Real-time dynamic fault diagnosis system and method for auxiliary control system of thermal power plant

Abstract

The invention relates to the technical field of fault diagnosis, and particularly discloses a real-time dynamic fault diagnosis system and method for an auxiliary control system of a thermal power plant, and the system carries out the abnormality diagnosis of the operation data of a boiler system through the pre-configuration of an abnormality rule and the generation of a target monitoring program based on the abnormality rule. Specifically, firstly, a deep learning-based data processing technology is utilized to carry out time sequence analysis on operation data of the boiler system to extract a time sequence change mode of the operation data of the boiler system, and then, time sequence operation mode characteristics of the boiler system and a pre-configured abnormal rule are subjected to rapid query positioning semantic interaction matching, so that the abnormal rule of the boiler system is obtained. Therefore, the abnormality diagnosis of the boiler system is realized, and the operation data of the preset time period before and after the abnormality is stored. By means of the mode, abnormal changes in the operation process of the boiler system can be more accurately captured, and the accuracy and timeliness of fault diagnosis are improved.

Description

Technical Field

[0001] This application relates to the technical field of fault diagnosis, and more specifically, to a real-time dynamic fault diagnosis system and method for the auxiliary control system of a thermal power plant. Background Art

[0002] Thermal power plants are an important part of electric power production, and their operating efficiency and safety are crucial for the country's energy supply. During the thermal power generation process, as one of the core devices, the boiler undertakes the key task of converting the chemical energy of fuel into heat energy. However, due to the complex structure of the boiler system, harsh working environment, and long-term continuous operation, the risk of faults is relatively high, which not only affects the normal operation of the power plant but also may pose potential safety hazards.

[0003] The existing monitoring means for boiler systems mainly rely on traditional SCADA (Supervisory Control And Data Acquisition) systems. Real-time operating parameters are collected through sensors, and an abnormal situation is detected based on a fixed threshold alarm mechanism. This method has obvious limitations. For example, the operating state of the boiler system is often affected by multiple complex factors, such as fuel quality, load changes, environmental temperature, etc. These factors may cause the actual operating parameters to fluctuate within the normal range, but they may also hide potential fault risks. This fixed threshold alarm mechanism often has difficulty accurately capturing these subtle abnormal changes, resulting in frequent false alarms or missed alarms.

[0004] Therefore, an optimized real-time dynamic fault diagnosis system and method for the auxiliary control system of a thermal power plant are expected. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide a real-time dynamic fault diagnosis system and method for the auxiliary control system of a thermal power plant, which perform abnormal diagnosis on the operating data of the boiler system by pre-configuring abnormal rules and generating a target monitoring program based on the abnormal rules. Specifically, first, time series analysis is performed on the operating data of the boiler system by using data processing technology based on deep learning to extract the time series change patterns of the operating data of the boiler system. Then, through rapid query and positioning semantic interaction matching between the time series operating mode characteristics of the boiler system and the pre-configured abnormal rules, abnormal diagnosis of the boiler system is realized, and the operating data in a preset time period before and after the abnormality is saved. In this way, abnormal changes during the operation of the boiler system can be captured more accurately, and the accuracy and timeliness of fault diagnosis can be improved.

[0006] Correspondingly, according to one aspect of this application, a real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant is provided, which includes:

[0007] Pre-configure parameters and exception rules;

[0008] Generate a target monitoring program according to the exception rules;

[0009] Configure the target monitoring program to the controller of the boiler system;

[0010] After loading and running the target monitoring program, the controller collects the operation data of the boiler system;

[0011] The controller monitors the operation data of the boiler system in real time according to the exception rules to obtain a judgment result;

[0012] When the judgment result is an exception, save the operation data in a preset time period before and after the exception.

[0013] In the above real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, the controller monitors the operation data of the boiler system in real time according to the exception rules to obtain a judgment result, including: extracting the sequential operation mode characteristics of the operation data of the boiler system to obtain a boiler system operation sequential characteristic coding vector; respectively extracting the semantic characteristics of each exception rule to obtain a set of exception rule semantic coding vectors; performing a fast location semantic query response coding on the boiler system operation sequential characteristic coding vector and the set of exception rule semantic coding vectors to obtain a boiler system operation state query response semantic coding vector; and determining the judgment result based on the boiler system operation state query response semantic coding vector.

[0014] According to another aspect of the present application, there is provided a real-time dynamic fault diagnosis system for the auxiliary control system of a thermal power plant, which includes:

[0015] An exception rule configuration module, configured to pre-configure parameters and exception rules;

[0016] A monitoring program generation module, configured to generate a target monitoring program according to the exception rules;

[0017] A program configuration module, configured to configure the target monitoring program to the controller of the boiler system;

[0018] An operation data collection module, configured to, after loading and running the target monitoring program, the controller collects the operation data of the boiler system;

[0019] An operation data exception judgment module, configured to the controller monitors the operation data of the boiler system in real time according to the exception rules to obtain a judgment result;

[0020] An abnormal operation data saving module, configured to save the operation data in a preset time period before and after the exception when the judgment result is an exception.

