Prediction of a temperature parameter associated with a component of a wind turbine

Abstract

Disclosed is a method, performed by an electronic device, for predicting a temperature parameter associated with a component of a wind turbine. The wind turbine comprises a plurality of sensors associated with the component. The plurality of sensors comprises a first sensor and a second sensor. The method comprises obtaining an energy input parameter indicative of an energy dissipation of the component. The method comprises obtaining a plurality of temperature measurements comprising a first temperature measurement of the component provided by the first sensor, and a second temperature measurement of the component provided by the second sensor. The method comprises predicting, based on the energy input parameter and a cross-sensor correlation model, a temperature parameter associated with the first sensor. The cross-sensor correlation model is configured to characterize a correlation between the first temperature measurement and the second temperature measurement. Fig. 2 to be published.

Description

[0001] PREDICTION OF A TEMPERATURE PARAMETER ASSOCIATED WITH A

[0002] COMPONENT OF A WIND TURBINE

[0003] The present disclosure pertains to the field of wind turbine control. The present disclosure relates to a method for prediction of a temperature parameter associated with a component of a wind turbine and related electronic device.

[0004] BACKGROUND

[0005] A wind turbine may be operated according to readings obtained by one or more sensors of the wind turbine. For example, a wind turbine may adjust its power output based on a temperature reading obtained from a temperature sensor of the wind turbine e.g., to maximise power output while preventing overheating. However, this may leave the wind turbine vulnerable to sensor malfunctions that may result in obtention of inaccurate or erroneous readings, potentially resulting in operation of the wind turbine being based on inaccurate or erroneous data. This can for example result in unnecessary derating of the wind turbine or incurring damage to the wind turbine.

[0006] It is of benefit to provide ways of mitigating the effect of a faulty sensor reading on the operation of a wind turbine.

[0007] SUMMARY

[0008] It is an object of the present invention to improve the ability of predicting a temperature parameter, such as indicative of a temperature, associated with a component of wind turbine.

[0009] Accordingly, it would be a benefit to provide an electronic device and a method that may allow prediction of a temperature parameter associated with a component of a wind turbine. Accordingly, it would be a benefit to provide an electronic device and a method for controlling operation of a wind turbine, which mitigate, alleviate, or address the existing shortcomings and predict the temperature parameter associated with a component of a wind turbine.

[0010] Disclosed is a method, performed by an electronic device, for predicting a temperature parameter associated with a component of a wind turbine. The wind turbine comprises a plurality of sensors associated with the component. The plurality of sensors comprises a first sensor and a second sensor. The method comprises obtaining an energy input parameter indicative of an energy dissipation of the component. The method comprises obtaining a plurality of temperature measurements comprising a first temperature measurement of the component provided by the first sensor, and a second temperature measurement of the component provided by the second sensor. The method comprises predicting, based on the energy input parameter and a cross-sensor correlation model, a temperature parameter associated with the first sensor. The cross-sensor correlation model is configured to characterize a correlation between the first temperature measurement and the second temperature measurement.

[0011] The cross-sensor correlation model may further comprise a function configured to determine a temperature parameter, such as a first temperature parameter, associated with the first sensor and / or a temperature parameter, such as a second temperature parameter, associated with the second sensor based on any of the first temperature measurement and the second temperature measurement, the energy input parameter, and one or more coefficients characterizing a relation between the energy input parameter and any of the first temperature measurement and the second temperature measurement.

[0012] Each of the one or more coefficients of the cross-sensor correlation model may be common to both the first sensor and the second sensor, i.e. the same one or more coefficients provide the relationship between the energy input parameter and the first temperature measurement as well as the second temperature measurement.

[0013] The first and second temperature parameters may be seen as the predicted quantities corresponding to the first and second temperature measurements.

[0014] Other coefficients of the cross-sensor correlation model account for possible differences between the first and second temperature parameters. Such other coefficients may not be common to the first and second sensors.

[0015] Disclosed is a wind turbine controller comprising a memory circuitry, a processor circuitry, and a wireless interface, wherein the wind turbine controller is configured to perform any of the methods according to the disclosed methods.

[0016] Disclosed is a wind turbine comprising the disclosed wind turbine controller. Disclosed is a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by an electronic device cause the electronic device to perform any of the methods according to the disclosed methods.

[0017] Since the cross-sensor correlation model includes a correlation between the first temperature measurement and the second temperature measurement, predictions of the temperature parameter associated with the first sensor advantageously may also depend on historical second temperature measurements in addition to historical first temperature measurements. Therefore, for example, when the first sensor is defective, the predictions of the (first) temperature parameter associated with the first sensor are still reliable since the cross-sensor correlation model includes a correlation between the temperature parameter associated with the first sensor and second temperature measurement(s). In this way, when a faulty first sensor is detected, e.g. due to unreliable first temperature measurements, reliable temperature parameter associated with the first sensor may still be provided.

[0018] The disclosed method, the disclosed electronic device (such as the disclosed wind turbine controller) may enable the prediction and / or estimation of a temperature of the wind turbine, such as a temperature of a component of the wind turbine. This may advantageously allow for the wind turbine to be controlled based on the predicted temperature parameter, such as when a temperature sensor has been determined to provide erroneous measurements, thereby enabling an improved reliability and robustness of the operations of the wind turbine.

[0019] In other words, the present disclosure can be seen as enabling the wind turbine to reduce the likelihood of unnecessary mitigations caused by erroneous temperature measurements. It may be appreciated that the disclosed method may enable a reduction in the loss of energy production, e.g., as indicated by a Lost Production Factor (LPF), for example caused by derating of the wind turbine, incorrectly heating up oil, activating and / or deactivating fans and / or pumps incorrectly, etc.

[0020] In some examples, the disclosed method and electronic device may advantageously reduce the likelihood of the wind turbine operating in conditions prone to lead to damage of the wind turbine, e.g., due to malfunctioning sensors. For example, when a malfunctioning sensor is indicating a temperature lower than the actual temperature, the wind turbine may fail to derate in order to control the temperature. The control of the wind turbine based on predicted temperature parameter may thereby advantageously prevent damage to the wind turbine caused by operation of the wind turbine based on erroneous sensor readings. In other words, the disclosed method and electronic device may enable an increased lifecycle (e.g., lifespan and / or lifetime) of the wind turbine, such as an increased lifecycle of a component and / or a part of the wind turbine.

