Intelligent monitoring method for running state of hydraulic power plant equipment
By building a target sample library and real-time data analysis, combining historical information and expert knowledge, intelligent monitoring and fault warning of hydropower plant equipment is achieved, and the problem of inaccurate diagnosis results in the existing technology is solved, and the operation efficiency and reliability of the equipment is improved. .
Patent Information
- Application Number
- CN202411156388.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-05
AI Technical Summary
The diagnosis results of existing hydropower plant equipment fault diagnosis methods are incomplete and inaccurate, resulting in long downtime and many downtimes.
A method of intelligent monitoring of the operating status of hydropower plant equipment is adopted. By building a target sample library, equipment parameters are collected in real time, feature quantity is extracted, and the comparison and analysis is carried out. Combining historical fault cycle information and expert knowledge, maintenance strategies are determined to achieve equipment health status. Evaluation and fault warning.
It realizes equipment status awareness, fault warning and health management, reduces unplanned downtime, improves power generation efficiency, reduces maintenance costs, and improves equipment reliability and availability.
Smart Images

Figure CN118897956A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of hydropower, and in particular to an intelligent monitoring method for the operating status of hydropower plant equipment. Background Art
[0002] As a key infrastructure for national economic and social development, the safe and stable operation of hydropower plant equipment is crucial to the reliability of power supply. Traditional equipment maintenance relies on regular manual inspections and simple status monitoring, but it is often difficult to detect early signs of equipment failure in a timely manner.
[0003] At present, a large number of studies have been carried out at home and abroad to improve equipment fault diagnosis. These studies mainly obtain characteristic parameters through equipment status monitoring, build a fault knowledge base based on expert knowledge and fault cases, and use pattern recognition and other methods to achieve fault diagnosis. However, due to the complexity of hydropower units and the diversity of fault manifestations, the existing diagnosis methods based on limited fault samples have limited application effects in actual projects, and the diagnosis results are incomplete and inaccurate, resulting in long downtime and high number of downtimes. Summary of the invention
[0004] In order to overcome the shortcomings of the prior art, the present invention provides an intelligent monitoring method for the operating status of hydropower plant equipment to solve the problems of the prior art such as incomplete and inaccurate diagnosis results leading to long downtime and frequent downtime.
[0005] The technical solution adopted by the present invention to solve the above problems is:
[0006] A method for intelligently monitoring the operating status of hydropower plant equipment comprises the following steps:
[0007] S100, building a target sample library of hydropower plant equipment;
[0008] S200, collecting real-time operating parameters of the device, and then extracting feature quantities based on the real-time operating parameters;
[0009] S300, comparing the characteristic quantity with the target sample library to obtain operation prediction information of the device;
[0010] S400, obtaining equipment health status assessment results based on operation prediction information and historical failure cycle information;
[0011] S500, determining a maintenance strategy based on the equipment health status assessment result, the expert knowledge base, and the equipment maintenance history data information.
[0012] As a preferred technical solution, step S100 includes the following steps:
[0013] S110, determining key characteristic parameters of the device;
[0014] S120, determining the operating condition type according to the operating characteristics of the equipment;
[0015] S130, obtaining first historical data of key characteristic parameters under various operating conditions;
[0016] S140, acquiring statistical characteristic information based on the first historical data;
[0017] Wherein, the statistical characteristic information includes time domain characteristic information, frequency domain characteristic information and probability distribution characteristic information;
[0018] The time domain characteristic information includes one or more of mean, standard deviation, maximum, minimum, median, and peak-to-peak value;
[0019] The frequency domain characteristic information includes one or more of the dominant frequency, the amplitude of the dominant frequency, the spectrum entropy, and the frequency multiplication component;
[0020] The probability distribution characteristic information includes one or more of distribution type, skewness, and kurtosis;
[0021] S150, determining the standard range and control limit information of key characteristic parameters of the equipment under corresponding operating conditions based on the statistical characteristic information;
[0022] The standard range and control limit information of key characteristic parameters of several types of equipment under different operating conditions is the target sample library.
[0023] As a preferred technical solution, step S200 includes the following steps:
[0024] S210, collecting the operating parameters of the equipment in real time and storing them in time series;
[0025] S220, preprocessing the operating parameters;
[0026] S230, extracting a feature quantity reflecting the state of the device from the preprocessed operating parameters;
[0027] Among them, the feature quantities include time domain features and frequency domain features; the time domain features include one or more of mean, peak-to-peak value, root mean square, waveform factor, and peak factor; the frequency domain features include one or more of spectrum center, frequency band energy, and characteristic frequency amplitude.
[0028] As a preferred technical solution, step S300 includes the following steps:
[0029] A100, preprocessing the actual parameter information corresponding to the current running state collected, so that the actual parameter information after preprocessing is consistent with the numerical range of the processed sample;
[0030] A200, extracts actual time domain features and actual frequency domain features from the preprocessed actual parameter information;
[0031] A300, uses sample data to train the constructed SVM classification model;
[0032] The sample data includes a first sample set data and a second sample set data, wherein the first sample set data is N steady-state vibration data samples selected from the normal operation of the unit within one year, and the second sample set data is M vibration data selected under various fault conditions, where N≥10M;
[0033] A400, analyzes the feature quantity based on the trained SVM classification model and the target sample library to obtain the operation prediction information of the corresponding equipment.
[0034] As a preferred technical solution, step S300 includes the following steps:
[0035] B100, taking the time series composed of the collected real-time operating parameters as the observed variables, and establishing the state space model;
[0036] B200, using historical monitoring data to identify unknown parameters in the state space model through the expectation maximization algorithm;
[0037] B300, uses the identified state space model to make one-step-ahead predictions of feature quantities;
[0038] B400, calculating the residual between the newly collected data of the device within a preset period and the prediction one step ago; obtaining the mean or variance of the residual; if the mean or variance exceeds the corresponding preset threshold, it is determined that the corresponding device is abnormal, and the operation prediction information is obtained;
[0039] The operation prediction information includes current component abnormality information and potential abnormality information of associated components.
[0040] As a preferred technical solution, in step B100, new real-time operating parameters of the equipment are regularly collected through a set sensor monitoring network;
[0041] Based on the real-time operating parameters collected at time t, the predicted value at time t+1 is predicted through the state space model;
[0042] Based on the predicted value at time t+1 and the measured data at time t+1, the prediction error is obtained;
[0043] Dynamically optimize the state space model based on prediction error and forgetting factor.
[0044] As a preferred technical solution, step S400 includes the following steps:
[0045] S410, constructing an analysis model;
[0046] S420, selecting a number of samples from the target sample library as a training set, each sample including a number of feature variables and a corresponding state label;
[0047] S430, normalizing all feature variables, and using the processed training set to train the analysis model;
[0048] S440, preprocessing the actual parameter information corresponding to the current running state collected, so that the preprocessed actual parameter information is consistent with the numerical range of the processed sample;
[0049] S450, inputting the pre-processed actual parameter information, operation prediction information, and historical failure cycle information into the trained analysis model to obtain the equipment health status assessment result.
[0050] As a preferred technical solution, step S500 includes the following steps:
[0051] C100, based on operation prediction information and historical fault cycle information, preliminarily determines the fault type and fault location;
[0052] C200, based on the normal operating parameter information of the fault part and the operating parameter information when the fault exists, obtains the fault change trend;
[0053] C300, based on the fault change trend, obtains the equipment health status assessment results;
[0054] The equipment health status assessment result includes fault development stage information, fault severity information and fault trend change information.
[0055] As a preferred technical solution, step S500 includes the following steps:
[0056] D100, matching the equipment health status assessment result with the expert knowledge base to obtain the first matching degree;
[0057] D200, matching the equipment health status assessment result with the equipment maintenance history data information to obtain a second matching degree;
[0058] If the second matching degree and the first matching degree are both greater than the matching threshold, the first maintenance strategy is obtained from the expert knowledge base, and the second maintenance strategy is obtained from the equipment maintenance history data information;
[0059] Generate a maintenance strategy based on the first maintenance strategy and the second maintenance strategy;
[0060] If the second matching degree and the first matching degree are both less than the matching threshold, the transition strategy is triggered to execute;
[0061] If the second matching degree is greater than the matching threshold and the first matching degree is less than the matching threshold, the maintenance strategy obtained from the equipment maintenance history data information is used as the maintenance strategy;
[0062] If the second matching degree is smaller than the matching threshold and the first matching degree is larger than the matching threshold, the inspection and repair strategy obtained from the expert knowledge base is used as the maintenance strategy.
[0063] As a preferred technical solution, the transition strategy includes: triggering the monitoring system to monitor the overall operating status and operating parameters of the corresponding equipment in real time;
[0064] If the operating status monitored within the preset period is normal, and each operating parameter is within the range of the target sample library, then the corresponding device is determined to be normal.
