Thermal power plant equipment management method and system based on data fusion

Abstract

The invention discloses a thermal power plant equipment management method and system based on data fusion, and relates to the technical field of equipment management, and the method comprises the steps: collecting the real-time data of equipment through a sensor, and evaluating the health state of the equipment through a data fusion algorithm; based on the real-time operation data of the equipment, the intelligent rule engine dynamically adjusts the evaluation standard; predicting a potential fault and generating a maintenance priority through an intelligent fault prediction algorithm by using historical data and a health state evaluation result; and according to fault prediction and an equipment state, scheduling maintenance resources are optimized, and timely maintenance and repair of key equipment are ensured. According to the thermal power plant equipment management method based on data fusion provided by the invention, comprehensive monitoring of the equipment operation state is realized, and the accuracy and comprehensiveness of data are remarkably improved by combining a multi-stage data fusion technology based on dynamic weighted average and principal component analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment management, and in particular to a thermal power plant equipment management method and system based on data fusion. Background Art

[0002] With the continuous improvement of industrial automation, thermal power plants, as important facilities for energy production, have attracted much attention for their equipment safety, reliability and economy. Thermal power plants have a wide variety of equipment, including core components such as generator sets, boilers, steam turbines and condensers. These equipment are prone to failures due to vibration, temperature rise, pressure fluctuations and other reasons during long-term operation. Traditional equipment management methods usually rely on periodic inspections and human experience for maintenance, but this method has obvious limitations.

[0003] Traditional equipment monitoring relies on data collection from a single type of sensor, lacks comprehensive perception of multi-dimensional operating status, and is difficult to accurately reflect the health status of the equipment.

[0004] Health status assessment is delayed: In existing technologies, health status assessment is mostly based on fixed rules and static thresholds, which fails to fully consider the dynamic changes in the equipment operating environment and historical status, and is prone to misjudgment or missed judgment.

[0005] Traditional fault prediction methods mostly rely on linear models or simple statistical methods, which cannot fully utilize the complex historical operation data and real-time status of the equipment, and the accuracy of the prediction results is poor.

[0006] In existing equipment management, the allocation and scheduling of maintenance resources usually rely on manual decision-making and lack scientific optimization methods, which can easily lead to waste of resources or untimely maintenance of key equipment.

[0007] The existing system lacks comprehensive feedback and data accumulation on the maintenance process and results, and is unable to form a dynamic optimization mechanism, making it difficult to continuously improve equipment management efficiency. Summary of the invention

[0008] In view of the above-mentioned problems, the present invention is proposed.

[0009] Therefore, the technical problem solved by the present invention is: the existing thermal power plant equipment management method has the problems of single monitoring data, delayed health assessment, low fault prediction accuracy and unscientific maintenance resource scheduling, as well as how to achieve comprehensive optimization problems such as accurate monitoring of multi-dimensional equipment operating status, efficient fault prediction and dynamic optimization of resource scheduling.

[0010] In order to solve the above technical problems, the present invention provides the following technical solutions: a thermal power plant equipment management method based on data fusion, comprising:

[0011] Collect real-time data from equipment through sensors and use data fusion algorithms to evaluate equipment health status;

[0012] Based on the real-time operation data of the equipment, the intelligent rule engine dynamically adjusts the evaluation criteria;

[0013] Using historical data and health status assessment results, intelligent fault prediction algorithms are used to predict potential failures and generate maintenance priorities;

[0014] Optimize the scheduling of maintenance resources based on fault prediction and equipment status to ensure timely maintenance and repair of key equipment.

[0015] As a preferred solution of the thermal power plant equipment management method based on data fusion described in the present invention, wherein: the real-time data of the equipment is collected by deploying multiple types of sensors at key operating parts of the equipment, and the multiple types of sensors include temperature sensors, pressure sensors, vibration sensors and flow sensors;

[0016] The acquisition frequency is set in sections, the acquisition frequency of vibration and flow parameters is set to high frequency, and the acquisition frequency of temperature and pressure parameters is set to low frequency, and the collected raw data is stored in the local buffer;

[0017] Perform calibration operations on the raw data, use the preset threshold range of the equipment operating parameters to eliminate abnormal data that exceeds the range, and filter out high-frequency noise through a denoising algorithm based on wavelet transform;

[0018] The calibrated and denoised data were normalized, and the data in different units were converted into a unified numerical range using the maximum-minimum normalization method.

