Patent Application
20100004878
This invention relates to the monitoring of wind turbines.
Many advanced wind turbines continuously acquire and transmit performance measurement data to a remote location. A wide range of parameters are typically measured covering both input and output parameters associated with the turbine. Input parameters measured include wind conditions, yaw angle, blade pitch angle and many more parameters. These parameters provide information about the configuration of the turbine and the conditions in which it is operating. Output parameters often measured include generator power, rotor speed, lubricant temperatures, and vibrations and provide information about how the turbine and its key constituent components are performing at any moment in time. The input parameters can be viewed as the ‘cause’ and the output parameters as the ‘effect’.
Measurement data is used for a variety of different purposes. Control systems within the turbine utilise input data to optimise the turbine configuration, for example adjusting the turbine yaw to track changes in the wind direction. Measurement data from turbine output parameters are used for performance monitoring, condition monitoring and fault protection. Performance monitoring provides an analysis of how a turbine is operating and enables comparison with expectation and with other turbines. Condition monitoring enables maintenance and intervention to be scheduled in a timely manner. Fault protection provides a fail safe mechanism to avoid or reduce turbine damage in the event of component failures or overloads.
In a typical condition monitoring system, data from numerous sensors and other instrumentation in the turbine are acquired by a central monitoring unit located in the nacelle of the turbine. The monitoring unit acquires data several times per second and performs signal processing on the measurements. Data can be statistically analysed, converted into the frequency domain for analysis, or combined with data from other sensors. Processed data is then sent on to a remote server. Since the bandwidth of the link between the monitoring unit and the server is limited, the monitoring unit summarises the measurement data prior to onward transmission.
The server stores threshold levels for key measurement parameters and can raise alarm and warning messages via email or SMS. The server transmits summary data received from the monitoring unit on to the control room, where data from other turbines is also collected. Following an alarm event, a short burst of data from the alarming sensor can also be sent to the control room. Software running in the remote control room enables data from all turbines connected to the system to be viewed and compared.
Blades for wind turbines are typically constructed of glass-reinforced plastics (GRP) on a sub-structure, which may be formed of wood, glass fibre, carbon fibre, foam or other materials. A typical wind turbine blade may have a length of between 20 and 60 metres or more. It is known, for example from U.S. Pat. No. 4,297,076, to provides the blades of a wind turbine with strain gauges and to adjust the pitch of portions of the blades in response to the bending moment on the blades measured by the strain gauges. Manufacturers of wind turbines are now installing strain sensors in turbine blades for real time measurement of blade bending moments. The blade load information is used for both cyclic pitch control and condition monitoring. Information about a wind turbine's condition can be monitored remotely to ensure continued effective operation of the turbine.
Wind turbines may also include drive train monitoring systems which use accelerometers and displacement sensors on key components of the drive train to identify any degradation of the drive train components.
It would be desirable to maximise the available performance information that can be generated from the sensors installed in a wind turbine.
Accordingly, viewed from a first aspect, this invention provides a method of monitoring the performance of a wind turbine. The wind turbine has at least one turbine blade mounted to a rotor and is provided with at least a first strain sensor for measuring mechanical strain of the turbine blade. The method comprises:
a) processing an output signal of the first strain sensor to identify a periodic component of the output signal indicative of mechanical strain due to the effect of gravity on the turbine blade;
b) generating a signal representing at least the speed of rotation of the turbine blade about the axis of the rotor by reference to the identified periodic component of the output signal of the first strain sensor.
Thus, in accordance with the invention, strain sensors that have been used typically for condition monitoring of wind turbines can be used to quantify the speed of rotation of the wind turbine by simple analysis of the strain sensor output signals.
The signal representing the speed of rotation of the turbine blade may be a simple speed indication, such as a value or a pulse. However, preferably the step of generating a signal representing the speed of rotation of the turbine blade about the axis of the rotor includes generating a signal representing the angular position of the turbine blade about axis of the rotor. Thus, the method extends to generating an indication of the angular position of the turbine blade, which in itself will indicate speed of rotation, because the angular position will change for a rotating turbine.
The signals generated in accordance with the invention may be relative indications of changes in rotation speed and/or other parameters of the wind turbine performance.
However, it is preferred for the output signals to be calibrated into accurate physical parameters. In one arrangement the measured values from the strain sensors are processed to generate bending moments for the turbine blades. For example, the turbine blade may be provided with at least a second strain sensor. The first and second strain sensors may be arranged to measure strain in a first direction and may be spaced on the turbine blade in a direction substantially orthogonal to the first direction. The difference in mechanical strain measured by the first and second strain sensors may be representative of a bending moment on the turbine blade. Other arrangements of strain sensors may be used to generate bending moment information.
