Method and system for evaluating health degree of power generation equipment based on data mining

Abstract

The invention provides a health degree assessment method and system for power generation equipment based on data mining, and relates to the technical field of health degree assessment, and the method comprises the steps: obtaining key performance index data of the power generation equipment in a normal operation state, so as to obtain the internal characteristics of the power generation equipment; performing clustering processing on the internal characteristics by using a Gaussian mixture model to obtain a first model parameter describing health state distribution of the power generation equipment under different working conditions; performing function principal component analysis on the first model parameters of the power generation equipment in different degradation stages, and constructing a multi-stage degradation model; dynamically updating parameters of the multi-stage degradation model by calculating posterior probability distribution to obtain second model parameters; and performing calculation processing on the second model parameter through a mahalanobis distance method to obtain a health degradation index capable of evaluating the health condition of the power generation equipment in real time and reflecting the degradation degree of the power generation equipment. The operation efficiency and safety of the power generation equipment are improved, the risk of accidental shutdown is reduced, and meanwhile the maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of health assessment, and in particular to a health assessment method and system for power generation equipment based on data mining. Background Art

[0002] In the field of power generation equipment health assessment, traditional technologies mainly rely on regular inspections and maintenance. This method often lacks accurate grasp of the real-time status of the equipment, making it difficult to detect and prevent potential failures in a timely manner. With the development of big data and machine learning technologies, it is now possible to predict the health status and remaining service life of equipment by analyzing its operating data. However, existing technologies often require a lot of manual intervention and feature engineering when processing large amounts of complex data, which is not only inefficient but also difficult to adapt to rapidly changing operating environments.

[0003] In traditional thermal power plants, the operating conditions of key equipment such as boilers, turbines and generators are directly related to the overall power generation efficiency and safety. During long-term operation, the performance of these equipment will gradually degrade due to factors such as wear, corrosion, and load changes, resulting in an increased probability of failure. Traditional periodic maintenance strategies often fail to detect equipment health problems in a timely manner, while condition-based maintenance strategies require real-time and accurate assessment of the health status of equipment. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for evaluating the health of power generation equipment based on data mining to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] In a first aspect, the present application provides a method for evaluating the health of power generation equipment based on data mining, comprising:

[0006] Acquire key performance indicator data of the power generation equipment under normal operating conditions, and perform unsupervised learning feature extraction through a deep learning autoencoder to obtain intrinsic features of the power generation equipment, wherein the intrinsic features are deep representations of the key performance indicators of the power generation equipment;

[0007] Clustering the intrinsic features using a Gaussian mixture model to obtain first model parameters describing the distribution of health states of the power generation equipment under different working conditions, wherein the Gaussian mixture model can capture and describe the statistical features of the power generation equipment under various health states, including feature means and covariances;

[0008] Performing a functional principal component analysis on the first model parameters of the power generation equipment at different degradation stages to extract the main degradation features and construct a multi-stage degradation model, wherein the multi-stage degradation model is used to describe the evolution process of the power generation equipment from a healthy state to a fault state;

[0009] Adopting a Bayesian reasoning framework, combining prior knowledge and real-time monitoring data, dynamically updating the parameters of the multi-stage degradation model by calculating the posterior probability distribution, so as to obtain a second model parameter that can reflect the degradation state of the power generation equipment in real time;

[0010] The parameters of the second model are calculated and processed by the Mahalanobis distance method to obtain a health decline index that can evaluate the health status of the power generation equipment in real time and reflect its degradation degree.

[0011] Preferably, the key performance indicator data of the power generation equipment under normal operating conditions is obtained, and unsupervised learning feature extraction is performed through a deep learning autoencoder to obtain the intrinsic features of the power generation equipment, including:

[0012] The original performance index data of the power generation equipment under normal operation is obtained, and the denoised data sequence is obtained through wavelet transform processing, wherein the wavelet transform is used to remove high-frequency noise in the data and retain low-frequency trend information related to the health status of the power generation equipment;

[0013] The denoised data is sequence trained through the encoding layer of the autoencoder to obtain a compressed data representation, wherein the encoding layer of the autoencoder maps high-dimensional data to a low-dimensional space through nonlinear transformation to extract key features of the health status of the power generation equipment;

[0014] According to the output of the encoding layer of the autoencoder, the reconstructed data of the operating status of the power generation equipment is obtained through nonlinear mapping of the decoding layer, wherein the decoding layer reconstructs the original data from the compressed representation;

[0015] According to the loss function of the autoencoder, the weights and biases of the autoencoder are iteratively adjusted through an optimization algorithm to obtain the intrinsic characteristics of the final power generation equipment. The optimization algorithm is stochastic gradient descent, which is used to minimize the reconstruction error and the loss of feature representation, while reflecting the health status of the equipment.

