Fault detection processing method and system for flue gas waste heat recovery device

Abstract

The invention discloses a fault detection processing method and system for a flue gas waste heat recovery device. The fault detection processing method comprises the steps of obtaining operation data of the flue gas waste heat recovery device in a preset period; the operation data is preprocessed; inputting the preprocessed operation data into a pre-trained fault detection model for detection, and determining a detection result; obtaining a solution strategy corresponding to the detection result; processing the fault of the flue gas waste heat recovery device based on the solving strategy; the device operation state can be monitored in real time, compared with hysteresis of manual inspection, early abnormal signals of the device can be captured in time, subjectivity and inaccuracy of manual judgment are avoided, the fault detection accuracy is greatly improved, and various fault hidden dangers can be recognized more rapidly and accurately.

Description

Technical Field

[0001] The present invention relates to the field of energy-saving equipment, and in particular to a method and system for fault detection and processing of a flue gas waste heat recovery device. Background Art

[0002] With the continuous expansion of industrial production scale and the increasing requirements for energy utilization efficiency and environmental protection, various energy-saving devices play a crucial role in various fields. In industrial production, energy-saving pumps, energy-saving gas compressors, energy-saving hydraulic and pneumatic components, etc., effectively reduce energy consumption in the production process and improve the overall energy efficiency of the industrial system by optimizing their own operating mechanisms. In daily life and commercial scenarios, energy-saving refrigeration and air-conditioning equipment, energy-saving refrigerators and freezers, energy-saving air conditioners, and dual-condition solar heat pump air-conditioning units, etc., with their high-efficiency and energy-saving characteristics, meet people's comfort needs while significantly reducing the use of energy such as electricity. And the flue gas waste heat recovery device, as an important part of industrial energy recovery and utilization, also occupies a crucial position in many industrial fields. Such devices can effectively recover a large amount of waste heat carried by the flue gas discharged during industrial production, convert it into utilizable energy, significantly improve energy utilization efficiency, reduce the energy consumption and production costs of enterprises, and at the same time reduce the thermal pollution to the environment caused by the direct discharge of high-temperature flue gas, which is in line with the concept of sustainable development.

[0003] However, in the actual operation process, the flue gas waste heat recovery device faces many complex and severe working conditions. On the one hand, the flue gas generated by industrial production has a complex composition, which may contain corrosive gases, dust particles, etc. These substances will continuously erode and wear the internal structure and heat exchange components of the device, resulting in frequent equipment failures. For example, acidic gases may cause corrosion of metal components, reducing the strength and service life of the equipment; dust accumulation will affect the heat exchange efficiency, thereby reducing the performance of the device. On the other hand, the device is in a harsh environment of high temperature and high pressure for a long time, and mechanical components of the equipment are prone to fatigue damage, seal failure and other problems. At present, traditional fault detection methods often rely on manual inspections. This method is not only inefficient but also greatly affected by human factors, and it is difficult to detect early potential faults in a timely and accurate manner. Once a fault occurs, due to the lack of effective fault diagnosis and processing strategies, it often leads to a long period of shutdown for maintenance, which not only seriously affects the continuity of industrial production, causes huge economic losses, but also may cause more serious equipment damage accidents due to delayed maintenance. Therefore, there is an urgent need for an efficient and intelligent fault detection and processing method. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the above technologies to a certain extent. To this end, the first aspect of the present invention aims to propose a fault detection and handling method for a flue gas waste heat recovery device, which can monitor the operation status of the device in real time. Compared with the lag of manual inspection, it can capture early abnormal signals of the device in time, avoid the subjectivity and inaccuracy of manual judgment, greatly improve the accuracy of fault detection, and can identify various potential fault hazards more quickly and accurately.

[0005] The second aspect of the present invention aims to propose a fault detection and handling system for a flue gas waste heat recovery device.

[0006] To achieve the above object, the first aspect embodiment of the present invention proposes a fault detection and handling method for a flue gas waste heat recovery device, including:

[0007] Obtain the operation data of the flue gas waste heat recovery device within a preset period;

[0008] Preprocess the operation data;

[0009] Input the preprocessed operation data into a pre-trained fault detection model for detection to determine the detection result;

[0010] Obtain the solution strategy corresponding to the detection result;

[0011] Process the fault of the flue gas waste heat recovery device based on the solution strategy.

[0012] Preferably, preprocessing the operation data includes:

[0013] Perform data cleaning on the operation data to obtain a data cleaning result;

[0014] Use the data cleaning result as the preprocessed operation data.

