System for analyzing boiler fault and diagnosis method

Abstract

A system for analyzing a boiler fault and a diagnosis method, the system comprising: a sensor system, a data acquisition and analysis system, and an early warning system. The sensor system is used to monitor temperature, pressure, and water level indicators inside of a boiler and transmit data. The data acquisition and analysis system is used to acquire, count, and analyze the temperature, the pressure, and the water level inside of the boiler in real time. The data acquisition and analysis system comprises a data acquisition module and an intelligent algorithm module. The data acquisition module is used to acquire real-time temperature, pressure, and water level data inside of the boiler. The intelligent algorithm module is used to analyze and process data and to predict and diagnose an operation state of the boiler. The early warning system comprises an early warning module. The early warning module is used to receive a signal of a fault model and to send, on the basis of the signal, a first-stage early warning signal and second-stage early warning information. Thus, real-time monitoring and early warning of the operation state of the boiler is achieved, and problems are promptly discovered and measures can be taken, therefore normal operation of the boiler is guaranteed, and production safety is guaranteed.

Description

A system and diagnostic method for analyzing boiler failure Technical Field

[0001] The present application relates to the technical field of boiler monitoring and early warning, and more particularly to a system and a diagnostic method for analyzing boiler failures. Background Art

[0002] With the continuous development of industrialization, boilers have become indispensable equipment in many industrial sites. However, due to various reasons, boilers may encounter various problems during operation, such as incomplete combustion, excessively high or low water levels, leakage, etc. These problems may cause boiler failure and even endanger personnel safety.

[0003] Therefore, it is particularly important to carry out early warning and monitoring of boilers at all times. Technical issues

[0004] How to achieve real-time early warning and monitoring of boilers. Technical Solutions

[0005] In response to the above-mentioned deficiencies in the prior art, the present application provides a system for analyzing boiler failures, which can realize real-time monitoring and early warning of boiler operating status, detect problems in a timely manner and take measures, thereby ensuring the normal operation of the boiler and ensuring production safety.

[0006] In order to solve the above technical problems, the technical solution adopted in this application is: a system for analyzing boiler failures, which includes a sensor system, a data acquisition and analysis system, and an early warning system;

[0007] The sensor system is used to monitor the temperature, pressure and water level indicators inside the boiler and transmit the data to the data acquisition system;

[0008] The data acquisition and analysis system is used to collect the temperature, pressure, and water level inside the boiler in real time, and to perform statistics and analysis on the collected data. The data acquisition and analysis system includes a data acquisition module and an intelligent algorithm module. The data acquisition module is used to collect real-time data on the temperature, pressure, and water level inside the boiler transmitted from sensors, and the intelligent algorithm module is used to analyze and process the data and predict and diagnose the operating status of the boiler through a fault model.

[0009] The early warning system includes an early warning module, which is used to receive signals from the fault model and issue a first-level early warning signal and a second-level early warning information according to the signals.

[0010] By adopting the above structure, real-time monitoring of the boiler's operating status and fault warning can be achieved, and the fault type and location can be accurately diagnosed, so that timely measures can be taken to ensure production safety; among them, the operating status of the boiler can be monitored in time by sensors, so that problems can be quickly discovered; by using intelligent algorithms to analyze data through intelligent algorithm modules, the operating status of the boiler can be predicted and faults can be diagnosed more accurately; then, through the use of the fault warning system, the fault type and location can be accurately diagnosed, and targeted measures can be taken; therefore, the above scheme of the present application can avoid the occurrence of production safety accidents and improve safety through the use of the fault warning system, and provide a safe and efficient boiler fault warning system, which can play an important role in industrial production and effectively prevent and reduce the safety and economic losses caused by boiler failures.