[0016] Preferably, the Gaussian mixture model is used to cluster the intrinsic features to obtain first model parameters describing the health status distribution of the power generation equipment under different working conditions, including:

[0017] According to the intrinsic feature representation, after initialization of the Gaussian mixture model, including randomly selected mean, variance and mixing coefficient, an initial parameter estimate of the Gaussian mixture model is obtained;

[0018] The initial parameter estimation of the Gaussian mixture model is iteratively optimized using the expectation maximization algorithm, including the expectation step to calculate the response variable and the maximization step to update the mean, variance and mixing coefficient, to obtain the parameters of the optimized Gaussian mixture model, where the parameters represent the health status distribution of the power generation equipment under different working conditions;

[0019] Based on the optimized Gaussian mixture model parameters, after model verification and selection, the Akaike information criterion was used to evaluate the model fit goodness, and the final first model parameters were obtained to describe the health status distribution of power generation equipment under different operating conditions.

[0020] Preferably, the Bayesian reasoning framework is used, combined with prior knowledge and real-time monitoring data, to dynamically update the parameters of the multi-stage degradation model by calculating the posterior probability distribution to obtain the second model parameters that can reflect the degradation state of the power generation equipment in real time, including:

[0021] According to the historical health data of power generation equipment and expert experience, the initial multi-stage degradation model parameters are obtained through the prior probability distribution setting of Bayesian reasoning. The multi-stage degradation model parameters include the mean, variance and mixing coefficient of each health state of the equipment.

[0022] The performance indicator data obtained from real-time monitoring of the equipment is processed through data preprocessing and normalization to obtain a real-time data set for model updating, wherein the real-time data set reflects the current operating status of the equipment;

[0023] Based on the performance indicator data and the probability update of the multi-stage degradation model parameters under the Bayesian reasoning framework, the posterior probability of each data point in each health state is obtained, that is, the probability of the device being in a healthy state when the current data is observed;

[0024] Based on the posterior probability distribution, the updated multi-stage degradation model parameters, namely the second model parameters, are obtained through the expectation maximization algorithm or the variational inference iterative optimization method.

[0025] In a second aspect, the present application also provides a health assessment system for power generation equipment based on data mining, comprising:

[0026] Acquisition module: used to acquire key performance indicator data of power generation equipment under normal operating conditions, and perform unsupervised learning feature extraction through deep learning autoencoder to obtain intrinsic features of the power generation equipment, wherein the intrinsic features are deep-level representations of key performance indicators of the power generation equipment;

[0027] A processing module: used for clustering the intrinsic features using a Gaussian mixture model to obtain a first model parameter describing the distribution of health states of the power generation equipment under different working conditions, wherein the Gaussian mixture model can capture and describe the statistical features of the power generation equipment under various health states, including feature means and covariances;

[0028] Construction module: used for performing functional principal component analysis on the first model parameters of the power generation equipment at different degradation stages to extract the main degradation characteristics and construct a multi-stage degradation model, wherein the multi-stage degradation model is used to describe the evolution process of the power generation equipment from a healthy state to a fault state;

[0029] Analysis module: used to adopt a Bayesian reasoning framework, combine prior knowledge and real-time monitoring data, and dynamically update the parameters of the multi-stage degradation model by calculating the posterior probability distribution to obtain a second model parameter that can reflect the degradation state of the power generation equipment in real time;

[0030] Calculation module: used to calculate and process the second model parameters by using the Mahalanobis distance method to obtain a health decline index that can evaluate the health status of the power generation equipment in real time and reflect its degradation degree.

[0031] In a third aspect, the present application also provides a health assessment device for power generation equipment based on data mining, comprising:

[0032] Memory for storing computer programs;

[0033] A processor is used to implement the steps of the health assessment method of power generation equipment based on data mining when executing the computer program.

[0034] In a fourth aspect, the present application further provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned method for assessing the health of power generation equipment based on data mining are implemented.

[0035] The beneficial effects of the present invention are:

[0036] The present invention combines deep learning autoencoders for feature extraction, and the present invention can capture the deep-level features of key performance indicators of power generation equipment, so as to more accurately evaluate the health status of the equipment; the Bayesian reasoning framework and real-time monitoring data can be used to dynamically update the multi-stage degradation model parameters of the power generation equipment, so as to realize real-time monitoring of the health status of the equipment and accurate prediction of future degradation trends; the health decline index calculated based on the Mahalanobis distance can provide scientific maintenance decision support for the power generation equipment, and help the maintenance team to take preventive maintenance measures in time to reduce unexpected downtime; through predictive maintenance, unnecessary excessive maintenance can be avoided, and the emergency repair costs caused by equipment failure can be reduced, so as to achieve optimal allocation of maintenance resources; real-time health assessment and accurate fault prediction help to discover potential fault risks in advance, improve the reliability and operation safety of power generation equipment, and ensure the stability of power supply.

