20 °C on the south coast, falls down to 14–16 °C on the west coast, and fluctuates 4–18 °C in the central parts. The yearly average solar radiation is 3.6 kW h/m² day, and the total yearly radiation period is 2610 h. The main focus of this study is put forward to solar energy potential in Turkey using artificial neural networks (ANNs). Scaled conjugate gradient (SCG), Pola-Ribiere conjugate gradient (CGP), and Levenberg–Marquardt (LM) learning algorithms and logistic sigmoid transfer function were used in the network. In order to train the neural network, meteorological data for last 4 years (2000–2003) from 12 cities (Çanakkale, Kars, Hakkari, Sakarya, Erzurum, Zonguldak, Balıkesir, Artvin, Çorum, Konya, Siirt, Tekirdağ) spread over Turkey were used as training (nine stations) and testing (three stations) data. Meteorological and geographical data (latitude, longitude, altitude, month, mean sunshine duration, and mean temperature) is used as input to the network. Solar radiation is the output. The maximum mean absolute percentage error was found to be less than 6.78% and R² values to be about 99.7768% for the testing stations. These values were found to be 5.283 and 99.897% for the training stations. The trained and tested ANN models show greater accuracy for evaluating solar resource posibilities in regions where a network of monitoring stations have not been established in Turkey. The predictions from ANN models could enable scientists to locate and design solar energy systems in Turkey and determine the best solar technology.     

Artificial neural networks used for the performance prediction of a thermosiphon solar water heater

Artificial Neural Networks (ANN) are widely accepted as a technology offering an alternative way to tackle complex and ill-defined problems. They can learn from examples, are fault tolerant, are able to deal with non-linear problems, and once trained can perform prediction at high speed. ANNs have been used in diverse applications and they have shown to be particularly effective in system modelling as well as for system identification. The objective of this work is to train an artificial neural network (ANN) to learn to predict the performance of a thermosiphon solar domestic water heating system. This performance is measured in terms of the useful energy extracted and of the stored water temperature rise. An ANN has been trained using performance data for four types of systems, all employing the same collector panel under varying weather conditions. In this way the network was trained to accept and handle a number of unusual cases. The data presented as input were, the storage tank heat loss coefficient (U-value), the type of system (open or closed), the storage volume, and a total of fifty-four readings from real experiments of total daily solar radiation, total daily diffuse radiation, ambient air temperature, and the water temperature in storage tank at the beginning of the day. The network output is the useful energy extracted from the system and the water temperature rise. The statistical coefficient of multiple determination (R²-value) obtained for the training data set was equal to 0.9914 and 0.9808 for the two output parameters respectively. Both values are satisfactory because the closer R²-value is to unity the better is the mapping. Unknown data for all four systems were subsequently used to investigate the accuracy of prediction. These include performance data for the systems considered for the training of the network at different weather conditions. Predictions with maximum deviations of 1 MJ and 2.2°C were obtained respectively. Random data were also used both with the performance equations obtained from the experimental measurements and with the artificial neural network to predict the above two parameters. The predicted values thus obtained were very comparable. These results indicate that the proposed method can successfully be used for the estimation of the performance of the particular thermosiphon system at any of the different types of configuration used here. The greatest advantage of the present model is the capacity of the network to learn from examples and thus gradually improve its performance. This is done by embedding experimental knowledge in the network.     

Application of artificial neural networks for forecasting photovoltaic system parameters

The main element which justifies the installation of a photovoltaic system is the solar energy potential. Various structures of artificial neural networks (ANNs) are used for predicting the sun location, the global solar radiation (GSR) at horizontal and inclined plans. Real meteorological data have been exploited in order to validate the computation results. The ANNs are also carried out to predict the current-voltage characteristics of the photovoltaic module. It can be concluded that the ANNs effectively predict the behavior of photovoltaic system parameters with good a coefficient of determination.     

