System and method for detecting anomalies during asset operation

Abstract

A method for detecting anomalies during operation of an asset. The method includes collecting data associated with operation of the asset. The data includes operating parameters of the asset and environmental parameters around the asset. The method also includes selecting a monitoring parameter and one or more classification parameters. The method also includes selecting an anomaly function for the monitoring parameters given one or more classification parameters. Given one or more classification parameters, an anomaly function is determined based on a respective conditional probability distribution of the monitored parameters. The method further includes determining an anomaly threshold for the monitoring parameter based on the anomaly function and the one or more classification parameters. The method also includes generating an alarm event when the monitored parameter exceeds the anomaly threshold. The method further includes actuating a human-machine interface to output an alarm based on comparing the score derived from the alarm event to an alarm threshold.

Description

Technical Field

[0001] This disclosure generally relates to wind farms, and more specifically, to systems and methods for detecting anomalies during the operation of one or more wind farm assets. Background Technology

[0002] Wind power is considered one of the cleanest and most environmentally friendly energy sources available today, and wind turbines have gained increasing attention in this area. Modern wind turbines typically consist of a tower, generator, gearbox, nacelle, and one or more rotor blades. The rotor blades capture the kinetic energy of the wind using known airfoil principles. For example, rotor blades typically have an airfoil cross-sectional profile, causing air to flow over the blades during operation, creating a pressure difference between the two sides. Therefore, lift acts on the blades from the pressure side towards the suction side. This lift generates torque on the main rotor shaft, which meshes with the generator used to produce electricity.

[0003] Multiple wind turbines are often used in combination to generate electricity and are commonly referred to as a "wind farm". During operation, it is advantageous to utilize various analyses to evaluate the performance of the wind turbines and / or wind farm to ensure that (one or more) the wind turbines and / or wind farm are operating correctly. Many analyses are trained on multi-parameter time-series data of the asset or group of assets and then applied to the asset. Such analyses can include, for example, anomaly detection analyses that utilize various machine learning methods to identify anomalous operation of (one or more) wind turbines in the wind farm.

[0004] However, existing anomaly detection analytics have certain drawbacks. For example, machine learning methods are data-driven and may not consider the physical characteristics of the operation of one or more assets (i.e., one or more wind turbines and / or their various components). Therefore, machine learning methods may not provide accurate anomaly detection analytics for one or more assets for which their machine learning models have not been trained. Consequently, to apply these anomaly detection analytics, the machine learning methods must be trained using data from each asset, which requires significant time and training data. Furthermore, training machine learning methods on data from each asset may require customization for each anomaly detection analytics.

[0005] In view of the foregoing, this disclosure relates to systems and methods for detecting anomalies during asset operation by utilizing conditional probability distributions conditioned on operational and / or environmental parameters, enabling the detection of anomalous asset behavior from historical data of the asset. Summary of the Invention

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

[0007] In this aspect, the present disclosure relates to a method for detecting anomalies during the operation of an asset. The method includes collecting data associated with the operation of the asset via a controller. The data includes operational parameters of the asset and environmental parameters surrounding the asset. The method also includes selecting monitoring parameters and one or more classification parameters. The monitoring parameter is one of the operational parameters, and the one or more classification parameters are at least one of one or more other operational parameters or one or more environmental parameters. The method further includes selecting an anomaly function for the monitoring parameter via the controller, given one or more classification parameters. Given one or more classification parameters, the anomaly function is determined based on a corresponding conditional probability distribution of the monitoring parameter. The method also includes determining an anomaly threshold for the monitoring parameter via the controller based on the anomaly function and the one or more classification parameters. The method further includes generating an alarm event via the controller when the monitoring parameter exceeds the anomaly threshold. The method also includes actuating a human-machine interface via the controller to output an alarm based on a score derived from the alarm event compared to the alarm threshold.

[0008] In another aspect, this disclosure relates to a system for detecting anomalies during the operation of an asset. The system includes a controller communicatively coupled to the asset, the controller being configured to perform a plurality of operations, including but not limited to selecting monitoring parameters and one or more classification parameters, the monitoring parameters being one of the operation parameters, and the one or more classification parameters being at least one of one or more other operation parameters or one or more environmental parameters; selecting an anomaly function for the monitoring parameters via the controller, given the one or more classification parameters; determining the anomaly function based on the corresponding conditional probability distribution of the monitoring parameters, given the one or more classification parameters; determining an anomaly threshold for the monitoring parameters via the controller based on the anomaly function and the one or more classification parameters; generating an alarm event via the controller when the monitoring parameter exceeds the anomaly threshold; and actuating a human-machine interface via the controller to output an alarm based on a score derived from the alarm event compared to the alarm threshold.