[0037] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 It is a schematic flow chart of a method for evaluating the health of power generation equipment based on data mining according to an embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the structure of a health assessment system for power generation equipment based on data mining according to an embodiment of the present invention;

[0041] Figure 3 It is a schematic diagram of the structure of a health assessment device for power generation equipment based on data mining described in an embodiment of the present invention.

[0042] In the figure: 701, acquisition module; 702, processing module; 703, construction module; 704, analysis module; 705, calculation module; 800, health assessment equipment of power generation equipment based on data mining; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0045] Embodiment 1:

[0046] This embodiment provides a method for evaluating the health of power generation equipment based on data mining.

[0047] See also Figure 1 , the figure shows that the method includes step S100, step S200, step S300, step S400 and step S500.

[0048] S100, obtaining key performance indicator data of the power generation equipment under normal operating conditions, and performing unsupervised learning feature extraction through a deep learning autoencoder to obtain intrinsic characteristics of the power generation equipment, wherein the intrinsic characteristics are deep-level representations of the key performance indicators of the power generation equipment.

[0049] It can be understood that the step S100 includes S101, S102, S103 and S104, wherein:

[0050] S101, obtaining original performance indicator data of the power generation equipment in normal operation, and processing it by wavelet transform to obtain a denoised data sequence, wherein the wavelet transform is used to remove high-frequency noise in the data and retain low-frequency trend information related to the health status of the power generation equipment;

[0051] S102, performing sequence training on the denoised data through an encoding layer of an autoencoder to obtain a compressed data representation, wherein the encoding layer of the autoencoder maps high-dimensional data to a low-dimensional space through a nonlinear transformation to extract key features of the health status of the power generation equipment;

[0052] S103, according to the encoding layer output of the autoencoder, through the nonlinear mapping of the decoding layer, to obtain the reconstructed data of the operating state of the power generation equipment, wherein the decoding layer reconstructs the original data from the compressed representation;

[0053] S104. According to the loss function of the autoencoder, the weight and bias of the autoencoder are iteratively adjusted through an optimization algorithm to obtain the final intrinsic characteristics of the power generation equipment, wherein the optimization algorithm is stochastic gradient descent, which is used to minimize the reconstruction error and the loss of feature representation, and at the same time reflect the health status of the equipment.

[0054] It should be noted that in the above decomposition, wavelet transform is a mathematical method used to process time series data, which can separate different components of the signal in the time-frequency domain, which is particularly important in the vibration analysis of power generation equipment. The autoencoder is a deep learning algorithm, and the structure of its encoding layer and decoding layer is designed to learn the effective representation of data, which is creative when processing complex data of power generation equipment. The parameters of the autoencoder are adjusted by the optimization algorithm to ensure the accuracy and reliability of the feature extraction process, which is crucial for the health assessment of power generation equipment.

[0055] S200. Clustering the intrinsic features using a Gaussian mixture model to obtain first model parameters describing the health status distribution of the power generation equipment under different operating conditions, wherein the Gaussian mixture model can capture and describe the statistical characteristics of the power generation equipment under various health states, including characteristic means and covariances.

[0056] It can be understood that step S200 includes S201, S202 and S203, wherein:

[0057] S201, according to the intrinsic feature representation, after initialization of the Gaussian mixture model, including randomly selected mean, variance and mixing coefficient, obtaining initial parameter estimation of the Gaussian mixture model;

[0058] S202, iteratively optimizing the initial parameter estimation of the Gaussian mixture model through an expectation maximization algorithm, including calculating the response variable in an expectation step and updating the mean, variance and mixing coefficient in a maximization step, to obtain the parameters of the optimized Gaussian mixture model, wherein the parameters represent the health status distribution of the power generation equipment under different working conditions;

[0059] S203. Based on the optimized Gaussian mixture model parameters, after model verification and selection, the Akaike information criterion is used to evaluate the model fit goodness to obtain the final first model parameters for describing the health status distribution of the power generation equipment under different working conditions.

[0060] It should be noted that the initial mean estimates are obtained based on the key performance indicator data of the power generation equipment through random selection or K-means++ algorithm initialization. These means, as the center of the Gaussian distribution, are the key to model initialization. Random selection is usually selected from the data set, while the K-means++ algorithm selects points that can represent the diversity of the data as the initial mean to improve the quality and speed of clustering; the data is grouped according to the initial mean, and the initial variance estimate is obtained by calculating the sum of the squares of the distances between the data points in each group and the group center. The variance estimate reflects the shape and diffusion of each Gaussian distribution, which is crucial for capturing the variability of the data; the mixing coefficient is obtained after normalization based on the probability that each data point belongs to each Gaussian distribution. These mixing coefficients represent the proportion of each Gaussian distribution in the overall model and are an important part of the model parameters; through the expectation maximization (EM) algorithm, the posterior probability of each data point belonging to each Gaussian distribution is calculated through the E step, and then the mean, variance and mixing coefficient are updated in the M step. This iterative process continuously optimizes the model parameters until convergence, that is, the change in the model parameters is less than the set threshold, to ensure that the model can accurately reflect the health status distribution of the power generation equipment under different working conditions. Through these four steps, GMM can gradually adjust its parameters to best fit the health status data of the power generation equipment, thereby providing accurate model parameters for the health assessment of the power generation equipment.