Modelling of solar energy potential in Nigeria using an artificial neural network model

In this study, an artificial neural network (ANN) based model for prediction of solar energy potential in Nigeria (lat. 4–14°N, log. 2–15°E) was developed. Standard multilayered, feed-forward, back-propagation neural networks with different architecture were designed using neural toolbox for MATLAB. Geographical and meteorological data of 195 cities in Nigeria for period of 10 years (1983–1993) from the NASA geo-satellite database were used for the training and testing the network. Meteorological and geographical data (latitude, longitude, altitude, month, mean sunshine duration, mean temperature, and relative humidity) were used as inputs to the network, while the solar radiation intensity was used as the output of the network. The results show that the correlation coefficients between the ANN predictions and actual mean monthly global solar radiation intensities for training and testing datasets were higher than 90%, thus suggesting a high reliability of the model for evaluation of solar radiation in locations where solar radiation data are not available. The predicted solar radiation values from the model were given in form of monthly maps. The monthly mean solar radiation potential in northern and southern regions ranged from 7.01–5.62 to 5.43–3.54 kW h/m² day, respectively. A graphical user interface (GUI) was developed for the application of the model. The model can be used easily for estimation of solar radiation for preliminary design of solar applications.     

Comparison of the different artificial neural networks in prediction of biomass gasification products

In this study, artificial neural networks (ANNs) and a nonlinear autoregressive exogenous (NARX) neural network model were employed in order to model a fixed bed downdraft gasification. The relation between the feature group and the regression performance was investigated. First, feature group consists of the equivalence ratio (ER), air flow rate (AF), and temperature distribution (T0‐T5) obtained from the fixed bed downdraft gasifiers, while the second group includes ultimate and proximate values of biomasses, ER, AF, and the reduction temperature (T0). Models constructed to predict the syngas composition (H₂, CO₂, CO, CH₄) and calorific value. Experimental gasification data that involve 3831 data samples that belong to pinecone and wood pellet were used for training the ANNs. Different ANN architecture and NARX time series model have been constructed to examine the prediction accuracy of the models. The results of the ANN models were consistent with the experimental data (R² > 0.99). The overall score of NARX time series networks is found to be higher than other architecture types. A successful method is proposed to reduce the number of features, and the effect of the features on the prediction capability was examined by calculating the relative importance index using the Garson's equation.     

Modeling global solar radiation using Particle Swarm Optimization (PSO)

The quantity of solar radiation received by the earth’s surface is very important to numerous renewable energy applications. However, direct measurement of solar data is not widely available, especially in developing countries. This paper uses Particle Swarm Optimization (PSO) to train an artificial neural network (PSO–ANN) using data from available measurement stations to estimate monthly mean daily Global Solar Radiation (GSR) at locations where no measurement stations are available. The inputs to the networks are: month of the year, latitude, longitude, altitude, and sunshine duration, and the output is the monthly mean daily GSR at the specified location. Using training data from 31 stations and testing data from 10 locations, the PSO–ANN outperforms a neural network trained using the standard backpropagation (BP) algorithm (BP–ANN) with an average Mean Absolute Percentage Error (MAPE) of 8.85% for the PSO–ANN and 12.61% for the BP–ANN. The performance is improved significantly, when we use the leave-one-out method, where data from 40 locations is used for training and data from the 41st station is used for assessing the performance. In this case the average of MAPE on data from the 10 testing stations is about 7%. We used the same method to assess the performance of the PSO–ANN on testing data from each of the 41 stations with an overall average MAPE of about 10.3%. Comparison with BP–ANN and an empirical model showed the superiority of the PSO–ANN.     

Neural network approach for food temperature prediction during solar drying

In the present study, the application of artificial neural network (ANN) for prediction of temperature variation of food product during solar drying is investigated. The important climatic variables namely, solar radiation intensity and ambient air temperature are considered as the input parameters for ANN modeling. Experimental data on potato cylinders and slices obtained with mixed mode solar dryer for 9 typical days of different months of the year were used for training and testing the neural network. A methodology is proposed for development of optimal neural network. Results of analysis reveal that the network with 4 neurons and logsig transfer function and trainrp back propagation algorithm is the most appropriate approach for both potato cylinders and slices based on minimum measures of error. In order to test the worthiness of ANN model for prediction of food temperature variation, the analytical heat diffusion model with appropriate boundary conditions and statistical model are also proposed. Based on error analysis results, the prediction capability of ANN model is found to be the best of all the prediction models investigated, irrespective of food sample geometry.     

