Tank area digital twin model construction method and tank area digital twin application system

Abstract

The embodiment of the application provides a kind of tank area digital twin model construction method and corresponding digital twin model, belong to digital twin technical field and safety engineering technical field.The construction method includes: based on tank area whole process twinborn data, the data model suitable for tank area digital twin model is constructed;The data model is corrected based on tank area real-time field data, including making the data model: utilize deep learning algorithm to process the tank area real-time field data, to predict the change of the tank area real-time field data in future period;And corresponding entity three-dimensional model of tank area is loaded into the corrected data model, to form tank area digital twin model.The application realizes the real-time, dynamic, whole process twinborn system modeling for tank area.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of Chinese invention patent application 202210238446.2 filed on March 11, 2022, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present invention relates to the fields of digital twin technology and safety engineering technology, and in particular to a method for constructing a digital twin model of a tank area and a corresponding digital twin model. Background Art

[0004] Storage tanks are essential infrastructure for industries such as petroleum, chemical engineering, grain and oil, food, fire protection, transportation, metallurgy, and national defense. With the continuous development of the national economy, the volume and number of individual storage tanks are constantly increasing, and the scale of tank farms is also expanding accordingly. In the complex chemical industry, storage tanks often contain large quantities of flammable, explosive, and toxic chemicals. Accidents such as fires, explosions, poisoning, and asphyxiation can cause significant economic losses and casualties, negatively impacting both businesses and society.

[0005] However, current emergency response to storage tank accidents often relies solely on pre-planned responses due to the uncertainty of tank operating conditions, which lacks specificity. In an effort to address this issue, the inventors discovered the potential for building a digital twin model of the tank farm, which could be used to predict potential changes in the tank farm during an accident.

[0006] Based on this, this application proposes a solution for building a digital twin model of a tank area. Summary of the Invention

[0007] The purpose of the embodiments of the present invention is to provide a method for constructing a digital twin model of a tank area, a digital twin model, a digital twin application system for a tank area, and a machine-readable storage medium, so as to at least partially solve the problem that emergency treatment of storage tanks can only be carried out according to plans.

[0008] In order to achieve the above-mentioned purpose, an embodiment of the present invention provides a method for constructing a digital twin model of a tank area, including: constructing a data model suitable for the digital twin model of the tank area based on the full-process twin data of the tank area; correcting the data model based on the real-time field data of the tank area, and the correction includes enabling the data model to use a deep learning algorithm to process the real-time field data of the tank area to predict the changes in the real-time field data of the tank area in future time periods; and loading the physical three-dimensional model corresponding to the tank area into the corrected data model to form a digital twin model of the tank area.

[0009] Furthermore, the construction method also includes: obtaining tank area body data, and performing multiple preprocessings on the tank area body data for digital twins to obtain the full-process twin data of the tank area.

[0010] Furthermore, the tank farm entity data includes: historical data of tank farm equipment from various control systems of the tank farm; process system information from various system analysis processes of the tank farm, wherein the process system information includes any one or more of design data, experience data, online data, fault data, panel data, analysis data and expert knowledge data; and real-time field data of the tank farm.

[0011] Furthermore, the multiple preprocessings include any one or more of the following: analyzing the dimensions and correlations of the tank area body data; detecting and filtering out errors and noise in the tank area body data; detecting missing data in the tank area body data, and correcting the missing data based on the material and energy balance of the tank area process and / or using the time series data feature vector to supplement the missing data; maintaining the dynamics and real-time nature of the tank area real-time field data; and studying the nonlinear characteristics of the tank area body data to ensure the consistency of the formed tank area full-process twin data with the tank area body data.

[0012] Furthermore, the following steps are adopted to analyze the correlation of the tank area body data: the time series data feature vector of the tank area body data is used as the input of the partial correlation coefficient or the extension correlation function, so as to measure the correlation of the corresponding data by using the partial correlation coefficient method or the extension correlation function method.

[0013] Furthermore, the time series data feature vector includes qualitative trend features and quantitative trend features, and is obtained by the following steps: establishing trend primitives for the tank area body data in time periods using a triangular time model, and extracting the qualitative trend features of single variables for the trend primitives using the interval half-division method or the fuzzy similarity method; and using a multi-scale wavelet analysis method to extract the frequency domain sequence statistical features of the tank area body data as quantitative trend features.

[0014] Furthermore, the data model is configured as a five-dimensional data model for tank farm process big data, wherein the five dimensions correspond to the time domain, the object domain, the relationship domain, the attribute and the attribute value respectively.

[0015] Furthermore, the use of a deep learning algorithm to process the real-time on-site data of the tank area includes: configuring a time series neural network, a time series neural network whose current output is affected by the result of the previous moment; and using the time series neural network to process the real-time on-site data of the tank area to predict changes in the real-time on-site data of the tank area in future time periods.

[0016] Furthermore, the construction method also includes: when the real-time on-site data of the tank area is disaster site data, constructing a disaster simulation model based on the predicted changes of the disaster site data in future time periods; and integrating the disaster simulation model into the digital twin model of the tank area.

[0017] Furthermore, the construction method also includes: using a deep learning algorithm to construct a tank area abnormality warning model for variable trend changes and disturbances in the real-time on-site data of the tank area, and / or using a graph deep neural network to construct a tank area abnormality tracing model for abnormal operating condition data in the real-time on-site data of the tank area; and integrating the tank area abnormality warning model and / or the tank area abnormality tracing model into the tank area digital twin model.