[0061] It can be understood that the E-step and the M-step are two key components of the expectation-maximization algorithm during the training process of the Gaussian mixture model (GMM). The EM algorithm is an iterative algorithm used to find the maximum likelihood estimates of the parameters of a probability model in the presence of hidden variables (i.e. data that cannot be directly observed). The following is a detailed explanation of the E-step and the M-step:

[0062] Among them, the E step is the expectation step. In the E step, the algorithm calculates the posterior probability that each data point belongs to each Gaussian distribution component. These probabilities are also called "responsibility". Specifically, the E step includes the following: according to the current model parameters (mean, covariance matrix and mixing coefficient), the probability of each data point belonging to each Gaussian distribution component is calculated; these probabilities reflect the "responsibility" or contribution of each component to the generation of the data point under the current model parameters.

[0063] Therefore, in the M step, the algorithm uses the posterior probability calculated in the E step to update the model parameters to maximize the log-likelihood function of the data. Specifically, the maximization step includes the following: updating the mean μk, covariance matrix Σk, and mixing coefficient πk of each Gaussian distribution component. The update of these parameters is achieved by maximizing the expected log-likelihood function, which is calculated based on the posterior probability obtained in the expectation step. The expectation step and the maximization step are performed alternately until the model parameters converge, that is, the model's fit to the data is no longer significantly improved, or the predetermined number of iterations is reached. Through this iterative process, the Gaussian mixture model is able to gradually adjust its parameters to better fit the data distribution.

[0064] S300, performing a principal component analysis of the first model parameters of the power generation equipment at different degradation stages to extract main degradation features, and constructing a multi-stage degradation model, wherein the multi-stage degradation model is used to describe the evolution process of the power generation equipment from a healthy state to a fault state.

[0065] It can be understood that step S300 includes S301, S302 and S303, wherein:

[0066] S301. Data preprocessing and basis function selection: Based on the operating data recorded by the power generation equipment under different working conditions, after appropriate preprocessing such as denoising and normalization, the data set used for FPCA is obtained; appropriate basis functions are selected to expand these data. These basis functions can be Fourier series, B-splines or other functions suitable for describing the operating characteristics of the equipment. The selection of basis functions is crucial for subsequent analysis because they will be used to construct the coordinate system in the function space.

[0067] S302, constructing covariance function and solving eigenvalue problem: Based on the preprocessed data and the selected basis function, construct the covariance function, which describes the intrinsic correlation of the data in time or usage cycle, and find the main degradation characteristics by solving the eigenvalue problem. This process involves solving an integral equation that contains the inner product of the covariance function and the basis function. The eigenvalues ​​and eigenfunctions (principal components) obtained will be used to describe the main change patterns in the equipment degradation process.

[0068] S303. Construction and parameter estimation of multi-stage degradation models: Use the characteristic functions obtained by FPCA to construct multi-stage degradation models, where these models represent the health status of the equipment as a function of time and can capture each stage of the equipment from health to failure; estimate model parameters by minimizing the difference between actual observed data and model predictions. This usually involves optimization algorithms such as gradient descent or Newton's method to find the best fit parameters. It combines the characteristic functions obtained by FPCA and the actual degradation data of the equipment to construct a multi-stage degradation model. In the model, the degradation process of the equipment is represented as a linear combination of the mean function and several main characteristic functions, that is, the calculation formula is as follows:

[0069]

[0070] Where μ(t) is the mean function of the degradation process, φk(t) is the kth feature function, and ∈ik is the score of device i on the kth feature function.

[0071] Using the Bayesian reasoning framework, combined with prior knowledge and real-time monitoring data, the parameters of the multi-stage degradation model are dynamically updated by calculating the posterior probability distribution, and the model parameters that can reflect the degradation state of the power generation equipment in real time are obtained.

[0072] S400, adopting a Bayesian reasoning framework, combining prior knowledge and real-time monitoring data, and dynamically updating the parameters of the multi-stage degradation model by calculating the posterior probability distribution to obtain a second model parameter that can reflect the degradation state of the power generation equipment in real time.

