Prediction and Optimization of Wind Power Intelligent Performance Enhancement Systems Based on Big Data and Deep Learning

The suppression of edgewise vibration is critical for enhancing the operational safety and power generation efficiency of large-scale wind turbines. Conventional control strategies, such as PID, often lack the robustness required to handle the inherent model uncertainties and external disturbances in such complex systems. This paper proposes a novel application of an Active Disturbance Rejection Control (ADRC) framework for actively suppressing the edgewise vibration of wind turbine blades. A comprehensive physical model of the blade is first established, deriving the transfer function from actuator torque to edgewise displacement. A systematic frequency-domain analysis reveals the system's resonant characteristics, and an uncertainty analysis is conducted to quantify its sensitivity to parameter variations. The ADRC controller is then meticulously designed, leveraging its core capability to estimate and cancel the "total disturbance" in real-time. Through comparative simulations with a finely-tuned PID controller, the ADRC demonstrates superior performance, characterized by significantly smaller overshoot (36.66%), faster settling time, and enhanced resistance to load disturbances. The ADRC-controlled system settles down within 1.07 seconds while sustaining a 49.47° phase margin which provides sufficient stability protection. The root-mean-square (RMS) tracking error decreases to 0.015 under ongoing harmonic disturbances which shows a 63.4% improvement over the PID controller. The proposed method demonstrates better dynamic response capabilities and higher stability margins together with improved disturbance rejection performance according to these quantitative performance indicators. Crucially, the robustness of the proposed method is rigorously validated via extensive uncertainty analysis, showing that the closed-loop performance remains stable and effective despite ±10% parameter perturbations, whereas the PID control exhibits performance regression.