[0019] As a preferred solution of the thermal power plant equipment management method based on data fusion described in the present invention, wherein: the evaluation of the equipment health status includes fusing the normalized sensor data at the sensor level, integrating the data of the same type of sensors by a dynamic weighted average method, and the weight value is dynamically adjusted based on the historical error rate of the sensor and the real-time environmental conditions;

[0020] In feature-level fusion, the main features of different sensor types are extracted based on principal component analysis. The extraction process calculates the covariance of various sensor data and selects the principal component with the highest contribution rate to form a feature matrix.

[0021] In decision-level fusion, the feature matrix is ​​input into the fuzzy logic health assessment model, and the model quantifies the comprehensive health status of the equipment into normal, caution, warning, and fault through membership functions.

[0022] As a preferred solution of the thermal power plant equipment management method based on data fusion described in the present invention, wherein: the dynamic adjustment evaluation standard includes, when the equipment health status is evaluated as normal, by calling the default standard in the rule engine, performing a regular calibration on the current comprehensive health status index, and adjusting the membership function parameter range of the fuzzy logic model;

[0023] When the equipment health status is assessed as caution, the rule engine combines the real-time data trend and the parameters with large weight changes in the principal component matrix to moderately relax or tighten the equipment operation threshold;

[0024] When the equipment health status is assessed as a warning, the real-time operation data is serialized and analyzed through a recurrent neural network to recalculate the sensitivity parameters of the equipment status model;

[0025] When the equipment health status is assessed as a fault, the deep diagnosis module is triggered to match the current assessment criteria with the fault patterns in the historical data. If no matching pattern is found, a new standard record is generated and marked as an update pair.

[0026] As a preferred solution of the thermal power plant equipment management method based on data fusion described in the present invention, wherein: generating maintenance priorities includes inputting the dynamically adjusted health assessment criteria, together with the historical operation data and real-time operation parameters of the equipment, into the long short-term memory network model;

[0027] The input layer of the LSTM network model receives multi-dimensional time series data, the hidden layer captures the long-term trend and short-term fluctuation characteristics of the equipment status, and the output layer calculates the potential fault type and its occurrence probability;

[0028] According to the probability of occurrence of the fault and the changes in important parameters in the equipment operation status, combined with the equipment health status level, the fault types are sorted using a priority generation algorithm;

[0029] Output a maintenance priority list, including equipment name, fault type, estimated occurrence time, and recommended repair sequence.

[0030] As a preferred solution of the thermal power plant equipment management method based on data fusion described in the present invention, wherein: the optimization and scheduling of maintenance resources includes taking the maintenance priority list as an input condition, performing optimization calculation through a mixed integer programming model, and the model objective function includes the number of maintenance resources, the repair time limit and the equipment health status level;

[0031] Generate a maintenance plan based on the optimization calculation results. The maintenance plan is distributed to the corresponding maintenance team in the form of an electronic work order. The work order content includes the equipment name, expected fault location, fault type and maintenance instructions.

[0032] The scheduling system monitors the execution process of the work order in real time. The monitoring data includes the start time of maintenance, actual operation records and expected completion time. If the task is delayed, the automatic adjustment module is triggered to reallocate resources.

[0033] After the maintenance is completed, the dispatch system stores the work order completion information in the maintenance database.

[0034] As a preferred solution of the thermal power plant equipment management method based on data fusion described in the present invention, the real-time operation data of the repaired equipment is recollected, the data is input into the fuzzy logic health assessment model, and the health status level and comprehensive health status index of the equipment are updated;

[0035] Compare the health assessment results before and after maintenance. If the difference exceeds the preset standard, the feedback mechanism is triggered to automatically adjust the weight parameters of the long short-term memory network model and update the historical fault mode.

[0036] Record the newly added fault modes and parameters in the rule base, and update the membership function definition range and corresponding rules in the fuzzy logic evaluation model;

[0037] All updated data and models are stored in the database of the equipment management system.

[0038] Another object of the present invention is to provide a thermal power plant equipment management system based on data fusion, which can solve the problems of limited monitoring range, evaluation results relying on static thresholds leading to misjudgment and missed judgment, poor adaptability of fault prediction models to complex equipment conditions, and low efficiency of maintenance decisions relying on manual experience in existing equipment management by constructing a thermal power plant equipment management system based on data fusion, thereby significantly improving the intelligence and scientific level of equipment management.