[0021] Furthermore, the present disclosure may enable a reduction in the frequency of maintenance required to retain satisfactory operations of the wind turbine.

[0022] BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The above and other features and advantages of the present disclosure will become readily apparent to those skilled in the art by the following detailed description of exemplary embodiments thereof with reference to the attached drawings, in which:

[0024] Fig. 1 is a graph illustrating temperature measurements from a malfunctioning sensor, Fig. 2 is a diagram illustrating schematically of an example system according to this disclosure,

[0025] Fig. 3 is a diagram illustrating schematically an example process or system where the disclosed technique for predicting a temperature parameter associated with a component of a wind turbine is carried out by an electronic device according to this disclosure, Figs. 4A-4B show a flow-chart illustrating an exemplary method, performed by an electronic device, for predicting a temperature parameter associated with a component of a wind turbine according to this disclosure,

[0026] Fig. 5 is a block diagram illustrating an exemplary electronic device according to this disclosure,

[0027] Figs. 6A-B show example curves for temperature measurements and predicted temperature parameters and status graphs of a wind turbine according to this disclosure, and

[0028] Figs. 7A-B show example curves corresponding temperature measurements and predictions and power graphs of a wind turbine in idle according to this disclosure.

[0029] DETAILED DESCRIPTION

[0030] Various exemplary embodiments and details are described hereinafter, with reference to the figures when relevant. It should be noted that the figures are only intended to facilitate the description of the embodiments. They are not intended as an exhaustive description of the disclosure or as a limitation on the scope of the disclosure. In addition, an illustrated embodiment needs not have all the aspects or advantages shown. An aspect or an advantage described in conjunction with a particular embodiment is not necessarily limited to that embodiment and can be practiced in any other embodiments even if not so illustrated, or if not so explicitly described.

[0031] A wind turbine may for example be controlled according to temperature readings obtained by one or more temperature sensors of a wind turbine. However, the temperature readings may be erroneous due to a malfunctioning sensor. For example, when a temperature sensor malfunctions, e.g. providing incorrect temperature readings, operations of the wind turbine may be controlled inaccurately (e.g. based on data that does not correspond with reality).

[0032] The inventors have determined that one of the major LPF contributor on the wind turbine is failure or malfunction of a temperature sensor associated with a component of the wind turbine. For example, the temperature sensors associated with certain components (e.g. in the gearbox, generator, and / or for cooling water) are prone to malfunction and to providing erroneous temperature measurements. The temperature sensors can be used within several modules or associated with components, for monitoring the module or temperature of the component (e.g. min / max values) to activate corrective measures (e.g., derating) when temperature measurements are outside expected ranges. When a temperature sensor fails, the temperature sensor often begins to report spurious measurements outside these ranges, forcing the turbine to perform unnecessary operations, such as derate, heating up oil, activate or deactivate fans and pumps etc, and / or schedule a repair visit from a technician. Repair visits can lead to one or more days where the wind turbine is not operational.

[0033] Fig. 1 shows a graph illustrating temperature measurements from a malfunctioning sensor associated with a component of a wind turbine according to this disclosure. Fig. 5 shows curves of the temperature measurements obtained from various sensors over time. The malfunctioning sensor provides temperature measurements of curve 5 which suddenly drop while the remaining sensors of the same group (e.g. for the same component) indicate no drop. Curve 5 provides the highest temperature before the drop. The wind turbine controller uses curve 5 to determine if additional cooling is required or if derating is needed. Since the sensor with the highest value is no longer working as shown by 5, the decision whether to increase cooling capacity or derate is based on a wrong assumption.

[0034] Fig. 2 is a diagram illustrating schematically a system 1 according to this disclosure. The system 1 comprises an electronic device 300, a wind turbine controller 30, and / or wind turbines 20, 22, 24, and 26. In some examples, the wind turbines 20, 22, 24 and / or 26 are a group of wind turbines arranged as a wind farm or a wind park. The wind turbines may be onshore wind turbine or offshore wind turbine, hereunder fixed-bottom and floating wind turbines.

[0035] The wind turbine controller 30 is for example configured to control the operations of (e.g., to operate) one or more of wind turbines 20, 22, 24 and / or 26. In some examples, the controller 30 can be seen as a central control system.

[0036] The wind turbine controller 30 may form part of one or more of the wind turbines 20, 22, 24 and / or 26. For example, the controller 30 can be part of a controller internal to one or more wind turbines 20, 22, 24 and / or 26. The wind turbine controller 30 may be external to one or more of the wind turbines 20, 22, 24 and / or 26. In one or more examples, the wind turbine controller 30 may be a part of the electronic device 300. In one or more examples, the wind turbine controller 30 and the electronic device 300 are the same entity.

[0037] The wind turbine comprises for example a component (e.g. gearbox, generator, cooling component, transformer, etc.) and a plurality of sensors associated with the component. The plurality of sensors comprises a first sensor and a second sensor. For example, the gearbox may have a plurality of temperature sensors arranged to measure temperatures of gearbox components and gearbox lubrication oil. The generator may include temperature sensors arranged to measure bearing temperatures.

[0038] The electronic device 300 is for example the electronic device configured to perform the method disclosed herein. The electronic device 300 may be configured to obtain (via link 32 and / or 34) an energy input parameter indicative of an energy dissipation of the component.

[0039] The electronic device 300 is configured to obtain (e.g. via link 32 and / or 34) a plurality of temperature measurements comprising a first temperature measurement of the component provided by the first sensor, and a second temperature measurement of the component provided by the second sensor.

[0040] The electronic device 300 is configured to predict, based on the energy input parameter and a cross-sensor correlation model, a temperature parameter associated with the first sensor. The cross-sensor correlation model is configured to characterize a correlation between the first temperature measurement and the second temperature measurement.