[0011] Furthermore, the sensor system includes a temperature sensor for monitoring the internal temperature of the boiler, a pressure sensor for monitoring the internal pressure of the boiler, and a water level sensor for monitoring the internal water level of the boiler; adopting the above structure, each corresponding indicator and specific position inside the boiler can be accurately monitored, providing a guarantee for the accurate judgment of subsequent boiler failures.

[0012] Furthermore, the boiler failure includes at least a steam temperature failure, a steam pressure failure and a boiler water level failure inside the boiler; by setting the above at least three failure problems of the boiler, the operation safety of the boiler can be further guaranteed.

[0013] Furthermore, the early warning procedure for the steam temperature failure is as follows: the steam temperature value inside the boiler is collected, and the temperature sensor is checked at this temperature value to see whether it is normal. If it is normal, it is further determined whether the steam temperature is higher than the highest flue gas temperature (threshold) of normal boiler operation in the past ten days. If it is higher than the above flue gas temperature, it is determined that the steam temperature of the boiler is too high and an early warning is issued; if the temperature sensor is abnormal, it is determined to be a temperature sensor failure.

[0014] Furthermore, the early warning procedure for the steam pressure failure is as follows: collecting the steam pressure value inside the boiler, observing whether the pressure sensor is normal under this pressure value, and if normal, further determining whether the steam pressure is higher than a preset maximum steam pressure (threshold value); if higher than the above maximum steam pressure, determining that the steam pressure in the boiler is too high and issuing an early warning; if the pressure sensor is abnormal, determining that the pressure sensor is faulty.

[0015] Furthermore, the judgment procedure for the boiler water level failure is as follows: collecting the water level value inside the boiler, observing whether the pressure sensor is normal under the water level value, and if normal, further judging whether the water level value is higher than a preset high water level (high threshold value). If it is higher than the above-mentioned preset high water level, judging that the water level in the boiler is too high and issuing an early warning; or, further judging whether the water level value is lower than a preset low water level (low threshold value). If it is lower than the above-mentioned preset low water level, judging that the water level in the boiler is too low and issuing an early warning; if the water level sensor is abnormal, it is judged that the water level sensor is faulty.

[0016] Furthermore, the present application also provides a method for fault diagnosis using a system for analyzing boiler faults, comprising:

[0017] (1) Monitor the temperature, pressure, and water level inside the boiler through the sensor system and transmit the data to the data acquisition system;

[0018] (2) The real-time detection data of (1) is collected through the data acquisition and analysis system, and the collected data is statistically analyzed and analyzed; the real-time data of the temperature, pressure, and water level inside the boiler transmitted by the sensor is obtained through the data acquisition module, and the data statistics and analysis are analyzed and processed by the intelligent algorithm module, and the operating status of the boiler is predicted and diagnosed through the fault model;

[0019] (3) The early warning system is activated based on the prediction and diagnosis results of (2). The early warning module of the early warning system receives the signal of the fault model and issues a first-level early warning signal and a second-level early warning information according to the signal.

[0020] Furthermore, the first-level warning signal can be dialed by phone, and the second-level warning information can be pushed to a WeChat public account to remind operators to pay attention to the operating status of the boiler.

[0021] Furthermore, the intelligent algorithm module performs analysis and processing, specifically: time series prediction is carried out by collecting and analyzing the historical operating data of the equipment, and establishing a corresponding prediction model to represent the potential relationship between historical observations and future values, and finally inferring the operating conditions in a certain period of time in the future through the time series.

[0022] Furthermore, the prediction is to use an LSTM neural network algorithm to predict the future operation data of the boiler, using 70% of the historical data as a training set and 30% as a test set to evaluate the performance of the model.

[0023] Furthermore, the implementation of the LSTM neural network algorithm consists of three main steps: 1) Forward calculation: output vector of each neuron; 2) Backward calculation: error term δ of each neuron; 3) Update of weight gradients. Based on the needs of network learning, LSTM learns weight matrices and their corresponding biases for various gate structures and cell states.

