Fault prediction method and device, storage medium and electronic device

By building a neural network model and using the fan's historical operation data and alarm log, the problem of unreasonable fan failure prediction is solved, more accurate fault prediction is achieved, and the management efficiency of wind farms is improved.

CN119046793APending Publication Date: 2024-11-29HUANENG RENEWABLES SHANGHAI POWER GENERATION CO LTD +1
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202411131053.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The failure prediction scheme for fans in the prior art is unreasonable, resulting in inaccurate prediction.

Method used

A neural network model is built through the historical operation data of the fan, and a model is trained using the alarm log to generate a fuzzy evaluation matrix for failure prediction, including determining the set of evaluation indicators, evaluation level and membership, and building a second model for prediction.

Benefits of technology

It improves the accuracy of fan failure prediction, reduces unnecessary maintenance costs, and enhances the management efficiency of wind farms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119046793A_ABST
    Figure CN119046793A_ABST
Patent Text Reader

Abstract

The invention discloses a fault prediction method and device, a storage medium and an electronic device.The method comprises the steps that operation data of a draught fan are input into a first model so as to indicate the first model to conduct fault prediction on the draught fan, the first model is obtained by training a second model through an alarm log included in historical operation data of the fan, and the second model is a neural network model constructed by taking the historical operation data as an input sample and taking a fuzzy evaluation matrix corresponding to the historical operation data as an output sample. The problem that a fan fault prediction scheme is unreasonable is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of wind power generation, and in particular to a fault prediction method and device, a storage medium and an electronic device. Background Art

[0002] In the related art, the operating data (i.e., historical operating data) obtained by monitoring the fan is used to train a fault prediction model to predict the fan failure. However, the operating data obtained by monitoring cannot accurately express the state of the fan, resulting in an unreasonable fault prediction scheme for the fan.

[0003] In view of the problem of unreasonable fault prediction scheme for wind turbines in related technologies, no effective solution has been found yet. Summary of the invention

[0004] The embodiments of the present application provide a fault prediction method and device, a storage medium, and an electronic device to at least solve the problem of unreasonable fault prediction schemes for wind turbines in the prior art.

[0005] According to one embodiment of the embodiments of the present application, a fault prediction method is provided, comprising: inputting operating data of a wind turbine into a first model to instruct the first model to perform fault prediction on the wind turbine, wherein the first model is obtained by training a second model with an alarm log included in historical operating data of the wind turbine, and the second model is a neural network model constructed with the historical operating data as input samples and a fuzzy evaluation matrix corresponding to the historical operating data as output samples.

[0006] In an exemplary embodiment, before the operating data of the wind turbine is input into the first model to instruct the first model to predict faults of the wind turbine, the method also includes: obtaining the fuzzy evaluation matrix through the historical operating data, and determining the model structure of the second model; constructing the second model with the model structure using the historical operating data as input samples and the fuzzy evaluation matrix as output samples; extracting the alarm log in the historical operating data, and training the second model through the alarm log to obtain the first model.

[0007] In an exemplary embodiment, obtaining the fuzzy evaluation matrix through the historical operation data includes: determining an evaluation index set corresponding to the historical operation data, and determining an evaluation level corresponding to the historical operation data; determining a membership degree between the evaluation index set and the evaluation level; and determining the fuzzy evaluation matrix through the membership degree.

[0008] In an exemplary embodiment, determining the evaluation indicator set corresponding to the historical operating data includes: determining the evaluation indicator set corresponding to the historical operating data, wherein the evaluation indicator set includes at least: the average failure-free time of the wind turbine operating in a target mode, the average repair time of the wind turbine operating in a target mode, wherein the target mode includes one of the following: a failure mode, an operating mode.

[0009] In an exemplary embodiment, determining the model structure of the second model includes: determining a first activation function of a hidden layer of the second model, and determining a second activation function of an output layer of the second model; and determining the model structure of the second model by means of the first activation function and the second activation function.

[0010] In an exemplary embodiment, the second model is trained by the alarm log to obtain the first model, including: determining a mapping relationship between the alarm log and a fault category label; labeling the alarm log according to the mapping relationship; and training the second model by the labeled alarm log to obtain the first model.