[0021] Compared with the prior art, the real-time dynamic fault diagnosis system and method for the auxiliary control system of a thermal power plant provided by the present application pre-configures abnormal rules and generates a target monitoring program based on the abnormal rules to perform abnormal diagnosis on the operation data of the boiler system. Specifically, first, a time series analysis is performed on the operation data of the boiler system by using a data processing technology based on deep learning to extract the time series change patterns of the operation data of the boiler system. Then, by performing a fast query and positioning semantic interaction match between the time series operation mode features of the boiler system and the pre-configured abnormal rules, the abnormal diagnosis of the boiler system is realized, and the operation data in a preset time period before and after the abnormality is saved. In this way, the abnormal changes during the operation of the boiler system can be captured more accurately, and the accuracy and timeliness of fault diagnosis can be improved. Description of the Drawings

[0022] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0023] Figure 1 It is a flowchart of the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application.

[0024] Figure 2 It is a flowchart of step S5 in the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application.

[0025] Figure 3 It is a schematic diagram of data flow in step S5 of the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application.

[0026] Figure 4 It is a flowchart of step S53 in the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application.

[0027] Figure 5 It is a block diagram of the real-time dynamic fault diagnosis system for the auxiliary control system of a thermal power plant according to an embodiment of the present application. Detailed Description of the Embodiments

[0028] Next, exemplary embodiments according to the present application will be described in detail with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0029] Figure 1 This is a flowchart of a real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application. As Figure 1 shown, the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application includes the steps of: S1, pre-configuring parameters and anomaly rules; S2, generating a target monitoring program according to the anomaly rules; S3, configuring the target monitoring program to the controller of the boiler system; S4, after loading and running the target monitoring program, the controller collects the operation data of the boiler system; S5, the controller monitors the operation data of the boiler system in real time according to the anomaly rules to obtain a judgment result; S6, when the judgment result is an anomaly, save the operation data in a preset time period before and after the anomaly.

[0030] In the above real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, in step S1, parameters and anomaly rules are pre-configured. It should be understood that the auxiliary control system of a thermal power plant includes many complex devices and technological processes, and the operating parameters of different devices are different. For example, the normal range of parameters such as the pressure, temperature, and water level of a boiler needs to be clearly defined. Therefore, based on in-depth research on the auxiliary control system of thermal power generation and the accumulation of a large amount of operation data, combined with the device design specifications and safety operation requirements, this application determines the judgment criteria for abnormal situations such as too high temperature and sudden pressure change and the corresponding parameter thresholds, forms the anomaly judgment rules of the system, and defines alarm result variables such as alarm signal triggering, completes the configuration of parameters and anomaly rules, helps to provide an accurate basis for subsequent fault diagnosis, standardizes the fault judgment criteria from the source, and improves the pertinence and effectiveness of system fault diagnosis.

[0031] Specifically, in terms of parameter configuration, for the boiler system, it is necessary to refer to the technical manuals and design specifications provided by the equipment manufacturer, and detail the determination of the normal value ranges and fluctuation thresholds of key operating parameters such as steam pressure, steam temperature, water level height, and combustion efficiency. For example, according to the model and rated working conditions of the boiler, the normal operating range of steam pressure may be set at 8 - 10 MPa, and the water level height should be maintained between specific upper and lower limits, such as -50 mm to +50 mm. Through in-depth analysis of historical operation data and using statistical methods to calculate statistical characteristics such as the mean and standard deviation of each parameter, the reasonable range of parameters is further optimized based on this.

[0032] During the abnormal rule configuration process, all possible fault situations should be comprehensively considered. For the situation of abnormal temperature rise, it can be set that when the boiler steam temperature rises by more than 10°C within a short period (such as 5 minutes), or when the steam turbine inlet steam temperature exceeds the design upper limit by 5°C, the abnormal rule is triggered. This rule is set based on the consideration of the high-temperature resistance performance of equipment materials and the safe operation of the system. Excessive temperature may cause damage to equipment components and a decrease in system efficiency. For abnormal pressure fluctuations, it may indicate pipeline leakage, valve failure or other serious problems. For example, when the boiler steam pressure fluctuates by more than 0.5 MPa within 10 minutes, or when the steam turbine exhaust pressure suddenly drops by more than 0.3 MPa, it is determined as abnormal. In addition, abnormal changes in water level may cause serious accidents such as water hammer and dry burning, threatening the safety of equipment. Therefore, for abnormal water level, it can be set that when the boiler water level continuously drops by more than 30 mm or rises by more than 40 mm within 15 minutes, the abnormal rule is started. In terms of combustion, if the combustion efficiency is lower than 90% (referring to the designed combustion efficiency) within 30 minutes of continuous operation, or the oxygen content in the flue gas deviates from the normal range (such as the normal range is 3%-5%, outside this range), it is considered abnormal. This helps to promptly detect problems in the combustion process, such as incomplete fuel combustion and improper air volume adjustment, and avoid energy waste and environmental pollution.

[0033] During the implementation process, an advanced monitoring and analysis software platform can be used to integrate real-time data and historical data from various equipment sensors. Through the software's visualization interface, configuration personnel can conveniently input and adjust parameters and abnormal rules. At the same time, the software should have data verification and error correction functions to check the rationality of the input parameters and rules during the configuration process, such as checking whether the parameter thresholds are within a reasonable range and whether the logic of the abnormal rules is rigorous. Once an error or irrationality is found, the configuration personnel are promptly prompted to make corrections.

