Thermal power generation equipment fault diagnosis system and method based on big data analysis

Abstract

The invention relates to the technical field of thermal power generation equipment monitoring, and provides a thermal power generation equipment fault diagnosis system and method based on big data analysis, and the system comprises a data collection module, a data processing module, an analysis and diagnosis module, and an early warning module. According to the system, the operation state monitoring of the thermal power generation main equipment and the auxiliary equipment is realized through the data acquisition module, and the thermal power plant SCADA unit, the turbine TSI unit, the auxiliary equipment vibration monitoring unit and the electrical monitoring unit work cooperatively, so that the acquisition of various key indexes of the thermal power generation equipment is realized, and basic support is provided for big data analysis. The analysis and diagnosis module analyzes and pre-warns equipment faults, realizes automatic classification of fault types through a support vector machine, realizes early warning through an isolated forest algorithm, captures a dependency relationship of equipment performance degradation through a long and short-term memory network, and realizes a complete process from fault identification to trend prediction.

Description

Technical Field

[0001] This invention relates to a fault diagnosis system, and more particularly to a fault diagnosis system and method for thermal power generation equipment based on big data analysis, belonging to the field of thermal power generation equipment monitoring technology. Background Technology

[0002] Thermal power generation systems are complex in structure and contain numerous devices. When critical power generation equipment malfunctions and is not addressed promptly, it can easily lead to unexpected shutdowns, affecting the stability of the power supply. Traditional fault diagnosis methods mostly rely on threshold alarms or manual experience, which are insufficient for quickly handling the complex situations encountered during the joint operation of thermal power generation equipment.

[0003] With the continuous upgrading of thermal power generation systems, artificial intelligence technology is being gradually applied to fault diagnosis of thermal power production equipment. For example, existing patent CN120145261A discloses a method and system for diagnosing fault causes in thermal power production equipment. This method includes acquiring abnormal results of time-series data from measuring points in a thermal power plant; inputting these abnormal results into a trained fault diagnosis language model; analyzing the abnormalities in the input measuring point data based on the model; and analyzing and inferring possible fault types, causes, handling suggestions, and subsequent maintenance suggestions based on the meaning and type of the measuring points, thereby generating a diagnostic report. This method assists users in providing decision-making suggestions, improving user-friendliness and the interpretability of specialized tasks.

[0004] However, most existing technologies focus on diagnosing faults that have already occurred, lacking in real-time performance and early warning capabilities. The time lag between the occurrence of a fault and its diagnosis leads to delayed handling and potentially more severe economic losses. Furthermore, faults in thermal power generation equipment are typically gradual processes; in the early stages, equipment parameters change only slightly, making it difficult for existing diagnostic technologies to accurately analyze fault risks, further increasing maintenance costs.

[0005] To address this, a fault diagnosis system and method for thermal power generation equipment based on big data analysis is proposed. Summary of the Invention

[0006] In view of this, the present invention provides a fault diagnosis system and method for thermal power generation equipment based on big data analysis, so as to solve or alleviate one of the technical problems existing in the prior art, and at least provide a beneficial option.

[0007] The technical solution of this invention is implemented as follows: A fault diagnosis system for thermal power generation equipment based on big data analysis, comprising:

[0008] The data acquisition module is used to collect operating data from the main and auxiliary equipment of thermal power generation.

[0009] The data processing module, connected to the data acquisition module, is used to perform timestamp alignment, cleaning, standardization, and feature extraction on data from different sampling frequencies to form a unified panoramic view of the equipment operation.

[0010] The analysis and diagnosis module is connected to the data processing module and is used to perform fault classification, anomaly detection, and degradation trend prediction on the processed data based on a machine learning model.

[0011] The early warning module is connected to the analysis and diagnosis module and is used to visualize the health status of the equipment, push alarm information, and generate diagnostic reports.

[0012] A further preferred embodiment: the data acquisition module includes:

[0013] The SCADA unit in a thermal power plant is used to collect temperature, pressure, flow rate, and load process parameters of the boiler, steam turbine, and generator.

[0014] The turbine TSI unit is used to collect parameters such as turbine shaft vibration, shaft displacement, differential expansion, speed, eccentricity, and key phase mechanical quantities.

