Chemical production data anomaly detection and safety early warning system based on deep learning

Abstract

The invention discloses a chemical production data anomaly detection and safety early warning system based on deep learning, relates to the field of computer and chemical crossed data anomaly detection, and solves the problem that a deep learning model for analyzing acquired data is lacked. The method comprises the following steps of: transmitting acquired temperature, pressure and material flow data to a data processing module for preprocessing, so as to obtain qualified values of the temperature, the pressure and the material flow; giving different weights to construct a parameter total deviation evaluation value in combination with a deep learning model prediction value and a measurement value; and according to the parameter total deviation evaluation value and the parameter total error score evaluation value, constructing a parameter total anomaly calculation value, and according to the parameter total anomaly evaluation calculation value, carrying out grade division on the key equipment position and sending out a corresponding safety early warning signal.

Description

Technical Field

[0001] The present invention belongs to the field of data anomaly detection at the intersection of computers and chemical industry, and specifically relates to a chemical production data anomaly detection and safety early warning system based on deep learning. Background Art

[0002] With the continuous development of artificial intelligence technology, deep learning, as a powerful data analysis tool, has important research value and practical significance in the application of anomaly detection in chemical production. Data analysis and anomaly detection are important links to ensure production safety and product quality. It is very important to accurately control the temperature, pressure and material flow data of key equipment locations. According to the total deviation evaluation value of the parameter and the total error score value of the parameter, the total anomaly calculation value of the parameter is constructed with corresponding weights. The total anomaly evaluation calculation value of the parameter is used to analyze and start the safety warning signal level, which effectively ensures the stability and safety of the chemical production process.

[0003] Although the existing chemical production data anomaly detection and safety warning system has achieved chemical production data anomaly detection and safety warning to a certain extent, the traditional anomaly detection method lacks a deep learning model to analyze the collected data, and it is difficult to comprehensively consider the deviation of multiple parameters, whether the detection equipment may have potential risks, and issue different levels of safety warning signals for key equipment locations. It is difficult to adapt to the complex and changeable chemical production environment. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a chemical production data anomaly detection and safety early warning system based on deep learning, which is used to solve the following technical problems:

[0005] Although the existing chemical production data anomaly detection and safety warning system has achieved chemical production data anomaly detection and safety warning to a certain extent, the traditional anomaly detection method lacks a deep learning model to analyze the collected data, and it is difficult to comprehensively consider the deviation of multiple parameters, whether the detection equipment may have potential risks, and issue different levels of safety warning signals for key equipment locations. It is difficult to adapt to the complex and changeable chemical production environment.

[0006] To solve the above problems, the first aspect of the present invention provides a chemical production data anomaly detection and safety early warning system based on deep learning, comprising the following modules:

[0007] Data acquisition module: transmits the temperature, pressure and material flow data collected by the temperature sensor, pressure sensor and flow sensor to the data processing module;

[0008] Data processing module: pre-process the collected temperature, pressure and material flow data, including cleaning and filtering, to obtain qualified values ​​of temperature, pressure and material flow data;

[0009] Data analysis module: Use deep learning models to train the processed data, and obtain three data deviation evaluation values ​​based on the model's parameter deviations of the predicted values, qualified values, and measured values ​​of temperature, pressure, and material flow data. The total parameter deviation evaluation value is obtained by assigning different data weights;

[0010] Anomaly detection module: Use the trained deep learning model to perform anomaly detection on real-time data to detect whether its output shows an abnormal state. If not, the process ends. Otherwise, the total abnormal score of the parameter is calculated to execute the safety warning module.

[0011] Safety warning module: Based on the calculated value of the total abnormal evaluation of parameters, the abnormal conditions of chemical production data are classified into different levels and safety warnings are issued.

[0012] As a further solution of the present invention: the data acquisition module comprises the following steps:

[0013] Install temperature sensors, pressure sensors and flow sensors at key equipment locations in the chemical production process to obtain temperature, pressure and material flow data and transmit them to the data processing module;

[0014] The key equipment locations include the control area, reactor area, separation equipment area, heat transfer equipment area and storage equipment area.