[0034] In addition, to ensure the reliability of the configuration, sufficient testing and verification are also required. In a simulated environment, the configured parameters and abnormal rules are tested using historical fault data and simulated fault scenarios. Observe whether the system can accurately identify faults and issue alarms in a timely manner, and conduct a detailed analysis and evaluation of the test results. According to the test feedback, further optimize and improve the configuration of parameters and abnormal rules to ensure that the fault diagnosis function can be effectively exerted during actual operation and guarantee the safe and stable operation of the auxiliary control system of the thermal power plant.

[0035] In the above real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, in step S2, a target monitoring program is generated according to the abnormal rules. That is, by converting the abnormal rules into program codes that can be executed on the controller, automated and intelligent fault monitoring is achieved, reducing the burden of manual monitoring. In specific implementation, specific programming tools and compilation environments are used to parse the configured abnormal rules. For example, for the abnormal judgment rule of boiler water level, if the water level is lower than the set lower limit value, an alarm is triggered. A corresponding conditional judgment statement and logical control flow are programmed and generated, and they are converted into a machine language instruction sequence that can be recognized and processed by the controller. The water level data storage address and the alarm variable address are associated, and a target monitoring program that can run on the controller is linked and generated. In this way, the executability of the abnormal rules can be achieved, enhancing the timeliness and reliability of system fault monitoring.

[0036] In the above real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, in step S3, the target monitoring program is configured to the controller of the boiler system. Specifically, using network communication technology, a stable data transmission channel is established between the upper computer and the controller, and the target monitoring program is transmitted to the storage unit of the controller and installed and deployed to ensure that the program can run normally, enabling the controller to analyze and judge the operation data of the boiler system according to the established program, realizing the local processing of fault monitoring, reducing data transmission delay, and thus helping to detect fault signs in a timely manner and improving the system response speed.

[0037] In the above real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, in step S4, after loading and running the target monitoring program, the controller collects the operation data of the boiler system. It should be understood that the operation data of the boiler system is the basis for judging whether it is operating normally. The controller uses its built-in data acquisition module to connect with various sensors of the boiler system (such as temperature sensors, pressure sensors, flow sensors, etc.), and obtains the real-time operation data of the system according to the predetermined sampling frequency and data transmission protocol, and transmits it to the internal buffer area of the controller for the monitoring program to analyze, thus providing real-time and accurate data support for fault diagnosis and ensuring the timeliness and reliability of the monitoring program to analyze data.

[0038] In specific implementation, a dedicated data acquisition hardware circuit is integrated inside the controller and connected to various sensors of the boiler system. At startup, the controller initializes the data acquisition channels according to the pre-set parameters. For example, for a temperature sensor, parameters such as the sampling accuracy and sampling frequency of its corresponding analog signal input channel are set. Usually, for the acquisition of key temperature parameters, the sampling accuracy may be set to 12 bits or higher, and the sampling frequency may be set to 10 times per second or even more densely to ensure that the subtle changes in temperature can be accurately captured.

[0039] After the hardware configuration is completed, the controller will establish a stable communication connection with the sensors of the boiler system. There are a large number of sensors distributed in the boiler system, such as pressure sensors, water level sensors, flow sensors, etc., which transmit data to the controller in a wired or wireless manner. For wired connections, common communication protocols such as industrial Ethernet and RS485 are often used. The controller will send handshake signals and configuration instructions to the sensors according to the specifications of the corresponding protocols to ensure that both parties can correctly identify and communicate. For example, when using the RS485 protocol, the controller will set appropriate baud rates (such as 9600bps), data bits (usually 8 bits), stop bits (1 bit), and parity bits (such as no parity), and send address recognition instructions to the sensors so that the sensors can respond to the data requests of the controller.

[0040] Once the communication connection is successfully established, the controller will start the data acquisition task according to the pre-determined program logic. During the acquisition process, the controller will obtain data from each sensor in sequence according to the set time interval and order. Taking the boiler pressure data acquisition as an example, the controller first sends a data reading instruction to the pressure sensor. After receiving the instruction, the sensor converts the currently detected pressure value into an electrical signal (such as a 4-20mA current signal or a 0-5V voltage signal) and transmits it to the controller. The data acquisition module of the controller will convert the received analog signal into a digital quantity through analog-to-digital conversion and store and process it in a certain data format.

[0041] During the data acquisition process, in order to ensure the accuracy and integrity of the data, the controller will also implement a series of data verification and error handling mechanisms. For the acquired data, the controller will perform range verification. For example, if the normal range of the boiler pressure is known to be 8-10MPa, and the acquired data exceeds this range, the controller will mark it as suspicious data and further analyze and verify it. Multiple repeated acquisitions may be adopted. If the results of multiple acquisitions all exceed the reasonable range, an alarm signal will be triggered to prompt the operation and maintenance personnel to check whether there is a fault in the sensor or the system. At the same time, the controller will detect and correct errors during the data transmission process. For example, when using industrial Ethernet communication, the checksum mechanism in the network protocol is used to perform integrity verification on the received data. If it is found that the data has errors during transmission, the sensor will be requested to re-send the data to ensure reliable reception of the data.