[0015] Auxiliary machine vibration monitoring unit is used to collect high-frequency vibration data of fans, water pumps, and coal mill auxiliary equipment;

[0016] The electrical monitoring unit is used to collect electrical parameters such as voltage, current, and power factor.

[0017] A further preferred embodiment: the data processing module includes a processing unit and a storage unit;

[0018] The processing unit is used to align data from different sources and frequencies on timestamps, and to clean, standardize, and extract features from the data to form a unified panoramic view of the equipment's operation.

[0019] The storage unit includes a database for storing real-time data and processed data, and a distributed big data platform for storing historical data.

[0020] A further preferred embodiment: the analysis and diagnosis module includes a fault classification unit, an anomaly detection unit, and a degradation trend prediction unit.

[0021] A further preferred embodiment: The specific working steps of the analysis and diagnosis module include:

[0022] The machine learning models in the analysis and diagnosis module are trained using historical data;

[0023] During equipment monitoring, the anomaly detection unit uses the isolated forest algorithm to calculate the input real-time equipment feature data and generate an anomaly score for each data point.

[0024] Real-time monitoring of changes in anomaly scores; when the anomaly score of the device data exceeds a preset threshold, the device is determined to be abnormal, triggering an early warning signal.

[0025] When the early warning signal continues to exceed the preset time threshold, the fault classification unit is activated. It inputs the current device feature vector into the support vector machine model, calculates the decision function, outputs the fault type classification result, and completes the fault identification.

[0026] After fault identification is completed, the degradation trend prediction unit inputs the historical time series of the key performance parameters of the equipment, uses the long short-term memory network to predict the changes in the future, compares the predicted parameter curve with the preset failure threshold, finds the time point when the parameter exceeds the threshold, and calculates the remaining service life of the equipment from the current time to that time point.

[0027] A further preferred embodiment: The fault classification model uses a support vector machine to automatically identify fault types. The support vector machine obtains the optimal classification hyperplane by solving a quadratic programming problem, and its decision function is:

[0028]

[0029] Where x is the feature vector of the device to be diagnosed, x i Let be the feature vector of the i-th training sample, and n be the number of support vectors. i For the Lagrange multipliers obtained through training, y i Let b be the class label of the training samples, and K(x) be the bias term. i ,x) is the kernel function.

[0030] A further preferred embodiment: the support vector machine uses a radial basis function as its kernel function, the expression of which is:

[0031]

[0032] Here, γ is the parameter of the kernel function, which is determined through grid search and cross-validation.

[0033] A further preferred embodiment: The anomaly detection model uses the isolated forest algorithm to discover deviations from the normal operation mode, thereby achieving early warning.

[0034] A further preferred embodiment: The isolated forest algorithm evaluates device data by constructing an ensemble of multiple isolated trees, and calculates an anomaly score s(x, n) for each data point;

[0035] The outlier score is determined by the average path length E(h(x)) of the data point across all isolated trees, and its calculation formula is as follows:

[0036]

[0037] Where h(x) is the path length, E(h(x)) is the expected value of h(x) in all isolated trees, and c(n) is the average path length given a sample size n;

[0038] When the abnormal score s(x,n) is greater than the preset threshold, x is determined to be an abnormal point and an early warning is triggered.

[0039] A further preferred embodiment: The degradation trend prediction model uses a long short-term memory network to predict the future trends of key performance parameters and estimate the remaining service life of the equipment;

[0040] The Long Short-Term Memory (LSTM) network comprises three gating units: an input gate, a forget gate, and an output gate. Through the gating units, information is selectively memorized, forgotten, and output. The network learns the long-term time dependencies in the process of device performance degradation, predicts the changing trends of key performance parameters, and calculates the remaining service life of the device through a linear degradation equation when the predicted value of the key performance parameter exceeds a preset failure threshold.

[0041] In addition, this invention also provides a fault diagnosis method for thermal power generation equipment based on big data analysis, applied to a fault diagnosis system for thermal power generation equipment based on big data analysis, including the following steps:

[0042] Step 1: The data acquisition module acquires real-time and historical data, and performs data cleaning, normalization, and feature extraction.

[0043] Step 2: Calculate the equipment operation data using the anomaly detection model in the analysis and diagnosis module, and output the anomaly score;

[0044] Step 3: When the abnormal score exceeds the preset threshold, an early warning is triggered;

[0045] Step 4: When the early warning continues to exceed the time threshold, the fault type is matched using the fault classification model, and the diagnostic results are output.

