A robust De-Noising Autoencoder imputation and VMD algorithm based deep learning technique for short-term wind speed prediction ensuring cyber resilience

Abstract

The intermittent and stochastic characteristics of wind speed make its predictions difficult. Further, forecasting systems are dependent on the communication network for coordination and control that are far more susceptible to cyber-attacks. In order to ensure system reliability, safety & cyber resilience, different situations that may arise as a result of cyber-attacks like unexpected device power outages, sensor or communication failures, etc. which may further lead to corrupted data/outliers in datasets/missing of actual values in the datasets required for forecasting, are investigated and addressed here with the help of De-Noising Autoencoder (DAE) algorithm. Afterwards, with the aim to increase the forecast accuracy, Variational mode decomposition (VMD) algorithm is coupled with the Bidirectional long short term memory (BiDLSTM) deep learning algorithm. In order to evaluate the precision of suggested forecasting approach, this research scrutinizes number of datasets related to wind speed for two Indian cities i.e. Sadodar and Bhogat, located in State of Gujarat. In comparison to other hybrid decomposition-based forecasting algorithms presented in the literature, empirical findings acquired using the suggested hybrid DAE-VMD-BiDLSTM approach is found to be more significant and accurate.