[0011] In an exemplary embodiment, after the operating data of the wind turbine is input into the first model to instruct the first model to perform fault prediction on the wind turbine, the method further includes: determining a prediction result obtained by performing fault prediction on the wind turbine, wherein the prediction result includes one of the following: presence of a fault, no fault; and generating an alarm event based on the prediction result.

[0012] According to another embodiment of the embodiment of the present application, a fault prediction device is also provided, including: an input module for inputting operating data of the fan into a first model to instruct the first model to perform fault prediction on the fan, wherein the first model is obtained by training a second model with an alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and a fuzzy evaluation matrix corresponding to the historical operating data as output samples.

[0013] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above method when running.

[0014] According to another aspect of the embodiments of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the method through the computer program.

[0015] According to another aspect of the embodiments of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0016] In an embodiment of the present application, the operation data of the fan is input into the first model to instruct the first model to predict the fault of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operation data of the fan, and the second model is a neural network model constructed with the historical operation data as input samples and the fuzzy evaluation matrix corresponding to the historical operation data as output samples. Through the above embodiment, the problem of unreasonable fault prediction scheme for the fan in the prior art is solved, making the fault prediction of the fan more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 It is a hardware structure block diagram of a computer terminal of a fault prediction method according to an embodiment of the present application;

[0019] Figure 2 is a flow chart of a fault prediction method according to an embodiment of the present application;

[0020] Figure 3 is another flow chart of a fault prediction method according to an embodiment of the present application;

[0021] Figure 4 It is a structural block diagram of a fault prediction device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices; "plurality" means two or more.

[0024] The method embodiments provided in the embodiments of the present application can be executed in a computer terminal or similar computing device or cloud platform or independent physical server or software platform, wherein the above software platform is run through one or more servers. Taking running on a computer terminal as an example, Figure 1 1 is a hardware structure block diagram of a computer terminal of a fault prediction method according to an embodiment of the present application. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown in the figure) processor 102 and memory 104 for storing data. In an exemplary embodiment, the computer terminal may also include a transmission device 106 for communication functions and an input and output device 108, wherein the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA. A person skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Equivalent functions or comparisons shown Figure 1 A different configuration with more features is shown.

[0025] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as computer programs corresponding to the methods in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, that is, the above-mentioned methods are implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0026] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0027] In this embodiment, a fault prediction method is provided, which is applied to the above-mentioned computer terminal. Figure 2 is a flowchart of a fault prediction method according to an embodiment of the present application, the process comprising the following steps:

[0028] Step S202, inputting the operating data of the fan into the first model to instruct the first model to predict faults of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and the fuzzy evaluation matrix corresponding to the historical operating data as output samples.

[0029] Optionally, the second model may be an artificial neural network (ANN for short) model.

[0030] In an embodiment of the present application, the operation data of the fan is input into the first model to instruct the first model to predict the fault of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operation data of the fan, and the second model is a neural network model constructed with the historical operation data as input samples and the fuzzy evaluation matrix corresponding to the historical operation data as output samples. Through the above embodiment, the problem of unreasonable fault prediction scheme for the fan in the prior art is solved, making the fault prediction of the fan more accurate.

[0031] In an exemplary embodiment, before the operating data of the wind turbine is input into the first model to instruct the first model to predict faults of the wind turbine, the method also includes: obtaining the fuzzy evaluation matrix through the historical operating data, and determining the model structure of the second model; constructing the second model with the model structure using the historical operating data as input samples and the fuzzy evaluation matrix as output samples; extracting the alarm log in the historical operating data, and training the second model through the alarm log to obtain the first model.

[0032] Optionally, the model structure may be a model structure corresponding to a common ANN model.

[0033] In an embodiment of the present application, historical operation data is collected from a SCADA system for monitoring wind turbines; a mid- to long-term reliability fuzzy evaluation model is constructed; and a fuzzy evaluation matrix of the historical operation data is determined through the reliability fuzzy evaluation model.

