Method and system for predicting residual life of large power generation equipment, and storage medium

Abstract

The invention discloses a large power generation equipment residual life prediction method and system and a storage medium, and relates to the field of equipment life prediction. The method comprises the following steps: acquiring life-related data of large power generation equipment, including design and manufacturing data, historical maintenance data and historical use state data, preprocessing the life-related data, and constructing a life-related data set of the large power generation equipment; based on the design and manufacturing data, constructing a digital twinborn model of the initial life of the large power generation equipment, performing deep learning on historical maintenance data and historical use data by using a deep learning model, generating a historical maintenance data-historical use data correlation curve, and determining correlation parameters of a use state and maintenance; and generating a life prediction digital twinborn model, inputting the current life related parameters of the to-be-predicted large power generation equipment into the life prediction digital twinborn model, and generating the residual life of the to-be-predicted large power generation equipment. According to the invention, the availability and production efficiency of equipment can be improved, so that the operation cost of an enterprise is reduced.

Description

Technical Field

[0001] The present invention relates to the field of equipment life prediction, and more particularly to a method, system, and storage medium for predicting the remaining life of large-scale power generation equipment. Background Art

[0002] With the increase in the service life, large-scale power generation equipment will inevitably experience problems such as wear and aging, which significantly increase the probability of equipment failures. Research shows that the proportion of failures caused by equipment aging in the total number of failures shows an increasing trend year by year, seriously threatening the safe and stable operation of the power system. Therefore, accurately predicting the remaining life of large-scale power generation equipment can not only help power enterprises make advance equipment maintenance and replacement plans in advance, reduce the possibility of failures, but also effectively optimize the operation management of equipment, extend the actual service life of equipment, and thus significantly reduce the operation cost.

[0003] Currently, the prediction methods for the remaining life of large-scale power generation equipment are mainly divided into three types: based on physical models, data-driven, and a combination of the two. The method based on physical models usually requires an in-depth understanding of the internal structure and operation mechanism of the equipment, and establishes an accurate physical model to simulate the aging process of the equipment. However, the structure of large-scale power generation equipment is extremely complex, and the operating environment is also ever-changing, resulting in great difficulty in establishing an accurate physical model, high computational cost of the model, and poor generality. The data-driven method relies on a large amount of historical data and uses machine learning and deep learning algorithms to mine the potential laws in the data. However, such methods are easily affected by data quality and noise, and the reliability and interpretability of the prediction results are poor. The method of combining physical models with data-driven, although integrating the advantages of both to a certain extent, still faces many technical problems in the process of model fusion, resulting in the prediction accuracy and stability being difficult to meet the requirements of actual engineering.

[0004] Therefore, developing a more accurate, efficient, and adaptable method for predicting the remaining life of large-scale power generation equipment has important practical significance for improving the operation reliability of the power system and reducing the operation and maintenance costs. Summary of the Invention

[0005] In view of this, the present invention provides a method, system, and storage medium for predicting the remaining life of large-scale power generation equipment, reducing equipment failure downtime, improving equipment availability and production efficiency, and thus reducing the operation cost of enterprises.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for predicting the remaining life of large-scale power generation equipment, comprising the following steps:

[0008] Obtain the life-related data of large-scale power generation equipment, including design and manufacturing data, historical maintenance data, and historical usage status data. Preprocess the life-related data to construct a dataset related to the life of large-scale power generation equipment;

[0009] Based on the design and manufacturing data, construct an initial life digital twin model of large-scale power generation equipment. Use a deep learning model to perform deep learning on historical maintenance data and historical usage data to generate a correlation curve of historical maintenance data - historical usage data, and determine the correlation parameters between the usage status and maintenance;

[0010] Input the correlation parameters into the initial life digital twin model to generate a life prediction digital twin model. Input the current life-related parameters of the large-scale power generation equipment to be predicted into the life prediction digital twin model to generate the remaining life of the large-scale power generation equipment to be predicted;

[0011] Based on the current operating vibration data of the large-scale power generation equipment to be predicted, determine the current operating condition of the equipment. According to the operating condition, judge whether the prediction of the remaining life of the large-scale power generation equipment to be predicted is accurate. If not, correct the loss function of the life prediction digital twin model and predict the remaining life again.