[0065] Compared with the prior art, the present invention has the following beneficial effects:
[0066] The present invention can realize the state perception, fault warning and health management of hydropower plant equipment, reduce unplanned downtime and improve power generation efficiency; by constructing a target sample library including the standard range and control limit information of various characteristic parameters of various types of equipment under different operating conditions, as well as accurate feature extraction, it can realize comprehensive monitoring and accurate prediction of the equipment operation status, avoiding the problem of incomplete and inaccurate diagnosis results in traditional methods; by predicting equipment performance degradation and potential failures in advance, the possible failure trend of the equipment can be discovered in advance, the probability of sudden failure can be reduced, and preventive maintenance measures can be taken to avoid long-term downtime and frequent maintenance caused by equipment failure; by regularly evaluating the health status of equipment and formulating optimized maintenance strategies, it can effectively reduce maintenance costs, reduce unnecessary maintenance and downtime, and improve the overall reliability and availability of equipment; by combining expert knowledge and a large amount of historical data to determine the maintenance strategy, it can make more rational and data-based decisions, which can effectively improve the intelligence level of equipment management and significantly improve the operating efficiency and reliability of hydropower plants. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] Figure 1 A schematic diagram of a flow chart of a method for intelligently monitoring the operating status of hydropower plant equipment provided in an embodiment of the present disclosure;
[0068] Figure 2 A schematic diagram of a process for constructing a target sample library provided by an embodiment of the present disclosure;
[0069] Figure 3 A flowchart of a method for extracting feature quantities provided in an embodiment of the present disclosure;
[0070] Figure 4A flowchart of a method for obtaining operation prediction information of various types of equipment provided in an embodiment of the present disclosure;
[0071] Figure 5 A flowchart of a method for obtaining a device health status assessment result provided in an embodiment of the present disclosure;
[0072] Figure 6 A flowchart of a method for obtaining operation prediction information of various types of equipment provided in a second embodiment of the present disclosure;
[0073] Figure 7 A flowchart of a method for obtaining a device health status assessment result provided by a second embodiment of the present disclosure;
[0074] Figure 8 A flowchart of a method for obtaining a maintenance strategy provided by an embodiment of the present disclosure;
[0075] Fig. 9 A principle block diagram of an intelligent monitoring system for the operation status of hydropower plant equipment provided by an embodiment of the present disclosure;
[0076] Fig.10 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure.
[0077] Marks and their corresponding names in the attached drawings: 10. Intelligent monitoring system for equipment operation status of hydropower plant, 11. Construction module, 12. Extraction module, 13. Comparison module, 14. Analysis module, 15. Generation module. DETAILED DESCRIPTION
[0078] The present invention will be further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.
[0079] Example 1
[0080] like Figures 1 to 10 As shown, refer to Figure 1 The first aspect of the present application discloses a method for intelligently monitoring the operating status of hydropower plant equipment, which specifically comprises the following steps:
[0081] S100, constructing a target sample library of hydropower plant equipment; the target sample library includes standard ranges and control limit information of various characteristic parameters of various equipment under different operating conditions.
[0082] In this step, it is first necessary to collect and organize the characteristic parameter data of various hydropower plant equipment under different operating conditions, including various performance parameters, working status, operating data, etc. of the equipment. The purpose of establishing the target sample library is to define the standard range and control limit information of the normal operation of the equipment, and provide a basis for subsequent operation prediction and health assessment. The target sample library in this embodiment refers to the healthy sample library.
[0083] This step can provide the standard range of each characteristic parameter of the equipment, which helps to identify whether the equipment is in the expected working state. It can also provide the possible abnormal state boundary during equipment operation, facilitating subsequent fault prediction and formulation of maintenance strategies.
[0084] S200, extracting feature quantities based on a plurality of collected real-time operating parameters of a plurality of devices.
[0085] In this step, important feature quantities are extracted from the real-time operating data of several devices collected, where the feature quantities may be device performance indicators or status parameters obtained through data processing and feature engineering.
[0086] Through this step, massive real-time data can be converted into a small amount of representative feature quantities, which helps to improve the efficiency of subsequent data processing and analysis. In addition, the extracted feature quantities can more accurately reflect the operating status of the equipment, laying the foundation for subsequent operation prediction and fault diagnosis.
[0087] S300, comparing the characteristic quantity with the target sample library to obtain operation prediction information of various types of equipment; the operation prediction information includes performance degradation trend information, stable operation prediction cycle information, and associated potential fault information.
[0088] In this step, the extracted feature quantities are compared and analyzed with the target sample library established in S100. The operating trend, performance degradation, stable operating cycle and potential fault information of the equipment can be predicted based on real-time data. Through prediction, possible problems can be discovered in advance, so that preventive maintenance measures can be taken.
[0089] Specifically, based on machine learning models and big data analysis, the performance changes and failure risks of equipment can be accurately predicted, and possible failure trends of equipment can be discovered in advance, which helps to take preventive maintenance measures and reduce downtime and repair costs.
[0090] S400, obtaining equipment health status assessment results based on operation prediction information and historical failure cycle information.
[0091] This step evaluates and analyzes the health status of the equipment based on the operation prediction information of S300 and the historical failure cycle data of the equipment. By combining multi-dimensional data analysis, it can more comprehensively evaluate the health status of the equipment, formulate personalized maintenance strategies based on the evaluation results, and optimize the maintenance management efficiency of the equipment, that is, it can more comprehensively judge the current working status of the equipment and possible problems in the future.
[0092] S500, determining a maintenance strategy based on the equipment health status assessment result, the expert knowledge base, and the equipment maintenance history data information.
[0093] Specifically, the determined maintenance strategy may include preventive maintenance, corrective maintenance, or recommendations for equipment replacement to ensure long-term stable operation of the equipment and minimize downtime.
[0094] In this step, a specific maintenance strategy is formulated based on the equipment health status assessment results, expert knowledge base, and equipment maintenance history data. The maintenance plan can be adjusted according to actual conditions to ensure a balance between maintenance costs and equipment availability. Appropriate maintenance measures can be taken for different equipment states, which can significantly improve the equipment's operating efficiency and lifespan.
[0095] Through the intelligent monitoring method for the operating status of hydropower plant equipment disclosed in the present application, it is possible to realize the state perception, fault warning and health management of hydropower plant equipment, reduce unplanned downtime and improve power generation efficiency; by constructing a target sample library including the standard range and control limit information of various characteristic parameters of various equipment under different operating conditions, as well as accurate feature extraction, it is possible to realize comprehensive monitoring and accurate prediction of the equipment operating status, avoiding the problem of incomplete and inaccurate diagnosis results in traditional methods; by predicting equipment performance degradation and potential failures in advance, it is possible to discover the possible failure trend of the equipment in advance, reduce the probability of sudden failures, and take preventive maintenance measures to avoid long-term downtime and frequent maintenance caused by equipment failures; by regularly evaluating the health status of equipment and formulating optimized maintenance strategies, it is possible to effectively reduce maintenance costs, reduce unnecessary maintenance and downtime, and improve the overall reliability and availability of equipment; by combining expert knowledge and a large amount of historical data to determine maintenance strategies, it is possible to make more rational and data-based decisions, which can effectively improve the intelligence level of equipment management and significantly improve the operating efficiency and reliability of hydropower plants.
[0096] Reference Figure 2 , the method for constructing the target sample library specifically includes the following steps:
[0097] S110, determining several key characteristic parameters of each type of equipment.
[0098] Specifically, the key characteristic parameters of the equipment can be determined through literature research, manufacturer's data, or on-site testing, such as the rated power, speed, efficiency of the motor, and the rated flow rate, head of the pump, etc. For different equipment, the corresponding key characteristic parameters are different, that is, the key characteristic parameters match the type of equipment.
[0099] In this embodiment, the key characteristic parameters refer to the parameters closely related to the normal operation of the equipment.
[0100] For example, for a hydro-generator set, the key characteristic parameters usually include vibration parameters, temperature parameters, pressure parameters, electrical parameters, efficiency parameters, and other parameters.
[0101] Among them, the vibration parameters include generator bearing vibration, thrust bearing vibration, upper guide bearing swing, etc.; the temperature parameters include stator winding temperature, rotor temperature, main shaft temperature, etc.; the pressure parameters include turbine inlet pressure, draft tube pressure, etc.; the electrical parameters include active power, reactive power, voltage, current, etc.; the efficiency parameters include turbine efficiency, generator efficiency, etc.; the other parameters include main shaft displacement, oil pressure, water level, etc.
[0102] According to the unit model and the monitoring system configuration, 20 - 30 of the most contributive monitoring quantities to the determination of the equipment status can be selected from hundreds of monitoring quantities as the key characteristics.
[0103] S120. Determine the type of operating conditions according to the operating characteristics of various equipment.
[0104] Specifically, the working state and operating mode of the equipment can be analyzed, such as the start-up, normal operation, load change of the motor, etc. For example, typical operating conditions include rated operation, overload operation, no-load operation, etc.
[0105] Specifically, according to the operating characteristics of the unit, the operating condition factors such as load and head that affect the characteristic parameters can be divided into several intervals. For example, the load can be divided into no-load operation, low-load operation, medium-load operation, and high-load operation. Among them, no-load operation means: power P < 5% of the rated; low-load operation means: 5% < power P < 50% of the rated; medium-load operation means: 50% < power P < 80% of the rated; high-load operation means: power P > 80% of the rated.
[0106] For example, according to the head, it can be divided into low-head operation, medium-head operation, and high-head operation. Among them, low-head operation means: H < 80% of the design head; medium-head operation means: 80% < H < 95% of the design head; high-head operation means: H > 95% of the design head.
[0107] Taking the combination of load and head, the unit operating conditions can be divided into: 4x3 = 12 intervals.
[0108] S130, obtaining first historical data of each key characteristic parameter under each operating condition.
[0109] In this embodiment, the first historical data refers to health data.
[0110] S140, based on the first historical data, obtaining statistical characteristic information; the statistical characteristic information includes time domain characteristic information, frequency domain characteristic information and probability distribution characteristic information;
[0111] The time domain characteristic information includes one or more of the mean, standard deviation, maximum value, minimum value, median, and peak-to-peak value; the frequency domain characteristic information includes one or more of the dominant frequency, the amplitude of the dominant frequency, spectral entropy, and harmonic components; the probability distribution characteristic information includes one or more of the distribution type, skewness, and kurtosis.
[0112] Through this step, the operating status and performance characteristics of the equipment can be quantified based on objective analysis of the data, revealing the changing patterns of the equipment under different working conditions, and providing a basis for formulating operation and maintenance strategies.
[0113] Assuming that the characteristic parameters follow a normal distribution, and taking "mean ± 3 times the standard deviation" as the healthy value range, take the stator winding temperature of a certain unit under a certain operating condition as an example: 1000 samples are counted, the mean is 75.2°C, the standard deviation is 6.8°C, the minimum is 65.1°C, the maximum is 88.5°C, and the distribution type is normal distribution, then the obtained healthy range is: [54.8, 95.6].