[0046] Step 5: Based on the degradation trend prediction model, predict the failure development speed and the remaining service life of the equipment, and generate a diagnostic report;

[0047] Step 6: Feed the diagnostic results and maintenance records back to the system's knowledge base, and regularly train and optimize the machine learning model of the diagnostic module.

[0048] The embodiments of the present invention have the following advantages due to the adoption of the above technical solutions:

[0049] I. This invention uses a data acquisition module to monitor the operating status of main and auxiliary equipment in thermal power generation. The SCADA unit, turbine TSI unit, auxiliary vibration monitoring unit and electrical monitoring unit of the thermal power plant work together to acquire various key indicators of thermal power generation equipment, providing basic support for big data analysis.

[0050] Second, this invention analyzes and warns of equipment faults through an analysis and diagnosis module, achieves early warning through the isolated forest algorithm, automatically classifies fault types through support vector machines, and captures the dependencies of equipment performance degradation through long short-term memory networks, thus realizing a complete process from fault identification to trend prediction.

[0051] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a structural diagram of the present invention. Detailed Implementation

[0054] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0055] It should be understood that the following specific examples illustrate the implementation of this disclosure, and those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0056] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0057] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The drawings only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0058] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0059] Example 1

[0060] like Figure 1 As shown, this embodiment of the invention provides a fault diagnosis system for thermal power generation equipment based on big data analysis, including:

[0061] The data acquisition module is used to collect operating data from the main and auxiliary equipment of thermal power generation.

[0062] The data processing module, connected to the data acquisition module, is used to perform timestamp alignment, cleaning, standardization, and feature extraction on data from different sampling frequencies to form a unified panoramic view of the equipment's operation.

[0063] The analysis and diagnosis module is connected to the data processing module and is used to perform fault classification, anomaly detection, and degradation trend prediction on the processed data based on machine learning models.

[0064] The early warning module is connected to the analysis and diagnosis module to visualize the health status of equipment, push alarm information, and generate diagnostic reports.

[0065] In this embodiment, specifically: the data acquisition module includes:

[0066] The SCADA unit in a thermal power plant is used to collect process parameters such as temperature, pressure, flow rate, and load of the boiler, steam turbine, and generator. The SCADA unit includes data acquisition equipment such as thermocouples, pressure transmitters, differential pressure flow meters, radar level gauges, and valve opening transmitters. The SCADA unit continuously collects process parameters reflecting the thermodynamic cycle at a low sampling frequency (interval time 1-60 minutes), including boiler steam temperature and pressure, steam turbine inlet steam flow rate and temperature, generator active power and load, etc., to assess the overall operating status of the unit.

[0067] The turbine TSI unit is used to collect mechanical parameters of the turbine, such as shaft vibration, shaft displacement, differential expansion, speed, eccentricity, and key phase. The turbine TSI unit includes data acquisition equipment such as eddy current sensors, speed sensors, and linear displacement sensors. The turbine TSI unit collects mechanical parameters at a high sampling frequency (interval time 1-60 seconds), such as the vibration amplitude and phase of the shaft, the axial displacement of the rotor relative to the stator, the thermal expansion difference between the rotor and the cylinder, as well as the speed and eccentricity, to assess early mechanical failures of the turbine.

[0068] The auxiliary machine vibration monitoring unit is used to collect high-frequency vibration data of auxiliary machines such as fans, water pumps, and coal mills. By installing vibration sensors in key auxiliary machines such as fans, water pumps, and coal mills, high-frequency vibration data is collected. By analyzing the vibration data, local defects such as bearing damage, blade wear, and gear breakage can be diagnosed.

[0069] The electrical monitoring unit is used to collect electrical parameters such as voltage, current, and power factor. It obtains electrical parameters from the power plant's electrical system, such as three-phase voltage, current, power factor, and harmonic content, and analyzes the health status of electrical equipment such as generators, transformers, and motors.