[0015] As a further solution of the present invention: the data processing module pre-processes the collected temperature, pressure and material flow data, including cleaning and filtering, to obtain qualified values ​​of the temperature, pressure and material flow data, including the following steps:

[0016] S201: receiving temperature, pressure and material flow data from the data acquisition module, building a cleaning plug-in that meets a specific preset value range through the FineDataLink data integration platform, and performing cleaning operations on the collected data;

[0017] S202: Build a filter plug-in that meets a specific preset value range through the FineDataLink data integration platform to operate on the cleaned data;

[0018] S203: Record the minimum and maximum values ​​of temperature, pressure and material flow rate through the cleaned and filtered data, and calculate the qualified values ​​of the corresponding data.

[0019] As a further solution of the present invention: the qualified value of the corresponding data in step S203 includes the following steps:

[0020] Through the minimum and maximum temperature values, the temperature qualified value calculation formula is obtained:

[0021]

[0022] Where Tmax is the maximum temperature, Tmin is the minimum temperature, Ti is the temperature measurement value, n is the number of measurements, and T is the qualified temperature value;

[0023] Through the minimum and maximum pressure values, the pressure qualified value calculation formula is obtained:

[0024]

[0025] The safety factor α is a constant value of 1.5 in the calculation formula, Pmax is the maximum pressure, Pmin is the minimum pressure, Pi is the pressure measurement value, n is the number of measurements, and P is the pressure qualification value;

[0026] Through the minimum and maximum values ​​of material flow, the qualified calculation formula of material flow is obtained:

[0027]

[0028] Among them, Qmax is the maximum value of material flow, Qmin is the minimum value of material flow, Qi is the measured value of logistics flow, n is the number of measurements, and Q is the qualified value of material flow.

[0029] As a further solution of the present invention: the data analysis module comprises the following steps:

[0030] A deep learning model is used to predict the processed temperature, pressure and material flow values, and the deviations between the predicted values, qualified values ​​and measured values ​​are calculated to obtain three data deviation values;

[0031] Parameter prediction value calculation formula:

[0032]

[0033] Among them, Xs is the predicted value of temperature, pressure and material flow, Xmax is the maximum value of the corresponding parameter, Xmin is the minimum value of the corresponding parameter, Xi is the measured value of the corresponding parameter, and β is a constant of 0.5 in this calculation formula;

[0034] The temperature deviation evaluation calculation formula is obtained through the temperature prediction value, qualified value and measured value:

[0035]

[0036] Where T is the qualified temperature value, Ti is the measured temperature value, Ts is the predicted temperature value, θ T is the temperature deviation evaluation value;

[0037] The pressure deviation evaluation calculation formula is obtained through the pressure prediction value, qualified value and measured value:

[0038]

[0039] Among them, P is the qualified pressure value, Pi is the pressure measurement value, Ps is the pressure prediction value, θ P is the pressure deviation evaluation value;

[0040] Through the material flow prediction value, qualified value and measured value, the material flow deviation evaluation calculation formula is obtained:

[0041]

[0042] Among them, Q is the qualified value of material flow, Qi is the measured value of material flow, Qs is the predicted value of material flow, θ Q is the material flow deviation evaluation value;

[0043] The total evaluation calculation formula for parameter deviation is constructed based on the temperature, pressure and material flow deviation values:

[0044] D1=θ T ×W T +θ P ×W P +θ Q ×W Q

[0045] Among them, D1 is the total deviation evaluation value of the parameter, θ T is the temperature deviation evaluation value, θ P is the pressure deviation evaluation value, θ Q is the material flow deviation evaluation value, W T , W P , W Q are the corresponding weights, W T =50%, W P 30%, W Q 20%;

[0046] As a further solution of the present invention: the abnormality detection module comprises the following steps:

[0047] The real-time preprocessed data is input into the trained model, and the model outputs the prediction results. The total error score evaluation value and the total deviation evaluation value of the parameters of the real-time data are calculated according to the output results of the model. The total abnormal score calculation value of the parameters is compared with the preset threshold. If it exceeds the preset threshold, it is judged as an abnormal state and the safety warning module is triggered;

[0048] Temperature, pressure, and material flow error fraction calculation formula:

[0049] E X =||X original -X reconstructed ||

[0050] Among them, X corresponds to the three variables of temperature, pressure and material flow, E T is the temperature error fraction value, E P is the pressure error fraction value, E Q is the material flow error fraction value, X original is the original input data corresponding to various parameters, X reconstracted Reconstructed data output by the autoencoder corresponding to each type of data;