[0034] The historical operation data and the module evaluation matrix are used as input samples and output samples respectively, and a second model with a determined model structure is constructed. The second model is trained through the alarm log in the historical operation data to obtain the first model, which is used to predict the fault of the fan according to the operation data of the fan. Optionally, the operation data is real-time operation data. Specifically, the second model is a model constructed by training an untrained model with the model structure through the historical operation data and the module evaluation matrix to obtain trained parameters.

[0035] It should be noted that the SCADA (Supervisory Control and Data Acquisition) system is a software system used to monitor and control industrial processes. It collects data through sensors and controllers, and then transmits this data to a central computer system so that operators can monitor and control industrial processes. SCADA systems are commonly used to monitor industrial fields such as power systems, water treatment plants, and transportation systems, and can monitor and remotely control the operation of these systems in real time. SCADA systems can help improve the efficiency, safety, and reliability of industrial processes.

[0036] Furthermore, obtaining the fuzzy evaluation matrix through the historical operation data includes: determining an evaluation index set corresponding to the historical operation data, and determining an evaluation level corresponding to the historical operation data; determining a membership degree between the evaluation index set and the evaluation level; and determining the fuzzy evaluation matrix through the membership degree.

[0037] Determining the evaluation index set corresponding to the historical operation data includes: determining the evaluation index set of the historical operation data according to a pre-established evaluation index system. The evaluation level refers to a set, for example: V = [1 2 3 4 5], where V is used to represent the evaluation level. The membership degree reflects the mapping relationship between the evaluation index set and the evaluation level.

[0038] More specifically, the degree of membership is used to indicate the mapping relationship between any evaluation indicator in the evaluation indicator set and a specific level in the evaluation level.

[0039] After the membership is determined, or at the same time as the membership is determined, weights may be calculated for each group of sub-run data being analyzed. Optionally, the weights may be calculated using Analytic Hierarchy Process (AHP)-entropy weights.

[0040]

[0041] Among them, W j represents the weight of each group of sub-run data, ω j is the weight calculated by AHP method, g j is the difference coefficient calculated for the entropy weight. Optionally, the historical operation data includes multiple groups of sub-operation data.

[0042] Further, determining the evaluation indicator set corresponding to the historical operation data includes: determining the evaluation indicator set corresponding to the historical operation data, wherein the evaluation indicator set includes at least: the average failure-free time of the wind turbine operating in the target mode, the average repair time of the wind turbine operating in the target mode, wherein the target mode includes one of the following: failure mode, operation mode.

[0043] Specifically, the evaluation index set of the historical operation data is determined according to the pre-established evaluation index system, including: according to the state of the wind turbine, the evaluation index system is constructed in the fault mode and the operation mode respectively. The evaluation index set of the fault and operation modes is as follows.

[0044] U failure =[MTBF.MTTR,C,D].

[0045] U operating =[MTBF.MTTR, N, AEP].

[0046] Among them, U failure The evaluation index set representing the failure mode, U operating It represents the evaluation index set of the operation mode, and MTBF, MTTR, C, D, N and AEP represent the mean time between failures, mean time to repair, criticality, fault detection difficulty, number of failure downtimes and annual energy production, respectively.

[0047]

[0048] Where T represents the total number of hours corresponding to the statistical historical operation data, T SCADA Indicates the number of hours the wind turbine is not connected to the SCADA system, T shutdown Indicates the number of downtime hours caused by faults in the fault mode, or the total number of downtime hours caused by faults in the operating mode, N fault Indicates the total number of faults that occurred in the Fault mode, or the total number of faults in the Run mode.

[0049] It should be noted that the data corresponding to any mode should be used to calculate the indicator data corresponding to any mode. For example, to calculate the MTBF corresponding to the failure mode, T shutdown represents the number of downtime hours caused by the failure in the failure mode, and N fault Indicates the total number of failures in the failure mode. The MTBF corresponding to the operation mode, the MTTR corresponding to the failure mode, the MTTR corresponding to the operation mode, and so on, are not repeated here.

