Forecasting of Solar Radiation in India Using Various ANN Models

• Published in: 2018 5th IEEE Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON) pp. 1 - 6
• Main Authors: Srivastava, Rachit, Tiwari, A. N., Giri, V. K.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2018
• Subjects: Artificial Neural Network; Atmospheric modeling; Biological neural networks; Deep Learning Neural Network; Feeds; Forecasting; Model Averaged Neural Network; Predictive models; R software; Solar radiation
• DOI: 10.1109/UPCON.2018.8597170

Solar Radiation Forecasting plays a very important role for integration of solar power plant with conventional power plant. As it can predict how much power can be generated by any solar powered plant in next few days. For short time load management few day-a-head forecasting is required. In this paper, 6-day-a-head solar radiation forecasting has been done using various multivariate ANN models. For this, Feed forward Neural Network, Back Propagation Neural Network, Deep Learning Neural Network and Model Averaged Neural Network have been compared on the basis of various Statistical Indicators. One year data has been used for this analysis which has been collected from Solar Radiation Resource setup in Gorakhpur, India. Nine parameters namely Time, Average Temperature, Minimum Temperature, Maximum Temperature, Rain, Wind, Dew, Atmospheric and Azimuth have been selected as input variable to ANN. To accurately examine the models, models have been applied for January to December month forecasting. From the results, it has been found that Model Averaged Neural Network presents best results whereas Back Propagation Neural Network presents worst results.