[0015] Preferably, performing data cleaning on the operation data to obtain a data cleaning result includes:

[0016] Align the data in multiple dimensions in the operation data based on the time series;

[0017] Perform anomaly detection and marking on the data points in each dimension data of the aligned operation data to obtain the anomaly marks corresponding to the data points in each dimension data;

[0018] Obtain the number of anomaly marks and the data type of the dimension where the anomaly marks are located in the data of multiple dimensions corresponding to each time series point;

[0019] Judging the multi-dimensional data corresponding to each time series point based on the number of anomaly marks in the multi-dimensional data corresponding to each time series point, and determining a number of first anomaly data points and a number of second anomaly data points in the multi-dimensional data corresponding to each time series point;

[0020] Regarding the number of first anomaly data points and the number of second anomaly data points as target data points to be cleaned; regarding the data corresponding to the target data points to be cleaned as target data to be cleaned;

[0021] Obtaining the data type corresponding to the target data to be cleaned of the target data points to be cleaned;

[0022] Obtaining the target data cleaning method corresponding to the target data to be cleaned based on the data type;

[0023] Cleaning the target data to be cleaned based on the target data cleaning method;

[0024] Traversing the multi-dimensional data corresponding to all time series points to obtain the data cleaning result.

[0025] Preferably, performing anomaly detection and marking on the data points in each dimension data of the aligned operation data to obtain the anomaly marks corresponding to the data points in each dimension data, including:

[0026] Regarding the data of any one dimension in the aligned operation data as the target data set;

[0027] Dividing the data in the target data set into several target sub-data sets on average;

[0028] Calculating the mean value of the data values corresponding to each data point in each target sub-data set respectively to obtain several target mean values;

[0029] Selecting a data point in any one first sub-data set as the first target data point;

[0030] Calculating the differences between the data value of the first target data point and several target mean values to obtain several differences;

[0031] Calculating the mean value of the several differences to obtain the anomaly degree value of the first target data point;

[0032] Comparing the anomaly degree value with a preset anomaly degree threshold, and when it is determined that the anomaly degree value is greater than or equal to the preset anomaly degree threshold, marking the first target data point as an anomaly;

[0033] Traversing all the data points in the target data set, and marking the data points as anomalies based on the anomaly degree values of each data point to obtain several anomaly marks;

[0034] Traverse the operation data of all dimensions to obtain the anomaly markers corresponding to the data points in the data of each dimension.

[0035] Preferably, based on the number of anomaly markers in the data of multiple dimensions corresponding to each time series point, judge the data of multiple dimensions corresponding to each time series point, and determine several first anomaly data points and several second anomaly data points in the data of multiple dimensions corresponding to each time series point, including:

[0036] Arbitrarily take the data of multiple dimensions corresponding to a time series point;

[0037] If the number of anomaly markers in the data of multiple dimensions is 1, then take the data point in the dimension where the anomaly marker is located as the second target data point; take the second target data point as the center and a preset distance as the radius to determine the target area; calculate the correlation coefficient between the second target data point and other data points in the target area; when the sum of the correlation coefficients is less than or equal to the preset correlation threshold, take the second target data point as the first anomaly data point;

[0038] If the number of anomaly markers in the data of multiple dimensions is greater than 1, then determine the anomaly evaluation value of the combination of the dimensions where the anomaly markers are located based on the first preset algorithm; compare the anomaly evaluation value with the preset anomaly evaluation threshold; when it is determined that the anomaly evaluation value is greater than or equal to the preset anomaly evaluation threshold, take the data point corresponding to the anomaly marker as the second anomaly data point;

[0039] Traverse each time series point to obtain several first anomaly data points and several second anomaly data points.

[0040] Preferably, the first preset algorithm includes:

[0041]

[0042] Among them, Y a,m represents the anomaly evaluation value of the m-th anomaly marker dimension combination corresponding to the a-th time series point; M represents the total number of anomaly marker dimensions; represents the mean value of the data values of all data points within the preset range of the anomaly marker data points in the i-th anomaly marker dimension in the m-th anomaly marker dimension combination, and δ m,i,j represents the fluctuation coefficient of the data value of the j-th anomaly data point in the i-th anomaly marker dimension in the m-th anomaly marker dimension combination, and τ m,i,j represents the data value of the j-th anomaly marker data point in the i-th anomaly marker dimension in the m-th anomaly marker dimension combination; f i represents the weight value of the data of the i-th anomaly marker dimension; exp() represents the exponential function with the natural constant as the base.

[0043] Preferably, the method for constructing a fault detection model includes:

[0044] Obtain a fault detection training data set;

[0045] Train a neural network model based on the obtained fault detection training data set to obtain an initial fault detection model;

[0046] Obtain a fault detection test data set;

[0047] Input the fault detection test data set into the initial fault detection model for testing to obtain a trained fault detection model.

[0048] Preferably, after processing the fault of the flue gas waste heat recovery device based on the solution strategy, it further includes:

[0049] Evaluate the flue gas waste heat recovery device after the fault handling is completed based on a second preset algorithm to obtain an evaluation value of the flue gas waste heat recovery device;

[0050] Compare the evaluation value of the flue gas waste heat recovery device with a preset evaluation threshold. When it is determined that the evaluation value of the flue gas waste heat recovery device is greater than or equal to the preset evaluation threshold, the fault handling is completed.

