A deep learning assisted adaptive nonlinear deloading strategy for wind turbine generator integrated with an interconnected power system for enhanced load frequency control

Abstract

The existing linear and quadratic deloading strategies with constant deloading factor, fail to effectively handle the nonlinear characteristics of WTGs. This work proposes a novel deep learning assisted adaptive nonlinear deloading (DL-AND) methodology based on a Newtonian interpolated polynomial for WTG integrated with an interconnected power system to provide effective load frequency control (LFC). The key feature of the proposed technique is its ability to adapt the deloading factor in accordance with wind speed to optimize the reserve power margin of the WTG. In this work, a deep learning-based recurrent neural network (RNN) with long short-term memory (LSTM) technique has been proposed for wind speed forecasting, as using a wind speed measurement device is expected to incorporate measurement lag, leading to the deterioration of the deloading operation. The proposed novel DL-AND technique for WTGs is used along with a fractional-order fuzzy-based PID (FFOPID) control structure as a supplementary controller for handling uncertainties in order to provide effective LFC. Further, Exhaustive simulation studies have been carried out to investigate the proposed technique and results show the effectiveness of proposed novel DL-AND strategy with FFOPID in terms WTG reserve power margin, frequency support and performance index for all the considered case studies.