METHOD FOR DETERMINING THE OPERATING STATES OF A WIND TURBINE
The method improves the detection of underperformance in wind turbines by generating and converting signal series based on probabilistic distributions and theoretical models, providing precise identification and reducing energy loss.
Patent Information
- Application Number
- FR2022007987
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-08-01
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2042-08-01
AI Technical Summary
Existing methods for detecting underperformance states in wind turbines are imprecise due to averaging techniques that obscure actual performance deviations, leading to inaccurate assessments of power generation and potential material damage.
A method utilizing a control and data acquisition system to generate and convert signal series based on probabilistic distributions and theoretical models to reconstruct possible scenarios of operating parameters, allowing for precise detection of underperformance by comparing measured and estimated values.
Enhances the accuracy of identifying underperformance states in wind turbines, enabling better energy loss assessment and preventing structural damage by accounting for various operational scenarios.
Smart Images

Figure 00000030_0000 
Figure 00000031_0000 
Figure 00000032_0000
Abstract
Description
Title of the invention: METHOD FOR DETERMINING OPERATING STATES OF A WIND TURBINE FIELD OF THE INVENTION
[0001] The invention relates to a method for determining operating states of a wind turbine. STATE OF THE ART
[0002] A wind turbine is a device that transforms kinetic energy from the wind into electrical energy.
[0003] A wind turbine generally comprises a mast, a nacelle and a rotor rotatably mounted on the nacelle. The rotor comprises a hub and blades extending radially from the hub. The rotor is adapted to be rotated by the action of the wind on the blades. The rotor in turn drives a generator by means of a transmission mechanism. When rotated, the generator produces an electric current.
[0004] The rotation speed of the rotor being too low to create electric current at the frequency of the electrical network, the transmission mechanism comprises a gearbox or "multiplier" making it possible to reach a rotation speed required by the generator to create the electric current at the frequency of the network.
[0005] The electrical power generated by the wind turbine depends on weather conditions and more particularly on wind speed and direction.
[0006] In order to maximize the electrical power generated, the wind turbine includes a control-command system for adjusting the operating parameters of the wind turbine, such as the pitch angle of the blades (or "pitch angle") and the orientation of the nacelle relative to the mast (or "yaw").
[0007] In normal operation, the electrical power generated depends on the wind speed, and follows a predefined theoretical power curve.
[0008] The theoretical power curve has an S shape: it is generally divided into two parts. The first part corresponds to a rise in power until reaching the nominal power. The second part is a plateau corresponding to reaching the nominal power.
[0009] However, there are situations where the electrical power generated by the wind turbine is lower than the theoretical power defined by the theoretical power curve. In these situations, the wind turbine operates in "underperformance".
[0010] An operating situation in "underperformance" may be voluntary. This is the case, for example, when the operator wishes to reduce the electrical power generated by the wind turbine, for example for regulatory reasons (limiting the noise level generated by the wind turbine) or to limit the forces undergone by certain mechanical components of the wind turbine. For this purpose, the control system sends instructions for power or pitch angle (angular adjustment of the blades). The wind turbine is restricted.
[0011] However, in certain cases, an “underperformance” operating situation is unintentional, for example in the event of a failure of a component of the wind turbine (for example a failure of the blade angular adjustment system or the nacelle orientation system), or due to certain climatic conditions (for example the presence of frost on the blades in winter) which can lead to a reduction in the efficiency of the wind turbine.
[0012] Wind farm owners want to be able to assess the power losses of wind turbines due to these “underperforming” operating states.
[0013] In addition, "underperforming" operating states due to material damage must be detected so as not to cause more serious consequences within the structure of the wind turbine.
[0014] In order to be able to detect situations of “underperformance” operation of a wind turbine, it is possible to compare the active power actually produced by the wind turbine with a theoretical active power defined by the theoretical power curve of the wind turbine. An “underperformance” operating situation is detected when the difference between the power actually produced by the wind turbine and the theoretical active power becomes too large.
[0015] However, this technique does not allow precise detection of “underperformance” states of wind turbines.
[0016] Indeed, the value of the active power and the value of the wind speed measured on the wind turbine are averages over time intervals (typically 10 minutes), which has the effect of dispersing the measurement points around the theoretical power curve. This implies that for wind speeds located in the first part (increasing part) of the theoretical power curve, the points where the wind turbine is underperforming merge with the points where the wind turbine is operating normally.
[0017] Furthermore, for wind speeds located in the second part of the theoretical power curve (plateau), the averaged values distort the analysis of the performance of the wind turbines: the average power measured may be located below the theoretical power curve, because over a time interval considered (typically 10 minutes), the wind turbine has undergone periods where the wind speed was higher than the maximum value beyond which the rotor of the wind turbine is immobilized ("cut-out speed"). This has the result of lowering the value of the average power measured for an average wind speed remaining high. Summary of the invention
[0018] An aim of the invention is to propose a method which allows more precise detection of the operating state of a wind turbine.
[0019] This aim is achieved within the framework of the present invention by means of a method for determining operating states of a wind turbine, from a set of data recorded in a wind turbine data acquisition and control system, the set of data comprising a plurality of operating points of the wind turbine, each operating point of the wind turbine being defined by a time interval, one or more characteristic values of a first operating parameter of the wind turbine measured over the time interval and one or more characteristic values of a second operating parameter of the wind turbine measured over the time interval, in which, for each operating point, the method comprises steps of: - generate a first series of signals, each signal of the first series of signals being representative of a possible evolution of a value of the first operating parameter as a function of time, over the time interval, each signal of the first series of signals having one or more characteristic values of the first operating parameter of the wind turbine measured over the time interval, - converting the first series of signals into a second series of signals representative of a change in a value of the second operating parameter as a function of time, over the time interval, each signal of the second series of signals being obtained by converting a signal of the first series of signals by means of a theoretical model of operation of the wind turbine making it possible to estimate a value of the second operating parameter as a function of a value of the first operating parameter, - for each signal of the second series of signals, calculate an estimated characteristic value of the second parameter over the time interval, - compare the characteristic value of the second operating parameter of the wind turbine measured over the time interval with the estimated characteristic value of the second operating parameter over the time interval, and - deduce an operating state of the wind turbine associated with the operating point.
[0020] In other words, for the same operating point, the proposed method makes it possible to reconstruct a first series of signals and a second series of signals in ex using characteristic values recorded in the wind turbine's control and data acquisition system.