[0114] S150, based on the statistical characteristic information, determining the standard range and control limit information of each key characteristic parameter of each type of equipment under corresponding operating conditions;
[0115] The target sample library is composed of the standard range and control limit information of each key characteristic parameter of several types of equipment under different operating conditions.
[0116] This step can provide a standard reference range for safe and efficient operation of the equipment, which can help operation and maintenance personnel monitor the equipment status, identify abnormalities in a timely manner and take preventive measures to reduce failures and downtime.
[0117] Specifically, the statistical model of characteristic parameters under each working condition is summarized to construct a healthy sample library of unit status. The data format of the healthy sample is: {"Equipment No.": "Unit 1", "Collection Time": "2023-04-15 08:00:00", "Operation Condition": {"Active Power": 320MW, "Unit Head": 100m, "Turbine Opening": 80%...}, "Characteristic Parameters": [{"Name": "Stator Winding Temperature", "Current Value": 76.8, "Healthy Range": [54.8, 95.6], "Reference Value": 75.2, "Status": "Normal"}, {"Name": "Generator Bearing Vibration", "Current Value": 18.6, "Healthy Range": [2.4, 38.2], "Reference Value": 16.8, "Status": "Normal"}]}.
[0118] The method for constructing a target sample library disclosed in this embodiment can ensure analysis and decision-making based on real data, improve prediction accuracy and operational reliability; statistical characteristic information provides a tool for in-depth understanding of equipment performance and operating conditions, which helps to optimize maintenance plans and resource allocation; through clear standard ranges and control limits, maintenance costs can be effectively reduced, and unnecessary equipment damage and downtime can be reduced; the established target sample library can serve as the basis for continuous improvement and optimization, and support long-term monitoring and management of equipment performance.
[0119] In this embodiment, the target sample library can also be stored in a time series database to facilitate rapid retrieval and analysis of massive data.
[0120] Furthermore, the target sample library (i.e., the healthy sample library) can also be updated. Specifically, a data management platform can be built to provide functions such as data labeling, query, and export to support engineers in reviewing and revising sample data; using newly collected operating data, the statistical model of the healthy sample library can be regularly updated to keep it consistent with the actual health status of the unit.
[0121] After the healthy sample library is built, it can be used for the following purposes: 1) Unit status monitoring: compare the characteristic parameters collected in real time with the healthy sample library to detect abnormal deviations in a timely manner; 2) Fault diagnosis: after discovering the abnormality, use the fault template library to locate the cause of the fault and generate a diagnostic report; 3) Life prediction: statistically analyze the historical trends of health parameters of key components, evaluate the current health level, and predict the remaining service life; 4) Maintenance decision-making: combine the unit status evaluation results and maintenance cost-benefit to optimize the maintenance strategy and formulate an annual maintenance plan.
[0122] Reference Figure 3 The feature extraction method specifically includes the following steps:
[0123] S210, collecting a number of operating parameters of a number of devices in real time and storing them in time series.
[0124] Specifically, sensors, monitoring systems or device interfaces can be used to collect key operating parameters of the equipment in real time, such as temperature, pressure, vibration, current, etc.; storing the data in time series can ensure the continuity and timeliness of the data.
[0125] S220, preprocessing the operating parameters;
[0126] Preprocessing includes one or more of outlier detection, noise removal, and data normalization.
[0127] Among them, outlier detection can use statistical methods or machine learning algorithms to detect and process abnormal data points to ensure data quality; noise elimination can apply filtering techniques (such as moving average, median filtering) or signal processing methods to reduce the impact of measurement noise on feature extraction; data normalization can unify data of different dimensions to the same scale, such as linear normalization or standardization, to ensure that the feature extraction process is not affected by the data magnitude.
[0128] Through this step, the data quality can be improved, the misjudgment caused by data anomalies or noise can be reduced, and the accuracy and stability of feature extraction can be ensured.
[0129] S230, extracting a feature quantity reflecting the state of the device from the preprocessed operating parameters;
[0130] The feature quantities include time domain features and frequency domain features. The time domain features include: one or more of mean, peak-to-peak value, root mean square, waveform factor, and peak factor. The frequency domain features include: one or more of spectrum center, frequency band energy, and characteristic frequency amplitude.
[0131] This step can provide a deeper understanding of the equipment's operating status, where the feature quantities can effectively capture key information about equipment status changes, helping to predict equipment failures or optimize equipment performance.
[0132] The feature quantity extraction method disclosed in this embodiment is based on the feature quantities extracted from real-time data, can promptly detect equipment operation abnormalities or trend changes, and support predictive maintenance and fault warnings; it can provide objective and quantitative equipment status information to help operation and maintenance personnel formulate more effective maintenance strategies and operational decisions; by accurately identifying equipment status characteristics, it can reduce unnecessary maintenance operations, optimize the allocation of maintenance resources, and reduce downtime losses caused by equipment failures.
[0133] The method of extracting equipment characteristics based on real-time operating parameters can not only enhance the monitoring capability of equipment operating status, but also effectively improve equipment reliability and operating efficiency. It is one of the important tools for modern industrial equipment management.
[0134] Taking the vibration monitoring of a hydro-turbine generator set as an example, assuming that a vibration time series {x(n), n=0, 1, ..., N-1} with a length of N=1000 is collected at a certain measuring point, and the sampling frequency fs=1000 Hz, the following feature quantities can be extracted.
[0135] Specifically, the mean (Mean) is μ, which reflects the average amplitude of the signal and represents the basic level of vibration;
[0136]
[0137] The root mean square value is RMS, which reflects the energy of the signal and is often used to characterize the overall intensity of the vibration;
[0138]
[0139] The peak-to-peak value is PPV, which reflects the dynamic range of the signal and represents the maximum amplitude of vibration; PPV = max[x(n)] - min[x(n)], 0≤n<N.
[0140] The waveform factor is WF, which reflects the complexity of the signal waveform and is sensitive to pulse-type vibration;
[0141]
[0142] The peak factor is CF, which reflects the impact characteristics of the signal and can be used for early fault symptom diagnosis; Among them, 0≤n<N.
[0143] Example: For a certain section of vibration data collected, the following time domain characteristics are calculated: mean μ = 0.02g, root mean square value RMS = 0.28g, peak-to-peak value PPV = 1.52g, waveform factor WF = 1.38, and peak factor CF = 3.52.
[0144] The frequency domain feature is obtained by fast Fourier transform (FFT) of the time domain signal x(n) to obtain its spectrum X(k), k=0, 1, ..., N-1, which is used to reflect the main frequency components of the signal and is related to the excitation frequency of the vibration source.
[0145] Spectral entropy reflects the uniformity of the spectrum and can characterize the chaos of vibration.
[0146] The Band Energy Ratio reflects the relative strength of different frequency components and can indicate the type of local fault.
[0147] Specifically, set several frequency bands of interest, such as f1~f2 corresponding to k1~k2, and calculate the energy ratio BER in the frequency band12 ,
[0148] In the first embodiment, referring to Figure 4 , the method for obtaining the operation prediction information of various types of equipment specifically includes the following steps:
[0149] A100, preprocessing the actual parameter information corresponding to the current running state collected, so that the preprocessed actual parameter information is consistent with the numerical range of the processed sample.
[0150] Specifically, data normalization or standardization can be performed to convert actual parameter information into a numerical range consistent with the sample library data, avoid errors caused by different data magnitudes or units, ensure the consistency and accuracy of subsequent processing, and improve the accuracy of the prediction model.
[0151] A200 extracts a number of actual time domain features and a number of actual frequency domain features from the preprocessed actual parameter information.
[0152] A300, uses sample data to train the constructed SVM classification model; the sample data includes the first sample set data and the second sample set data, the first sample set data is N steady-state vibration data samples selected from the normal operating conditions of the unit within one year, and the second sample set data is M vibration data under various fault conditions, N≥10M.
[0153] In this step, the SVM model can effectively process high-dimensional data and has good generalization ability.
[0154] A400, analyzes the feature quantity based on the trained SVM classification model and the target sample library to obtain the operation prediction information of the corresponding equipment.
[0155] In this step, the feature quantities extracted from the actual operating data are compared with the standard range and control limits in the target sample library; based on the comparison results, the current operating status of the equipment is judged and predictive information is provided, such as whether it is in a normal state, whether there is a potential failure, etc. It can provide instant equipment status evaluation and prediction information, support real-time decision-making and operation, and based on data-driven prediction capabilities, help prevent equipment failures in advance and optimize maintenance plans.
[0156] The method for obtaining operation prediction information of various types of equipment disclosed in this embodiment can realize real-time monitoring and accurate prediction of equipment operation status based on feature extraction and comparison of actual operation data; can identify potential equipment problems in advance, support preventive maintenance and fault warning, and reduce downtime risks and maintenance costs; can provide objective and quantitative equipment status information to help decision makers make data-based operation and maintenance strategies; and can continuously optimize models and sample libraries based on feedback data to improve prediction accuracy and system reliability.
[0157] Specifically, if the operation prediction information is abnormal, the current operation status is compared with the historical fault cycle information to obtain the potential fault type and trigger an alarm.
[0158] In the first embodiment, referring to Figure 5 The method for obtaining the equipment health status evaluation result specifically includes the following steps:
[0159] S410, constructing an analysis model.
[0160] Specifically, a suitable machine learning or statistical analysis model may be selected, such as a regression model, a decision tree model, a neural network or an integrated learning model.
[0161] In this step, the appropriate algorithm and model structure can be determined based on the device characteristics and data features.
[0162] S420, select a number of samples from the target sample library as a training set, each sample including a number of feature variables and a corresponding state label.