[0009] These and other features, aspects, and advantages of the invention will become more readily understood with reference to the following description and the appended claims. Embodiments of the invention are illustrated in conjunction with the accompanying drawings, which are incorporated in and form part of this specification, and together with the description serve to explain the principles of the invention. Attached Figure Description

[0010] The invention (including its preferred mode) is fully disclosed and can be practiced by one of ordinary skill in the art in the description with reference to the accompanying drawings, in which: Figure 1The illustration is a perspective view of an embodiment of a wind farm according to the present disclosure; Figure 2 The illustration is a perspective view of an embodiment of a wind turbine according to the present disclosure; Figure 3 The figure shows a block diagram of an embodiment of a controller for a wind turbine and / or wind farm according to the present disclosure; Figure 4 The illustration is a flowchart of an embodiment of a method for detecting anomalies during asset operation according to the present disclosure; Figure 5 The figure is a schematic diagram of an embodiment of a system for detecting anomalies during the operation of an asset according to the present disclosure; Figure 6 The illustration shows multiple exemplary conditional probability distributions of exemplary monitoring parameters given exemplary classification parameters, according to this disclosure; Figure 7 The figure depicts a graph representing an exemplary anomaly function of a corresponding percentile threshold according to an exemplary conditional probability distribution of this disclosure; and Figure 8 The figure depicts a graph comparing exemplary monitoring parameters and corresponding anomaly thresholds over time according to this disclosure. Detailed Implementation

[0011] Reference will now be made in detail to embodiments of the invention, one or more examples of which are illustrated in the accompanying drawings. Each example is provided by way of explanation rather than limitation of the invention. Indeed, it will be apparent to those skilled in the art that various modifications and variations may be made to the invention without departing from the scope or spirit of the invention. For example, features illustrated or described as part of an embodiment may be used with another embodiment to produce yet another embodiment. Therefore, it is intended that the invention cover such modifications and variations as those falling within the scope of the appended claims and their equivalents.

[0012] Generally, this disclosure relates to systems and methods for detecting anomalies during the operation of one or more wind farm assets using conditional probability distributions, such that an alarm is output in response to the detection of anomalous asset behavior. Utilizing conditional probability distributions improves system performance during application by allowing dynamic selection of anomaly thresholds based on classification of asset operation data, eliminating the need for time-consuming manual anomaly threshold selection. Additionally, in embodiments, historical data from one or more assets is used to generate the conditional probability distribution. The conditional probability distribution is then applied to new data about the assets to detect non-specific mechanical, operational, or performance anomalies related to the assets.

[0013] Now refer to the attached diagram, Figure 1The illustration shows an embodiment of a wind farm 100 comprising a plurality of wind turbines 102 according to aspects of this disclosure. The wind turbines 102 can be arranged in any suitable manner. As an example, the wind turbines 102 can be arranged in an array of rows and columns, a single row, or a random arrangement. Furthermore, Figure 1 The illustration shows an example layout of an embodiment of wind farm 100. Typically, the arrangement of wind turbines in a wind farm is determined based on numerous optimization algorithms to maximize annual power generation (AEP) for the corresponding site's wind climate. It should be understood that any wind turbine arrangement can be achieved, such as on uneven land, without departing from the scope of this disclosure. Furthermore, while there are benefits to applying this method to turbines from a single farm, it can also be applied to a group of turbines from several farms.

[0014] Furthermore, it should be understood that the wind turbine 102 of the wind farm 100 can have any suitable configuration, such as, for example, Figure 2 As shown, the wind turbine 102 includes a tower 114 extending from a supporting surface, a nacelle 116 mounted on the tower 114, and a rotor 118 coupled to the nacelle 116. The rotor 118 includes a rotatable hub 120 having a plurality of rotor blades 112 mounted thereon, which is in turn connected to a main rotor shaft, which is coupled to a generator (not shown) housed within the nacelle 116. Thus, the generator produces electrical power from the rotational energy generated by the rotor 118. It should be understood that... Figure 2 The wind turbine 102 is provided for illustrative purposes only. Therefore, those skilled in the art will understand that the invention is not limited to any particular type of wind turbine configuration.

[0015] As in Figure 1-3 As generally shown, each wind turbine 102 of the wind farm 100 may further include a turbine controller 104, which is communicatively coupled to the wind farm controller 108. Furthermore, in embodiments, as... Figure 1 As shown, the wind farm controller 108 can be coupled to the turbine controller 104 via network 110 to facilitate communication between various wind farm components. The wind turbine 102 may also include one or more sensors 105, 106, 107 (…). Figure 1 and Figure 3 It is configured to monitor various operating conditions, wind and / or load conditions of the wind turbine 102.