[0021] The first series of signals and the second series of signals generated represent as many possible scenarios of evolution of the first parameter (for example the wind speed) which could have been encountered by the wind turbine during the time interval considered.
[0022] In this way, the estimated characteristic value of the second operating parameter can take into account one or more of these possible scenarios.
[0023] In particular, the estimated characteristic value of the second operating parameter may take into account the worst scenario and / or the best scenario and / or an average scenario possibly encountered by the wind turbine over the time interval.
[0024] The characteristic value of the second operating parameter of the wind turbine measured (for example the active power generated by the generator of the wind turbine) is then compared with the characteristic value of the second operating parameter of the wind turbine estimated by taking into account one or more of these possible scenarios.
[0025] In this way, the operating state of the wind turbine associated with the operating point can be determined more accurately.
[0026] The proposed method may further have the following particularities:
[0027] In one embodiment of the method, the method comprises, for each operating point, a step of: - generating a plurality of possible values of the first operating parameter, distributed according to a predefined probabilistic distribution function, as a function of the characteristic value(s) of the first operating parameter of the wind turbine, and the first series of signals is generated by randomly drawing successive values of the first operating parameter from among the possible values of the first operating parameter.
[0028] In one embodiment of the method, the predefined probabilistic distribution function is chosen from a pseudo-Gaussian distribution function, a pseudo-Weibull distribution function or a uniform distribution function.
[0029] In one embodiment of the method, the step of generating a first series of signals representative of the evolution of a value of the first operating parameter as a function of time, over the time interval, comprises a sub-step of: - eliminate randomly drawn values of the first operating parameter which are not included in a predefined interval.
[0030] In one embodiment of the method, the method comprises, for each point of operation, stages of: - generate a first low-frequency component representative of a variation of the first operating parameter over the time interval, and - generating a plurality of second high frequency components representative of possible relative variations of the first operating parameter over the time interval, each signal of the first series of signals being generated by summing the first low frequency component and a second high frequency component of the plurality of second high frequency components.
[0031] In one embodiment of the method, the first low-frequency component is obtained by interpolation of characteristic values of the first operating parameter measured over several consecutive time intervals, including the time interval considered.
[0032] In one embodiment of the method, for each operating point, the characteristic value(s) of the first operating parameter comprise an average value of the first operating parameter measured over the time interval, and the first low-frequency component is generated by interpolation of the different average values of the first operating parameter.
[0033] In one embodiment of the method, each second high-frequency component is generated by a random process having a covariance function, the covariance function having been calculated beforehand on learning data comprising real values of the first operating parameter having been measured with a time step between two measurement instants less than the duration of the time interval.
[0034] In one embodiment of the method, for each operating point, the characteristic value(s) of the first operating parameter comprise a standard deviation value of the first operating parameter measured over the time interval, and each signal of the first series of signals is generated by summing the first low-frequency component and the second high-frequency component and denormalizing the resulting signal obtained so that the denormalized signal has a standard deviation equal to the standard deviation value of the first operating parameter over the time interval.
[0035] In one embodiment of the method, the characteristic value(s) of the first operating parameter comprise a minimum value, a maximum value, an average value and / or a standard deviation of the first operating parameter measured over the time interval.
[0036] In one embodiment of the method, the characteristic value(s) of the second operating parameter comprise an average value of the second operating parameter measured over the time interval.
[0037] In one embodiment of the method, the wind turbine comprises a mast, a nacelle, a rotor, an electric current generator, the rotor comprising a hub and blades mounted on the hub, the blades being capable of driving the rotor in rotation relative to the nacelle under the effect of the wind, and the electric current generator being capable of being driven in rotation by the rotor, and the first operating parameter and the second operating parameter are chosen from: - respectively a wind speed and an active power generated by the generator (12), or - respectively a pitch angle of the blades (6) relative to the hub (5) and a wind speed, or - respectively a motor torque transmitted by the rotor (4) to the generator (12) and an active power generated by the generator (12), or - respectively a wind orientation angle and an orientation angle of the nacelle (3) relative to the mast (2).
[0038] In one embodiment of the method, the reference value of the second operating parameter is determined as: - a minimum value among the set of estimated characteristic values of the second operating parameter over the time interval, and / or - an average of the estimated characteristic values of the second operating parameter over the time interval, and / or - a maximum value among the set of estimated characteristic values of the second operating parameter over the time interval.
[0039] In one embodiment of the method, during the step of deducing an operating state of the wind turbine associated with the operating point, the operating state of the wind turbine is selected from a first operating state and a second operating state, one of the first operating state being a normal operating state and the other of the first operating state and the second operating state being an underperforming operating state.
[0040] In one embodiment of the method, for each operating point: - if the characteristic value of the second operating parameter of the wind turbine measured over the time interval is greater than or equal to a threshold which depends on the reference value of the second operating parameter over the time interval, then the operating state of the wind turbine associated with the operating point is the first operating state, and - if the characteristic value of the second operating parameter of the wind turbine measured over the time interval is lower than the threshold, then the operating state of the wind turbine associated with the operating point is the second operating state.
[0041] In one embodiment of the method, the threshold is equal to the reference value of the second operating parameter plus or minus a margin.
[0042] In one embodiment of the method, the margin is defined as a predefined number multiplied by a standard deviation calculated on the set of estimated characteristic values of the second operating parameter over the time interval.
[0043] In one embodiment of the method, the model for estimating a value of the second operating parameter as a function of a value of the first operating parameter was obtained from learning data comprising a plurality of operating points of the wind turbine, the learning data having been previously acquired and recorded over a period of operation of the wind turbine during which the wind turbine is only in a normal operating state.
[0044] In one embodiment of the method, the operating period of the wind turbine has a duration of several months, preferably ranging from 3 to 6 months.
[0045] In one embodiment of the method, the model making it possible to estimate a value of the second operating parameter as a function of a value of the first operating parameter was obtained by steps of: - partition the learning data into several successive elementary slices, each slice containing all the operating points for which the first operating parameter is included in a predefined interval associated with the slice, and - calculate, from all the operating points contained in the slice, an average value of the second operating parameter over the predefined interval associated with the slice.
[0046] In one embodiment of the method, the method comprises a step of: - from the set of operating points and the respective associated operating states, calculate an electrical energy loss value as a difference between an expected theoretical energy production value and an actual energy production value.
