Big data-based intelligent monitoring fault prediction system for thermal power plant

Abstract

The invention discloses a big data-based intelligent monitoring fault prediction system for a thermal power plant, relates to the technical field of thermal power plants, and solves the technical problems in the prior art that after fault early warning, personnel need to spend time in searching problems and overhaul, the efficiency of repair and maintenance work is low, and the operation of the power plant is affected. The method comprises the following steps: acquiring a plurality of operation data of a thermal power plant in real time, analyzing and evaluating an abnormal data sequence in a period, and performing data rejection to obtain a new operation data sequence; establishing a digital twinborn model of the thermal power plant, and obtaining fault information; key influence factors corresponding to the fault information of the thermal power plant are analyzed; and analyzing a fault level based on the abnormal data sequence, based on the key influence factors and the fault levels, constructing a prison disk fault prediction model; the early warning of the thermal power plant is realized, the fault occurrence position is accurately identified, the efficiency of the maintenance operation is improved, and the normal operation of the thermal power plant is ensured.

Description

Technical Field

[0001] The present invention relates to the field of thermal power plants, and in particular to a big data-based intelligent monitoring panel fault prediction system for thermal power plants. Background Art

[0002] The invention patent application with publication number CN115425749A discloses an intelligent monitoring and fault warning system for a smart power plant, including: an intelligent interface module, which is equipped with different intelligent interfaces; a data acquisition, storage and processing module, which collects real-time data and historical data of unit equipment operation from a DCS system or a SIS system in seconds; a configured data modeling module, which is configured through existing algorithms, graphical configuration modeling or code interaction modeling; an intelligent analysis module, which realizes parameter warning and fault analysis; an intelligent monitoring module, which displays warning and fault information in real time; a monitoring assistant module, which records operating data, automatically reads meters under typical operating conditions, and records historical data of characteristic values ​​of important parameters when parameters on the monitoring screen are abnormal.

[0003] The above-mentioned prior art realizes intelligent early warning of power plants, avoids human misjudgment, and reduces the tasks of on-duty personnel, but after the fault warning, personnel need to spend time to find the problem and repair it, which is inefficient in repair and maintenance work and affects the operation of the power plant. Therefore, the present invention provides a thermal power plant intelligent monitoring panel fault prediction system based on big data. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a thermal power plant intelligent monitoring panel fault prediction system based on big data, which is used to solve the technical problems that the prior art requires personnel to spend time to find and repair problems after fault warning, and the efficiency of repair and maintenance work is low, which affects the operation of the power plant.

[0005] To achieve the above purpose, the present invention provides a thermal power plant intelligent monitoring panel fault prediction system based on big data, comprising:

[0006] A data acquisition module is connected to a plurality of sensors deployed on key components of target equipment in a thermal power plant, and is used to collect a number of operating data on the key components of the target equipment in real time, and transmit the operating data collected within an evaluation period to an abnormal data analysis module based on a preset evaluation period, wherein the operating data includes temperature, voltage and current;

[0007] The abnormal data analysis module organizes various types of operating data on each key component collected during the evaluation period into corresponding operating data sequences, analyzes the false abnormal data and abnormal data sequences during the evaluation period, marks the abnormal data sequences, deletes the false abnormal data in each operating data sequence, and obtains the corresponding new operating data sequence;

[0008] The keyboard fault prediction module predicts the fault information and fault level of the target device evaluation cycle based on the new operation data sequence transmitted by the abnormal data analysis module and the abnormal data sequence marked in the new operation data sequence, specifically:

[0009] ① Establish a digital twin model of the thermal power plant and run it based on the new operation data sequence to obtain fault information, where the fault information includes equipment name, model, serial number and fault location;

[0010] ②Analyze the key influencing factors corresponding to each fault information of the thermal power plant;

[0011] ③ Extract the marked abnormal data sequence from the new running data sequence, and analyze the fault level based on the abnormal data sequence;

[0012] ④ Based on key influencing factors and fault levels, a monitoring fault prediction model is constructed to predict the fault information and fault level of the evaluation period;

[0013] The fault processing module dispatches a maintenance unit matching the fault level to repair the fault based on the fault information and fault level predicted by the keyboard fault prediction module.

[0014] Preferably, the evaluation cycle is a number of time periods evenly divided from a day, wherein each time period is an evaluation cycle;

[0015] In addition, when the data acquisition module transmits the operating data collected during the evaluation cycle to the abnormal data analysis module, different encryption algorithms are used to encrypt and transmit the operating data collected during each evaluation cycle. After receiving the data, the abnormal data analysis module uses the secret key corresponding to the encryption algorithm to decrypt the data and then analyze it.

[0016] Preferably, different categories of operating data on the same key component in the abnormal data analysis module form corresponding operating data sequences according to the categories they belong to; the same category of operating data on different key components form corresponding operating data sequences according to the key components they belong to;

[0017] In addition, the specific steps for analyzing false abnormal data and abnormal data sequences in the abnormal data analysis module are as follows:

[0018] S11, comparing a number of data in each operation data sequence with the corresponding preset standard threshold in sequence, and when the data in the operation data sequence exceeds the corresponding preset standard threshold, marking the data as abnormal data;

[0019] S12. When the time points of several abnormal data collection in a certain operation data sequence within the evaluation period are continuous and the continuous duration exceeds the corresponding preset duration, these several abnormal data constitute an abnormal data sequence; otherwise, they are false abnormal data.