[0051] Preferably, the second preset algorithm includes:

[0052]

[0053] Wherein, φ represents the evaluation value of the flue gas waste heat recovery device; U represents the voltage value at both ends of the flue gas waste heat recovery device during operation; I represents the current value at both ends of the flue gas waste heat recovery device during operation; R represents the internal resistance value of the flue gas waste heat recovery device itself; t 1 represents the time value that the flue gas waste heat recovery device has been used, t 2 represents the ideal use time of the flue gas waste heat recovery device; μ represents the flow index of the flue gas waste heat recovery device; γ represents the heat exchange efficiency of the flue gas waste heat recovery device: represents the theoretical remaining service life of the flue gas waste heat recovery device; δ represents the number of faults of the flue gas waste heat recovery device.

[0054] To achieve the above object, the second aspect embodiment of the present invention proposes a fault detection and handling system for a flue gas waste heat recovery device, including:

[0055] A first acquisition module for acquiring the operation data of the flue gas waste heat recovery device within a preset period;

[0056] A preprocessing module for preprocessing the operation data;

[0057] A detection module, configured to input the preprocessed operation data into a pre-trained fault detection model for detection and determine the detection result;

[0058] A second acquisition module, configured to acquire a solution strategy corresponding to the detection result;

[0059] A processing module, configured to process the faults of the flue gas waste heat recovery device based on the solution strategy.

[0060] The present invention provides a method and system for fault detection and processing of a flue gas waste heat recovery device. By acquiring operation data within a preset period, the operation state of the flue gas waste heat recovery device can be monitored in real time; subtle changes during the operation of the device, such as abnormal fluctuations in parameters such as temperature, pressure, and flow rate, can be captured in a timely manner; compared with traditional manual inspections, the timeliness of fault detection is greatly improved, thus winning precious time for subsequent fault processing. At the same time, by inputting the data into a pre-trained fault detection model and utilizing the powerful data processing and analysis capabilities of the model, the fault type can be accurately identified, avoiding misjudgment or missed judgment caused by manual judgment errors and improving the accuracy of fault detection; by preprocessing the operation data, the data can be screened, cleaned, and sorted to make it more suitable for input into the fault detection model, thereby accelerating the detection speed of the model. Once the detection result is determined, the corresponding solution strategy can be quickly acquired and implemented for processing; it can avoid waste heat waste caused by insufficient heat exchange and subsequent operation problems of environmental protection equipment, effectively shorten the fault processing time, and improve the overall operation efficiency of the device; timely and accurate fault detection and efficient processing can effectively prevent the deterioration of equipment faults.

[0061] Other features and advantages of the present invention will be described in the subsequent description, and some of them will be obvious from the description or understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written description and the drawings.

[0062] The technical solutions of the present invention will be further described in detail below through the drawings and embodiments. Description of the Drawings

[0063] The drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0064] Figure 1 is a flowchart of a method for fault detection and processing of a flue gas waste heat recovery device according to an embodiment of the present invention;

[0065] Figure 2 is a flowchart of preprocessing of operation data according to an embodiment of the present invention;

[0066] Figure 3 It is a block diagram of a fault detection and processing system for a flue gas waste heat recovery device according to an embodiment of the present invention. Detailed implementation manners

[0067] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not intended to limit the present invention.

[0068] Embodiment 1

[0069] As Figure 1 shown, a fault detection and processing method for a flue gas waste heat recovery device includes steps S1 - S5:

[0070] S1: Obtain the operation data of the flue gas waste heat recovery device within a preset period;

[0071] S2: Preprocess the operation data;

[0072] S3: Input the preprocessed operation data into a pre - trained fault detection model for detection to determine the detection result;

[0073] S4: Obtain the solution strategy corresponding to the detection result;

[0074] S5: Process the faults of the flue gas waste heat recovery device based on the solution strategy.

[0075] In this embodiment, the operation data of the flue gas waste heat recovery device includes, but is not limited to: temperature data: including the flue gas inlet temperature, flue gas outlet temperature, inlet temperature and outlet temperature of the heated medium (such as water, air); pressure data: related to the flue gas side pressure and the heated medium side pressure; flow data: including the flue gas flow rate and the heated medium flow rate; equipment operation parameters: such as the fan speed, the head and flow rate of the pump, etc.; energy consumption data: mainly referring to the electricity consumption and fuel consumption during the operation of the device; heat recovery amount data: that is, the heat value recovered by the device per unit time.

[0076] In this embodiment, the detection results include: crystallization, low refrigerant temperature, low heat source water temperature, hot water over - temperature, solution over - temperature, abnormal solution liquid level, abnormal refrigerant liquid level, etc.

[0077] In this embodiment, the solution strategy is the specific solution and implementation steps set based on experience for each detection result.

[0078] The working principle of the above technical solution is as follows: within a preset period, various sensors are used to comprehensively collect the operation data of the flue gas waste heat recovery device; the original operation data collected may have problems such as noise, missing values, or inconsistent formats; the preprocessed operation data is input into a pre-trained fault detection model. This model is usually constructed based on machine learning or deep learning algorithms. During the training phase, the model learns the operation data characteristics in a large number of normal and fault states, thereby establishing a mapping relationship between the data characteristics and the fault types. When new operation data is input, the model analyzes the data according to the learned feature patterns to determine whether the current operation state of the device is normal. If an anomaly is detected, the fault type and severity are determined; once the fault detection model determines the detection result, the system will obtain the corresponding solution strategy according to the pre-set rules and knowledge base. This knowledge base is summarized from the experience of domain experts and historical fault handling cases; based on the obtained solution strategy, the fault of the flue gas waste heat recovery device is processed through an automated control system or manual operation.