[0070] In this embodiment, specifically: the data processing module includes a processing unit and a storage unit;

[0071] The processing unit aligns data from different sources and frequencies on timestamps, cleans, standardizes, and extracts features to form a unified panoramic view of the equipment's operation. Since the data sampling frequencies from units such as SCADA and TSI differ, this unit uses linear interpolation to unify all data onto the same timestamp sequence, forming a synchronized dataset. During data processing, hard safety ranges for each parameter are set according to the equipment design manual and operating procedures. Noise and outliers significantly exceeding the range are removed from the data. Z-Score standardization is used to scale the data to a similar scale, extracting meaningful features from the original signal. For example, temporal features such as effective values, peak values, and kurtosis are extracted from vibration signals, and frequency domain features such as first and second harmonics are extracted using fast Fourier transform. Frequency domain features are used for support vector machine classification, and temporal features are used for long short-term memory network prediction. After timestamp alignment and standardization, all features together form a fixed-dimensional feature vector. The feature list of the feature vector can be adjusted according to the specific equipment.

[0072] The storage unit includes a database for storing real-time and processed data, and a distributed big data platform for storing historical data. The database stores recent (within a few days) processed high-frequency real-time data, supporting fast querying and calculation by the analysis and diagnosis module. The distributed big data platform uses Hadoop HDFS technology to store massive amounts of historical data, feature data, and diagnostic results for a long time, providing a data foundation for subsequent model training and optimization.

[0073] In this embodiment, the analysis and diagnosis module specifically includes a fault classification unit, an anomaly detection unit, and a degradation trend prediction unit.

[0074] In this embodiment, the specific working steps of the analysis and diagnosis module include:

[0075] The machine learning models in the analysis and diagnosis module are trained using historical data;

[0076] During equipment monitoring, the anomaly detection unit uses the isolated forest algorithm to calculate the input real-time equipment feature data and generate an anomaly score for each data point.

[0077] Real-time monitoring of changes in anomaly scores; when the anomaly score of the device data exceeds a preset threshold, the device is determined to be abnormal, triggering an early warning signal.

[0078] When the early warning signal continues to exceed the preset time threshold, the fault classification unit is activated. It inputs the current device feature vector into the support vector machine model, calculates the decision function, outputs the fault type classification result, and completes the fault identification.

[0079] After fault identification is completed, the degradation trend prediction unit inputs the historical time series of the key performance parameters of the equipment, uses the long short-term memory network to predict the changes in the future, compares the predicted parameter curve with the preset failure threshold, finds the time point when the parameter exceeds the threshold, and calculates the remaining service life of the equipment from the current time to that time point.

[0080] When training the machine learning models in the analysis and diagnosis module, historical data of the device under known health conditions or containing historical failure cases are used as the training set.

[0081] In this embodiment, specifically: the fault classification model uses a support vector machine to automatically identify fault types. The support vector machine obtains the optimal classification hyperplane by solving a quadratic programming problem, and its decision function is:

[0082]

[0083] Where x is the feature vector of the device to be diagnosed, x i Let be the feature vector of the i-th training sample, and n be the number of support vectors. i For the Lagrange multipliers obtained through training, y i Let b be the class label of the training samples, and K(x) be the bias term. i ,x) is the kernel function.

[0084] In this embodiment, specifically: the support vector machine uses a radial basis function as its kernel function, and its expression is:

[0085]

[0086] Here, γ is the parameter of the kernel function, which is determined through grid search and cross-validation.

[0087] In this embodiment, specifically: the anomaly detection model uses the isolated forest algorithm to discover deviations from the normal operation mode, thereby achieving early warning.

[0088] When a device issues a fault warning, the fault classification unit classifies the fault type using a support vector machine with radial basis functions as the kernel function. In practical applications, when solving multi-class classification problems, a binary support vector machine sub-classifier is trained for every two fault types, and the final classification result is determined through a voting mechanism. During model training, the parameter γ is determined on the training set through grid search and k-fold cross-validation to select the model with the best generalization ability. During decision-making, the current feature vector x of the device to be diagnosed is input into the trained support vector machine, and the value of the decision function f(x) is calculated to output the fault category label.

[0089] In this embodiment, specifically: the isolated forest algorithm evaluates device data by constructing an ensemble of multiple isolated trees, and calculates an anomaly score s(x, n) for each data point;

[0090] The outlier score is determined by the average path length E(h(x)) of the data point across all isolated trees, and its calculation formula is as follows:

[0091]

[0092] Where h(x) is the path length, E(h(x)) is the expected value of h(x) in all isolated trees, and c(n) is the average path length given a sample size n;

[0093] When the abnormal score s(x,n) is greater than the preset threshold, x is determined to be an abnormal point and an early warning is triggered.