[0020] Preferably, the specific steps of the digital twin model of a thermal power plant are as follows:

[0021] S21. Obtain various historical operation data and perform preprocessing, wherein the preprocessing includes removing invalid data, outliers and missing values, and converting the data into a unified format and unit;

[0022] S22. Extract corresponding historical new operation data sequences from various historical operation data, construct various historical new operation data sequences within the same evaluation period into an operation data matrix, and obtain several historical operation data matrices, wherein the operation data matrix is ​​as follows:

[0023]

[0024] In the formula, A1~A k 、Y1~Y k A1~Y1, A k ~Y k It is various types of operation data collected at the same time point within the evaluation period;

[0025] S23. Construct a physical model of a thermal power plant based on physical principles and engineering knowledge;

[0026] S24. Combine the historical operation data matrix and the corresponding physical model to build a digital twin model of the thermal power plant.

[0027] Preferably, the specific steps of analyzing the key influencing factors corresponding to the fault information are:

[0028] 1) Extracting several operation data matrices corresponding to several identical fault information from the historical data, and counting the number of times abnormal data sequences appear in each new operation data sequence of the several operation data matrices;

[0029] 2) Divide the number of times obtained by statistics by the total number of times abnormal data sequences appear in several operation data matrices to obtain the abnormal proportion of operation data;

[0030] 3) When the abnormal proportion of operating data exceeds the proportion threshold, the operating data corresponding to the abnormal proportion of operating data is the key influencing factor of the corresponding fault information, and the fault information is associated with the corresponding key influencing factor.

[0031] Preferably, if several abnormal data sequences appear in the new running data sequence, they are counted only once.

[0032] Preferably, the specific steps of analyzing the fault level based on the abnormal data sequence are: performing mean processing on the data in the abnormal data sequence, and performing difference processing with the preset standard threshold to obtain the deviation difference. When the deviation difference is greater than the deviation threshold, the fault level is level one; otherwise, the fault level is level two, where the fault degree is level one > level two.

[0033] Preferably, when there are several abnormal data sequences in the new operation data sequence, the maximum deviation difference is compared with the deviation threshold to evaluate the fault level.

[0034] Preferably, the specific steps of constructing a monitoring disk fault prediction model based on key influencing factors and fault levels are:

[0035] 1) Extract several abnormal data sequences, corresponding fault information and fault levels from historical data; use the abnormal data sequences as input data of the artificial intelligence model, and the corresponding fault information and fault levels as output data of the artificial intelligence;

[0036] 2) The artificial intelligence model is trained through input data and output data to obtain a monitoring fault prediction model, where the artificial intelligence model is built based on the LSTM model;

[0037] 3) Input the abnormal data sequence of the evaluation period into the monitoring fault prediction model, and predict and output the fault information and fault level of the evaluation period.

[0038] Preferably, the steps for evaluating the grade of each maintenance unit are as follows:

[0039] S31. Based on the number of times each maintenance unit successfully repairs a primary fault, calculate the maintenance success ratio of the maintenance unit. The calculation formula is as follows:

[0040]

[0041] In the formula, Δ i represents the maintenance success rate of the i-th maintenance unit, E i represents the number of times the i-th maintenance unit successfully repairs the first-level fault, and e represents the number of maintenance units;

[0042] S32. Based on the skill types and corresponding skill proficiency of all maintenance workers in each maintenance unit, the average skill proficiency of all maintenance workers in each maintenance unit is calculated, wherein the skill proficiency is obtained by expert scoring, and the calculation formula is as follows:

[0043]

[0044] In the formula, represents the average skill proficiency of the jth maintenance worker in the i-th maintenance unit, f krepresents the proficiency of the kth skill type, where k represents the number of skill types;

[0045] S33. Based on the average skill proficiency of all maintenance workers in each maintenance unit, the average skill proficiency of each maintenance unit is calculated. The calculation formula is as follows:

[0046]

[0047] In the formula, R i represents the average skill proficiency of the maintenance group i, j i represents the number of maintenance workers in the i-th maintenance unit;

[0048] S34: Based on the maintenance success ratio of each maintenance unit in step S31 and the average skill proficiency of each maintenance unit in step S33, the level of each maintenance unit is evaluated, specifically:

[0049] If the maintenance success ratio of each maintenance unit is greater than the ratio threshold, and the average skill proficiency of each maintenance unit is greater than the proficiency threshold, the maintenance unit is evaluated as a first-level maintenance unit and repairs a first-level fault; otherwise, the maintenance unit is evaluated as a second-level maintenance unit and repairs a second-level fault.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. The present invention can quickly and accurately identify data that exceeds the normal range by comparing the operating data with the corresponding preset standard threshold, thereby marking it as abnormal in time, improving the accuracy and consistency of abnormal evaluation processing, and by combining the continuity of abnormal data in time, it can more accurately identify the real abnormal data, avoid the problem of misjudgment caused by short-term fluctuations or errors in single abnormal data, and affect subsequent fault analysis.