[0073] It can be understood that the step S400 includes S401, S402, S403 and S404, wherein:

[0074] S401. According to the historical health data of the power generation equipment and the expert experience, the initial multi-stage degradation model parameters are obtained through the prior probability distribution setting of Bayesian reasoning, wherein the multi-stage degradation model parameters include the mean, variance and mixing coefficient of each health state of the equipment;

[0075] S402, processing the performance indicator data obtained by real-time monitoring of the equipment through data preprocessing and normalization to obtain a real-time data set for model updating, wherein the real-time data set reflects the current operating status of the equipment;

[0076] S403, based on the performance indicator data and the multi-stage degradation model parameters, the probability of updating under the Bayesian reasoning framework is used to obtain the posterior probability of each data point in each health state, that is, the probability of the device being in a healthy state when the current data is observed;

[0077] S404. Based on the posterior probability distribution, the updated multi-stage degradation model parameters, i.e., the second model parameters, are obtained by processing with an expectation maximization algorithm or a variational inference iterative optimization method.

[0078] It should be noted that in these four steps, the Bayesian reasoning framework allows us to combine prior knowledge and real-time data to dynamically update model parameters by calculating the posterior probability distribution. This method is particularly important in the health assessment of power generation equipment because it can provide a real-time assessment of the current health status of the equipment and predict its future performance changes.

[0079] It can be understood that in step S403, when the current data point is observed, the posterior probability P(S|D) that the device is in a certain health state S can be calculated by the Bayesian formula:

[0080]

[0081] Where P(S) is the prior probability of the health state S, P(D|S) is the probability of observing data D in the health state S, also known as the likelihood function, and P(D) is the marginal probability of observing data DD.

[0082] In step S404, the posterior probability of the latent variable (e.g., the health state of the device) is calculated, which usually involves calculating the soft assignment of each data point to each health state. For each data point i and health state, the calculation formula is as follows:

[0083]

[0084] Where P(Sk) is the prior probability of the healthy state Sk, P(D|Sk) is the likelihood probability of observing data D in the healthy state Sk, and P(D) is the marginal probability of data D.

[0085] Using the posterior probability obtained in step E, the parameters of the multi-stage degradation model are updated. The update formula is as follows:

[0086]

[0087] Where Q is an auxiliary function used to construct an optimization problem about θ by expecting the posterior probability γ under the current parameter θold.

[0088] S500: Calculate and process the second model parameters by using the Mahalanobis distance method to obtain a health decline index that can evaluate the health status of the power generation equipment in real time and reflect the degree of its degradation.

[0089] It can be understood that in this step, for each data point monitored in real time, the Mahalanobis distance between it and the healthy benchmark model is calculated, wherein the calculation formula of the Mahalanobis distance is as follows:

[0090]

[0091] Among them, x is the data point monitored in real time, μ is the mean vector in the healthy state, Σ is the covariance matrix, ∑ -1 is the inverse of the covariance matrix.

[0092] Among them, the Mahalanobis distance provides a quantitative indicator for evaluating the difference between the current state of the device and the healthy state. The larger the distance, the further the device deviates from the healthy state and the higher the degree of degradation. A threshold can be set. When the Mahalanobis distance exceeds this threshold, it indicates that the device may need maintenance or inspection.

[0093] Therefore, as new monitoring data is continuously obtained, the above calculation process is repeated and the Mahalanobis distance is updated in real time, thereby realizing continuous monitoring of the health status of the equipment. Through such a process, the health status of the power generation equipment can be effectively evaluated and potential degradation problems can be identified in time, thereby improving the reliability and safety of the equipment and reducing the risk of unexpected downtime. Among them, the introduction of the Mahalanobis distance method makes the health assessment process more scientific and accurate, and provides a powerful tool for predictive maintenance of power generation equipment.

[0094] Embodiment 2:

[0095] like Figure 2 As shown, this embodiment provides a health assessment system for power generation equipment based on data mining, see Figure 2 The system comprises:

[0096] Acquisition module 701: used to acquire key performance indicator data of the power generation equipment under normal operating conditions, and perform unsupervised learning feature extraction through a deep learning autoencoder to obtain intrinsic features of the power generation equipment, wherein the intrinsic features are deep-level representations of the key performance indicators of the power generation equipment;

[0097] Processing module 702: for clustering the intrinsic features using a Gaussian mixture model to obtain first model parameters describing the distribution of health states of the power generation equipment under different working conditions, wherein the Gaussian mixture model can capture and describe the statistical features of the power generation equipment under various health states, including feature means and covariances;

[0098] Construction module 703: used to perform functional principal component analysis on the first model parameters of the power generation equipment at different degradation stages to extract the main degradation characteristics and construct a multi-stage degradation model, wherein the multi-stage degradation model is used to describe the evolution process of the power generation equipment from a healthy state to a fault state;

[0099] Analysis module 704: used to adopt a Bayesian reasoning framework, combine prior knowledge and real-time monitoring data, and dynamically update the parameters of the multi-stage degradation model by calculating the posterior probability distribution to obtain a second model parameter that can reflect the degradation state of the power generation equipment in real time;

[0100] Calculation module 705: used to calculate and process the second model parameters by using the Mahalanobis distance method to obtain a health decline index that can evaluate the health status of the power generation equipment in real time and reflect the degree of its degradation.