[0094] The anomaly detection unit does not rely on known fault labels. Instead, it learns the distribution patterns of historical normal data to discover anomalies that deviate slightly from the normal pattern. The Isolation Forest algorithm randomly samples data features and randomly segments them. By constructing multiple isolation trees, each data point is isolated. Because anomalies differ from the normal group, they can be isolated by a shorter path. An anomaly score s(x,n) is calculated for each data point. The closer the score is to 1, the higher the probability that the point is an anomaly. In this embodiment, the 99th percentile is used as the threshold. When the anomaly score of the real-time data continuously exceeds this threshold, an early warning is triggered.

[0095] In this embodiment, specifically: the degradation trend prediction model uses a long short-term memory network to predict the future trend of key performance parameters and estimate the remaining service life of the equipment;

[0096] Long Short-Term Memory (LSTM) networks consist of three gating units: an input gate, a forget gate, and an output gate. Through these gating units, information is selectively memorized, forgotten, and output. The network learns the long-term time dependencies in the process of device performance degradation, predicts the changing trends of key performance parameters, and calculates the remaining service life of the device using a linear degradation equation when the predicted values ​​of key performance parameters exceed a preset failure threshold.

[0097] The Long Short-Term Memory (LSTM) network employs a two-layer LSTM structure, with each layer containing 128 hidden units. Its gating structure effectively captures long-term dependencies in time series data. The model takes the performance parameter sequence of a device over a historical period (e.g., the past 30 days) as input and learns to predict parameter values ​​for a future period (e.g., the next 7 days). The system sets a preset failure threshold for key parameters, calculated based on device design standards or historical fault data. By continuously predicting future parameter values, the system finds the point in time when the predicted curve first exceeds the failure threshold. The time interval from the current time to the point when the failure threshold is first exceeded is the predicted remaining service life.

[0098] The early warning module's functions include visualizing equipment health status, pushing alarm information, and generating diagnostic reports. By using dashboards and trend curves on the monitoring screen, it comprehensively displays the real-time status of the equipment, the anomaly score change curve, and the predicted change trend. When an early warning is triggered or a specific fault is diagnosed, information such as the alarm level, equipment name, and fault type is pushed to relevant maintenance personnel via SMS, email, etc. At the same time, it regularly generates structured diagnostic reports, including descriptions of abnormal events, fault cause analysis, and development trend predictions, providing data support for maintenance decisions.

[0099] When this invention is in operation: the data acquisition module is activated, and the power plant SCADA unit, turbine TSI unit, auxiliary machine vibration monitoring unit, and electrical monitoring unit begin collecting various operating data of the power generation equipment. The power plant SCADA unit continuously acquires process parameters such as temperature, pressure, flow rate, and load of the main equipment such as boilers, turbines, and generators at a set low sampling frequency; the turbine TSI unit acquires mechanical parameters such as shaft vibration and shaft displacement of the turbine at a higher sampling frequency; the auxiliary machine vibration monitoring unit acquires high-frequency vibration data of auxiliary machines such as fans, pumps, and coal mills; and the electrical monitoring unit obtains electrical parameters such as voltage, current, and power factor from the power plant's electrical system.

[0100] The collected data is transmitted to the data processing module. After receiving the data from different units, the processing unit uses linear interpolation to unify all the data onto the same timestamp sequence to form a synchronized dataset. The data is then cleaned to remove noise and outliers that are significantly out of range. The Z-Score normalization method is used to scale the data to a similar scale for subsequent analysis to extract meaningful features from the original signal.

[0101] The processed data is partly stored in a database to support rapid querying and calculation by the analysis and diagnosis module, and partly stored long-term on a distributed big data platform using Hadoop HDFS technology to provide a data foundation for subsequent model training and optimization.

[0102] The processed data is sent to the analysis and diagnosis module. The fault classification unit uses a support vector machine with radial basis functions as kernel functions to classify the fault types of the equipment. The current feature vector x of the equipment to be diagnosed is input into the trained support vector machine. By calculating the value of the decision function f(x), the fault category label is output.