[0047] In one embodiment of the method, the method comprises a step of: - from all the operating points and the respective associated operating states, calculate an operating time of the wind turbine in each of the operating states of the wind turbine.
[0048] The invention further relates to a computer program product, comprising instructions which, when the program is executed by a computer, cause the latter to implement the steps of the method as defined previously. PRESENTATION OF THE DRAWINGS
[0049] Other characteristics and advantages will emerge from the following description, which is purely illustrative and not limiting, and must be read in conjunction with the appended figures, among which:
[0050] - [Fig.l] schematically represents a wind turbine,
[0051] - [Fig.2] schematically represents a power curve of a wind turbine,
[0052] - [Fig.3] schematically represents steps of a method of determining determination of the operating state of a wind turbine according to a first embodiment of the invention,
[0053] - [Fig.4A] and [Fig.4B] schematically represent the processed data during different stages of the process of determining the operating status of the wind turbine,
[0054] - [Fig.5] schematically represents a theoretical power curve enabling a theoretical value of the second operating parameter of the wind turbine to be determined as a function of the first operating parameter of the wind turbine,
[0055] - [Fig.6] schematically represents data recorded in the real-time data acquisition and control system connected to the wind turbine,
[0056] - [Fig.7] schematically represents the average value of the values averages of the signals of the first series of signals, as a function of the number N of signals generated representative of the evolution of a value of the first operating parameter,
[0057] - [Fig.8] schematically represents the average value of the values averages of the signals of the second series of signals, as a function of the number N of signals generated representative of the evolution of a value of the second operating parameter,
[0058] - [Fig.9] schematically represents variations of determined thresholds for the values of the second parameter over a period of operation of the wind turbine to be analyzed,
[0059] - [Fig. 10] schematically represents steps of a method of determining determination of the operating state of a wind turbine according to a second embodiment of the invention,
[0060] - [Fig.l 1] is a curve representative of a relative variation in the value of the first operating parameter over time over the operating period of the wind turbine to be analyzed,
[0061] - [Fig.12] is a representative curve of a denormalized signal obtained from the normalized signal of [Fig.l 1] to take into account standard deviation values of the first operating parameter,
[0062] - [Fig. 13] is a curve representing an estimated variation in the value average of the first operating parameter over time over the operating period of the wind turbine to be analyzed,
[0063] - [Fig. 14] is a curve representative of a signal obtained by summation of the values of the curves in figures 12 and 13. DETAILED DESCRIPTION OF AN EMBODIMENT
[0064] In [Fig.l], the wind turbine 1 comprises a mast 2 and a nacelle 3 mounted at one end of the mast 2. The nacelle 3 is capable of pivoting relative to the mast 2 about a vertical axis X in order to allow adjustment of the orientation (or "yaw angle") of the nacelle 3 depending on the direction of the wind. The wind turbine 1 comprises a mechanism for orienting the nacelle 3 allowing the nacelle 3 to be pivoted relative to the mast 2 in order to orient the nacelle 3 according to the direction of the wind.
[0065] The wind turbine 1 further comprises a rotor 4. The rotor 4 comprises a hub 5 and blades 6. The rotor 4 is rotatably mounted relative to the nacelle about a horizontal Y axis. The rotor 4 may comprise three blades 6, but only two blades 6 are visible in [Fig.l].
[0066] The blades 6 are mounted on the hub 5. Each blade 6 extends radially relative to the Y axis.
[0067] The wind turbine 1 further comprises an angular adjustment mechanism for the blades 6 making it possible to modify the pitch of the blades. More precisely, each blade is pivotally mounted relative to the hub 5 around a radial axis Z making it possible to modify an angle of the blade 6 relative to the hub 5 (or “pitch angle”).
[0068] Inside the nacelle 3, the wind turbine 1 comprises a first transmission shaft 7 (or slow shaft), a multiplier 8 (or gearbox), a second transmission shaft 9 (or fast shaft) and a generator 10.
[0069] The generator 10 comprises a stator 11 fixedly mounted on the nacelle 3 and a rotor 12 rotatably mounted relative to the stator 11.
[0070] The first transmission shaft 7 connects the hub 5 to an input of the multiplier 8. The second transmission shaft 9 connects an output of the multiplier 8 to the rotor 12 of the generator 10.
[0071] In operation, under the effect of the wind, the rotor 4 is rotated relative to the nacelle 2 around the Y axis. The rotor 4 rotates the first transmission shaft 7 at a first rotational speed. The first transmission shaft 7 rotates the second transmission shaft 9 via the multiplier 8.
[0072] Thanks to the multiplier 8, the second transmission shaft 9 is driven in rotation at a second rotational speed higher than the first rotational speed. The multiplication ratio of the multiplier (output speed / input speed) is for example of the order of 100. The second transmission shaft 9 drives the rotor 12 of the generator 10 in rotation. Under the effect of the rotation of the rotor 12 relative to the stator 11, the generator 10 generates an alternating electric current at variable frequency.
[0073] A power converter makes it possible to convert the variable frequency alternating current generated by the generator 10 into an alternating current at a constant frequency equal to the frequency of an electrical distribution network. The multiplier 8 makes it possible to achieve a rotation speed required by the generator 10 to generate the electric current at a frequency suitable so that once converted, the electric current can propagate on the electrical distribution network.
[0074] A brake 13 makes it possible to immobilize the rotor 4 of the wind turbine 1 in the event of an emergency or during maintenance work on the wind turbine 1.
[0075] The wind turbine 1 is equipped with a plurality of sensors making it possible to measure operating parameters of the wind turbine 1.
[0076] In the example illustrated in [Fig.l], the wind turbine 1 comprises an anemometer 14 suitable for measuring a wind speed and a wind vane 15 suitable for measuring a wind direction relative to the nacelle 3.
[0077] The wind turbine 1 further comprises a first tachometer for measuring a rotation speed of the first shaft 7, and a second tachometer for measuring a rotation speed of the second shaft 9.
[0078] The wind turbine 1 may further comprise a sensor making it possible to measure the angle of the blades 6 (or “pitch angle”).
[0079] The wind turbine 1 may further comprise a sensor making it possible to measure parameters relating to a main bearing supporting the first transmission shaft 7, for example a thermocouple making it possible to measure a temperature of the main bearing, or an accelerometer making it possible to measure the frequency of its natural modes.
