Systems and methods for operating a wind turbine

Abstract

A system and method for operating and maintaining a wind turbine are provided. Accordingly, a plurality of data inputs are received. The plurality of data inputs represent a plurality of monitored attributes of components of the wind turbine. The plurality of monitored attributes are used to determine a comprehensive risk index for the component and to predict a range of potential risk indices. A remaining useful life distribution is determined based on a damage potential and an end-of-life damage threshold. If the remaining useful life distribution is below a shutdown threshold, the wind turbine is shut down or idled.

Description

Technical Field

[0001] The present disclosure generally relates to wind turbines, and more particularly to systems and methods for operating a wind turbine based on a predicted range of potential risk indices. Background Art

[0002] Wind power is considered to be one of the cleanest and most environmentally friendly energy sources currently available, and wind turbines have received increased attention in this regard. Modern wind turbines typically include a tower, a generator, a gearbox, a nacelle, and one or more rotor blades. The nacelle includes a rotor assembly coupled to the gearbox and the generator. The rotor assembly and the gearbox are mounted on a platen support frame located within the nacelle. One or more rotor blades use known airfoil principles to obtain the kinetic energy of the wind. The rotor blades transfer the kinetic energy in the form of rotational energy to rotate a shaft that couples the rotor blades to the gearbox (or directly to the generator if no gearbox is used). The generator then converts the mechanical energy into electrical energy, and the electrical energy can be transmitted to a converter and / or transformer housed within the tower and subsequently deployed to the utility grid. Modern wind power generation systems typically take the form of a wind farm having a plurality of such wind turbine generators that are operable to supply power to a transmission system that supplies power to the electrical grid.

[0003] Typically, wind turbines are designed to operate at a rated power output over a predetermined or expected service life. For example, a typical wind turbine may be designed for a 20-year life. In many instances, the expected overall operating life may be limited based on the expected fatigue life of one or more wind turbine components. Additionally, the operational use of a component can result in a wear rate of the component that is different from the expected rate. Thus, a component monitoring system can be used to detect anomalies within various components of a wind turbine, such as a gearbox.

[0004] Typically, the component monitoring system is configured to detect component failures in a diagnostic mode. Since component failures can occur at any time, maintenance actions and parts planning are typically based on prior knowledge and past field observations rather than on quantifiable predictions. Thus, a wind farm operator typically has two options for maintaining a wind turbine. The operator can schedule preventive maintenance at or before the expected service life of a component. When a maintenance action cannot be performed due to wear of a component, replacing the component can result in a significant increase in cost. Alternatively, the operator can seek to delay maintenance activities. This can result in an unexpected failure of a component and cause the wind turbine to shut down. Since the shutdown can be unexpected, it can be prolonged due to the availability of replacement parts, or it may be necessary to maintain an excessive quantity of parts in inventory. Additionally, an unexpected failure of a component can result in inefficient maintenance actions at the wind farm, such as multiple visits by a ground crane, resulting in increased costs.

[0005] Accordingly, there is a continuing quest in the art for new and improved systems for operating and maintaining wind turbines. Accordingly, the present disclosure relates to systems and methods for controlling a wind turbine based on a predicted remaining useful life distribution. SUMMARY OF THE INVENTION

[0006] Aspects and advantages of the invention will be set forth in part in the following description, or may be obvious from the description, or may be learned by practice of the invention.

[0007] In one aspect, the present disclosure relates to a method for operating a wind turbine. The method may include receiving, by a controller, a plurality of data inputs from at least one source. The plurality of data inputs may represent a plurality of monitored attributes of a component of the wind turbine. The method may further include determining, by the controller, a comprehensive risk index for the component using the plurality of monitored attributes. The risk index may define a deviation from a nominal behavior of the component. The method may further include predicting, by the controller, a range of potential risk indices evolving from the comprehensive risk index over a defined plurality of component cycles. The range of potential risk indices may be related to a historical wind turbine dataset and may be defined between an upper confidence interval and a lower confidence interval. Each potential risk index may be associated with a damage potential. The method may additionally include determining, by the controller, a remaining useful life distribution based on the damage potential and a life-ending damage threshold. Further, the method may include stopping or idling the wind turbine if the remaining useful life distribution is below a shutdown threshold.

[0008] In an embodiment, receiving the plurality of data inputs may further include receiving, by the controller, a plurality of time series data inputs from at least one sensor configured to monitor the component in operation.

[0009] In a further embodiment, determining the comprehensive risk index for the component may include downselecting, by the controller, the plurality of time series data inputs to establish a plurality of monitored attributes. The method may further include applying, by the controller, an upper normal limit and a lower normal limit to each of the plurality of monitored attributes. The upper normal limit and the lower normal limit may define a range of data inputs consistent with a baseline health component. The upper normal limit and the lower normal limit may be based on a historical wind turbine dataset. The method may further include detecting, by the controller, a deviation of at least one of the monitored attributes from the range defined by the normal limits. Additionally, the method may include defining, by the controller, a squared prediction error for each of the plurality of monitored attributes. The squared prediction error may reflect the amount of deviation of the corresponding monitored attribute from the normal limit of interest. Further, the method may include summing, by the controller, the squared prediction errors for each of the plurality of monitored attributes in order to calculate the comprehensive risk index for the component. A non-zero value may indicate operation outside of the component normal limits.

[0010] In additional embodiments, the method may further include defining a risk threshold, wherein defining the risk index threshold balances early fault detection with the likelihood of false alarms. The method may further include: prior to predicting a range of potential risk indices, detecting that the risk index threshold is crossed by a combined risk index for the component.

[0011] In an embodiment, predicting a range of potential risk indices may include selecting, for each of a plurality of wind turbines, a historical risk index data set from a historical fleet turbine data set over a defined number of component cycles. The method may include using a controller to determine an average fleet risk index over the defined number of component cycles. Additionally, the method may include using the controller to determine a covariance of the historical risk index data set for each of the plurality of wind turbines relative to the average fleet risk index. The method may include using the controller to model a range of potential risk indices for the component, such as via a Karhunen Loeve Expansion. The method may further include using the controller to determine an optimal fit risk index series for the component, such as via Markov Chain Monte Carlo sampling. Additionally, the method may include using the controller to determine a 95% confidence band for the prediction.

[0012] In an embodiment, when the risk index for each of the plurality of wind turbines crosses a predetermined threshold, the defined number of component cycles may be set to zero.

[0013] In an embodiment, the method may further include using a controller to convert the range of potential risk indices to a damage potential via a correlation between damage and the risk index. The correlation between damage and the risk index may be determined by performing a regression on a historical fleet turbine failure data set. The historical fleet turbine failure data set may depict a given damage value by inspection and the recorded risk index for the component at the time of inspection.

[0014] In additional embodiments, determining the remaining useful life distribution may further include establishing a life - termination damage threshold based on a damage level where the likelihood of catastrophic failure of the component or secondary damage to the wind turbine exceeds an acceptable limit. The method may include using a controller to determine the number of prediction cycles required to reach the life - termination damage threshold for each sample of the sampling. The method may further include using the controller to combine the determined number of prediction cycles for each sample in order to produce a remaining useful life distribution between an upper confidence interval and a lower confidence interval. The upper confidence interval and the lower confidence interval may define a 95% confidence band.

[0015] In an embodiment, the method may further include using a controller to determine an interpolation of the component cycles for a specified time interval based on a historical operation data set regarding a wind turbine. The method may include using the controller to predict a distribution of the remaining useful life over time via a ratio.

[0016] In a further embodiment, the method may further include using the controller to generate an output table. The output table may indicate the probability of component failure within each of a plurality of time intervals.

[0017] In an embodiment, the method may include performing an inspection of a component and assigning a damage level corresponding to the observed degree of damage of the component. The method may include providing the damage level to the controller. Additionally, the method may include using the controller to determine a difference between the damage level and a predicted damage level based on a risk index. The method may further include using the controller to refine a model for predicting a range of potential risk indices with respect to a defined plurality of component cycles. The refinement may be based on a determined difference between a graded observed degree of damage and a predicted damage potential based on a risk index.

[0018] In yet another embodiment, the controller may be a field controller and the wind turbine may be one of a plurality of wind turbines in a wind farm.

[0019] In an embodiment, the method may include using the field controller to group maintenance activities regarding each of the wind turbines based on a distribution of the remaining useful life within a specified time interval. Additionally, the method may include using the field controller to generate a maintenance plan for the wind farm. The maintenance plan may be calculated to maximize the maintenance performed during maintenance operations while minimizing premature maintenance operations.