[0024] Furthermore, the implementation of the LSTM neural network algorithm is as follows:

[0025] (1) First, preprocess the data to remove some abnormal values ​​and null values; then perform differential transformation on the data, and then standardize the data. The calculation formula (I) for standardization is:

[0026] (I)

[0027] Where x* is the normalized value, x is the original value, min is the minimum value, and max is the maximum value;

[0028] (2) Let z t is the time series generated by the characteristic variables after the above processing; it is assumed that an a-order nonlinear regression model can be used to describe the historical observations (z t-1 、z t-2 、……、z t-a ) and predicted values ​​(z t 、z t+1 、……、z t+b-1 ), specifically the relationship between the following formula (II):

[0029] (II)

[0030] Where: f is a nonlinear equation; ε is a random error;

[0031] (3) Use LSTM to approximate the unknown equation f; then the observed value and the predicted value are expressed as follows when the relationship is established using LSTM:

[0032] (III)

[0033] Where: fˆ is the approximate nonlinear equation of f; θ is the LSTM parameter set;

[0034] (4) LSTM uses historical data to train and update the parameters θ to minimize the prediction error E, which is shown in the following formula (VI): (VI)

[0035] In order to evaluate the model concisely and intuitively, the mean square error is selected as the model evaluation index. The smaller the mean square error, the closer the model prediction value is to the actual value. The mean square error MSE is shown in the following formula (V): (V)

[0036] Where yi is the true value, ˆyi is the predicted value, and n is the total number of samples;

[0037] (5) After training, the obtained model is within a reasonable error range. Then, the steam pressure change level value and the steam temperature change level value are calculated respectively; the steam pressure change level value is the mode of the predicted error data of the steam pressure, and the steam temperature change level value is the mode of the predicted error data of the steam temperature. If the real-time collected data exceeds the predicted level change threshold, an early warning message is issued. Beneficial effects

[0038] 1. Real-time monitoring: Sensors monitor the operating status of the boiler in real time and can quickly detect problems.

[0039] 2. Intelligent analysis: Using intelligent algorithms to analyze data can more accurately predict the operating status of the boiler and diagnose faults.

[0040] 3. Accurate diagnosis: Through the use of fault warning systems, the type and location of faults can be accurately diagnosed, and targeted measures can be taken.

[0041] 4. Improve safety: The use of fault warning system can avoid production safety accidents and improve safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 Overall boiler early warning flow chart.

[0043] Figure 2 Steam temperature fault warning flow chart.

[0044] Figure 3 Steam pressure fault warning flow chart.

[0045] Figure 4 Boiler water level fault warning flow chart.

[0046] Figure 5 is a schematic diagram of the connection diagram of the system of this application. Best Mode for Carrying Out the Invention

[0047] The present application is further described in detail below through specific examples, but the present application is not limited to the following examples.

[0048] As shown in FIG5 , the present application provides a system for analyzing boiler failures, which includes a sensor system, a data acquisition and analysis system, and an early warning system;

[0049] The sensor system is used to monitor the temperature, pressure and water level indicators inside the boiler and transmit the data to the data acquisition system;

[0050] The data acquisition and analysis system is used to collect the temperature, pressure, and water level inside the boiler in real time, and to perform statistics and analysis on the collected data. The data acquisition and analysis system includes a data acquisition module and an intelligent algorithm module. The data acquisition module is used to collect real-time data on the temperature, pressure, and water level inside the boiler transmitted from sensors, and the intelligent algorithm module is used to analyze and process the data and predict and diagnose the operating status of the boiler through a fault model.

[0051] The early warning system includes an early warning module, which is used to receive signals from the fault model and issue a first-level early warning signal and a second-level early warning information according to the signals.