[0080] The wind turbine 1 may further comprise sensors making it possible to measure parameters relating to the multiplier 8, such as an accelerometer making it possible to measure natural mode frequencies of the different parts making up the multiplier, a thermocouple making it possible to measure a temperature of the multiplier, for example a temperature of a lubricating fluid contained in the multiplier, or a particle counter making it possible to measure a number of particles per unit of volume contained in the lubricating fluid.
[0081] The sensors make it possible to measure values of operating parameters of the wind turbine.
[0082] The measured values make it possible on the one hand to control the func in real time operation of the wind turbine 1. For this purpose, the wind turbine 1 includes a control-command system capable of adjusting the pitch of the blades and the orientation of the nacelle.
[0083] The measured values can furthermore be collected over a predefined period of time to evaluate the performance of the wind turbine 1 over this period of time.
[0084] [Fig.2] is a diagram schematically representing a power curve of wind turbine 1.
[0085] In this figure, the power curve comprises a plurality of operating points.
[0086] Each operating point of the wind turbine 1 is defined by a time interval and a series of characteristic values of operating parameters of the wind turbine measured during the time interval.
[0087] For example, for each operating point of the wind turbine 1, the characteristic values of an operating parameter may comprise a minimum value, a maximum value, an average value and / or a standard deviation of the operating parameter measured over the time interval.
[0088] Certain operating parameters can be measured directly using sensors.
[0089] These operating parameters may include one or more of the following parameters: - wind speed, - active power generated by the generator (i.e. at the output of the power converter), - blade pitch angle relative to the hub, - engine torque transmitted by the rotor to the generator, - wind direction angle, - angle of orientation of the nacelle relative to the mast (“yaw angle”).
[0090] The characteristic values of these operating parameters are recorded in a real-time control and data acquisition system (called “Supervisory Control and Data Acquisition” or “SCADA”) connected to the wind turbine 1.
[0091] More specifically, the characteristic values of the operating parameters recorded in the control and data acquisition system may include, for each operating point: - a minimum value, a maximum value, an average value and / or a standard deviation of a first operating parameter measured over the time interval, and - average value of the second operating parameter measured over the time interval.
[0092] The characteristic values of the first operating parameter and the second operating parameter can be measured directly by the sensors or be calculated from values measured by the sensors.
[0093] Each time interval can have a duration of 10 minutes for example.
[0094] In the example illustrated in [Fig.2], the first operating parameter is the wind speed and the second operating parameter is the active power generated by the generator 10.
[0095] However, other pairs (first operating parameter, second operating parameter) can be chosen such as: - (blade pitch angle relative to the hub, wind speed), or - (engine torque transmitted by the rotor to the generator, active power generated by the generator), or - (wind orientation angle, an angle of orientation of the nacelle relative to the mast).
[0096] The engine torque transmitted by the rotor 4 to the generator 10 can be calculated as being equal to the active power divided by the rotational speed of the first shaft 7.
[0097] In the diagram in [Fig.2], each operating point has as its abscissa an average value of the wind speed (in meters per second) and as its ordinate an average value of the active power generated by the generator (in kilowatts).
[0098] In normal operation, the power curve has a general “S” shape. The “S” curve is representative of the theoretical power delivered by wind turbine 1 as a function of wind speed.
[0099] When the wind speed is lower than a predetermined minimum speed (for example a minimum speed of approximately 3 meters per second), the active power produced by the wind turbine 1 is substantially zero.
[0100] When the wind speed is between the minimum speed and an intermediate speed (for example an intermediate speed of approximately 12 meters per second), the active power produced by the wind turbine increases with the wind speed.
[0101] When the wind speed is greater than the intermediate speed, the active power produced by the wind turbine 1 is substantially constant and equal to the nominal power of the wind turbine.
[0102] However, the diagram of [Fig. 2] also includes operating points which are located under the “S” curve and are distant from this curve. In other words, these points represent an operation of the wind turbine 1 in which the wind turbine provides an active power lower than the theoretical power expected for a given wind speed. These operating points may be representative of a degraded operating state of the wind turbine 1, referred to as an “underperforming” operating state. lier embodiment
[0103] [Fig. 3] schematically represents steps of a method 100 for determining mination of an operating state of the wind turbine 1, in accordance with a first embodiment of the invention.
[0104] The method 100 makes it possible to determine, for each operating point of the wind turbine recorded in the real-time control and acquisition system (SCADA), whether the wind turbine 1 is in a “normal” operating state or in an “underperforming” operating state.
[0105] In the example illustrated in [Fig.3], the method 100 comprises a first phase 110 (or “learning phase”) and a second phase 120 (or “analysis phase”).
[0106] The learning phase 110 aims to develop a theoretical operating model of the wind turbine which will subsequently be used during the analysis phase to determine whether the wind turbine 1 is in a “normal” operating state or in an “underperforming” operating state.
[0107] This theoretical model of operation of the wind turbine can be represented in the form of a curve defining a value of the second operating parameter as a function of the first operating parameter.
[0108] More precisely, in this example, the theoretical operating model of the wind turbine is a power curve defining a value of the active power generated by the generator 10 of the wind turbine 1 as a function of a value of the wind speed.
[0109] The learning phase 110 comprises the following steps.
[0110] According to a first step 111, learning data are collected. The collected learning data are for example data which have been recorded in the real-time control and acquisition system (SCADA) connected to the wind turbine 1.
[0111] The learning data comprises a plurality of operating points of the wind turbine 1. Each operating point of the wind turbine 1 is defined by a series of values of operating parameters of the wind turbine measured during the same time interval, during at least one operating period of the wind turbine, this operating period being considered as a reference period.
[0112] The operating period(s) of the wind turbine may include an operating period during which the wind turbine is only in a normal operating state. The operating period of the wind turbine may have a duration of several months, preferably ranging from 3 to 6 months.
[0113] The series of values of operating parameters of the wind turbine are measured with a constant time interval between two measurements. The duration of a time interval is several minutes, for example 10 minutes.
[0114] The operating parameters of wind turbine 1 include: - a first operating parameter (namely in this example, the wind speed), and - a second operating parameter (namely in this example, the active power generated by the generator).
[0115] According to a second step 112, the learning data which have been collected are partitioned into several successive elementary slices, each slice containing all of the operating points for which the first operating parameter is included in a predefined interval associated with the slice.