The step of processing the output signal of the first strain sensor may include generating a signal representing a bending moment on the turbine blade by reference the output signal from the second strain sensor. Thus, the rotational speed signal may be generated by reference to a sinusoidal signal indicative of a bending moment, rather than simply strain, due to gravity.
The turbine blade may be provided with at least a third strain sensor spaced from, and not collinear with, the first and second strain sensors. In this way, signals representing bending moments on the turbine blade in two orthogonal directions can be generated from the differences in the mechanical strain measured by the first, second and third strain sensors. In a typical arrangement, at least four strain sensors are provided. The four strain sensors may be arranged in two collinear pairs along respective, substantially orthogonal axes.
The method may comprise determining the angle of inclination of the turbine blade about an axis extending radially from the rotor by comparison of the components of the bending moments in the two orthogonal directions.
This in itself is believed to be a novel method and, thus, viewed from further aspect, the invention provides a method of monitoring the performance of a wind turbine, the wind turbine having at least one turbine blade mounted to a rotor and provided with at least a first strain sensor and a second strain sensor for measuring mechanical strain of the turbine blade, wherein the first strain sensor and the second strain sensor are arranged on the turbine blade to provide output signals representative of mechanical strain on the turbine blade in two non-parallel directions, the method comprising:
Thus, in accordance with this aspect of the invention, strain sensors that have been used typically for condition monitoring of wind turbines can be used to quantify the pitch of the wind turbine blades by simple analysis of the strain sensor output signals.
The comparison of the components of the moments in the two non-parallel directions may comprise calculating a ratio of the components.
Viewed from a yet further aspect, the invention provides a method of monitoring the performance of a wind turbine, the wind turbine having at least one turbine blade mounted to a rotor and provided with at least a first strain sensor and a second strain sensor for measuring mechanical strain of the turbine blade, wherein the first strain sensor and the second strain sensor are arranged on the turbine blade to provide output signals representative of bending moments on the turbine blade in two non-parallel directions, the method comprising:
According to this aspect of the invention the drive torque about the axis of the rotor can be measured directly. This has the very significant advantage that the input power to the turbine can be calculated from the drive torque and the rotational speed. If the input power is known the efficiency of the wind turbine can be calculated from the output power.
The wind turbine may comprise a plurality of turbine blades distributed evenly about the rotor. Each blade may have respective first and second sensors. The step of generating a signal indicative of the torque about the axis of the rotor may include summing the bending moments about the axis of the rotor due to each turbine blade, whereby the effect of gravity is cancelled out.
The method may further comprise the step of generating a signal from the bending moment signals indicative of the resultant torque about an axis orthogonal to the axis of the rotor of the wind turbine.
The invention extends to computer software adapted to process output signals from strain sensors in accordance with the method described and to data processing apparatus adapted to process output signals from strain sensors in accordance with the method.
An embodiment of the invention will now be described by way of example only and with reference to the accompanying drawings, in which:
By far the most influential input or cause of degradation in the drive train of a wind turbine is the loads on the drive shaft as a result of wind-induced forces on the blades. Monitoring of the blade loads and the loads transmitted into the drive shaft provides information about why degradation has occurred and further enables action to be taken to reduce wear.
Monitoring of blade loads has traditionally been conducted using resistive strain gauges and has been restricted to blade testing and qualification applications due to the poor reliability and fatigue performance of resistive gauges. The present applicant has introduced a long term reliable blade load monitoring system, based on Bragg fibre grating strain sensors, as described in WO 2004/056017. The blade monitoring system is installed within turbines for long term structural health monitoring and cyclic pitch control applications. Optical fibre sensors are installed in the root of each blade to measure flapwise and edgewise bending moment.
As shown schematically in
The edgewise axis E runs generally parallel to the longest transverse dimension, i.e. the width, of the turbine blade. Thus, the sensors 4 located on the edgewise axis E measure the strain in the edges of the turbine blade 2 that cut through the air as the rotor 3 rotates. From the difference in the strain measurements from the two sensors 4 located on the edgewise axis and the fixed mechanical properties of the turbine blade, the bending moment on the turbine blade 2 in the plane defined by the edgewise axis and the radial axis can be calculated.