[0012] Optionally, install a vibration signal acquisition device RFID on the large-scale power generation equipment to be predicted. Collect the vibration signal of the large-scale power generation equipment to be predicted through the vibration signal acquisition device RFID and transmit it to the control terminal. The control terminal preprocesses the vibration signal to determine the current operating condition of the equipment.

[0013] Optionally, determining the current operating condition of the equipment specifically includes the following steps:

[0014] Determine the current life stage of the equipment according to the vibration signal, and at the same time perform time-frequency analysis on the vibration signal to generate a time-frequency diagram;

[0015] Combine the distinguished life stage with the characteristic information in the time-frequency diagram to form an input dataset;

[0016] Based on the time series prediction model of the Transformer architecture, introduce the attention mechanism to generate a life prediction model based on the vibration signal. Input the input dataset into the life prediction model to obtain the life condition of the current equipment.

[0017] Optionally, it also includes denoising the vibration signal. The denoising process includes: performing mean removal on the vibration signal data of the current equipment; performing Hilbert transform on the signal after mean removal to obtain the signal envelope value.

[0018] Optionally, the remaining life prediction digital twin model uses a multimodal deep learning framework to perform fusion analysis on historical operation data and real-time operation data. Based on the results of the fusion analysis, a hybrid remaining life prediction digital twin model that includes a physical model and a data-driven model is constructed.

[0019] Optionally, determine the correlation parameters between the usage status and maintenance. Specifically, extract the characteristic parameters of historical maintenance data and historical usage data from the historical maintenance data - historical usage data correlation curve, and use the Pearson correlation coefficient to perform correlation analysis on the characteristic parameters of historical maintenance data and historical usage data to obtain the correlation parameters.

[0020] A remaining life prediction system for large-scale power generation equipment includes the following steps:

[0021] Remaining life-related data collection module: used to obtain the remaining life-related data of large-scale power generation equipment, including design and manufacturing data, historical maintenance data, and historical usage status data, preprocess the remaining life-related data, and construct a remaining life-related data set for large-scale power generation equipment;

[0022] Correlation parameter generation module: used to construct an initial remaining life digital twin model for large-scale power generation equipment based on design and manufacturing data, use a deep learning model to perform deep learning on historical maintenance data and historical usage data, generate a historical maintenance data - historical usage data correlation curve, and determine the correlation parameters between the usage status and maintenance;

[0023] Remaining life prediction module: used to input the correlation parameters into the initial remaining life digital twin model to generate a remaining life prediction digital twin model, and input the current remaining life-related parameters of the large-scale power generation equipment to be predicted into the remaining life prediction digital twin model to generate the remaining life of the large-scale power generation equipment to be predicted;

[0024] Remaining life correction module: used to determine the current operation status of the equipment based on the current operation vibration data of the large-scale power generation equipment to be predicted, judge whether the remaining life prediction of the large-scale power generation equipment to be predicted is accurate according to the operation status. If it is not accurate, correct the loss function of the remaining life prediction digital twin model and predict the remaining life again.

[0025] A computer storage medium stores a computer program thereon, and when the computer program is executed by a processor, it implements the steps of any one of the remaining life prediction methods for large-scale power generation equipment.

[0026] Through the above technical solutions, compared with the prior art, the present invention provides a remaining life prediction method, system and storage medium for large-scale power generation equipment, and has the following beneficial effects:

[0027] 1. By obtaining design and manufacturing data, historical maintenance data, and historical usage status data, various factors affecting the equipment life can be comprehensively considered. Using a deep learning model to mine the potential relationships between these data, generating correlation curves and determining correlation parameters, and then constructing a life prediction digital twin model to achieve accurate prediction of the remaining life of the equipment.

[0028] 2. Inputting the current life-related parameters of the equipment to be predicted into the model, the remaining life prediction results can be generated in real time according to the current state of the equipment, which can timely reflect the actual aging degree and remaining service life of the equipment, providing a timely and accurate basis for the operation and maintenance decision-making of the equipment.