[0116] According to a third step 113, for each elementary slice, an average value of the second operating parameter over the predefined interval associated with the slice is calculated from all of the operating points contained in the slice.
[0117] According to a fourth step 114, the theoretical operating model of the wind turbine is obtained as a theoretical curve (namely in this example a theoretical power curve) defining a value of the second parameter (namely in this example, the active power generated by the generator) as a function of a value of the first parameter (namely, in this example the wind speed).
[0118] This theoretical curve can be obtained by polynomial regression from the average values of the second operating parameter obtained during the third step 113. The theoretical curve is thus estimated by approximation using a polynomial function, for example a polynomial function of degree greater than 3.
[0119] Alternatively, instead of steps 112 to 114, it would be possible to obtain the theoretical operating curve of the wind turbine by calculating an average value or a median value of the second operating parameter over a sliding interval of the first parameter.
[0120] [Fig.5] schematically represents the different operating points of the wind turbine forming the learning data and a theoretical curve obtained from these operating points.
[0121] The theoretical curve defines a theoretical value of the active power generated by the generator 10 as a function of a value of the wind speed.
[0122] The theoretical curve shown in [Fig.5] was obtained by calculating an average value of the active power generated by the generator 10 over a sliding interval of the wind speed. The sliding interval may, for example, have a width equal to 1 meter per second.
[0123] This learning phase 110 makes it possible to obtain a theoretical model of operation of the wind turbine which can then be used during the analysis phase 120.
[0124] Alternatively, instead of the learning phase 110, the theoretical model used may be a pre-established theoretical model.
[0125] The analysis phase 120 comprises the following steps.
[0126] According to a first step 121, monitoring data is collected.
[0127] The monitoring data collected are, for example, data which have been recorded in the real-time control and acquisition system (SC AD A), during a period of operation of the wind turbine 1, this period of operation being considered as a period of time to be analyzed.
[0128] The operating period of the wind turbine may be a period lasting several months, for example 1 year, over which the performance of the wind turbine is to be evaluated.
[0129] The monitoring data comprises a plurality of operating points of the wind turbine 1 recorded during the time period to be analyzed. Each operating point of the wind turbine 1 is defined by a given time interval i and a series of characteristic values of operating parameters of the wind turbine 1 measured during the time interval i.
[0130] The duration of a time interval is for example ten minutes.
[0131] In other words, the series of characteristic values of operating parameters of wind turbine 1 are recorded in the control and acquisition system with a constant time interval between two recordings. In this example, the time interval is 10 minutes.
[0132] In this example, the operating parameters of wind turbine 1 include a first operating parameter and a second operating parameter. The first operating parameter is the wind speed. The second operating parameter is the active power generated by the generator.
[0133] More precisely, for each operating point of the wind turbine, the series of characteristic values of operating parameters include: - a minimum value, a maximum value, an average value and a standard deviation of the first operating parameter (namely in this example, the wind speed) measured over the time interval considered, and - an average value of the second operating parameter (namely in this example, the active power generated by the generator) measured over the time interval considered.
[0134] [Fig.6] schematically represents the data recorded in the real-time control and acquisition system (SCADA) connected to wind turbine 1.
[0135] Each data line includes data defining an operating point (i = 0, 1, 2, ...).
[0136] Each operating point i of wind turbine 1 is defined by an interval of given time (identified by the start date and time), an average value of the active power measured over the time interval (ActivePower_Avg-SQL-10minAvg), an average value of the wind speed measured over the time interval (Amb_WindSpeed_Avg-SQL-10minAvg), a minimum value of the wind speed measured over the time interval (Amb_WindSpeed_Min-SQL-10minAvg), a maximum value of the wind speed measured over the time interval (Amb_WindSpeed_Max-SQL-10minAvg) and a standard deviation value of the wind speed measured over the time interval (Amb_WindSpeed_Std-SQL-10minAvg).
[0137] In this example, each time interval has a duration of 10 minutes.
[0138] [Fig.4A] schematically illustrates the monitoring data 21 recorded in the real-time control and acquisition system (SCADA), this monitoring data comprising a plurality of operating points of the wind turbine 22 to be analyzed.
[0139] Each operating point of the wind turbine 22 to be analyzed is defined by a given time interval and a series of characteristic values of operating parameters of the wind turbine 23 measured during the time interval.
[0140] According to a second step 122, for each operating point (i = 0, 1, 2, ...), a plurality of possible values of the first operating parameter is generated, as a function of the characteristic values of the first operating parameter recorded in the real-time control and acquisition system (SCADA).
[0141] More specifically, the possible values of the first operating parameter are generated such that the possible values of the first operating parameter are distributed according to a predefined probabilistic distribution function.
[0142] The probabilistic distribution function may be a pseudo-Gaussian distribution function, a pseudo-Weibull distribution function, or a uniform distribution function.
[0143] For example, a pseudo-Gaussian distribution function can be defined as follows: . where is the standard deviation and is the mean of the first parameter f ( x} — •—;=■ € operating over the time interval associated with the operating point considered.
[0144] According to a third step 123, for each operating point (i = 0, 1, 2, ...), a first series of signals representative of the evolution of a value of the first operating parameter as a function of time, over the time interval i considered, is generated.
[0145] Each signal of the first series of signals is generated by randomly drawing successive values of the first operating parameter among the possible values of the first operating parameter which were determined in the second step 122.
[0146] However, during this draw, the values of the first operating parameter drawn randomly which are not included in a predefined interval [Vvent_min; V vent_max] are eliminated.
[0147] For each operating point, the interval [Vvent_min; Vvent_max] can be equal to [Amb_WindSpeed_Min-SQL- lOminAvg; Amb_WindSpeed_Max-SQL- lOminAvg].
[0148] Each signal of the first series of signals may comprise a predefined number n of values of the first operating parameter drawn randomly, each value of the first operating parameter being associated with a given instant of the time interval.
[0149] For example, for a time interval of 10 minutes, the number n of values of the first operating parameter drawn randomly can be equal to 60, which makes it possible to generate a signal consisting of a series of values having an interval of 10 seconds between two successive values.
[0150] According to another example, for a time interval of 10 minutes, the number n of values of the first operating parameter drawn randomly can be equal to 600, which makes it possible to generate a signal consisting of a series of values having an interval of 1 second between two successive values.
[0151] The number N of signals generated may be greater than 50. For example, the number of signals in the first series of signals may be equal to 100.