[0016] For example, one or more sensors 105, 106, 107 may include: a blade sensor for monitoring rotor blades 112; a generator sensor 105 for monitoring generator load, torque, speed, acceleration, and / or generator power output; a wind sensor 106 for monitoring one or more wind conditions; and / or a shaft sensor for measuring rotor shaft load and / or rotor shaft rotational speed. Additionally, the wind turbine 102 may include one or more tower sensors for measuring load transmitted through tower 114 and / or tower 114 acceleration. In various embodiments, one or more sensors 105, 106, 107 may be any one or a combination of the following: temperature sensor, accelerometer, pressure sensor, angle of attack sensor, vibration sensor, miniature inertial measurement unit (MIMU), camera system, fiber optic system, anemometer, wind vane, sonic detection and ranging (SODAR) sensor, infrared laser, optical detection and ranging (LIDAR) sensor, radiometer, pitot tube, radiosonde anemometer, other optical sensor, virtual sensor, estimates derived from multiple sensors, and / or any other suitable sensor.

[0017] Now for reference Figure 3 The illustration shows a block diagram of an embodiment that may include suitable components within a field controller 108, one or more turbine controllers 104, and / or other suitable controllers according to the present disclosure. As shown, the controllers 104, 108 may include one or more processors 150 (or servers) and associated memory devices 152 configured to perform various computer-implemented functions (e.g., performing methods, steps, calculations, etc., and storing related data as disclosed herein). Additionally, the controllers 104, 108 may also include a communication module 154 for facilitating communication between the controllers 104, 108 and various components of the wind turbine 102. Furthermore, the communication module 154 may include a sensor interface 156 (e.g., one or more analog-to-digital converters) for allowing signals transmitted from one or more sensors 105, 106, 107 to be converted into signals that can be understood and processed by the processors 150. It should be understood that the sensors 105, 106, 107 may be communicatively coupled to the communication module 154 using any suitable means. For example, as shown, one or more sensors 105, 106, 107 are coupled to sensor interface 156 via a wired connection. However, in other embodiments, one or more sensors 105, 106, 107 may be coupled to sensor interface 156 via a wireless connection, such as by using any suitable wireless communication protocol known in the art.

[0018] As used herein, the term "processor" refers not only to integrated circuits known in the art as included in a computer, but also to controllers, microcontrollers, microcomputers, programmable logic controllers (PLCs), servers, application-specific integrated circuits (ASICs), and other programmable circuits. Additionally, memory device(s) 152 may typically include memory elements(s), 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 disc read-only memory (CD-ROM), magneto-optical disks (MOD), digital versatile discs (DVDs), and / or other suitable memory elements. Such memory devices(s) 152 may typically be configured to store suitable computer-readable instructions that, when implemented by processor(s) 150, configure controller(s) 104, 108 to perform the various functions described herein.

[0019] Furthermore, the network 110 coupling the wind farm controller 108, turbine controller 104, and / or one or more sensors 105, 106, 107 in the wind farm 100 can include any known communication network, such as wired or wireless networks, optical networks, etc. Moreover, the network 110 can be connected in any known topology (such as ring, bus, or hub) and can have any known contention resolution protocol without departing from the art. Therefore, the network 110 is configured to provide near real-time data communication between the turbine controller (one or more) and the wind farm controller 108.

[0020] As commonly understood, wind turbines typically include multiple operational analyses, which generally refer to data modules collected and analyzed in association with the operation of a wind turbine, which are categorized, stored, and / or analyzed, or can be categorized, stored, and / or analyzed to study various trends or patterns in the data. Therefore, in embodiments, as an example, the analyses(s) described herein may include anomaly detection analyses, which can be used to identify anomalies within the operational data of a wind turbine or a set of wind turbines. Thus, as Figure 4 and Figure 5 As shown, this disclosure relates to a method 200 and a system 300 for detecting anomalies during the operation of an asset.

[0021] More specifically, Figure 4 The diagram illustrates a flowchart of a method 200 for detecting anomalies during asset operations according to this disclosure. Figure 5 The diagram illustrates a system 300 for detecting anomalies during asset operation, according to this disclosure. Typically, such as... Figure 4As shown, the method 200 described herein can be implemented using the aforementioned wind turbine 102 and / or wind farm 100. However, it should be understood that the disclosed method 200 can be used with any other suitable asset having any suitable configuration. Additionally, although... Figure 4 The steps are depicted in a particular order for illustrative and descriptive purposes, but the methods discussed herein are not limited to any particular order or arrangement. Using the disclosure provided herein, those skilled in the art will appreciate that the various steps of the methods can be omitted, rearranged, combined, and / or modified in various ways.

