Flow separation on wind turbines blades

Abstract

In the year 2000, 15GW of wind power was installed throughout the world, producing 100PJ of energy annually. This contributes to the total electricity demand by only 0.2%. Both the installed power and the generated energy are increasing by 30% per year world-wide. If the airflow over wind turbine blades could be controlled fully, the generation efficiency and thus the energy production would increase by 9%. Power Control To avoid damage to wind turbines, they are cut out above 10 Beaufort (25 m/s) on the wind speed scale. A turbine could be designed in such a way that it converts as much power as possible in all wind speeds, but then it would have to be to heavy. The high costs of such a design would not be compensated by the extra production in high winds, since such winds are rare. Therefore turbines usually reach maximum power at a much lower wind speed: the rated wind speed, which occurs at about 6 Beaufort (12.5 m/s). Above this rated speed, the power intake is kept constant by a control mechanism. Two different mechanisms are commonly used. Active pitch control, where the blades pitch to vane if the turbine maximum is exceeded or, passive stall control, where the power control is an implicit property of the rotor. Stall Control The flow over airfoils is called ‘attached’ when it flows over the surface from the leading edge to the trailing edge. However, when the angle of attack of the flow exceeds a certain critical angle, the flow does not reach the trailing edge, but leaves the surface at the separation line. Beyond this line the flow direction is reversed, i.e. it flows from the trailing edge backward to the separation line. A blade section extracts much less energy from the flow when it separates. This property is used for stall control. Stall controlled rotors always operate at a constant rotation speed. The angle of attack of the flow incident to the blades is determined by the blade speed and the wind speed. Since the latter is variable, it determines the angle of attack. The art of designing stall rotors is to make the separated area on the blades extend in such a way, that the extracted power remains precisely constant, independent of the wind speed, while the power in the wind at cut-out exceeds the maximum power of the turbine by a factor of 8. Since the stall behaviour is influenced by many parameters, this demand cannot be easily met. However, if it can be met, the advantage of stall control is its passive operation, which is reliable and cheap. Problem Definition In practical application, stall control is not very accurate and many stall-controlled turbines do not meet their specifications. Deviations of the design-power in the order of tens of percent are regular. In the nineties, the aerodynamic research on these deviations focussed on: profile aerodynamics, computational fluid dynamics, rotational effects on separation and pressure measurements on test turbines. However, this did not adequately solve the actual problems with stall turbines. In this thesis, we therefore formulated the following as the essential question: ‘Does the separated blade area really extend with the wind speed, as we predict?’ To find the answer a measurement technique was required, which 1) was applicable on large commercial wind turbines, 2) could follow the dynamic changes of the stall pattern, 3) was not influenced by the centrifugal force and 4) did not disturb the flow. Such a technique was not available, therefore we decided to develop it. Stall Flag Method For this method, a few hundred indicators are fixed to the rotor blades in a special pattern. These indicators, called ‘stall flags’ are patented by the Netherlands Energy Research Foundation (ECN). They have a retro-reflective area which, depending on the flow direction, is or is not covered. A powerful light source in the field up to 500m behind the turbine illuminates the swept rotor area. The uncovered reflectors reflect the light to the source, where a digital video camera records the dynamic stall patterns. The images are analysed by image processing software that we developed. The program extracts the stall pattern, the blade azimuth angles and the rotor speed from the stall flags. It also measures the yaw error and the wind speed from the optical signals of other sensors, which are recorded simultaneously. We subsequently characterise the statistical stall behaviour from the sequences of thousands of analysed images. For example, the delay in the stall angle by vortex generators can be measured within 1° of accuracy from the stall flag signals. Properties of the Stall Flag The new indicators are compared to the classic tufts. Stall flags are pressure driven while tufts are driven by frictional drag, which means that they have more drag. The self-excited motion of tufts, due to the Kelvin-Helmholtz instability, complicates the interpretation and gives more drag. We designed stall flags in such a way that this instability is avoided. An experiment with a 65cm diameter propeller confirms the independence of stall flags from the centrifugal force and that stall flags respond quickly to changes in the flow. We developed an optical model of the method to find an optimum set-up. With the present system, we can take measurements on turbines of all actual diameters. The stall flag responds to separated flow with an optical signal. The contrast of this signal exceeds that of tuft-signals by a factor of at least 1000. To detect the stall flag signal we need a factor of 25 fewer pixels of the CCD chip than is necessary for tufts. Stall flags applied on fast moving objects may show light tracks due to motion blur, which in fact yields even more information. In the case of tuft visualisations, even a slight motion blur is fatal. Principal Results In dealing with the fundamental theory of wind turbines, we found a new aspect of the conversion efficiency of a wind turbine, which also concerns the stall behaviour. Another new aspect concerns the effects of rotation on stall. By using the stall flag method, we were able to clear up two practical problems that seriously threatened the performance of stall turbines. These topics will be described briefly. 