[0050] In an exemplary embodiment, determining the model structure of the second model includes: determining a first activation function of a hidden layer of the second model, and determining a second activation function of an output layer of the second model; and determining the model structure of the second model by means of the first activation function and the second activation function.

[0051] Optionally, the neural network model is constructed with historical operation data as input samples and fuzzy evaluation matrix as output samples. The activation function of the hidden layer of the neural network model can be set to tan sigmoid, and the activation function (i.e., activation function) of the output layer of the neural network model can be set to log sigmoid. The neural network uses the backpropagation algorithm (BP) algorithm to learn nonlinear mapping relationships.

[0052] It should be noted that the BP algorithm is mainly used to train multi-layer feedforward neural networks (Multi-Layer Perceptron, MLP) to minimize the output error by adjusting the weights and biases in the network. The core idea of ​​the BP algorithm is to use the gradient descent method to update the network parameters, thereby achieving adaptive learning of the network.

[0053] In an exemplary embodiment, the second model is trained by the alarm log to obtain the first model, including: determining a mapping relationship between the alarm log and a fault category label; labeling the alarm log according to the mapping relationship; and training the second model by the labeled alarm log to obtain the first model.

[0054] The mapping relationship between the alarm log and the fault category label will assist in marking whether the fault corresponding to the alarm log is included in the time window of each group of sub-operation data corresponding to the alarm log.

[0055] In an exemplary embodiment, after the operating data of the wind turbine is input into the first model to instruct the first model to perform fault prediction on the wind turbine, the method further includes: determining a prediction result obtained by performing fault prediction on the wind turbine, wherein the prediction result includes one of the following: presence of a fault, no fault; and generating an alarm event based on the prediction result.

[0056] Optionally, when the prediction result is that there is a fault, it is allowed to generate a first-level alarm event, which is used to indicate that the fan has an operating fault within the time window corresponding to the operating data. When the prediction result is that there is no fault, it is allowed to generate a second-level alarm event, which is used to indicate that the fan is temporarily free of faults within the time window corresponding to the operating data.

[0057] In order to better understand the process of the above-mentioned fault prediction method, the above-mentioned fault prediction method is described below in combination with an optional embodiment, but it is not used to limit the technical solution of the embodiment of the present application.

[0058] The rapid development of the wind power industry requires a comprehensive evaluation of wind turbines and effective fault prediction. In response to the challenges of integration and optimization of reliability evaluation and fault prediction models, a fuzzy evaluation and fault prediction method for system reliability based on SCADA data (equivalent to the above-mentioned fault prediction method) is proposed. Specifically, a medium- and long-term reliability fuzzy evaluation model is established using a fuzzy comprehensive evaluation method. At the same time, an alarm log (equivalent to the alarm log in the above-mentioned embodiment) is used to study fault prediction, because the alarm log can better express the status of the wind turbine than monitoring data. In general, reliability evaluation and fault prediction complement each other, which may affect future wind farm management and avoid unnecessary maintenance costs.

[0059] Specifically, Figure 3 As shown, the above method can be implemented by the following steps 1 to 4, including:

[0060] Step 1: First, collect raw data (equivalent to historical operation data) from the SCADA system.

[0061] Step 2: Construct a medium- and long-term reliability fuzzy evaluation model. The evaluation results of the original data (i.e., the fuzzy evaluation matrix) are obtained through the medium- and long-term reliability fuzzy evaluation model. The original ANN model is trained based on the original data and the evaluation results, and the trained parameters are used to construct the ANN model (i.e., the second model). The ANN model is trained based on the alarm log in the original data to obtain the first model to evaluate the wind turbine. Finally, for wind turbines with low reliability, fault prediction is implemented based on the neural network algorithm. Specifically, the evaluation results of the original data are obtained through the following steps 21 to 25:

[0062] Step 21: Establish an evaluation index system and evaluation levels.

[0063] According to the status of the wind turbine, an evaluation index system is constructed in the fault mode and operation mode respectively. The evaluation index set of the fault and operation modes is as follows.

[0064] U failure =[MTBF.MTTR, C, D.]