[0022] As shown at (202), method 200 includes collecting data associated with the operation of an asset (e.g., wind turbine 102) via a controller. For example, as Figure 5 As shown, system 300 may include controller 302 (such as field controller 108). Controller 302 may collect data associated with one or more assets 304. In one embodiment, as shown, controller 302 collects data from one asset 304 (e.g., wind turbine 102). In another embodiment, controller 302 collects data from a group of assets (e.g., multiple wind turbines 102 in wind farm 100). In such an embodiment, when collecting data from a group of assets, each asset in the group may be similar because each asset is expected to behave or perform in substantially the same way (e.g., have similar data variation patterns).

[0023] While method 200 is applied to wind turbine 102 in this disclosure, it should be understood that method 200 can be applied to other types of assets, components, or devices to be monitored and expected to perform or function in substantially the same manner (i.e., where multiple instances exist). Therefore, method 200 can be applied to solar panels, energy storage devices or systems, engines, vehicles, trucks, and / or aircraft. Furthermore, method 200 can be applied to sub-components of larger systems, such as valves, gearboxes, circuits, power converters, bearings, or any other system components.

[0024] Data associated with the operation of asset 304 can be collected by one or more of sensors 105, 106, and 107. The collected data can then be organized (e.g., based on asset identifiers such as serial numbers) and stored (e.g., in memory device 152 of controller 302). In embodiments, the collected data includes operating parameters of the asset and environmental parameters surrounding the asset. Operating parameters include values ​​of various parameters that define the operating state of the asset. As an example, the collected data may include sensed values ​​of various operating parameters, such as rotor speed, rotor pitch, nacelle yaw, actual power output, generator speed, etc. These sensed values ​​of operating parameters can be used to calculate or determine (e.g., via model-based estimation) values ​​of other operating parameters associated with the asset (e.g., mechanical load, component stress and strain, expected power output, etc.). The calculated values ​​can be included in the data associated with the operation of asset 304.

[0025] Environmental parameters include values ​​for various parameters that define the environmental conditions surrounding the asset. As an example, the collected data may include sensed values ​​of various environmental parameters such as wind speed, wind direction, ambient temperature, etc. These sensed values ​​can be used to calculate or determine values ​​for other environmental parameters associated with the asset (e.g., wind turbulence, wind effects, etc.). The calculated values ​​can be included in the data associated with the operation of asset 304.

[0026] Return to reference Figure 4 As shown at (204), method 200 includes selecting a monitoring parameter and one or more classification parameters via controller 302. The monitoring parameter is one of the operating parameters. The classification parameter (one or more) is at least one of one or more other operating parameters or one or more environmental parameters. As an example, such as... Figure 5 As shown, controller 302 can select monitoring parameters and (one or more) classification parameters based on user input 308. In such an example, as described below, the user can provide input to the human-machine interface (HMI) 316 specifying parameters for selection. Therefore, controller 302 can use user input 308 to select 306 monitoring parameters and (one or more) classification parameters. That is, controller 302 can select the operating parameters specified by user input 308 as monitoring parameters, and select at least one of one or more other operating parameters or environmental parameters specified by user input 308 as classification parameters. As another example, controller 302 can access lookup tables, etc., that associate various monitoring parameters with various classification parameters (e.g., stored in memory device 152 of controller 302). In such an example, controller 302 can iteratively select various monitoring parameters and corresponding classification parameters.

[0027] Return to reference Figure 4 As shown at (206), method 200 includes selecting anomaly function 314 for the monitoring parameter via controller 302, given one or more classification parameters. Anomaly function 314 represents an anomaly threshold for the monitoring parameter as a function of one or more classification parameters. As described above and as... Figure 5 As illustrated, controller 302 can receive user input 308 specifying monitoring parameters and one or more classification parameters. Controller 302 can select anomaly functions 314 associated with the specified monitoring parameters and the specified classification(s) parameters(s). For example, controller 302 can access a lookup table (e.g., stored in memory device 152 of controller 302) that associates various anomaly functions 314 with various monitoring parameters given corresponding classification parameters. The anomaly functions 314 for various monitoring parameters given various classification parameters can be stored by controller 302 (e.g., in its memory device 152).

