Adaptive Artificial Neural Network-Based Proportional Integral Controllers and Extremum Seeking Energy Optimizer for Wind Systems

Abstract

Proportional-Integral (PI) controllers are widely used in various industrial applications due to their simplicity, affordability and performance in linear system control. However, their effectiveness deacreases when applied to nonlinear systems, such as Wind Energy Systems (WES) based on Doubly-Fed Induction (DFIG). Indeed, the WES operates in a complex and highly disturbed environment, its model is non-linear, multivariable and with varying parameters depending on the operating conditions. To address these issues, this paper presents an adaptive Artificial Neural Network(ANN)-based PI controller for active and reactive power for WES. The proposed controller consists of two stages. The first stage is made up of an ANN performing the online identification of the WES, while the second stage adjusts the PI parameters in real time. For the WES optimization, the search for the Maximum Power Point (MPP) is ensured by the Extremum Seeking (ES) technique. This non-based model strategy provides flexibility and robustness in the presence of uncertainties in the WES parameters or without knowledge of the wind turbine characteristics.Simulations in MATLAB/Simulink show that the proposed controller significantly outperforms both the conventional PI controller and the Model Predictive Controller with Particle Swarm Optimization (MPC-PSO) in terms of accuracy, quality and robustness under various wind energy operating conditions. It shows significant improvements in active and reactive power control and robustness, significantly reducing errors metrics compared to the PI and MPC-PSO controllers.

Keywords

• Adaption control
• robustness
• correlation matrix
• active power
• reactive power
• wind energy