Active control, by complementing the primary technologies in wind turbines, can help significantly in attaining—perhaps even improving—the industry's objectives of safe and cost-efficient energy production. For example, although the need to understand and improve material properties will continue, active control can alleviate the requirements on materials at any stage of the technology. Discussed here are the control requirements for wind turbines, together with the control means and sensors typically employed in modern wind turbines. This will provide background and serve as a focal point for subsequent discussion of the applicability of modern control theory to wind turbines.
The wind turbine control problem has at least three important requirements:
setting upper bounds on and limiting the torque and power experienced by the drive train, principally the low-speed shaft;
minimizing the fatigue life extraction from the rotor drive train and other structural components due to changes in wind direction, speed (including gusts), and turbulence, as well as start-stop cycles of the wind turbine; and
maximizing the energy production.
The control problem is the judicious balancing of these requirements.
The control theory discussion here will be cast in terms of a pitch-controlled, variable-speed wind turbine. The discussion is general in the sense that the combination of sensors and control means is thought to span the space of current, practical control techniques. We recognize that stall-controlled and other wind turbines may not require all of the sensors and control means discussed. To the extent this is true, these architectures represent simpler cases. However, for all wind turbine architectures, research may reveal other analogous techniques for control that are functionally equivalent to the particular techniques listed here. In any case, there will remain the general requirements for control of the torque and power in the drive train, the minimization of fatigue life extraction, and the maximization of energy production.
The control effectors (actuators) and commands, together with the sensors and the quantities they measure for a pitch-controlled, variable-speed wind turbine, are summarized in Table 6-1. By control effectors we mean the physical devices that implement angular or linear motion of a wind turbine control surface or component, examples of which are electric motors and hydraulic pistons. The control effectors (actuators) execute motion in response to commands from the control computer. The control computer executes the algorithms derived from the application of control theory. The algorithms, in turn, utilize information from the wind turbine sensors to generate the commands and assess their effects.
The control effectors listed in Table 6-1 include a contactor, nacelle yaw drive, and blade pitch actuator. These are listed along with typical command inputs. The contactor is an electrical relay that connects and disconnects the generator to the load. As shown in Figure 6-1, the generator
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology 6 ACTIVE CONTROL IN WIND TURBINES THE CONTROL PROBLEM FOR WIND TURBINES Active control, by complementing the primary technologies in wind turbines, can help significantly in attaining—perhaps even improving—the industry's objectives of safe and cost-efficient energy production. For example, although the need to understand and improve material properties will continue, active control can alleviate the requirements on materials at any stage of the technology. Discussed here are the control requirements for wind turbines, together with the control means and sensors typically employed in modern wind turbines. This will provide background and serve as a focal point for subsequent discussion of the applicability of modern control theory to wind turbines. The wind turbine control problem has at least three important requirements: setting upper bounds on and limiting the torque and power experienced by the drive train, principally the low-speed shaft; minimizing the fatigue life extraction from the rotor drive train and other structural components due to changes in wind direction, speed (including gusts), and turbulence, as well as start-stop cycles of the wind turbine; and maximizing the energy production. The control problem is the judicious balancing of these requirements. The control theory discussion here will be cast in terms of a pitch-controlled, variable-speed wind turbine. The discussion is general in the sense that the combination of sensors and control means is thought to span the space of current, practical control techniques. We recognize that stall-controlled and other wind turbines may not require all of the sensors and control means discussed. To the extent this is true, these architectures represent simpler cases. However, for all wind turbine architectures, research may reveal other analogous techniques for control that are functionally equivalent to the particular techniques listed here. In any case, there will remain the general requirements for control of the torque and power in the drive train, the minimization of fatigue life extraction, and the maximization of energy production. The control effectors (actuators) and commands, together with the sensors and the quantities they measure for a pitch-controlled, variable-speed wind turbine, are summarized in Table 6-1. By control effectors we mean the physical devices that implement angular or linear motion of a wind turbine control surface or component, examples of which are electric motors and hydraulic pistons. The control effectors (actuators) execute motion in response to commands from the control computer. The control computer executes the algorithms derived from the application of control theory. The algorithms, in turn, utilize information from the wind turbine sensors to generate the commands and assess their effects. The control effectors listed in Table 6-1 include a contactor, nacelle yaw drive, and blade pitch actuator. These are listed along with typical command inputs. The contactor is an electrical relay that connects and disconnects the generator to the load. As shown in Figure 6-1, the generator
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology TABLE 6-1 Control Effectors and Sensors for a Pitch-Controlled Wind Turbine Control Effectors and COMMANDS Sensors and MEASURED PARAMETERS Generator contactor Contactor status ON ON OFF OFF Nacelle yaw drive Nacelle orientation ON ANGLE OFF DIRECTION Wind direction RATE ANGLE END POINT Wind speed Blade pitch actuator SPEED ON OFF Blade pitch angle DIRECTION ANGLE RATE END POINT Generator power REAL POWER Generator torque REACTIVE POWER TORQUE ANGLE Reactive power SPEED REACTIVE POWER Figure 6-1 Wind turbine drive train.
