An Efficient Predictor of Renewable Energy Based on Deep Learning Technique (DGBM) and Multi-Objectives Optimization Function

This paper produces a predictor called Zero to Max Energy Predictor Model based on Deep Embedded Intelligence Techniques (ZME-DEI) to predict Dc-power which is the maximum energy generation from renewable resources that does not cause environmental pollution. ZME-DEI model consists of many stages which flow sequentially in stepwise style; the first stage presents collecting data from sensors of weather and solar plant (i.e., the sensors give a stream of data that contain multi-features). The second stage is preprocessing which contains multi-steps such as (a) Merging between two datasets. (b) Using correlation to the new dataset. (c) Splitting readings into intervals, and deleting duplication intervals. In the third stage, the ZME-DEI model is constructed based on adopting gradient boosting techniques through replacing its kernel (i.e., Decision Tree function) with multi parameters optimization functions. The stage begins with dividing the dataset into two sets using five cross-validation methods, the training dataset is used to construct the ZME-DEI model, while the testing dataset is for evaluating in the final stage. Finally, the fourth ZME-DEI stage is used for evaluation results of the testing dataset based on three measures such as coefficient of determination (R2), Root Mean Square Error (RMSE), and Accuracy(A). the results explain the robust of ZME-DEI and reduce the computation complexity while the best results get from the optimization objective function that based on three parameters Irritation, AC-Power and Temperature.