[0163] In this step, representative samples are selected from the target sample library to ensure coverage of various equipment states and working conditions. Each sample contains a characteristic variable and a corresponding state label (such as normal or faulty).
[0164] Establishing a training set can reflect the actual equipment operation status and historical failure conditions, improve the generalization ability of the model, and enable it to adapt to various practical scenarios.
[0165] S430, normalizing all feature variables, and using the processed training set to train the analysis model.
[0166] Through normalization, feature variables can be unified into the same numerical range, such as linear normalization or standardization; the constructed analysis model is trained using the normalized training set data. Normalization ensures that the contribution of each feature to model training is relatively balanced, and the trained model can more accurately capture the relationship between feature variables and the changing rules of device status.
[0167] S440, preprocessing the collected actual parameter information corresponding to the current running state so that the preprocessed actual parameter information is consistent with the numerical range of the processed sample.
[0168] The actual parameter information is normalized in the same way as the training set to ensure data range consistency and model applicability.
[0169] This step ensures that the data used in the evaluation process is consistent with the data processing method during model training, which can improve the accuracy and stability of the evaluation.
[0170] S450, inputting the pre-processed actual parameter information, operation prediction information, and historical failure cycle information into the trained analysis model to obtain the equipment health status assessment result.
[0171] This step can provide a comprehensive equipment health status assessment, combining real-time data and historical information for comprehensive analysis, supporting operation and maintenance personnel to make timely maintenance decisions and avoid equipment failures and downtime.
[0172] The method for obtaining the equipment health status assessment results disclosed in this embodiment, combined with real-time data, prediction information and historical fault data, can comprehensively assess the current status and future trends of the equipment, and based on the data-driven analysis model, can detect potential equipment failures in advance and perform predictive maintenance; optimize maintenance resource allocation based on the assessment results to reduce maintenance costs and downtime; continuously optimize the model and sample library based on feedback data to improve the accuracy of the assessment and the reliability of the system.
[0173] In this embodiment, the historical trend and current status of the characteristic parameters are comprehensively analyzed to quantitatively evaluate the health level of the equipment and determine whether the equipment needs maintenance. The expert knowledge base and the equipment maintenance history data are combined to infer the cause of the failure, give maintenance suggestions, and guide maintenance. The health status and maintenance information of the equipment can then be fed back to the equipment management system to optimize the management of spare parts and arrange maintenance plans reasonably.
[0174] Specifically, a device state recognition model based on SVM can be constructed. Support Vector Machine (SVM) is a commonly used machine learning algorithm, which is particularly suitable for small sample and nonlinear binary classification problems. Its basic idea is to find an optimal classification hyperplane in the feature space so that samples of different categories can be separated by the maximum interval.
[0175] For equipment status recognition, the healthy state and fault state can be regarded as two categories, the multidimensional feature quantities collected from the equipment can be regarded as sample points, and the SVM can be used to construct the optimal state discrimination model in the feature space.
[0176] The SVM model is trained using historical data in the healthy sample library. The main steps are as follows: a certain number of samples are selected from the healthy sample library as the training set, each sample contains m feature variables and corresponding state labels, the healthy state label is +1, and the fault state label is -1; all feature variables are normalized so that their numerical range is scaled to [0, 1] or [-1, +1] to avoid the influence of features of different dimensions.
[0177] For the linearly separable case, the goal of SVM is to find an optimal hyperplane so that the two types of samples can be separated by the hyperplane with the maximum interval.
[0178] For the case of linear inseparability, slack variables and penalty factors can be introduced to allow a small number of samples to be misclassified, while the cost of misclassification can be controlled by the penalty factor.
[0179] The above problem can be solved by solving a convex quadratic programming problem to obtain the optimal solution, and the obtained optimal hyperplane parameters can be used to construct the decision function of SVM.
[0180] For nonlinear classification problems, the kernel technique can be used to map samples from the original space to a high-dimensional feature space, making them linearly separable.
[0181] Commonly used kernel functions include polynomial kernel and Gaussian kernel (RBF kernel).
[0182] Using the training set data, the optimal parameters of the model are obtained by solving the convex quadratic programming problem of SVM.
[0183] For nonlinear SVM, it is necessary to select the parameters of the kernel function. The commonly used method is cross-validation + grid search to select the parameter combination with the best average performance.
[0184] During the training process, it is also necessary to balance the model's fitting ability and generalization performance, and control the complexity of the model by adjusting the penalty factor or kernel function parameters to avoid overfitting.
[0185] Once the SVM state recognition model is trained, it can be used to perform real-time diagnosis on the online monitoring data of the equipment.
[0186] For the collected monitoring signals such as equipment vibration, multi-dimensional features are extracted in the same way as in the training phase. The features are normalized and aligned with the numerical range of the training samples.
[0187] Substitute the preprocessed feature vector into the decision function of the SVM model to obtain the judgment result of the current state of the device (+1 or -1).
[0188] Combined with the operating conditions of the equipment, the status judgment result is further interpreted: if it is +1, the equipment is operating normally; if it is -1, the equipment may have a fault or performance degradation.
[0189] For equipment monitoring data identified as faulty or abnormal, its feature vector is compared with the support vector of SVM to find the most similar fault mode and preliminarily infer the fault type and cause.
[0190] With the help of expert knowledge base and historical equipment operation data, the preliminary diagnosis results are verified and corrected, and the possible fault location, severity and development trend are given.
[0191] If the same equipment is identified as a fault state multiple times in a row, or the degree of the fault continues to increase, corresponding maintenance decisions need to be made, such as adjusting operating parameters, arranging maintenance plans, etc., to eliminate equipment hazards in a timely manner.
[0192] Record each fault diagnosis result in the equipment health file to improve the knowledge base and provide reference for subsequent equipment management.
[0193] Take the vibration monitoring of generator set bearings as an example. A power plant installed a vibration sensor on the steam end bearing of the steam turbine generator, and collected a set of bearing vibration data every 10 minutes, including time series signals such as axial displacement, longitudinal and lateral speed.
[0194] 500 groups of normal state data were selected from historical data as healthy samples, and 80% of them were randomly selected as training sets and 20% as test sets. At the same time, 100 groups of bearing early failure data were obtained through artificial accelerated life tests as abnormal samples, of which minor wear and minor looseness accounted for 80%, and serious wear failures accounted for 20%.
[0195] Feature extraction is performed on each set of data to obtain a 10-dimensional feature vector, including time domain features (mean axial displacement, peak-to-peak axial displacement, root mean square value of longitudinal velocity, longitudinal velocity peak factor, root mean square value of lateral velocity, and lateral velocity peak factor) and frequency domain features (spectral amplitude at the bearing characteristic frequency, and frequency band energy near the bearing characteristic frequency).
[0196] Using the Gaussian kernel function, the optimal model parameters were selected through cross-validation, and the SVM state recognition model was trained. The diagnostic accuracy on the test set was 96%, of which the recognition accuracy of healthy samples was 100% and the recognition accuracy of fault samples was 85%, which can well distinguish between minor abnormalities and serious faults.
[0197] The SVM model was embedded in the online monitoring system of the steam turbine unit to perform real-time diagnosis on the newly collected bearing vibration data. When the bearing status was found to be abnormal for five consecutive times, the system automatically issued an early warning and diagnosed the initial cause as abnormal increase in bearing clearance. After on-site inspection, the bearing cover bolts were loose, and the vibration characteristics returned to normal after tightening. If the system continues to identify it as a fault state, it may indicate that the bearing has been worn and needs to be replaced in time.
[0198] In summary, the intelligent identification method of equipment status based on SVM can automatically determine the health status of equipment by extracting and classifying feature quantities, detect the degradation trend of equipment performance early, provide a basis for condition inspection and predictive maintenance decisions, and effectively improve equipment safety and reliability. However, its performance depends largely on the quality and quantity of training samples. In practical applications, it is necessary to continuously optimize the feature index system and sample library to improve the generalization ability and robustness of the model. In the future, with the continuous accumulation of data and the further development of machine learning theory, intelligent diagnosis technologies such as SVM will surely be more widely used in the field of condition monitoring and fault diagnosis.
[0199] Furthermore, the present application also includes: a sensor detection unit matching each type of hydropower plant equipment is provided, and a plurality of the sensor detection units constitute a sensor monitoring network.
[0200] Specifically, in this application, industrial Ethernet and bus technologies are used to build a sensor monitoring network covering the core equipment of the unit; a massive monitoring data collection and management platform is built to provide data cleaning, storage, retrieval and other functions; and an equipment intelligent monitoring App can also be developed to achieve data visualization, health status evaluation, intelligent alarm push, etc. Cloud computing technology is used to implement machine learning algorithms in a distributed manner to improve data processing efficiency.
[0201] In the second embodiment, referring to Figure 6 , the method for obtaining the operation prediction information of various types of equipment specifically includes the following steps:
[0202] B100, takes the time series consisting of several real-time operating parameters collected as observation variables to establish a state space model.
[0203] Among them, the state space model includes state equations and observation equations.
[0204] A time series is constructed using several real-time operating parameters collected as observation variables to establish a state space model to describe the dynamic changes in the equipment operating status.
[0205] This step can provide a mathematical description of the dynamic changes in the equipment's operating state and can effectively capture the relationship between the equipment's internal state and external observation data.
[0206] In this embodiment, the state space model established is a two-dimensional state space model, which is a commonly used time series modeling method, and is particularly suitable for describing and predicting multivariable, non-stationary complex time series. The core idea is to regard the time series as a dynamic system, use a few intrinsic state variables to characterize the evolution law of the system, and establish a mapping relationship between the state variables and the observed variables.
[0207] B200, using historical monitoring data, identifies unknown parameters in the state space model through the expectation maximization algorithm.
[0208] Using historical monitoring data, the expectation maximization (EM) algorithm is used to identify the unknown parameters in the state space model, where the parameter identification can include determining the state transfer matrix, observation matrix, process noise, and covariance of observation noise.