[0052] 2. The present invention extracts operation data matrices corresponding to several identical fault information from historical data, ensuring the richness and diversity of the data and laying a solid foundation for subsequent statistical analysis; the number of times abnormal data sequences appear in each operation data sequence in several operation data matrices is counted, and the abnormal proportion of the operation data is evaluated, which reflects the contribution of each operation data in the occurrence of the fault, realizes the accurate screening of key influencing factors, and lays the foundation for the subsequent establishment of the relationship between fault information and key influencing factors, which is conducive to the subsequent precise positioning of the source of the fault and the acquisition of accurate fault information.

[0053] 3. The present invention collects a number of operating data of thermal power plant equipment in real time, analyzes and evaluates the abnormal data sequence within a period; deletes the abnormal data in the operating data sequence except the abnormal data sequence to obtain a new operating data sequence; establishes a digital twin model of the thermal power plant, and operates based on the new operating data sequence to obtain fault information; analyzes the key influencing factors corresponding to each fault information of the thermal power plant; and analyzes the fault level based on the abnormal data sequence; constructs a monitoring panel fault prediction model based on the key influencing factors and the fault level; realizes early warning of thermal power plant faults, accurately locates the source of the fault, improves the maintenance efficiency, and helps to reduce the downtime and maintenance cost of the thermal power plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0056] Figure 2 It is a schematic diagram of the process of the present invention;

[0057] Figure 3 The figure is a flow chart of the method for analyzing key influencing factors of various faults of the present invention. DETAILED DESCRIPTION

[0058] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] See also Figure 1-Figure 3 The present invention provides a thermal power plant intelligent monitoring panel fault prediction system based on big data, including a data acquisition module, an abnormal data analysis module, a monitoring panel fault prediction module and a fault processing module.

[0060] The data acquisition module involved in the present invention is connected to multiple sensors deployed on key components of target equipment in a thermal power plant, and is used to collect a number of operating data on the key components of the target equipment in real time, and transmit the operating data collected within an evaluation period to an abnormal data analysis module based on a preset evaluation period, wherein the operating data includes temperature, voltage and current.

[0061] It should be noted that in this application, the temperature, voltage and current data of the equipment are collected by deploying temperature, voltage and current sensors on the equipment, but in actual operation, pressure, flow, vibration and other sensors can also be deployed on the equipment to collect the pressure, flow, vibration and other operating data of the equipment. In addition, the target equipment in the thermal power plant includes but is not limited to boilers, steam turbines, and generators. Since the key components of each equipment are different, the key components of the target equipment are not limited in this application.

[0062] Taking the steam turbine as an example, the key components include the steam chamber, exhaust area, outer cylinder, main steam valve, middle main door, and generator. The temperature data of the steam chamber, exhaust area, outer cylinder, main steam valve, and middle main door are collected by temperature sensors deployed in the steam chamber, exhaust area, outer cylinder, main steam valve, and middle main door. The voltage and current data of the generator are collected by voltage and current sensors deployed in the generator.

[0063] In the present application, an evaluation cycle is a number of time periods after a day is evenly divided, where each time period is an evaluation cycle.

[0064] For example, a day, that is, 24 hours, is evenly divided into 6 time periods, and each time period is an evaluation cycle, that is, each evaluation cycle is 4 hours.

[0065] It should be noted that, in actual application, the size of each evaluation cycle is set according to actual conditions.

[0066] In the present application, the data acquisition module transmits the operating data collected within the evaluation cycle to the abnormal data analysis module, and uses different encryption algorithms to encrypt and transmit the operating data collected within each evaluation cycle. After receiving the data, the abnormal data analysis module uses the secret key corresponding to the encryption algorithm to decrypt the data and then analyze it.

[0067] In the embodiment of the present application, the encryption algorithm includes a symmetric encryption algorithm and an asymmetric encryption algorithm.

[0068] The present invention effectively ensures the security of data by encrypting the data transmission process, preventing tampering and causing monitoring errors.

[0069] The abnormal data analysis module involved in the present invention forms corresponding operation data sequences from various types of operation data on key components collected during an evaluation period, analyzes the false abnormal data and abnormal data sequences within the evaluation period, marks the abnormal data sequences, deletes the false abnormal data in each operation data sequence, and obtains the corresponding new operation data sequence.

[0070] In the present application, different categories of operating data on the same key component in the abnormal data analysis module form corresponding operating data sequences according to the categories they belong to; the same category of operating data on different key components form corresponding operating data sequences according to the key components they belong to.

[0071] Taking a steam turbine as an example, after deleting the false abnormal temperature data in the temperature operation data sequence of the steam chamber, a new temperature operation data sequence of the steam chamber is obtained; after deleting the false abnormal temperature data in the temperature operation data sequence of the exhaust zone, a new temperature operation data sequence of the exhaust zone is obtained; after deleting the false abnormal voltage data in the current operation data sequence of the generator, a new current operation data sequence of the generator is obtained; after deleting the false abnormal voltage data in the voltage operation data sequence of the generator, a new voltage operation data sequence of the generator is obtained.