[0101] Specifically, the acquisition module 701 includes:

[0102] The first processing unit is used to obtain the original performance index data of the power generation equipment under normal operating conditions, and obtain a denoised data sequence through wavelet transform processing, wherein the wavelet transform is used to remove high-frequency noise in the data and retain low-frequency trend information related to the health status of the power generation equipment;

[0103] Training unit: used to perform sequence training on the denoised data through the encoding layer of the autoencoder to obtain a compressed data representation, wherein the encoding layer of the autoencoder maps high-dimensional data to a low-dimensional space through nonlinear transformation to extract key features of the health status of the power generation equipment;

[0104] Mapping unit: used to obtain the reconstructed data of the operating status of the power generation equipment according to the output of the encoding layer of the autoencoder through nonlinear mapping of the decoding layer, wherein the decoding layer reconstructs the original data from the compressed representation;

[0105] Adjustment unit: used to iteratively adjust the weight and bias of the autoencoder according to the loss function of the autoencoder through the optimization algorithm to obtain the intrinsic characteristics of the final power generation equipment, wherein the optimization algorithm is stochastic gradient descent, which is used to minimize the reconstruction error and the loss of feature representation, and at the same time reflect the health status of the equipment.

[0106] Specifically, the processing module 702 includes:

[0107] Initialization unit: used for obtaining initial parameter estimation of the Gaussian mixture model through initialization of the Gaussian mixture model according to the intrinsic feature representation, including randomly selected mean, variance and mixing coefficient;

[0108] Optimization unit: used to iteratively optimize the initial parameter estimation of the Gaussian mixture model through the expectation maximization algorithm, including the expectation step to calculate the response variable and the maximization step to update the mean, variance and mixing coefficient, to obtain the parameters of the optimized Gaussian mixture model, where the parameters represent the health status distribution of the power generation equipment under different working conditions;

[0109] Fitting unit: Based on the optimized Gaussian mixture model parameters, after model verification and selection, the Akaike information criterion is used to evaluate the model fit goodness, and the final first model parameters are obtained to describe the health status distribution of power generation equipment under different working conditions.

[0110] Specifically, the analysis module 704 includes:

[0111] Setting unit: used to obtain initial multi-stage degradation model parameters according to historical health data of power generation equipment and expert experience through prior probability distribution setting of Bayesian reasoning, where the multi-stage degradation model parameters include the mean, variance and mixing coefficient of each health state of the equipment;

[0112] The second processing unit is used to process the performance indicator data obtained by real-time monitoring of the equipment through data preprocessing and normalization to obtain a real-time data set for model updating, wherein the real-time data set reflects the current operating status of the equipment;

[0113] Update unit: used to update the probability of each data point in each health state based on the performance indicator data and the multi-stage degradation model parameters through the Bayesian reasoning framework, that is, the probability of the device being in a healthy state when the current data is observed;

[0114] The third processing unit is used to obtain updated multi-stage degradation model parameters, namely, second model parameters, based on the posterior probability distribution through an expectation maximization algorithm or a variational inference iterative optimization method.

[0115] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0116] Embodiment 3:

[0117] Corresponding to the above method embodiment, this embodiment also provides a health assessment device for power generation equipment based on data mining. The health assessment device for power generation equipment based on data mining described below and the health assessment method for power generation equipment based on data mining described above can refer to each other.

[0118] Figure 3 FIG. 8 is a block diagram of a device 800 for evaluating the health of a power generation device based on data mining according to an exemplary embodiment. Figure 3As shown, the health assessment device 800 for power generation equipment based on data mining includes: a processor 801 and a memory 802. The health assessment device 800 for power generation equipment based on data mining also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0119] The processor 801 is used to control the overall operation of the health assessment device 800 for power generation equipment based on data mining, so as to complete all or part of the steps in the health assessment method for power generation equipment based on data mining. The memory 802 is used to store various types of data to support the operation of the health assessment device 800 for power generation equipment based on data mining, and these data may include, for example, instructions for any application or method operating on the health assessment device 800 for power generation equipment based on data mining, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules may be a keyboard, a mouse or buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the health assessment device 800 of the power generation equipment based on data mining and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 may include: a Wi-Fi module, a Bluetooth module or an NFC module.

[0120] In an exemplary embodiment, the health assessment device 800 for power generation equipment based on data mining can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned health assessment method for power generation equipment based on data mining.

[0121] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned method for assessing the health of a power generation device based on data mining are implemented. For example, the computer-readable storage medium may be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions may be executed by the processor 801 of the device 800 for assessing the health of a power generation device based on data mining to complete the above-mentioned method for assessing the health of a power generation device based on data mining.

[0122] Embodiment 4:

[0123] Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. The readable storage medium described below and the health assessment method of power generation equipment based on data mining described above can refer to each other.

