Parameter Optimization Applied toUse ofAdaptive Blades on aVariable Speed WindTurbine

The value of adding adaptive blades to a variable speed turbine is studied by applying a general purpose optimization scheme to a baseline turbine configuration. The turbine is the AWT-26. Adaptive-blade effects involve elastic twist coupling tied to centrifugal and flap loadings. Two baseline variable speed approaches are examined. In the “aggressive” approach the speed control is assumed to limit power output perfectly at the rated level, producing the theoretical maximum energy. In the “conservative” approach, speed control is not used at all to limit power; the power is limited by passive stall regulation. The adaptive blades do not substantially improve the performance of the aggressive, theoretical maximum. However, adaptive blades with centrifugal coupling are shown to make up 45°/0of the difference between the aggressive and conservative baseline cases when applied to the conservative speed control case. Simple pitch adjustments can also compensate for the conservative speed control in low wind sites, but in moderate to high wind sites the adaptive blades achieve substantially more. Cases including simultaneous centrifugal and flap coupling could not be studied with the optimization procedure because they amount to an over parameterization of the problem; i.e., there are many local maxima in the system performance criterion. A parameter optimization study to improve average annual energy capture was conducted on a computational model of the two-bladed AWT-26 turbine. The thrusts of the study were to examine variable speed operation and adaptive blade measures. It is felt that present and future technology developments will make variable speed operation advantageous. Interest in adaptive blade measures has been fueled via recent studies that have shown marked improvements in system performance by coupling blade twist with rotor loads. Optimization was applied as a “wrapper” around the industry-standard PROP performance analysis code. PROP models wind turbine operation as a quasi-steady process combining geometry, twist, and aerodynamic data to compute force, moment, and generated power as a function of tip speed ratio. The original code was modified to accept rpm’s, local wind speed, and a blade pitch functional. Tip speed ratio was computed and power was extracted from PROP over a range of local wind speeds. These values were combined with a Rayleigh probability density function for a given average annual wind speed and integrated over local wind speed to compute an average annual power. The optimization problem was cast to maximize this average power metric for a given average annual wind speed while maintaining an upper limit on power over the local wind speeds. The Modified Method of Feasible Directions nonlinear programming technique was used to do the parameter optimization. The decision parameters were embedded in functional for rotor speed and blade pitch. Blade pitch adaption models include dependencies on a constant offset, current rpm value, and a nominal bending moment value. Two variable-speed operational modes were examined via optimization. The first was an “aggressive” mode allowing rpm’s to vary at will as a function of local wind speed. The second was a “conservative” mode in which rpm’s were allowed to increase with wind speed. In high winds, however, the speed is fixed at a speed that would allow stall regulation at a self-imposed 200-kilowatt power limit. Therefore, the speed increases linearly in low winds up to the speed at which the turbine will stall regulate at 200kW and then will operate at constant speed. Since this power limit is severe compared to the nominal turbine operation, the optimized conservative mode served as a baseline. Average power was maximized for 5,7, and 9 meterslsec average annual wind speeds while maintaining the upper power limit. The aggressive mode produced rotor speeds that were well over the nominal AWT-26 speed (57.1rpm’s) and a 7-10’%increase in average power over average wind speed range versus the conservative mode. The final polynomial functional for blade pitch included a constant offset and a quadratic dependence on rpm’s. Blade adaption was more effective in the conservative case, where rotor speed was constrained. An attempt to add bending-moment dependency into the blade pitch functional produced no gain, as it appeared to over-parameterize the problem. Comparison of the variable-speed and adaptive-blade approaches to increase average power show that their effects are intertwined. If one is constrained, the other can compensate. In the aggressive speed case, rotor speed was essentially unconstrained and blade adaption had minimal effect.