[0020] In an additional embodiment, the method may include using the field controller to re - distribute at least a portion of the power generation demand from a wind turbine to at least one other wind turbine in the wind farm of the wind turbine. The method may further include alternating an idle period of the wind turbine with an active power generation period in order to reduce the number of component cycles per unit time. Reducing the number of component cycles per unit time may delay the approach to a shutdown threshold. Delaying the approach to the shutdown threshold may facilitate grouping repair activities regarding each of the wind turbines.

[0021] In another aspect, the present disclosure relates to a system for operating and maintaining a wind turbine. The system may include a sensor operably coupled to a component of the wind turbine to detect an attribute of the component. The system may also include a controller communicatively coupled to the sensor. The controller may include at least one processor configured to perform a plurality of operations. The plurality of operations may include receiving a plurality of data inputs from the sensor. The plurality of data inputs represent a plurality of monitored attributes of the component. The plurality of operations may also include determining, via a risk index module, a risk index regarding the component using the plurality of monitored attributes. The risk index may define a deviation from a nominal behavior of the component. The plurality of operations may also include predicting, via a prediction module, a range of potential risk indices evolving from the comprehensive risk index within a defined plurality of component cycles. The range of potential risk indices may be related to a historical wind turbine dataset and defined between an upper confidence interval and a lower confidence interval. Each potential risk index may be related to a damage potential. The plurality of operations may include determining a remaining useful life distribution based on the damage potential and an end-of-life damage threshold. The plurality of operations may also include shutting down or idling the wind turbine if the remaining useful life distribution is below a shutdown threshold.

[0022] Technical solution 1. A method for operating a wind turbine, the method comprising:

[0023] Receiving, by a controller, a plurality of data inputs from at least one source, the plurality of data inputs representing a plurality of monitored attributes of a component of the wind turbine;

[0024] Using, by the controller, the plurality of monitored attributes to determine a comprehensive risk index regarding the component, the risk index defining a deviation from a nominal behavior of the component;

[0025] Predicting, by the controller, a range of potential risk indices evolving from the comprehensive risk index within a defined plurality of component cycles, wherein each potential risk index is related to a damage potential;

[0026] Determining, by the controller, a remaining useful life distribution based on the damage potential and an end-of-life damage threshold; and

[0027] Shutting down or idling the wind turbine if the remaining useful life distribution is below a shutdown threshold.

[0028] Technical solution 2. The method according to technical solution 1, wherein receiving the plurality of data inputs further comprises:

[0029] Receiving, by the controller, a plurality of time series data inputs from at least one sensor configured to monitor the component during operation.

[0030] Solution 3. The method according to Solution 2, wherein determining the comprehensive risk index for the component further includes:

[0031] Using the controller to select the plurality of time series data inputs downward to establish the plurality of monitoring attributes;

[0032] Using the controller to apply an upper normal limit and a lower normal limit to each of the plurality of monitoring attributes, the upper normal limit and the lower normal limit defining a data input range consistent with the baseline of a healthy component, the upper normal limit and the lower normal limit being based on a historical unit turbine data set;

[0033] Using the controller to detect a deviation of at least one of the monitoring attributes from the range defined by the normal limits;

[0034] Using the controller to define a prediction error for each of the plurality of monitoring attributes, wherein the prediction error reflects the amount of deviation of the corresponding monitoring attribute from the corresponding normal limit;

[0035] Using the controller to square the prediction error for each of the plurality of monitoring attributes; and

[0036] Using the controller to sum the squared prediction errors for each of the plurality of monitoring attributes in order to calculate the comprehensive risk index for the component, wherein a non-zero value indicates operation outside the normal limits for the component.

[0037] Solution 4. The method according to Solution 3, wherein the method further includes:

[0038] Defining a risk index threshold, wherein defining the risk index threshold balances the likelihood of early fault detection and false alarms; and

[0039] Before predicting the potential risk index range, detecting that the risk index threshold is crossed by the comprehensive risk index for the component.

[0040] Solution 5. The method according to Solution 1, wherein predicting the potential risk index range further includes:

[0041] Selecting a historical risk index data set for each of the plurality of wind turbines from a historical unit turbine data set within the defined plurality of component cycles;

[0042] Using the controller to determine the average unit risk index within the defined plurality of component cycles;

[0043] Using the controller to determine the covariance of the historical risk index data set for each of the plurality of wind turbines with respect to the average unit risk index;

[0044] Model the potential risk index range for the component using the controller;

[0045] Determine the best - fit risk index series for the component using the controller; and

[0046] Determine the confidence band for the potential risk index range using the controller.

[0047] Technical solution 6. The method according to technical solution 5, wherein when the risk index for each of the plurality of wind turbines crosses a predetermined threshold, the defined plurality of component cycles are set to zero.

[0048] Technical solution 7. The method according to technical solution 1, wherein the method further comprises:

[0049] Convert the potential risk index range to the damage potential using the controller via the correlation between damage and the risk index, wherein the correlation between damage and the risk index is determined by performing a regression on a historical wind turbine failure dataset, wherein the historical wind turbine failure dataset depicts the damage values given by inspections and the recorded risk indices for the component at the time of inspection.

[0050] Technical solution 8. The method according to technical solution 1, wherein determining the remaining useful life distribution further comprises:

[0051] Establish the end - of - life damage threshold based on a damage level at which the probability of failure of the component or secondary damage to the wind turbine exceeds an acceptable limit;

[0052] Determine, using the controller, the number of prediction periods required to reach the end - of - life damage threshold for each sampled sample; and

[0053] Combine, using the controller, the determined number of prediction periods for each sample so as to produce a remaining useful life distribution between an upper confidence interval and a lower confidence interval.

[0054] Technical solution 9. The method according to technical solution 1, wherein the method further comprises:

[0055] Determine, using the controller, the interpolation of the component cycles for a specified time interval based on a historical operation dataset of the wind turbine; and

[0056] Predict, using the controller, the distribution of the remaining useful life over time via a ratio.

[0057] Technical solution 10. The method according to technical solution 9, wherein the method further comprises:

[0058] Generate an output table using the controller, the output table indicating the probability of component failure within each of a plurality of time intervals.

[0059] Technical solution 11. The method according to technical solution 1, wherein the method further comprises:

[0060] Perform an inspection of the component;

[0061] Assign a damage level corresponding to the observed degree of damage to the component;

[0062] Provide the damage level to the controller;

[0063] Use the controller to determine the difference between the damage level and the predicted damage level based on the risk index; and

[0064] Use the controller to improve a model for predicting a range of potential risk indices relative to the defined plurality of component cycles, wherein the improvement is based on the determined difference between the classified observed degree of damage and the predicted damage potential based on the risk index.

[0065] Technical solution 12. The method according to technical solution 1, wherein the controller is a field controller, and the wind turbine is one of a plurality of wind turbines in a wind farm.

[0066] Technical solution 13. The method according to technical solution 12, wherein the method further comprises:

[0067] Use the field controller to group maintenance activities for each of the wind turbines based on the remaining useful life distribution of each of the plurality of wind turbines within a specified time interval; and

[0068] Use the field controller to generate a maintenance plan for the wind farm.

[0069] Technical solution 14. The method according to technical solution 13, wherein the method further comprises:

[0070] Use the field controller to reallocate at least a portion of the power generation demand from the wind turbine to at least one other wind turbine in the wind farm; and

[0071] Alternate idle periods of the wind turbine with active power generation periods to reduce the number of component cycles per unit time, wherein reducing the number of component cycles per unit time delays the approach to the shutdown threshold to facilitate grouping repair activities for each of the wind turbines.

[0072] Technical solution 15. A system for operating a wind turbine, comprising:

[0073] at least one sensor operably coupled to a component of the wind turbine to detect an attribute of the component; and

[0074] a controller communicatively coupled to the at least one sensor, the controller including at least one processor configured to perform a plurality of operations, the plurality of operations including:

[0075] receiving a plurality of data inputs from the at least one sensor, the plurality of data inputs representing a plurality of monitored attributes of the component,

[0076] determining a comprehensive risk index for the component via a risk index module using the plurality of monitored attributes, the risk index defining a deviation from a nominal behavior of the component,

[0077] predicting, via a prediction module, a range of potential risk indices evolving from the comprehensive risk index within a defined plurality of component cycles, wherein each potential risk index is associated with a damage potential,

[0078] determining a remaining useful life distribution based on the damage potential and a end-of-life damage threshold, and

[0079] if the remaining useful life distribution is below a shutdown threshold, shutting down or idling the wind turbine.