[0039] To solve the above technical problems, the present invention provides the following technical solutions: a thermal power plant equipment management system based on data fusion, comprising: an analysis module, an evaluation module, a sorting module and a maintenance module; the analysis module is used to collect real-time data of the equipment through sensors, and use a data fusion algorithm to evaluate the health status of the equipment; the evaluation module is used to dynamically adjust the evaluation criteria based on the real-time operation data of the equipment by an intelligent rule engine; the sorting module is used to use historical data and health status evaluation results to predict potential faults and generate maintenance priorities through an intelligent fault prediction algorithm; the maintenance module is used to optimize the scheduling of maintenance resources according to fault prediction and equipment status to ensure timely maintenance and repair of key equipment.

[0040] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the thermal power plant equipment management method based on data fusion as described above when executing the computer program.

[0041] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the thermal power plant equipment management method based on data fusion as described above.

[0042] Beneficial effects of the present invention: The thermal power plant equipment management method based on data fusion provided by the present invention realizes comprehensive monitoring of the equipment operation status through the optimized arrangement of multiple types of sensors and segmented frequency acquisition, and combines the multi-level data fusion technology based on dynamic weighted average and principal component analysis to significantly improve the accuracy and comprehensiveness of the data; through the fuzzy logic health assessment model, the equipment health status is quantified into four levels, which can dynamically adapt to different operating conditions and effectively reduce the probability of misjudgment and missed judgment; the long short-term memory network is used to perform deep learning on the historical operation data and real-time status of the equipment, and high-precision prediction of potential fault types and occurrence probabilities is achieved, and support is provided for scientific decision-making of maintenance tasks through the priority generation algorithm; the maintenance resource optimization scheduling scheme combined with the mixed integer programming model improves maintenance efficiency, avoids resource waste and key equipment maintenance delays; through the closed-loop feedback mechanism, the evaluation criteria and fault mode library are automatically updated after the maintenance is completed, and the evaluation and prediction accuracy are continuously optimized to form a dynamically optimized equipment management process, which significantly improves the safety, reliability and economy of the operation of thermal power plant equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0044] Figure 1 An overall flow chart of a thermal power plant equipment management method based on data fusion provided by an embodiment of the present invention.

[0045] Figure 2 An overall structural diagram of a thermal power plant equipment management system based on data fusion provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0047] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0048] Example 1

[0049] Reference Figure 1 , which is an embodiment of the present invention, provides a thermal power plant equipment management method based on data fusion, comprising:

[0050] Collect real-time data from equipment through sensors and use data fusion algorithms to evaluate equipment health status;

[0051] Based on the real-time operation data of the equipment, the intelligent rule engine dynamically adjusts the evaluation criteria;

[0052] Using historical data and health status assessment results, intelligent fault prediction algorithms are used to predict potential failures and generate maintenance priorities;

[0053] Optimize the scheduling of maintenance resources based on fault prediction and equipment status to ensure timely maintenance and repair of key equipment.

[0054] The real-time data of the equipment is collected by deploying multiple types of sensors at key operating locations of the equipment, wherein the multiple types of sensors include temperature sensors, pressure sensors, vibration sensors, and flow sensors;

[0055] The acquisition frequency is set in sections, the acquisition frequency of vibration and flow parameters is set to high frequency, and the acquisition frequency of temperature and pressure parameters is set to low frequency, and the collected raw data is stored in the local buffer;

[0056] Perform calibration operations on the raw data, use the preset threshold range of the equipment operating parameters to eliminate abnormal data that exceeds the range, and filter out high-frequency noise through a denoising algorithm based on wavelet transform;

[0057] The calibrated and denoised data were normalized, and the data in different units were converted into a unified numerical range using the maximum-minimum normalization method.

[0058] It should be noted that the real-time data of the equipment involved involves the arrangement of multiple types of sensors in specific areas of the equipment, and the arrangement of sensors needs to be optimized according to the operating characteristics of the equipment. For example, at the bearing of the generator set, the vibration sensor should be installed in a stable support area near the bearing seat to capture the mechanical vibration characteristics in real time; at the superheater tube wall of the boiler, the temperature sensor needs to be fixed by welding or high-temperature bonding to ensure accurate sensing of the surface temperature changes of the tube wall; at the inlet and outlet steam pipelines of the turbine, the pressure sensor needs to be fixed with a shock-proof bracket and maintain smooth contact with the inner wall of the pipeline; and at the water inlet of the condenser, the flow sensor needs to be configured at a position where the fluid turbulence has the least impact to ensure measurement accuracy.