The flapwise axis F is substantially orthogonal to the edgewise axis E, such that the sensors 4 located on the flapwise axis measure the strain on opposed surface of the turbine blade 2 over which air passes as the rotor 3 rotates. From the difference in the strain measurements from the two sensors 4 located on the flapwise axis F and the fixed mechanical properties of the turbine blade, the bending moment on the turbine blade 2 in the plane defined by the flapwise axis and the radial axis can be calculated. The edgewise axis E and the flapwise axis F are substantially orthogonal to the radial axis R. As shown in
As the wind turbine rotates, the radial axis R of each blade 2 describes a circle about the axis A of the rotor 3. During this rotation, the bending moments measured by the edgewise and flapwise strain sensors 4 on each blade due to the effect of gravity vary sinusoidally as the relative orientation of the edgewise and flapwise axes vary with respect to the absolute vertical direction. This results in a sinusoidal component of the bending moment data of frequency ω, i.e. the rotational speed of the wind turbine, as shown in
Moreover, the pitch α of the turbine blade determines the proportion of the sinusoidal bending moment that appears in each of the edgewise and flapwise bending moment signals. If the pitch is zero (and the rotor axis A is substantially horizontal), all of the bending moment due to gravity will appear in the edgewise bending moment signal. Where the pitch α of the turbine blade 2 is non-zero, the ratio of the sinusoidal components from the flapwise and the edgewise bending moments (corrected for any angle between the axis A of the rotor and the horizontal) represents the tangent of the pitch angle α. Thus, the pitch α of the turbine blade can also be calculated from the bending moment information. In order to calculate the pitch angle correctly, the ratio of the in-phase sinusoidal components of the flapwise and edgewise bending moments is calculated, as it is possible for each signal to include two sinusoidal components, for example a component due to differing winds speed at the highest and lowest points of the rotation cycle, as well as the components due to the effect of gravity.
Where the instantaneous rotational position θ and the pitch α of each blade 2 is known, the instantaneous bending moments in the edgewise and flapwise planes can be resolved into coordinate system relative to the orientation of the rotor 3. In this way, in addition to indicating blade health, the blade root bending moments can be combined to calculate input loads to the drive shaft including drive torque, load on tower and resultant offset load on the rotor shaft. Thus, for example, the sum of the bending moments from all of the turbine blades 2 resolved into the plane normal to the rotational axis A of the rotor 3 represents the drive torque on the rotor.
Using some basic assumptions regarding the expected distribution of the forces on the blade 2 that cause the bending moments, the force on each blade causing each bending moment can be calculated. The forces can be resolved in any desired direction using the instantaneous rotational position θ and the pitch α of each blade 2.
For the calculation of loads and moments on the rotor from the blade bending moment data, the instantaneous rotational position θ and the pitch α of each blade 2 may be determined otherwise than from the blade bending moment data, as described above. For example, the instantaneous rotational position θ and the pitch α of each blade 2 may be received from the control system of the wind turbine.
By resolving the edgewise and flapwise loads in the plane of the rotor, the input torque to the drive shaft can be calculated as a function of time showing the magnitude and variability of the drive torque. Frequency domain analysis of the drive torque for a particular turbine highlighted a strong harmonic at the rotor rotation frequency. If the rotor were perfectly balanced, then all blades would be generating equally when at the same point in space, except for variations in wind conditions. Variations in wind conditions are both systematic (for example wind shear, tower shadowing) and non-systematic (for example gusts) and for a balanced rotor neither of these should generate a drive torque that varies at the same frequency as the rotor rotates. The large harmonic is therefore an indication of rotor imbalance. Examination of the phase of the frequency domain information revealed the particular blade that was out of balance.
By resolving blade root bending moments in the horizontal and vertical direction the offset loads on the drive shaft can be determined. The bearings in the drive shaft are designed to accommodate axial loads, but continual dynamic offset loads can lead to wear. The resultant magnitude and direction of the resultant load for the turbine monitored is shown in
The load vector is almost vertical as one blade travels through the top of its sweep and then drops rapidly in magnitude. For the turbine monitored the magnitude of the offset load was examined in the frequency domain. The response is almost entirely at three times the rotor frequency, as would be expected due to systematic wind variations.
All of the above derived parameters representing both blade performance and rotor loads are initially calculated 30 times per second and stored in short term memory within the instrumentation 5. Further data compression and statistical analysis of both time domain and frequency domain data is performed to generate a limited number of summary parameters for onward transmission to a condition monitoring system or a separate data logger unit.