[0042] In addition, while collecting data, the controller also performs preliminary preprocessing and collation on the data. For the collected raw data, filtering may be performed to remove noise interference. For example, a moving average filtering algorithm is used to calculate the average of the data at multiple consecutive sampling points to smooth the data curve and reduce data fluctuations caused by the noise of the sensor itself or external interference. At the same time, the controller sorts and stores the collected data according to the time series, facilitating subsequent analysis and processing by the target monitoring program.

[0043] In the above real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, in step S5, the controller monitors the operation data of the boiler system in real time according to the abnormal rules to obtain a judgment result. That is, the controller runs the target monitoring program, compares and analyzes the collected real-time operation data with the pre-stored abnormal rules (such as the safety rule list), and determines whether the data is within the normal range according to the set logical judgment conditions, thereby obtaining a judgment result on whether the system operation is abnormal.

[0044] Specifically, Figure 2 FIG. is a flowchart of step S5 in the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application. Figure 3 FIG. is a schematic diagram of data flow in step S5 in the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application. As Figure 2 and Figure 3 shown, step S5 includes: S51, extracting the time-series operation mode characteristics of the operation data of the boiler system to obtain a boiler system operation time-series feature coding vector; S52, respectively extracting the semantic characteristics of each abnormal rule to obtain a set of abnormal rule semantic coding vectors; S53, performing a fast location semantic query response coding on the boiler system operation time-series feature coding vector and the set of abnormal rule semantic coding vectors to obtain a boiler system operation state query response semantic coding vector; S54, determining the judgment result based on the boiler system operation state query response semantic coding vector.

[0045] Specifically, in step S51, the temporal operation mode features of the operation data of the boiler system are extracted to obtain the boiler system operation temporal feature coding vector. In a specific example of the present application, the operation data of the boiler system is input into a boiler system operation mode feature extractor based on the RNN-LSTM model to obtain the boiler system operation temporal feature coding vector. It should be understood that since the operation data of the boiler system has time series characteristics, such as parameters like temperature and pressure changing dynamically over time, the temporal changes in this operation data often contain important information about the equipment operation state. Therefore, in the technical solution of the present application, the RNN-LSTM model is used to perform time series analysis on the operation data of the boiler system to extract the temporal operation mode features of the boiler system. It should be known that the RNN (Recurrent Neural Network) model can process sequence data through its recurrent connection structure, and the internal hidden layer state can transmit the information of the previous time step to the next time step, thereby capturing the temporal dependence relationship in the data. The LSTM (Long Short-Term Memory) model, by introducing a gating mechanism (input gate, forget gate, and output gate), can better handle the long-term dependence problem in long sequences and avoid gradient vanishing or gradient explosion. Based on this, by using the RNN-LSTM model to construct a boiler system operation mode feature extractor in the present application, the recurrent connection structure of the RNN model and the gating mechanism of the LSTM model can be comprehensively utilized to deeply mine the operation data of the boiler system, effectively learn and remember the temporal change trend of the operation data of the boiler system, capture the temporal dependence relationship in the data, and then accurately depict the operation mode of the boiler system, thereby generating the boiler system operation temporal feature coding vector.

[0046] Specifically, in step S52, semantic features of each of the abnormal rules are extracted respectively to obtain a set of semantic encoding vectors of the abnormal rules. In a specific example of the present application, a semantic encoder based on the Bert model is used to perform semantic encoding on each of the abnormal rules to obtain the set of semantic encoding vectors of the abnormal rules. It should be understood that since abnormal rules usually exist in the form of text or logical expressions, it is relatively difficult to directly interact and compare with the operation data of the boiler system. Therefore, in order to achieve fast and accurate matching between the abnormal rules and the operation data of the boiler system, the present application further uses a semantic encoder based on the Bert model to perform semantic encoding on each of the abnormal rules respectively, so as to capture the deep semantic information in the abnormal rule text, generate corresponding semantic encoding vectors of the abnormal rules, and enable them to perform operations and comparisons in the same mathematical space as the operation characteristics of the boiler system, thereby achieving more accurate fault judgment. It should be known that the Bert model (Bidirectional Encoder Representations from Transformers) is a bidirectional encoding representation model based on the Transformer structure. Through pre-training on a large corpus, it can deeply understand the text semantics and generate high-quality semantic representation vectors. In the present application, using the Bert model to perform semantic encoding on the abnormal rules can effectively extract the deep context semantic meaning of the abnormal rules, achieve accurate semantic understanding of the abnormal rules, and convert them into high-dimensional vector forms, thereby providing strong semantic support for subsequent fault judgment and improving the intelligent level and accuracy of fault diagnosis.

[0047] Specifically, in step S53, fast positioning semantic query response encoding is performed on the set of the operation time-series feature encoding vectors of the boiler system and the set of the semantic encoding vectors of the abnormal rules to obtain a semantic encoding vector of the query response for the operation state of the boiler system. Further, by performing semantic query matching on the set of the operation time-series feature encoding vectors of the boiler system and the set of the semantic encoding vectors of the abnormal rules, it is determined whether the current operation state of the boiler system conforms to the abnormal rules. In particular, in order to improve the efficiency and accuracy of semantic query matching, the present application proposes an efficient semantic query positioning mechanism to achieve fast search and matching interaction between the operation data features and the abnormal rule semantics, and dynamically generate a query response, thereby providing a key information basis for the final fault judgment.