[0124] A computer program is stored on the readable storage medium, and when the computer program is executed by the processor, the steps of the method for assessing the health of power generation equipment based on data mining in the above method embodiment are implemented.

[0125] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0127] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for evaluating the health of power generation equipment based on data mining, characterized in that: include: Acquire key performance indicator data of the power generation equipment under normal operating conditions, and perform unsupervised learning feature extraction through a deep learning autoencoder to obtain intrinsic features of the power generation equipment, wherein the intrinsic features are deep representations of the key performance indicators of the power generation equipment; Clustering the intrinsic features using a Gaussian mixture model to obtain first model parameters describing the distribution of health states of the power generation equipment under different working conditions, wherein the Gaussian mixture model can capture and describe the statistical features of the power generation equipment under various health states, including feature means and covariances; Performing a functional principal component analysis on the first model parameters of the power generation equipment at different degradation stages to extract the main degradation features and construct a multi-stage degradation model, wherein the multi-stage degradation model is used to describe the evolution process of the power generation equipment from a healthy state to a fault state; Adopting a Bayesian reasoning framework, combining prior knowledge and real-time monitoring data, dynamically updating the parameters of the multi-stage degradation model by calculating the posterior probability distribution, so as to obtain a second model parameter that can reflect the degradation state of the power generation equipment in real time; The parameters of the second model are calculated and processed by the Mahalanobis distance method to obtain a health decline index that can evaluate the health status of the power generation equipment in real time and reflect its degradation degree.

2. The method for evaluating the health of power generation equipment based on data mining according to claim 1, characterized in that: The key performance indicator data of the power generation equipment under normal operating conditions is obtained, and unsupervised learning feature extraction is performed through a deep learning autoencoder to obtain the intrinsic features of the power generation equipment, including: The original performance index data of the power generation equipment under normal operation is obtained, and the denoised data sequence is obtained through wavelet transform processing, wherein the wavelet transform is used to remove high-frequency noise in the data and retain low-frequency trend information related to the health status of the power generation equipment; The denoised data is sequenced through the encoding layer of the autoencoder to obtain a compressed data representation, wherein the encoding layer of the autoencoder maps high-dimensional data to a low-dimensional space through nonlinear transformation to extract key features of the health status of the power generation equipment; According to the output of the encoding layer of the autoencoder, the reconstructed data of the operating status of the power generation equipment is obtained through nonlinear mapping of the decoding layer, wherein the decoding layer reconstructs the original data from the compressed representation; According to the loss function of the autoencoder, the weights and biases of the autoencoder are iteratively adjusted through an optimization algorithm to obtain the intrinsic characteristics of the final power generation equipment. The optimization algorithm is stochastic gradient descent, which is used to minimize the reconstruction error and the loss of feature representation, while reflecting the health status of the equipment.

3. The method for evaluating the health of power generation equipment based on data mining according to claim 1, characterized in that: The Gaussian mixture model is used to cluster the intrinsic features to obtain first model parameters describing the health status distribution of the power generation equipment under different working conditions, including: According to the intrinsic feature representation, after initialization of the Gaussian mixture model, including randomly selected mean, variance and mixing coefficient, an initial parameter estimate of the Gaussian mixture model is obtained; The initial parameter estimation of the Gaussian mixture model is iteratively optimized using the expectation maximization algorithm, including the expectation step to calculate the response variable and the maximization step to update the mean, variance and mixing coefficient, to obtain the parameters of the optimized Gaussian mixture model, where the parameters represent the health status distribution of the power generation equipment under different working conditions; Based on the optimized Gaussian mixture model parameters, after model verification and selection, the Akaike information criterion was used to evaluate the model fit goodness, and the final first model parameters were obtained to describe the health status distribution of power generation equipment under different operating conditions.

4. The method for evaluating the health of power generation equipment based on data mining according to claim 1, characterized in that: The Bayesian reasoning framework is used to combine prior knowledge and real-time monitoring data to dynamically update the parameters of the multi-stage degradation model by calculating the posterior probability distribution to obtain the second model parameters that can reflect the degradation state of the power generation equipment in real time, including: According to the historical health data of power generation equipment and expert experience, the initial multi-stage degradation model parameters are obtained through the prior probability distribution setting of Bayesian reasoning. The multi-stage degradation model parameters include the mean, variance and mixing coefficient of each health state of the equipment. The performance indicator data obtained from real-time monitoring of the equipment is processed through data preprocessing and normalization to obtain a real-time data set for model updating, wherein the real-time data set reflects the current operating status of the equipment; Based on the performance indicator data and the probability update of the multi-stage degradation model parameters under the Bayesian reasoning framework, the posterior probability of each data point in each health state is obtained, that is, the probability of the device being in a healthy state when the current data is observed; Based on the posterior probability distribution, the updated multi-stage degradation model parameters, namely the second model parameters, are obtained through the expectation maximization algorithm or the variational inference iterative optimization method.