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology load is the input switch matrix of the power electronics. For a constant-speed wind turbine—one without power electronics—the generator load is the utility grid. The nacelle yaw drive may be an electric or hydraulic motor acting through gearing to rotate the nacelle and rotor. The function of the yaw drive is to orient the wind turbine relative to the prevailing wind direction. The blade pitch actuator effects the rotation of the rotor blades about their pitch axis. With the blade pitch angle set at the full-power angle, maximum power is extracted from the incident wind flow field. As the blade pitch angle is rotated toward the full feather position, the blades become less efficient at converting the power in the wind flow field to shaft power. Also listed in the first column of Table 6-1 are two controls that are effected by the power electronics. The generator torque may be controlled by the power electronics acting on the electrical characteristics of the generator. The power electronics, acting as the load for the generator, can electronically vary the load and thus the drive train torque. The reactive power delivered to or received from the utility grid may also be controlled by the power electronics. This is important for maintenance of one of the desired electrical characteristics (power factor) of the power delivered to the utility system. The second column of Table 6-1 lists the typical sensor complement associated with a wind turbine. Given also is the physical information measured by these sensors. The contactor status sensor indicates whether the contactor is open or closed, that is, whether the generator is connected to the load. The nacelle orientation sensor measures the angular position of the nacelle, either relative to a fixed reference or to the wind direction. The wind direction sensor measures the angular direction from which the wind blows. The wind speed sensor measures the wind speed. This sensor may be shared by a number of wind turbines or may be absent altogether. If absent, wind speed may be derived from knowledge of the blade airfoil characteristics, the power level, and the drive train rotational speed. The blade pitch angle sensor measures the pitch angle of the blades. The generator power sensor measures the real power flow into, or delivered by, the generator. Typically, the reactive power flow is also measured. The generator speed sensor measures the generator rotational speed and, through knowledge of the gearbox ratio, the speed of the low-speed shaft. Of importance to control are two quantities that may be estimated from the sensor measurements. These are the wind speed and the torque in the low-speed shaft. Estimation of the wind speed value was discussed above. Estimation of the torque value proceeds from knowledge of the power, the rotational speed, and the gearbox ratio. RECENT TRENDS IN CONTROL SYSTEM THEORY The 1980s witnessed tremendous strides in the development of theory and algorithms applicable to the design of dynamic compensators for both single-input single-output (SISO) and multiple-input multiple-output (MIMO) systems under different sets of assumptions about the process to be controlled and with a variety of performance/robustness criteria satisfied. In the area of linear time invariant (LTI) systems with parameters assumed to be known, perhaps the greatest breakthrough in the early phase was the unification of frequency and time domains so as to allow for the design of compensators that satisfy frequency domain specifications and robustness constraints using time domain synthesis tools (i.e., solutions to Riccati equations). Clearly, the advantage of using time domain mathematics, which is finite dimensional and numerically robust, has greatly facilitated control system designs for arbitrary plant dimensions both for SISO and MIMO systems. In the H2 framework, a two-norm formulation was adopted, admitting the interpretation that root mean square (rms) tracking errors are to be minimized in a fixed spectrum, flat power spectral density (PSD), disturbance environment. Other measures of performance may be desirable, such as minimizing the worst possible error (in the frequency domain), captured by an infinity norm (H)