[0048] Figure 4 The flowchart of step S53 in the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to an embodiment of the present application is as follows. As Figure 4 shown, step S53 includes:

[0049] S531. Determine the central abnormal rule semantic coding vector of the set of abnormal rule semantic coding vectors based on the semantic differences between the operation timing feature coding vector of the boiler system and each abnormal rule semantic coding vector in the set of abnormal rule semantic coding vectors; S532. Determine the fine-grained matching search window for abnormal rules based on the central abnormal rule semantic coding vector; S533. Perform semantic query response interaction coding on the operation timing feature coding vector of the boiler system and the abnormal rule semantic coding vector within the fine-grained matching search window for abnormal rules to obtain the query response semantic coding vector for the operation state of the boiler system.

[0050] Specifically, step S531 includes: calculating the cross entropy of the operation timing feature coding vector of the boiler system with respect to each abnormal rule semantic coding vector in the set of abnormal rule semantic coding vectors to obtain a set of abnormal rule quick query positioning factors; taking the abnormal rule semantic coding vector corresponding to the minimum value in the set of abnormal rule quick query positioning factors as the central abnormal rule semantic coding vector, which is expressed by the formula:

[0051] D = {d1, d2,..., d i ,..., d n},

[0052]

[0053] F = {H(q, d1), H(q, d2),..., H(q, d i ),..., H(q, d n )},

[0054]

[0055] where D represents the set of abnormal rule semantic coding vectors, d1, d2, d i and d n respectively represent the 1st, 2nd, ith, and nth abnormal rule semantic coding vectors in the set of abnormal rule semantic coding vectors, m is the number of feature vectors in the set of abnormal rule semantic coding vectors, represents the feature value at the kth position in the ith abnormal rule semantic coding vector, q represents the operation timing feature coding vector of the boiler system, q k represents the feature value at the kth position in the operation timing feature coding vector of the boiler system, log2(·) represents the logarithmic function with base 2, and H(q, d1), H(q, d2), H(q, d i ) and H(q, d n ) respectively represent the H of q with respect to d1, d2, di and the said d n cross-entropy, F represents the set of anomaly rule fast query location factors, argmin represents taking the anomaly rule semantic encoding vector corresponding to the minimum value in the set of anomaly rule fast query location factors, and d * represents the location center anomaly rule semantic encoding vector.

[0056] Specifically, considering that there may be some rules among a large number of anomaly rules that are not directly related or have a low correlation with the current operating state of the boiler system. Therefore, in order to further optimize the fault monitoring process and reduce unnecessary computational overhead, this application first uses the cross-entropy metric to quantify the semantic distance and mismatch degree between the operation time series feature encoding vector of the boiler system and each anomaly rule semantic encoding vector, generating a set of anomaly rule fast query location factors to guide the search process to make it closer to the target query area of the anomaly rule and provide an accurate reference for subsequent fast query location. Then, by selecting the minimum value in the set of anomaly rule fast query location factors obtained from the above calculation, the anomaly rule that best matches the current operating state of the boiler system is determined to ensure that the initially screened anomaly rule semantic encoding vector has a high semantic correlation with the current operating state of the boiler system.

[0057] Specifically, in the step S532, the vector at the center position of the anomaly rule fine-grained matching search window is the location center anomaly rule semantic encoding vector, and each anomaly rule semantic encoding vector in the anomaly rule fine-grained matching search window is defined as a fine-grained query anomaly rule semantic encoding vector. The construction method of the anomaly rule fine-grained matching search window is expressed by the formula:

[0058]

[0059] W = {d j-m / 2 , d j-m / 2+1 ,..., d * ,..., d j+m / 2-1 , d j+m / 2},

[0060] where, ‖·‖ represents calculating the norm of the vector, represents rounding down, m represents the width of the anomaly rule fine-grained matching search window, j represents the position index of the location center anomaly rule semantic encoding vector in the set of anomaly rule semantic encoding vectors, and d j-m / 2 , d j-m / 2+1 , d j+m / 2-1 and d j+m / 2Semantically encode vectors for each fine-grained query anomaly rule in the fine-grained matching search window for the anomaly rules. W is a set composed of semantically encoded vectors for each fine-grained query anomaly rule in the fine-grained matching search window for the anomaly rules.

[0061] That is, taking the preliminarily screened semantically encoded vector of the anomaly rule as the positioning center, a fine-grained matching search window for the anomaly rules is delimited. Since the distance between each semantically encoded vector of the anomaly rule within the window and the positioning center in the semantic feature space is small, it represents potential matching items for the current operating state of the boiler system. Therefore, in this application, by designating each semantically encoded vector of the anomaly rule within the window as a fine-grained query semantically encoded vector of the anomaly rule, it serves as a candidate object for the next semantic interaction response encoding, for more refined local context awareness and semantic query interaction, which helps to concentrate resources for in-depth analysis in a smaller and more targeted data subset, improving both retrieval accuracy and maintaining computational efficiency.