[0028] Furthermore, given corresponding values ​​of one or more classification parameters, an anomaly function 314 is determined based on the corresponding conditional probability distribution 310 indicating the probability of the values ​​of the monitoring parameters. As an example, Figure 6 Provides a classification parameter P c Monitoring parameter P under the example values ​​x1, x2, x3, and x4 m Multiple conditional probability distributions 310a, 310b, 310c, and 310d. As depicted, each conditional probability distribution 310a, 310b, 310c, and 310d indicates the condition given a classification parameter P. c The monitoring parameter P under the corresponding values ​​x1, x2, x3, and x4. m The probability P of the value can be calculated using known data preprocessing techniques. m The values ​​are binned (i.e., categorized based on the range of values ​​and represented by values ​​that represent that range).

[0029] Furthermore, percentile thresholds 400 can be specified for multiple conditional probability distributions 310a, 310b, 310c, and 310d. The percentile thresholds 400 can be specified via user input. For example, the HMI 316 can receive user input specifying the percentile thresholds 400. The controller 302 can then store the percentile thresholds 400 (e.g., in its memory device 152). The percentile thresholds 400 define the monitoring parameter P. m The value is greater than the monitoring parameter P of the corresponding conditional probability distributions 310a, 310b, 310c, and 310d. mThe percentile threshold 400 is a given percentage of the value. The percentile threshold 400 can be any suitable percentile (e.g., the 50th percentile, 75th percentile, 97th percentile, 99.7th percentile, etc.). For example, the percentile threshold 400 can be specified based on the sampling rate of data associated with the asset's operations, the number of expected alert events within a time period, and / or the sample size of historical data used to generate conditional probability distributions 310a, 310b, 310c, 310d. As a non-limiting example, a user could specify the percentile threshold 400 as the 99.3rd percentile, which corresponds to one expected alert event per day at a data sampling rate of once every ten minutes.

[0030] When determining the percentile threshold 400, the controller 302 can, for example, plot the monitoring parameter P corresponding to the percentile threshold 400 for the corresponding conditional probability distributions 310a, 310b, 310c, 310d. m The corresponding values ​​of the points are defined. Then, by applying one or more regression techniques to the points, points can be generated via controller 302 based on the given classification parameters P. c Monitoring parameter P under favorable conditions m The exception function is 314. As an example... Figure 7 Provides a description as a classification parameter P c The graph 404 shows an exemplary anomaly function 314 of the exemplary values ​​x1, x2, x3, and x4, which represents the monitoring parameter P corresponding to the percentile threshold 400 of the corresponding conditional probability distributions 310a, 310b, 310c, and 310d. m The corresponding values. Then any one or more suitable regression techniques can be used to generate an outlier function relative to the points 314. One or more linear regression and / or nonlinear regression techniques can be used to generate an initial regression line relative to the data points.

[0031] In an embodiment, method 200 may include generating corresponding conditional probability distributions 310a, 310b, 310c, and 310d via controller 302 based on historical data 312 associated with asset operations. For example, historical data 312 may be aggregated for the asset over a period of time. A classification parameter P may be obtained based on the corresponding conditional probability distributions 310a, 310b, 310c, and 310d. cThe statistically significant number of data points for the exemplary values ​​x1, x2, x3, and x4 is used to determine the time period. The statistically significant number of data points can be determined based on a percentile threshold of 400. Historical data 312 can be associated with operations of an asset group that includes assets. Collecting data from an asset group that includes assets can reduce the amount of time required to train the conditional probability distribution by aggregating (e.g., simultaneously) data collected from assets that are expected to perform or be performed in substantially the same way.

[0032] Historical data 312 can be, for example, simulated data. In such an example, conditional probability distributions 310a, 310b, 310c, and 310d can be generated based on data obtained through computer simulations such as Monte Carlo simulations. As another example, historical data 312 can be measurement data. In such an example, conditional probability distributions 310a, 310b, 310c, and 310d can be generated based on data sensed / computed via one or more sensors.

[0033] Return to reference Figure 4 As shown at (208), method 200 includes, via controller 302, based on an anomaly function 314 and one or more classification parameters P c The corresponding value is used to determine the monitoring parameter P. m The abnormal threshold is 402. For example... Figure 7 As shown, the anomaly threshold 402 is determined by the anomaly function 314 for a given classification parameter P. c Output monitoring parameter P m The value of the classification parameter P. For example, controller 302 can assign one or more classification parameters P. c The sensed / calculated value is input into the selected anomaly function 314, and the selected anomaly function 314 outputs the monitoring parameter P. m The anomaly threshold 402. As another example, the controller 302 can access various anomaly thresholds 402 and their corresponding classification parameters P. c A lookup table, etc., associated with various values ​​of a given exception function 314 (e.g., stored in memory device 152). In such an example, controller 302 can access the lookup table for the selected exception function 314, and can then select one or more classification parameters P from the lookup table. c The abnormal threshold 402 is associated with the sensed / calculated value.

