A Genetic Algorithm-based Robust Control Approach for Wind Turbine System Test Benches

Testing wind turbines on system test benches is an important part in today's development process of turbine manufacturers as it is the only way to excite the whole physically present system on ground level. Due to the altered drivetrain characteristics resulting from the dismounted rotor, hardware-in-the-loop systems are used to guarantee stable operation with the turbine controller in the loop. As model-plant mismatches are common due to the plant's sensitivity to assembly discrepancies and temperature related effects, we need a robust controller to compensate for that. This work presents a procedure in which H
_∞
-synthesis is embedded in a genetic algorithm for weight optimization. The derived weights are then incorporated in a DK-synthesis subject to structured dynamic uncertainties which approximate the model's uncertain shifted natural frequencies and altered structural damping properties. The resulting controller is fitted by a stable, low order system to recover from complex-rational transfer functions. The procedure results in a robust performant controller of order 10 and a structured singular value of 0.94. Motivated by experimental experiences, we allow an uncertain eigenfrequency region of one third Hertz and a communication induced time delay of 30 ms. The controller is validated by means of a nonlinear test bench model, derived from experimental data using a least-squares initialization and a grey-box estimation.