Assessment of diffuse solar energy under general sky condition using artificial neural network

In this paper, artificial neural network (ANN) models are developed for estimating monthly mean hourly and daily diffuse solar radiation. Solar radiation data from 10 Indian stations, having different climatic conditions, all over India have been used for training and testing the ANN model. The coefficient of determination (R²) for all the stations are higher than 0.85, indicating strong correlation between diffuse solar radiation and selected input parameters. The feedforward back-propagation algorithm is used in this analysis. Results of ANN models have been compared with the measured data on the basis of percentage root-mean-square error (RMSE) and mean bias error (MBE). It is found that maximum value of RMSE in ANN model is 8.8% (Vishakhapatnam, September) in the prediction of hourly diffuse solar radiation. However, for other stations same error is less than 5.1%. The computation of monthly mean daily diffuse solar radiation is also carried out and the results so obtained have been compared with those of other empirical models. The ANN model shows the maximum RMSE of 4.5% for daily diffuse radiation, while for other empirical models the same error is 37.4%. This shows that ANN model is more accurate and versatile as compared to other models to predict hourly and daily diffuse solar radiation.     

Estimation of monthly-mean daily global solar radiation based on MODIS and TRMM products

Global solar radiation (GSR) is required in a large number of fields. Many parameterization schemes are developed to estimate it using routinely measured meteorological variables, since GSR is directly measured at a limited number of stations. Even so, meteorological stations are sparse, especially, in remote areas. Satellite signals (radiance at the top of atmosphere in most cases) can be used to estimate continuous GSR in space. However, many existing remote sensing products have a relatively coarse spatial resolution and these inversion algorithms are too complicated to be mastered by experts in other research fields. In this study, the artificial neural network (ANN) is utilized to build the mathematical relationship between measured monthly-mean daily GSR and several high-level remote sensing products available for the public, including Moderate Resolution Imaging Spectroradiometer (MODIS) monthly averaged land surface temperature (LST), the number of days in which the LST retrieval is performed in 1 month, MODIS enhanced vegetation index, Tropical Rainfall Measuring Mission satellite (TRMM) monthly precipitation. After training, GSR estimates from this ANN are verified against ground measurements at 12 radiation stations. Then, comparisons are performed among three GSR estimates, including the one presented in this study, a surface data-based estimate, and a remote sensing product by Japan Aerospace Exploration Agency (JAXA). Validation results indicate that the ANN-based method presented in this study can estimate monthly-mean daily GSR at a spatial resolution of about 5 km with high accuracy.     

Estimation of monthly average daily global solar irradiation using artificial neural networks

This study explores the possibility of developing a prediction model using artificial neural networks (ANN), which could be used to estimate monthly average daily global solar irradiation on a horizontal surface for locations in Uganda based on weather station data: sunshine duration, maximum temperature, cloud cover and location parameters: latitude, longitude, altitude. Results have shown good agreement between the estimated and measured values of global solar irradiation. A correlation coefficient of 0.974 was obtained with mean bias error of 0.059 MJ/m² and root mean square error of 0.385 MJ/m². The comparison between the ANN and empirical method emphasized the superiority of the proposed ANN prediction model.     