[0051] Parameter total error score evaluation calculation formula:

[0052] D2=E T ×W T +E P ×W P +E Q ×W Q

[0053] Among them, D2 is the total error score evaluation value of the parameter, E T 、E P 、E Q are the error fraction values ​​corresponding to temperature, pressure, and material flow, respectively, W T , W P , W Q are the corresponding weights, W T =50%, W P 30%, W Q 20%;

[0054] The total parameter anomaly evaluation calculation formula is obtained based on the parameter total deviation evaluation value and the parameter total error score evaluation value:

[0055] D=(D1×W1+D2×W2)×100%

[0056] Among them, D1 is the parameter total deviation evaluation value, D2 is the parameter total error score evaluation value, D is the parameter total abnormality calculation evaluation value, W1 and W2 are the weights corresponding to the parameter total deviation evaluation value and the parameter total error score value, respectively, W1 is 40%, and W2 is 60%.

[0057] As a further solution of the present invention: the safety warning module comprises the following steps:

[0058] According to the total abnormal evaluation calculation value, the safety warning signals can be divided into three levels:

[0059] If 0%<D<30%, the safety warning signal is a mild warning;

[0060] If 30%<D<70%, the safety warning signal is a moderate warning;

[0061] If 70% < D, the safety warning signal is a high warning;

[0062] There are three levels of adjustment:

[0063] In case of a mild warning, a level 3 safety warning signal will be issued for key locations;

[0064] In case of moderate warning, a secondary safety warning signal will be issued for key locations;

[0065] When high alert is in effect, a first-level safety warning signal will be issued for key locations.

[0066] Beneficial effects of the present invention:

[0067] The present invention discloses a chemical production data anomaly detection and safety early warning system based on deep learning, which involves the field of data anomaly detection at the intersection of computers and chemical industry, solves the technical problem of accurately calculating the total anomaly evaluation value and analyzing it, and obtains the corresponding qualified value by transmitting the collected temperature, pressure and material flow data to the data processing module for cleaning and denoising; combining the deep learning model prediction value and the measured value to give different weights to construct the parameter total deviation evaluation value; constructing the parameter total anomaly calculation value according to the parameter total deviation evaluation value and the parameter total error score value, and realizing the classification of key equipment positions and issuing corresponding safety early warning signals. Real-time data processing and analysis ensures the timely discovery of abnormal situations in the production process, and comprehensively considers the deviation and error values ​​of multiple parameters to detect whether there are potential risks in the equipment, thereby improving the accuracy and efficiency of detecting potential risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 Schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] See also Figure 1 As shown, the present invention is a chemical production data anomaly detection and safety early warning system based on deep learning, including the following modules:

[0072] Data acquisition module: transmits the temperature, pressure and material flow data collected by the temperature sensor, pressure sensor and flow sensor to the data processing module;

[0073] Data processing module: pre-process the collected temperature, pressure and material flow data, including cleaning and filtering, to obtain qualified values ​​of temperature, pressure and material flow data;

[0074] Data analysis module: Use deep learning models to train the processed data, and obtain three data deviation evaluation values ​​based on the model's parameter deviations of the predicted values, qualified values, and measured values ​​of temperature, pressure, and material flow data. The total parameter deviation evaluation value is obtained by assigning different data weights;

[0075] Anomaly detection module: Use the trained deep learning model to perform anomaly detection on real-time data to detect whether its output shows an abnormal state. If not, the process ends. Otherwise, the total abnormal score of the parameter is calculated to execute the safety warning module.

[0076] Safety warning module: Based on the calculated value of the total abnormal evaluation of parameters, the abnormal conditions of chemical production data are classified into different levels and safety warnings are issued.

[0077] Specifically, the collected temperature, pressure and material flow data are transmitted to the data processing module for preprocessing to obtain qualified values ​​of temperature, pressure and material flow; different weights are assigned to the predicted value and measured value of the deep learning model to construct the average value of the total deviation of the parameter; the total abnormality calculation value of the parameter is constructed according to the total deviation evaluation value of the parameter and the total error score of the parameter, and the key equipment locations are graded and corresponding safety warning signals are issued through the total abnormality evaluation calculation value of the parameter. The introduction of deep learning technology improves the accuracy and automation level of anomaly detection.