[0052] By adopting the above structure, real-time monitoring of the boiler's operating status and fault warning can be achieved, and the fault type and location can be accurately diagnosed, so that timely measures can be taken to ensure production safety; among them, the operating status of the boiler can be monitored in time by sensors, so that problems can be quickly discovered; by using intelligent algorithms to analyze data through intelligent algorithm modules, the operating status of the boiler can be predicted and faults can be diagnosed more accurately; then, through the use of the fault warning system, the fault type and location can be accurately diagnosed, and targeted measures can be taken; therefore, the above scheme of the present application can avoid the occurrence of production safety accidents and improve safety through the use of the fault warning system, and provide a safe and efficient boiler fault warning system, which can play an important role in industrial production and effectively prevent and reduce the safety and economic losses caused by boiler failures.

[0053] As an example, the sensor system described in this application includes a temperature sensor for monitoring the internal temperature of the boiler, a pressure sensor for monitoring the internal pressure of the boiler, and a water level sensor for monitoring the water level inside the boiler; by adopting the above structure, each corresponding indicator and specific position inside the boiler can be accurately monitored, providing a guarantee for the accurate judgment of subsequent boiler failures.

[0054] As an example, the boiler failure described in this application includes at least steam temperature failure, steam pressure failure and boiler water level failure inside the boiler; by setting at least three of the above-mentioned boiler failure problems, the safe operation of the boiler can be further guaranteed.

[0055] As shown in Figure 2, this is the early warning procedure for steam temperature failure described in this application: the steam temperature value inside the boiler is collected, and the temperature sensor is checked at this temperature value to see whether it is normal. If it is normal, it is further determined whether the steam temperature is higher than the highest flue gas temperature (threshold) of normal boiler operation in the past ten days. If it is higher than the above flue gas temperature, it is determined that the steam temperature of the boiler is too high and an early warning is issued; if the temperature sensor is abnormal, it is determined to be a temperature sensor failure.

[0056] As shown in Figure 3, the early warning procedure for steam pressure failure described in this application is: collecting the steam pressure value inside the boiler, observing whether the pressure sensor is normal under this pressure value, and if normal, further determining whether the steam pressure is higher than the preset maximum steam pressure (threshold). If it is higher than the above-mentioned maximum steam pressure, it is determined that the steam pressure in the boiler is too high and an early warning is issued; if the pressure sensor is abnormal, it is determined to be a pressure sensor failure.

[0057] As shown in Figure 4, the judgment procedure for the boiler water level failure described in this application is as follows: collecting the water level value inside the boiler, observing whether the pressure sensor is normal under the water level value, and if normal, further judging whether the water level value is higher than a preset high water level (high threshold). If it is higher than the above-mentioned preset high water level, it is judged that the water level in the boiler is too high and an early warning is issued; or, further judging whether the water level value is lower than a preset low water level (low threshold). If it is lower than the above-mentioned preset low water level, it is judged that the water level in the boiler is too low and an early warning is issued; if the water level sensor is abnormal, it is judged that the water level sensor is faulty.

[0058] With reference to FIG1 , the present application further provides a method for fault diagnosis using a system for analyzing boiler faults, comprising:

[0059] (1) Monitor the temperature, pressure, and water level inside the boiler through the sensor system and transmit the data to the data acquisition system;

[0060] (2) The real-time detection data of (1) is collected through the data acquisition and analysis system, and the collected data is statistically analyzed and analyzed; the real-time data of the temperature, pressure, and water level inside the boiler transmitted by the sensor is obtained through the data acquisition module, and the data statistics and analysis are analyzed and processed by the intelligent algorithm module, and the operating status of the boiler is predicted and diagnosed through the fault model;

[0061] (3) The early warning system is activated based on the prediction and diagnosis results of (2). The early warning module of the early warning system receives the signal of the fault model and issues a first-level early warning signal and a second-level early warning information according to the signal.

[0062] The first-level warning signal described in this application can be dialed by phone, and the second-level warning information can be pushed to a WeChat public account to remind operators to pay attention to the operating status of the boiler.

