Optimized Convolutional Neural Network-Based Adaptive Controller for 6-Phase DFIG-Based Wind Energy Systems

Abstract

This article proposes an optimized convolutional neural network-based adaptive control scheme (OCAC) for DFIG-based wind energy conversion systems. While linear systems can function successfully with the help of a PI controller, the behavior of the system becomes unstable when physical variations are present, rendering the PI controller ineffective. The purpose of this research is to guarantee that the proposed OCAC acquires self-adaptation under all conditions. By accounting for crucial circumstances such as changes in wind speed, fluctuations in generator parameters, and asymmetrical grid faults, the efficiency of OCAC control is proven. The hyperparameters of the deep convolutional neural network are optimized using the flower pollination algorithm, which boosts the network’s speed and precision. In comparison to the non-optimized technique, an overall improvement of 6.3% in training accuracy was attained through the use of the optimized method. Furthermore, the suggested OCAC would forecast the next systemic state and update control strategies of DFIG-based wind energy systems in real-time. The effectiveness of the OCAC is assessed with five different state-of-the-art algorithms under two distinct test scenarios. The simulation was conducted using MATLAB software. A comparison with a PID controller showed that the total harmonic distortion of the grid current decreased by 16.57% and that of the generator current decreased by 12.07%.