Optimization of an artificial neural network dedicated to the multivariate forecasting of daily global radiation

This paper presents an application of Artificial Neural Networks (ANNs) to predict daily solar radiation. We look at the Multi-Layer Perceptron (MLP) network which is the most used of ANNs architectures. In previous studies, we have developed an ad-hoc time series preprocessing and optimized a MLP with endogenous inputs in order to forecast the solar radiation on a horizontal surface. We propose in this paper to study the contribution of exogenous meteorological data (multivariate method) as time series to our optimized MLP and compare with different forecasting methods: a naïve forecaster (persistence), ARIMA reference predictor, an ANN with preprocessing using only endogenous inputs (univariate method) and an ANN with preprocessing using endogenous and exogenous inputs. The use of exogenous data generates an nRMSE decrease between 0.5% and 1% for two stations during 2006 and 2007 (Corsica Island, France). The prediction results are also relevant for the concrete case of a tilted PV wall (1.175 kWp). The addition of endogenous and exogenous data allows a 1% decrease of the nRMSE over a 6 months-cloudy period for the power production. While the use of exogenous data shows an interest in winter, endogenous data as inputs on a preprocessed ANN seem sufficient in summer.     

An analysis of thermal and solar zone radiation models using an Angstrom–Prescott equation and artificial neural networks

The correlation between the clearness index and sunshine duration is useful in the estimation of the solar radiation for areas where measured solar radiation data are unavailable. Regression techniques and artificial neural networks were used to investigate the correlations between daily global solar radiation (GSR) and sunshine duration for different climates in China. Measurements made during the 30-year period (1971–2000) from 41 measuring stations covering 9 thermal and 7 solar climate zones and sub-zones across China were gathered and analysed. The performance of the regression and the ANN models in the thermal and solar zones was analysed and compared. The coefficient of determination (R²), Nash–Sutcliffe efficiency coefficient (NSEC), mean bias error (MBE) and root-mean-square error (RMSE) were determined. It was found that the regression models in both the thermal and the solar climate zones showed a strong correlation between the clearness index and sunshine duration (R²=0.79–88). There appeared to be an increasing trend of larger MBE and RMSE from colder climates in the north to warmer climates in the south. In terms of the thermal and solar climate zone models, there was very little to choose between the two models.     

A flexible neural network-fuzzy data envelopment analysis approach for location optimization of solar plants with uncertainty and complexity

This study presents a flexible neuro-fuzzy approach for location optimization of solar plants with possible complexity and uncertainty. The flexible approach is composed of artificial neural network (ANN) and fuzzy data envelopment analysis (FDEA). The intelligent approach of this study is applied for location optimization of solar plants in Iran. First, FDEA is validated by DEA, and then it is used for ranking of solar plant units (SPUs) and the best α-cut is selected based on the test of Normality. Also, several ANNs are developed through multi layer perceptron (MLP) for ranking of solar plants and the best one with minimum mean absolute percentage of error (MAPE) is selected for further considerations. Finally, the preferred model (FDEA or ANN) is selected based on test of Normality. The implementation of the flexible approach for solar plants in Iran identifies the preferred FDEA at α = 0.3, where is the level of data uncertainty. This indicates that the data are collected from the uncertain and fuzzy environment. This is the first study that presents a flexible approach for identification of optimum location of solar plants with possible noise, non-linearity, complexity and environmental uncertainty. This would help policy makers to identify the preferred Strategy for location optimization problems associated with solar plant units.     

Performance analysis of a hybrid photovoltaic thermal double pass air collector using ANN

This paper presents the use of artificial neural network for performance analysis of a semi transparent hybrid photovoltaic thermal double pass air collector for four weather conditions (a, b, c and d type) of New Delhi. The MATLAB 7.1 neural networks toolbox has been used for defining and training of ANN for calculations of thermal energy, electrical energy, overall thermal energy and overall exergy. The ANN model uses ambient air temperature, global solar radiation, diffuse radiation and number of clear days as input parameters for four weather conditions. The transfer function, neural network configuration and learning parameters have been selected based on highest convergence during training and testing of network. About 2000 sets of data from four weather stations (Bangalore, Mumbai, Srinagar, and Jodhpur) have been given as input for training and data of the fifth weather station (New Delhi) has been used for testing purpose. It has been observed that the best transfer function for a given configuration is logsig. The feedforward back-propagation algorithm has been used in this analysis. Further the results of ANN model have been compared with analytical values on the basis of root mean square error.