[0209] This step can improve the accuracy and adaptability of the state-space model, enabling it to better reflect the actual operating characteristics of the equipment. The parameter identification capability based on historical data contributes to the stability and reliability of the model.
[0210] B300, uses the identified state space model to make one-step-ahead predictions of feature quantities.
[0211] By using the identified state-space model to predict the equipment status at the current moment and obtain the state value predicted one step ago, we can provide an estimate of the future state of the equipment and provide a benchmark for the residual calculation of the subsequent steps. Through the predictive ability of the state-space model, real-time tracking and early warning of the equipment operating status can be achieved.
[0212] B400, calculates a number of residuals between the newly collected data of the device within a preset period and the one-step-ahead prediction;
[0213] Get the mean or variance of several residuals;
[0214] If the mean or variance exceeds the corresponding preset threshold, it is determined that the corresponding equipment is abnormal and operation prediction information is obtained; the operation prediction information includes the abnormal information of the current component and the potential abnormal information of other associated components.
[0215] In this step, the residual between the newly collected data within the preset period and the prediction one step ago is calculated, such as the mean or variance of the residual, and a preset threshold is set. If the residual mean or variance exceeds the threshold, it is determined that the equipment is abnormal. Through residual analysis, changes and abnormalities in the operating status of the equipment can be effectively detected, and targeted abnormal warning information can be provided to help operation and maintenance personnel take timely measures to prevent equipment failure or damage.
[0216] The method for obtaining operation prediction information of various types of equipment disclosed in this embodiment realizes real-time monitoring and prediction of the equipment operation status based on the prediction ability of the state space model; through residual analysis, potential anomalies in equipment operation can be accurately identified, including associated anomalies of a single component and multiple components; early detection of equipment anomalies helps to reduce downtime risks and maintenance costs, and supports the formulation and implementation of predictive maintenance strategies; combined with operation prediction information, it can comprehensively evaluate the health status of the equipment, optimize the use of equipment resources and the formulation of maintenance plans.
[0217] The method of comparing the feature quantity with the target sample library, by combining the state space model and residual analysis technology, can provide efficient and accurate tools and methods for equipment management and maintenance, and is suitable for operating status monitoring and prediction of various industrial equipment.
[0218] The two-dimensional state space model includes state space equations, and the state space equations include:
[0219] x t =Fx t-1 +Bu t +w t ,y t =Hx t +v t Among them, y t is the observed variable, x t is the hidden state variable, u t is the control input, w t and v t are process noise and observation noise respectively, which obey Gaussian distribution, F is the state transfer matrix, B is the control input matrix, H is the observation matrix, x t The dimension is smaller than y t Dimension.
[0220] Furthermore, the hidden state variable x t The dimension is usually much smaller than the observed variable y t , can be regarded as reducing the dimensionality of multivariate time series and extracting its inherent dynamic features. These features are often closely related to the physical mechanism and failure mode of the equipment, and contain the essential information of the equipment status.
[0221] Using the historical monitoring data of the equipment, the unknown parameters F, B, H and the noise covariance matrix in the state space model are estimated through the expectation maximization algorithm (EM algorithm). The EM algorithm maximizes the likelihood function of the observed data by iteratively solving the following two steps:
[0222] (1) Step E: Calculate the posterior probability distribution of the hidden state variables under the current parameter estimation;
[0223] (2) Step M: Use the posterior probability of the hidden state variables to update the model parameter estimates to maximize the likelihood function.
[0224] The EM algorithm is an unsupervised learning method that does not require prior labeling of equipment status and is very suitable for situations where sample labels are missing in industrial scenarios.
[0225] Example: A centrifugal compressor has temperature, pressure, flow and other sensors installed at its inlet and outlet, and collects data once a minute to form a set of multivariate time series. Now use these monitoring data, through the two-dimensional state space model and EM algorithm, to identify the internal working state of the compressor and realize its health monitoring.
[0226] The specific steps are as follows: 1) Data preprocessing: The collected raw sensor data is preprocessed by denoising and normalizing to remove obvious outliers and unify different physical quantities into a similar numerical range. 2) Construction of the state space model: Select the four monitoring quantities of the compressor's inlet temperature, inlet pressure, outlet temperature, and outlet pressure as observation variables to construct a two-dimensional state space model. Assume that the hidden state variable is 2-dimensional, reflecting the internal working conditions of the compressor (such as compression efficiency, gas dynamic parameters, etc.). The state equation describes how the hidden state evolves over time, and the observation equation describes how the hidden state affects the observed quantity. The model also includes process noise and observation noise, which reflect the random disturbance of the system and the measurement error of the sensor, respectively. 3) Initialization of the EM algorithm: Set initial values for the unknown parameters of the model (state transfer matrix, observation matrix, noise covariance, etc.). It is usually selected based on experience or prior knowledge, and can also be obtained using some heuristic methods (such as K-means clustering).
[0227] 4) Step E: Under the current parameter estimation, use the forward-backward recursive algorithm to calculate the posterior probability distribution of the hidden state variables at each time. Forward recursion is to predict the state distribution at time t based on the state estimate at time t-1 and the observed data at time t; backward recursion is to correct the state distribution at time t based on the state estimate at time t+1 and the observed data at time t. Combining the two, a smooth estimate of the hidden state can be obtained.
[0228] 5) Step M: Use the hidden state posterior probability obtained in step E to update the parameter estimation of the state space model. Specifically, replace the hidden state values in the estimation formulas of the state transfer matrix, observation matrix, and noise covariance with their posterior means to obtain new parameter values so that the likelihood function of the entire observation sequence is maximized.
[0229] 6) Repeated iteration: Alternately execute the E step and the M step until the model parameters converge or the preset maximum number of iterations is reached. 7) Model evaluation: Apply the identified state space model to the monitoring data of the compressor, calculate the one-step-ahead prediction error and the state estimation error, and evaluate the model fitting effect. You can also use methods such as cross-validation to evaluate the model's generalization ability to new data.
[0230] Through the EM algorithm, we identified the dynamic model of the compressor's internal working state from the multivariate monitoring data, and obtained key parameters such as state transition rules, observation mechanism, and noise characteristics. These parameters describe the physical process inside the compressor, reveal the internal mechanism of its performance degradation, and lay the foundation for subsequent fault diagnosis and health prediction.
[0231] In practical applications, we found that the identified hidden state variables are closely related to the working conditions of the compressor: one of the state variables is positively correlated with the efficiency of the compressor. When the mean of the state decreases abnormally, it often indicates that the performance of the compressor is degraded; the other state variable is related to the dynamic parameters of the gas. The increase in its variance often indicates that the airflow has turbulence or instability, which may cause the compressor to vibrate or surge. Therefore, by tracking the evolution of the state variables, abnormal signs of the compressor can be discovered in time, and its predictive maintenance can be achieved.
[0232] In addition, we also found that the statistical characteristics of the residuals of the state space model identified by the EM algorithm are more sensitive to the type of compressor fault: when periodic residuals of certain specific frequencies appear, it often indicates that the rotor or bearing of the compressor is unbalanced or loose; while residuals with strong randomness may indicate impurities or cavitation in the gas path. Therefore, spectral analysis and statistical tests on the residual signal can further refine the diagnosis of compressor faults.
[0233] It should be noted that the EM algorithm is an iterative optimization method, and its results depend to some extent on the selection of initial values. In practice, we usually try multiple sets of different initial values and select the solution with the largest likelihood function as the final result. In addition, the EM algorithm sometimes converges to a local extreme point. For this reason, we can introduce some global optimization strategies (such as simulated annealing) to improve the robustness of the algorithm.
[0234] In summary, the state space model identification based on the EM algorithm provides a novel and effective method for compressor health monitoring. By mining the inherent dynamic laws of multivariate monitoring data, this method can adaptively track the evolution of equipment status and detect potential anomalies and failures early, thereby achieving predictive maintenance and improving the reliability and energy efficiency of the compressor; this method can also be extended to other types of rotating machinery and industrial equipment, and has broad application prospects.
[0235] For online equipment status monitoring and fault diagnosis, it specifically includes: using the identified state space model to make a one-step-ahead prediction of the newly collected equipment monitoring data
[0236] in, is the number of observations y at time t-1 1:t-1 Estimated state variables at time t.
[0237] Calculate the new observation data y t One-step-ahead prediction The residual between t :
[0238] If the residual exceeds the preset threshold, it is determined that the device status is abnormal.
[0239] Update the state estimate at time t Among them, K t is the Kalman gain, which reflects the degree of correction of the state estimation by the new information.
[0240] By abnormal residual e t Reflects the dynamic characteristics, combined with state variables The inherent laws contained therein can be used to make a preliminary diagnosis of the type and severity of equipment failure.
[0241] Example: In a large hydropower station, the main shaft of the generator set is monitored online. Multiple vibration and displacement sensors, as well as some temperature, pressure, strain and other sensors are arranged on the main shaft. The machine learning algorithm based on the state space model is used to conduct real-time evaluation of the health status of the main shaft and fault warning.
[0242] The specific steps are as follows: 1) Data collection and preprocessing: When the generator set is running, the sensor data of each monitoring point of the main shaft is collected every 10 minutes and transmitted to the server in the central control room. The collected data is preprocessed by filtering, synchronization, and standardization to eliminate the influence of noise and outliers. 2) State estimation and anomaly detection: The newly collected monitoring data is processed in real time using the state space model trained offline.
[0243] Based on the state estimation at the previous moment and the observed data at the current moment, the Kalman filter algorithm is used to recursively estimate the health status of the spindle at the current moment, including key indicators such as bearing vibration, axial displacement, and temperature rise.
