Comparison of experimental and semi-experimental models for predicting solar thermal power plants with artificial neural network

Abstract

Researchers have recently accomplished many studies on using renewable energy sources without pollution due to the significant development in demand for energy access. The fact that these energy sources are pure and offer free renewable energy, like solar and wind, has attracted the interest of several governments and enterprises. Solar energy may be viewed as a clean, sustainable, and renewable energy source that will be very important for the planet's energy needs in the future. Accurate knowledge of the amount of solar radiation must be accessible in the initial stage to utilize solar energy to its fullest potential. As a result, scientists have been seeking other methods to gauge the amount of solar radiation in recent years. The use of satellite data has advanced in recent years due to its many benefits, including broad coverage level, repeatability, ease of data processing, and access to field data. It is also crucial to anticipate solar radiation using various methods. Since it is typically not feasible to assess directly. Different techniques are used to monitor radiation levels. These strategies use hybrids, regression relationships, experimental connections, and linear interpolation or neural networks for remote sensing. Solar radiation data is essential in many fields. The requirement for water for plants is one of several waters and soil processes that are impacted by solar radiation, a meteorological variable. However, as was already said, solar radiation measurements are sometimes quite limited due to the high expense of the equipment used to measure it. In addition, issues with the calibration of measuring instruments provide challenges in measuring solar radiation. This study's primary goal is to offer a neural network-based approach for predicting the amount of solar radiation. For this goal, many experimental and semi-experimental associations and the linear multiple regression approach have already been presented. However, this study will look at the high-accuracy usage of the neural technique. Future human energy must come from a sustainable, clean energy source that uses new technology with little harm to the environment. Solar energy will naturally be used in numerous technologies. Solar radiation is the most important natural energy source guiding all natural processes and reactions on Earth. Solar energy may be calculated using various techniques, such as artificial neural networks and regression analysis, experimental approaches, solar radiation measuring equipment, manual mathematics's computations, experimental and semi-experimental relationships, and more recent techniques. For at least one to two years, I gathered data and information from meteorological stations to be used as input to an artificial neural network. I develop a precise and ideal model to predict how much solar radiation will be present in a city, comparing the obtained model with other models and comparing the neural network method with other methods of predicting solar radiation, such as linear regression and experimental and semi-experimental methods, and finally evaluating the results. The ability of the neural network model and meteorological parameters to be effective in net radiation, as well as the estimation of daily net radiation in the hot and dry climate of Yazd city, were investigated in this study using 15 experimental and semi-experimental models. The accuracy of each model, as mentioned earlier with the measurement data by the net radiation logger at the meteorological stations in Yazd, was assessed over 36 months. Thus, the layout of 1-2-11 for the Yazd region represented 11 parameters in the input layer, two neurons in a hidden layer, and an output layer to produce the network's best structure. The outcomes demonstrated the neural network's strong performance in radiation prediction, with a coefficient of determination of 0.95 and error profiles for RMSE, MAE, FB, and MBE with values of 393, 850, 0.006, and 49 watts per square meter per day, respectively. The principles and generalizations of solar radiation, as well as strategies for forecasting solar radiation, are examined in this study. The sorts of in-country and international research in this area are discussed in the second chapter. The third chapter briefly overviews the research topic and details the data collection process. This part introduces research methodologies such as artificial neural networks, multivariate linear regression, and experimental and semi-experimental models. The fourth chapter calculates solar radiation using an artificial neural network and compares the findings to other methods. The solar radiation prediction algorithm was designed and implemented using the MATLAB program and its neural network simulator toolkit. The Irmak model was shown to be the best option in the Yazd region during the winter season. Due to its superior accuracy compared to other models, the basic regression model 1 is the best model for the spring season. In summer and autumn, the basic regression model 3 is chosen as the optimal model. Also, on the annual scale, the basic regression model 3 had higher accuracy in predicting net radiation and was chosen as the optimal model in the Yazd region.