[0063] The intelligent algorithm module described in this application performs analysis and processing, specifically: time series prediction is achieved by collecting and analyzing the historical operating data of the equipment, and establishing a corresponding prediction model to represent the potential relationship between historical observations and future values, and finally inferring the operating conditions in a certain period of time in the future through time series.

[0064] The prediction described in this application uses the LSTM neural network algorithm to predict the future operation data of the boiler, using 70% of the historical data as a training set and 30% as a test set to evaluate the performance of the model.

[0065] The implementation of the LSTM neural network algorithm described in this application mainly involves the following three steps: 1) Forward calculation: output vector of each neuron; 2) Backward calculation: error term δ value of each neuron; 3) Update of weight gradient. According to the needs of network learning, LSTM needs to learn the weight matrix and corresponding bias of various gate structures and unit states. Specifically, the implementation of the LSTM neural network algorithm described in this application is as follows:

[0066] (1) First, preprocess the data to remove some outliers and null values; then perform differential transformation on the data, and then perform standardization on the data. The calculation formula (I) for standardization is: (1) First, preprocess the data to remove some outliers and null values; then perform differential transformation on the data, and then perform standardization on the data. The calculation formula (I) for standardization is:

[0067] (I)

[0068] Where x* is the normalized value, x is the original value, min is the minimum value, and max is the maximum value;

[0069] (2) Let z t is the time series generated by the characteristic variables after the above processing; it is assumed that an a-order nonlinear regression model can be used to describe the historical observations (z t-1 、z t-2 、……、z t-a ) and predicted values ​​(z t 、z t+1 、……、z t+b-1 ), specifically the relationship between the following formula (II):

[0070] (II)

[0071] Where: f is a nonlinear equation; ε is a random error;

[0072] (3) Use LSTM to approximate the unknown equation f; then the observed value and the predicted value are expressed as follows when the relationship is established using LSTM:

[0073] (III) Where: fˆ is the approximate nonlinear equation of f; θ is the set of LSTM parameters;

[0074] (4) LSTM uses historical data to train and update the parameters θ to minimize the prediction error E, which is shown in the following formula (VI): (VI)

[0075] In order to evaluate the model concisely and intuitively, the mean square error is selected as the model evaluation index. The smaller the mean square error, the closer the model prediction value is to the actual value. The mean square error MSE is shown in the following formula (V): (V) Where yi is the true value, ˆyi is the predicted value, and n is the total number of samples;

[0076] (5) After training, the obtained model is within a reasonable error range. Then, the steam pressure change level value and the steam temperature change level value are calculated respectively; the steam pressure change level value is the mode of the predicted error data of the steam pressure, and the steam temperature change level value is the mode of the predicted error data of the steam temperature. If the real-time collected data exceeds the predicted level change threshold, an early warning message is issued. Modes for Carrying Out the Invention

[0077] Specifically, the implementation of the LSTM neural network algorithm described in this application is as follows:

[0078] (1) First, preprocess the data to remove some outliers and null values; then perform differential transformation on the data, and then perform standardization on the data. The calculation formula (I) for standardization is: (1) First, preprocess the data to remove some outliers and null values; then perform differential transformation on the data, and then perform standardization on the data. The calculation formula (I) for standardization is:

[0079] (I)

[0080] Where x* is the normalized value, x is the original value, min is the minimum value, and max is the maximum value;

[0081] (2) Let z t is the time series generated by the characteristic variables after the above processing; it is assumed that an a-order nonlinear regression model can be used to describe the historical observations (z t-1 、z t-2 、……、z t-a ) and predicted values ​​(zt 、z t+1 、……、z t+b-1 ), specifically the relationship between the following formula (II):

[0082] (II)

[0083] Where: f is a nonlinear equation; ε is a random error;

[0084] (3) Use LSTM to approximate the unknown equation f; then the observed value and the predicted value are expressed as follows when the relationship is established using LSTM:

[0085] (III) Where: fˆ is the approximate nonlinear equation of f; θ is the set of LSTM parameters;