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology minimization, which may be important for some systems/operating conditions. Certainly, in the wind turbine environment, a reasonable designer's objective would be to minimize the cost of energy. Furthermore, if performance specifications are expressed in the time domain (as, for example, in terms of an upper bound on the system impulse response), the absolute integral norm is the appropriate measure (Dahleh and Pearson, 1987). Each design method is geared toward a particular aspect of system performance as expressed by the corresponding norm, and it is conceivable that simultaneous suboptimization of all three norms could be achieved, by appropriately modifying/combining the solutions to each individual norm optimization. Results along these lines are presently becoming available (Bernstein and Haddad, 1989; Doyle et al., 1989). However, in processes where parametric variation drastically affects their nominal characteristics, serious performance degradation or even instability can occur. The need for design under parametric uncertainty/ (slow) time variation gave rise to the area of adaptive control, which allows for real-time, on-line ''learning'' and subsequent compensator parameter adjustment for performance, while maintaining overall system stability. This class of algorithms is very distinct from "gain-scheduled" control algorithms, which are quite often perceived and referred to as adaptive. Gain-scheduled controllers change the compensator characteristics as the operating conditions change, in an open-loop fashion, from table look-ups, based on the scheduling parameter(s). On the other hand, truly adaptive algorithms are real-time learning to update compensator parameters in a feedback fashion. Impressive theoretical advances were achieved, by the late 1970s, for adaptive algorithms designed to operate with the assumption of no modeling (unstructured) errors. However, in the early 1980s such algorithms were shown to be nonrobust in the presence of unmodeled dynamics and persistent disturbances (Rohrs et al., 1985). Subsequently, various methods were developed for robustification, with a newly evolved notion for robust adaptive control, which encompasses robust identification with robust control (Middleton et al., 1988). A systematic study of the former has resulted in algorithms that identify the processes with guaranteed frequency domain error bounds but at the expense of excessive—but not prohibitive, nowadays—computational burden that requires real-time spectral monitoring of pertinent signals in the feedback loop (LaMaire et al., 1991). The combination of robust identification with a robust controller structure as discussed above, with only infrequent compensator parameter update, as indicated by the quality of the identification process information, has resulted in the study of performance/stability robustness of time-varying systems, in particular as they arise in this context. A number of results only recently became available (Dahleh et al., 1990; Meyer, 1988). At the same time, analysis and design methods have been developed for nonlinear systems via the use of appropriate transformations for "nonlinear inversion" and feedback (input-output) linearization. In conclusion, the current state of the art in control theory provides the designer with a rich environment of design modules/algorithms that can be ingeniously combined to meet large classes of design challenges. The wind turbine rotor certainly belongs to the class of problems that can be readily addressed by the existing state of the art, without the need for new theoretical developments. What is needed in the present problem is a good modeling effort of the wind turbine/tower combination as well as the environment within which it operates and by which it is affected. EXISTING CONTROL TECHNOLOGY FOR WIND TURBINES As discussed earlier in this report, wind turbines are classified into two major categories: vertical axis and horizontal axis machines. There are two general modes of operation: constant speed (HAWT) and variable speed; the latter is under active investigation in the United States and Europe. The existing machines have been operating in either of the two modes with an industry-projected life-cycle cost of energy (LCCOE) of approximately 5/kWh. At the present time, however, the cost remains in the range of 7 to 9/kWh.