[0244] The residuals between the estimated state and the actual observed value are calculated, and the residual sequence is statistically analyzed; when the mean or variance of the residual exceeds the preset control limit, an abnormal alarm is issued, indicating that the state of the spindle may have shifted.
[0245] 3) Fault diagnosis and cause analysis: When the spindle status is abnormal, the fault diagnosis program is automatically triggered.
[0246] Compare the state parameters at the time of abnormality with historical fault cases to preliminarily determine the type and location of the fault. For example, if the vibration amplitude and spectrum of a bearing are similar to typical bearing wear faults, the bearing may be worn.
[0247] Analyze the changing trend of the state parameters before and after the fault, and infer the development stage and severity of the fault. For example, if the vibration level of the bearing has continued to rise in the past month, and the rising speed is accelerating, the wear of the bearing may have entered the accelerated period and needs to be replaced as soon as possible. Combined with the background information such as the unit's operating conditions and maintenance records, analyze the possible causes of the fault. For example, bearing wear may be caused by long-term overload operation, poor lubrication, seal failure, etc.
[0248] 4) Maintenance decision and subsequent processing: According to the results of fault diagnosis, timely formulate maintenance strategies and optimize operation and maintenance management.
[0249] For faults that seriously affect safety or endurance, an order to shut down and inspect will be issued immediately, and the power grid dispatching center will be notified to adjust the power production plan.
[0250] For moderate failures, assess their impact on equipment life and efficiency, arrange maintenance time reasonably, and complete maintenance as much as possible without affecting power generation tasks.
[0251] For early minor faults, pay close attention to their development trends, optimize operating parameters (such as reducing load, increasing cooling, etc.), slow down their deterioration rate, prepare spare parts and personnel, and activate emergency plans at any time.
[0252] Fault diagnosis reports and maintenance plans are archived in the electronic health record of the equipment for reference in subsequent reliability analysis and maintenance decision-making.
[0253] In the actual application of this hydropower station, we found that the online monitoring method based on the state space model is very sensitive to abnormal vibration caused by rotor imbalance. In one case, the system warned of an imbalance fault in Unit 2 a week in advance. After on-site verification, it was found that the rotation center of the rotor deviated by 0.2 mm. Although it was not beyond the standard, it had affected the safety margin of the unit. By carrying out dynamic balancing correction in a timely manner, more serious accidents were avoided.
[0254] In addition, many problems such as abnormal guide vane opening and lubrication system blockage were successfully diagnosed, providing reliable guarantees for the safe and economical operation of the generator set.
[0255] In short, online equipment monitoring technology with state space model as the core has opened up a new path for intelligent operation and maintenance of hydropower stations. In the future, with the further development of technologies such as the Internet of Things and big data, and the deep integration of mechanism modeling and data-driven methods, this technology will play an increasingly important role in the safe, economical and environmentally friendly operation of hydropower systems. At the same time, its application experience can also provide useful reference and inspiration for the health management of other types of power stations and large equipment.
[0256] This application also includes online learning and updating of the state space model. In order to adapt to the long-term evolution of the equipment state, the state space model needs to be learned and updated online. The model parameters can be re-identified regularly using the newly collected monitoring data. During the online learning process, a forgetting factor can be introduced to make the model parameter estimation focus on the latest data, gradually forget the past state, and improve the algorithm's adaptive ability.
[0257] In summary, the two-dimensional state space model is used to dynamically model the equipment monitoring data, the EM algorithm is used to realize unsupervised learning of model parameters, and online state estimation and fault diagnosis are performed based on methods such as Kalman filtering to form a complete set of equipment intelligent monitoring solutions. This solution can fully explore the inherent correlation and dynamic laws of multivariate time series, detect abnormal signs of equipment status early, and warn and diagnose potential faults, thereby realizing health management of the equipment throughout its life cycle and improving its reliability and economy.
[0258] Compared with the traditional equipment monitoring method based on static data and shallow learning, this scheme has the following advantages: (1) It takes into account the dynamic correlation between multiple monitoring indicators, explores the inherent laws of time series, and is more in line with the physical mechanism of equipment degradation; (2) It adopts an unsupervised learning framework and does not rely on a large amount of labeled data, which reduces the burden of data labeling in engineering applications; (3) It introduces an online learning mechanism that can adaptively track changes in equipment status and improve the robustness of the algorithm.
[0259] In the second embodiment, referring to Figure 7 The method for obtaining the equipment health status evaluation result specifically includes the following steps:
[0260] C100, based on operation prediction information and historical fault cycle information, preliminarily determines the fault type and fault location.
[0261] Specifically, based on the operation prediction information and historical fault cycle information, the operating status of the equipment and the possible fault types are preliminarily analyzed; using the fault prediction model or rule engine, the real-time operating parameters are compared with the preset fault mode to preliminarily determine the possible fault type; the possible fault location can be inferred based on the fault mode and equipment structure.
[0262] This step provides a quick assessment of the current status of the equipment, helping to quickly locate potential faults; it can provide early warning before a fault occurs to prevent further damage or downtime of the equipment.
[0263] C200 obtains the fault change trend based on the normal operating parameter information of the fault location and the operating parameter information when the fault exists.
[0264] Based on the normal operating parameter information of the fault location and the operating parameter information when the fault exists, the changing trend of the fault parameters is analyzed. Time series analysis or trend monitoring technology can be used to identify the changing pattern and rate of the fault parameters.
[0265] This step provides a detailed understanding of the fault development process, including the rate and pattern of fault change, which helps to assess the severity of the fault and the impact on the health of the equipment.
[0266] C300 obtains equipment health status assessment results based on fault change trends; equipment health status assessment results include fault development stage information, fault severity information, and fault trend change information.
[0267] Based on the fault change trend and prediction information, the health status of the equipment is comprehensively assessed, which may include the fault development stage information (such as early, middle, late), fault severity information (such as minor, serious) and fault trend change information (such as aggravation, stability, relief).
[0268] This step provides a comprehensive equipment health status assessment that is not limited to the current fault status, but also takes into account the development trend of the fault, supporting operation and maintenance personnel to formulate more accurate maintenance strategies and decisions, and optimize equipment maintenance cycles and resource allocation.
[0269] The method for obtaining the equipment health status assessment results disclosed in this embodiment, combined with operation prediction information and historical failure cycle information, can quickly and accurately diagnose the type and location of equipment failures; by analyzing the failure change trend, it can deeply understand the development process of the failure and the impact on equipment performance; based on the assessment results, timely warn of potential failures, support the implementation of predictive maintenance and failure prevention strategies; based on the comprehensive assessment results, optimize maintenance plans and resource allocation, and improve equipment reliability and operating efficiency.
[0270] In summary, equipment health status assessment based on operation prediction information and historical failure cycle information provides efficient and accurate analysis tools and methods for modern industrial equipment management and maintenance, which is suitable for health status monitoring and fault management of various types of equipment.
[0271] In the second embodiment, referring to Figure 8 , the method for obtaining the maintenance strategy specifically includes the following steps:
[0272] D100, matches the equipment health status assessment result with the expert knowledge base to obtain a first matching degree.
[0273] In this step, the results of the equipment health status assessment are used as input and matched with various health status patterns preset in the expert knowledge base to determine the similarity or match between the assessment results and the patterns in the knowledge base. Specifically, based on the matching of the expert knowledge base, it is possible to quickly identify whether the current state of the equipment conforms to the known health patterns, and it can also provide the first level of verification and reference for the assessment results, which is the basis for formulating subsequent maintenance strategies.
[0274] D200, matching the equipment health status assessment result with the equipment maintenance history data information to obtain a second matching degree;
[0275] If the second matching degree and the first matching degree are both greater than the matching threshold, the first maintenance strategy is obtained from the expert knowledge base, and the second maintenance strategy is obtained from the equipment maintenance history data information;
[0276] Generate a maintenance strategy based on the first maintenance strategy and the second maintenance strategy;
[0277] If the second matching degree and the first matching degree are both less than the matching threshold, the transition strategy is triggered to execute;
[0278] If the second matching degree is greater than the matching threshold and the first matching degree is less than the matching threshold, the maintenance strategy obtained from the equipment maintenance history data information is used as the maintenance strategy;
[0279] If the second matching degree is less than the matching threshold and the first matching degree is greater than the matching threshold, the inspection strategy obtained from the expert knowledge base is used as the maintenance strategy.
[0280] In this step, the equipment health status assessment results are compared with historical maintenance data information, the correspondence between the current assessment results and similar situations in the past is analyzed, and the similarity or matching degree between the assessment results and the historical data is calculated. Based on the matching of the equipment maintenance history data, it is possible to more accurately understand whether similar problems have ever occurred in the current state of the equipment, which provides a second level of verification and reference for the assessment results, and helps to judge the severity and urgency of the current problem.
[0281] Specifically, 1) Matching threshold setting: setting a threshold for judging the degree of matching, such as a percentage of similarity or matching. 2) Generate strategy rules: Scenario 1: If both the second degree of matching and the first degree of matching are greater than the matching threshold, the first maintenance strategy is obtained from the expert knowledge base, and the second maintenance strategy is obtained from the equipment maintenance history data information, and then the maintenance strategy is generated. Scenario 2: If the second degree of matching is greater than the matching threshold and the first degree of matching is less than the matching threshold, the maintenance strategy obtained from the equipment maintenance history data information is used as the maintenance strategy. Scenario 3: If the second degree of matching is less than the matching threshold and the first degree of matching is greater than the matching threshold, the maintenance strategy obtained from the expert knowledge base is used as the maintenance strategy. Scenario 4: If both the second degree of matching and the first degree of matching are less than the matching threshold, the execution of the transition strategy is triggered, such as increasing the monitoring frequency or temporarily enhancing maintenance.