5. A health assessment system for power generation equipment based on data mining, based on the health assessment method for power generation equipment based on data mining according to claim 1, characterized in that: include: Acquisition module: used to acquire key performance indicator data of power generation equipment under normal operating conditions, and perform unsupervised learning feature extraction through deep learning autoencoders to obtain intrinsic features of the power generation equipment, wherein the intrinsic features are deep representations of the key performance indicators of the power generation equipment; A processing module: used for clustering the intrinsic features using a Gaussian mixture model to obtain a first model parameter describing the distribution of health states of the power generation equipment under different working conditions, wherein the Gaussian mixture model can capture and describe the statistical features of the power generation equipment under various health states, including feature means and covariances; Construction module: used for performing functional principal component analysis on the first model parameters of the power generation equipment at different degradation stages to extract the main degradation characteristics and construct a multi-stage degradation model, wherein the multi-stage degradation model is used to describe the evolution process of the power generation equipment from a healthy state to a fault state; Analysis module: used to adopt a Bayesian reasoning framework, combine prior knowledge and real-time monitoring data, and dynamically update the parameters of the multi-stage degradation model by calculating the posterior probability distribution to obtain a second model parameter that can reflect the degradation state of the power generation equipment in real time; Calculation module: used to calculate and process the second model parameters by using the Mahalanobis distance method to obtain a health decline index that can evaluate the health status of the power generation equipment in real time and reflect its degradation degree.

6. The health assessment system for power generation equipment based on data mining according to claim 5 is characterized in that: The acquisition module includes: The first processing unit is used to obtain the original performance index data of the power generation equipment under normal operating conditions, and obtain a denoised data sequence through wavelet transform processing, wherein the wavelet transform is used to remove high-frequency noise in the data and retain low-frequency trend information related to the health status of the power generation equipment; Training unit: used to perform sequence training on the denoised data through the encoding layer of the autoencoder to obtain a compressed data representation, wherein the encoding layer of the autoencoder maps high-dimensional data to a low-dimensional space through nonlinear transformation to extract key features of the health status of the power generation equipment; Mapping unit: used to obtain the reconstructed data of the operating status of the power generation equipment according to the output of the encoding layer of the autoencoder through nonlinear mapping of the decoding layer, wherein the decoding layer reconstructs the original data from the compressed representation; Adjustment unit: used to iteratively adjust the weight and bias of the autoencoder according to the loss function of the autoencoder through the optimization algorithm to obtain the intrinsic characteristics of the final power generation equipment, wherein the optimization algorithm is stochastic gradient descent, which is used to minimize the reconstruction error and the loss of feature representation, and at the same time reflect the health status of the equipment.

7. The health assessment system for power generation equipment based on data mining according to claim 5, characterized in that: The processing module includes: Initialization unit: used for obtaining initial parameter estimation of the Gaussian mixture model through initialization of the Gaussian mixture model according to the intrinsic feature representation, including randomly selected mean, variance and mixing coefficient; Optimization unit: used to iteratively optimize the initial parameter estimation of the Gaussian mixture model through the expectation maximization algorithm, including the expectation step to calculate the response variable and the maximization step to update the mean, variance and mixing coefficient, to obtain the parameters of the optimized Gaussian mixture model, where the parameters represent the health status distribution of the power generation equipment under different working conditions; Fitting unit: Based on the optimized Gaussian mixture model parameters, after model verification and selection, the Akaike information criterion is used to evaluate the model fit goodness, and the final first model parameters are obtained to describe the health status distribution of power generation equipment under different working conditions.

8. The health assessment system for power generation equipment based on data mining according to claim 5, characterized in that: The analysis module includes: Setting unit: used to obtain initial multi-stage degradation model parameters according to historical health data of power generation equipment and expert experience through prior probability distribution setting of Bayesian reasoning, where the multi-stage degradation model parameters include the mean, variance and mixing coefficient of each health state of the equipment; The second processing unit is used to process the performance indicator data obtained by real-time monitoring of the equipment through data preprocessing and normalization to obtain a real-time data set for model updating, wherein the real-time data set reflects the current operating status of the equipment; Update unit: used to update the probability of each data point in each health state based on the performance indicator data and the multi-stage degradation model parameters through the Bayesian reasoning framework, that is, the probability of the device being in a healthy state when the current data is observed; The third processing unit is used to obtain updated multi-stage degradation model parameters, namely, second model parameters, based on the posterior probability distribution through an expectation maximization algorithm or a variational inference iterative optimization method.

9. A health assessment device for power generation equipment based on data mining, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the health assessment method for power generation equipment based on data mining as described in any one of claims 1 to 4 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for evaluating the health of a power generation equipment based on data mining as claimed in any one of claims 1 to 4 is implemented.

Images