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology But with the cost of power electronics decreasing, the variable-speed mode may be an attractive mode of operation for such machines in the future. In order for the LCCOE to be kept sufficiently low for wind power to be a competitive energy alternative, the lifetime of a machine would extend over a period of up to 30 years and should be operating at a maximum of "economic" performance. This presupposes that fatigue during "normal" operation is minimized while appropriate measures are taken to avoid accidents (e.g., blade breakage, tower hits) during abrupt environmental changes. Assuming that the structural, material, mechanical, and manufacturing aspects of the process are sufficient, additional benefits can be gained by appropriately controlling the blade for minimization of wind-loading effects and gust alleviation as well as maximizing efficiency of power production under variable wind speed profiles. Aerodynamic and mechanical solutions in the form of stall-controlled blades that limit CLmas, teetered hub versus rigid hub, and optimum choices of materials have been successful as one method of addressing problems cited above. In addition, other less complex, potentially less costly control schemes are under development, including passive control schemes. However, the full achievable potential offered by active control for improvement in the overall performance of wind turbines has not yet been fully exploited. The existing active (feedback) control strategies known to the committee have for the most part addressed the issue of energy production (McNerney, 1981; Vachon, 1987; Ralph, 1989), although Vachon (1987) also addresses fatigue issues primarily for vertical axis machines. Others are deemed proprietary and were not available to the committee. The designed controllers referred to above are of the proportional integral (P-I) type and are rather ad hoc. Typically, P-I action is taken on the basis of wind speed information—averaged over sampling windows of various lengths—in conjunction with the turbine power curve. The latter provides the equivalent of a command input to the overall system for desired performance, usually in terms of kilowatt-hours. At Sandia National Laboratories, different algorithms have been tested for evaluation in a controlled simulation environment. These are discrete wind speed; moving wind speed; moving power; and discrete, double-power. These four algorithms have two adjustable parameters for optimum operation: cut-in/cut-off threshold (of wind speed) and test window (for averaging). As they stand, no real safeguards for robustness are built in to the algorithms, particularly in response to severe wind gusts. This is so because the system is set to respond to averaged time information of wind speed profiles, in which high-frequency information is suppressed; certainly this is true for the fast wind gust characteristic times. Thus, although statistical predictive capability exists, application of a modern control design can further help anticipate or, at a minimum, initiate a fast system response on an individual system basis so as to alleviate a severe wind profile that may be building up. Thus, alleviating wind-loading effects under nominal operation can be more systematically addressed in real time for each specific wind turbine. ROLE OF CONTROL TECHNOLOGY IN THE WIND POWER INDUSTRY A systematic control approach to the overall system operation can favorably impact the economics of wind power systems while assuring safe operation with a minimal number of system failures. Within the wind industry, there is no general consensus as to whether active versus passive control is preferable. Both are employed at various levels of sophistication; however, the committee is not aware of any systematic study that has quantified the benefits of either. All available advances in structural tailoring, advanced airfoil design, and advanced machine configurations need to be applied for passive load alleviation. Simultaneously, active control techniques should be employed to assess the performance capabilities of an existing structure while also establishing achievable limits and indicating modifications necessary to realize further benefits. Thus, a parallel development of both strategies is advisable in the design stage, with a number of iterations before a final overall system design is arrived at. Such a process will reduce the
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology performance requirements for control and have, as a by-product, a system operational fault tolerance resulting from simpler control functions and implementation; moreover, it will be cost-effective in the long run. Consequently, active and passive control strategies are not mutually exclusive, but, when both are synergistically employed, they are capable of enhancing the overall system capabilities by alleviating the requirements on either or both component strategies. Thus, actuator loads in an active setting could be greatly alleviated if operating in a passively tailored structure. Suffice it to say that neither approach alone is capable of delivering performance equivalent to that obtainable by the synergy of both. A passive approach, on the one hand, is effected through a judicious hardware/material design and thus may not be adaptable to changing requirements or dramatically changing operating conditions; "passive tailoring" of the system cannot be done on-line, if the thresholds of a passive design have been exceeded. Moreover, passively designed structures are not easily modifiable to changing requirements/forcing functions with any reliable guarantees for consistent performance. Furthermore, the design margins of a passive control structure in response to dynamic exogenous inputs can be severely limited in comparison with an active control computer-implemented strategy that is easily modifiable as needed. On the other hand, an active controller may be considerably limited by state-of-the-art actuator/sensor bandwidths, which may constitute an "unbreakable barrier" to system performance. Certainly, weight, cost, and compatibility