1. Inherent Heat Generation The classic result for an actuator disk representing a wind turbine is that the power extracted equals the kinetic power transferred. This is a consequence of disregarding the flow around the disk. When this flow is included, we need to introduce a heat generation term in the energy balance. This has the practical consequence that an actuator disk at the Lanchester-Betz limit transfers 50% more kinetic energy than it extracts. This surplus is dissipated in heat. Using this new argument, together with a classic argument on induction, we see no reason to introduce the concept of edge-forces on the tips of the rotor blades (Van Kuik, 1991). We rather recommend following the ideas of Lanchester (1915) on the edge of the actuator disk and on the wind speed at the disc. We analyse the concept induction, and show that correcting for the aspect ratio, for induced drag and application of Blade Element Momentum Theory all have the same significance for a wind turbine. Such corrections are sometimes made twice (Viterna & Corrigan, 1981). 2. Rotational Effects on Flow Separation In designing wind turbine rotors, one uses the aerodynamic characteristics measured in the wind tunnel on fixed aerodynamic profiles. These characteristics are corrected for the effects of rotation and subsequently used for wind turbine rotors. Such a correction was developed by Snel (1990-1999). This correction is based on boundary layer theory, the validity of which we question in regard to separated flow. We estimated the effects of rotation on flow separation by arguing that the separation layer is thick so the velocity gradients are small and viscosity can be neglected. We add the argument that the chord-wise speed and its derivative normal to the wall is zero at the separation line, which causes the terms with the chord-wise speed or accelerations to disappear. The conclusion is that the chord-wise pressure gradient balances the Coriolis force. By doing so we obtain a simple set of equations that can be solved analytically. Subsequently, our model predicts that the convective term with the radial velocity (vrvr/r) is dominant in the equation for the r-direction, precisely the term that was neglected in Snel’s analysis. 3. Multiple Power Levels Several large commercial wind turbines demonstrate drops in maximum power levels up to 45%, under apparently equal conditions. Earlier studies attempting to explain this effect by technical malfunctioning, aerodynamic instabilities and blade contamination effects estimated with computational fluid dynamics, have not yet yielded a plausible result. We formulated many hypotheses, three of which were useful. By taking stall flag measurements and making two other crucial experiments, we could confirm one of those three hypotheses: the insect hypothesis. Insects only fly in low wind, impacting upon the blades at specific locations. In these conditions, the insectual remains are located at positions where roughness has little influence on the profile performance, so that the power is not affected. In high winds however, the flow around the blades has changed. As a result, the positions at which the insects have impacted at low winds are very sensitive to contamination. So the contamination level changes at low wind when insects fly and this level determines the power in high winds when insects do not fly. As a consequence we get discrete power levels in high winds. The other two hypotheses, which did not cause the multiple power levels for the case we studied, gave rise to two new insights. First, we expect the power to depend on the wind direction at sites where the shape of the terrain concentrates the wind. In this case the power level of all turbine types, including pitch regulated ones, will be affected. Second, we infer heuristically that the stalled area on wind turbine blades will adapt continuously to wind variations. Therefore, the occurrence of strong bi-stable stall-hysteresis, which most blade sections demonstrate in the wind tunnel, is lost. This has been confirmed by taking special stall flag measurements. 4. Deviation of Specifications The maximum power of stall controlled wind turbines often shows large systematic deviations from the design. We took stall flag measurements on a rotor, the maximum power of which was 30% too high, so that the turbine had to be cut out far below the designed cut-out wind speed. We immediately observed the blade areas with deviating stall behaviour. Some areas that should have stalled did not and caused the excessive power. We adapted those areas by shifting the vortex generators. In this way we obtained a power curve that met the design much more closely and we realised a production increase of 8%.