[0079] The beneficial effects of the above technical solution are as follows: by obtaining the operation data within a preset period, the operation state of the flue gas waste heat recovery device can be monitored in real time; subtle changes during the device operation, such as abnormal fluctuations in parameters such as temperature, pressure, and flow rate, can be captured in a timely manner; compared with traditional manual inspections, the timeliness of fault detection is greatly improved, winning precious time for subsequent fault handling. At the same time, by inputting the data into a pre-trained fault detection model and utilizing the model's powerful data processing and analysis capabilities, the fault type can be accurately identified, avoiding misjudgment or missed judgment caused by human judgment errors and improving the accuracy of fault detection; preprocessing the operation data can screen, clean, and organize the data to make it more suitable for input into the fault detection model, accelerating the detection speed of the model. Once the detection result is determined, the corresponding solution strategy can be quickly obtained and implemented for processing; avoiding waste heat waste caused by insufficient heat exchange and subsequent problems in the operation of environmental protection equipment, effectively shortening the fault handling time and improving the overall operation efficiency of the device; timely and accurate fault detection and efficient handling can effectively prevent the deterioration of equipment faults.

[0080] Embodiment 2

[0081] As Figure 2 shown, preprocessing the operation data includes steps S21 - S22:

[0082] S21: Perform data cleaning on the operation data to obtain a data cleaning result;

[0083] S22: Use the data cleaning result as the preprocessed operation data.

[0084] Embodiment 3

[0085] Clean the operation data to obtain the data cleaning result, including:

[0086] Align the data in multiple dimensions in the operation data based on the time series;

[0087] Perform anomaly detection and marking on the data points in each dimension data of the aligned operation data to obtain the anomaly marks corresponding to the data points in the data of each dimension;

[0088] Obtain the number of anomaly marks and the data types of the dimensions where the anomaly marks are located in the data of multiple dimensions corresponding to each time series point;

[0089] Based on the number of anomaly marks in the data of multiple dimensions corresponding to each time series point, judge the data of multiple dimensions corresponding to each time series point to determine several first anomaly data points and several second anomaly data points in the data of multiple dimensions corresponding to each time series point;

[0090] Take several first anomaly data points and several second anomaly data points as target data points to be cleaned; take the data corresponding to the target data points to be cleaned as target data to be cleaned;

[0091] Obtain the data types corresponding to the target data to be cleaned of the target data points to be cleaned;

[0092] Obtain the target data cleaning method corresponding to the target data to be cleaned based on the data type;

[0093] Clean the target data to be cleaned based on the target data cleaning method;

[0094] Traverse the data of multiple dimensions corresponding to all time series points to obtain the data cleaning result.

[0095] In this embodiment, the first anomaly data point is that when the number of anomaly marks in the data of multiple dimensions corresponding to the time series point is 1, and the sum of the correlations between the data point with the anomaly mark and other data points within the preset distance range in the dimension where the data point with the anomaly mark is located is less than or equal to the preset correlation threshold, the data point with the anomaly mark is taken as the first anomaly data point.

[0096] In this embodiment, the second anomaly data point is that when the number of anomaly marks in the data of multiple dimensions corresponding to the time series point is greater than 1, and the anomaly evaluation value of the dimension combination where the anomaly mark is located is greater than or equal to the preset anomaly evaluation value, the data points corresponding to the anomaly marks are taken as the second anomaly data points.

[0097] The working principle of the above technical solution is as follows: The operation data of the flue gas waste heat recovery device comes from multiple sensors, such as temperature, pressure, and flow sensors. The data collection frequencies and start times of these sensors may vary. Time series alignment processing is to uniformly calibrate data of different dimensions in chronological order. Taking time as the reference axis, the data of each dimension are made to correspond one by one at the same time point. For the time series data after alignment of each dimension, it is judged whether the data points deviate from the normal range. If a certain temperature data point is significantly higher or lower than the normal fluctuation range, the algorithm will identify it as an abnormal point and mark it accordingly. Each data point in each dimension of data undergoes such detection. After completing the abnormal marking of each dimension data point, for each time series point, count the number of abnormal marks in the corresponding multiple dimension data, and record the data type of the dimension data where the abnormal mark is located. According to the number of abnormal marks corresponding to each time series point, set different judgment thresholds to determine the first abnormal data point and the second abnormal data point. Take the determined first abnormal data point and the second abnormal data point as the target data points to be cleaned, and extract the data corresponding to these data points as the target data to be cleaned. At the same time, obtain the data types corresponding to these target data points to be cleaned. According to the data types of the target data to be cleaned, select the corresponding target data cleaning method from the preset data cleaning method library. For example, for numerical data, if the abnormal data is an obvious deviation caused by a sensor failure, methods such as mean filling, median filling, or model prediction-based methods can be used for repair; for boolean data, if the abnormality is a state reversal, it can be corrected after checking with the actual state of the device. According to the selected target data cleaning method, perform cleaning operations on the target data to be cleaned. After completing the traversal and cleaning of the multiple dimension data corresponding to all time series points, obtain the cleaned data set, that is, the data cleaning result.