with an existing system also factor in but to a more manageable extent (i.e., provided the system setup/architecture is modifiable for tolerance of such factors) once satisfactory performance can be assured by the components in question. Ultimately, this is an area where trade-offs have to be systematically worked out and weighed. Before any detailed study of relative benefits can proceed, however, developing a good model of the overall process intended for control is of utmost importance. This seems to be the crux of a successful engineering endeavor, particularly in the wind turbine environment, which involves interaction of fluids (possibly unsteady aerodynamics) with electromechanical structures. For the purposes of alleviating wind gust effects and/or undesirable vibrations due to the interaction of the blade and unsteady aerodynamics (turbulence increases with deeper location in the wind farm), mass flow or pressure sensors need to be placed at judiciously chosen locations (perhaps even in the blade), in order to measure (precursors to) turbulent upcoming events (i.e., wind gusts). Such sensor bandwidth should be extremely high given the nature of the latter. The sensor signal can be used to generate an appropriate actuator command, via a microprocessor-implemented control algorithm. Actuation can be effected through incorporation of aileron-like surfaces on the main blade structure. Thus, fast response can be achieved locally, with very limited control actions, which will be effective in modifying the boundary layer of the unsteady fluid process, thus alleviating excessive loading on the blade. On a more global actuation scale, control of variable-speed machines through their power electronics effected via the rotor does appear to have the potential of tailoring the blade structure to operating in close to optimal configuration in adverse wind profiles. Based on experience with higher harmonic control (HHC) for helicopter vibrations (Goldenthal et al., 1987) and, more recently, on active control of rotating stall and surge in compressors (Paduano et al., 1990), such models of the fluid processes are readily obtainable. Sensor data of mass flow perturbations (velocity, pressure) can be used while the perturbations are small (i.e., the nonlinear [unsteady] process is still at its inception and it evolves in a close to linear fashion for a short period of time). Control designs based on models thusly obtained have proven successful for HHC and for active surge and rotating stall control. Experience gained there can be brought to bear in the wind turbine environment, given the basic similarities of the processes involved. In fact, the dynamic models obtained for the fluid processes in the areas mentioned above are as simple as low-order, linear, time-invariant (even scalar) systems that admit to the simplest control algorithm synthesis: fixed
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology parameter compensation that can be derived via the simplest theory, even classical approaches. Certainly, an ingeniously designed lead/lag network will work effectively, whether derived via classical considerations or, more formally, via a modern methodology such as the linear quadratic Gaussian with loop transfer recovery. Computational and implementation/hardware requirements are within existing state-of-the-art capabilities, and, as already stated, no new control theory is necessary for a successful outcome. The emphasis here is primarily on a good modeling effort. However, the literature is severely lacking in complete methodologies for systematic modeling of the effect on the machine of such fluid processes. Given its importance, this is an area that needs to be mathematically formulated and formalized. The foregoing suggestions are immediately implementable as a technology transfer from other related industries (Paduano et al., 1990; Goldenthal et al., 1987). To be precise, the proof of concept in a realistic engine (laboratory) setting was so recent that industrial production can only be in the very near future and is certainly not in existence at the present time. However, implementation details in an industrial setting need to be worked out specifically within the particular application industry at hand. Further down the line, for future development, smart materials (e.g., piezoelectric sensors and actuators embedded in them) may result in similar performance/operation improvements for wind turbines (Spangler and Hall, 1990). In such a case, an essential continuum (or distributed network) of sensors and actuators results in appropriate blade twisting so as to alleviate the effect of the aforementioned undesirable wind disturbances. This technology is under development, actively being researched at this time in large space structures groups and materials science laboratories. Preliminary results are encouraging (deLuis et al., 1989), although the problem of forcing such a distributed structure into a desirable coordinated configuration via local sensor feedback entails complexities and performance robustness risks. This is an area that requires further investigation. One should also note that the existing preliminary results in large space structures rely on the overall passivity of a flexible structure. However, the wind turbine environment is considerably more severe. In conclusion, a systematic application of readily available control technology can offer substantial advantages in enabling the wind power industry to realize its long-term objectives and become a competitive energy alternative. More specifically, the following are recommended control technology areas for research and development in the immediate (near) term: Continue investigating viable models for the fluid structure interaction, together with signal-processing algorithms for valid signal extraction. Develop other active control techniques and incorporate the best response means (e.g., aileron surfaces and boundary layer control). Design control algorithms for wind loading/gust alleviation. Develop models and algorithms for control of variable-speed machines to optimize energy production versus fatigue life.