[0078] In one embodiment of the present invention, the data acquisition module comprises the following steps:

[0079] Install temperature sensors, pressure sensors and flow sensors at key equipment locations in the chemical production process to obtain temperature, pressure and material flow data and transmit them to the data processing module;

[0080] The key equipment locations include the control area, reactor area, separation equipment area, heat transfer equipment area and storage equipment area.

[0081] Specifically, the control area equipment includes sensors, instruments and automatic control system equipment, the reactor area equipment includes reactors, stirring devices and heating equipment, the separation equipment area includes distillation towers, extraction towers and crystallization towers, the heat transfer equipment area includes heat exchangers, evaporators and condensing gas equipment, and the storage equipment area includes key equipment such as storage tanks and tank trucks. Each device has its specific function and importance. The temperature, pressure and material flow data are collected by installing temperature sensors, pressure sensors and flow sensors on the key equipment and transmitted to the data processing module, thereby realizing the monitoring and control of the production process.

[0082] In one embodiment of the present invention, the data processing module pre-processes the collected temperature, pressure and material flow data, including cleaning and filtering, to obtain qualified values ​​of the temperature, pressure and material flow data, including the following steps:

[0083] S201: receiving temperature, pressure and material flow data from the data acquisition module, building a cleaning plug-in that meets a specific preset value range through the FineDataLink data integration platform, and performing cleaning operations on the collected data;

[0084] S202: Build a filter plug-in that meets a specific preset value range through the FineDataLink data integration platform to operate on the cleaned data;

[0085] S203: Record the minimum and maximum values ​​of temperature, pressure and material flow rate through the cleaned and filtered data, and calculate the qualified values ​​of the corresponding data.

[0086] Specifically, the received temperature, pressure and material flow data are cleaned and filtered through the FineDataLink data integration platform, key data are screened out according to the analysis purpose, unimportant data are excluded, and statistical analysis is performed on the pre-processed data to obtain historical data of temperature, pressure and material flow values. The maximum and minimum values ​​of the data are screened out for corresponding calculations, providing a solid foundation for the monitoring and control of chemical production.

[0087] In one embodiment of the present invention, the qualified value of the corresponding data in step S203 includes the following steps:

[0088] Through the minimum and maximum temperature values, the temperature qualified value calculation formula is obtained:

[0089]

[0090] Where Tmax is the maximum temperature, Tmin is the minimum temperature, Ti is the temperature measurement value, n is the number of measurements, and T is the qualified temperature value;

[0091] Through the minimum and maximum pressure values, the pressure qualified value calculation formula is obtained:

[0092]

[0093] The safety factor α is a constant value of 1.5 in the calculation formula, Pmax is the maximum pressure, Pmin is the minimum pressure, Pi is the pressure measurement value, n is the number of measurements, and P is the pressure qualification value;

[0094] Through the minimum and maximum values ​​of material flow, the qualified calculation formula of material flow is obtained:

[0095]

[0096] Among them, Qmax is the maximum value of material flow, Qmin is the minimum value of material flow, Qi is the measured value of logistics flow, n is the number of measurements, and Q is the qualified value of material flow.

[0097] Specifically, by collecting n measurement times, n measurement values, maximum and minimum values ​​of temperature, pressure and material flow data, a calculation formula for qualified values ​​of temperature, pressure and material flow is constructed to obtain qualified values ​​of corresponding data. Accurate qualified values ​​help ensure that these parameters are strictly controlled at the optimal level, thereby ensuring the quality of the final product.