[0103] The anomaly detection unit uses the isolated forest algorithm to find deviations from the normal operation mode. By constructing multiple isolated trees and integrating them, an anomaly score s(x, n) is calculated for each data point. When the anomaly score s(x, n) is greater than the preset threshold (99th percentile in this embodiment), x is determined to be an anomaly point and an early warning is triggered.

[0104] The degradation trend prediction unit utilizes a Long Short-Term Memory (LSTM) network to predict the future trends of key performance parameters and estimate the remaining service life of the equipment. The LSM network comprises three gating units: an input gate, a forget gate, and an output gate. Through selective memorization, forgetting, and output of information by these gating units, it learns the long-term time dependencies in the equipment's performance degradation process. The model takes a sequence of performance parameters from a historical period (e.g., the past 30 days) as input and learns to predict parameter values ​​for a future period (e.g., the next 7 days). When the predicted value of a key performance parameter exceeds a preset failure threshold, the remaining service life of the equipment is calculated using a linear degradation equation.

[0105] The results from the analysis and diagnosis module are transmitted to the early warning module, which comprehensively displays the real-time status of the equipment, the anomaly score change curve, and the predicted trend on the monitoring screen using dashboards and trend curves. When an early warning is triggered or a specific fault is diagnosed, information such as the alarm level, equipment name, and fault type is pushed to relevant maintenance personnel via SMS, email, etc. Simultaneously, structured diagnostic reports are generated regularly to provide data support for maintenance decisions.

[0106] Example 2

[0107] like Figure 1 As shown, this embodiment of the invention provides a fault diagnosis system for thermal power generation equipment based on big data analysis, including:

[0108] The data acquisition module is used to collect operating data from the main and auxiliary equipment of thermal power generation.

[0109] The data processing module, connected to the data acquisition module, is used to perform timestamp alignment, cleaning, standardization, and feature extraction on data from different sampling frequencies to form a unified panoramic view of the equipment's operation.

[0110] The analysis and diagnosis module is connected to the data processing module and is used to perform fault classification, anomaly detection, and degradation trend prediction on the processed data based on machine learning models.

[0111] An early warning module, connected to the analysis and diagnosis module, is used to visualize the equipment health status, push alarm information, and generate diagnostic reports.

[0112] In this embodiment, the working status of the steam turbine is detected by a fault diagnosis system for thermal power generation equipment based on big data analysis, and early warnings are given for local faults.

[0113] In this embodiment, when the fault diagnosis system for thermal power generation equipment is working, the TSI unit of the steam turbine continuously monitors the shaft vibration data in the X and Y directions of the bearing at a sampling frequency of 10 times per second, and the SCADA unit of the thermal power plant collects process data such as unit load and steam parameters at a frequency of 1 time per minute.

[0114] The data processing module processes the high-frequency vibration signals transmitted by the TSI, completes data cleaning and standardization, and then extracts key features, including the effective value of vibration, peak value, and kurtosis. The effective value of vibration of a certain bearing rises from 4.5 mm / s to 5.8 mm / s, and the kurtosis rises from 3.2 to 6.5. The spectrum is extracted through fast Fourier transform to obtain information that the vibration energy is mainly concentrated at the fundamental frequency and the amplitude of the second harmonic increases slightly.

[0115] The anomaly detection unit is trained based on the normal vibration characteristic data of a certain bearing in the past three months. For each newly collected vibration feature vector x, the anomaly detection unit calculates its path length h(x) in each isolation tree, that is, the number of edges required to reach this data point from the root node, and then calculates the expected value E(h(x)) of the path lengths in all trees. Combining with the normalization constant c(n) corresponding to the sample size n, the following formula is used to calculate the anomaly score:

[0116]

[0117] When the real-time feature vector continuously generates a high anomaly score, the system determines that the operating mode has deviated from the normal range and triggers an early warning.

[0118] After the early warning lasts for 1 hour, the fault classification unit is activated, and the current feature vector x [effective value = 5.8, kurtosis = 6.5, fundamental frequency amplitude = 85 μm, second harmonic amplitude = 15 μm,...] is input into the trained support vector machine. Through the decision function:

[0119]

[0120] The input features are compared and calculated with the support vectors. Since the mode that the fundamental frequency amplitude dominates in the feature vector is obvious, the decision function f(x) outputs the classification result as dynamic imbalance. Combining with the maintenance relational knowledge base, the system further classifies the fault cause as the deterioration of the rotor dynamic balance caused by the loosening of the bearing seat.