[0059] Furthermore, the segmented setting of the sensor acquisition frequency is based on the data characteristics and acquisition purpose. Vibration sensors and flow sensors need to collect high-frequency data (such as 100Hz to 200Hz) to capture rapidly changing signals, such as abnormal bearing vibration or instantaneous fluctuations in flow; temperature sensors and pressure sensors need to collect low-frequency data (such as 1Hz to 10Hz) to record slowly changing trends, such as temperature rise of the pipe wall or fluctuations in steam pressure. The collected data is transmitted to the local buffer in sequence through the data acquisition module at the set frequency.

[0060] Furthermore, when performing calibration operations on raw data, the preset threshold ranges of equipment operating parameters are defined according to the specific equipment and sensor type. For example, the normal range of bearing vibration is 0.02 to 0.2 mm / s, and data exceeding this range is marked as anomalies. The marked abnormal data is corrected using an interpolation algorithm, such as selecting the nearest normal data on both sides of the abnormal point in the time series for linear interpolation to fill the gaps caused by the abnormal data. The denoising algorithm uses wavelet transform to decompose the signal into different frequency components, and performs soft threshold processing by setting a threshold on the high-frequency component to remove noise while retaining the key features in the low-frequency signal.

[0061] The device health status evaluation includes fusing the normalized sensor data at the sensor level, integrating the data of similar sensors by a dynamic weighted average method, and the weight value is dynamically adjusted based on the sensor's historical error rate and real-time environmental conditions;

[0062] In feature-level fusion, the main features of different sensor types are extracted based on principal component analysis. The extraction process calculates the covariance of various sensor data and selects the principal component with the highest contribution rate to form a feature matrix.

[0063] In decision-level fusion, the feature matrix is ​​input into the fuzzy logic health assessment model, and the model quantifies the comprehensive health status of the equipment into normal, caution, warning, and fault through membership functions.

[0064] It should be noted that the evaluation of the health status of the equipment includes multi-layer fusion of the normalized sensor data, which is divided into sensor-level fusion, feature-level fusion and decision-level fusion. Among them, the dynamic weighted average method of sensor-level fusion integrates the data of similar sensors by assigning different weights. For example, the weight of the temperature sensor can be dynamically adjusted according to its historical error rate and the current operating environment. The operating environment includes the external environment temperature range, humidity level and sensor aging degree.

[0065] Furthermore, in feature-level fusion, the main features of different sensor types are extracted based on principal component analysis. The process of principal component analysis includes calculating the covariance matrix of sensor data, identifying the eigenvectors with the largest contribution in the data, and retaining the top N main features. For example, when the co-correlation of vibration and temperature sensors is high, principal component analysis can extract the feature representation of their common influence, thereby reducing data redundancy and highlighting key information. Each row in the feature matrix represents the main feature of a sensor type, and each column corresponds to a specific moment in the time series.

[0066] Furthermore, decision-level fusion achieves quantification of the comprehensive health status of the equipment through a fuzzy logic health assessment model. The specific settings of the membership functions in the model are based on the operating conditions and historical data of the equipment. Taking temperature as an example, the "normal" range of the membership function may be defined as 300 to 500°C, while the "caution" range is 500 to 600°C. When the measured temperature is 550°C, its health status level can be determined by a weighted calculation of the membership of the two levels of "normal" and "caution". The final health status level is stored in a distributed database and marked as the latest status for subsequent analysis.

[0067] The dynamic adjustment of the evaluation criteria includes, when the health status of the equipment is evaluated as normal, performing a regular calibration on the current comprehensive health status indicator by calling the default criteria in the rule engine, and adjusting the membership function parameter range of the fuzzy logic model;

[0068] When the equipment health status is assessed as caution, the rule engine combines the real-time data trend and the parameters with large weight changes in the principal component matrix to moderately relax or tighten the equipment operation threshold;

[0069] When the equipment health status is assessed as a warning, the real-time operation data is serialized and analyzed through a recurrent neural network to recalculate the sensitivity parameters of the equipment status model;

[0070] When the equipment health status is assessed as a fault, the deep diagnosis module is triggered to match the current assessment criteria with the fault patterns in the historical data. If no matching pattern is found, a new standard record is generated and marked as an update pair.

[0071] It should be noted that the rule engine for dynamically adjusting the evaluation criteria includes a preset rule base and a real-time learning module. The rule base has built-in parameter adjustment strategies for different health states. For example, when the health state is "normal", the rule engine will call the default standard and set the sensitivity parameter of the device's comprehensive health state indicator to the lowest value to avoid frequent triggering of unnecessary adjustments.