[0098] Example 4

[0099] Perform abnormal detection and marking on the data points in each dimension of the aligned operation data to obtain the abnormal marks corresponding to the data points in each dimension of data, including:

[0100] Take the data of any one dimension in the aligned operation data as the target data set;

[0101] Divide the data in the target data set into several target sub-data sets on average;

[0102] Calculate the mean values of the data values corresponding to each data point in each target sub-data set respectively to obtain several target mean values;

[0103] Arbitrarily take a data point in a first sub-data set as the first target data point;

[0104] Calculate the differences between the data values of the first target data points and several target means to obtain several differences;

[0105] Calculate the mean of the several differences to obtain the abnormality degree value of the first target data point;

[0106] Compare the abnormality degree value with a preset abnormality degree threshold. When it is determined that the abnormality degree value is greater than or equal to the preset abnormality degree threshold, perform an abnormality mark on the first target data point;

[0107] Traverse all data points in the target dataset, and perform abnormality marks on the data points based on the abnormality degree values of each data point to obtain several abnormality marks;

[0108] Traverse the operation data of all dimensions to obtain the abnormality marks corresponding to the data points in the data of each dimension.

[0109] The beneficial effects of the above technical solution are as follows: The operation data of each dimension is refined into multiple target sub-datasets, and the data means in the sub-datasets are calculated respectively. Based on this, the abnormality degree values of each data point are calculated. This refined calculation method can fully consider the characteristic changes of the data in different local intervals; by calculating the mean value in sub-intervals, compared with calculating the mean value as a whole, it can capture the abnormal conditions of data points in their respective local intervals more sensitively, avoiding the masking of local abnormalities by overall characteristics, thereby significantly improving the recognition accuracy of abnormal data and providing a more reliable data basis for subsequent fault judgment; such detailed abnormality detection and marking operations are performed on the operation data of each dimension, ensuring all-round monitoring of the device operation status; because the data of different dimensions (such as temperature, pressure, flow, etc.) reflect the operation of the device from different angles, any abnormality in a dimension may imply potential faults in the device; when calculating the abnormality degree value of a single data point, by comparing it with the preset abnormality degree threshold in real time, once it is determined that the abnormality degree value is greater than or equal to the threshold, immediately perform an abnormality mark on this data point. This immediate processing method can quickly capture the abnormal changes in the data and detect them in a timely manner when the abnormal situation just appears.

[0110] Embodiment 5

[0111] Based on the number of abnormality marks in the data of multiple dimensions corresponding to each time series point, judge the data of multiple dimensions corresponding to each time series point, and determine several first abnormal data points and several second abnormal data points in the data of multiple dimensions corresponding to each time series point, including:

[0112] Arbitrarily take the data of multiple dimensions corresponding to a time series point;

[0113] If the number of anomaly markers in the data of multiple dimensions is 1, then the data point in the dimension where the anomaly marker is located is used as the second target data point; with the second target data point as the center and a preset distance as the radius, a target area is determined; the correlation coefficient between the second target data point and other data points in the target area is calculated; when the sum of the correlation coefficients is less than or equal to a preset correlation threshold, the second target data point is used as the first anomaly data point;

[0114] If the number of anomaly markers in the data of multiple dimensions is greater than 1, then an anomaly evaluation value of the combination of the dimensions where the anomaly markers are located is determined based on a first preset algorithm; the anomaly evaluation value is compared with a preset anomaly evaluation threshold; when it is determined that the anomaly evaluation value is greater than or equal to the preset anomaly evaluation threshold, the data points corresponding to the anomaly markers are used as the second anomaly data points;

[0115] Each time series point is traversed to obtain a number of first anomaly data points and a number of second anomaly data points.

[0116] In this embodiment, the correlation coefficient includes but is not limited to the Pearson correlation coefficient.

[0117] The working principle of the above technical solution is as follows: A time series point is randomly selected from the entire time series data, and the data of multiple dimensions corresponding to this point is obtained; when there is only 1 anomaly marker in the data of multiple dimensions corresponding to this time series point, the data point in the dimension with the anomaly marker is defined as the second target data point; with the second target data point as the center and a preset distance as the radius, a target area is delimited. This target area contains other data points that are close to the second target data point in terms of time and value; when the sum of the correlation coefficients is less than or equal to a preset correlation threshold, it indicates that the association degree between the second target data point and the surrounding data points is low, and its anomaly is more prominent. At this time, the second target data point is upgraded to the first anomaly data point. The preset correlation threshold is set based on a large amount of historical data and experience and is used as a standard for judging the anomaly degree of data points; when the number of anomaly markers in the data of multiple dimensions is greater than 1, a first preset algorithm is used to determine the anomaly evaluation value of the combination of the dimensions where the anomaly markers are located. The calculated anomaly evaluation value is compared with the preset anomaly evaluation threshold. The preset anomaly evaluation threshold is also set based on historical data and experience and is used to distinguish whether the anomaly degree reaches a level that requires special attention. When the anomaly evaluation value is greater than or equal to the preset anomaly evaluation threshold, it indicates that the anomaly degree of this dimension combination is relatively high, and the data points corresponding to these anomaly markers are determined as the second anomaly data points; the above steps are repeated for each time series point in the entire time series, and the data of multiple dimensions corresponding to each time series point is judged in turn to determine the first anomaly data point and the second anomaly data point among them.

