Short-term prediction of wind power density using convolutional LSTM network

Efficient extraction of renewable energy from wind depends on the reliable estimation characteristics and optimization farm installation operation conditions. There exists uncertainty in prediction tapping potential based variability behavior. Thus power density empirical models demand subsequent data processing to ensure accuracy reliability computations. Present study analyses ANN-based machine learning approach predicting for five stations (Chennai, Coimbatore, Madurai, Salem, Tirunelveli) state Tamil Nadu, India using different non-linear models. The selected such as Convolutional Neural Network (CNN), Dense (DNN), Recurrent (RNN), Bidirectional Long Short Term Memory (LSTM) Network, linear regression are employed comparing a period Jan 1980 May 2018. Based results, it was found that performance (1->Conv1D|2->LSTM|1-dense) is better than other estimating with minimum error values (based mean absolute root squared error).