[0062] Specifically, step S533 includes: performing a linear transformation on the encoded vector of the operating time series characteristics of the boiler system to obtain a query vector and a value vector, and taking each fine-grained query semantically encoded vector of the anomaly rule in the fine-grained matching search window for the anomaly rules as a key vector, and inputting the query vector, the value vector, and the key vector into a fine-grained query encoding module based on a heterogeneous transformer structure to obtain the semantically encoded vector of the query response for the operating state of the boiler system, which is expressed by the formula:

[0063] v q = qW q + b q ,

[0064] v v = qW v + b v ,

[0065]

[0066] where W q and W v represent the query embedding matrix and the value embedding matrix respectively, b q and b v represent different bias terms respectively, v q and v v represent the query vector and the value vector respectively, d l represents the l-th fine-grained query semantically encoded vector of the anomaly rule in the fine-grained matching search window for the anomaly rules, represents matrix multiplication, softmax represents the softmax function, (·) Trepresents the transpose of a vector, S represents the feature scale value of the semantic encoding vector of the l-th fine-grained query anomaly rule, v r represents the semantic encoding vector of the boiler system operation status query response.

[0067] That is, a linear transformation is performed on the boiler system operation time series feature encoding vector to convert it into a query vector and a value vector suitable for the attention mechanism. At the same time, each fine-grained query anomaly rule semantic encoding vector within the fine-grained matching search window of the anomaly rule is used as a key vector, and the fine-grained query encoding module based on the heterogeneous transformer structure is used to simulate the attention interaction between the query, key, and value. Through the multi-head self-attention layer in the fine-grained query encoding module, the semantic associations between the current operation status of the boiler system and each anomaly rule are processed in parallel to achieve the deep semantic interaction and fusion between the current operation status of the boiler system and each anomaly rule, generating a semantic encoding vector of the boiler system operation status query response that synthesizes the semantic association information of both, thereby providing accurate semantic information support for the final anomaly judgment.

[0068] Specifically, in step S54, based on the semantic encoding vector of the boiler system operation status query response, the judgment result is determined. In a specific example of the present application, the semantic encoding vector of the boiler system operation status query response is input into the anomaly monitoring module based on a classifier to obtain the judgment result, and the judgment result is used to indicate whether an anomaly has occurred in the boiler system. Here, the anomaly monitoring module based on the classifier performs deep learning and pattern recognition on the input semantic encoding vector of the boiler system operation status query response through a pre-trained classification algorithm, so as to utilize the deep semantic association information between the operation status of the boiler system and the anomaly rule contained therein, and output a judgment result indicating whether the boiler system is currently in an abnormal state. In this way, real-time and intelligent monitoring of the operation status of the boiler system can be achieved, potential faults or anomalies can be discovered in a timely and accurate manner, and reliable decision-making support can be provided for maintenance personnel.

[0069] In the above real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, in step S6, when the judgment result is that an anomaly has occurred, the operation data for a preset time period before and after the anomaly is saved. That is, considering that the operation data before and after a fault plays a key role in fault analysis, cause tracing, and formulating improvement measures, therefore, when the judgment result indicates that the boiler system is currently in an abnormal state, the present application further records and stores the operation data for a preset time period before and after the anomaly, so as to provide comprehensive and detailed data for subsequent specific fault analysis, facilitate maintenance personnel to deeply study the cause of the fault, optimize the system operation and maintenance strategy, and improve the overall reliability and stability of the system.

[0070] Based on this, in the design architecture of the controller, a dedicated area for data storage needs to be reserved, usually by combining a cache and a large-capacity storage medium. The cache is used to temporarily store recently collected operation data to ensure quick response and data preservation in case of anomalies, while large-capacity storage media such as hard disks or solid-state drives are used to store complete operation data records for a long time. During the system initialization phase, these storage areas are formatted and parameter-configured to ensure their stable and efficient operation. For example, according to the generation rate of system operation data and the expected preservation time length, the size of the cache is reasonably allocated, generally set to range from several hundred megabytes to several gigabytes, to meet the requirement of quickly storing a large amount of data in a short time.

[0071] When the controller obtains an anomaly judgment result according to the monitoring program, it immediately activates the data preservation mechanism. The data preservation mechanism first triggers a time-backtracking program, whose purpose is to determine the starting time point of a preset time period before the anomaly occurs. For example, if it is preset to save operation data for 60 seconds before and after the anomaly, and the system sampling period is 1 second, then the time-backtracking program will calculate that it needs to backtrack to the 60th sampling point before the anomaly occurs. During this process, the controller refers to its internal high-precision clock module, which usually uses an atomic clock or a high-precision quartz crystal oscillator as the time reference to ensure the accuracy of time calculation.

[0072] After determining the starting time point, the controller starts to extract data for the corresponding time period from the cache. Since the cache adopts a first-in-first-out (FIFO) or similar storage management strategy, data extraction is relatively efficient. During the extraction process, integrity checks are performed on the data, through methods such as checksum algorithms or verification of data packet header and footer information, to ensure that each data record is complete. For example, for each data packet, calculate its CRC (Cyclic Redundancy Check) value and compare it with the CRC value pre-stored in the data packet. If inconsistency is found, mark the data packet as suspicious data and try to perform data recovery or mark it as unavailable.