[0098] In one embodiment of the present invention, the data analysis module comprises the following steps:

[0099] A deep learning model is used to predict the processed temperature, pressure and material flow values, and the deviations between the predicted values, qualified values ​​and measured values ​​are calculated to obtain three data deviation values;

[0100] Parameter prediction value calculation formula:

[0101]

[0102] Among them, Xs is the predicted value of temperature, pressure and material flow, Xmax is the maximum value of the corresponding parameter, Xmin is the minimum value of the corresponding parameter, Xi is the measured value of the corresponding parameter, and β is a constant of 0.5 in this calculation formula;

[0103] The temperature deviation evaluation calculation formula is obtained through the temperature prediction value, qualified value and measured value:

[0104]

[0105] Where T is the qualified temperature value, Ti is the measured temperature value, Ts is the predicted temperature value, θ T is the temperature deviation evaluation value;

[0106] The pressure deviation evaluation calculation formula is obtained through the pressure prediction value, qualified value and measured value:

[0107]

[0108] Among them, P is the qualified pressure value, Pi is the pressure measurement value, Ps is the pressure prediction value, θ P is the pressure deviation evaluation value;

[0109] Through the material flow prediction value, qualified value and measured value, the material flow deviation evaluation calculation formula is obtained:

[0110]

[0111] Among them, Q is the qualified value of material flow, Qi is the measured value of material flow, Qs is the predicted value of material flow, θ Q is the material flow deviation evaluation value;

[0112] The total evaluation calculation formula for parameter deviation is constructed based on the temperature, pressure and material flow deviation values:

[0113] D1=θ T ×W T +θ P ×W P +θ Q ×W Q

[0114] Among them, D1 is the total deviation evaluation value of the parameter, θ T is the temperature deviation evaluation value, θ P is the pressure deviation evaluation value, θ Q is the material flow deviation evaluation value, W T , W P , W Q are the corresponding weights, W T =50%, W P 30%, W Q 20%;

[0115] Specifically, the patterns in historical data are analyzed through the deep learning model, and the temperature, pressure and material flow data values ​​are predicted according to the data prediction calculation formula to obtain the corresponding prediction values. The corresponding parameter deviation evaluation values ​​are obtained through the temperature, pressure and material flow prediction values, qualified values ​​and measured values. Different weights of the deviation evaluation values ​​are assigned according to the importance of the parameters, and the total deviation evaluation value of the parameters is obtained by adding them up. This can more accurately reflect the actual status of the production process and help to conduct overall assessment and management of deviations in the chemical production process.

[0116] In one embodiment of the present invention, the anomaly detection module comprises the following steps:

[0117] The real-time preprocessed data is input into the trained model, and the model outputs the prediction results. The total error score evaluation value and the total deviation evaluation value of the parameters of the real-time data are calculated according to the output results of the model. The total abnormal score calculation value of the parameters is compared with the preset threshold. If it exceeds the preset threshold, it is judged as an abnormal state and the safety warning module is triggered;

[0118] Temperature, pressure, and material flow error fraction calculation formula:

[0119] E X =||X original -X reconstructed ||

[0120] Among them, X corresponds to the three variables of temperature, pressure and material flow, E T is the temperature error fraction value, E P is the pressure error fraction value, E Q is the material flow error fraction value, X original is the original input data corresponding to various parameters, X reconstracted Reconstructed data output by the autoencoder corresponding to each type of data;

[0121] Parameter total error score evaluation calculation formula:

[0122] D2=E T ×W T +E P ×W P +E Q ×W Q

[0123] Among them, D2 is the total abnormal score evaluation value of the parameter, E T 、E P 、E Q are the error fraction values ​​corresponding to temperature, pressure, and material flow, respectively, W T , W P , W Q are the corresponding weights, W T=50%, W P 30%, W Q 20%;

[0124] The total parameter anomaly evaluation calculation formula is obtained based on the parameter total deviation evaluation value and the parameter total error score evaluation value:

[0125] D=(D1×W1+D2×W2)×100%

[0126] Among them, D1 is the parameter total deviation evaluation value, D2 is the parameter total error score evaluation value, D is the parameter total abnormality calculation evaluation value, W1 and W2 are the weights corresponding to the parameter total deviation evaluation value and the parameter total error score evaluation value, respectively, W1 is 40%, and W2 is 60%.

[0127] Specifically, the temperature, pressure and material flow anomaly score calculation formula is constructed based on the reconstructed data output by the autoencoder of the original temperature, pressure and material flow input data and the predicted corresponding data of the trained deep learning model. The error score values ​​with different weights are added according to the importance of the parameters to obtain the total error score value of the parameters. The total deviation evaluation value of the parameters is combined with different weights to obtain the total anomaly evaluation calculation value of the parameters. When the parameters deviate from the normal range, the total anomaly evaluation calculation value will increase, which is of great significance for realizing real-time monitoring and safety early warning of chemical production, and provides a scientific basis.