[0065] U operating =[MTBF.MTTR, N, AEP].

[0066] MTBF, MTTR, C, D, N and AEP represent mean time between failures, mean time to repair, criticality, difficulty of fault detection, number of failure downtimes and annual energy output, respectively.

[0067]

[0068] Among them, T represents the total number of hours counted, T SCADAIndicates the number of hours without connection to the SCADA system, T shutdown Indicates the number of downtime hours caused by faults in the fault mode, or the total number of downtime hours caused by faults in the operation mode; N fault Indicates the total number of faults that occurred in the Fault mode, or the total number of faults in the Run mode.

[0069] Step 22: Standardize evaluation indicators.

[0070] The data matrix in the above evaluation index set U is normalized so that the value is in the range of [0, 1].

[0071]

[0072] x ij is the normalized value, x is the above parameter value, x max , x min is the maximum and minimum value of the parameter. Optionally, x can also represent running data

[0073] Step 23: Determine the degree of membership.

[0074] Membership degree r ij Reflects the evaluation index u i and level v j The mapping relationship between them is that level V = [1 2 3 45], and the collected operation data is directly divided into the level of V. i Indicates U failure , U operating The elements in v j Represents an element in V.

[0075] Step 24: Set weights. Optionally, set a weight for each set of run data being analyzed.

[0076] The weight calculation adopts the analytic hierarchy process (AHP)-entropy weight.

[0077]

[0078] Among them, where ω i is the weight calculated by AHP method, g j Difference coefficient calculated for entropy weight.

[0079] Step 25: Calculate the reliability fuzzy evaluation matrix.

[0080] In the time window t k Period x ij The evaluation result can be calculated by the following formula.

[0081]

[0082] Among them, R(x ij ) is r ij The matrix formed.

[0083] In order to improve the accuracy of long-term evaluation results, a time-varying factor is applied, which can be calculated by a monotonically increasing function F(x). k The time-varying factor of can be calculated as follows:

[0084]

[0085] The value of α is determined according to the actual situation and its value range is [0, 1].

[0086] After grouping, the long-term reliability fuzzy evaluation matrix is ​​expressed as follows.

[0087]

[0088] The maximum membership principle is used to analyze the reliability of fuzzy evaluation results.

[0089] Step 3: Model building and training.

[0090] The original data is used as the input sample and the fuzzy evaluation matrix is ​​used as the output signal (i.e., the output sample) to construct a neural network model (i.e., the second model). The activation function of the hidden layer is set to tan sigmoid, and the activation function of the output layer is set to log sigmoid. The neural network uses the BP algorithm to learn nonlinear mapping relationships. 70% of the data in the original data and the fuzzy evaluation matrix are selected as training samples, and 30% of the data in the original data and the fuzzy evaluation matrix are selected as test samples; the original ANN model is trained in this way, and the ANN model (i.e., the second model) is constructed by the parameters obtained after training. The reliability of the evaluation model is then measured by the mean square error (MSE).

[0091] In order to perform fault prediction, each alarm log that will cause the wind turbine to restart is first selected as a sample. The alarm logs are aggregated by week and converted to the corresponding average values. The alarm logs (equivalent to alarm logs) are divided into training and test data sets. Since fault prediction is a few-shot learning problem and is prone to overfitting, the data set division time varies for different faults. When processing data, null values ​​are inevitable. If the null ratio exceeds 50%, the indication will be omitted. Then, the mapping relationship between the input variables and the category labels is mined, and it is marked whether the fault is included in the time window. If a fault occurs, the corresponding label will be assigned a value of 1, and if not, a value of 0 will be assigned. Finally, the prediction fault models were obtained based on the neural network algorithm.

[0092] Training the second model through alarm logs can be divided into the following steps:

[0093] Step 31, obtain alarm data (equivalent to alarm log) and divide it into training set and test set;

[0094] Step 32: alarm data processing, if the empty ratio exceeds 50%, the indication is omitted;

[0095] Step 33: Map alarm data to category labels, mark whether the fault is in the time window, fault is marked as 1, no mark is 0;

[0096] Step 34: Input standard post-alert data and train the model.