[0118] The beneficial effects of the above technical solution are as follows: By analyzing the number of anomaly markers in the multi-dimensional data of each time series point, this technology can more accurately locate the anomaly data points, laying a foundation for subsequent data cleaning.

[0119] Embodiment 6

[0120] The first preset algorithm includes:

[0121]

[0122] Where Y a,m represents the anomaly evaluation value of the m-th anomaly marker dimension combination corresponding to the a-th time series point; M represents the total number of anomaly marker dimensions; represents the mean value of the data values of all data points within the preset range of the anomaly marker data points in the i-th anomaly marker dimension in the m-th anomaly marker dimension combination, and δ m,i,j represents the fluctuation coefficient of the data value of the j-th anomaly data point in the i-th anomaly marker dimension in the m-th anomaly marker dimension combination, and τ m,i,j represents the data value of the j-th anomaly marker data point in the i-th anomaly marker dimension in the m-th anomaly marker dimension combination; f i represents the weight value of the data of the i-th anomaly marker dimension; exp() represents the exponential function with the natural constant as the base.

[0123] In this embodiment, the specific calculation only involves numerical calculation, ignoring the dimension.

[0124] In this embodiment, the specific method for obtaining the fluctuation coefficient is as follows: The difference between the data value of the anomaly marker data point and the mean value of the data values of all data points within the preset range is used as the first difference; the differences between the data value of the anomaly marker data point and the data values of all data points within the preset range are obtained to get several second differences; the sum value of the several second differences is used as the target difference; the ratio of the first difference to the target difference is used as the fluctuation coefficient.

[0125] The beneficial effects of the above technical solution are as follows: When judging that anomalies occur simultaneously in multiple dimensions such as temperature and pressure, it can comprehensively analyze the degree and mutual relationship of anomalies in each dimension; by using different weight values for data in different dimensions, the algorithm can adapt to the importance differences of different parameters of the flue gas waste heat recovery device in a complex industrial environment, and further improve the accuracy of identifying anomaly data points through the anomaly evaluation value of the anomaly combination.

[0126] Embodiment 7

[0127] A method for constructing a fault detection model includes:

[0128] Obtain a fault detection training data set;

[0129] Training a neural network model based on obtaining a fault detection training data set to obtain an initial fault detection model;

[0130] Obtaining a fault detection test data set;

[0131] Inputting the fault detection test data set into the initial fault detection model for testing to obtain a trained fault detection model.

[0132] Embodiment 8

[0133] After processing the faults of the flue gas waste heat recovery device based on the solution strategy, it further includes:

[0134] Evaluating the flue gas waste heat recovery device after fault handling based on a second preset algorithm to obtain an evaluation value of the flue gas waste heat recovery device;

[0135] Comparing the evaluation value of the flue gas waste heat recovery device with a preset evaluation threshold, and when it is determined that the evaluation value of the flue gas waste heat recovery device is greater than or equal to the preset evaluation threshold, the fault handling is completed.

[0136] In this embodiment, evaluating the flue gas waste heat recovery device after fault handling is mainly to evaluate the operating state of the flue gas waste heat recovery device after fault handling.

[0137] The working principle of the above technical solution is: evaluating the flue gas waste heat recovery device after fault handling based on a second preset algorithm; this algorithm will comprehensively consider multiple operating parameters and performance indicators of the device to calculate the evaluation value; comparing the calculated evaluation value of the flue gas waste heat recovery device with a preset evaluation threshold; the preset evaluation threshold is determined based on various factors such as the design standard of the device, historical operating data, and industry best practices; it represents the performance level that an ideal device after fault handling should reach; when it is determined that the evaluation value of the flue gas waste heat recovery device is greater than or equal to the preset evaluation threshold, it indicates that the device after fault handling has reached the expected standards in all key indicators and is in good operating condition, thus completing the fault handling.

[0138] The beneficial effects of the above technical solution are: evaluating the device after fault handling through a second preset algorithm, comprehensively considering various operating indicators and performance parameters of the device, generating a quantitative evaluation value; accurately judging whether the fault handling is completed, ensuring that the device is always in good operating condition, and maximizing the realization of waste heat recovery and utilization. Reducing energy waste caused by device failures or improper handling, and further deepening the energy conservation and emission reduction effect.