[0152] This third step 123 thus makes it possible to recreate N possible scenarios for the evolution of the first operating parameter over the time interval.
[0153] [Fig.4A] schematically illustrates the N signals 24 of the first series of signals generated during step 123.
[0154] According to a fourth step 124, for each operating point (i = 0, 1, 2, ...), the first series of signals is converted into a second series of signals representative of an evolution of a value of the second operating parameter as a function of time, over the time interval considered.
[0155] Each signal of the second series of signals is obtained by converting a signal of the first series of signals by means of the theoretical operating model of the wind turbine which was previously obtained during the learning phase 110 or by another means.
[0156] The theoretical operating model of the wind turbine may be a theoretical curve defining a value of the second operating parameter as a function of a value of the first operating parameter.
[0157] In this example, the theoretical operating model is a curve defining a theoretical value of the active power generated by the generator as a function of a value of the wind speed, such as that illustrated in [Fig.5].
[0158] Each signal of the second series of signals obtained can comprise the same number n of values as the signal of the first series of signals, each value of the second operating parameter being associated with a given instant of the time interval.
[0159] This fourth step 124 makes it possible to convert the N scenarios for the evolution of the first operating parameter determined in step 123 into N corresponding scenarios for the evolution of the second operating parameter over the time interval considered.
[0160] [Fig.4A] schematically illustrates the theoretical operating model of the wind turbine 25 as well as the N signals 26 of the second series of signals generated during step 124.
[0161] According to a fifth step 125, for each operating point (i = 0, 1, 2, ...), and for each signal of the second series of signals, an estimated characteristic value of the second operating parameter over the time interval is calculated.
[0162] For example, for each signal of the second series of signals, the estimated characteristic value can be an average value of the second operating parameter calculated from the n values constituting the signal of the second series of signals.
[0163] This fifth step 125 makes it possible to obtain for each operating point (i = 0, 1, 2, ...), N estimated characteristic values of the second operating parameter over the time interval.
[0164] [Fig.7] schematically represents an average value of the characteristic values of the signals of the first series of signals, as a function of the number N of signals of the first series generated (the signals being representative of the evolution of the wind speed over the time interval).
[0165] Similarly, [Fig.8] schematically represents an average value of the characteristic values of the signals of the second series of signals, as a function of the number N of signals of the second series generated (the signals being representative of the evolution of the active power generated by the generator over the time interval).
[0166] Figures 7 and 8 show that in this example, it is possible to obtain an average value of the wind speed and an average value of the active power characteristics over 10 minutes stabilized from approximately N = approximately 100 prints of wind speed signals lasting 10 minutes.
[0167] However, in order to limit the calculation time required to implement steps 123 and 124, it is possible to choose N < 100.
[0168] According to a sixth step 126, for each operating point (i = 0, 1, 2, ...), the measured characteristic value of the second operating parameter (i.e. the characteristic value of the second operating parameter which has been recorded in the real-time control and acquisition monitoring system and which is part of the data to be analyzed) is compared with a threshold value which depends on one or more of the N estimated characteristic values of the second operating parameter over the time interval.
[0169] For example, the threshold value considered can be calculated from: - a minimum value among the N estimated characteristic values of the second operating parameter over the time interval, and / or - an average of the N estimated characteristic values of the second operating parameter over the time interval, and / or - a maximum value among the N estimated characteristic values of the second operating parameter over the time interval.
[0170] For example, the threshold value can be defined as the minimum value among the N estimated characteristic values of the second operating parameter over the time interval, from which a margin is subtracted.
[0171] In the example illustrated in [Fig.9], curve A represents a variation of a first threshold value as a function of time, during the operating period of the wind turbine.
[0172] The first threshold value is defined on each time interval, as the minimum value among the N estimated characteristic values of the second operating parameter, less a margin, the margin being equal to 2 times the standard deviation calculated on all the estimated characteristic values of the second operating parameter over the time interval.
[0173] In other words, curve A in [Fig.9] represents the worst operating scenario of the wind turbine, from which the margin is subtracted.
[0174] Curve B of [Fig.9] represents a variation of a second threshold value as a function of time, during the operating period of the wind turbine.
[0175] The second threshold value is defined on each time interval, as the maximum value among the N estimated characteristic values of the second operating parameter, plus a margin, the margin being equal to 2 times the standard deviation calculated on all the estimated characteristic values of the second operating parameter over the time interval.
[0176] In other words, curve B represents the best operating scenario of the wind turbine, to which the margin is added.
[0177] Curve C of [Fig.9] represents a variation of a third threshold value as a function of time, during the operating period of the wind turbine.
[0178] The third threshold value is defined as the average value of the N estimated characteristic values of the second operating parameter over the time interval.
[0179] In other words, curve C represents the average operating scenario of the wind turbine.
[0180] [Fig.4B] schematically illustrates the worst operating scenario of the wind turbine 27, the best operating scenario of the wind turbine 28, the average operating scenario of the wind turbine 29 and the margin o(P) applied.
[0181] Curve D of [Fig.9] represents a variation of the measured characteristic value of the second operating parameter as a function of time, during the operating period of the wind turbine.
[0182] In other words, curve D represents the actual operating scenario of the wind turbine. Curve D can be compared to curve A, curve B and / or curve C.
[0183] [Fig.4B] schematically illustrates the characteristic values of the second operating parameter 30 from the monitoring data collected during step 121. For each operating point of the wind turbine, the characteristic value of the second operating parameter 30 associated with a given time interval is compared to one or more of the threshold values associated with the same time interval.
[0184] According to a seventh step 127, for each operating point (i = 0, 1, 2, ...), a state of the wind turbine is determined based on the result of the comparison.
[0185] For example, for each operating point: - if the characteristic value of the second operating parameter of the wind turbine measured over the time interval (curve D in [Fig.9]) is greater than or equal to the first threshold value (curve A in [Fig.9]) then the operating state of the wind turbine associated with the operating point is determined to be a normal operating state, and - if the characteristic value of the second operating parameter of the wind turbine measured over the time interval (curve D in [Fig.9]) is lower than the first threshold value (curve A in [Fig.9]), then the operating state of the wind turbine associated with the operating point is determined to be an underperforming operating state.
[0186] According to an eighth step 128, it is possible to calculate a value of electrical energy loss over the operating period of the wind turbine (period to be analyzed), from all the operating points and the respective associated operating states which were determined in step 127.