[0121] Meanwhile, the degradation trend prediction unit uses the historical data sequence of the effective value of a bearing vibration over the past 30 days as input and makes predictions through a long short-term memory network. The model predicts that the vibration value will exceed the failure threshold for safe operation of the equipment within the next 15 days (in this embodiment, the failure threshold is 10 mm / s).

[0122] Based on this predicted trend, the system calculates the predicted remaining service life of the bearing to be 15 days and issues an early warning through the early warning module.

[0123] Example 3

[0124] This invention also provides a fault diagnosis method for thermal power generation equipment based on big data analysis, applied to a fault diagnosis system for thermal power generation equipment based on big data analysis, comprising the following steps:

[0125] Step 1: The data acquisition module acquires real-time and historical data, and performs data cleaning, normalization, and feature extraction.

[0126] Step 2: Calculate the equipment operation data using the anomaly detection model in the analysis and diagnosis module, and output the anomaly score;

[0127] Step 3: When the abnormal score exceeds the preset threshold, an early warning is triggered;

[0128] Step 4: When the early warning continues to exceed the time threshold, the fault type is matched through the fault classification model and the diagnosis result is output. The time threshold is preset according to the development speed of different fault modes and combined with operation and maintenance experience.

[0129] Step 5: Based on the degradation trend prediction model, predict the failure development speed and the remaining service life of the equipment, and generate a diagnostic report;

[0130] Step 6: Feed the diagnostic results and maintenance records back to the system's knowledge base, and regularly train and optimize the machine learning model of the diagnostic module, for example, automatically perform training and optimization once a month.

[0131] This invention uses a data acquisition module to monitor the operating status of main and auxiliary equipment in thermal power generation. The SCADA unit, turbine TSI unit, auxiliary vibration monitoring unit and electrical monitoring unit of the thermal power plant work together to acquire various key indicators of thermal power generation equipment, providing basic support for big data analysis.

[0132] This invention analyzes and provides early warnings of equipment faults through an analysis and diagnosis module, achieves early warning through the Isolation Forest algorithm, automatically classifies fault types through Support Vector Machine, and captures the dependencies of equipment performance degradation through Long Short-Term Memory Network, thus realizing a complete process from fault identification to trend prediction.

[0133] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0134] In this disclosure, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The block diagrams of devices, apparatuses, devices, and systems involved in this disclosure are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as "comprising," "including," "having," etc., are open-ended terms meaning "including but not limited to," and are used interchangeably with them. The terms "or" and "and" as used herein refer to the terms "and / or," and are used interchangeably with them unless the context clearly indicates otherwise. The term "such as" as used herein refers to the phrase "such as but not limited to," and is used interchangeably with it.

[0135] It should also be noted that in the system disclosed herein, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0136] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A fault diagnosis system for thermal power generation equipment based on big data analysis, characterized in that, include: The data acquisition module is used to collect operating data from the main and auxiliary equipment of thermal power generation. The data processing module, connected to the data acquisition module, is used to perform timestamp alignment, cleaning, standardization, and feature extraction on data from different sampling frequencies to form a unified panoramic view of the equipment operation. The analysis and diagnosis module is connected to the data processing module and is used to perform anomaly detection, fault classification and degradation trend prediction on the processed data based on a machine learning model. The early warning module is connected to the analysis and diagnosis module and is used to visualize the health status of the equipment, push alarm information, and generate diagnostic reports.

2. The fault diagnosis system for thermal power generation equipment based on big data analysis according to claim 1, characterized in that, The data acquisition module includes: The SCADA unit in a thermal power plant is used to collect temperature, pressure, flow rate, and load process parameters of the boiler, steam turbine, and generator. The turbine TSI unit is used to collect parameters such as turbine shaft vibration, shaft displacement, differential expansion, speed, eccentricity, and key phase mechanical quantities. Auxiliary machine vibration monitoring unit is used to collect high-frequency vibration data of fans, water pumps, and coal mill auxiliary equipment; The electrical monitoring unit is used to collect electrical parameters such as voltage, current, and power factor.

