Abstract
•The present study introduces a new Deep Recurrent Neural Network with a stacked Convolution Neural Network dropout paradigm for the WE prediction.•The proffered stacked framework and the integration of dropout layers remain beneficial for precision enhancement in the energy prediction paradigm.•The experimental result has been executed in Python software and the parameters used for analysis include MSE, RMSE.•Renewable energy sources remain one of the very important substitutes to the standard unrenewable energy generating systems.
Renewable energy sources remain one of the very important substitutes to the standard unrenewable energy generating systems. Amongst various replenish able power sources, the installed wind power ability presents nearly half of the whole ability. Nevertheless, the changeability and seasonality in wind speed, direction, atmospheric pressure, relative humidity, and precipitation result in wind power generation being greatly capricious. On this basis, this work intends to establish a wind speed prediction strategy employing Deep Recurrent Neural Network (DRNN) approaches. Single-step and multi-step DRNNs will be applied. The stacked Convolution Neural Network (sCNN) and rectified linear unity (ReLU) activation functions will be regarded in Mayfly Optimized Deep Recurrent Neural Network (MO_DRNN). It is observed that this attained 0.0062 of MSE, 0.0786 of RMSE, and 0.0651 of MAE with FS for testing data and 0.0138 of MSE, 0.1175 of RMSE, and 0.0877 of MAE with FS for training data. It has been noticed that the proffered MO_DRNN attains 0.0196 of MSE, 0.14 of RMSE, and 0.1033 of MAE devoid of the FS procedure.