[0080] Technical solution 16. The system according to technical solution 15, wherein determining a comprehensive risk index for the component via the risk index module further includes:

[0081] downselecting a plurality of time series data inputs via a filtering module to establish the plurality of monitored attributes;

[0082] applying an upper normal limit and a lower normal limit to each of the plurality of monitored attributes via the filtering module, the upper normal limit and the lower normal limit defining a range of data inputs consistent with a baseline of a healthy component, the upper normal limit and the lower normal limit being based on a historical fleet turbine data set;

[0083] detecting, via the risk index module, a deviation of at least one of the monitored attributes from the range defined by the normal limits;

[0084] defining, via the risk index module, a prediction error for each of the plurality of monitored attributes, wherein the prediction error reflects an amount of deviation of a corresponding monitored attribute from a corresponding normal limit;

[0085] Square the prediction error for each of the plurality of monitored attributes via the risk index module; and

[0086] Sum the squared prediction errors for each of the plurality of monitored attributes via the risk index module to calculate a comprehensive risk index for the component, where a non-zero value indicates operation outside the normal limits for the component.

[0087] Aspect 17. The system according to Aspect 15, wherein predicting the potential risk index range via the prediction module further includes:

[0088] Select a historical risk index data set for each of the plurality of wind turbines from a historical unit turbine data set within a defined plurality of component cycles, wherein when the risk index for each of the plurality of wind turbines crosses a predetermined threshold, the defined plurality of component cycles is set to zero;

[0089] Determine the average unit risk index within the defined plurality of component cycles;

[0090] Determine the covariance of the historical risk index data set for each of the plurality of wind turbines relative to the average unit risk index;

[0091] Model the potential risk index range for the component;

[0092] Determine the best fit risk index series for the component; and

[0093] Determine the confidence band for the potential risk index range.

[0094] Aspect 18. The system according to Aspect 15, wherein the plurality of operations further includes:

[0095] Convert the potential risk index range to the damage potential via the correlation between damage and the risk index, wherein the correlation between damage and the risk index is determined by performing a regression on a historical unit turbine failure data set, wherein the historical unit turbine failure data set depicts the damage values given by inspections and the recorded risk indices for the component at the time of inspection.

[0096] Aspect 19. The system according to Aspect 15, wherein determining the remaining useful life distribution further includes:

[0097] Establish a life termination damage threshold based on a damage level at which the probability of failure of the component or secondary damage to the wind turbine exceeds an acceptable limit;

[0098] Determine the number of prediction cycles required to reach the end-of-life damage threshold for each sample regarding sampling; and

[0099] Utilize the controller to combine the determined number of prediction cycles for each sample so as to generate a remaining useful life distribution between an upper confidence interval and a lower confidence interval.

[0100] Aspect 20. The system according to aspect 19, wherein the plurality of operations further includes:

[0101] Determine an interpolation of component cycles to a specified time interval based on a historical operation data set regarding the wind turbine;

[0102] Predict the distribution of the remaining useful life over time via a ratio; and

[0103] Generate an output table that indicates the probability of component failure within each of a plurality of time intervals.

[0104] With reference to the following description and the appended claims, these and other features, aspects, and advantages of the present invention will become better understood. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present invention and, together with the description, serve to explain the principles of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] A complete and enabling disclosure of the present invention (including its best mode) for a person of ordinary skill in the art is set forth in the specification with reference to the drawings, in which:

[0106] Figure 1 A perspective view of an embodiment of a wind turbine in accordance with the present disclosure is shown;

[0107] Figure 2 An internal perspective view of an embodiment of a nacelle of a wind turbine in accordance with the present disclosure is shown;

[0108] Figure 3 A schematic view of an embodiment of a wind farm having a plurality of wind turbines in accordance with the present disclosure is shown;

[0109] Figure 4 A schematic view of an embodiment of a controller in accordance with the present disclosure is shown;

[0110] Figure 5 A schematic view of an embodiment of the control logic of a system for operating and maintaining a wind turbine in accordance with the present disclosure is shown;

[0111] Figure 6 A schematic view of another embodiment of the control logic of a system for operating and maintaining a wind farm in accordance with the present disclosure is shown;

[0112] Figure 7 Show Figure 5 A graphical representation of a part of the control logic, particularly showing a system for determining an integrated risk index for a component according to the present disclosure;

[0113] Figure 8 A graphical representation showing the predicted range of potential risk indices according to the present disclosure;

[0114] Figure 9 A graphical representation showing the correlation between component damage and component risk index during replacement according to the present disclosure; and

[0115] Figure 10 A graphical representation showing the remaining service life distribution according to the present disclosure.

[0116] The repeated use of reference signs in this specification and the drawings is intended to represent the same or similar features or elements of the present invention. Detailed Description

[0117] Reference will now be made in detail to embodiments of the present invention, one or more examples of which are shown in the drawings. Each example is provided by way of explanation of the present invention and not a limitation thereof. In fact, it will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the scope or spirit thereof. For example, features shown or described as part of one embodiment can be used with another embodiment to yield still further embodiments. Accordingly, it is intended that the present invention cover such modifications and variations as come within the scope of the appended claims and their equivalents.

[0118] As used herein, the terms "first", "second", and "third" may be used interchangeably to distinguish one component from another and are not intended to denote the position or importance of each component.

[0119] Unless otherwise specified herein, the terms "coupled", "attached to", etc. denote direct coupling, fixing, or attachment as well as indirect coupling or attachment through one or more intermediate components or features.

[0120] As used throughout this specification and the claims, approximating language is applicable to modify any quantitative expression that can permissibly vary without causing a change in the basic function to which it relates. Thus, a value modified by one or more terms such as "about", "approximately", and "substantially" is not limited to the precise value specified. In at least some instances, the approximating language may correspond to the precision of the instrument for measuring the value, or the precision of the method or machine for constructing or manufacturing the component and / or system. For example, the approximating language may represent within a 10% margin.

[0121] Here and throughout the specification and claims, unless the context or language indicates otherwise, range limitations are combined and interchanged, such ranges are identified and include all sub-ranges subsumed therein. For example, all ranges disclosed herein include the endpoints, and the endpoints can be combined independently of each other.

[0122] Generally, the present disclosure relates to systems and methods for operating and maintaining a wind turbine. In particular, the present disclosure can include systems and methods for predicting the remaining useful life of a component once a first indication of a problem is detected. In particular, the present disclosure can include receiving sensor information indicative of a plurality of attributes of a component. In some instances, the amount of information received can be large. The information can be filtered or downselected to establish a plurality of monitored attributes. When data is received indicating that at least one monitored attribute has deviated from a normal limit, the deviations of each monitored attribute can be combined to determine an overall risk index for the component. The risk index can define a deviation from the nominal or expected behavior of the component. A controller can use the risk index and a historical wind turbine dataset to predict a range of potential risk indices during future component cycles. Based on historical data, the range of potential risk indices can be correlated with a range of failure potential. Thus, an end-of-life failure threshold can be established, and a remaining useful life distribution can be determined. The remaining useful life distribution can indicate the probability of reaching the end-of-life failure threshold within various accumulations of component cycles or time intervals.

[0123] It should be appreciated that predicting the remaining useful life of a component can reduce unplanned maintenance due to catastrophic failures, which can be caused by a direct failure of the component or its consequences. At the same time, the remaining useful life distribution can facilitate obtaining the maximum useful life from a component by allowing the component to continue to be used until component failure is predicted. This can preclude premature replacement of components having a long remaining useful life. Additionally, having a predicted component failure window can reduce the inventory requirements for replacement parts and can also reduce potential delays due to part availability in the event of an unexpected component failure. When extended to the wind farm level, knowledge of the predicted remaining useful life of each wind turbine in a wind farm can facilitate the comprehensive planning of maintenance activities. For example, significant cost savings can be achieved if all maintenance requiring a particular ground crane can be performed simultaneously rather than scheduling multiple visits of the ground crane.