[0073] At the same time, while saving the data before the anomaly, the controller does not stop collecting and storing the current operation data. It continues to store the newly generated operation data into another temporary buffer according to the normal sampling frequency and process. The design purpose of this temporary buffer is to ensure that the data within a preset time period after the anomaly occurs can be completely saved. For example, within 60 seconds after the anomaly occurs, the newly collected data is continuously written into this temporary buffer, and data integrity checks and error handling are also performed.

[0074] After the preset time period following the occurrence of an anomaly ends, the controller combines and organizes the pre-anomaly data previously fetched from the cache and the post-anomaly data in the temporary buffer. During this process, the data is sorted in chronological order to ensure the temporal continuity of the data. At the same time, detailed time tags and event markers are added to this data so that subsequent analysts can clearly understand the generation time of the data and the corresponding anomaly events. For example, a timestamp accurate to the millisecond level is added to each data record, as well as a code indicating the type of anomaly, such as "abnormal boiler temperature too high", etc.

[0075] Next, the organized data is stored in a mass storage medium. During the storage process, data compression technology is adopted to reduce the occupancy of data storage space and improve storage efficiency. Common data compression algorithms such as ZIP, LZ4, etc. are applied to the storage of running data. These algorithms can effectively compress the data without losing key information. For example, for some parameter data with relatively slow continuous changes, such as boiler water level data, the storage volume can be significantly reduced through the compression algorithm while ensuring that the data can be accurately restored when needed.

[0076] To further ensure the security and reliability of the data, after the data is stored in the mass storage medium, a data backup operation can also be performed. For example, redundant storage technology such as RAID (Redundant Array of Independent Disks) technology is adopted to store the data on multiple disks simultaneously. Even if a certain disk fails, the data can still be recovered from other disks. In addition, the data can be regularly backed up to an external storage device or a remote storage server to prevent data loss in case of a catastrophic failure of the local storage medium. For example, the daily running data is backed up to an external hard drive during the system idle time or transmitted over the network to a remote data center for storage every day.

[0077] During the entire data saving process, detailed log records are simultaneously generated to record information such as the start time, end time, amount of data saved, data source, and destination. At the same time, any errors or abnormal situations that occur during the data saving process, such as data verification failure, storage medium write error, etc., are detailedly recorded in the log so that the operation and maintenance personnel can promptly discover the problems and take corresponding measures to repair them. For example, if a write error occurs when storing data on the hard disk, the log will record the specific location of the error, the error code, and the possible reasons to help the operation and maintenance personnel quickly locate and solve the problem.

[0078] Finally, after completing data saving and backup, the system automatically sends a notification message to the monitoring center or relevant operation and maintenance personnel. The notification message includes information such as a brief description of the abnormal event, the location and time range of data saving, so that the operation and maintenance personnel can obtain the data in a timely manner and conduct subsequent fault analysis and handling. For example, the notification is sent to the operation and maintenance personnel by means of text messages, emails or alarm pop-ups within the system, informing them that an abnormality has occurred in the boiler system and the relevant operation data has been saved to a specified location for their further investigation and analysis.

[0079] In summary, the real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to the embodiments of the present application is elucidated. It pre-configures abnormal rules and generates a target monitoring program based on the abnormal rules to perform abnormal diagnosis on the operation data of the boiler system. Specifically, first, the time series analysis of the operation data of the boiler system is carried out by using the data processing technology based on deep learning to extract the time series change pattern of the operation data of the boiler system. Then, by quickly querying and locating semantic interaction matching between the time series operation mode characteristics of the boiler system and the pre-configured abnormal rules, the abnormal diagnosis of the boiler system is realized, and the operation data in the preset time period before and after the abnormality is saved. In this way, the abnormal changes in the operation of the boiler system can be captured more accurately, and the accuracy and timeliness of fault diagnosis can be improved.

[0080] Furthermore, the present application also provides a real-time dynamic fault diagnosis system for the auxiliary control system of a thermal power plant. [[ID=B]]

[0081] Figure 5 The block diagram of the real-time dynamic fault diagnosis system for the auxiliary control system of a thermal power plant according to the embodiments of the present application is as follows. As Figure 5 shown, the real-time dynamic fault diagnosis system 100 for the auxiliary control system of a thermal power plant according to the embodiments of the present application includes: an abnormal rule configuration module 110 for pre-configuring parameters and abnormal rules; a monitoring program generation module 120 for generating a target monitoring program according to the abnormal rules; a program configuration module 130 for configuring the target monitoring program to the controller of the boiler system; an operation data acquisition module 140 for the controller to acquire the operation data of the boiler system after loading and running the target monitoring program; an operation data abnormality judgment module 150 for the controller to monitor the operation data of the boiler system in real time according to the abnormal rules to obtain a judgment result; and an abnormal operation data saving module 160 for saving the operation data in the preset time period before and after the abnormality when the judgment result is that an abnormality has occurred.