[0187] Electrical energy loss is calculated as a deviation between an expected theoretical energy production value and an actual energy production value over the operating period of the wind turbine for the operating points for which the wind turbine is underperforming.
[0188] The expected theoretical energy production value can be obtained by determining for each underperforming operating point, a theoretical active power value as a function of the average wind speed value recorded in the SCADA for the operating point, using a theoretical model of wind turbine operation.
[0189] The theoretical active power values are multiplied by the duration of a time interval and then summed to derive the expected theoretical energy production value.
[0190] It is also possible to calculate an operating time of the wind turbine in each of the operating states of the wind turbine, from all the operating points and the respective associated operating states. 2nd embodiment
[0191] [Fig. 10] schematically represents steps of a method for determining the operating state of a wind turbine according to a second embodiment of the invention.
[0192] In the second embodiment, the method may have a first phase 110 (or “learning phase”) and a second phase 120 (or “analysis phase”), as in the first embodiment illustrated in [Fig.3].
[0193] However, in the case of the second embodiment, step 122 of generating a plurality of possible values of the first operating parameter and step 123 of generating a first series of signals are replaced by steps 131 to 135 illustrated in [Fig.10].
[0194] According to a first step 131, a first curve representative of the average value of the first parameter (namely, the wind speed) is generated over the period of time to be analyzed.
[0195] This first curve can be obtained by interpolation of the average values of the first parameter which were recorded in the SCADA during the period of time to be analyzed.
[0196] This step makes it possible to reconstruct from the average values of the first parameter, recorded with a first time interval (in this example 10 minutes between two consecutive operating points) between two consecutive recordings, a first low-frequency component representative of a variation in the average value of the first parameter with a second time interval (for example 30 seconds), less than the first time interval.
[0197] [Fig. 13] schematically represents a variation of the average value of the first parameter as a function of time, obtained by interpolation of the values of the first parameter recorded in the SCADA for the different operating points. In this example, the first parameter is the wind speed.
[0198] This first curve provides, for each 10-minute time interval, an estimated variation in the average value of the first parameter.
[0199] The interpolation method used for this step may be a Gaussian process interpolation method, for example a Gaussian process having a covariance function of Matém 5 / 2. This process makes it possible to randomly generate several curves passing through all the points to be interpolated, then to calculate the first curve as an average of the randomly generated curves.
[0200] According to a second step 132, a covariance function of the first parameter is calculated.
[0201] This covariance function can be calculated from real data, for example meteorological data making it possible to trace a variation in the wind speed measured over a predefined period (for example 1 or several hours), at a frequency higher than the frequency of the data recorded in the SCADA.
[0202] In this example, wind speed values were recorded with a time interval of 30 seconds between two measurements.
[0203] The covariance function may be a 5 / 2 Matérn covariance function, defined by a scale parameter and an amplitude parameter.
[0204] This covariance function represents how the value of the wind speed at a given instant is correlated with the wind speeds recorded previously.
[0205] According to a third step 133, a series of second normalized high-frequency components representative of a relative variation of the value of the first operating parameter as a function of time is generated.
[0206] Each second component is generated by a random process having the covariance function determined in step 132.
[0207] Each second component comprises successive values of the first operating parameter with a time interval of 30 seconds between two successive values.
[0208] The number N of second components generated may be greater than 50. For example, the number of N may be equal to 100.
[0209] [Fig. 11] schematically represents a second normalized component, i.e. a normalized relative variation of the value of the first parameter as a function of time, obtained by the random process. In this example, the first parameter is the wind speed. Since this is a relative variation, the average of this relative variation is zero.
[0210] According to a fourth step 134, a series of second denormalized high-frequency components is generated from the series of second high-frequency components normalized, so that for each 10-minute time interval, the value of the standard deviation of the denormalized signal is equal to the standard deviation recorded in the SCADA for the corresponding operating point.
[0211] [Fig. 12] schematically represents a denormalized signal obtained from the normalized signal of [Fig. 11].
[0212] This fourth step makes it possible to generate N second denormalized components of relative variation of the first parameter. Each second denormalized component has, for each 10-minute time interval, a standard deviation equal to the standard deviation recorded in the SCADA for the corresponding operating point.
[0213] According to a fifth step 135, a first series of signals is generated, each signal of the first series being obtained by summing the first low-frequency average component determined in step 131 and a respective second denormalized high-frequency component obtained in step 134.
[0214] [Fig. 14] schematically represents a signal obtained by summing the first component illustrated in [Fig. 13] and the second component illustrated in [Fig. 12].
[0215] This fifth step 135 thus makes it possible to recreate N possible scenarios for the evolution of the first operating parameter over each time interval.
[0216] Each signal obtained at the end of the fifth step 135 presents for each 10-minute time interval, an average and a standard deviation equal to the average and the standard deviation recorded in the SCADA for the corresponding operating point.
[0217] In the case of the second embodiment, the series of second high-frequency components is generated taking into account a covariance function calculated from real data.
[0218] Thus, the possible scenarios are generated taking into account the fact that in reality, the variations of the first operating parameter over each time interval are not totally random.
[0219] This makes it possible to obtain a first series of signals whose variations are closer to real variations of the first operating parameter.
Claims
1. Claims Method for determining operating states of a wind turbine (1), from a set of data (21) recorded in a wind turbine data acquisition and control system, the set of data (21) comprising a plurality of operating points (22) of the wind turbine, each operating point (22) of the wind turbine being defined by a time interval, one or more characteristic values of a first operating parameter of the wind turbine measured over the time interval and one or more characteristic values of a second operating parameter of the wind turbine measured over the time interval, in which, for each operating point, the method comprises steps of: - generating a first series of signals (24), each signal of the first series of signals being representative of a possible evolution of a value of the first operating parameter as a function of time, over the time interval, each signal of the first series of signals having one or more characteristic value(s) equal to the characteristic value(s) of the first operating parameter of the wind turbine measured over the time interval, - converting the first series of signals (24) into a second series of signals (26) representative of an evolution of a value of the second operating parameter as a function of time, over the time interval,each signal of the second series of signals being obtained by converting a signal of the first series of signals (24) by means of a theoretical model of operation of the wind turbine (25) making it possible to estimate a value of the second operating parameter as a function of a value of the first operating parameter, - for each signal of the second series of signals (26), calculating an estimated characteristic value of the second parameter over the time interval, - compare the characteristic value of the second operating parameter of the wind turbine measured over the time interval with the estimated characteristic value of the second operating parameter over the time interval, and - deduce an operating state of the wind turbine associated with the operating point.
