Deep Learning–Driven Solar Radiation Estimation Using Meteorological Time-Series Data

Today, renewable energy sources are becoming increasingly important due to the continuous increase in energy demand and the limited availability of fossil fuel resources. Solar energy is becoming popular in many countries as a clean and sustainable energy source. Accurate estimation of solar radiation plays a critical role for efficient design, operation and management of solar energy systems. In this study, solar radiation estimation is performed using meteorological data of Gümüşhane province. In the estimation process, Multi Layer Perceptron (MLP) artificial neural network method, which has a multilayer structure from deep learning approaches, was used. In the MLP model, 5 inputs, a hidden layer with 14 neurons and 1 output layer are structured. Error criteria such as Mean Absolute Percentage Error (MAPE) and Mean Square Error (MSE) were used to evaluate the performance of the model. The results showed that the MAPE value was 1.3% and the MSE value was 0.026. In addition, regression analysis was performed to examine the relationship between model outputs and actual values and the coefficient of determination R2 was calculated. The results of the study are in line with the results reported in similar studies in the literature and reveal that the developed MLP model provides high accuracy in solar radiation prediction.