Wind turbine rotational speed control using two proximity sensors for rotational speed measurement

The invention relates to a method for speed control in a wind turbine according to claim 1.

As wind turbines get larger, the loads in the drive train naturally increase, which normally consists of a wind rotor, a gearbox and the shaft sections between the wind rotor and gearbox as well as the gearbox and generator. The drive train is exposed to considerable bending and torsional loads. In addition to the static loads, there are dynamic ones, which place particular stress on the bearings and the gearbox. It has therefore already become known to electronically dampen vibrations in the drive train by means of appropriate control of the converter. This requires precise speed measurement. Various methods for measuring the speed in wind energy plants have become known.

A known method uses a so-called incremental encoder, which counts the number of pulses per unit of time. Such an incremental encoder is usually arranged on a shaft section on the side of the generator opposite the gear. Incremental encoders are complex and prone to failure. If possible, a speed measuring arrangement should work precisely and undisturbed over a very long period of time. Since an incremental encoder should be arranged at the output of the gearbox for safety reasons, installation is difficult and time-consuming.

As an alternative to the incremental encoder, it is known to use an encoder disk which has a row of equally spaced cams on its circumference and which is seated in a rotationally fixed manner on the associated shaft section at the output of the transmission. A stationary proximity sensor detects the passing of the cams, and the time between two adjacent or more cams is detected by means of a computer. The speed is calculated using the measured time according to the formula<i>n</i>1 = <maths id="math0001" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><msub><mrow><mtext mathvariant="italic">t</mtext></mrow><mrow><mtext>1</mtext></mrow></msub><mtext>*</mtext><mtext mathvariant="italic">Number of cams</mtext></mrow></mfrac></mrow></math><img file="EP1524433A1_D0001.tif" /></maths> calculated, where t₁ is the time that elapses between the approach or the passage of two adjacent cams. It is understood that the accuracy in the manufacture of the encoder disk influences the precision of the speed measurement. For a load-optimized control, an accuracy of &lt;0.1% and a phase shift &lt;5 ° in relation to a 5 Hz oscillation should be aimed for. A sensor disk that achieves this accuracy would have to be manufactured with a tolerance of ≤0.01 mm. This requires a not inconsiderable manufacturing effort. In addition, the disc can be damaged by external influences, so that the required accuracy is no longer guaranteed.

The invention has for its object to provide a method for speed control in wind turbines that can be carried out with little effort and provides the required accuracy for a load-optimized control.

This object is achieved by the features of patent claim 1.

In the method according to the invention, an encoder arrangement is also used, for example an encoder disk of conventional construction. It is essential to the invention that the encoder sections, which are preferably arranged at a uniform distance on the encoder arrangement, each interact with two sensors arranged one behind the other in the direction of rotation of the encoder arrangement or sensor arrangement and at a fixed distance from one another, with a computer determining the time which at least takes place when the encoder arrangement rotates a transmitter section between the first and second sensors is required. If the angular distance α between the sensors is known, the speed results <i>n</i> = <maths id="math0002" num=""><math display="inline"><mrow><mfrac><mrow><mtext>α</mtext></mrow><mrow><mtext>360°*</mtext><mtext mathvariant="italic">t</mtext></mrow></mfrac></mrow></math><img file="EP1524433A1_D0002.tif" /></maths>. If the circumference U of the sensor arrangement or sensor arrangement is known, ie the length of the circle which the effective sensor sections or sensors describe when rotating and the length s of the chord between the sensors or sensor sections, the speed can be approximated from the formula n * t = Calculate s / U, where t is again the measured time that a sensor section needs for the distance between the sensors. An actual speed signal corresponding to the calculated speed is sent to a controller, which sends an actuating signal to an adjusting device for a blade adjustment of a rotor if the actual speed signal deviates from a desired speed value.

Recent wind turbines place extremely high demands on the technical design of the control system. On the one hand, extremely large masses are in motion in wind turbines. At present, rotor blades with a weight of up to 20 tons and a length of up to 60 meters are very elastic components that can bend strongly. The drive train, with its large rotating masses (rotor weight currently at 120 tonnes, gearbox and generator weights at up to 70 tonnes) is also a very vibration-prone system and reliable execution of all components. Large masses have to be moved when adjusting the blades of a wind turbine. To regulate the power consumption from the wind, the blades are rotated about the longitudinal axis with electrical or hydraulic blade adjustment systems. The blade adjustment, together with a control of the generator torque, has the task of regulating or keeping the speed and the power of the wind energy installation constant. If the speed increases at nominal power, the blade angle is adjusted in the direction of the flag position, for example. As the speed drops, the blade angle is reduced.