3. The fault diagnosis system for thermal power generation equipment based on big data analysis according to claim 1, characterized in that, The data processing module includes a processing unit and a storage unit; The processing unit is used to align data from different sources and frequencies on timestamps, and to clean, standardize, and extract features from the data to form a unified panoramic view of the equipment's operation. The storage unit includes a database for storing real-time data and processed data, and a distributed big data platform for storing historical data.

4. The fault diagnosis system for thermal power generation equipment based on big data analysis according to claim 1, characterized in that, The analysis and diagnosis module includes a fault classification unit, an anomaly detection unit, and a degradation trend prediction unit.

5. The fault diagnosis system for thermal power plants based on big data analysis according to claim 4, characterized in that, The specific working steps of the analysis and diagnosis module include: The machine learning models in the analysis and diagnosis module are trained using historical data; During equipment monitoring, the anomaly detection unit uses the isolated forest algorithm to calculate the input real-time equipment feature data and generate an anomaly score for each data point. Real-time monitoring of changes in anomaly scores; when the anomaly score of the device data exceeds a preset threshold, the device is determined to be abnormal, triggering an early warning signal. When the early warning signal continues to exceed the preset time threshold, the fault classification unit is activated. It inputs the current device feature vector into the support vector machine model, calculates the decision function, outputs the fault type classification result, and completes the fault identification. After identification, the degradation trend prediction unit inputs the historical time series of the key performance parameters of the device, uses the long short-term memory network to predict the changes in the future, compares the predicted parameter curve with the preset failure threshold, finds the time point when the parameter exceeds the threshold, and calculates the remaining service life of the device from the current time to that time point.

6. The fault diagnosis system for thermal power plants based on big data analysis according to claim 5, characterized in that, The isolated forest algorithm evaluates device data by constructing an ensemble of multiple isolated trees and calculates an anomaly score s(x, n) for each data point; The outlier score is determined by the average path length E(h(x)) of the data point across all isolated trees, and its calculation formula is as follows: Where h(x) is the path length, E(h(x)) is the expected value of h(x) in all isolated trees, and c(n) is the average path length given a sample size n; When the abnormal score s(x,n) is greater than the preset threshold, x is determined to be an abnormal point and an early warning is triggered. 7.The thermal power plant fault diagnosis system based on big data analysis of claim 5, wherein, The fault classification model uses a support vector machine (SVM) to automatically identify fault types. The SVM obtains the optimal classification hyperplane by solving a quadratic programming problem, and its decision function is: Where x is the feature vector of the device to be diagnosed, x i Let be the feature vector of the i-th training sample, and n be the number of support vectors. i For the Lagrange multipliers obtained through training, y i Let b be the class label of the training samples, and K(x) be the bias term. i ,x) is the kernel function. 8.The thermal power plant fault diagnosis system based on big data analysis of claim 7, wherein, The support vector machine uses radial basis functions as its kernel function, and its expression is: Here, γ is the parameter of the kernel function, which is determined through grid search and cross-validation. 9.The thermal power plant fault diagnosis system based on big data analysis of claim 5, wherein, The Long Short-Term Memory (LSTM) network comprises three gating units: an input gate, a forget gate, and an output gate. Through the gating units, information is selectively memorized, forgotten, and output. The network learns the long-term time dependencies in the process of device performance degradation, predicts the changing trends of key performance parameters, and calculates the remaining service life of the device through a linear degradation equation when the predicted value of the key performance parameter exceeds a preset failure threshold.

10. A method for diagnosing faults of a thermal power plant based on big data analysis, applied to a system for diagnosing faults of a thermal power plant based on big data analysis according to any one of claims 1-9, characterized in that, Includes the following steps: Step 1: The data acquisition module acquires real-time and historical data, and performs data cleaning, normalization, and feature extraction. Step 2: Calculate the equipment operation data using the anomaly detection model in the analysis and diagnosis module, and output the anomaly score; Step 3: When the abnormal score exceeds the preset threshold, an early warning is triggered; Step 4: When the early warning continues to exceed the time threshold, the fault type is matched using the fault classification model, and the diagnostic results are output. Step 5: Based on the degradation trend prediction model, predict the failure development speed and the remaining service life of the equipment, and generate a diagnostic report; Step 6: Feed the diagnostic results and maintenance records back to the system's knowledge base, and regularly train and optimize the machine learning model of the diagnostic module.

Images