[0097] Step 4: Output the trained model (i.e. the first model).

[0098] In summary, in the optional embodiment of the present application, reliability assessment and fault prediction are integrated. In the related art, the research on reliability assessment and fault prediction is relatively independent. The optional embodiment of the present application is based on the diversity of reliability indicators and the comprehensiveness of the evaluation results, and the deep learning neural network can find the nonlinear relationship of "multiple inputs and multiple outputs", and realizes the effective integration of reliability assessment and fault prediction through the deep neural network.

[0099] Furthermore, in the related art, there is no consistent principle or standard applicable to fans in different states. The optional embodiment of the present application is based on the fuzzy comprehensive evaluation (FCE) theory, by establishing an evaluation index system and evaluation level, using time-varying factors to collect evaluation information at each stage, and constructing a medium- and long-term reliability fuzzy evaluation model. The proposed model can help map the overall state of the fan in fault and operation modes, and determine the overall health of the fan in a complementary way.

[0100] In addition, the optional embodiment of the present application optimizes fault prediction. Mining SCADA log files is as important as analyzing real-time sensor data. Due to the lack of labels for component failures, this method conducts in-depth analysis of SCADA logs and ledgers, corresponds the operation failures of the fan to the data time window, simplifies the labeling process, and makes it easier to obtain labeled sample data.

[0101] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0102] The present application also provides a structural block diagram of a fault prediction device. Figure 4 is a structural block diagram of a fault prediction device according to an embodiment of the present application. Figure 4 As shown, including:

[0103] The input module 42 is used to input the operating data of the fan into the first model to instruct the first model to predict faults of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and the fuzzy evaluation matrix corresponding to the historical operating data as output samples.

[0104] In an embodiment of the present application, the operation data of the fan is input into the first model to instruct the first model to predict the fault of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operation data of the fan, and the second model is a neural network model constructed with the historical operation data as input samples and the fuzzy evaluation matrix corresponding to the historical operation data as output samples. Through the above embodiment, the problem of unreasonable fault prediction scheme for the fan in the prior art is solved, making the fault prediction of the fan more accurate.

[0105] In an exemplary embodiment, before the operating data of the wind turbine is input into the first model to instruct the first model to predict faults of the wind turbine, the device also includes a training module, which is used to: obtain the fuzzy evaluation matrix through the historical operating data, and determine the model structure of the second model; construct the second model with the model structure using the historical operating data as input samples and the fuzzy evaluation matrix as output samples; extract the alarm log in the historical operating data, and train the second model through the alarm log to obtain the first model.

[0106] In an exemplary embodiment, the training module is also used to: determine the evaluation indicator set corresponding to the historical operation data, and determine the evaluation level corresponding to the historical operation data; determine the membership between the evaluation indicator set and the evaluation level; and determine the fuzzy evaluation matrix through the membership.

[0107] In an exemplary embodiment, the training module is also used to: determine a set of evaluation indicators corresponding to the historical operating data, wherein the set of evaluation indicators at least includes: the average failure-free time of the wind turbine operating in a target mode, the average repair time of the wind turbine operating in a target mode, wherein the target mode includes one of the following: a failure mode, an operating mode.

[0108] In an exemplary embodiment, the training module is also used to: determine a first activation function of a hidden layer of the second model, and determine a second activation function of an output layer of the second model; and determine a model structure of the second model through the first activation function and the second activation function.

[0109] In an exemplary embodiment, the training module is further used to: determine the mapping relationship between the alarm log and the fault category label; annotate the alarm log according to the mapping relationship; and train the second model through the annotated alarm log to obtain the first model.

[0110] In an exemplary embodiment, after the operating data of the wind turbine is input into the first model to instruct the first model to perform fault prediction on the wind turbine, the device also includes a generation module for: determining a prediction result obtained by performing fault prediction on the wind turbine, wherein the prediction result includes one of the following: presence of a fault, no fault; and generating an alarm event based on the prediction result.