[0072] In this application, the specific steps of analyzing false abnormal data and abnormal data sequences in the abnormal data analysis module are as follows:

[0073] S11, comparing a number of data in each operation data sequence with the corresponding preset standard threshold in sequence, and when the data in the operation data sequence exceeds the corresponding preset standard threshold, marking the data as abnormal data;

[0074] S12. When the time points of several abnormal data collection in a certain operation data sequence within the evaluation period are continuous and the continuous duration exceeds the corresponding preset duration, these several abnormal data constitute an abnormal data sequence; otherwise, they are false abnormal data.

[0075] By comparing the actual operation data with the preset standard threshold, the present invention can quickly and accurately identify data that exceeds the normal range, thereby timely marking it as abnormal, avoiding the subjectivity and uncertainty of human judgment, and improving the accuracy and consistency of data processing. In practical applications, abnormalities often do not exist in isolation, but appear continuously in some form. By judging the continuity of abnormal data in time and whether the continuous duration exceeds the preset value, the abnormal data sequence is screened out to avoid the false alarm of a single abnormal data caused by accidental factors, such as sensor errors, data transmission errors, etc., so as to improve the accuracy and reliability of abnormal detection.

[0076] It should be noted that the preset standard thresholds and preset durations of temperature, voltage, and current on key components can be obtained from the technical manual of the device, or can be set by yourself according to actual needs.

[0077] Assuming that the steam chamber temperature data, generator current data, and generator voltage data of a steam turbine are collected during a certain evaluation period, then:

[0078] 1) The steam chamber temperature data collected during the evaluation period are combined into a steam chamber temperature operation data sequence, and the steam chamber temperature data are compared with the preset standard threshold of the steam chamber temperature in turn. When a certain steam chamber temperature data exceeds the preset standard threshold of the steam chamber temperature, the steam chamber temperature data is marked as abnormal temperature data; when the time points of collecting several abnormal temperature data are continuous and the continuous time exceeds the preset time of the steam chamber temperature, this part of the abnormal temperature data constitutes an abnormal temperature data sequence;

[0079] 2) Several generator voltage data collected during the evaluation period constitute a voltage operation data sequence, and several generator voltage data are compared with the preset standard threshold of the generator voltage in turn. When a certain generator voltage data exceeds the preset standard threshold of the generator voltage, the generator voltage data is marked as abnormal voltage data; when several abnormal voltage data collection time points are continuous and the continuous duration exceeds the preset duration of the generator voltage, this part of the abnormal voltage data constitutes an abnormal voltage data sequence;

[0080] 3) Several generator current data collected during the evaluation period constitute a generator current operation data sequence, and several generator current data are compared with the preset standard threshold of the generator current in turn. When a certain generator current data exceeds the preset standard threshold of the generator current, the generator current data is marked as abnormal current data; when several abnormal current data collection time points are continuous and the continuous duration exceeds the preset duration of the generator current, this part of the abnormal current data constitutes an abnormal current data sequence.

[0081] Taking the temperature operation data sequence as an example, if the data is collected once every 1 second, the collection time point of each abnormal data is extracted. If the collection time points of n abnormal data are i, i+1, i+2, ..., i+n, respectively, then the collection time points of the n abnormal data are continuous, and n exceeds the preset temperature duration, then the n abnormal data are an abnormal data sequence.

[0082] The keyboard fault prediction module of the present invention predicts the fault information and fault level of the target device evaluation cycle based on the new operation data sequence transmitted by the abnormal data analysis module and the abnormal data sequence marked in the new operation data sequence, specifically:

[0083] ① Establish a digital twin model of the thermal power plant and run it based on the new operating data sequence to obtain fault information, where the fault information includes equipment name, model, serial number and fault location.

[0084] It should be noted that the serial number is a unique identification code assigned by the manufacturer to each device for tracking and warranty purposes.

[0085] Preferably, the specific steps of the digital twin model of the thermal power plant are as follows:

[0086] S21. Obtain various historical operation data and perform preprocessing, wherein the preprocessing includes removing invalid data, outliers and missing values, and converting the data into a unified format and unit.

[0087] It should be noted that the various types of historical operation data in step S21 include temperature historical operation data, voltage historical operation data and current historical operation data.

[0088] S22. Extract corresponding historical new operation data sequences from various historical operation data, construct various historical new operation data sequences within the same evaluation period into an operation data matrix, and obtain several historical operation data matrices, wherein the operation data matrix is ​​as follows:

[0089]

[0090] In the formula, A1~A k 、Y1~Y k A1~Y1, A k ~Y k It refers to various operating data collected at the same time point within the evaluation period.