[0072] Furthermore, when the health status is assessed as "caution", the rule engine makes adjustments based on the features with the largest contribution rate in the principal component matrix. For example, when the contribution rate of the vibration sensor exceeds 80%, the rule engine will dynamically tighten the health assessment threshold of the vibration to detect potential bearing abnormalities earlier. At the same time, the real-time data trend analysis module monitors the rate of change of the health status of the "caution" state equipment and triggers a "warning" warning in advance when the state deteriorates.

[0073] Furthermore, when the health status is evaluated as "fault", the deep diagnosis module compares the current data with the historical fault patterns through a clustering algorithm. For example, if the vibration characteristics of the equipment do not match the known pattern, it will be marked as a new fault and the characteristic value will be recorded. The real-time learning module will update the new fault information to the rule base and generate corresponding adjustment rules to ensure that the evaluation criteria are consistent with the actual situation.

[0074] Generating maintenance priorities includes inputting the dynamically adjusted health assessment criteria, together with the historical operation data and real-time operation parameters of the equipment, into a long short-term memory network model;

[0075] The input layer of the LSTM network model receives multi-dimensional time series data, the hidden layer captures the long-term trend and short-term fluctuation characteristics of the equipment status, and the output layer calculates the potential fault type and its occurrence probability;

[0076] According to the probability of occurrence of the fault and the changes in important parameters in the equipment operation status, combined with the equipment health status level, the fault types are sorted using a priority generation algorithm;

[0077] Output a maintenance priority list, including equipment name, fault type, estimated occurrence time, and recommended repair sequence.

[0078] It should be noted that the long short-term memory network model adopts a recursive neural network architecture, and the input layer receives time-series sensor data, such as normalized multidimensional data such as temperature, vibration and flow. The hidden layer captures the long-term trend characteristics of the equipment operation status through memory units and filters irrelevant information through the forget gate. For example, when there are atypical fluctuations in the input sequence, the forget gate will automatically reduce its weight to avoid interference with the output results.

[0079] Furthermore, the calculation of potential fault types is based on the time step of the hidden layer and the activation function of the output layer. For example, for abnormal changes in vibration frequency, the model can determine whether there is a possibility of bearing failure through the state information accumulated in multiple time steps and calculate the probability of occurrence. The priority generation algorithm performs weighted sorting based on the probability of fault occurrence, health status level, and equipment importance parameters. For example, critical equipment with a health status of "warning" has a higher weight in the priority sorting.

[0080] Furthermore, the generation process of the maintenance priority list supports the dynamic allocation of maintenance tasks. For example, each task in the list contains the equipment identifier, fault type, predicted occurrence time value, and recommended maintenance sequence. The list interacts with the data of the scheduling module to update the priority in real time to adapt to changes in the equipment operating environment.

[0081] The optimization scheduling of maintenance resources includes taking the maintenance priority list as input conditions and performing optimization calculations through a mixed integer programming model, wherein the model objective function includes the number of maintenance resources, the repair time limit, and the equipment health status level;

[0082] Generate a maintenance plan based on the optimization calculation results. The maintenance plan is distributed to the corresponding maintenance team in the form of an electronic work order. The work order content includes the equipment name, expected fault location, fault type and maintenance instructions.

[0083] The scheduling system monitors the execution process of the work order in real time. The monitoring data includes the start time of maintenance, actual operation records and expected completion time. If the task is delayed, the automatic adjustment module is triggered to reallocate resources.

[0084] After the maintenance is completed, the dispatch system stores the work order completion information in the maintenance database.

[0085] It should be noted that the constraints of the mixed integer programming model include the type and quantity of maintenance resources, task priority, and estimated completion time. The optimization goal of the model is to minimize the maintenance time and resource waste, and globally optimize resource scheduling through linear constraint equations. For example, for equipment with a health status level of "fault", the constraints for maintenance resource allocation will be set to "mandatory satisfaction" to ensure that critical tasks are executed first.

[0086] Furthermore, the generated electronic work order includes the equipment name, specific fault location, recommended repair steps and estimated completion time. The work order is stored in JSON format and distributed to the maintenance team via wireless network. The status of each work order is updated in real time as the task progresses, such as "accepted", "in progress" or "completed".

[0087] Furthermore, the real-time monitoring module locates and records the maintenance team's task execution through wireless communication devices. For example, Bluetooth beacons are used to confirm whether maintenance personnel are performing operations at designated locations, and the maintenance start and completion times are recorded in the database. If a task is behind schedule, the scheduling system will re-optimize task allocation by adjusting maintenance resources and send a new electronic work order to the nearest team.