[0041] The electronic device 300 may provide (via link 32 and / or 34) to the one or more wind turbines (e.g. directly and / or via the wind turbine controller 30) the predicted temperature parameters, for example when a sudden significant drop of temperature is reported in the temperature measurements. In some examples, the one or more wind turbines 20, 22, 24 and / or 26 can then operate according to the predicted temperature parameter(s) instead of the temperature measurements. In some examples, the wind turbine controller 30 can control 34 the one or more wind turbines 20, 22, 24 and / or 26 according to the predicted temperature parameter(s).

[0042] In some examples, the wind turbine controller 30 is configured to act as electronic device 300 for predicting temperature of a component of a wind turbine.

[0043] The present disclosure provides a wind turbine (e.g. wind turbine 20, 22, 24, 26) comprising a controller configured to predict temperature of a component of a wind turbine according to any of the methods disclosed herein.

[0044] The present disclosure provides a method for estimating and / or predicting a temperature of a component of a wind turbine. The method obtains temperature measurements via sensors associated with the component and an energy input parameter (e.g. power) to predict, using a cross-sensor correlation model, the temperature parameter associated with the sensor. The wind turbine may then be operated by controlling the provision of temperature to further application, to replace the temperature measurement by the predicted temperature parameter, when a sensor has been determined to be providing incorrect temperature measurements.

[0045] Fig. 3 is a diagram illustrating schematically an example electronic device 300 where the disclosed technique for predicting a temperature parameter associated with a component of a wind turbine is carried out according to this disclosure. The wind turbine comprises a plurality of sensors associated with the component, wherein the plurality of sensors comprises a first sensor and a second sensor.

[0046] The electronic device 300 comprises memory circuitry, processing circuitry, and an interface 40. The memory circuitry and / or processing circuitry can be seen as configured to execute a cross-sensor correlation model 60. In some examples, the memory circuitry and / or processing circuitry comprise one or more of: an input pre-processor 50, a signal switching logic 70, an output processor 80, and a signal estimation logic 90.

[0047] The interface 40 may be configured to obtain input data and to provide the input data to the input processor 50 (such as a pre-processor). The input data comprises, for example, one or more of: an energy input parameter 42 indicative of an energy dissipation of the component, a plurality of temperature measurements 46, 48, 49, and a reference temperature 44 (such as an ambient temperature, such as ambient temperature of the nacelle of the wind turbine). For example, the plurality of temperature measurements comprises a first temperature measurement 46 (T1) of the component provided by the first sensor, and a second temperature measurement 48 (T2) of the component provided by the second sensor.

[0048] In some examples, the input pre-processor 50 applies a fixed window median filter 51 to a part of the input data, such as to the temperature measurements and the energy input parameter.

[0049] The energy input parameter can be seen as representative of the energy dissipation at the component. In some examples, the energy input parameter comprises one or more of: a power parameter, an energy capacity parameter, a torque parameter, and a rotor speed parameter.

[0050] The input processor 50 provides for example one or more of: the processed energy input parameter 52, the processed ambient temperature 54, and the temperature measurements 56 (e.g., T1, T2, ... , TN) to the cross-sensor correlation model 60.

[0051] The cross-sensor correlation model 60 can be seen as adapting a multiple-input-multiple- output (MIMO) model that estimates each of the input signals. For example, when one or more sensors fail, the disclosed technique can continue providing a temperature estimate by applying the cross-sensor correlation model 60 to the input data. The cross-sensor correlation model 60 can be seen as an adaptive model used for fault detection by monitoring the difference (specifically the difference statistics e.g., rolling mean and variance) for the individual temperature sensors.

[0052] In some examples, the cross-sensor correlation model 60 comprises an adaptive model configured to estimate, based on the energy input parameter, one or more of: model parameters of the cross-sensor correlation model, the sensor-specific coefficients, the temperature parameter. For example, the cross-sensor correlation model 60 comprises an adaptive model executed by an adaptation logic 62 for determining when to iterate on the model parameters of cross-sensor correlation model, and when to lock or keep the model parameters for an upcoming period. The cross-sensor correlation model 60 comprises for example a temperature predictor 64 and a parameter estimator 66 configured to iterate to determine the model parameters of the cross-sensor correlation model. The parameter estimator 66 is configured to estimate the model parameters, such as cross-sensor coefficients and sensor-specific coefficients.

[0053] In some examples, the parameter estimator 66 monitors a variance metric 91 of the predicted temperature parameter, e.g. to evaluate the prediction error in the predicted temperature parameters so that the adaptation logic 62 can lock the model parameters and / or iterate to update the model parameters of the cross-sensor correlation model 60. In some examples, the cross-sensor correlation model 60 is based on one or more of: a recursive non-linear estimator, a Kalman filter, and a recursive least squared estimator.

[0054] In some examples, the temperature predictor 64 predicts the temperature parameters of a group of sensors. The temperature sensors are likely to have similar behavior: the temperature measurements all increase and decrease together on the same time scale. This can be applied for specific systems like the generator or gearbox. The first sensor and the second sensor disclosed herein are seen as part of the same group in some examples.

[0055] In some examples, the nacelle temperature measurements can correlate with power production captured by the energy input parameter disclosed herein. The relationship between temperature and the energy input parameter is included in the cross-sensor correlation model to improve estimation accuracy and allow estimation even in the single- sensor case. For example, the wind turbine is in production when the energy input parameter is used as input to the temperature predictor.

[0056] The cross-sensor correlation model 60 for example provides the predicted temperature parameter(s) 69 and optionally the “raw” temperature measurements in 68 to the signal switching logic 70.

[0057] The signal switching logic 70 can use the predicted temperature parameter(s) 69 and optionally the “raw” temperature measurements in 68 to switch (in 72) from raw measurements to the predicted temperature parameter(s) 69, for example when error flags are raised. The signal switching 70 can, in 74, switch to raw measurements from the predicted temperature parameter(s) 69, e.g. with a smooth switching for example when no error is present. The signal switching 70 can generate in 76 an average temperature signal if the predicted temperature parameters are not provided by the cross-sensor correlation model.

[0058] The signal switching logic 70 can provide a temperature estimate 78 to the output processor 80 which generates a temperature output: e.g. a filtered temperature output 86 by filtering 82 and / or the min or max value 88 of the temperature output by calculation 84. The temperature output can be provided to an application.