[0086] (4) LSTM uses historical data to train and update the parameters θ to minimize the prediction error E, which is shown in the following formula (VI): (VI)

[0087] In order to evaluate the model concisely and intuitively, the mean square error is selected as the model evaluation index. The smaller the mean square error, the closer the model prediction value is to the actual value. The mean square error MSE is shown in the following formula (V): (V) Where yi is the true value, ˆyi is the predicted value, and n is the total number of samples;

[0088] (5) After training, the obtained model is within a reasonable error range. Then, the steam pressure change level value and the steam temperature change level value are calculated respectively; the steam pressure change level value is the mode of the predicted error data of the steam pressure, and the steam temperature change level value is the mode of the predicted error data of the steam temperature. If the real-time collected data exceeds the predicted level change threshold, an early warning message is issued. Industrial Applicability

[0089] 1. Real-time monitoring: Sensors monitor the operating status of the boiler in real time and can quickly detect problems.

[0090] 2. Intelligent analysis: Using intelligent algorithms to analyze data can more accurately predict the operating status of the boiler and diagnose faults.

[0091] 3. Accurate diagnosis: Through the use of fault warning systems, the type and location of faults can be accurately diagnosed, and targeted measures can be taken.

[0092] 4. Improve safety: The use of fault warning system can avoid production safety accidents and improve safety. Sequence Listing Free Content

[0093] Type your sequence listing free description paragraph here.

Claims

1. A system for analyzing boiler failure, characterized in that: The system includes a sensor system, a data acquisition and analysis system, and an early warning system; The sensor system is used to monitor the temperature, pressure and water level indicators inside the boiler and transmit the data to the data acquisition system; The data acquisition and analysis system is used to collect the temperature, pressure and water level inside the boiler in real time, and to perform statistics and analysis on the collected data; the data acquisition and analysis system includes a data acquisition module and an intelligent algorithm module, the data acquisition module is used to collect the real-time data of the temperature, pressure and water level inside the boiler transmitted from the sensor, the intelligent algorithm module is used to analyze and process the data, and to predict and diagnose the operating status of the boiler through the fault model; The early warning system comprises an early warning module, which is used to receive a signal of a fault model and send out a first-level early warning signal and a second-level early warning information according to the signal.

2. The system for analyzing boiler failure according to claim 1, characterized in that: The sensor system includes a temperature sensor for monitoring the temperature inside the boiler, a pressure sensor for monitoring the pressure inside the boiler, and a water level sensor for monitoring the water level inside the boiler.

3. The system for analyzing boiler failure according to claim 1, characterized in that: The boiler failure at least includes a steam temperature failure, a steam pressure failure and a boiler water level failure inside the boiler.

4. The system for analyzing boiler failure according to claim 3, characterized in that: The early warning procedure for the steam temperature failure is as follows: collecting the steam temperature value inside the boiler, observing whether the temperature sensor is normal under this temperature value, and if normal, further judging whether the steam temperature is higher than the highest smoke temperature of the normal boiler operation in the past ten days. If it is higher than the above smoke temperature, judging that the steam temperature of the boiler is too high and issuing an early warning; if the temperature sensor is abnormal, judging that the temperature sensor is faulty.

5. The system for analyzing boiler failure according to claim 3, characterized in that: The early warning procedure for the steam pressure failure is as follows: collecting the steam pressure value inside the boiler, observing whether the pressure sensor is normal under the pressure value, and if normal, further determining whether the steam pressure is higher than the preset maximum steam pressure. If it is higher than the above maximum steam pressure, determining that the steam pressure in the boiler is too high and issuing an early warning; if the pressure sensor is abnormal, determining that the pressure sensor is faulty.