[0091] For example, when the operating data only includes temperature, voltage and current, assuming that the evaluation period is 1 hour and the operating data is collected every 5 seconds, the operating data matrix constructed by various historical new operating data sequences within the evaluation period is as follows:

[0092]

[0093] At this time, A1~A k B1 to B2 are temperature data arranged by collection time during the entire evaluation period. k are several voltage data arranged by collection time during the entire evaluation period, C1~C k A1~C1 are the temperature, voltage and current data collected at the 0th second, A2~C2 are the temperature, voltage and current data collected at the 5th second, A k ~C k Temperature, voltage and current data collected in the first hour.

[0094] It should be noted that in this application, various types of operating data are arranged in order in the operating data matrix, that is, row A represents temperature data, row B represents voltage data, and row C represents current data. In actual application, a certain type of data represented by each row in the operating data matrix is ​​set according to actual needs. If the collected temperature data is temperature data at different key parts, such as steam chamber temperature data and exhaust zone temperature data, the steam chamber temperature data is arranged in a row according to the collection time, and the exhaust zone temperature data is also arranged in a row according to the collection time.

[0095] S23. Construct a physical model of a thermal power plant based on physical principles and engineering knowledge.

[0096] In the embodiment of the present application, the physical model of the thermal power plant includes models of equipment such as boilers, steam turbines, and generators.

[0097] S24. Combine the historical operation data matrix and the corresponding physical model to build a digital twin model of the thermal power plant.

[0098] It should be noted that the physical model corresponds to the equipment of the operating data. If the operating data is collected from the boiler, the physical model of the boiler is constructed, and the digital twin model of the thermal power plant is constructed by combining the historical operating data matrix and the physical model of the boiler; if the operating data is collected from the steam turbine, the physical model of the steam turbine is constructed, and the digital twin model of the thermal power plant is constructed by combining the historical operating data matrix and the physical model of the steam turbine.

[0099] The faulty equipment and fault location information output by the digital twin model of the present invention can provide powerful decision-making support for the management of the thermal power plant. The management can formulate more reasonable production plans, equipment update plans and personnel training plans based on this information to improve the overall operational efficiency of the power plant. Through this model, the power plant can realize real-time monitoring and data analysis of the operating status of the equipment, and promote the development of the power plant towards intelligence and automation.

[0100] ②Analyze the key influencing factors corresponding to each fault information of the thermal power plant. The specific steps are as follows:

[0101] 1) Extracting several operation data matrices corresponding to several identical fault information from the historical data, and counting the number of times abnormal data sequences appear in each new operation data sequence of the several operation data matrices;

[0102] 2) Divide the number of times obtained by statistics by the total number of times abnormal data sequences appear in several operation data matrices to obtain the abnormal proportion of operation data;

[0103] 3) When the abnormal proportion of operating data exceeds the proportion threshold, the operating data corresponding to the abnormal proportion of operating data is the key influencing factor of the corresponding fault information, and the fault information is associated with the corresponding key influencing factor.

[0104] The present invention evaluates the abnormal proportion of each operating data by counting the number of occurrences of abnormal data sequences of each operating data of a certain fault information; based on the abnormal proportion, the operating data that affects the fault information is extracted from a number of operating data, that is, the key influencing factors, and the association between the fault information and the key influencing factors is established, so as to more accurately judge the root cause of the fault, provide a strong basis for subsequent operation and maintenance management work, help optimize the operation and maintenance strategy, and improve the operation and maintenance efficiency.

[0105] Preferably, if several abnormal data sequences appear in the new running data sequence, they are counted only once.

[0106] A new operating data of the present invention may contain several abnormal data sequences. When the data volume is huge and the abnormal data sequences appear frequently, if the total number of occurrences is counted, the abnormal proportion will be excessively magnified, thus failing to accurately reflect the true situation of the data. By extracting it only once for calculation, the deviation caused by such repeated calculation can be avoided.

[0107] For example: extract several operation data matrices corresponding to a certain fault information from historical data, such as operation data matrix 1, operation data matrix 2, ..., operation data matrix N, where N is large enough; count the number of times abnormal data sequences appear in new operation data sequences such as temperature, voltage, and current in operation data matrices 1 to N, assuming that the abnormal data sequence containing temperature appears a times, the abnormal data sequence containing voltage appears b times, the abnormal data sequence containing current appears c times, and the number of times abnormal data sequences containing other operation data appear respectively;

[0108] Based on the statistical data, it is calculated that the sum of the number of abnormal data sequences of temperature to the number of abnormal data sequences of voltage, current, and other operating data is a / (a+b+c+…), that is, the abnormal proportion of temperature; accordingly, the abnormal proportion of voltage is b / (a+b+c+…), the abnormal proportion of current is C / (a+b+c+…), and so on; then, a proportion threshold is set. If the abnormal proportion of temperature and voltage exceeds the proportion threshold, temperature and voltage are the key influencing factors of the fault information, which means that there are multiple factors that lead to the occurrence of the fault, and temperature and voltage play a decisive role in the occurrence of the fault.

[0109] ③ Extract the marked abnormal data sequence from the new running data sequence, and analyze the fault level based on the abnormal data sequence.