[0059] It may be appreciated that the cross-sensor correlation model 60 takes as input temperature measurements as provided by the sensors, and can thereby not readily distinguish between erroneous temperature measurements (e.g. from a faulty sensor) and temperature measurement from a functioning sensor. In such scenario, the cross-sensor correlation model 60 would inconveniently adapt towards the values coming from erroneous temperature measurements. To mitigate such incorrect adaptation, the present disclosure provides a signal estimation logic 90 that can decide when to activate or disactivate the adaption of the cross-sensor correlation model 60. For example, the signal estimation logic 90 can indicate to the cross-sensor correlation model to keep, maintain, and / or lock the model parameters for a failing sensor until it has recovered.

[0060] The signal estimation logic 90 can take as input the temperature measurements (56, and / or 46, 48, 49) and determines whether a first criterion is met by the temperature measurements 56 and / or when a second criterion is met by the temperature measurements 56 and / or by the variance metric. The signal estimation logic 90 can monitor for when the temperature measurements do not meet a first criterion (e.g. selected from various first criteria e.g. spike(s) 92 in temperature measurements, and / or when the temperature measurements are below a first threshold or above a second threshold 94, and / or when the temperature measurements provide no value 95, dead signal. The signal estimation logic 90 can monitor for when the variance metric 91 of the cross-sensor correlation model does not meet a second criterion 96. The signal estimation logic 90 can, based on the comparison of temperature measurements with the first criterion and / or of the variance metric 91 with the second criterion, provide input 97 to the cross-sensor correlation model to update the model parameters of the cross-sensor correlation model and / or to the signal switching 70 for switching to e.g. predicted temperature parameter(s).

[0061] In other words, for example, the signal estimation logic 90 examines the incoming temperature measurements 46, 48, 49 and / or pre-processed temperature measurements 56 by checking if the temperature measurements meet a first criterion: e.g. temperature measurements too low or too high, signal spikes, signal dead over long time periods. For example, the signal estimation logic 90 examines whether a variance metric of the crosssensor correlation model meets a second criterion, e.g., to evaluate if the model error variance too high (e.g. absolute and / or relative).

[0062] For example, the signal estimation logic 90 can turn off the adaption (e.g. update) of the model parameters for a specific sensor whenever the first and / or the second is met. In some examples, the signal error and model error types are associated with a hold timer to avoid activating the adaptation immediately after an incident. An example from an actual wind turbine is illustrated in Figs. 7A-7B.

[0063] Figs. 4A-B show a flow diagram of an example method 100, performed by an electronic device, e.g., for predicting a temperature parameter associated with a component of a wind turbine. The wind turbine comprises a plurality of sensors associated with the component (e.g., a gearbox of the wind turbine, a generator of the wind turbine, a cooling water component of the wind turbine). The plurality of sensors comprise a first sensor and a second sensor. The electronic device is the electronic device disclosed herein, such as the electronic device 300 of Figs. 2, 3 and 5.

[0064] The method 100 comprises obtaining (e.g., receiving and / or generating) S102 an energy input parameter indicative of an energy dissipation of the component. In one or more example methods, the energy input parameter comprises one or more of: a power parameter, an energy capacity parameter, a torque parameter, and a rotor speed parameter. For example, the power parameter indicates power produced by the wind turbine. For example, the torque parameter indicates the torque of the main shaft of the wind turbine. For example, the rotor speed parameter indicates the speed of the rotor of the wind turbine. For example, the energy capacity parameter can be seen as quantifying an amount of energy needed for cooling and / or heating a component.

[0065] The method 100 comprises obtaining S104 a plurality of temperature measurements comprising a first temperature measurement of the component provided by the first sensor, and a second temperature measurement of the component provided by the second sensor. The temperature measurements can comprise as “raw” temperature measurements obtained from sensors of the wind turbine and / or pre-processed (e.g., filtered using a fixed window median) temperature measurements. In some examples, the sensors, such as the first sensor and / or the second sensors, are temperature sensor such as thermocouples.

[0066] The method 100 comprises predicting S106, based on the energy input parameter and a cross-sensor correlation model, a temperature parameter associated with the first sensor, and optionally with a first time slot. The temperature parameter can be referred to as a first temperature parameter predicted for the first sensor, e.g. for the first time slot. In one or more example methods, the method 100 comprises predicting, based on the energy input parameter and the cross-sensor correlation model, a second temperature parameter associated with the second sensor, and optionally with the first time slot. For example, the second temperature parameter is predicted for the second sensor, e.g. for the first time slot. In one or more example methods, the temperature parameter comprises a temperature relative to a reference temperature. In other words, for example, the temperature parameter (of the first sensor, and / or of the second sensor) is relative to the reference temperature, such as ambient temperature of the environment of the component (e.g. nacelle temperature). Stated differently, the temperature parameter disclosed herein can be seen a relative temperature, such as a difference in temperature with respect to the reference temperature, such as a delta temperature above ambient (e.g. nacelle) temperature. This allows accounting for seasonal and / or daily patterns, such as the daily temperature cycle. The cross-sensor correlation model is configured to characterize a correlation between the first temperature measurement and the second temperature measurement. The crosssensor correlation model can be seen as characterizing the relation between the energy input parameter (e.g. power) and temperature, across sensors associated with the same component. An example of the cross-sensor correlation model is illustrated in Fig. 3. In some examples, the cross-sensor correlation model comprises model parameters for characterizing (e g. modelling) the correlation between the first temperature measurement and the second temperature measurement, such as the relation between the energy input parameter (e.g. power) and temperature, across sensors associated with the same component. In one or more example methods, the cross-sensor correlation model comprises a non-linear function based on a first coefficient characterizing a square relation between the energy input parameter and any of the first temperature measurement and the second temperature measurement. In one or more examples, the cross-sensor correlation model comprises a non-linear function based on a first coefficient characterizing a square relation between the energy input parameter and any of the first temperature parameter and the second temperature parameter. In one or more example methods, the cross-sensor correlation model comprises a non-linear function based on a second coefficient characterizing a linear relation between the energy input parameter and any of the first temperature measurement and the second temperature measurement. In one or more example methods, the cross-sensor correlation model comprises a nonlinear function based on a second coefficient characterizing a linear relation between the energy input parameter and any of the first temperature parameter and the second temperature parameter. In one or more example methods, the cross-sensor correlation model comprises a non-linear function based on the first coefficient (e.g., a of Equations (2)(3)) and the second coefficient (e.g., b of Equations (2)(3)), and optionally a common bias term.