[0128] In one embodiment of the present invention, the safety warning module includes the following steps:

[0129] According to the total abnormal evaluation calculation value, the safety warning signals can be divided into three levels:

[0130] If 0%<D<30%, the safety warning signal is a mild warning;

[0131] If 30%<D<70%, the safety warning signal is a moderate warning;

[0132] If 70% < D, the safety warning signal is a high warning;

[0133] There are three levels of adjustment:

[0134] In case of a mild warning, a level 3 safety warning signal will be issued for key locations;

[0135] In case of moderate warning, a secondary safety warning signal will be issued for key locations;

[0136] When high alert is in effect, a first-level safety warning signal will be issued for key locations.

[0137] Specifically, by calculating the total abnormal evaluation value, different levels of safety warning signals are issued to key locations at different warning levels, and corresponding analysis and adjustments are made, which can promptly remind operators and managers to pay attention to potential risks, ensure the safety of personnel and equipment, and optimize resource utilization and improve overall operational efficiency.

[0138] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A chemical production data anomaly detection and safety early warning system based on deep learning, characterized in that: Includes the following modules: Data acquisition module: transmits the temperature, pressure and material flow data collected by the temperature sensor, pressure sensor and flow sensor to the data processing module; Data processing module: pre-process the collected temperature, pressure and material flow data, including cleaning and filtering, to obtain qualified values ​​of temperature, pressure and material flow data; Data analysis module: Use deep learning models to train the processed data, and obtain three data deviation evaluation values ​​based on the model's parameter deviations of the predicted values, qualified values, and measured values ​​of temperature, pressure, and material flow data. The total parameter deviation evaluation value is obtained by assigning different data weights; Anomaly detection module: Use the trained deep learning model to perform anomaly detection on real-time data to detect whether its output shows an abnormal state. If not, the process ends. Otherwise, the total abnormal score of the parameter is calculated to execute the safety warning module. Safety warning module: Based on the calculated value of the total abnormal evaluation of parameters, the abnormal conditions of chemical production data are classified into different levels and safety warnings are issued.

2. The chemical production data anomaly detection and safety early warning system based on deep learning according to claim 1 is characterized in that: The data acquisition module comprises the following steps: Install temperature sensors, pressure sensors and flow sensors at key equipment locations in the chemical production process to obtain temperature, pressure and material flow data and transmit them to the data processing module; The key equipment locations include the control area, reactor area, separation equipment area, heat transfer equipment area and storage equipment area.

3. The chemical production data anomaly detection and safety early warning system based on deep learning according to claim 1 is characterized in that: The data processing module pre-processes the collected temperature, pressure and material flow data, including cleaning and filtering, to obtain qualified values ​​of the temperature, pressure and material flow data, including the following steps: S201: receiving temperature, pressure and material flow data from the data acquisition module, building a cleaning plug-in that meets a specific preset value range through the FineDataLink data integration platform, and performing cleaning operations on the collected data; S202: Build a filter plug-in that meets a specific preset value range through the FineDataLink data integration platform to operate on the cleaned data; S203: Record the minimum and maximum values ​​of temperature, pressure and material flow rate through the cleaned and filtered data, and calculate the qualified values ​​of the corresponding data.

4. The chemical production data anomaly detection and safety early warning system based on deep learning according to claim 3 is characterized in that: The qualified value of the corresponding data in step S203 includes the following steps: Through the minimum and maximum temperature values, the temperature qualified value calculation formula is obtained: Where Tmax is the maximum temperature, Tmin is the minimum temperature, Ti is the temperature measurement value, n is the number of measurements, and T is the qualified temperature value; Through the minimum and maximum pressure values, the pressure qualified value calculation formula is obtained: The safety factor α is a constant value of 1.5 in the calculation formula, Pmax is the maximum pressure, Pmin is the minimum pressure, Pi is the pressure measurement value, n is the number of measurements, and P is the pressure qualification value; Through the minimum and maximum values ​​of material flow, the qualified calculation formula of material flow is obtained: Among them, Qmax is the maximum value of material flow, Qmin is the minimum value of material flow, Qi is the measured value of logistics flow, n is the number of measurements, and Q is the qualified value of material flow.