[0029] 3. Judging the accuracy of the remaining life prediction based on the current operating vibration data, and being able to further verify the prediction results by considering the actual vibration situation during the operation of the equipment. If it is inaccurate, the loss function is corrected and predicted again to make the prediction results more in line with the actual operating conditions of the equipment, improving the reliability and credibility of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0031] Figure 1 It is a schematic flowchart of the method of the present invention:

[0032] Figure 2 It is a schematic structural diagram of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0034] An embodiment of the present invention discloses a method for predicting the remaining life of a large-scale power generation equipment, as Figure 1 shown, including the following steps:

[0035] Step 1: Obtain the life-related data of the large-scale power generation equipment, including design and manufacturing data, historical maintenance data, and historical usage status data, preprocess the life-related data, and construct a life-related data set of the large-scale power generation equipment;

[0036] Step 2: Build a digital twin model of the initial life of large-scale power generation equipment based on design and manufacturing data. Use a deep learning model to perform deep learning on historical maintenance data and historical usage data to generate a correlation curve of historical maintenance data - historical usage data, and determine the correlation parameters between the usage status and maintenance.

[0037] Step 3: Input the correlation parameters into the initial life digital twin model to generate a life prediction digital twin model. Input the current life-related parameters of the large-scale power generation equipment to be predicted into the life prediction digital twin model to generate the remaining life of the large-scale power generation equipment to be predicted.

[0038] Step 4: Determine the current operating condition of the large-scale power generation equipment to be predicted based on the current operating vibration data of the equipment. Judge whether the prediction of the remaining life of the large-scale power generation equipment to be predicted is accurate according to the operating condition. If it is not accurate, correct the loss function of the life prediction digital twin model and predict the remaining life again.

[0039] Further, in Step 4, install a vibration signal acquisition device RFID on the large-scale power generation equipment to be predicted. Collect the vibration signal of the large-scale power generation equipment to be predicted through the vibration signal acquisition device RFID and transmit it to the control terminal. The control terminal preprocesses the vibration signal to determine the current operating condition of the equipment.

[0040] Furthermore, determining the current operating condition of the equipment specifically includes the following steps:

[0041] Determine the current life stage of the equipment according to the vibration signal, and at the same time perform time-frequency analysis on the vibration signal to generate a time-frequency diagram.

[0042] Step 4.1: Combine the distinguished life stages with the characteristic information in the time-frequency diagram to form an input data set.

[0043] Step 4.2: Based on a time series prediction model based on the Transformer architecture, introduce an attention mechanism to generate a life prediction model based on vibration signals. Input the input data set into the life prediction model to obtain the life condition of the current equipment.

[0044] The life prediction model includes:

[0045] Encoder part: Encode the input time series data (i.e., the constructed input data set) to extract the key features in the time series. The encoder consists of multiple stacked self-attention modules, and each layer of the module contains a multi-head self-attention mechanism and a feed-forward neural network.

[0046] Decoder part: Predict the remaining life of large-scale power generation equipment based on the output of the encoder. The decoder also contains multiple layers of self-attention modules, and positional encoding is introduced during the prediction process to retain the temporal information of the time series.

[0047] Improvement of the attention mechanism: The Informer introduces a probabilistic sparse attention mechanism, which calculates by randomly sampling the features of some time points, thereby reducing the computational amount while retaining important time dependencies.

[0048] Feature fusion: Fuse the time-frequency features and temporal features in the input dataset as the input of the model. This can make full use of the information in the time-frequency diagram and improve the model's perception ability of the state changes of large-scale power generation equipment.

[0049] In this embodiment, it also includes denoising the vibration signal, and the denoising process includes: performing mean removal processing on the vibration signal data of the current device; performing Hilbert transform on the signal after mean removal processing to obtain the signal envelope value.

[0050] Furthermore, the life prediction digital twin model uses a multi-modal deep learning framework to perform fusion analysis on historical operation data and real-time operation data. Based on the results of the fusion analysis, a hybrid life prediction digital twin model including a physical model and a data-driven model is constructed.