[0067] In one or more example methods, predicting S106, based on the energy input parameter and the cross-sensor correlation model, the temperature parameter comprises predicting S106A the temperature parameter based on a previous temperature measurement provided by the first sensor (e.g. Tn k-of Equation (1)). For example, the previous temperature measurement is a temperature measurement provided in the preceding time slot by the first sensor. In one or more example methods, predicting, based on the energy input parameter and the cross-sensor correlation model, the temperature parameter comprises predicting S106B the temperature parameter based on the energy input parameter, the cross-sensor correlation model, and sensor-specific coefficients (e.g. cnand dnof Equation (2)). For example, the temperature parameter for the first sensor is predicted based on based on the energy input parameter, the cross-sensor correlation model, the previous temperature measurement provided by the first sensor and the sensor-specific coefficients. Stated differently, for example, the cross-sensor correlation model is based on the non-linear function that takes as input the energy input parameter, and parameterized with the first coefficient, the second coefficient. The prediction of the temperature parameter of the first sensor can be further based on the previous temperature and the sensor-specific coefficients.

[0068] In some examples, the relation between the energy input parameter (e.g. power) and temperature measurement (e.g. the first temperature measurement and / or the second temperature measurement)) used in the prediction can be seen as a first-order system combined with a nonlinear function: where: x denotes predictor variable: the energy input parameter (e.g., one or more of: power, energy capacity, torque, and rotor speed).

[0069] Tn kdenotes the temperature of the 'th sensor at the fc'th timestep (e.g., k’th time slot)

[0070] 8t denotes the model timestep (e.g., fixed parameter)

[0071] T denotes the first-order time constant (e.g., fixed parameter)

[0072] Tn(x) denotes nonlinear energy input parameter (e.g. power) to temperature measurement function (see Equation (2))

[0073] The nonlinear energy input parameter (e.g. power) to temperature function can be expressed as e.g.: fn(x) = (a • x2+ b • x) • cn+ dn+ e (2) where: a denotes the first coefficient (e.g. quadratic coefficient) describing the energy input parameter to temperature measurement relation; b denotes the second coefficient (e.g. linear coefficient) describing the energy input parameter to temperature measurement relation; cndenotes a first sensor coefficient (e.g. gain term) describing the sensitivity of the n'th sensor; dndenotes a second sensor coefficient (e.g., relative offset, bias term) describing the relative offset of the n'th sensor; e denotes a common bias term.

[0074] In one or more example methods, the first coefficient and the second coefficient are common to the first sensor and the second sensor. In other words, the first coefficient and the second coefficient are the same across sensors, e.g., shared between the sensors. For example, the model parameters a, b, and e are shared across sensors, while the parameters cnand dnare unique for each sensor.

[0075] In some examples, Equation (2) can be simplified by excluding the sensor-specific coefficients, and fit into a second order polynomial function, e.g.:

[0076] T = (a • x2+ b • x) + e (3)

[0077] Equations (2) and / or (3) can be seen as a baseline model for the cross-sensor correlation model that relates power to temperature variations. In one or more examples, the crosssensor correlation model comprises an adaptive model for estimating the model parameters. For example, the adaptive model aims at converging to a shared power to temperature relation across sensors. In some examples, the sensor-specific coefficients cnand dndescribe the slightly different characteristics that each sensor inhibits to changes in power.

[0078] In one or more examples, the prediction provides a first temperature parameter and a second temperature parameter which can be expressed as e.g.: f2( ) = (a • x2+ b • x) • c2+ d2+ e

[0079] For examples, model parameters a, b, c1c2are adaptively determined in the model. For example, the cross-sensor parameter a, b, are determined based on the first temperature measurement and the second temperature measurement, then a, b each depends on historical values of the first temperature measurement and the second temperature measurement. In other words, for example, the predicted temperature parameter for the first sensor and the estimate ^(x) depends on the first temperature measurement and the second temperature measurement via the parameters a,b. Thus, since the parameters a, b of the cross-sensor correlation model includes a correlation between the first temperature measurement and the second temperature measurement, the cross-sensor correlation model is able to predict, e.g. Tltbased on the energy input parameter x, even when measurements of the first temperature sensor are incorrect due to a faulty first temperature sensor.

[0080] It may be appreciated that the predicted temperature parameter for the first sensor (used when the first sensor is defect) are more reliable as the cross-sensor correlation model includes a common model parameter and is still able to predict the predicted temperature parameter for the first sensor based on a reliable second temperature measurements (in addition to the energy input parameter).

[0081] It may be appreciated that the difference in predicting the temperature for the first sensor and for the second sensor is captured by the sensor-specific coefficients c2.

[0082] In one or more example methods, the cross-sensor correlation model comprises an adaptive model configured to estimate, based on the energy input parameter, one or more of: model parameters of the cross-sensor correlation model, the sensor-specific coefficients, the temperature parameter.

[0083] In one or more example methods, the adaptive model comprises a temperature predictor and a parameter estimator. In one or more examples, the temperature predictor is configured to predict the temperature parameter. In one or more examples, the parameter estimator is configured to estimate the model parameters of the cross-sensor correlation model, and the sensor-specific coefficients. In one or more example methods, the adaptive model is one or more of: a recursive nonlinear estimator, a Kalman filter, and a recursive least squared estimator. In some examples, the adaptive model is a least mean squares algorithm and / or other method for adaptive determination of model parameters, such as the model parameters a, b, c2.