The strain sensor instrument 5 is located in the turbine hub and converts optical signals received from the blade load sensors 4 to digital data. The instrument 5 generates about 500 measurements every second which is too much to be directly input to a typical condition monitoring unit, in addition to data arriving from all the other sensors recorded. The strain sensor instrument 5 therefore processes the raw data as described above to calculate key parameters for both the blades 2 and the rotor 3. Parameters include blade bending moment, blade fatigue, drive torque and offset load vector and are calculated 30 times every second. Time domain data of the derived parameters is still too much information to be transmitted to the condition monitoring unit so the data is summarised using key time domain and frequency domain statistics as described above. Blocks of data 1 minute in length are summarised into a total of 32 numbers that are transmitted to the monitoring unit. The strain sensor instrumentation has therefore summarised a total of 30,000 strain measurements acquired during a minute into 32 numbers for onward transmission to the monitoring unit.
The monitoring unit typically transmits data on to the control room a few times each day. Transmission bandwidth from the monitoring unit to the control room is also limited, so further processing of the data is performed in the monitoring unit. Further data reduction is performed specific to each parameter.
For accumulated fatigue measurements sent from the strain-sensing instrumentation 5, the most recent value is sent to the control room. For blade bending moments, the strain-sensing instrumentation transmits maximum, minimum, average and RMS summary values for each minute. The monitoring unit uses these values to generate blade load histograms and it is these histograms that are uploaded to the control room to summarise the loading on the blades since the last data upload.
In addition to processing and summarising data from the strain-sensing instrumentation 5, the monitoring unit also summarises and transmits data from all the other sensors. The control room server therefore receives summary data from both input and output instrumentation. With all the data stored via a single system it becomes simple to view cause and effect on a single screen.
Linking the input blade loads to other monitoring systems within a turbine provides a number of different benefits. Firstly the monitoring unit can raise alarms based on measurement data from the strain-sensing instrumentation, alerting the operator that key threshold levels stored within the monitoring unit have been exceeded. Secondly any unscheduled maintenance or intervention may be planned through condition monitoring of the blade and rotor along with the output drive train parameters. Thirdly the data provides the opportunity to monitor and improve our understanding of how dynamic blade loads lead to degradation of both blades and drive train components and to monitor the effectiveness of load reduction methods such as cyclic pitch control to ultimately reduce levels of wear.
Modern turbines contain extensive instrumentation monitoring a wide variety of parameters. Drive train monitoring during turbine operation has been limited to monitoring the output or result of blade loads on the drive shaft. Frequency analysis of accelerometer responses provides information about degradation of different parts of the drive train.
Blade load sensors monitor the blade bending moments that are the individual input loads to the drive shaft. The system has been interfaced with a typical condition monitoring system to link the input loads to the output responses. The system can be installed in new turbines alongside an existing condition monitoring system, or can just as easily be retrofitted to an existing turbine that contains a condition monitoring system. Alternatively, the system can operate as a stand-alone data logger.
The strain-sensing instrumentation processes the blade root bending moments to calculate information about both the blades and the rotor. Key parameters calculated include blade fatigue, rotor drive torque and offset load on the drive shaft. Frequency analysis of the drive torque and offset loads can infer how much the rotor is out of balance and can further identify which blade is incorrectly configured.
Linking the drive train inputs and outputs via a single system enables the condition of both the blades and the drive train to be monitored via a single system. It also enables drive shaft degradation to be correlated with the blade load conditions that cause the degradation which will lead to improved design of turbines and load reduction methods such as cyclic pitch control.
In summary, embodiments of the invention include installing load sensors in turbine blades for real time measurement of blade bending moments. Blade load information is used for both cyclic pitch control and condition monitoring applications. The blade loads are the inputs that lead to drive train degradation, which are linked to the output signals measuring degradation on the turbines drive train by combining the blade load data with drive train data via a single condition monitoring system.
The instrumentation calculates key parameters for the blades such as load history spectra and fatigue. It also combines measurements from each blade to calculate key rotor parameters including drive torque and offset load on the drive shaft. Statistical analysis of both time-domain and frequency-domain responses are used to summarise data prior to onward submission to the condition monitoring system.
With the input and output parameters all recorded via a single system, the cause of drive train degradation can be identified, enabling improved design of turbine components and dynamic load reduction methods. Furthermore, with both blade health and drive train health monitored via a single system, unscheduled maintenance and intervention may be better identified and planned.
A method of monitoring the performance of a wind turbine 1 uses bending moment data from strain sensors 4 in the turbine blades 2 to calculate rotational speed of the turbine 1, angular position of the turbine blades 2, drive torque and resultant load on the rotor 3. The method has the advantage that the inputs to the drive train of the wind turbine can be measure directly.
| Number | Date | Country | Kind |
|---|---|---|---|
| 0708749.7 | May 2007 | GB | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/GB08/50325 | 5/2/2008 | WO | 00 | 3/17/2009 |