6. The system for analyzing boiler failure according to claim 3, characterized in that: The judgment procedure for the boiler water level failure is as follows: collect the water level value inside the boiler, observe whether the pressure sensor is normal under the water level value, and if normal, further judge whether the water level value is higher than a preset high water level. If it is higher than the above preset high water level, judge that the water level in the boiler is too high and issue an early warning; or, further judge whether the water level value is lower than a preset low water level. If it is lower than the above preset low water level, judge that the water level in the boiler is too low and issue an early warning; if the water level sensor is abnormal, it is judged to be a water level sensor failure.

7. A method for fault diagnosis using the system for analyzing boiler faults according to any one of claims 1 to 6, characterized in that: include: (1) Monitor the temperature, pressure, and water level indicators inside the boiler through the sensor system and transmit the data to the data acquisition system; (2) The real-time detection data of (1) is collected through the data collection and analysis system, and the collected data is counted and analyzed; the real-time data of the temperature, pressure and water level inside the boiler transmitted by the sensor is obtained through the data collection module, and the data statistics and analysis are analyzed and processed by the intelligent algorithm module, and the operating status of the boiler is predicted and diagnosed through the fault model; (3) The early warning system is activated based on the prediction and diagnosis results of (2). The early warning module of the early warning system receives the signal of the fault model and issues a first-level early warning signal and a second-level early warning information based on the signal.

8. The fault diagnosis method according to claim 7, characterized in that: The first-level warning signal can be dialed by phone, and the second-level warning information can be pushed to a WeChat public account to remind the operator to pay attention to the operating status of the boiler.

9. The fault diagnosis method according to claim 7, characterized in that: The intelligent algorithm module performs analysis and processing, specifically: time series prediction collects and analyzes the historical operation data of the equipment, and establishes a corresponding prediction model to represent the potential relationship between historical observations and future values, and finally infers the operating conditions in a certain period of time in the future through the time series.

10. The fault diagnosis method according to claim 9, characterized in that: The prediction adopts LSTM neural network algorithm to predict the future operation data of the boiler, and uses 70% of the historical data as the training set and 30% as the test set to evaluate the performance of the model; the implementation of the LSTM neural network algorithm mainly includes the following three steps: 1) forward calculation: each neuron output vector; 2) reverse calculation: each neuron error term δ value; 3) weight gradient update; according to the needs of network learning, LSTM needs to learn the weight matrix of various gate structures and unit states and their corresponding biases; specifically: (1) First, preprocess the data to remove some outliers and null values; then perform differential transformation on the data, and then standardize the data. The calculation formula (I) for standardization is: (I) Where x* is the standardized value, x is the original value, min is the minimum value, and max is the maximum value; (2) Let z t is the time series of the characteristic variables after the above processing; it is assumed that an a-order nonlinear regression model can be used to describe the historical observations (z t-1 、z t-2 ,……,z t-a ) and predicted value (z t 、z t+1 ,……,z t+b-1 ), specifically, the relationship between them is as follows (II): (II) Where: f is a nonlinear equation; ε is a random error; (3) Use LSTM to approximate the unknown equation f; then the observed value and the predicted value are expressed by the following formula (III) when using LSTM to establish a relationship: (III) Where: fˆ is the approximate nonlinear equation of f; θ is the LSTM parameter set; (4) LSTM uses historical data to train and update the parameter θ to minimize the prediction error E, which is shown in the following formula (VI): (VI) In order to evaluate the model concisely and intuitively, the mean square error is selected as the model evaluation index. The smaller the mean square error, the closer the model prediction value is to the actual value. The mean square error MSE is shown in the following formula (V): (V) Where yi is the true value, ˆyi is the predicted value, and n is the total number of samples; (5) After training, the obtained model is within a reasonable error range, and then the steam pressure change level value and the steam temperature change level value are calculated respectively; the steam pressure change level value is the mode of the predicted error data of the steam pressure, and the steam temperature change level value is the mode of the predicted error data of the steam temperature; if the real-time collected data is greater than the predicted level change threshold, an early warning message is issued.

Images