[0282] The method for obtaining maintenance strategies disclosed in this embodiment, combined with equipment health assessment results, expert knowledge and historical maintenance data, can intelligently generate targeted maintenance strategies; use historical data to verify current assessment results, thereby enhancing the reliability and practicality of decision-making; quickly determine appropriate maintenance measures based on matching judgments, which helps to quickly respond to equipment problems and reduce downtime; generate maintenance strategies based on specific circumstances, which can optimize resource allocation, reduce maintenance costs, and improve equipment reliability and efficiency.
[0283] In summary, the maintenance strategy determination method based on equipment health status assessment results, expert knowledge base and equipment maintenance history data information provides effective decision support and operation guidance for equipment operation and maintenance management, and is a key link in intelligent maintenance in modern industrial environments.
[0284] Among them, the transition strategy includes: triggering the monitoring system to monitor the overall operating status and various operating parameters of the corresponding equipment in real time;
[0285] If the operating status monitored within the preset period is normal and all operating parameters are within the range of the target sample library, the corresponding device is determined to be normal.
[0286] Specifically, the operating status of the equipment can be monitored in real time, triggering the monitoring system to monitor the overall operating status and various key operating parameters of the target equipment in real time; the operating parameters collected in real time are compared with the normal range in the target sample library; if the operating status of the equipment monitored within the preset period is normal and all operating parameters are within the reasonable range of the target sample library, the equipment is judged to be in normal operating state to prevent misjudgment.
[0287] The transition strategy disclosed in this embodiment can promptly detect any changes in the operating status of the equipment through real-time monitoring; automatically determine whether the equipment is normal based on a preset operating parameter range, reducing the time cost of manual intervention and judgment; and quickly detect problems with the equipment and take timely measures, helping to reduce the risk of production interruptions and equipment damage.
[0288] Through the implementation of the transition strategy, 1) Instant response and accurate judgment can be achieved: real-time monitoring enables operation and maintenance personnel to understand the equipment status in a timely manner, make quick judgments, and improve the accuracy and efficiency of fault diagnosis; 2) The risk of downtime can be reduced: timely detection of equipment abnormalities and judgments can help reduce the risk of unexpected downtime and production interruptions; 3) Automated operation and maintenance management can be achieved: through automated monitoring and judgment, the need for manual intervention is reduced, and the efficiency and reliability of operation and maintenance management are improved.
[0289] In summary, the application of transition strategy in equipment operation and maintenance management not only enhances the monitoring capability of equipment status changes, but also improves operation and maintenance efficiency and production safety.
[0290] Furthermore, the method for intelligently monitoring the operating status of hydropower plant equipment disclosed in the present application further includes: regularly collecting new monitoring data of the equipment through a set sensor monitoring network;
[0291] Based on the monitoring data collected at time t, the predicted value at time t+1 is predicted through the state space model;
[0292] Based on the predicted value at time t+1 and the measured data at time t+1, the prediction error is obtained;
[0293] The state equation parameters and observation equation parameters in the state space model are dynamically optimized based on the prediction error and the forgetting factor.
[0294] Specifically, a sensor monitoring network can be set up to regularly collect monitoring data of hydropower plant equipment, including various key parameters such as temperature, pressure, flow, etc.; ensure that the sensors are reasonably arranged to fully cover the key parts and operating status of the equipment.
[0295] The prediction based on the state space model includes: using the state space model to analyze and predict the collected monitoring data; at time t, using the state space model to predict the values of each monitoring parameter at time t+1.
[0296] The prediction error calculation and dynamic optimization parameters include: comparing the predicted value at time t+1 with the measured data to calculate the prediction error; dynamically optimizing the state equation parameters and observation equation parameters in the state space model based on the prediction error and the set forgetting factor.
[0297] By regularly collecting monitoring data and predicting based on the state space model, real-time monitoring and prediction of the equipment operating status can be achieved, improving the accuracy of fault warning; according to the prediction error and forgetting factor, the model parameters are dynamically adjusted so that the model can more accurately reflect the actual operating status changes of the equipment; by discovering the trend of equipment status changes in advance, preventive maintenance can be carried out to avoid losses to production caused by equipment failures.
[0298] The combination of sensor monitoring, state space models and dynamic optimization strategies makes the operation and maintenance management of hydropower plant equipment more intelligent and refined; the automated data collection and predictive optimization process reduces the need for manual operation and judgment, and improves the reliability and consistency of operations; through predictive maintenance and dynamic optimization parameters, it can reduce equipment maintenance costs and downtime, and improve equipment reliability and service life.
[0299] In summary, the predictive optimization method based on the state space model is of great significance in the intelligent monitoring of the operating status of hydropower plant equipment. It not only improves the accuracy and timeliness of equipment monitoring, but also provides a scientific basis for operation and maintenance decisions, which helps to achieve safe and stable operation of equipment and continuous production efficiency.
[0300] In order to better illustrate the implementation effect of the present invention, the present invention is described in detail below in conjunction with specific application examples.
[0301] Embodiment 1: Vibration status monitoring of a hydro-turbine generator set. 1) Arrange vibration sensors to collect vibration data of the generator structure in real time, with a sampling frequency of 1kHz. 2) Preprocess the collected vibration data: remove obvious wild points and outliers, use wavelet transform to reduce noise on the data, normalize it, and eliminate the influence of data of different dimensions. 3) Extract time-frequency domain features; time domain features: extract 12 feature quantities such as peak-to-peak value, root mean square value, waveform factor, and peak factor of the data; frequency domain features: perform FFT transformation on the data to extract 10 feature quantities such as predominant frequency and its amplitude, spectrum entropy, and frequency band energy ratio. 4) Construct an SVM (support vector machine) classification model based on historical data: select steady-state vibration data samples (a total of 1,000) of the unit under normal operating conditions for one year, and manually mark them as normal samples; select vibration data under fault conditions such as main shaft misalignment and imbalance (100 fault samples of each type), and manually mark them as abnormal samples; use sample data to train and optimize the SVM model to obtain a fault diagnosis model. 5) Online vibration state intelligent identification: extract vibration data feature in real time and input it into the trained SVM model; the SVM model determines the abnormal probability of the data and gives the health status of the monitored quantity; if it is determined to be abnormal, it is compared with the typical fault characteristics, the fault type is inferred, an alarm is issued and diagnostic suggestions are generated. 6) Application effect: The system was installed and applied in 67 units. During the one-year operation, three potential faults were successfully warned, avoiding serious unplanned downtime accidents; the diagnostic accuracy rate reached more than 95%, which was 50% higher than the efficiency of traditional manual inspections; the unplanned downtime of generators was reduced by 20%, and the average availability rate was increased by 1.5 percentage points.
[0302] Example 2: Analysis of turbine flow field and cavitation detection. 1) High-frequency pressure sensors are arranged at key locations of flow channels such as the turbine water inlet, volute, and tailwater pipe. 2) The collected pressure pulsation data is subjected to spectrum analysis, and the time-frequency characteristics of the pressure signal are extracted using STFT (short-time Fourier transform); the pressure pulsation PSD (power spectrum density) changes in the range of 0.2 to 5 times the blade passing frequency are focused on; and the pressure pulsation amplitude alarm thresholds in different frequency bands are set. 3) The CFD model of the internal flow field of the turbine is established using fluid mechanics simulation to obtain the theoretical characteristics of pressure pulsation under different working conditions. 4) The measured pressure pulsation spectrum is compared with the theoretical model to evaluate the flow field stability: the measured pressure pulsation spectrum components are basically consistent with the theoretical predictions, and the amplitude change does not exceed 10%, and the flow field is considered normal; abnormal frequency components that cannot be explained by the theoretical model appear, or the amplitude exceeds the alarm threshold, and it is judged as unstable flow; broadband noise and a sudden increase in amplitude appear in the tailwater pipe pressure pulsation, and severe cavitation is judged to have occurred. 5) Cavitation fault diagnosis and degree assessment: When cavitation occurs, the pressure pulsation amplitude at the fundamental frequency and its multiples will increase significantly, and the statistical increase can determine the degree of cavitation; combined with the material damage model of the cavitation site, the impact of cavitation on the efficiency and service life of the turbine is evaluated; the turbine start-up and shutdown and adjustment process are optimized to avoid long-term operation under high cavitation risk conditions. 6) Application effect: The quantitative analysis of the complex flow field inside the turbine is realized, revealing the cavitation risk of different operating conditions; by optimizing the scheduling strategy, the incidence of cavitation failures is reduced by 80%; the turbine maintenance cycle is reduced from once a year to once every three years, and the average annual maintenance cost is reduced by 700,000 yuan.
[0303] Example 3: Generator winding fault diagnosis and life prediction. 1) Install an optical fiber temperature measuring device at the end of the generator stator winding rod to monitor the winding temperature in real time. 2) Perform statistical analysis on the winding rod temperature: calculate the highest temperature, lowest temperature, and average temperature of 120 temperature measuring rods; find the "hot spot" area of temperature distribution and focus on monitoring; analyze the dynamic response characteristics of temperature changes with load. 3) Combine the generator vibration and power data to build a multivariate anomaly detection model: select 10 characteristic parameters such as winding temperature, casing vibration, and stator current; use the isolation forest algorithm to train the anomaly detection model to determine the degree of outliers in the data; outlier data appears frequently or with a large amplitude, and warns of potential winding faults. 4) Prediction of the remaining service life of the winding based on machine learning: segmentally count the daily average temperature rise of the winding rod to estimate the aging degree of the winding insulation material; combine historical maintenance records and use the GPR (Gaussian process regression) algorithm to establish a winding loss-operating time model; calculate the remaining service life of the winding under the current operating conditions and decide the best maintenance time. 5) Application effect: It realizes early warning of winding faults, discovers problems an average of 3 months in advance, and avoids operating accidents; optimizes maintenance strategies, and each generator can reduce maintenance downtime by 1 day per year, increasing power generation by 1.5 million kWh; extends the average service life of the windings, extending the maintenance cycle from 6 years to 8 years, saving 3 million yuan in maintenance costs.