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology REFERENCES AND BIBLIOGRAPHY Bernstein, D. S., and W. M. Haddad. 1989. LQG Control with an H Performance Bound. IEEE Trans A-C, Vol. 34, No. 3, February, pp. 293-305. Dahleh, M. A., and J. B. Pearson. 1987. l1 Optimal Feedback Controllers for MIMO Discrete-Time Systems. IEEE Trans A-C, Vol. AC-32, April. Dahleh, M. A., P. Voulgaris, and L. Valavani. 1990. Optimal Rejection of Bounded Persistent Disturbances in Periodic Systems. Proceedings of the Decision and Control Conference, Honolulu, Hawaii, December. deLuis, J., E. F. Crawley, and S. R. Hall. 1989. Design and Implementation of Optimal Controllers for Intelligent Structures Using Infinite Order Structural Models. SSL #3-89, MIT Space Systems Laboratory, Cambridge, Massachusetts, January. Doyle, J. C. 1985. Structured Uncertainty in Control Design. Proceedings of the 24th Conference on Decision and Control, Ft. Lauderdale, Florida, December, pp. 260-265. Doyle, J. C., K. Glover, P. Khargonekar, and B. Francis. 1988. State Space Solutions to Standard H2 and H Control Problems. Proceedings of American Control Conference, Atlanta, Georgia, June, pp. 1691-1896. Doyle, J. C., K. Zhou, and B. Bodenheimer. 1989. Optimal Control with Mixed H2 and H Performance Objectives. Proceedings of the American Control Conference, Pittsburgh, Pennsylvania, June, pp. 2065-2070. Goldenthal, W., V. Valavani, P. Motyka, S. R. Hall, and B. Eberman. 1987. Development and Evaluation of a Dynamic Model Active Vibration Control Concept. C. S. Draper Laboratory Technical Report, March. LaMaire, R. O., V. Valavani, M. Athans, and G. Stein. 1991. Robust Time and Frequency Domain Estimation Methods in Adaptive Control. Automatica, January, pp. 23-39. McNerney, G. M. 1981. Vertical Axis Wind Turbine Control Strategy. SAND 81-1156, SANDIA National Laboratories Report, August. Meyer, D. G. 1988. Shift-Invariant Equivalents for a New Class of Shift-Varying Operators with Applications to Multi-Rate Digital Control. Proceedings of the 27th Conference on Decision and Control, Austin, Texas, December, pp. 1697-1701. Middleton, R., G. C. Goodwin, and D. Mayne. 1988. Adaptive Robust Control. IEEE Transaction on Automatic Control, Vol. AC-33. Paduano, J., L. Valavani, A. H. Epstein, E. M. Greitzer, and J. Guennette. 1990. Modeling for Control of Rotating Stall. Invited Paper, IEEE Conference on Decision and Control, December. Submitted to Automatica. Ralph, M. E. 1989. Control of the Variable Speed Generator on the SANDIA 34-Metre Vertical Axis Wind Turbine. Presented at Windpower '89, San Francisco, California, September. Rohrs, C. E., L. Valavani, M. Athans, and G. Stein. 1985. Robustness of Adaptive Control Algorithms in the Presence of Unmodeled Dynamics. IEEE Transaction on Automatic Control, Vol. AC-30, No. 9, September, pp. 881-889. Spangler R. L., and S. R. Hall. 1990. Piezoelectric Actuators for Helicopter Rotor Control. 31st Structures, Structural Dynamics and Materials Conference, Long Beach, California, April 2-4.
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology Stein, G. 1985. Beyond Singular Values and Loop Shapes. Proceedings of the American Control Conference, Boston, Massachusetts, June. Stein G., and M. Athans. 1987. The LQG/LTR Procedure for Multivariable Feedback Control Design. IEEE Transactions on Automatic Control, Vol. AC-32, No. 2, February, pp. 105-114. Vachon, W. A. 1987. The Effects of Control Algorithms on Fatigue Life and Energy Production of Vertical Axis Wind Turbines. Vachon & Associates Report .
OCR for page 91
Assessment of Research Needs for Wind Turbine Rotor Materials Technology This page in the original is blank.