[0088] Recollect real-time operation data of the repaired equipment, input the data into the fuzzy logic health assessment model, and update the health status level and comprehensive health status index of the equipment;

[0089] Compare the health assessment results before and after maintenance. If the difference exceeds the preset standard, the feedback mechanism is triggered to automatically adjust the weight parameters of the long short-term memory network model and update the historical fault mode.

[0090] Record the newly added fault modes and parameters in the rule base, and update the membership function definition range and corresponding rules in the fuzzy logic evaluation model;

[0091] All updated data and models are stored in the database of the equipment management system as the basic input for subsequent fault prediction and resource scheduling, forming a dynamic optimization closed-loop management process.

[0092] It should be noted that the feedback evaluation module includes two parts: health status re-evaluation and fault prediction model parameter optimization. After the maintenance is completed, the real-time data collected by the sensor is input into the health evaluation model to generate the latest health status level. If the evaluation result shows that the health status has not returned to "normal", the feedback module will record the evaluation deviation and trigger the deep diagnosis function.

[0093] Furthermore, the parameter optimization of the LSTM network model is based on the back propagation algorithm, and the weights of the hidden layer and the output layer are adjusted in combination with the evaluation results before and after maintenance. For example, if the prediction error of a certain fault type is large, the feedback mechanism will increase the training weight of the corresponding time step and reset the threshold value of the forget gate.

[0094] Furthermore, the newly added fault modes are updated to the rule base through the automated rule generation module. The rules include fault feature vectors, matching algorithm parameters, and adjusted health assessment criteria. All updated content is stored in a distributed database and synchronized to other management modules to form a dynamic closed-loop management process to improve the accuracy of subsequent predictions and assessments.

[0095] Example 2

[0096] Reference Figure 2 , which is an embodiment of the present invention, provides a thermal power plant equipment management system based on data fusion, including:

[0097] Analysis module 100, evaluation module 200, sorting module 300 and maintenance module 400;

[0098] The analysis module 100 is used to collect real-time data of the device through sensors and evaluate the health status of the device using a data fusion algorithm;

[0099] The evaluation module 200 is used for dynamically adjusting the evaluation criteria by the intelligent rule engine based on the real-time operation data of the equipment;

[0100] The sorting module 300 is used to use historical data and health status assessment results to predict potential faults and generate maintenance priorities through intelligent fault prediction algorithms;

[0101] The maintenance module 400 is used to optimize the scheduling of maintenance resources according to fault prediction and equipment status to ensure timely maintenance and repair of key equipment.

[0102] Example 3

[0103] An embodiment of the present invention is different from the first two embodiments in that:

[0104] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0105] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0106] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0107] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0108] Example 4

[0109] As an embodiment of the present invention, a thermal power plant equipment management method based on data fusion is provided. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0110] In order to verify the technical advantages of the method of the present invention over the traditional method, two bearing-driven generator sets were selected as experimental objects, of which equipment A was managed by the traditional method and equipment B was managed by the method of the present invention. The experimental environment was carried out in the actual operation workshop of a thermal power plant. The experimental period was 30 days, and the equipment was operated for 10 hours a day to simulate the state changes and potential failures that may occur in the actual operation of the equipment.

[0111] Before the experiment, sensors were placed at key locations on the two devices. Device A installed a vibration sensor on the bearing seat, with the acquisition frequency set to 10 Hz; Device B used multiple types of sensors, including a vibration sensor (200 Hz) on the bearing seat, a temperature sensor (5 Hz) on the superheater tube wall, a pressure sensor (1 Hz) on the steam pipeline, and a flow sensor (50 Hz) at the condenser inlet. The sensor data was transmitted to the local computing unit via a wireless module.

[0112] After the experiment started, the vibration sensor of device A directly outputted raw data to the local database without data processing; the data acquisition unit of device B calibrated all sensor data in real time, eliminated outliers through a dynamic threshold algorithm, and combined with a wavelet transform denoising algorithm to filter out high-frequency noise. At the same time, all data were normalized to a uniform value range (0-1) through maximum-minimum normalization.

[0113] Device A uses the traditional fixed threshold judgment method, which is based only on the output data of the vibration sensor. When the vibration amplitude exceeds 0.2 mm / s, it is directly marked as an abnormal state. The data fusion module of device B first performs dynamic weighted averaging on the data of similar sensors (the weight is adjusted according to the historical error rate of the sensor), and then extracts the main features of multi-type data through principal component analysis. The features are input into the fuzzy logic health assessment model, and the comprehensive health status is quantified into four levels: "normal", "caution", "warning" and "fault" and stored in real time.