[0111] Optionally, when the prediction result is that there is a fault, it is allowed to generate a first-level alarm event, which is used to indicate that the fan has an operating fault within the time window corresponding to the operating data. When the prediction result is that there is no fault, it is allowed to generate a second-level alarm event, which is used to indicate that the fan is temporarily free of faults within the time window corresponding to the operating data.

[0112] An embodiment of the present application further provides a storage medium, which includes a stored program, wherein the program executes any of the above methods when it is run.

[0113] Optionally, in this embodiment, the storage medium may be configured to store program codes for executing the following steps:

[0114] S1. Inputting the operating data of the fan into the first model to instruct the first model to predict faults of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and the fuzzy evaluation matrix corresponding to the historical operating data as output samples.

[0115] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0116] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0117] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0118] S1. Inputting the operating data of the fan into the first model to instruct the first model to predict faults of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and the fuzzy evaluation matrix corresponding to the historical operating data as output samples.

[0119] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.

[0120] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0121] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0122] S1. Inputting the operating data of the fan into the first model to instruct the first model to predict faults of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and the fuzzy evaluation matrix corresponding to the historical operating data as output samples.

[0123] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0124] An embodiment of the present application also provides a computer program, which includes computer instructions stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in any one of the above method embodiments.

[0125] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0126] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0127] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A fault prediction method, characterized in that: include: The operating data of the fan is input into the first model to instruct the first model to predict faults of the fan, wherein the first model is obtained by training the second model with the alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and the fuzzy evaluation matrix corresponding to the historical operating data as output samples.

2. The fault prediction method according to claim 1, characterized in that: Before inputting the operation data of the wind turbine into the first model to instruct the first model to perform fault prediction on the wind turbine, the method further includes: Acquiring the fuzzy evaluation matrix through the historical operation data, and determining the model structure of the second model; Constructing the second model having the model structure, using the historical operation data as input samples and the fuzzy evaluation matrix as output samples; Extract the alarm log in the historical operation data, and train the second model through the alarm log to obtain the first model.

3. The fault prediction method according to claim 2, characterized in that: Acquiring the fuzzy evaluation matrix through the historical operation data includes: Determining an evaluation indicator set corresponding to the historical operation data, and determining an evaluation level corresponding to the historical operation data; Determining the degree of membership between the evaluation indicator set and the evaluation level; The fuzzy evaluation matrix is ​​determined by the membership degree.

4. The fault prediction method according to claim 3, characterized in that: Determining an evaluation indicator set corresponding to the historical operation data includes: Determine an evaluation indicator set corresponding to the historical operation data, wherein the evaluation indicator set includes at least: the average failure-free time of the wind turbine operating in a target mode, the average repair time of the wind turbine operating in a target mode, wherein the target mode includes one of the following: a failure mode, an operation mode.

5. The fault prediction method according to claim 2, characterized in that: Determining a model structure of the second model includes: determining a first activation function for a hidden layer of the second model, and determining a second activation function for an output layer of the second model; The model structure of the second model is determined by the first activation function and the second activation function.

6. The fault prediction method according to claim 2, characterized in that: Training the second model by using the alarm log to obtain the first model includes: Determine the mapping relationship between the alarm log and the fault category label; Annotate the alarm log according to the mapping relationship; The second model is trained by using the annotated alarm logs to obtain the first model.

7. The fault prediction method according to claim 1, characterized in that: After inputting the operation data of the wind turbine into the first model to instruct the first model to perform fault prediction on the wind turbine, the method further includes: Determine a prediction result obtained by performing fault prediction on the wind turbine, wherein the prediction result includes one of the following: a fault exists, or there is no fault; An alarm event is generated according to the prediction result.

8. A fault prediction device, characterized in that: include: An input module is used to input the operating data of the fan into a first model to instruct the first model to predict faults of the fan, wherein the first model is obtained by training a second model with the alarm log included in the historical operating data of the fan, and the second model is a neural network model constructed with the historical operating data as input samples and the fuzzy evaluation matrix corresponding to the historical operating data as output samples.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method described in any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Citation Information

Patent Citations

  • Deep learning-based fan failure rate prediction method and device

    CN117473373A