If a "noise" occurs during the speed measurement, the blade adjustment controller is subjected to incorrect information about the speed. If you use this noise directly as an adjustment command for the rotor blades, there are constant small adjustment movements, depending on the frequency, there can even be a real "tremor". As the inertia of the adjustment system and blades is constantly accelerated, this leads to high loads and a reduction in the running time of the components involved. Since there is "play" in every mechanical system (eg tooth flanks in the gearbox, guide rods in the bolt eye, etc.), this leads to additional wear. This effect is reinforced by the fact that no play changes the blade angle and the controller would compensate for this by increasing the setpoint. That is why it has already become known to filter a speed signal. However, a filter results in delay times and also a phase shift. It is therefore only possible to react to the constantly fluctuating wind conditions with a delay, which worsens the control quality.

A phase shift can be avoided or reduced by an optimal speed signal. This is of crucial importance in the case of an increasing number of wind energy plants feeding into the electrical network, since this is the only way to generate the current with small power fluctuations and free from harmonics (flicker). Filtering the speed signal can then only be used to avoid resonance effects, e.g. to dampen the drive train vibrations mentioned.

The distance between the sensors should be small in relation to the diameter of the rotating arrangement, ie encoder or sensor arrangement, since otherwise the distance measurement via the circular chord leads to inaccuracies.

An extremely low-noise speed signal is obtained with the method according to the invention. The speed measurement is no longer dependent on the accuracy of the encoder arrangement. It is not essential whether the encoder sections have an exact circumferential distance from one another. Therefore, the method according to the invention is not affected by a mechanical deformation of the encoder arrangement, which can occur during assembly or due to service work or other processes.

Unavoidable inaccuracies in the geometry of encoder arrangements led to noise in the speed signal in conventional methods. They caused the display of changes in speed that did not actually occur. Such a noisy speed signal cannot be used for precise control without special, complex electronic processing. The method according to the invention allows a more precise speed measurement without significant noise, since only the actual speed and not speed fluctuations due to inaccuracies in the system are measured. The method according to the invention therefore leads to a steadier speed signal which can be used well for load-optimized control. A slight inaccuracy in the speed signal - in absolute terms - is acceptable. It is not important to measure the absolute speed precisely, but to detect fluctuations and phase shift. Because the rotation of the measured shaft section is not completely uniform due to the dynamic loads that occur, but is subject to changes dependent on vibrations. The low-frequency changes can be detected with sufficient accuracy using the method according to the invention.

The method according to the invention requires at least knowledge of the distance or angular distance of the sensors. The distance between the sensors can change under certain circumstances. It is also necessary to either set it precisely or to measure it precisely. Both can lead to inaccuracies. Therefore, an embodiment of the invention provides that a speed n1 is determined in a first step for calibration in a first step by the time t₁ between the activation of the sensor is measured by two or more sensor sections and in a second step the ratio s / U is determined according to the formula s / U = n1 * t using the time t required by a sensor section between the two sensors. s is the distance between the sensors and U is the circumference of the circle that the sensors or sensor sections describe. Since it is assumed here that the speed is unchanged from the first to the second step, the measurement of the time t is preferably carried out simultaneously or within the scope of the possible data acquisition rate of a computer used for the measuring method almost simultaneously with the first step (determination of the speed n1).

The accuracy of the calibration can be increased by repeating the method several times by determining the ratio s / U as the mean value of the results of the individual measurements. This value is saved and serves as the assumed quotient s / U until the next calibration.

In order to avoid that when the first start-up, for which there is still no calibrated value for the quotient s / U, the system must first be controlled using the measurement method based on only one sensor, an advantageous development of the invention provides that the system is determined in a first step with an approximate (estimated) value according to the method of claims 1 to 5. Thus, the noise-free speed signal maintains operational reliability and the risk of possible vibrations is avoided, but an absolute error (depending on the quality of the approximate quotient s / U less) is accepted in the speed detection. In a second step, a speed n1 is determined with the help of a single one of the sensors in each case by measuring the time between the activation of the sensor by two or more adjacent sensor sections. In a third step, the ratio s / U is newly determined from a measurement of the time t approximately at the same time as the second step, according to the formula s / U = n1 * t, and stored in the computer for the further measurement.