[0124] Now referring to the figures, Figure 1A perspective view of one embodiment of a wind turbine 100 in accordance with the present disclosure is shown. As shown, the wind turbine 100 generally includes a tower 102 extending from a support surface 104, a nacelle 106 mounted on the tower 102, and a rotor 108 coupled to the nacelle 106. The rotor 108 includes a rotatable hub 110 and at least one rotor blade 112 coupled to and extending outwardly from the hub 110. For example, in the illustrated embodiment, the rotor 108 includes three rotor blades 112. However, in alternative embodiments, the rotor 108 may include more or fewer than three rotor blades 112. Each rotor blade 112 may be spaced about the hub 110 to facilitate rotation of the rotor 108 such that kinetic energy can be converted from wind into useful mechanical energy and subsequently into electrical energy. For example, the hub 110 may be rotatably coupled to a generator 118 ( Figure 2 ) positioned within the nacelle 106 to permit generation of electrical energy.

[0125] The wind turbine 100 may also include a controller 200 centralized within the nacelle 106. However, in other embodiments, the controller 200 may be located within any other component of the wind turbine 100 or at a location external to the wind turbine. Additionally, the controller 200 may be communicatively coupled to any number of components of the wind turbine 100 to control the components. Thus, the controller 200 may include a computer or other suitable processing unit. Accordingly, in several embodiments, the controller 200 may include suitable computer-readable instructions that, when implemented, cause the controller 200 to be configured to perform various different functions, such as receiving, transmitting, and / or executing wind turbine control signals.

[0126] Now referring to Figure 2 , a simplified internal view of one embodiment of the nacelle 106 of the wind turbine 100 shown in Figure 1 is shown. As shown, the generator 118 may be coupled to the rotor 108 for generating electrical power from the rotational energy generated by the rotor 108. For example, as shown in the illustrated embodiment, the rotor 108 may include a rotor shaft 122 that is coupled to the hub 110 for rotation therewith. The rotor shaft 122 may be rotatably supported by a main bearing 144. The rotor shaft 122 may in turn be rotatably coupled to a high-speed shaft 124 of the generator 118 by a gearbox 126 that is connected to a platen support frame 136 by one or more torque arms 142. As generally understood, the rotor shaft 122 may provide a low-speed high-torque input to the gearbox 126 in response to rotation of the rotor blades 112 and the hub 110. The gearbox 126 may then be configured to convert the low-speed high-torque input into a high-speed low-torque output to drive the high-speed shaft 124 and thus the generator 118. In an embodiment, the gearbox 126 may be configured with multiple gear ratios to produce varying rotational speeds of the high-speed shaft for a given low-speed input and vice versa.

[0127] Each rotor blade 112 may further include a pitch control mechanism 120 configured to rotate each rotor blade 112 about its pitch axis 116. The pitch control mechanism 120 may include a pitch controller 150 configured to receive at least one pitch setpoint command from the controller 200. Similarly, the wind turbine 100 may include one or more yaw drive mechanisms 138 communicatively coupled to the controller 200, where each yaw drive mechanism 138 is configured to change the angle of the nacelle 106 relative to the wind (e.g., by engaging the yaw bearing 140 of the wind turbine 100).

[0128] Still referring to Figure 2 , one or more sensors 156, 158, 160 may be provided on the wind turbine 100 to monitor the performance of the wind turbine 100 and / or environmental conditions affecting the wind turbine 100. For example, the sensors 158, 160 may be elements of a component monitoring system configured to monitor the properties of components of the wind turbine 100. In at least one embodiment, the component may be the gearbox 126, and the one or more sensors 158, 160 may be positioned to detect vibrations of the gearbox 126. The sensors 158, 160 may include any number of sensors known to those of ordinary skill in the art, such as accelerometers, strain gauge sensors, velocity sensors, laser displacement sensors, and / or microphones. The component monitoring system may include a plurality of sensors 158, 160 mounted at various locations on, in, or around the component to be monitored, where each sensor measures a single property of the component. In an alternative embodiment, the component monitoring system may use a reduced number of sensors and may detect properties over a wide frequency spectrum, such as vibrations. The outputs of the sensors 158, 160 may then be filtered to extract portions of the detected frequency spectrum that may indicate specific properties to be monitored. As a result, a single sensor 158, 160 may deliver multiple data inputs to the controller 200. In at least one embodiment, the component monitoring system may deliver over one hundred sensor inputs to the controller 200. It should also be understood that, as used herein, the term "monitor" and its variants indicate that the various sensors of the wind turbine 100 may be configured to provide either a direct measurement or an indirect measurement of the parameter being monitored. Thus, the sensors described herein may be used, for example, to generate a signal related to the parameter being monitored, which signal may then be used by the controller 200 to determine the condition of the wind turbine 100.

[0129] Now referring to Figure 3, a schematic diagram showing an embodiment of a wind farm 152 controlled by a system and method according to the present disclosure. As shown, the wind farm 152 may include a plurality of wind turbines 100 and a controller 200 described herein. For example, as shown in the illustrated embodiment, the wind farm 152 may include twelve wind turbines 100. However, in other embodiments, the wind farm 152 may include any other number of wind turbines 100, such as fewer than twelve wind turbines 100 or more than twelve wind turbines 100. In one embodiment, the controller 200 of the turbine 100 may be communicatively coupled to the farm controller 202 via a wired connection (such as by connecting the controller 200 via a suitable communication link 154 (e.g., a suitable cable)). Alternatively, the controller 200 may be communicatively coupled to the farm controller 200 via a wireless connection, such as by using any suitable wireless communication protocol known in the art. Additionally, for each of the individual wind turbines 100 within the wind farm 152, the farm controller 200 may be configured generally similarly to the controller 200.

[0130] In several embodiments, the wind turbines 100 of the wind farm 152 may include one or more sensors, such as any of sensors 156, 158, 160, for monitoring various operating data of the wind turbine 100 and / or one or more wind parameters of the wind farm 152. For example, as shown, the sensors may include an environmental sensor 156 configured to collect data indicative of at least one environmental condition. Thus, in an embodiment, the environmental sensor 156 may be, for example, a wind vane, an anemometer, a lidar sensor, a thermometer, a barometer, or other suitable sensor. The data collected by the environmental sensor 156 may include measures of wind speed, wind direction, wind shear, gusts, wind veer, atmospheric pressure, and / or temperature. In at least one embodiment, the environmental sensor 156 may be mounted to the nacelle 106 at a downwind position of the rotor 108. It should be understood that the environmental sensor 156 may include a sensor network and may be located away from the turbine 100. It should be understood that the environmental conditions can vary significantly throughout the wind farm 152. Thus, the environmental sensor 156 may allow for local environmental conditions at each wind turbine 100, such as local wind speed, to be monitored individually by the respective turbine controller 200 and collectively by the farm controller 200.

[0131] Now refer to Figures 4 - 10 , a schematic diagram and graphical representation showing multiple embodiments of a system 300 for operating and maintaining a wind turbine according to the present disclosure. As Figure 4FIG. 0 schematically illustrates an example of suitable components that may be included within controller 200. For example, as shown, controller 200 may include one or more processors 206 and associated memory devices 208 configured to perform a variety of computer-implemented functions (e.g., execute methods, steps, calculations, etc., and store relevant data as disclosed herein). Additionally, controller 200 may further include a communication module 210 to facilitate communication between controller 200, 202 and various components of turbine 100. Further, communication module 210 may include a sensor interface 212 (e.g., one or more analog-to-digital converters) to allow signals transmitted from one or more sensors 156, 158, 160 to be converted into signals understandable and processable by processor 206. It should be understood that sensors 156, 158, 160 may be communicatively coupled to communication module 210 using any suitable means. For example, as Figure 4 shown in FIG., sensors 156, 158, 160 are coupled to sensor interface 212 via a wired connection. However, in other embodiments, sensors 156, 158, 160 may be coupled to sensor interface 212 via a wireless connection (such as by using any suitable wireless communication protocol known in the art). Additionally, communication module 210 may also be operatively coupled to an operating state control module 214 configured to change at least one wind turbine operating state.

[0132] As used herein, the term “processor” not only refers to integrated circuits conventionally considered to be included within a computer in the art, but also refers to controllers, microcontrollers, microcomputers, programmable logic controllers (PLCs), application specific integrated circuits, and other programmable circuits. Additionally, memory device 208 may generally include memory elements including, but not limited to: computer-readable media (e.g., random access memory (RAM)), computer-readable non-volatile media (e.g., flash memory), floppy disks, compact disk read-only memory (CD ROM), magneto-optical disk (MOD), digital versatile disk (DVD), and / or other suitable memory elements. Such memory device 208 may generally be configured to store suitable computer-readable instructions that, when implemented by processor 206, cause controller 200 to be configured to perform various functions including, but not limited to: detecting a current condition approaching a current subordinate limit and affecting the speed of generator 118 so as to change the rotor-stator balance of generator 118 such that the current subordinate limit is not exceeded and wind turbine 100 may operate at rated power, as described herein, and various other suitable computer-implemented functions.