2. Method according to claim 1, comprising, for each operating point, a step of: - generating a plurality of possible values of the first operating parameter, distributed according to a predefined probabilistic distribution function, as a function of the characteristic value(s) of the first operating parameter of the wind turbine, and in which the first series of signals is generated by randomly drawing successive values of the first operating parameter from among the possible values of the first operating parameter.
3. The method of claim 2, wherein the predefined probabilistic distribution function is selected from a pseudo-Gaussian distribution function, a pseudo-Weibull distribution function or a uniform distribution function.
4. Method according to one of claims 2 and 3, in which the step of generating a first series of signals representative of the evolution of a value of the first operating parameter as a function of time, over the time interval, comprises a sub-step of: - eliminating values of the first operating parameter drawn randomly which are not included in a predefined interval.
5. Method according to claim 1, comprising, for each operating point, steps of: - generating a first low frequency component representative of a variation of the first operating parameter over the time interval, and - generating a plurality of second high frequency components representative of possible relative variations of the first operating parameter over the time interval, wherein each signal of the first series of signals is generated by summing the first low frequency component and a second high frequency component of the plurality of second high frequency components.
6. Method according to claim 5, in which the first low frequency component is obtained by interpolation of characteristic values of the first operating parameter measured over several consecutive time intervals, including the time interval considered.
7. Method according to one of claims 5 and 6, in which for each operating point, the characteristic value(s) of the first operating parameter comprise an average value of the first operating parameter measured over the time interval, and in which the first low-frequency component is generated by interpolation of the different average values of the first operating parameter.
8. Method according to one of claims 5 to 7, in which each second high-frequency component is generated by a random process having a covariance function, the covariance function having been calculated beforehand on learning data comprising real values of the first operating parameter having been measured with a time step between two measurement instants less than the duration of the time interval.
9. Method according to one of claims 5 to 8, wherein for each operating point, the characteristic value(s) of the first operating parameter comprise a standard deviation value of the first operating parameter measured over the time interval, and wherein each signal of the first series of signals is generated by summing the first low-frequency component and the second high-frequency component and denormalizing the resulting signal obtained so that the denormalized signal has a standard deviation equal to the standard deviation value of the first operating parameter over the time interval.
10. Method according to one of claims 1 to 9, in which the characteristic value(s) of the first operating parameter comprise a minimum value, a maximum value, an average value and / or a standard deviation of the first operating parameter measured over the time interval.
11. Method according to one of claims 1 to 10, in which the characteristic value(s) of the second operating parameter comprise an average value of the second operating parameter measured over the time interval.
12. Method according to one of claims 1 to 11, in which the wind turbine (1) comprises a mast (2), a nacelle (3), a rotor (4), an electric current generator (12), the rotor (4) comprising a hub (5) and blades (6) mounted on the hub (5), the blades (6) being suitable for driving the rotor (4) in rotation relative to the nacelle (3) under the effect of the wind, and the electric current generator (12) being suitable for being driven in rotation by the rotor (4), and in which the first operating parameter and the second operating parameter are chosen from: - respectively a wind speed and an active power generated by the generator (12), or - respectively a pitch angle of the blades (6) relative to the hub (5) and a wind speed, or - respectively an engine torque transmitted by the rotor (4) to the generator (12) and an active power generated by the generator (12), or - respectively a wind orientation angle and an orientation angle of the nacelle (3) relative to the mast (2).
13. Method according to one of claims 1 to 12, in which the reference value of the second operating parameter is determined as: - a minimum value among the set of estimated characteristic values of the second operating parameter over the time interval, and / or - an average of the estimated characteristic values of the second operating parameter over the time interval, and / or - a maximum value among the set of estimated characteristic values of the second operating parameter over the time interval.
14. Method according to one of claims 1 to 13, wherein during the step of deducing an operating state of the wind turbine (1) associated with the operating point, the operating state of the wind turbine is selected from a first operating state and a second operating state, one of the first operating state being a normal operating state and the other of the first operating state and the second operating state being an underperforming operating state.
15. Method according to claim 14, wherein for each operating point: - if the characteristic value of the second operating parameter of the wind turbine measured over the time interval is greater than or equal to a threshold which depends on the reference value of the second operating parameter over the time interval, then the operating state of the wind turbine associated with the operating point is the first operating state, and - if the characteristic value of the second operating parameter of the wind turbine measured over the time interval is less than the threshold, then the operating state of the wind turbine associated with the operating point is the second operating state.
16. The method of claim 15, wherein the threshold is equal to the reference value of the second operating parameter plus or minus a margin.
17. The method of claim 16, wherein the margin is defined as a predefined number multiplied by a standard deviation calculated over the set of estimated characteristic values of the second operating parameter over the time interval.
18. Method according to one of claims 1 to 17, in which the model making it possible to estimate a value of the second operating parameter as a function of a value of the first operating parameter has been obtained from learning data comprising a plurality of operating points of the wind turbine, the learning data having been previously acquired and recorded over a period of operation of the wind turbine during which the wind turbine is only in a normal operating state.
19. A method according to claim 18, wherein the operating period of the wind turbine has a duration of several months, preferably ranging from 3 to 6 months.
20. Method according to one of claims 18 and 19, in which the model making it possible to estimate a value of the second operating parameter as a function of a value of the first operating parameter has been obtained by steps of: - partitioning the learning data into several successive elementary slices, each slice containing all the operating points for which the first operating parameter is included in a predefined interval associated with the slice, and - calculating, from all the operating points contained in the slice, an average value of the second operating parameter over the predefined interval associated with the slice.
21. Method according to one of claims 1 to 20, comprising a step of: - from the set of operating points and the respective associated operating states, calculating an electrical energy loss value as a difference between an expected theoretical energy production value and an energy production value real.
22. Method according to one of claims 1 to 21, comprising a step of: - from the set of operating points and the respective associated operating states, calculating an operating time of the wind turbine in each of the operating states of the wind turbine.
23. Computer program product, comprising instructions which, when the program is executed by a computer, cause the latter to implement the steps of the method according to one of claims 1 to 22.