The calibration method described has the advantage that an exact measurement of the distance between the sensors on the one hand and the circumference of the circle of the transmitter sections or sensors is not necessary.

The sensors are preferably proximity sensors and interact with sensor sections of a sensor arrangement which is arranged approximately on a shaft section of the drive train. It is also conceivable to rotate the sensors and to arrange the transmitter sections in a stationary manner. A sensor disk for a sensor arrangement can conventionally be provided with radial cams on the circumference, which are arranged at the same circumferential distance. However, it is possible to attach any other encoder sections to a encoder arrangement, for example axial cutouts or bores that interact with the sensors. The sensors preferably operate inductively and can be conventional proximity sensors.

An encoder disk can be formed in one piece and mounted in a rotationally fixed manner on a shaft section, for example in the area of ​​a clutch. Alternatively, it is possible to design the transmitter arrangement, for example a transmitter disk, in a divided manner and to attach it to a desired section of the drive train.

The transmitter arrangement is preferably arranged on the output shaft of the transmission, the rotational speed of which, owing to the transmission ratio, is significantly greater than the rotational speed of the shaft of the wind rotor. On the other hand, it is conceivable to design the so-called locking disk on the rotor shaft as an encoder disk. As is known, for safety reasons, a locking disk is arranged between the gearbox and the wind rotor, which interacts with a lock in order to hold the shaft of the rotor and thus the rotor itself in a certain position. For security reasons, this is done while performing maintenance work.

The effects described for the blade adjustment also apply analogously to the drive train, which, with its large rotating masses and its high susceptibility to vibrations, places very high demands on the control quality and the robust controller design. If the torque controller is operated with a noisy speed signal as an input variable, there is a great risk that dynamic instabilities and thus mechanical vibrations will occur. A smoothing or filtering of the speed signal is out of the question due to the delay time or the phase shift, since active damping of drive train vibrations can only be achieved with very fast control with satisfactory results. Although it is conceivable to generally dampen a periodic oscillation only with a delay of a few periods, the most effective and most economical control can only be achieved with a noise-free and reliable speed signal.

The invention will be explained in more detail below with reference to an embodiment shown in the drawings.<dl id="dl0001"><dt>Fig. 1</dt><dd>very schematically shows a block diagram of a wind turbine.</dd><dt>Fig. 2</dt><dd>shows schematically a speed measuring arrangement for the wind turbine according to FIG. 1.</dd><dt>Fig. 3</dt><dd>2 shows a pulse diagram of the measuring arrangement according to FIG. 2.</dd></dl>

1 shows a wind rotor 10 of a wind power plant, the shaft 12 of which is mounted in two bearings 14, 16. The shaft 12 is the input shaft of a transmission 18, which is not further described, and which translates the speed of the shaft 12 to a much higher speed, for example by a factor of 100. An output shaft 20 of the transmission 18 is connected to a shaft 24 of a three-phase generator 26 via a coupling 22 coupled. Between the wind rotor 10 and the first bearing 14, a locking disk 28 is arranged in a rotationally fixed manner on the shaft 12, which interacts with a locking element 30. If the locking element 30 is inserted, for example, into an opening or recess in the locking disk 28, the shaft 12 is thus prevented from rotating.

Near the clutch 22, an encoder disk 32 is arranged on the shaft section 20 in a rotationally fixed manner. It interacts with a sensor arrangement 34, the signals of which are sent to a computer 36. As a controller, the computer 36 outputs a torque control signal to a converter 38 for the alternating current generated by the generator 26. The converter generates alternating current with the standard parameters for the purpose of feeding into a network. However, this will not be dealt with in detail. It should only be mentioned that it is possible with the aid of a load-optimized control of the converter 38 to implement electronic vibration damping for the drive train according to FIG. 1. This will not be discussed in more detail here either. For such a control, however, an accurate detection of the speed on the drive train is required.