[0133] Specifically referring to Figure 5, in an embodiment, the controller 200 of the system 300 may be configured to receive a plurality of data inputs 302 from at least one source, such as sensors 158, 160. The plurality of data inputs 302 may represent a plurality of monitored attributes 304 of components of the wind turbine 100. The controller 200 may use the plurality of monitored attributes 304 at 306 via a risk index module 203 to determine a comprehensive risk index 308 for the component. The risk index 308 may define a deviation from the nominal behavior of the component. The controller 200 may predict at 310, via a prediction module 204, a range of potential risk indices 312 relative to a defined plurality of component cycles 314. The range of potential risk indices 312 may be related to a historical turbine dataset of the unit and defined between an upper confidence interval and a lower confidence interval. Each potential risk index may be associated with a damage potential 316. The controller 200 may also determine at 318 a remaining useful life distribution 320 based on the damage potential 316 and a end-of-life damage threshold 322. Additionally, if the remaining useful life distribution 320 is below a shutdown threshold, the controller 200 may cause the wind turbine 100 to shut down or idle at 324.

[0134] Specifically referring to Figure 5 and Figure 7 , in an embodiment, the plurality of data inputs 302 may include a plurality of time series data inputs from sensors 158, 160. The time series data inputs 302 may include data indicative of the health of the components in operation. For example, sensors 158, 160 may be configured to detect vibrations, speeds, and / or temperatures of the components in operation. In at least one embodiment, a plurality of sensors 158, 160 may be employed to detect various attributes of the components. In additional embodiments, a reduced number of sensors 158, 160 may be employed to detect data over a wide range. The data may then be filtered or otherwise processed to provide data to the controller 200 that highlights specific attributes of the components. For example, the output from sensors 158, 160 may be filtered such that the data inputs 302 include a number of narrow frequency bands that highlight specific attributes of the components in operation.

[0135] As indicated at 326, in at least one embodiment, the controller 200 may be configured to down-select a plurality of time series date inputs 302 via the filtering module 205 to establish a plurality of monitored attributes 304 of components of the wind turbine 100. In an embodiment, for each time interval, the plurality of data inputs 302 may include more than 50 data inputs. The more than 50 data inputs 302 may be down-selected to 10 or fewer (e.g., 5 or fewer) data inputs 302. In an embodiment, the down-selection may be based on historical data that indicates that a particular frequency, temperature, or other sensor output may indicate damage occurring within the component. In at least one embodiment, the down-selection may be accomplished by selecting irrelevant data inputs 302 via data transformation. For example, the transformation may include principal component analysis, independent component analysis, or any other suitable transformation.

[0136] In an embodiment, the controller 200 may apply an upper normal limit 328 and a lower normal limit 330 to each of the plurality of monitored attributes 304 at 325. The upper normal limit 328 and the lower normal limit 330 may define a range of data inputs 302 that is consistent with a baseline of a healthy component. The upper normal limit 328 and the lower normal limit 330 may be based on a historical wind turbine dataset. The historical wind turbine dataset may include data derived from the correlation of identification data related to replacement parts with data recorded from sensors 158, 160 prior to part replacement. For example, the historical wind turbine dataset may indicate that for a healthy component, a first amplitude of vibration at a specified frequency is normal, while a second amplitude of vibration at the specified frequency indicates wear or other damage to the component.

[0137] In an embodiment, the controller 200 may detect at 332 that at least one monitored attribute 304 deviates from the range 334 defined by the normal limits 328, 330. The monitored attribute 304 read may cross one of the normal limits 328, 330 at the detection point 336.

[0138] In a further embodiment, the controller 200 may define a prediction error 344 for each of the plurality of monitored attributes 304. The prediction error 340 may reflect the amount of deviation of the corresponding monitored attribute 304 from the corresponding normal limits 328, 330. Thus, the controller 200 may square and sum the prediction errors 340 for each attribute of the plurality of monitored attributes 304 at 338 in order to calculate a comprehensive risk index 308 for the component. A non-zero value of the risk index 308 may indicate operation outside of the normal limits 328, 330 for the component. Similarly, a risk index 308 of zero is an indication that the component is operating within the normal limits 328, 330 for the plurality of monitored attributes 304.

[0139] In at least one embodiment, system 300 may further include a risk index threshold 342. The risk index threshold 342 may be defined to balance early fault detection with the likelihood of false alarms. Additionally, in an embodiment, controller 200 may be configured to record but filter any combined risk index that does not exceed the risk index threshold.

[0140] With particular reference to Figure 5 and Figure 8 , in an embodiment, controller 200 of system 300 may be configured at 310 to predict a range of potential risk indices 312 relative to a defined plurality of component cycles 314. Thus, at 344, when the risk index threshold 342 is exceeded, the number of component cycles recorded for the wind turbines that make up the plurality of historical wind turbine data sets is zero for the plurality of wind turbines 346a-d. It should be understood that setting the cycle counter to zero when the risk index crosses the risk index threshold 342 may allow for consistent timing of series data from all unit turbines.

[0141] Still referring to 310, in an embodiment, system 300 may include selecting, at 346, a historical risk index data set for each of the plurality of wind turbines 346a-d from the historical unit turbine data sets within the defined plurality of component cycles 314. In an embodiment, the plurality of wind turbines 346a-d may be selected by determining which of the turbine units having non-zero risk indices within the defined plurality of component cycles 314 remain in operation. In additional embodiments, other criteria such as turbine type, gearbox type, maintenance date, etc. may also be used for the unit turbine data sets.

[0142] Still referring to 310, at 348, controller 200 may calculate an average unit risk index 350 within the defined plurality of component cycles 314. As shown at 352, controller 200 of system 300 may determine the covariance of the historical risk index data sets for each of the plurality of wind turbines 346a-d relative to the average unit risk index 350.

[0143] As also depicted at 310, at 354, controller 200 may model the range of potential risk indices 312 for the components, e.g., via a Karhunen - Loève expansion or other suitable modeling or sampling method. In such embodiments, the Karhunen - Loève expansion may be represented as follows:

[0144]

[0145] where RI corresponds to the risk index for a particular implementation, is the average unit risk index 350, λ and are the eigenvalues and eigenfunctions of the covariance matrix, and are uncorrelated random variables.

[0146] In an embodiment, uncorrelated random variables may be calculated at 356 by determining an optimal fit series 358 via, for example, Markov Chain Monte Carlo sampling. It should be understood that Markov Chain Monte Carlo is a technique for estimating statistical expectations in complex models via simulation. The controller 200 may also determine at 360 a confidence band 362 regarding the prediction, such as a 95% confidence level. The confidence band 362 may be defined between an upper confidence interval 364 and a lower confidence interval 366. It should be understood that all samples and confidence bands may be used to generate a prediction range for the potential risk index 312, and the range of the potential risk index 312 indicates the expected series of the combined risk index 308 when accumulated over additional cycles by component operation. It should also be understood that as the number of prediction cycles increases, the range of the potential risk index falling within the confidence band will also increase.

[0147] Specifically referring to Figure 5 and Figure 9 In an embodiment, the controller 200 of the system 300 may be configured at 368 to perform a regression on a historical unit turbine failure data set 370. The historical unit turbine failure data set 370 depicts damage values 372 given by an inspection and the risk index 308 recorded regarding the component at the time of the inspection. The inspection may be done when replacing the component. It should be understood that the damage values 372 may be given by the operator and / or the controller 200 using artificial intelligence.

[0148] In an embodiment, the regression may produce a regression line 374 within confidence limits 376, such as 95% confidence limits. The regression line 374, expressed as an equation, may provide a correlation between damage and the risk index. The correlation between damage and the risk index may provide an indication of the expected amount of physical damage to the component at the recorded or predicted risk index. Thus, at 378, the controller 200 may convert the potential risk index range into a damage potential 316 via the damage-risk index correlation.

