Insolation prediction for reliable energy storage using deep learning algorithms

Renewable energy in the form of insolation, is abundantly available in India, and one can reap huge profits by provisioning smart energy storage systems based on solar photovoltaic energy. Government and energy utility authorities incentivize the individual as well as industries for harnessing solar energy through multiple initiatives. However, consistent long-term revenues can be gained only when there is good performance with reliability. Profitable deployment of solar photovoltaic panels cannot be realized to its best unless the reliability is at par with the competitors in this domain. Hence, deep learning models are very resourceful for reliably accurate prediction associated with the availability of solar radiation to enable the intelligence of the energy storage management system to take suitable rectification measures whenever insolation varies. This provides robustness and flexibility to the energy storage system. In this work, open-source dataset of weather in Jaipur has been utilized to train the regression model using 5 deep learning techniques, and these methods have been compared using parameters like root mean square error, mean absolute error, as well as the time complexity of the algorithms. The efficacy of regression algorithm has been represented by the residuals as well as deviation of the predicted value from the true value. The choice between the accuracy, indicated by the error metrics, and the time complexity, indicated by the training speed can be made judiciously depending on the utility of prediction.