[0114] On the 12th day of the experiment, device A predicted that bearing wear would cause vibration to exceed the standard within 18 hours through simple linear regression analysis of historical vibration data; device B identified the potential risk of bearing wear through the long short-term memory network (LSTM) model combined with historical data and real-time status trends, and predicted that it would deteriorate to a dangerous state within 36 hours. At the same time, device B generated a maintenance priority list based on the output of the LSTM model, including device name, potential fault type, expected time of occurrence, and recommended maintenance sequence.

[0115] After the fault prediction, the workshop maintenance team of equipment A determined the priority maintenance tasks based on manual experience and allocated maintenance resources. The average time from the issuance to the completion of the maintenance task was 18.4 hours, and the resource utilization rate was not recorded during this period. Equipment B automatically generated electronic work orders based on the mixed integer programming model of the method of the present invention, assigned maintenance tasks to the optimal team, and monitored the progress of maintenance in real time. The average time from the issuance to the completion of the maintenance task was 9.6 hours, and the resource utilization rate reached 94.6%. After all repairs were completed, the health assessment module of equipment B re-evaluated the repair equipment, updated the health status level and optimized the fault prediction model parameters.

[0116] The experimental results are shown in Table 1.

[0117] Table 1 Comparison of experimental results

[0118]

[0119]

[0120] The accuracy of health status assessment of device B is significantly higher than that of device A. This is because the present invention adopts multi-level data fusion technology, organically combines sensor level, feature level and decision level, and optimizes the assessment criteria in real time through a dynamic adjustment mechanism, making data processing more comprehensive and accurate, thereby effectively eliminating noise interference in the data and being able to adapt to dynamic changes in the equipment operating environment. In contrast, device A only relies on a single type of sensor and a fixed threshold method, and is unable to integrate multi-dimensional data characteristics, resulting in a less accurate assessment result.

[0121] The significant improvement in the lead time and accuracy of fault prediction is due to the fact that the present invention adopts a long short-term memory network (LSTM) model, which can capture the time series characteristics of equipment operation data. Through deep learning, the model can not only identify the current operating status, but also accurately analyze potential faults in combination with historical trends, significantly improving the reliability and foresight of the prediction results. In contrast, the simple regression method used by device A can only perform linear trend analysis and cannot capture the complex characteristics of equipment state changes, resulting in significant deficiencies in both the lead time and accuracy of the prediction.

[0122] The shortening of the average maintenance completion time and the reduction of resource waste rate are due to the maintenance priority list and mixed integer programming scheduling model designed by the present invention. This method ensures that the execution efficiency of maintenance tasks is greatly improved through scientific task priority sorting and resource allocation optimization, while real-time monitoring of maintenance progress and dynamic adjustment of resource allocation strategies, avoiding the time waste and resource idleness caused by manual judgment delays and unreasonable resource scheduling in traditional methods.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A thermal power plant equipment management method based on data fusion, characterized in that: include: Collect real-time data from equipment through sensors and use data fusion algorithms to evaluate equipment health status; Based on the real-time operation data of the equipment, the intelligent rule engine dynamically adjusts the evaluation criteria; Using historical data and health status assessment results, intelligent fault prediction algorithms are used to predict potential failures and generate maintenance priorities; Optimize the scheduling of maintenance resources based on fault prediction and equipment status to ensure timely maintenance and repair of key equipment.

2. The thermal power plant equipment management method based on data fusion according to claim 1, characterized in that: The real-time data of the equipment is collected by deploying multiple types of sensors at key operating locations of the equipment, wherein the multiple types of sensors include temperature sensors, pressure sensors, vibration sensors, and flow sensors; The acquisition frequency is set in sections, the acquisition frequency of vibration and flow parameters is set to high frequency, and the acquisition frequency of temperature and pressure parameters is set to low frequency, and the collected raw data is stored in the local buffer; Perform calibration operations on the raw data, use the preset threshold range of the equipment operating parameters to eliminate abnormal data that exceeds the range, and filter out high-frequency noise through a denoising algorithm based on wavelet transform; The calibrated and denoised data were normalized, and the data in different units were converted into a unified numerical range using the maximum-minimum normalization method.