[0149] Specifically referring to Figure 5 and Figure 10, in an embodiment, the controller 200 of the system 300 can be configured to determine a remaining useful life distribution 320 at 318 based on a damage potential 316 and an end-of-life damage threshold 322. Thus, the controller 200 can be configured at 380 to establish the end-of-life damage threshold 322. In an embodiment, the end-of-life damage threshold 322 can be based on a damage level at which the likelihood of a catastrophic failure of a component or secondary damage to the wind turbine 100 exceeds an acceptable limit. It should be understood that the acceptable limit can be determined by the wind turbine operator and can depend at least in part on the operator's risk tolerance. For example, in an embodiment, the end-of-life damage threshold 322 can be set at a damage level where the likelihood of a catastrophic failure is greater than 50% or greater than 80%. In yet another embodiment, the operator can choose to reduce the risk of catastrophic failure at the expense of more frequent component replacement by commanding that the likelihood of a catastrophic failure be less than 30%.

[0150] As also depicted at 318, the controller 200 can determine, at 382, the number of prediction cycles required for each sample of a Markov chain Monte Carlo sampling between the upper confidence interval 364 and the lower confidence interval 366 of the confidence band 362 to reach the end-of-life damage threshold 322. The controller 200 can combine, at 384, the number of prediction cycles determined at 382 to produce the remaining useful life distribution 320. The remaining useful life distribution 320 can thus indicate the probability of reaching the end-of-life damage threshold 322 within a given number of cycles. For example, in an embodiment, the remaining useful life distribution 320 can indicate that the probability of reaching the end-of-life threshold 322 within 10,000 cycles is less than 5%. The same remaining useful life distribution 320 can also indicate that there is a 50% probability of reaching the end-of-life threshold 322 within 14,000 to 16,000 component cycles. Similarly, the same remaining useful life distribution 320 can also indicate that the probability of not exceeding the end-of-life threshold 322 within 20,000 component cycles can be less than 5%. It should be understood that the remaining useful life distribution 320 can be concentrated around the best-fit series 358.

[0151] At 318, the remaining useful life distribution 320 can be represented in relation to the predicted number of component cycles. However, in an embodiment, the controller 200 can be configured at 386 to predict the remaining useful life distribution 320 with respect to time. Thus, in an embodiment, the controller 200 can determine an interpolation of component cycles over a specified time interval, such as a ratio, based on historical operating data set for the wind turbine 100. The controller 200 can use this ratio to transform the remaining useful life distribution 320 in order to represent the probability of reaching the end-of-life threshold 322 within the specified time interval. The controller 200 can generate an output table at 388 that indicates the probability of component failure within each of a plurality of time intervals. It should be understood that the time intervals can be weeks, months, years, and / or any other suitable intervals.

[0152] Still referring to Figure 5 , in an embodiment, the controller 200 of the system 300 can be configured to determine at 390 whether the remaining useful life distribution 320 is below a shutdown threshold. In an embodiment, the controller 200 can be configured to, at 324, if the remaining useful life distribution 320 is below the shutdown threshold, shut down or idle the wind turbine 100. The shutdown threshold can be determined by the operator. Being below the shutdown threshold can indicate that the component will likely reach the end-of-life damage threshold 322 faster than otherwise desired. For example, in an embodiment, where the remaining useful life distribution 320 indicates that the probability of the component reaching the end-of-life damage threshold 322 within one month is greater than 50%, the controller 200 can shut down or idle the wind turbine 100.

[0153] Still referring to Figure 5 , in an embodiment, the system 300 can include an inspection feedback loop 392. The inspection feedback loop 392 can include performing an inspection on the component and assigning a damage level corresponding to the observed degree of damage of the component. The damage level can be provided to the controller 200. The controller 200 can determine the difference between the damage level and the predicted damage level based on a risk index. The controller 200 can be configured to improve the model at 310 for predicting the range of potential risk indices 312 over a defined plurality of component cycles. The improvement can be based on the determined difference between the classified observed degree of damage and the predicted damage potential 316 based on the risk index 308. It should be understood that in at least one embodiment, the grading of damage can be implemented by the controller 200 using an artificial intelligence system.

[0154] As Figure 3 , Figure 5 and Figure 6As depicted in, system 300 can be configured such that the field controller 202 groups maintenance activities regarding each of the wind turbines 100 at 394 based on the remaining useful life distribution 320 of each wind turbine 100 within a specified time interval. In such embodiments, the field controller 202 can also be configured to generate a maintenance plan for the wind farm 152 at 395. The maintenance plan can be calculated to maximize the maintenance performed during maintenance operations at the farm level while minimizing premature maintenance operations. For example, the maintenance plan can seek to maximize the amount of maintenance performed when a local crane or maintenance team is on site without prematurely replacing components.

[0155] In additional embodiments, the field controller 202 can reallocate at least a portion of the power generation demand from the wind turbine 100 to at least one other wind turbine of the wind farm 152 at 396. In an embodiment, reallocating at least a portion of the power generation burden can allow for the generation of an idle command for the wind turbine. In an embodiment, the field controller 202 can alternate idle periods with periods of active power generation of the wind turbine 100 at 398 in order to reduce the number of component cycles per unit time. It should be understood that reducing the number of component cycles per unit time can delay the approach to the shutdown threshold. Delaying the approach to the shutdown threshold can in turn facilitate the grouping of repair activities regarding each of the wind turbines.

[0156] Furthermore, those skilled in the art will recognize the interchangeability of various features from different embodiments. Similarly, the various method steps and features described, as well as other known equivalents for each such method and feature, can be mixed and matched by those of ordinary skill in the art to construct additional systems and techniques in accordance with the principles of this disclosure. Of course, it is to be understood that not all of the such objectives or advantages described above need to be achieved in accordance with any particular embodiment. Thus, for example, those skilled in the art will recognize that the systems and techniques described herein can be embodied or performed in a manner that achieves or optimizes one advantage or a group of advantages as taught herein without necessarily achieving other objectives or advantages as may be taught or suggested herein.

[0157] This written description uses examples to disclose the invention (including the best mode), and also enables any person skilled in the art to practice the invention, including making and using any device or system and performing any incorporated method. The patentable scope of the invention is defined by the claims, and may include other examples that occur to those skilled in the art. If such other examples include structural elements that are not different from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal language of the claims, such other examples are intended to be within the scope of the claims.

[0158] Additional aspects of the invention are provided by the subject matter of the following clauses:

[0159] Clause 1. A method for operating a wind turbine, the method comprising: receiving, by a controller, a plurality of data inputs from at least one source, the plurality of data inputs representing a plurality of monitored attributes of components of the wind turbine; using, by the controller, the plurality of monitored attributes to determine a comprehensive risk index for the component, the risk index defining a deviation from a nominal behavior of the component; predicting, by the controller, a range of potential risk indices evolving from the comprehensive risk index over a defined plurality of component cycles, wherein each potential risk index is associated with a damage potential; determining, by the controller, a remaining useful life distribution based on the damage potential and a end-of-life damage threshold; and stopping or idling the wind turbine if the remaining useful life distribution is below a shutdown threshold.

[0160] Clause 2. The method of any of the preceding clauses, wherein receiving the plurality of data inputs further comprises receiving, by the controller, a plurality of time series data inputs from at least one sensor configured to monitor the component during operation.

[0161] Clause 3. The method of any of the preceding clauses, wherein determining the comprehensive risk index for the component further comprises: downselecting, by the controller, the plurality of time series data inputs to establish a plurality of monitored attributes; applying, by the controller, an upper normal limit and a lower normal limit to each of the plurality of monitored attributes, the upper normal limit and the lower normal limit defining a range of data inputs consistent with a baseline of a healthy component, the upper normal limit and the lower normal limit being based on a historical wind turbine dataset; detecting, by the controller, a deviation of at least one of the monitored attributes from the range defined by the normal limits; defining, by the controller, a prediction error for each of the plurality of monitored attributes, wherein the prediction error reflects a deviation amount of the corresponding monitored attribute from the corresponding normal limit; squaring, by the controller, the prediction error for each of the plurality of monitored attributes; and summing, by the controller, the squared prediction errors for each of the plurality of monitored attributes in order to calculate the comprehensive risk index for the component, wherein a non-zero value indicates operation outside the normal limits for the component.

[0162] Clause 4. The method of any of the preceding clauses, further comprising: defining a risk index threshold, wherein defining the risk index threshold balances early fault detection with the likelihood of false alarms; and detecting, prior to predicting the range of potential risk indices, that the risk index threshold is crossed by the comprehensive risk index for the component.