[0034] Return to reference Figure 4 As shown at (210), method 200 includes monitoring parameter P m When the value exceeds the abnormal threshold 402, an alarm event is determined via controller 302. For example, controller 302 can monitor parameter P. mThe sensed / calculated value is compared with the anomaly threshold 402 determined at (208). Then, when monitoring parameter P... m When the abnormal threshold 402 is exceeded, the controller 302 can determine an alarm event 318. Alarm event 318 can be stored by the controller 302 (e.g., in its memory device 152). For example, alarm event 318 can be determined by a monitoring parameter P that includes exceeding the abnormal threshold 402. m The value is represented by the timestamp of the data.

[0035] Return to reference Figure 4 As shown at (212), method 200 includes actuating HMI 316 via controller 302 to output an alarm based on comparing an alarm event with an alarm threshold. For example, as Figure 5 As shown, controller 302 may also include HMI 316. HMI 316 enables a user to interact with controller 302. In some embodiments, HMI 316 may include one or more interfaces, such as a display screen, to display information to the user, and may also include one or more interfaces, such as a touchscreen component, a mouse component, a keyboard component, a stylus component, etc., to allow the user to interact with the information displayed on the screen. In some embodiments, HMI 316 may include one or more interfaces, such as a speaker, to present audio information to the user. That is, controller 302 may control HMI 316 to output audio and / or visual information to the user. In addition, HMI 316 may receive information from the user. For example, HMI 316 may receive user input that specifies information to controller 302 (e.g., via a sensor that detects the user pressing a virtual button on a touchscreen, via a mouse component that receives user input specifying a selection of information displayed on the screen, via a keyboard component that receives user input specifying alphanumeric information, etc.).

[0036] As an example, Figure 8 Provided the mapping monitoring parameter P m A graph 408 comparing the measured value (shown via solid line 410) with the corresponding value of the anomaly threshold (shown via dashed line 412) over time. Alerts may include audio and / or visual information indicating abnormal behavior of an asset or group of assets. In some embodiments, the HMI 316 may also be actuated to output previous alerts and timestamps associated with those previous alerts.

[0037] The alarm threshold can specify the maximum number of expected anomalies within the monitoring period 414, as described below. In an embodiment, the alarm threshold can be determined based on historical data used to generate conditional probability distributions 310a, 310b, 310c, and 310d. For example, the alarm threshold can be determined based on the maximum number of alarm events within the period during which historical data was collected. Alternatively, the alarm threshold can be predetermined based on the design and / or performance parameters of the asset (e.g., specified by the manufacturer of the asset or its components).

[0038] In one embodiment, method 200 may include actuating HMI 316 via controller 302 when a score derived from an alarm event exceeds an alarm threshold. In another embodiment, method 200 may include counting the number of alarm events within a monitoring period 414 to determine a score, and then comparing the score to an alarm threshold.

[0039] like Figure 8 As shown, monitoring period 414 can be defined, for example, by a time quantity (e.g., one minute, one hour, one day, one week, etc.). In such an example, monitoring period 414 can include a time period defined by a predetermined time quantity ending at the current time and preceding the current time (e.g., the time when the value of the monitoring parameter is captured). As another example, given the operational state of an asset (e.g., startup, normal power generation, shutdown, etc.), monitoring period 414 can be defined by the number of data collection instances expected to be collected within the time quantity for the monitoring parameter. The operational state of an asset can be defined by one or more parameters of the asset. One or more parameters can be the same as or different from one or more classification parameters. As an example, data can be collected at a certain sampling rate (i.e., a predetermined number of instances per unit time, such as 144 instances per day). Therefore, monitoring period 414 can include a time period ending at the current time and preceding the time quantity for collecting data for monitoring parameter P given the operational state of the asset. m The time period is defined by a specified number of instances of the data (e.g., 144).

[0040] Furthermore, in another embodiment, method 200 may include based on monitoring parameter P. m The difference between the value of the alarm and the anomaly threshold 402 determines the corresponding scale value for each alarm event within the monitoring period 414. For example, this can be achieved by monitoring parameter P. m The scaling value is determined by the ratio of the value to the anomaly threshold of 402. As another example, the scaling value can be determined by monitoring the parameter P. m The scaling value is determined by the ratio of the first difference between the value and the mean of the corresponding conditional probability distributions 310a, 310b, 310c, and 310d, and the second difference between the outlier threshold 402 and the mean of the corresponding conditional probability distributions 310a, 310b, 310c, and 310d.