5. The chemical production data anomaly detection and safety early warning system based on deep learning according to claim 1 is characterized in that: The data analysis module includes the following steps: A deep learning model is used to predict the processed temperature, pressure and material flow values, and the deviations between the predicted values, qualified values ​​and measured values ​​are calculated to obtain three data deviation values; Parameter prediction value calculation formula: Among them, Xs is the predicted value of temperature, pressure and material flow, Xmax is the maximum value of the corresponding parameter, Xmin is the minimum value of the corresponding parameter, Xi is the measured value of the corresponding parameter, and β is a constant of 0.5 in this calculation formula; The temperature deviation evaluation calculation formula is obtained through the temperature prediction value, qualified value and measured value: Where T is the qualified temperature value, Ti is the measured temperature value, Ts is the predicted temperature value, θ T is the temperature deviation evaluation value; The pressure deviation evaluation calculation formula is obtained through the pressure prediction value, qualified value and measured value: Among them, P is the qualified pressure value, Pi is the pressure measurement value, Ps is the pressure prediction value, θ P is the pressure deviation evaluation value; Through the material flow prediction value, qualified value and measured value, the material flow deviation evaluation calculation formula is obtained: Among them, Q is the qualified value of material flow, Qi is the measured value of material flow, Qs is the predicted value of material flow, θ Q is the material flow deviation evaluation value; The total evaluation calculation formula for parameter deviation is constructed based on the temperature, pressure and material flow deviation values: D1=θ T ×W T +θ P ×W P +θ Q ×W Q Among them, D1 is the total deviation evaluation value of the parameter, θ T is the temperature deviation evaluation value, θ P is the pressure deviation evaluation value, θ Q is the material flow deviation evaluation value, W T , W P , W Q are the corresponding weights, W T =50%, W P 30%, W Q is 20%.

6. The chemical production data anomaly detection and safety early warning system based on deep learning according to claim 1 is characterized in that: The anomaly detection module comprises the following steps: The real-time preprocessed data is input into the trained model, and the model outputs the prediction results. The total error score calculation value of the parameters of the real-time data and the total deviation evaluation calculation value of the parameters are calculated according to the output results of the model. The total abnormal score calculation value of the parameters is compared with the preset threshold. If it exceeds the preset threshold, it is judged as an abnormal state and the safety warning module is triggered; Temperature, pressure, and material flow error fraction calculation formula: E X =||X original -X reconstructed || Among them, X corresponds to the three variables of temperature, pressure and material flow, E T is the temperature error fraction value, E P is the pressure error fraction value, E Q is the material flow error fraction value, X original is the original input data corresponding to various parameters, X reconstracted Reconstructed data output by the autoencoder corresponding to each type of data; Parameter total error score evaluation calculation formula: D2=E T ×W T +E P ×W P +E Q ×W Q Among them, D2 is the total error score evaluation value of the parameter, E T 、E P 、E Q are the error fraction values ​​corresponding to temperature, pressure, and material flow, respectively, W T , W P , W Q are the corresponding weights, W T =50%, W P 30%, W Q 20%; The total parameter anomaly evaluation calculation formula is obtained based on the total parameter deviation evaluation value and the total parameter error score value: D=(D1×W1+D2×W2)×100% Among them, D1 is the parameter total deviation evaluation value, D2 is the parameter total error score evaluation value, D is the parameter total abnormality calculation evaluation value, W1 and W2 are the weights corresponding to the parameter total deviation evaluation value and the parameter total error score evaluation value, respectively, W1 is 40%, and W2 is 60%.

7. The chemical production data anomaly detection and safety early warning system based on deep learning according to claim 1 is characterized in that: The safety warning module includes the following steps: According to the total abnormal evaluation calculation value, the safety warning signal can be divided into three levels: if 0% < D < 30%, the safety warning signal is a mild warning; If 30%<D<70%, the safety warning signal is a moderate warning; If 70% < D, the safety warning signal is a high warning; There are three levels of adjustment: In case of a mild warning, a level 3 safety warning signal will be issued for key locations; In case of moderate warning, a secondary safety warning signal will be issued for key locations; When high alert is in effect, a first-level safety warning signal will be issued for key locations.

Images