[0163] Clause 5. The method of any of the foregoing clauses, wherein predicting the potential risk index range further includes: selecting a historical risk index data set for each of the plurality of wind turbines from a historical unit turbine data set within a defined plurality of component cycles; using a controller to determine an average unit risk index within the defined plurality of component cycles; using a controller to determine the covariance of the historical risk index data set for each of the plurality of wind turbines relative to the average unit risk index; using a controller to model the potential risk index range for the component; using a controller to determine the best-fit risk index series for the component; and using a controller to determine the confidence band for the potential risk index range.

[0164] Clause 6. The method of any of the foregoing clauses, wherein when the risk index for each of the plurality of wind turbines crosses a predetermined threshold, the defined plurality of component cycles is set to zero.

[0165] Clause 7. The method of any of the foregoing clauses, further including: using a controller to convert the potential risk index range to a damage potential via the correlation between damage and the risk index, wherein the correlation between damage and the risk index is determined by performing a regression on a historical unit turbine failure data set, wherein the historical unit turbine failure data set depicts the damage values given by inspections and the recorded risk index for the component at the time of inspection.

[0166] Clause 8. The method of any of the foregoing clauses, wherein determining the remaining useful life distribution further includes: establishing a life-ending damage threshold based on a damage level at which the probability of failure of the component or secondary damage to the wind turbine exceeds an acceptable limit; using a controller to determine the number of prediction cycles required to reach the life-ending damage threshold for each sample of the sampling; and using a controller to combine the determined number of prediction cycles for each sample so as to generate a remaining useful life distribution between an upper confidence interval and a lower confidence interval.

[0167] Clause 9. The method of any of the foregoing clauses, further including: using a controller to determine the interpolation of the component cycles to a specified time interval based on a historical operation data set of the wind turbine; and using a controller to predict the distribution of the remaining useful life over time via a ratio.

[0168] Clause 10. The method of Clause 9, further including using a controller to generate an output table indicating the probability of component failure within each of a plurality of time intervals.

[0169] Clause 11. The method of any of the foregoing clauses further includes performing an inspection of a component; assigning a damage level corresponding to the observed damage level of the component; providing the damage level to a controller; using the controller to determine a difference between the damage level and a predicted damage level based on a risk index; and using the controller to improve a model for predicting a range of potential risk indices relative to a defined plurality of component cycles, wherein the improvement is based on a determined difference between a graded observed damage level and a predicted damage potential based on a risk index.

[0170] Clause 12. The method of any of the foregoing clauses, wherein the controller is a field controller and the wind turbine is one of a plurality of wind turbines in a wind farm.

[0171] Clause 13. The method of any of the foregoing clauses further includes using the field controller to group maintenance activities for each of the wind turbines based on a remaining useful life distribution for each of the plurality of wind turbines over a specified time interval; and using the field controller to generate a maintenance plan for the wind farm.

[0172] Clause 14. The method of any of the foregoing clauses further includes using the field controller to reallocate at least a portion of a power generation demand from a wind turbine to at least one other wind turbine in the wind farm; and alternating an idle period of the wind turbine with an active power generation period to reduce the number of component cycles per unit time, wherein reducing the number of component cycles per unit time delays an approach to a shutdown threshold to facilitate grouping repair activities for each of the wind turbines.

[0173] Clause 15. A system for operating a wind turbine, comprising: at least one sensor operably coupled to a component of the wind turbine to detect an attribute of the component; and a controller communicatively coupled to the at least one sensor, the controller including at least one processor configured to perform a plurality of operations, the plurality of operations including: receiving a plurality of data inputs from the at least one sensor, the plurality of data inputs representing a plurality of monitored attributes of the component; determining, via a risk index module, a comprehensive risk index for the component using the plurality of monitored attributes, the risk index defining a deviation from a nominal behavior of the component; predicting, via a prediction module, a range of potential risk indices evolving from the comprehensive risk index over a defined plurality of component cycles, wherein each potential risk index is associated with a damage potential; determining a remaining useful life distribution based on the damage potential and a end-of-life damage threshold; and stopping or idling the wind turbine if the remaining useful life distribution is below a shutdown threshold.

[0174] Clause 16. The system of any of the foregoing clauses, wherein determining a comprehensive risk index for the component via the risk index module further comprises: selecting a plurality of time series data inputs downward via the filtering module to establish a plurality of monitoring attributes; applying an upper normal limit and a lower normal limit to each of the plurality of monitoring attributes via the filtering module, the upper normal limit and the lower normal limit defining a range of data inputs consistent with the baseline of a healthy component, the upper normal limit and the lower normal limit being based on a historical turbine dataset; detecting, via the risk index module, a deviation of at least one of the monitored attributes from the range defined by the normal limits; defining, via the risk index module, a prediction error for each of the plurality of monitoring attributes, wherein the prediction error reflects the amount of deviation of the corresponding monitoring attribute from the corresponding normal limit; squaring, via the risk index module, the prediction error for each of the plurality of monitoring attributes; and summing, via the risk index module, the squared prediction errors for each of the plurality of monitoring attributes in order to calculate a comprehensive risk index for the component, wherein a non-zero value indicates operation outside the normal limits for the component.

[0175] Clause 17. The system of any of the foregoing clauses, wherein predicting a potential risk index range via the prediction module further comprises: selecting a historical risk index dataset for each of the plurality of wind turbines from the historical turbine dataset within a defined plurality of component cycles, wherein when the risk index for each of the plurality of wind turbines crosses a predetermined threshold, the defined plurality of component cycles is set to zero; determining an average fleet risk index within the defined plurality of component cycles; determining the covariance of the historical risk index dataset for each of the plurality of wind turbines with respect to the average fleet risk index; modeling a potential risk index range for the component; determining an optimal fit risk index series for the component; and determining a confidence band for the potential risk index range.

[0176] Clause 18. The system of any of the foregoing clauses, wherein the plurality of operations further comprises converting the potential risk index range to a damage potential via a correlation between damage and the risk index, wherein the correlation between damage and the risk index is determined by performing a regression on a historical turbine failure dataset, wherein the historical turbine failure dataset depicts damage values given by inspections and the recorded risk index for the component at the time of inspection.

[0177] Clause 19. The system of any of the foregoing clauses, wherein determining a remaining useful life distribution further comprises: establishing a life-ending damage threshold based on a damage level at which the probability of failure of the component or secondary damage to the wind turbine exceeds an acceptable limit; determining the number of prediction cycles required to reach the life-ending damage threshold for each sample of the sampling; and combining, using a controller, the determined number of prediction cycles for each sample in order to generate a remaining useful life distribution between an upper confidence interval and a lower confidence interval.

[0178] Clause 20. A system of any of the foregoing clauses, wherein the plurality of operations further comprises: determining an interpolation of the component cycles over a specified time interval based on a historical operation data set regarding a wind turbine; predicting a distribution of the remaining useful life over time via a ratio; and generating an output table that indicates a probability of component failure within each of a plurality of time intervals.

Claims

1. A method for operating a wind turbine, the method comprising: The controller receives a plurality of data inputs from a plurality of sensors, the plurality of data inputs representing a plurality of different monitored attributes of a plurality of components of a gearbox of the wind turbine; The controller combines the plurality of different monitored attributes of a particular component among the plurality of components of the gearbox to calculate, for each of the plurality of components of the gearbox, a single comprehensive risk index for the particular component, each of the single comprehensive risk indices defining a deviation from a nominal behavior of the particular component of the plurality of components of the gearbox; The controller predicts, for each of the plurality of components of the gearbox, a range of potential risk indices evolving from each of the single comprehensive risk indices over a defined number of component cycles, wherein each potential risk index is associated with a damage potential; The controller determines a remaining useful life distribution based on the damage potential and a end-of-life damage threshold for each of the plurality of components of the gearbox; And If the remaining useful life distribution for at least one of the plurality of components of the gearbox is below a shutdown threshold, the wind turbine is shut down or idled.

2. The method according to claim 1, characterized in that, Receiving the plurality of data inputs further includes: The controller receives a plurality of time series data inputs from the plurality of sensors, the plurality of sensors configured to monitor the plurality of components of the gearbox during operation.