[0139] Embodiment 9

[0140] The second preset algorithm includes:

[0141]

[0142] Among them, φ represents the evaluation value of the flue gas waste heat recovery device; U represents the voltage value at both ends of the flue gas waste heat recovery device during operation; I represents the current value at both ends of the flue gas waste heat recovery device during operation; R represents the internal resistance value of the flue gas waste heat recovery device itself; t 1 represents the time value that the flue gas waste heat recovery device has been used, t 2 represents the ideal service time of the flue gas waste heat recovery device; μ represents the flow index of the flue gas waste heat recovery device; γ represents the heat exchange efficiency of the flue gas waste heat recovery device: θ represents the theoretical remaining service life of the flue gas waste heat recovery device; δ represents the number of failures of the flue gas waste heat recovery device.

[0143] The beneficial effects of the above technical solution are as follows: The algorithm incorporates the voltage value U, current value I, and internal resistance value R of the device during operation. These electrical parameters can intuitively reflect the electrical performance and energy consumption of the device. At the same time, the already used time t 1 and the ideal service time t 2 are considered, and the impact of the aging degree of the device on its performance can be evaluated; the evaluation value φ is obtained through complex and rigorous mathematical operations, providing a quantitative index for the device performance; enterprises can perform dynamic management throughout the life cycle of the device based on the evaluation value; after a fault is processed, using this algorithm to calculate the evaluation value can intuitively judge the processing effect.

[0144] As Figure 3 shown, to achieve the above object, the second aspect embodiment of the present invention proposes a fault detection and processing system for a flue gas waste heat recovery device, including:

[0145] A first acquisition module, configured to acquire the operation data of the flue gas waste heat recovery device within a preset period;

[0146] A preprocessing module, configured to preprocess the operation data;

[0147] A detection module, configured to input the preprocessed operation data into a pre-trained fault detection model for detection to determine the detection result;

[0148] A second acquisition module, configured to acquire the solution strategy corresponding to the detection result;

[0149] A processing module, configured to process the fault of the flue gas waste heat recovery device based on the solution strategy.

[0150] The beneficial effects of the above technical solution are as follows: By obtaining the operation data within a preset period, the operation status of the flue gas waste heat recovery device can be monitored in real time; subtle changes during the operation of the device, such as abnormal fluctuations in parameters like temperature, pressure, and flow rate, can be captured in a timely manner; compared with traditional manual inspections, the timeliness of fault detection is greatly improved, thus gaining valuable time for subsequent fault handling. At the same time, by inputting the data into a pre-trained fault detection model and leveraging the model's powerful data processing and analysis capabilities, the type of fault can be accurately identified, avoiding misjudgment or missed judgment caused by human judgment errors and enhancing the accuracy of fault detection; preprocessing the operation data can screen, clean, and organize the data to make it more suitable for input into the fault detection model, thereby accelerating the detection speed of the model. Once the detection result is determined, the corresponding solution strategy can be quickly obtained and implemented for handling; it can avoid waste heat waste caused by insufficient heat exchange and subsequent problems in the operation of environmental protection equipment, effectively shorten the fault handling time, and improve the overall operation efficiency of the device; timely and accurate fault detection and efficient handling can effectively prevent the deterioration of equipment faults.

[0151] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A method for detecting and processing a fault of a flue gas waste heat recovery device, characterized in that: include: Obtaining the operating data of the flue gas waste heat recovery device within a preset period; Preprocessing the operation data; Input the preprocessed operation data into the pre-trained fault detection model for detection and determine the detection result; Obtain the solution strategy corresponding to the test results; The failure of the flue gas waste heat recovery device is handled based on the solution strategy.

2. The fault detection and processing method for the flue gas waste heat recovery device according to claim 1 is characterized in that: Preprocessing the operation data includes: Performing data cleaning on the operation data to obtain a data cleaning result; The data cleaning results are used as the preprocessed running data.

3. The fault detection and processing method for the flue gas waste heat recovery device according to claim 2 is characterized in that: Performing data cleaning on the operation data to obtain data cleaning results includes: Align the data of multiple dimensions in the running data based on time series; Perform anomaly detection and marking on the data points in each dimension of the aligned running data to obtain anomaly marks corresponding to the data points in each dimension; Obtain the number of abnormal markers in the data of multiple dimensions corresponding to each time series point and the data type of the dimension where the abnormal markers are located; The multiple dimensional data corresponding to each time series point are judged based on the number of abnormal marks in the multiple dimensional data corresponding to each time series point, and a plurality of first abnormal data points and a plurality of second abnormal data points in the multiple dimensional data corresponding to each time series point are determined; Using a plurality of first abnormal data points and a plurality of second abnormal data points as target data points to be cleaned; using data corresponding to the target data points to be cleaned as target data to be cleaned; Obtain the data type corresponding to the target to-be-cleaned data of the target to-be-cleaned data point; Acquire a target data cleaning method corresponding to the target data to be cleaned based on the data type; Performing data cleaning on the target data to be cleaned based on the target data cleaning method; Traverse the data of multiple dimensions corresponding to all time series points to obtain the data cleaning results.

