Application of Artificial Neural Network to Solar Potential Estimation in Hilly Region of India

Abstract

The use of these conventional resources causes continuous depletion of fossil fuels and increased greenhouse effect. Solar power is the major renewable resource used for power generation across the globe. Solar energy activities depend on the available potential of any geographical location. Therefore, prior to the installation of solar technologies for these activities, estimation of solar potential is very important due to costly technologies. Data of solar potential is not present at every location in Himachal Pradesh (H. P.) due to the high cost of measurement instruments. The objective of this study includes the solar potential estimation for 12 cities of the H. P. The present study could be divided into two parts. Initially, Artificial Neural Networks (ANNs) are utilized to estimate global sun radiation utilizing meteorological and geographical data from 23 places. The ANN model with seven input parameters including latitude, longitude, altitude, air temperature, humidity, pressure, and wind speed were used to estimate the solar irradiation. Statistical indicators including Mean Absolute Percentage Error (MAPE) were used for the performance evaluation of these ANNs. The minimum MAPE value was obtained to be 2.39 % with Multi-Layer Perception (MLP) architecture 7-11-1. For the 12 districts of the H. P., the acquired network 7-11-1 was utilized to estimate Global Solar Radiation (GSR). The output of ANN model was implemented in Geographic Information System (GIS) environment to obtain the solar potential map for each month. The available map of the present study may be helpful for solar application in each district.

Keywords

• artificial neural networks (ann)
• global sun radiation
• solar
• mean absolute percentage error (mape)