[0304] In summary, the intelligent monitoring method for the operating status of hydropower plant equipment proposed in this application can effectively improve the safety and reliability of equipment, reduce maintenance costs, and improve power generation efficiency, which is of great significance for the refined management of hydropower plants; this technology can also be promoted and applied to thermal power, wind power, nuclear power and other fields, and has broad market prospects.
[0305] Reference Fig. 9 The second aspect of the present application discloses an intelligent monitoring system 10 for the operation status of hydropower plant equipment, comprising:
[0306] Construction module 11 is used to construct a target sample library of hydropower plant equipment; the target sample library includes standard ranges and control limit information of various characteristic parameters under different operating conditions of various types of equipment;
[0307] An extraction module 12, used for extracting characteristic quantities based on a plurality of real-time operating parameters of a plurality of devices collected;
[0308] The comparison module 13 is used to compare the characteristic quantity with the target sample library to obtain the operation prediction information of various types of equipment; the operation prediction information includes performance degradation trend information, stable operation prediction cycle information, and associated potential fault information;
[0309] An analysis module 14 is used to obtain equipment health status assessment results based on operation prediction information and historical failure cycle information;
[0310] The generating module 15 is used to determine the maintenance strategy based on the equipment health status assessment result, the expert knowledge base and the equipment maintenance history data information.
[0311] The computer device according to the embodiment of the present disclosure includes a memory and a processor. The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.
[0312] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the computer device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory, so that the computer device executes all or part of the steps of the aforementioned hydropower plant equipment operating status intelligent monitoring method of each embodiment of the present disclosure.
[0313] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.
[0314] like Fig.10 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure is shown, which is a schematic diagram of the structure of a computer device suitable for implementing the embodiment of the present disclosure. Fig.10 The computer device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.
[0315] like Fig.10 As shown, the computer device may include a processor (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage device into a random access memory (RAM). In the RAM, various programs and data required for the operation of the computer device are also stored. The processor, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.
[0316] Typically, the following devices can be connected to the I / O interface: input devices such as sensors or visual information acquisition devices; output devices such as display screens; storage devices such as tapes, hard disks, etc.; and communication devices. The communication device can allow the computer device to communicate with other devices (such as edge computing devices) wirelessly or by wire to exchange data. Fig.10 A computer device having various devices is shown, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.
[0317] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processor, all or part of the steps of the method for intelligently monitoring the operating status of hydropower plant equipment of an embodiment of the present disclosure are executed.
[0318] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.
[0319] According to the computer-readable storage medium of the embodiment of the present disclosure, non-transitory computer-readable instructions are stored thereon. When the non-transitory computer-readable instructions are executed by the processor, all or part of the steps of the method for intelligently monitoring the operating status of hydropower plant equipment in the above-mentioned embodiments of the present disclosure are executed.
[0320] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).
[0321] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.
[0322] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. According to the technical essence of the present invention, within the spirit and principles of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.
Claims
1. A method for intelligently monitoring the operating status of hydropower plant equipment, characterized in that: The following steps are involved: S100, building a target sample library of hydropower plant equipment; S200, collecting real-time operating parameters of the device, and then extracting feature quantities based on the real-time operating parameters; S300, comparing the characteristic quantity with the target sample library to obtain operation prediction information of the device; S400, obtaining equipment health status assessment results based on operation prediction information and historical failure cycle information; S500, determining a maintenance strategy based on the equipment health status assessment result, the expert knowledge base, and the equipment maintenance history data information.
2. The intelligent monitoring method for the operation status of hydropower plant equipment according to claim 1 is characterized in that: Step S100 includes the following steps: S110, determining key characteristic parameters of the device; S120, determining the operating condition type according to the operating characteristics of the equipment; S130, obtaining first historical data of key characteristic parameters under various operating conditions; S140, acquiring statistical characteristic information based on the first historical data; Wherein, the statistical characteristic information includes time domain characteristic information, frequency domain characteristic information and probability distribution characteristic information; The time domain characteristic information includes one or more of mean, standard deviation, maximum, minimum, median, and peak-to-peak value; The frequency domain characteristic information includes one or more of the dominant frequency, the amplitude of the dominant frequency, the spectrum entropy, and the frequency multiplication component; The probability distribution characteristic information includes one or more of distribution type, skewness, and kurtosis; S150, determining the standard range and control limit information of key characteristic parameters of the equipment under corresponding operating conditions based on the statistical characteristic information; The standard range and control limit information of key characteristic parameters of several types of equipment under different operating conditions is the target sample library.
3. The intelligent monitoring method for the operation status of hydropower plant equipment according to claim 1 is characterized in that: Step S200 includes the following steps: S210, collecting the operating parameters of the equipment in real time and storing them in time series; S220, preprocessing the operating parameters; S230, extracting a feature quantity reflecting the state of the device from the preprocessed operating parameters; Among them, the feature quantities include time domain features and frequency domain features; the time domain features include one or more of mean, peak-to-peak value, root mean square, waveform factor, and peak factor; the frequency domain features include one or more of spectrum center, frequency band energy, and characteristic frequency amplitude.
4. The intelligent monitoring method for the operation status of hydropower plant equipment according to claim 1 is characterized in that: Step S300 includes the following steps: A100, preprocessing the actual parameter information corresponding to the current running state collected, so that the actual parameter information after preprocessing is consistent with the numerical range of the processed sample; A200, extracts actual time domain features and actual frequency domain features from the preprocessed actual parameter information; A300, uses sample data to train the constructed SVM classification model; The sample data includes a first sample set data and a second sample set data, wherein the first sample set data is N steady-state vibration data samples selected from the normal operation of the unit within one year, and the second sample set data is M vibration data selected under various fault conditions, where N≥10M; A400, analyzes the feature quantity based on the trained SVM classification model and the target sample library to obtain the operation prediction information of the corresponding equipment.
5. The intelligent monitoring method for the operation status of hydropower plant equipment according to claim 1 is characterized in that: Step S300 includes the following steps: B100, taking the time series composed of the collected real-time operating parameters as the observed variables, and establishing the state space model; B200, using historical monitoring data to identify unknown parameters in the state space model through the expectation maximization algorithm; B300, uses the identified state space model to make one-step-ahead predictions of feature quantities; B400, calculating the residual between the newly collected data of the device within a preset period and the prediction one step ago; obtaining the mean or variance of the residual; if the mean or variance exceeds the corresponding preset threshold, it is determined that the corresponding device is abnormal, and the operation prediction information is obtained; The operation prediction information includes current component abnormality information and potential abnormality information of associated components.
6. The method for intelligent monitoring of the operating status of hydropower plant equipment according to claim 5, characterized in that: In step B100, new real-time operating parameters of the equipment are regularly collected through the set sensor monitoring network; Based on the real-time operating parameters collected at time t, the predicted value at time t+1 is predicted through the state space model; Based on the predicted value at time t+1 and the measured data at time t+1, the prediction error is obtained; Dynamically optimize the state space model based on prediction error and forgetting factor.
7. The intelligent monitoring method for the operating status of hydropower plant equipment according to any one of claims 1 to 6, characterized in that: Step S400 includes the following steps: S410, constructing an analysis model; S420, selecting a number of samples from the target sample library as a training set, each sample including a number of feature variables and a corresponding state label; S430, normalizing all feature variables, and using the processed training set to train the analysis model; S440, preprocessing the actual parameter information corresponding to the current running state collected, so that the preprocessed actual parameter information is consistent with the numerical range of the processed sample; S450, inputting the pre-processed actual parameter information, operation prediction information, and historical failure cycle information into the trained analysis model to obtain the equipment health status assessment result.
8. The method for intelligent monitoring of the operating status of hydropower plant equipment according to claim 1, characterized in that: Step S500 includes the following steps: C100, based on operation prediction information and historical fault cycle information, preliminarily determines the fault type and fault location; C200, based on the normal operating parameter information of the fault part and the operating parameter information when the fault exists, obtains the fault change trend; C300, based on the fault change trend, obtains the equipment health status assessment results; The equipment health status assessment result includes fault development stage information, fault severity information and fault trend change information.
9. The method for intelligent monitoring of the operating status of hydropower plant equipment according to claim 1, characterized in that: Step S500 includes the following steps: D100, matching the equipment health status assessment result with the expert knowledge base to obtain the first matching degree; D200, matching the equipment health status assessment result with the equipment maintenance history data information to obtain a second matching degree; If the second matching degree and the first matching degree are both greater than the matching threshold, the first maintenance strategy is obtained from the expert knowledge base, and the second maintenance strategy is obtained from the equipment maintenance history data information; Generate a maintenance strategy based on the first maintenance strategy and the second maintenance strategy; If the second matching degree and the first matching degree are both less than the matching threshold, the transition strategy is triggered to execute; If the second matching degree is greater than the matching threshold and the first matching degree is less than the matching threshold, the maintenance strategy obtained from the equipment maintenance history data information is used as the maintenance strategy; If the second matching degree is smaller than the matching threshold and the first matching degree is larger than the matching threshold, the inspection and repair strategy obtained from the expert knowledge base is used as the maintenance strategy.
10. The intelligent monitoring method for the operating status of hydropower plant equipment according to claim 9, characterized in that: The transition strategy includes: triggering the monitoring system to monitor the overall operating status and operating parameters of the corresponding equipment in real time; If the operating status monitored within the preset period is normal, and each operating parameter is within the range of the target sample library, then the corresponding device is determined to be normal.
Citation Information
Cited By
Fault processing method and system for distributed power supply
CN119323267A
A Fault Handling Method and System for Distributed Power Source
CN119323267B
Industrial park equipment operation state management method and system
CN119379272A
Unit non-stop supervision method, system and device based on model driving and storage medium
CN119397368A
Campus security equipment diagnosis system based on Internet of Things
CN119476726A