4. The fault detection and processing method for the flue gas waste heat recovery device according to claim 3 is characterized in that: Anomaly detection and labeling are performed on the data points in each dimension of the aligned running data to obtain the anomaly labels corresponding to the data points in each dimension, including: Use the data of any dimension in the aligned running data as the target data set; Divide the data in the target data set into several target sub-data sets evenly; Calculate the mean of the data values ​​corresponding to each data point in each target sub-dataset to obtain several target means; Randomly select a data point in a first sub-data set as a first target data point; Calculate the difference between the data value of the first target data point and a number of target means to obtain a number of difference values; Calculate the mean of several differences to obtain the abnormality value of the first target data point; Comparing the abnormality level value with a preset abnormality level threshold, and marking the first target data point as abnormal when it is determined that the abnormality level value is greater than or equal to the preset abnormality level threshold; Traverse all data points in the target data set, mark the data points as abnormal based on the abnormality value of each data point, and obtain several abnormal marks; Traverse the running data of all dimensions and obtain the abnormal labels corresponding to the data points in each dimension data.

5. The fault detection and processing method for a flue gas waste heat recovery device according to claim 4 is characterized in that: The multiple dimensional data corresponding to each time series point are judged based on the number of abnormal marks in the multiple dimensional data corresponding to each time series point, and a plurality of first abnormal data points and a plurality of second abnormal data points in the multiple dimensional data corresponding to each time series point are determined, including: Take any multiple dimensional data corresponding to a time series point; If the number of abnormal marks in the data of multiple dimensions is 1, the data point of the dimension where the abnormal mark is located is used as the second target data point; the target area is determined with the second target data point as the center and the preset distance as the radius; the correlation coefficient between the second target data point and other data points in the target area is calculated; when the sum of the correlation coefficients is less than or equal to the preset correlation threshold, the second target data point is used as the first abnormal data point; If the number of abnormal marks in the data of multiple dimensions is greater than 1, then determine the abnormal evaluation value of the combination of dimensions where the abnormal marks are located based on the first preset algorithm; compare the abnormal evaluation value with the preset abnormal evaluation threshold; when it is determined that the abnormal evaluation value is greater than or equal to the preset abnormal evaluation threshold, take the data point corresponding to the abnormal mark as the second abnormal data point; Traverse each time series point to obtain a number of first abnormal data points and a number of second abnormal data points.

6. The fault detection and processing method for a flue gas waste heat recovery device according to claim 5, characterized in that: The first preset algorithm includes: Among them, Y a,m represents the abnormal evaluation value of the mth abnormal marking dimension combination corresponding to the ath time series point; M represents the total number of abnormal marking dimensions; represents the mean value of all data points within the preset range of the abnormal marking data point in the ith abnormal marking dimension in the mth abnormal marking dimension combination, δ m,i,j represents the fluctuation coefficient of the data value of the jth abnormal data point in the ith abnormal marking dimension in the mth abnormal marking dimension combination within the preset range, τ m,i,j represents the data value of the jth abnormal marker data point in the ith abnormal marker dimension in the mth abnormal marker dimension combination; f i Represents the weight value of the data in the i-th abnormal marker dimension; exp() represents an exponential function with a natural constant as the base.

7. The fault detection and processing method for a flue gas waste heat recovery device according to claim 1 is characterized in that: The method for constructing a fault detection model includes: Obtain a fault detection training data set; The neural network model is trained based on the acquired fault detection training data set to obtain an initial fault detection model; Obtain a fault detection test data set; The fault detection test data set is input into the initial fault detection model for testing to obtain a trained fault detection model.

8. The fault detection and processing method for a flue gas waste heat recovery device according to claim 1, characterized in that: After the failure of the flue gas waste heat recovery device is handled based on the solution strategy, the method further includes: Evaluate the flue gas waste heat recovery device after the fault processing is completed based on the second preset algorithm to obtain an evaluation value of the flue gas waste heat recovery device; The evaluation value of the flue gas waste heat recovery device is compared with a preset evaluation threshold, and when it is determined that the evaluation value of the flue gas waste heat recovery device is greater than or equal to the preset evaluation threshold, the fault processing is completed.

9. The fault detection and processing method for a flue gas waste heat recovery device according to claim 8, characterized in that: The second preset algorithm includes: Among them, φ represents the evaluation value of the flue gas waste heat recovery device; U represents the voltage value at both ends of the flue gas waste heat recovery device when it is working; I represents the current value at both ends of the flue gas waste heat recovery device when it is working; R represents the internal resistance value of the flue gas waste heat recovery device; t1 represents the time value that the flue gas waste heat recovery device has been used, and t2 represents the ideal use time of the flue gas waste heat recovery device; μ represents the flow index of the flue gas waste heat recovery device; γ represents the heat exchange efficiency of the flue gas waste heat recovery device: θ represents the theoretical remaining service life of the flue gas waste heat recovery device; δ represents the number of failures of the flue gas waste heat recovery device.

10. A fault detection and processing system for a flue gas waste heat recovery device, characterized in that: include: The first acquisition module is used to acquire the operation data of the flue gas waste heat recovery device within a preset period; A preprocessing module, used for preprocessing the operation data; A detection module is used to input the pre-processed operation data into a pre-trained fault detection model for detection and determine the detection result; The second acquisition module is used to obtain the solution strategy corresponding to the detection result; A processing module is used to process the failure of the flue gas waste heat recovery device based on the solution strategy.

Images