The rotor 10 contains an adjusting device 39 for the blades of the rotor 10. At least one control signal ϑ is generated by the computer 36 <i>i</i> given on the adjusting device 39. With two or more blades of a rotor, a control signal can be generated for each blade. The controller is located in the computer 36. The actual speed signal is calculated in the computer 36 from the signals of the sensor arrangement 34 and compared with a speed setpoint signal for the purpose of determining at least one control signal ϑ<i>i</i> for the blade adjustment device 39.

In Fig. 2 it can be seen that the encoder disc 32 has on its circumference a series of radial cams 40 which are at a certain equal distance from each other. Two sensors 42, 44 are arranged at a distance s on the circumference of the cam disk 32. The sensors 42, 44 arranged one behind the other in the direction of rotation of the encoder disc essentially result in the sensor arrangement 34 according to FIG. 1. The distance between the cams 40 is preferably greater than the distance s between the sensors 42, 44. The sensors 42, 44 are preferably arranged radially (not shown), so that the distance is not a straight line but an arc with the arc angle α. Then the speed n results<maths id="math0003" num=""><math display="inline"><mrow><mfrac><mrow><mtext>α</mtext></mrow><mrow><mtext>360°*</mtext><mtext mathvariant="italic">t</mtext></mrow></mfrac></mrow></math><img file="EP1524433A1_D0003.tif" /></maths>, where α is the arc angle between the sensors and t is the time it takes a cam from the first to the second sensor.

The encoder disk 32 is made, for example, in one piece from metal, for example iron, and the sensors 42, 44 are proximity sensors which generate a signal when a cam 40 moves close to the sensor 42, 44. If one considers a single cam, for example cam 1, approximately rectangular pulses according to FIG. 3 result, whereby in the time diagram of FIG. 3rd the time t is the time that a cam 40 needs for the distance between sensor 42 and sensor 44 when the encoder disk 34 rotates. The time t is measured in the computer 36, and the speed is determined in the computer according to the formula<i>n</i> = <maths id="math0004" num=""><math display="inline"><mrow><mfrac><mrow><mtext>α</mtext></mrow><mrow><mtext>360°*</mtext><mtext mathvariant="italic">t</mtext></mrow></mfrac></mrow></math><img file="EP1524433A1_D0004.tif" /></maths> or <i>n</i> = <maths id="math0005" num=""><math display="inline"><mrow><mfrac><mrow><mtext mathvariant="italic">s</mtext></mrow><mrow><mtext mathvariant="italic">t</mtext><mtext>*</mtext><mtext mathvariant="italic">U</mtext></mrow></mfrac></mrow></math><img file="EP1524433A1_D0005.tif" /></maths>3 and U is the circumference of the encoder disk 34, specifically in the area of ​​the radial ends of the cams 40. Since there is always only one cam 40 for calculating the time t is considered, the true speed of the corresponding shaft section, here the shaft section 20, is measured in this way, regardless of whether the cams 40 are precisely machined with regard to their distance.

As can be seen, knowledge of the arc angle α or the parameters s and U is required for the described speed measurement. The arc angle α or the quotient s / U can be calculated by a calibration method without a precise measurement having to be carried out. In a first step, a speed n1 is determined by means of a sensor and in a second step, preferably at the same time, the ratio s / U is determined according to the formula s / U = n1 * t. This process can be repeated several times to increase the accuracy.

In order to avoid that when the first start-up, in which there is still no calibrated value for the quotient s / U, the system must first be controlled using the measurement method based on only one sensor, an advantageous development of the invention provides that the speed is determined in a first step with an approximate (estimated) value according to the method of claims 1 to 5. Thus, the noise-free speed signal maintains operational reliability and the risk of possible vibrations is avoided, but an absolute error (depending on the quality of the approximate quotient s / U less) is accepted in the speed detection. In a second step, a speed n1 is determined with the help of a single one of the sensors in each case by measuring the time between the activation of the sensor by two or more adjacent sensor sections. In a third step, the ratio s / U is newly determined from a measurement of the time t approximately at the same time as the second step, according to the formula s / U = n1 * t, and stored in the evaluation unit for the further measurement.

Such a calibration can be carried out as often as desired and, in particular, can be carried out before each start-up of the wind energy installation.

As can be seen, the sensors are stationary and the encoder disk is described as rotating. It goes without saying that a kinematic reversal is also possible. Furthermore, the sensors can be arranged with an axially directed axis and the sensor sections can be formed by recesses or bores in a sensor arrangement.