3. The thermal power plant equipment management method based on data fusion according to claim 2 is characterized in that: The device health status evaluation includes fusing the normalized sensor data at the sensor level, integrating the data of similar sensors by a dynamic weighted average method, and the weight value is dynamically adjusted based on the sensor's historical error rate and real-time environmental conditions; In feature-level fusion, the main features of different sensor types are extracted based on principal component analysis. The extraction process calculates the covariance of various sensor data and selects the principal component with the highest contribution rate to form a feature matrix. In decision-level fusion, the feature matrix is ​​input into the fuzzy logic health assessment model, and the model quantifies the comprehensive health status of the equipment into normal, caution, warning, and fault through membership functions.

4. The thermal power plant equipment management method based on data fusion according to claim 3 is characterized in that: The dynamic adjustment of the evaluation criteria includes, when the health status of the equipment is evaluated as normal, performing a regular calibration on the current comprehensive health status indicator by calling the default criteria in the rule engine, and adjusting the membership function parameter range of the fuzzy logic model; When the equipment health status is assessed as caution, the rule engine combines the real-time data trend and the parameters with large weight changes in the principal component matrix to moderately relax or tighten the equipment operation threshold; When the equipment health status is assessed as a warning, the real-time operation data is serialized and analyzed through a recurrent neural network to recalculate the sensitivity parameters of the equipment status model; When the equipment health status is assessed as a fault, the deep diagnosis module is triggered to match the current assessment criteria with the fault patterns in the historical data. If no matching pattern is found, a new standard record is generated and marked as an update pair.

5. The thermal power plant equipment management method based on data fusion according to claim 4 is characterized in that: Generating maintenance priorities includes inputting the dynamically adjusted health assessment criteria, together with the historical operation data and real-time operation parameters of the equipment, into a long short-term memory network model; The input layer of the LSTM network model receives multi-dimensional time series data, the hidden layer captures the long-term trend and short-term fluctuation characteristics of the equipment status, and the output layer calculates the potential fault type and its occurrence probability; According to the probability of occurrence of the fault and the changes in important parameters in the equipment operation status, combined with the equipment health status level, the fault types are sorted using a priority generation algorithm; Output a maintenance priority list, including equipment name, fault type, estimated occurrence time, and recommended repair sequence.

6. The thermal power plant equipment management method based on data fusion according to claim 5, characterized in that: The optimization scheduling of maintenance resources includes taking the maintenance priority list as input conditions and performing optimization calculations through a mixed integer programming model, wherein the model objective function includes the number of maintenance resources, the repair time limit, and the equipment health status level; Generate a maintenance plan based on the optimization calculation results. The maintenance plan is distributed to the corresponding maintenance team in the form of an electronic work order. The work order content includes the equipment name, expected fault location, fault type and maintenance instructions. The scheduling system monitors the execution process of the work order in real time. The monitoring data includes the start time of maintenance, actual operation records and expected completion time. If the task is delayed, the automatic adjustment module is triggered to reallocate resources. After the maintenance is completed, the dispatch system stores the work order completion information in the maintenance database.

7. The thermal power plant equipment management method based on data fusion according to claim 6 is characterized in that: Recollect real-time operation data of the repaired equipment, input the data into the fuzzy logic health assessment model, and update the health status level and comprehensive health status index of the equipment; Compare the health assessment results before and after maintenance. If the difference exceeds the preset standard, the feedback mechanism is triggered to automatically adjust the weight parameters of the long short-term memory network model and update the historical fault mode. Record the newly added fault modes and parameters in the rule base, and update the membership function definition range and corresponding rules in the fuzzy logic evaluation model; All updated data and models are stored in the database of the equipment management system.

8. A system using the thermal power plant equipment management method based on data fusion as claimed in any one of claims 1 to 7, characterized in that: include: An analysis module (100), an evaluation module (200), a sorting module (300), and a maintenance module (400); The analysis module (100) is used to collect real-time data of the equipment through sensors and evaluate the health status of the equipment using a data fusion algorithm; The evaluation module (200) is used for dynamically adjusting the evaluation criteria by an intelligent rule engine based on real-time operation data of the device; The sorting module (300) is used to use historical data and health status assessment results to predict potential faults and generate maintenance priorities through an intelligent fault prediction algorithm; The maintenance module (400) is used to optimize the scheduling of maintenance resources according to fault prediction and equipment status, so as to ensure timely maintenance and repair of key equipment.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the thermal power plant equipment management method based on data fusion according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the thermal power plant equipment management method based on data fusion described in any one of claims 1 to 7 are implemented.

Images