[0082] Here, those skilled in the art can understand that the specific operations of each module in the above real-time dynamic fault diagnosis system for the auxiliary control system of a thermal power plant have been described above with reference to Figures 1 to 4The description of the real-time dynamic fault diagnosis method for the auxiliary control system of thermal power plants has been introduced in detail, and therefore, its repeated description will be omitted.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant, characterized in that, Including: Pre-configured parameters and exception rules; Generating a target monitoring program according to the exception rules; Configuring the target monitoring program to the controller of the boiler system; After loading and running the target monitoring program, the controller collects the operation data of the boiler system; The controller monitors the operation data of the boiler system in real time according to the exception rules to obtain a judgment result, including: extracting the sequential operation mode features of the operation data of the boiler system to obtain a boiler system operation sequential feature coding vector; respectively extracting the semantic features of each exception rule to obtain a set of exception rule semantic coding vectors; performing a fast location semantic query response coding on the boiler system operation sequential feature coding vector and the set of exception rule semantic coding vectors to obtain a boiler system operation status query response semantic coding vector; determining the judgment result based on the boiler system operation status query response semantic coding vector; When the judgment result is that an exception occurs, save the operation data in a preset time period before and after the exception.

2. The real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to claim 1, wherein Extracting the sequential operation mode features of the operation data of the boiler system to obtain a boiler system operation sequential feature coding vector, including: Inputting the operation data of the boiler system into a boiler system operation mode feature extractor based on an RNN-LSTM model to obtain the boiler system operation sequential feature coding vector.

3. The real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to claim 2, characterized in that Respectively extracting the semantic features of each exception rule to obtain a set of exception rule semantic coding vectors, including: Using a semantic encoder based on a Bert model to perform semantic coding on each exception rule to obtain the set of exception rule semantic coding vectors.

4. The real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to claim 3, wherein Performing a fast location semantic query response coding on the boiler system operation sequential feature coding vector and the set of exception rule semantic coding vectors to obtain a boiler system operation status query response semantic coding vector, including: Determining a central exception rule semantic coding vector for locating the set of exception rule semantic coding vectors based on the semantic differences between the boiler system operation sequential feature coding vector and each exception rule semantic coding vector in the set of exception rule semantic coding vectors; Determining an exception rule fine-grained matching search window based on the central exception rule semantic coding vector for locating; Performing a semantic query response interaction coding on the boiler system operation sequential feature coding vector and the exception rule semantic coding vectors within the exception rule fine-grained matching search window to obtain the boiler system operation status query response semantic coding vector.

5. The real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to claim 4, wherein, Determining a central exception rule semantic coding vector for locating the set of exception rule semantic coding vectors based on the semantic differences between the boiler system operation sequential feature coding vector and each exception rule semantic coding vector in the set of exception rule semantic coding vectors, including: Calculating the cross entropy of the boiler system operation sequential feature coding vector with respect to each exception rule semantic coding vector in the set of exception rule semantic coding vectors to obtain a set of exception rule fast query location factors; Use the semantic encoding vector of the anomaly rule corresponding to the minimum value in the set of the anomaly rule quick query location factors as the semantic encoding vector of the location center anomaly rule.

6. The real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to claim 5, characterized in that The vector at the center position of the fine-grained matching search window of the anomaly rule is the semantic encoding vector of the location center anomaly rule, and each semantic encoding vector of the anomaly rule in the fine-grained matching search window of the anomaly rule is defined as the semantic encoding vector of the fine-grained query anomaly rule.

7. The real-time dynamic fault diagnosis method for the auxiliary control system of thermal power plants according to claim 6, characterized in that Within the fine-grained matching search window of the anomaly rule, perform semantic query response interaction encoding on the operation time series feature encoding vector of the boiler system and the semantic encoding vector of the anomaly rule to obtain the semantic encoding vector of the operation state query response of the boiler system, including: Perform a linear transformation on the operation time series feature encoding vector of the boiler system to obtain a query vector and a value vector, and use each fine-grained query anomaly rule semantic encoding vector in the fine-grained matching search window of the anomaly rule as a key vector, and input the query vector, the value vector, and the key vector into the fine-grained query encoding module based on the heterogeneous transformer structure to obtain the semantic encoding vector of the operation state query response of the boiler system.

8. The real-time dynamic fault diagnosis method for the auxiliary control system of a thermal power plant according to claim 7, characterized in that, Based on the semantic encoding vector of the operation state query response of the boiler system, determine the judgment result, including: Input the semantic encoding vector of the operation state query response of the boiler system into the anomaly monitoring module based on the classifier to obtain the judgment result, and the judgment result is used to indicate whether an anomaly occurs in the boiler system.

9. A real-time dynamic fault diagnosis system for the auxiliary control system of a thermal power plant, characterized in that, Include: Anomaly rule configuration module, used to pre-configure parameters and anomaly rules; Monitoring program generation module, used to generate a target monitoring program according to the anomaly rule; Program configuration module, used to configure the target monitoring program to the controller of the boiler system; Operation data acquisition module, used to collect the operation data of the boiler system by the controller after loading and running the target monitoring program; Operation data anomaly judgment module, used for the controller to monitor the operation data of the boiler system in real time according to the anomaly rule to obtain a judgment result; Anomaly operation data saving module, used to save the operation data in a preset time period before and after the anomaly when the judgment result is that an anomaly occurs.

Images