[0084] For example, the model parameters are estimated adaptively using a Kalman filter, such as an Extended Kalman Filter (EKF) as parameter estimator. In other words, for example, the EKF uses a formulation of the model described in Equations (1) and (2). For example, the model has a state vector, X, and input vector, U, which are described as follows e.g.: X =Ti>Tz> - . Tn. a. b. c^ c^, .... c^ d^ d^ ... , dn, e]T(4)

[0085] U = x (5)

[0086] For example, the state vector comprises the predicted temperature parameters (from the temperature predictor) and the parameter estimates (from the parameter estimator). The temperature predictor and parameter estimator estimate the temperature parameters and

[0087] It may be envisaged that model parameters (e.g., ak, bk, ek) are locked (e.g. kept) as constant when satisfactory according to the second. It may be envisaged that sensorspecific coefficients (e.g., cn fc, dn k) are constant. The temperature predictor and parameter estimator estimate the temperature parameters and model parameters with noise terms for each state, following e.g.:

[0088] The noises terms may be assumed to be drawn from a zero mean multivariate normal distribution with a specific covariance, Q.

[0089] In some examples, the adaptive model (e.g. EKF) can be evaluated based on uncertainty parameters (e.g. variance metrics) for the estimated variables (e.g. predicted temperature parameters), which determines how quickly the values of the temperature parameters are adapted. For example, by selecting a smaller variance metric, the estimated variables (e.g. predicted temperature parameters) adapt more slowly. For example, the extended Kalman filter assumes the true state evolves according to a nonlinear model, with the assumption that the covariance can be propagated using a locally linear approximation. For example, the local linear model is written in form e.g.: k = Fk^k-l + Bku(8)

[0090] In some examples, Equation (6) can be converted in the form of Equation (8) by linearization, where Fkand Bkare Jacobians in relation to the states and inputs, respectively. For example, the F matrix is referred to as the state transition Jacobian. It may be appreciated that the F matrix is sufficient to perform the EKF update equations. For example, for a system with n = 3 sensors, the F matrix can be constructed as e.g.:

[0091] For example, the first and second coefficients a, b and the common bias term e have nonzero entries in all rows, highlighting the cross-coupled (e.g. cross correlation) effect between temperature state estimates.

[0092] The prediction disclosed herein allows for the model parameters for the cross-sensor correlation model to be static for a period (e.g. after being selected by tuning the variance metric, e.g. by trial and error), while adapting the sensor-specific temperatures more quickly.

[0093] In one or more example methods, the method 100 comprises monitoring S108, using the parameter estimator, a variance metric of the predicted temperature parameter(s), e.g. in comparison with the temperature measurement(s). The variance metric for example quantifies the uncertainty (or confidence) in the predicted temperature parameter(s). For example, the variance metric can be determined based on a predicted first temperature parameter for the first sensor and a first temperature measurement from the first sensor. For example, the variance metric can include a variance estimate and / or a covariance estimate. For example, the variance metric is evaluated periodically to consider for update of the model parameters. The monitoring can take place for example in the signal estimation logic 90 of Fig. 3.

[0094] In one or more example methods, the method 100 comprises determining S110 whether the variance metric meets a second criterion. For example, the method 100 comprises determining S110 whether the variance metric meets the second criterion (e.g. below a third threshold). In other words, the variance metric meets the second criterion when the variance metric shows a satisfactory variance, showing limited prediction error.

[0095] In one or more example methods, the method 100 comprises updating S112 the model parameters of the cross-sensor correlation model, upon determining that the variance metric does not meet the second criterion. For example, when the variance metric does not meet the second criterion (e.g. equal or above the third threshold), the model parameters are updated, e.g. using the parameter estimator until the variance metric falls below the third threshold. For example, once the variance metric monitored meets the second criterion, the model parameters are locked or kept for a time period until the variance metric does not meet the second criterion.

[0096] In one or more example methods, the method 100 comprises refraining S114 from updating the model parameters of the cross-sensor correlation model, upon determining that the variance metric meets the second criterion. For example, when the variance metric meets the second criterion (e.g. equal or above the third threshold), the model parameters are not updated.

[0097] In one or more example methods, the sensor-specific coefficients comprise a first sensor coefficient indicative of gain of the first sensor, and a second sensor coefficient indicative of a relative offset of the first sensor with respect to the second sensor. For example, the first sensor coefficient may be seen as a gain parameter for sensitivity of the sensor. For example, the sensor-specific coefficients can be determined by applying constraints. In one or more example methods, the first sensor coefficient is based on a minimum relative offset value across the plurality of sensors. For example, the first sensor coefficient is selected to be the minimum sensor-specific bias to the baseline term: mm dr, d2, ..., dn) = e (10)

[0098] For example, the constraint applied to the first sensor coefficient ensures that the first sensor coefficient converges to the minimal temperature measurement, as opposed to [— oo, oo] .

[0099] In one or more example methods, the second sensor coefficient is based on a normalized gain value across the plurality of sensors. The normalization of the gain value can be performed using e.g.: mean(c1,c2, ..., cn) = 1 (11)

[0100] For example, the constraint applied to the second sensor coefficient ensures that the model parameters, such as polynomial parameters a and b, and the sensor-specific gains cnhave a unique solution with a correct scale.

[0101] In one or more example methods, the method 100 comprises determining S116 whether the first temperature measurement and / or the second temperature measurement meet a first criterion. For example, the method 100 comprises determining S116, using the signal estimation logic 90 of Fig. 3, whether the first and / or second temperature measurements meet the first criterion. For example, the first criterion can include one or more of first criteria: the (first and / or second) temperature measurements are above a first threshold, the obtained (first and / or second) temperature measurements are below a second threshold, the obtained (first and / or second) temperature measurements providing signal spikes or no signal for long periods. The first criterion can be selected from various first criteria where the first criterion is not met when e.g.: the obtained (first and / or second) temperature measurements are above a first threshold, the obtained (first and / or second) temperature measurements are below a second threshold, and / or the obtained (first and / or second) temperature measurement providing signal spikes or no signal for long periods.

[0102] In one or more example methods, the method 100 comprises: upon determining S116 that the first temperature measurement and / or the second temperature measurement do not meet the first criterion, replacing S118 the temperature measurements by the first sensor with the predicted temperature parameter of the first sensor. For example, when the first and / or second temperature measurements does not meet the first criterion, the temperature measurement(s) are replaced by the corresponding predicted temperature parameter(s).

