Optimal control theory applied to 5MW wind turbine model for region 3 control using minimal sensors

Wind turbines are complicated, fragile and non-linear systems that require carefully designed control systems to optimize power captured in region 2 (below rated wind speed) and regulate power captured in region 3 (above rated wind speed). With wind turbines increasing in size from hundreds of kilowatts capacity in the past to few megawatts capacities now, the structures will have low frequency lightly damped modes that will lead to larger vibrations in the structure and over time lead to failure due to fatigue. Hence the goal in region 3 is not only rotor rate regulation but also load mitigation for the drive train and wind tower.;This work discusses the design and simulation of optimal controllers for rotor rate regulation and load mitigation in region 3. The controllers designed uses a small sensor suite (only two). Having a larger sensor suite not only adds cost, but is also susceptible to failures. In this work, the focus is on quantifying the performance that can be achieved using two sensors, high speed shaft rate and tower fore-aft acceleration. Full state information case, Linear Quadratic Regulator (LQR), and state estimation cases, Linear Quadratic Gaussian (LQG) and Linear Quadratic Gaussian - Hybrid (LQG-H), control algorithms are discussed. The plant based on which the controllers are designed is the 5MW Offshore Wind Turbine Model obtained from National Renewable Energy Laboratory (NREL). In the model, lack of observability of the states compelled the development of hybrid solution consisting of state-space and frequency domain controller (LQG-H). Comparison between optimal controllers and frequency domain controllers, baseline PI - Gain Schedule controller (PI-GS), Fixed Gain Two Output controller (FGTO) and Variable Gain Two Output (VGTO) controller, are provided. The results obtained from the controllers using only two measurements are indeed favorable, and indicates a hybrid controller like the LQG-H provides good performance. The LQG-H controller reduces the normalized standard deviation in the rotor rate by 2%, power spectral density of the rotor rate at drivetrain natural frequency by more than 10dB and standard deviation in the tower fore-aft acceleration by 40% compared to the baseline PI-GS controller.