[0110] In the present application, the specific steps of analyzing the fault level based on the abnormal data sequence are: performing mean processing on the data in the abnormal data sequence, and performing difference processing with the preset standard threshold to obtain the deviation difference. When the deviation difference is greater than the deviation threshold, the fault level is level one; otherwise, the fault level is level two, where the fault degree is level one > level two.

[0111] Preferably, when there are several abnormal data sequences in the new operation data sequence, the maximum deviation difference is compared with the deviation threshold to evaluate the fault level.

[0112] ④ Based on key influencing factors and fault levels, a monitoring fault prediction model is constructed to predict the fault information and fault level of the evaluation period.

[0113] In this application, a monitoring fault prediction model is constructed based on key influencing factors and fault levels. The specific steps are as follows:

[0114] 1) Extract several abnormal data sequences, corresponding fault information and fault levels from historical data; use the abnormal data sequences as input data of the artificial intelligence model, and the corresponding fault information and fault levels as output data of the artificial intelligence;

[0115] 2) The artificial intelligence model is trained through input data and output data to obtain a monitoring fault prediction model, where the artificial intelligence model is built based on the LSTM model;

[0116] 3) Input the abnormal data sequence of the evaluation period into the monitoring fault prediction model, and predict and output the fault information and fault level of the evaluation period.

[0117] The present invention provides rich training samples for the artificial intelligence model by extracting abnormal data sequences and their corresponding fault information from historical data, so that the model can learn the characteristics of fault data, thereby improving the accuracy of prediction; secondly, the LSTM model is particularly suitable for processing time series data. Since fault data often has time series characteristics, the LSTM model can capture long-term dependencies in the data, which is crucial for predicting future fault trends; therefore, the monitoring fault prediction model constructed based on LSTM has significant advantages in prediction performance; the monitoring fault prediction model can realize the accurate prediction of fault information and fault level of the evaluation cycle, which helps operation and maintenance personnel to take timely measures to avoid further evolution of faults.

[0118] For example, if a steam turbine in a thermal power plant has a temperature anomaly during the evaluation period, the temperature anomaly data sequence is input into the monitoring panel fault prediction model, and the model predicts and outputs the fault information of the steam turbine: basic information of the steam turbine such as the model and serial number of the steam turbine, the fault location is the exhaust area, and the fault level is level two.

[0119] The fault processing module of the present invention dispatches a maintenance unit matching the fault level to repair the fault based on the fault information and fault level predicted by the keyboard fault prediction module.

[0120] In this application, the specific steps for the rating of each maintenance unit are as follows:

[0121] S31. Based on the number of times each maintenance unit successfully repairs a primary fault, calculate the maintenance success ratio of the maintenance unit. The calculation formula is as follows:

[0122]

[0123] In the formula, Δ i represents the maintenance success rate of the i-th maintenance unit, E i It represents the number of times the i-th maintenance unit successfully repairs the first-level fault, and e represents the number of maintenance units.

[0124] Assume that a thermal power plant has three maintenance units, which are numbered A1, A2, and A3 respectively. Each maintenance unit has three people. The personnel of Group A1 are numbered A11, A12, and A13; the personnel of Group A2 are numbered A21, A22, and A23; and the personnel of Group A3 are numbered A31, A32, and A33.

[0125] Assuming that the number of times each group repairs the first-level fault is: 3 times for Group A1, 2 times for Group A2, and 1 time for Group A3, the repair success rate of each unit is: Group A1 = (3 / (3+2+1))*100% = 50%, Group A2 = (2 / (3+2+1))*100% = 33.3%, Group A3 = (1 / (3+2+1))*100% = 16.7%.

[0126] S32. Based on the skill types and corresponding skill proficiency of all maintenance workers in each maintenance unit, the average skill proficiency of all maintenance workers in each maintenance unit is calculated, wherein the skill proficiency is obtained by expert scoring, and the calculation formula is as follows:

[0127]

[0128] In the formula, represents the average skill proficiency of the jth maintenance worker in the i-th maintenance unit, f k represents the proficiency of the kth skill type, and k represents the number of skill types.

[0129] It should be noted that the higher the skill proficiency score, the more proficient the maintenance worker's maintenance skills are.

[0130] In addition, there is a cap on the maximum expert score in this application.

[0131] S33. Based on the average skill proficiency of all maintenance workers in each maintenance unit, the average skill proficiency of each maintenance unit is calculated. The calculation formula is as follows:

[0132]

[0133] In the formula, R i represents the average skill proficiency of the maintenance group i, j i Represents the number of maintenance workers in the i-th maintenance unit.

[0134] Assume that there are three skills required for repairing a first-level fault, such as W1, W2, and W3; the highest score for the expert score is 10 points, and the experts score the proficiency of A11 in group A1 as 5, 2, and 3 respectively; A12: 2, 2, 3; A13: 3, 3, 2; the average proficiency of A1 in the W1 skill is (5+2+3) / 3=3.33, the average proficiency in the W2 skill is (2+2+3) / 3=2.33, and the average proficiency in the W2 skill is (3+3+2) / 3=2.66; the overall proficiency of group A1 in its three skills, that is, the average skill proficiency = (3.33+2.33+2.66) / 3=2.77; in addition, according to this method, groups A2 and A3 are calculated, and assuming that the average skill proficiency calculated is 2.12 and 2.5 respectively, it can be seen that group A1 has the highest average skill proficiency.