3. The method according to claim 2, characterized in that, The method further includes: The controller down-selects the plurality of time series data inputs to establish the plurality of different monitored attributes; The controller applies an upper normal limit and a lower normal limit to each of the plurality of different monitored attributes, the upper normal limit and the lower normal limit defining a range of data inputs consistent with a baseline of a healthy component, the upper normal limit and the lower normal limit being based on a historical fleet turbine data set; The controller detects a deviation of at least one of the different monitored attributes from the range defined by the upper normal limit and the lower normal limit; The controller defines a prediction error for each of the plurality of different monitored attributes, wherein the prediction error reflects the amount of deviation of the corresponding monitored attribute from the corresponding normal limit; The controller squares the prediction error for each of the plurality of different monitored attributes; and The controller sums the squared prediction errors for each of the plurality of different monitored attributes in order to calculate a single comprehensive risk index for the plurality of components of the gearbox, wherein a non-zero value indicates operation outside of the upper normal limit and the lower normal limit for one or more of the plurality of components.

4. The method according to claim 3, characterized in that, The method further includes: Defining a risk index threshold, wherein defining the risk index threshold balances early fault detection with the likelihood of false alarms; and Before predicting the range of potential risk indices, detecting that the risk index threshold is crossed by a single comprehensive risk index for each of the plurality of components of the gearbox.

5. The method according to claim 1, characterized in that, Predicting the range of potential risk indices further includes: Select a historical risk index data set for each of a plurality of wind turbines from a historical fleet turbine data set over a defined plurality of component cycles; Determine an average fleet risk index over the defined plurality of component cycles using the controller; Determine a covariance of the historical risk index data set for each of the plurality of wind turbines relative to the average fleet risk index using the controller; Model a potential risk index range for each of the plurality of components of the gearbox using the controller; Determine an optimal fit risk index series for each of the plurality of components of the gearbox using the controller; and Determine a confidence band for the potential risk index range using the controller.

6. The method according to claim 5, characterized in that, Set the defined plurality of component cycles to zero when the risk index for each of the plurality of wind turbines crosses a predetermined threshold.

7. The method according to claim 1, characterized in that, The method further includes: Convert the potential risk index range to a damage potential using the controller via a correlation of damage to the risk index, wherein the correlation of damage to the risk index is determined by performing a regression on a historical fleet turbine failure data set that depicts a given damage value from an inspection and the recorded risk index for the component at the time of the inspection.

8. The method according to claim 1, characterized in that, Determining the remaining useful life distribution further includes: Establishing a life - ending damage threshold based on a damage level at which the probability of failure or secondary damage to the wind turbine of each of the plurality of components of the gearbox exceeds an acceptable limit; Determining, using the controller, the number of prediction cycles required to reach the life - ending damage threshold for each sample taken; and Combining, using the controller, the determined number of prediction cycles for each sample to produce a remaining useful life distribution between an upper confidence interval and a lower confidence interval.

9. The method according to claim 1, wherein The method further includes: Determining, using the controller, an interpolation of the component cycles to a specified time interval based on a historical operation data set for the wind turbine; and Predicting, using the controller, the distribution of the remaining useful life relative to time via a ratio.

10. The method according to claim 9, wherein The method further includes: Generating, using the controller, an output table that indicates the probability of component failure within each of a plurality of time intervals.

11. The method according to claim 1, wherein The method further includes: Performing an inspection of each of the plurality of components of the gearbox; Assigning a damage grade corresponding to the observed degree of damage for each of the plurality of components of the gearbox; Providing the damage grade to the controller; Determining, using the controller, a difference between the damage grade and a predicted damage level based on the risk index; and Improving, using the controller, a model for predicting the potential risk index range relative to the defined plurality of component cycles, wherein the improvement is based on the determined difference between the graded observed degree of damage and the predicted damage potential based on the risk index.

12. The method according to claim 1, wherein The controller is a field controller, and the wind turbine is one of a plurality of wind turbines in a wind farm.

13. The method according to claim 12, wherein The method further includes: Using the field controller to group maintenance activities for each of the plurality of wind turbines based on the remaining useful life distribution for each of the plurality of wind turbines over a specified time interval; and Using the field controller to generate a maintenance plan for the wind farm.

14. The method according to claim 13, wherein The method further includes: Using the field controller to reallocate at least a portion of the power generation demand from the wind turbine to at least one other wind turbine of the wind farm; and Alternating an idle period of the wind turbine with an active power generation period to reduce the number of component cycles per unit time, wherein reducing the number of component cycles per unit time delays approaching the shutdown threshold to facilitate grouping repair activities for each of the wind turbines.

15. A system for operating a wind turbine, comprising: A plurality of sensors operatively coupled to a plurality of components of a gearbox of the wind turbine to detect attributes of the plurality of components; And A controller communicatively coupled to the plurality of sensors, the controller including at least one processor configured to perform a plurality of operations, the plurality of operations including: Receiving a plurality of data inputs from the plurality of sensors, the plurality of data inputs representing a plurality of different monitored attributes of each of the plurality of components of the gearbox; Combining, via a risk index module, each of the plurality of different monitored attributes of a particular component of the plurality of components of the gearbox to calculate a single comprehensive risk index for the particular component for each of the plurality of components of the gearbox, each of the single comprehensive risk indices defining a deviation from a nominal behavior of the particular component of the plurality of components of the gearbox; Predicting, via a prediction module, a range of potential risk indices evolving from each of the single comprehensive risk indices over a defined number of component cycles for each of the plurality of components of the gearbox, wherein each potential risk index is associated with a damage potential; Determining a remaining useful life distribution based on the damage potential and a life - ending damage threshold for each of the plurality of components of the gearbox, and If the remaining useful life distribution for at least one of the plurality of components of the gearbox is below a shutdown threshold, shutting down or idling the wind turbine.

16. The system according to claim 15, wherein The plurality of operations further includes: Receiving a plurality of time - series data inputs from the plurality of sensors configured to monitor the plurality of components of the gearbox during operation; Down - selecting, via a filtering module, the plurality of time - series data inputs to establish the plurality of different monitored attributes; Applying, via the filtering module, an upper normal limit and a lower normal limit to each of the plurality of different monitored attributes, the upper normal limit and the lower normal limit defining a range of data inputs consistent with a baseline of healthy components, the upper normal limit and the lower normal limit being based on a historical fleet turbine dataset; Detecting, via the risk index module, a deviation of at least one of the plurality of different monitored attributes from the range defined by the upper normal limit and the lower normal limit; Define a prediction error for each of the plurality of different monitoring attributes via the risk index module, wherein the prediction error reflects the amount of deviation of the corresponding monitoring attribute from the corresponding normal limit; Square the prediction error for each of the plurality of different monitoring attributes via the risk index module; and Sum the squared prediction errors for each of the plurality of different monitoring attributes to calculate a comprehensive risk index for each of the plurality of components of the gearbox, wherein a non-zero value indicates operation outside the normal limit for each of the plurality of components of the gearbox.

17. The system according to claim 15, wherein, Predicting the potential risk index range via the prediction module further includes: Selecting a historical risk index data set from a historical unit turbine data set for each of the plurality of wind turbines within a defined plurality of component cycles, wherein when the risk index for each of the plurality of wind turbines crosses a predetermined threshold, the defined plurality of component cycles is set to zero; Determining an average unit risk index within the defined plurality of component cycles; Determining the covariance of the historical risk index data set for each of the plurality of wind turbines relative to the average unit risk index; Modeling the potential risk index range for each of the plurality of components of the gearbox; Determining the best fit risk index series for each of the plurality of components of the gearbox; and Determining the confidence band for the potential risk index range.

18. The system according to claim 15, wherein, The plurality of operations further includes: Converting the potential risk index range to the damage potential via the correlation between damage and the risk index, wherein the correlation between damage and the risk index is determined by performing a regression on a historical unit turbine failure data set, wherein the historical unit turbine failure data set depicts the damage value given by an inspection and the recorded risk index for the component at the time of the inspection.

19. The system according to claim 15, wherein, Determining the remaining useful life distribution further includes: Establishing a life termination damage threshold based on a damage level at which the probability of failure or secondary damage to the wind turbine of each of the plurality of components of the gearbox exceeds an acceptable limit; Determining the number of prediction cycles required to reach the life termination damage threshold for each sample of the sampling; and Combining the determined number of prediction cycles for each sample using the controller to generate a remaining useful life distribution between an upper confidence interval and a lower confidence interval.

20. The system according to claim 19, wherein, The plurality of operations further includes: Determining an interpolation of the component cycle to a specified time interval based on a historical operation data set of the wind turbine; Predicting the distribution of the remaining useful life relative to time via a ratio; and Generating an output table that indicates the probability of component failure within each of a plurality of time intervals.

Images