[0041] Method 200 may further include determining a score based on combining (e.g., via addition or multiplication) corresponding scale values ​​within the monitoring period 414. Method 200 may also include actuating HMI 316, as described above, when the score exceeds an alarm threshold. In such an embodiment, alarm events may be weighted by corresponding scale values. Weighting alarm events can adjust the sensitivity of the output alarm, which can reduce instances of unwanted alarm outputs.

[0042] Various aspects and embodiments of the present invention are defined by the following numbered clauses: A method for detecting anomalies during the operation of an asset, the method comprising: collecting data associated with the operation of the asset via a controller, the data including operational parameters of the asset and environmental parameters surrounding the asset; selecting a monitoring parameter and one or more classification parameters, the monitoring parameter being one of the operational parameters, and the one or more classification parameters being at least one of one or more other operational parameters or one or more environmental parameters; selecting an anomaly function for the monitoring parameter via the controller, given the one or more classification parameters, and determining the anomaly function based on a corresponding conditional probability distribution of the monitoring parameter, given the one or more classification parameters; determining an anomaly threshold for the monitoring parameter via the controller based on the anomaly function and the one or more classification parameters; generating an alarm event via the controller when the monitoring parameter exceeds the anomaly threshold; and actuating a human-machine interface via the controller to output an alarm based on a score derived from the alarm event and the alarm threshold.

[0043] The method described under any of the foregoing provisions also includes generating the corresponding conditional probability distribution via the controller based on historical data associated with the operation of the asset.

[0044] The methods described under any of the foregoing provisions, wherein the historical data is also associated with the operation of the asset group that includes the asset.

[0045] The method described under any of the foregoing provisions also includes determining the alarm threshold based on the historical data.

[0046] According to any of the methods described in the foregoing clauses, the alarm threshold specifies the maximum number of expected anomalies within the monitoring period.

[0047] According to any of the foregoing provisions of the method, the monitoring period is defined, given the operational status of an asset, by the amount of time expected to be used for monitoring parameters or the number of data collection instances expected to be collected within said amount of time.

[0048] The method according to any of the foregoing provisions, wherein actuating the HMI via the controller to output an alarm based on comparing a score derived from an alarm event with an alarm threshold further includes: actuating the HMI when the score exceeds the alarm threshold.

[0049] The method described under any of the foregoing provisions also includes determining a score by counting the number of alarm events within a monitoring period, defined by the amount of time used for monitoring parameters or the number of data collection instances expected to be collected within said amount of time, given the operational status of an asset.

[0050] According to any of the foregoing provisions, wherein actuating the human-machine interface via the controller to output the alarm based on comparing the score derived from the alarm event with the alarm threshold further includes: determining a corresponding scale value for each alarm event within the monitoring period based on the difference between the monitoring parameters and the anomaly threshold; determining a score based on combining the corresponding scale values; and actuating the human-machine interface when the score exceeds the alarm threshold.

[0051] According to any of the foregoing provisions of the method, the monitoring period is defined by the amount of time used for the monitoring parameters or the number of data collection instances expected to be collected within the amount of time, given the operational status of the asset.

[0052] A system for detecting anomalies during the operation of an asset, the system comprising: a controller communicatively coupled to the asset, the controller being configured to perform a plurality of operations including: collecting data associated with the operation of the asset via the controller, the data including operational parameters of the asset and environmental parameters surrounding the asset; selecting monitoring parameters and one or more classification parameters, the monitoring parameters being one of the operational parameters, and the one or more classification parameters being at least one of one or more other operational parameters or one or more environmental parameters; selecting an anomaly function for the monitoring parameter via the controller, given the one or more classification parameters, and determining the anomaly function based on a corresponding conditional probability distribution of the monitoring parameter, given the one or more classification parameters; determining an anomaly threshold for the monitoring parameter via the controller based on the anomaly function and the one or more classification parameters; generating an alarm event via the controller when the monitoring parameter exceeds the anomaly threshold; and actuating a human-machine interface via the controller to output an alarm based on a score derived from the alarm event and an alarm threshold.

[0053] According to any of the foregoing provisions, the plurality of operations further include: generating a corresponding conditional probability distribution via the controller based on historical data associated with operations of the asset.