[0051] Furthermore, determine the correlation parameters between the usage status and maintenance, specifically including: extracting the historical maintenance data feature parameters and historical usage data feature parameters from the historical maintenance data - historical usage data correlation curve, and using the Pearson correlation coefficient to perform correlation analysis on the historical maintenance data feature parameters and historical usage data feature parameters to obtain the correlation parameters. Specifically, it includes the following steps:

[0052] Data cleaning: Clean the historical maintenance data and historical usage data to remove noise and outliers.

[0053] Feature extraction: Extract key features from historical usage data, such as operating time, load, temperature, etc.; extract maintenance time, maintenance type, maintenance duration, etc. from historical maintenance data.

[0054] Correlation analysis method: The Pearson correlation coefficient is used to measure the strength and direction of the linear relationship between two variables. Its value range is [-1, 1], where 1 represents a perfect positive correlation, -1 represents a perfect negative correlation, and 0 represents no correlation. Distance correlation is used to measure the non-linear relationship between two variables. It evaluates the correlation by calculating the distance matrix between variables. The maximum information coefficient (MIC) is used to detect the non-linear relationship between variables and can capture complex patterns.

[0055] Deep learning model: Using a deep learning model (such as a neural network) to perform feature embedding on historical maintenance data and historical usage data, and map the data to a low-dimensional space.

[0056] Correlation curve generation: By training a deep learning model, generate a correlation curve between historical maintenance data and historical usage data. The model can learn complex patterns and non-linear relationships in the data.

[0057] Determination of correlation parameters: Model output analysis: According to the correlation curve generated by the deep learning model, analyze the strength of the correlation between different usage states and maintenance.

[0058] Parameter extraction: Extract key parameters from the correlation curve, such as correlation peaks, correlation change rates, etc. These parameters can be used as quantitative indicators for the correlation between usage states and maintenance.

[0059] And Figure 1 Corresponding to the method shown, the present invention also discloses a remaining life prediction system for large-scale power generation equipment for Figure 1 Implementation of the method, the specific structure is as Figure 2 Shown, including the following steps:

[0060] Life-related data set acquisition module: Used to obtain life-related data of large-scale power generation equipment, including design and manufacturing data, historical maintenance data, and historical usage status data, preprocess the life-related data, and construct a life-related data set for large-scale power generation equipment;

[0061] Correlation parameter generation module: Used to construct an initial life digital twin model for large-scale power generation equipment based on design and manufacturing data, use a deep learning model to perform deep learning on historical maintenance data and historical usage data, generate a historical maintenance data - historical usage data correlation curve, and determine the correlation parameters between usage states and maintenance;

[0062] Remaining life prediction module: Used to input the correlation parameters into the initial life digital twin model to generate a life prediction digital twin model, and input the current life-related parameters of the large-scale power generation equipment to be predicted into the life prediction digital twin model to generate the remaining life of the large-scale power generation equipment to be predicted;

[0063] Remaining life correction module: Used to determine the current operating condition of the equipment based on the current operating vibration data of the large-scale power generation equipment to be predicted, judge whether the remaining life prediction of the large-scale power generation equipment to be predicted is accurate, and if not, correct the loss function of the life prediction digital twin model and predict the remaining life again.

[0064] This embodiment also discloses a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods for predicting the remaining life of a large-scale power generation device are implemented.

[0065] In this specification, the various embodiments are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method part.