[0135] S34: Based on the maintenance success ratio of each maintenance unit in step S31 and the average skill proficiency of each maintenance unit in step S33, the level of each maintenance unit is evaluated, specifically:

[0136] If the maintenance success ratio of each maintenance unit is greater than the ratio threshold, and the average skill proficiency of each maintenance unit is greater than the proficiency threshold, the maintenance unit is evaluated as a first-level maintenance unit and repairs a first-level fault; otherwise, the maintenance unit is evaluated as a second-level maintenance unit and repairs a second-level fault.

[0137] The present invention can more accurately evaluate the maintenance capability of the maintenance unit by extracting the number of successful repairs of primary faults and related skill information, which is an important basis for evaluating whether the unit is competent for specific maintenance tasks; and realizes a clear division of maintenance tasks; wherein, the primary maintenance unit focuses on the maintenance of primary faults, while the secondary maintenance unit is responsible for the maintenance of secondary faults. The clear division of labor not only improves the maintenance efficiency, but also ensures the maintenance quality, because each unit can focus on its own areas of expertise, thereby exerting the best performance.

[0138] The present application divides the maintenance units into levels and matches each level of fault with a maintenance unit of the corresponding level, i.e., a level one maintenance unit repairs level one faults, and a level two maintenance unit repairs level two faults. This helps to reasonably allocate maintenance resources, avoid waste of resources, and improve the overall efficiency of maintenance work.

[0139] It should be noted that if the maintenance success rate of each maintenance unit is calculated only based on the number of times each maintenance unit successfully repairs the first-level fault, and the maintenance unit level is divided according to the maintenance success rate and average skill proficiency of each maintenance unit, and these levels are directly matched with the fault level, it may cause the second-level maintenance unit to be "solidified" on the maintenance task of the second-level fault, lacking the opportunity to promote to a higher level of fault maintenance. This may not only inhibit the skill improvement and enthusiasm of the maintenance unit, but also lead to unreasonable and inefficient resource allocation. Therefore, it is also necessary to set up a method for each maintenance unit to improve the maintenance level. For example, the maintenance unit is re-evaluated regularly, not only based on the past maintenance success rate and skill proficiency, but also considering the performance of the unit in training, skill improvement, new technology application, etc. According to the evaluation results, the level of the maintenance unit is dynamically adjusted to give the second-level maintenance unit the opportunity to be promoted to the first-level fault maintenance. Since the methods for promoting the level of maintenance units in each plant are slightly different, in actual application, the methods for promoting the level of maintenance units of each level can be set according to the actual situation, and no specific restrictions are made in this application.

[0140] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0141] The above embodiments are only used to illustrate the technical method 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 method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. The intelligent monitoring panel fault prediction system for thermal power plants based on big data is characterized by: include: A data acquisition module is connected to a plurality of sensors deployed on key components of target equipment in a thermal power plant, and is used to collect a number of operating data on the key components of the target equipment in real time, and transmit the operating data collected within an evaluation period to an abnormal data analysis module based on a preset evaluation period, wherein the operating data includes temperature, voltage and current; The abnormal data analysis module organizes various types of operating data on each key component collected during the evaluation period into corresponding operating data sequences, analyzes the false abnormal data and abnormal data sequences during the evaluation period, marks the abnormal data sequences, deletes the false abnormal data in each operating data sequence, and obtains the corresponding new operating data sequence; The keyboard fault prediction module predicts the fault information and fault level of the target device evaluation cycle based on the new operation data sequence transmitted by the abnormal data analysis module and the abnormal data sequence marked in the new operation data sequence, specifically: ① Establish a digital twin model of the thermal power plant and run it based on the new operation data sequence to obtain fault information, where the fault information includes equipment name, model, serial number and fault location; ②Analyze the key influencing factors corresponding to each fault information of the thermal power plant; ③ Extract the marked abnormal data sequence from the new running data sequence, and analyze the fault level based on the abnormal data sequence; ④ Based on key influencing factors and fault levels, a monitoring fault prediction model is constructed to predict the fault information and fault level of the evaluation period; The fault processing module dispatches a maintenance unit matching the fault level to repair the fault based on the fault information and fault level predicted by the keyboard fault prediction module.

2. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 1 is characterized in that: The evaluation cycle is a number of time periods after a day is evenly divided, and each time period is an evaluation cycle; In addition, when the data acquisition module transmits the operating data collected during the evaluation cycle to the abnormal data analysis module, different encryption algorithms are used to encrypt and transmit the operating data collected during each evaluation cycle. After receiving the data, the abnormal data analysis module uses the secret key corresponding to the encryption algorithm to decrypt the data and then analyze it.

3. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 1 is characterized in that: Different categories of operating data on the same key component in the abnormal data analysis module form corresponding operating data sequences according to their categories; the same category of operating data on different key components form corresponding operating data sequences according to their key components; In addition, the specific steps for analyzing false abnormal data and abnormal data sequences in the abnormal data analysis module are as follows: S11, comparing a number of data in each operation data sequence with the corresponding preset standard threshold in sequence, and when the data in the operation data sequence exceeds the corresponding preset standard threshold, marking the data as abnormal data; S12. When the time points of several abnormal data collection in a certain operation data sequence within the evaluation period are continuous and the continuous duration exceeds the corresponding preset duration, these several abnormal data constitute an abnormal data sequence; otherwise, they are false abnormal data.

4. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 1 is characterized in that: The specific steps of the digital twin model of a thermal power plant are as follows: S21. Obtain various historical operation data and perform preprocessing, wherein the preprocessing includes removing invalid data, outliers and missing values, and converting the data into a unified format and unit; S22. Extract corresponding historical new operation data sequences from various historical operation data, construct various historical new operation data sequences within the same evaluation period into an operation data matrix, and obtain several historical operation data matrices, wherein the operation data matrix is ​​as follows: In the formula, A1~A k 、Y1~Y k A1~Y1, A k ~Y k It is various types of operation data collected at the same time point within the evaluation period; S23. Construct a physical model of a thermal power plant based on physical principles and engineering knowledge; S24. Combine the historical operation data matrix and the corresponding physical model to build a digital twin model of the thermal power plant.

5. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 1 is characterized in that: The specific steps for analyzing the key influencing factors corresponding to the fault information are as follows: 1) Extracting several operation data matrices corresponding to several identical fault information from the historical data, and counting the number of times abnormal data sequences appear in each new operation data sequence of the several operation data matrices; 2) Divide the number of times obtained by statistics by the total number of times abnormal data sequences appear in several operation data matrices to obtain the abnormal proportion of operation data; 3) When the abnormal proportion of operating data exceeds the proportion threshold, the operating data corresponding to the abnormal proportion of operating data is the key influencing factor of the corresponding fault information, and the fault information is associated with the corresponding key influencing factor.

6. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 5 is characterized in that: If several abnormal data sequences appear in the new running data sequence, they are only counted once.

7. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 1 is characterized in that: The specific steps of analyzing the fault level based on the abnormal data sequence are as follows: the data in the abnormal data sequence are processed by averaging, and the data are processed by difference with the preset standard threshold value to obtain the deviation difference. When the deviation difference is greater than the deviation threshold value, the fault level is level one; Otherwise, the fault level is level 2, where the fault degree is level 1 > level 2.

8. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 7 is characterized in that: When there are several abnormal data sequences in the new running data sequence, the maximum deviation difference is compared with the deviation threshold to evaluate the fault level.

9. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 1 is characterized in that: The specific steps of building a monitoring fault prediction model based on key influencing factors and fault levels are as follows: 1) Extract several abnormal data sequences, corresponding fault information and fault levels from historical data; use the abnormal data sequences as input data of the artificial intelligence model, and the corresponding fault information and fault levels as output data of the artificial intelligence; 2) The artificial intelligence model is trained through input data and output data to obtain a monitoring fault prediction model, where the artificial intelligence model is built based on the LSTM model; 3) Input the abnormal data sequence of the evaluation period into the monitoring fault prediction model, and predict and output the fault information and fault level of the evaluation period.

10. The intelligent monitoring panel fault prediction system for thermal power plants based on big data according to claim 1 is characterized in that: The specific steps for rating each maintenance unit are as follows: S31. Based on the number of times each maintenance unit successfully repairs a primary fault, calculate the maintenance success ratio of the maintenance unit. The calculation formula is as follows: In the formula, Δ i represents the maintenance success rate of the i-th maintenance unit, E i represents the number of times the i-th maintenance unit successfully repairs the first-level fault, and e represents the number of maintenance units; S32. Based on the skill types and corresponding skill proficiency of all maintenance workers in each maintenance unit, the average skill proficiency of all maintenance workers in each maintenance unit is calculated, wherein the skill proficiency is obtained by expert scoring, and the calculation formula is as follows: In the formula, r i j represents the average skill proficiency of the jth maintenance worker in the i-th maintenance unit, f k represents the proficiency of the kth skill type, where k represents the number of skill types; S33. Based on the average skill proficiency of all maintenance workers in each maintenance unit, the average skill proficiency of each maintenance unit is calculated. The calculation formula is as follows: In the formula, R i represents the average skill proficiency of the maintenance group i, j i represents the number of maintenance workers in the i-th maintenance unit; S34: Based on the maintenance success ratio of each maintenance unit in step S31 and the average skill proficiency of each maintenance unit in step S33, the level of each maintenance unit is evaluated, specifically: If the maintenance success ratio of each maintenance unit is greater than the ratio threshold, and the average skill proficiency of each maintenance unit is greater than the proficiency threshold, the maintenance unit is evaluated as a first-level maintenance unit and repairs a first-level fault; otherwise, the maintenance unit is evaluated as a second-level maintenance unit and repairs a second-level fault.

Images