[0054] The system as described in any of the foregoing provisions, wherein the historical data is also associated with the operation of the asset group that includes the asset.

[0055] According to any of the foregoing provisions, the plurality of operations further include: determining the alarm threshold based on the historical data.

[0056] According to any of the foregoing provisions, the alarm threshold specifies the maximum number of expected anomalies within a monitoring period.

[0057] According to any of the foregoing provisions, the monitoring period is defined by the amount of time used for the monitoring parameters or the number of data collection instances expected to be collected within the amount of time, given the operational status of an asset.

[0058] According to any of the foregoing provisions, the process of actuating the human-machine interface via the controller to output an alarm based on a comparison of a score derived from an alarm event with an alarm threshold further includes actuating the human-machine interface when the score exceeds the alarm threshold.

[0059] According to any of the foregoing provisions, the plurality of operations further include: determining, given the operational state of an asset, the number of alarm events during the monitoring period by counting the number of data collection instances defined by the amount of time used for the monitoring parameters or the amount of data expected to be collected within the amount of time.

[0060] According to any of the foregoing provisions, the system wherein actuating the human-machine interface via the controller to output the alarm based on comparing the score derived from the alarm event with the alarm threshold further includes: determining a corresponding scale value for each alarm event within a monitoring period based on the difference between the value of the monitoring parameter and the anomaly threshold; determining a score based on combining the corresponding scale values; and actuating the human-machine interface when the score exceeds the alarm threshold.

[0061] According to any of the foregoing provisions, the monitoring period is defined by the amount of time used for the monitoring parameters or the number of data collection instances expected to be collected within the amount of time, given the operational status of an asset.

[0062] This written description uses examples including the best mode to disclose the invention and also enables any person skilled in the art to practice the invention, including making and using any apparatus or system, and performing any combination method. The patentable scope of the invention is defined by the claims and may include other examples that may occur to those skilled in the art. Such other examples are expected to fall within the scope of the claims if they have structural units that are exactly the same as the wording of the claims, or if they contain equivalent structural units that have a non-substantially different wording from the claims.

Claims

1. A method for detecting anomalies during asset operations, the method comprising: Data related to the operation of the asset is collected via a controller, including the asset's operating parameters and environmental parameters surrounding the asset. Select a monitoring parameter and one or more classification parameters, wherein the monitoring parameter is one of the operating parameters, and the one or more classification parameters are at least one of one or more other operating parameters or one or more environmental parameters. Given one or more classification parameters, an anomaly function is selected for the monitoring parameters via the controller, and the anomaly function is determined based on the corresponding conditional probability distribution of the monitoring parameters, given one or more classification parameters. The controller determines an anomaly threshold for the monitoring parameters based on the anomaly function and the one or more classification parameters. When the monitored parameter exceeds the abnormal threshold, an alarm event is generated via the controller; as well as An alarm is output via the controller by comparing a score derived from the alarm event with an alarm threshold.

2. The method of claim 1, further comprising generating the corresponding conditional probability distribution via the controller based on historical data associated with the operation of the asset.

3. The method according to claim 2, wherein, The historical data is also associated with the operations of the asset group that includes the asset.

4. The method according to claim 2, further comprising determining the alarm threshold based on the historical data.

5. The method according to claim 1, wherein, The alarm threshold specifies the maximum number of expected anomalies within the monitoring period.

6. The method according to claim 5, wherein, Given the operational status of the asset, the monitoring period is defined by the amount of time used for the monitoring parameters or the number of data collection instances expected to be collected within the amount of time.

7. The method according to claim 1, wherein, The process of actuating the human-machine interface via the controller to output the alarm, based on a comparison between the score derived from the alarm event and the alarm threshold, further includes: The human-machine interface is activated when the score exceeds the alarm threshold.

8. The method of claim 7, further comprising determining the score by counting the number of alarm events within a monitoring period defined by the amount of time used for the monitoring parameters or the number of data collection instances expected to be collected within the amount of time, given the operational status of the asset.

9. The method according to claim 1, wherein, The process of actuating the human-machine interface via the controller to output the alarm, based on a comparison of the score derived from the alarm event with the alarm threshold, further includes: The corresponding scale value for each alarm event within the monitoring period is determined based on the difference between the monitoring parameter and the anomaly threshold. The score is determined based on the combination of the corresponding scale values; and The human-machine interface is activated when the score exceeds the alarm threshold.

10. The method according to claim 9, wherein, Given the operational status of the asset, the monitoring period is defined by the amount of time used for the monitoring parameters or the number of data collection instances expected to be collected within the amount of time.

Images