[0066] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the remaining life of a large-scale power generation equipment, characterized in that, It includes the following steps: Obtain the life-related data of large-scale power generation equipment, including design and manufacturing data, historical maintenance data, and historical usage status data, preprocess the life-related data, and construct a life-related data set for large-scale power generation equipment; Based on the design and manufacturing data, construct an initial life digital twin model for large-scale power generation equipment, use a deep learning model to perform deep learning on historical maintenance data and historical usage data, generate a historical maintenance data - historical usage data correlation curve, and determine the correlation parameters between the usage status and maintenance; Input the correlation parameters into the initial life digital twin model to generate a life prediction digital twin model, input the current life-related parameters of the large-scale power generation equipment to be predicted into the life prediction digital twin model, and generate the remaining life of the large-scale power generation equipment to be predicted; Based on the current operating vibration data of the large-scale power generation equipment to be predicted, determine the current operating condition of the equipment, and judge whether the prediction of the remaining life of the large-scale power generation equipment to be predicted is accurate. If it is not accurate, correct the loss function of the life prediction digital twin model and predict the remaining life again.

2. A method for predicting the remaining life of a large-scale power generation device according to claim 1, characterized in that, Install a vibration signal acquisition device RFID on the large-scale power generation equipment to be predicted, collect the vibration signal of the large-scale power generation equipment to be predicted through the vibration signal acquisition device RFID, and transmit it to the control terminal. The control terminal preprocesses the vibration signal to determine the current operating condition of the equipment.

3. A method for predicting the remaining life of a large-scale power generation device according to claim 2, characterized in that, Determine the current operating condition of the equipment, which specifically includes the following steps: Determine the current life stage of the equipment according to the vibration signal, and at the same time perform time-frequency analysis on the vibration signal to generate a time-frequency diagram; Combine the distinguished life stage with the characteristic information in the time-frequency diagram to form an input data set; Based on a time series prediction model based on the Transformer architecture, introduce an attention mechanism to generate a life prediction model based on the vibration signal, and input the input data set into the life prediction model to obtain the life condition of the current equipment.

4. A method for predicting the remaining life of a large-scale power generation device according to claim 3, characterized in that, It also includes denoising the vibration signal. The denoising process includes: performing mean removal processing on the vibration signal data of the current equipment; performing Hilbert transform on the signal after mean removal processing to obtain the signal envelope value.

5. A method for predicting the remaining life of a large-scale power generation device according to claim 1, characterized in that The life prediction digital twin model uses a multi-modal deep learning framework to perform fusion analysis on historical operation data and real-time operation data. Based on the fusion analysis results, construct a hybrid life prediction digital twin model including a physical model and a data-driven model.

6. A method for predicting the remaining life of a large-scale power generation device according to claim 1, characterized in that, Determine the correlation parameters between the usage status and maintenance, which specifically includes: extract the historical maintenance data characteristic parameters and historical usage data characteristic parameters from the historical maintenance data - historical usage data correlation curve, and use the Pearson correlation coefficient to perform correlation analysis on the historical maintenance data characteristic parameters and historical usage data characteristic parameters to obtain the correlation parameters.

7. A remaining life prediction system for large-scale power generation equipment, characterized in that, It includes the following steps: Life-related data set acquisition module: used to obtain the life-related data of large-scale power generation equipment, including design and manufacturing data, historical maintenance data, and historical usage status data, preprocess the life-related data, and construct a life-related data set for large-scale power generation equipment; Relevance parameter generation module: It is used to construct an initial life digital twin model of large-scale power generation equipment based on design and manufacturing data, and use a deep learning model to perform deep learning on historical maintenance data and historical usage data to generate a historical maintenance data-historical usage data correlation curve, and determine the relevance parameters between the usage status and maintenance. Remaining life prediction module: It is used to input the relevance parameters into the initial life digital twin model to generate a life prediction digital twin model, and input the current life-related parameters of the large-scale power generation equipment to be predicted into the life prediction digital twin model to generate the remaining life of the large-scale power generation equipment to be predicted. Remaining life correction module: It is used to determine the current operation condition of the equipment based on the current operation vibration data of the large-scale power generation equipment to be predicted, and judge whether the remaining life prediction of the large-scale power generation equipment to be predicted is accurate according to the operation condition. If it is inaccurate, correct the loss function of the life prediction digital twin model and predict the remaining life again.

8. A computer storage medium, characterized in that, A computer program is stored on the computer storage medium, and when the computer program is executed by a processor, it implements the steps of a method for predicting the remaining life of a large-scale power generation equipment according to any one of claims 1-6.

Images