[0103] It may be envisaged that the method 100 comprises refraining S114 from updating the model parameters of the cross-sensor correlation model, upon determining that the first / second temperature measurements do not meet the first criterion. For example, the model parameters are not updated when the first and / or second temperature measurements do not meet the first criterion. The first criterion can be selected from various first primary criteria where the first criterion is not met when e.g.: the obtained (first and / or second) temperature measurements are above a first threshold indicative of a maximum value, the obtained (first and / or second) temperature measurements are below a second threshold indicative of a minimum value, and / or the obtained (first and / or second) temperature measurement providing signal spikes or no signal for long periods. This way, the model parameters are not adapted to temperature measurements that are from faulty sensors.

[0104] In one or more example methods, the method 100 comprises: upon determining that the first temperature measurement and / or the second temperature measurement meet the first criterion, forgoing S120 the replacement of the first and / or second temperature measurement(s) with the predicted temperature parameter.

[0105] Fig. 5 shows a block diagram of an exemplary electronic device 300 according to the disclosure. The electronic device 300 comprises memory circuitry 301, processor circuitry 302, and an interface 303. The electronic device 300 is configured to perform any of the methods disclosed in Figs. 4A-B. In other words, the electronic device 300 is configured for predicting a temperature parameter associated with a component of a wind turbine. For example, the electronic device 300 may be a wind turbine controller of a wind turbine, such as an internal controller of the wind turbine. For example, the electronic device may be a device remote from the wind turbine. For example, the electronic device may be an external controller of the wind turbine. Electronic device 300 can be seen as a generic temperature estimator device. The electronic device 300 can reduce the likelihood of erroneous corrective actions taken unnecessarily. The electronic device 300 is configured to obtain (e.g., via processor circuitry 302 and / or interface 303) an energy input parameter indicative of an energy dissipation of the component.

[0106] The electronic device 300 is configured to obtain (e.g., via processor circuitry 302 and / or interface 303) a plurality of temperature measurements comprising a first temperature measurement of the component provided by the first sensor, and a second temperature measurement of the component provided by the second sensor.

[0107] The electronic device 300 is configured to predict (e.g., via processor circuitry 302), based on the energy input parameter and a cross-sensor correlation model, a temperature parameter associated with the first sensor. The cross-sensor correlation model is configured to characterize a correlation between the first temperature measurement and the second temperature measurement.

[0108] The processor circuitry 302 is optionally configured to perform any of the operations disclosed in Fig. 4A-B (such as any one or more of: S102, S104, S106, S106A, S106AA, S106B, S108, S110, S112, S114, S116, S118). The operations of the electronic device 300 may be embodied in the form of executable logic routines (e.g., lines of code, software programs, etc.) that are stored on a non-transitory computer readable medium (e.g., the memory circuitry 301) and are executed by the processor circuitry 302).

[0109] Furthermore, the operations of the electronic device 300 may be considered a method that the electronic device 300 is configured to carry out. Also, while the described functions and operations may be implemented in software, such functionality may as well be carried out via dedicated hardware or firmware, or some combination of hardware, firmware and / or software.

[0110] The memory circuitry 301 may be one or more of a buffer, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, a random-access memory (RAM), or other suitable device. In a typical arrangement, the memory circuitry 301 may include a non-volatile memory for long term data storage and a volatile memory that functions as system memory for the processor circuitry 302. The memory circuitry 301 may exchange data with the processor circuitry 302 over a data bus. Control lines and an address bus between the memory circuitry 301 and the processor circuitry 302 also may be present (not shown in Fig. 5). The memory circuitry 301 is considered a non-transitory computer readable medium. The memory circuitry 301 may be configured to store energy input parameter(s), temperature measurements, model parameters and criteria in a part of the memory.

[0111] Figs. 6A-B show example curves for temperature measurements and predicted temperature parameters for various sensors associated with the same component vs. time and status graphs of a wind turbine vs. time according to this disclosure.

[0112] Fig. 6A shows curve 601 of temperature measurements for a first sensor, curve 602 of temperature measurements for a second sensor, curve 603 of temperature measurements for a third sensor, and curve 604 of predicted temperature parameters for the third sensor. Fig. 6B shows the status 606 of the third sensor over time and failure score 607.

[0113] As illustrated, when the temperature measurement from the third sensor begins to give erroneous measurements, the fault is detected (in curve 606) and the predicted temperature of 604 is used instead of faulty measurements of 603. The disclosed electronic device (such as wind turbine controller) can keep operating on the predicted temperature parameter until the temperature measurements are satisfactory or valid again or until the sensors issue has been fixed at a scheduled service.

[0114] The present disclosure provides a prediction model for a wind turbine that is idle, such as inactive. For example, when the wind turbine enters Idle, the energy input parameter (e.g. power) cannot be used for predicting temperature parameters. In one or more examples, the temperature parameter is predicted using a temperature decay towards ambient temperature with a first-order differential equation (discrete form) e.g.:

[0115] Where:

[0116] Tn fedenotes the temperature parameter of the n'th sensor at the k'th timestep,

[0117] 8t denotes the model timestep, f denotes a slope term,

[0118] Todenotes steady-state temperature in Idle - assuming Toto be the ambient temperature. Fig. 7A shows a graph illustrating example temperature measurements and example predicted temperature parameters for an idle wind turbine. Fig. 7B show corresponding temperature and power graphs of a wind turbine according to this disclosure. Figs. 7A-B show the measured temperature (curve 506) of one of the generator phases together with the predicted temperature parameter (curve 507). As illustrated whenever there is no power, the temperature parameter simply decays towards the ambient (e.g. nacelle) temperature with a first-order differential equation.

[0119] It should further be noted that any reference signs do not limit the scope of the claims, that the exemplary embodiments may be implemented at least in part by means of both hardware and software, and that several "means", "units" or "devices" may be represented by the same item of hardware.

