Solar Tracking System by Utilized Optimized Algorithm Based Deep Learning

Solar tracking system is significant for most PV solar power systems in order to enhance the power production. In this study, a dual-axis solar tracking system-based solar tracking algorithm has been designed and implemented. However, several issues still need to be overcome for using a tracker system that includes power consumption related to the electrical part of the tracker and the estimation of the best path based on tracking technology that is utilized. The solar tracker predication algorithm can predicate the best solar path for each day in the year based on location. For this purpose, Deep Recurrent Neural Network (D-RNN) with Long-Term Short-Term units (LSTM) has been implemented to predict the path that the tracker should move with time for all the years. “Adam” optimizer has also been used in order to update network weights iteratively based on the training data. The proposed model achieved more accuracy than slandered algorithm, in which the average RMSE of proposed algorithm is 0.0512 in the training phase and 0.067 in the testing phase, while the average RMSE (Root Mean Square Error) of the traditional machine learning model is 0.0746 in training phase and 0.086 in testing phase.