[0120] The various exemplary methods, devices, nodes, and systems described herein are described in the general context of method steps or processes, which may be implemented in one aspect by a computer program product, embodied in a computer- readable medium, including computer-executable instructions, such as program code, executed by computers in networked environments. A computer-readable medium may include removable and non-removable storage devices including, but not limited to, Read Only Memory (ROM), Random Access Memory (RAM), compact discs (CDs), digital versatile discs (DVD), etc. Generally, program circuitries may include routines, programs, objects, components, data structures, etc. that perform specified tasks or implement specific abstract data types. Computer-executable instructions, associated data structures, and program circuitries represent examples of program code for executing steps of the methods disclosed herein. The particular sequence of such executable instructions or associated data structures represents examples of corresponding acts for implementing the functions described in such steps or processes.

[0121] It is to be noted that the term "indicative of" may be seen as “associated with”, “related to”, “descriptive of’, “characterizing”, and / or “defining”. The terms “indicative of’, “associated with” “related to”, “descriptive of’, “characterizing”, and “defining” can be used interchangeably. The term “indicative of” can be seen as indicating a relation. For example, weight data indicative of weight may comprise one or more weight parameters.

[0122] It is to be noted that the word "based on" may be seen as “as a function of’ and / or “derived from”. The terms “based on” and “as a function of” can be used interchangeably. For example, a parameter determined “based on” a data set can be seen as a parameter determined “as a function of” the data set. In other words, the parameter may be an output of one or more functions with the data set as an input.

[0123] A function may be characterizing a relation between an input and an output, such as mathematical relation, a database relation, a hardware relation, logical relation, and / or other suitable relations.

[0124] Although features have been shown and described, it will be understood that they are not intended to limit the claimed disclosure, and it will be made obvious to those skilled in the art that various changes and modifications may be made without departing from the scope of the claimed disclosure. The specification and drawings are, accordingly, to be regarded in an illustrative rather than restrictive sense. The claimed disclosure is intended to cover all alternatives, modifications, and equivalents.

Claims

CLAIMS1. A method, performed by an electronic device, for predicting a temperature parameter associated with a component of a wind turbine, wherein the wind turbine comprises a plurality of sensors associated with the component, wherein the plurality of sensors comprises a first sensor and a second sensor, the method comprising: obtaining (S102) an energy input parameter indicative of an energy dissipation of the component; obtaining (S104) a plurality of temperature measurements comprising a first temperature measurement of the component provided by the first sensor, and a second temperature measurement of the component provided by the second sensor; and predicting (S106), based on the energy input parameter and a cross-sensor correlation model, a temperature parameter associated with the first sensor; wherein the cross-sensor correlation model is configured to characterize a correlation between the first temperature measurement and the second temperature measurement.

2. The method according to claim 1 , wherein the cross-sensor correlation model comprises a function based on one or more coefficients characterizing a relation between the energy input parameter and any of the first temperature measurement and the second temperature measurement, wherein each of the one or more coefficients are common to the first sensor and the second sensor.

3. The method according to claim any of the previous claims, wherein the energy input parameter comprises one or more of: a power parameter, an energy capacity parameter, a torque parameter, and a rotor speed parameter.

4. The method according to any of the previous claims, the method comprising: determining (S116) whether the first temperature measurement and / or the second temperature measurement meet a first criterion; and upon determining that the first temperature measurement and / or the second temperature measurement do not meet the first criterion, replacing (S118) the temperature measurements by the first sensor with the predicted temperature parameter of the first sensor.

5. The method according to any of the previous claims, wherein the temperature parameter comprises a temperature relative to a reference temperature.

6. The method according to any of the previous claims, wherein the cross-sensor correlation model comprises a non-linear function based on a first coefficient characterizing a square relation between the energy input parameter and any of the first temperature measurement and the second temperature measurement, and / or a second coefficient characterizing a linear relation between the energy input parameter and any of the first temperature measurement and the second temperature measurement.

7. The method according to claim 6, wherein the first coefficient and the second coefficient are common to the first sensor and the second sensor.

8. The method according to any of the previous claims, wherein predicting (S106), based on the energy input parameter and the cross-sensor correlation model, the temperature parameter comprises predicting (S106A) the temperature parameter based on a previous temperature measurement provided by the first sensor.

9. The method according to any of the previous claims, wherein predicting (S106), based on the energy input parameter and the cross-sensor correlation model, the temperature parameter comprises predicting (S106B) the temperature parameter based on the energy input parameter, the cross-sensor correlation model, and sensor-specific coefficients.

10. The method according to any of the previous claims, wherein the cross-sensor correlation model comprises an adaptive model configured to estimate, based on the energy input parameter, one or more of: model parameters of the cross-sensor correlation model, the sensor-specific coefficients, the temperature parameter.

11. The method according to claim 10, wherein the adaptive model comprises a temperature predictor and a parameter estimator.

12. The method according to any of claims 10-11 , wherein the adaptive model is one or more of: a recursive non-linear estimator, a Kalman filter, and a recursive least squared estimator.

13. The method according to any of claims 11-12, the method comprising:- monitoring (S108), using the parameter estimator, a variance metric of the predicted temperature parameter.

14. The method according to claim 13, the method comprising:- determining (S110) whether the variance metric meets a second criterion;- updating (S112) the model parameters of the cross-sensor correlation model, upon determining that the variance metric does not meet the second criterion; and- refraining (S114) from updating the model parameters of the cross-sensor correlation model, upon determining that the variance metric meets the second criterion.

15. The method according to any of claims 7-14, wherein the sensor-specific coefficients comprise a first sensor coefficient indicative of gain of the first sensor, and a second sensor coefficient indicative of a relative offset of the first sensor with respect to the second sensor.

16. The method according to claim 15, wherein the first sensor coefficient is based on a minimum relative offset value across the plurality of sensors.

17. The method according to any of claims 15-16, wherein the second sensor coefficient is based on a normalized gain value across the plurality of sensors.

18. A wind turbine controller comprising a memory circuitry, a processor circuitry, and a wireless interface, wherein the wind turbine controller is configured to perform any of the methods according to any of claims 1-17.

19. A wind turbine comprising a wind turbine controller according to claim 18.

20. A computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by an electronic device cause the electronic device to perform any of the methods of claims 1-17.

Images