[0018] Furthermore, the construction method also includes: establishing a process mechanism model based on the physical properties of the tanks in the tank area, the material properties of the stored materials and the chemical process characteristics; using logistics analysis methods and information flow analysis methods to establish a data-driven model or soft measurement model of key process variables in the tank area; and integrating the process mechanism model, the data-driven model and / or the soft measurement model into the digital twin model of the tank area.

[0019] Furthermore, the construction method also includes: simulating the entire life cycle of the tank area based on the formed digital twin model of the tank area, and performing abnormality analysis, parameter optimization and / or operation control on the tank area according to the simulation results.

[0020] On the other hand, the present invention also provides a digital twin model of a tank farm, comprising: a physical three-dimensional model of the tank farm, which serves as a support for the digital twin model of the tank farm; a data model constructed based on the full-process twin data of the tank farm, and the data model is configured to process the real-time field data of the tank farm using a deep learning algorithm to predict changes in the real-time field data of the tank farm in future time periods; and any one or more of the following models:

[0021] In the case where the real-time on-site data of the tank farm is disaster on-site data, a disaster simulation model is constructed based on the predicted changes of the disaster on-site data in future time periods;

[0022] A process mechanism model based on the physical characteristics of the tanks in the tank farm, the material properties of the materials stored in the tanks, and the chemical process characteristics;

[0023] Data-driven models or soft-sensing models of key process variables in the tank farm constructed using logistics analysis methods and information flow analysis methods;

[0024] A tank farm abnormality warning model constructed using a deep learning algorithm based on variable trend changes and disturbances in the real-time field data of the tank farm; and

[0025] Aiming at the abnormal operating condition data in the real-time field data of the tank area, a tank area abnormality tracing model is constructed using a graph deep neural network.

[0026] On the other hand, the present invention also provides a tank area digital twin application system, including: a data twin layer, on which the above-mentioned tank area digital twin model is constructed, and the tank area digital twin model is used to simulate the entire life cycle process of the tank area; and a twin application layer, which communicates with the data twin layer and includes an abnormality analysis module, a parameter optimization module and / or an operation control module, wherein the abnormality analysis module, the parameter optimization module and the operation control module are respectively used to perform corresponding abnormality analysis, parameter optimization and / or operation control on the tank area according to the simulation results of the tank area digital twin model.

[0027] The present invention also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute any of the above-mentioned methods for constructing a digital twin model of a tank area.

[0028] Through the above technical solution, the present invention provides a highly real-time, multi-temporal, multi-scale, dynamic digital twin modeling solution for tank areas. Starting from the system perspective, it constructs a multi-dimensional, multi-temporal, and multi-scale twin model, and enables the twin model to have a high-confidence model prediction and analysis function, realizing real-time, dynamic, and full-process twin system modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings:

[0030] Figure 1 1 is a flow chart of a method for constructing a digital twin model of a tank farm according to the first embodiment of the present invention;

[0031] Figure 2 It is a schematic diagram of the general structure of a temporal neural network;

[0032] Figure 3 1 is a flow chart of a method for constructing a digital twin model of a tank farm according to a second embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of the principle of the method for acquiring and preprocessing tank farm data in the second embodiment of the present invention;

[0034] Figure 5 This is a schematic diagram of the exemplary method for extracting feature vectors from time series data;

[0035] Figure 6is a schematic diagram of the principle of forming a digital twin model of a tank farm in the sixth embodiment of the present invention; and

[0036] Figure 7 It is a structural diagram of the digital twin application system of the tank area of ​​embodiment seven of the present invention. DETAILED DESCRIPTION

[0037] The following describes the specific implementation of the embodiment of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiment of the present invention and is not used to limit the embodiment of the present invention.

[0038] Example 1

[0039] Figure 1 FIG. 1 is a flow chart of a method for constructing a digital twin model of a tank farm according to the first embodiment of the present invention. Figure 1 As shown, the method for constructing the digital twin model of the tank farm may include the following steps:

[0040] Step S100: Based on the full-process twin data of the tank area, a data model suitable for the digital twin model of the tank area is constructed.

[0041] A digital twin model is a model that maps the entire lifecycle of the corresponding physical equipment in a virtual space (also known as a twin space) by fully utilizing the physical model, sensor updates, and operational history of the equipment. This data is then integrated into a multidisciplinary, multi-physics, multi-scale, and multi-probability simulation process. This model then reflects the entire lifecycle of the corresponding physical equipment. Thus, full-process twin data for a tank farm refers to the data required to map the physical tank farm into its digital twin model.

[0042] It can be understood by those skilled in the art that, based on the information-physical mapping relationship between the physical space tank area scene and the virtual space tank area scene, when obtaining the tank area entity data (i.e., the tank area data under the physical space tank area scene), it is easy to form a data model that represents the twin data of the entire process of the tank area and is suitable for the digital twin model of the tank area.

[0043] In addition, in an embodiment of the present invention, the data model is configured as a five-dimensional data model for tank area process big data, wherein the five dimensions correspond to the time domain, object domain, relationship domain, attribute, and attribute value, respectively. For example, based on the multi-scale, multi-level, and difficult-to-unify characteristics of the twin data of the entire process of the tank area, the triple (subject, attribute, attribute value) knowledge description method of the unified resource description framework (RDF) is used to draw on the knowledge description method of the triple (subject, attribute, attribute value) of the unified resource description framework (RDF). A five-dimensional data model (time domain, object domain, relationship domain, attribute, attribute value) for the tank area process big data is proposed, wherein the relationship domain includes generalized relationships such as similarity relationship, indistinguishable relationship, order relationship, fuzzy relationship, functional similarity relationship, and projection. By unifying the description of the data model structure, the mutual conversion and unified description of information between structured and unstructured twin data (operating procedures, operating experience) are ultimately achieved. Furthermore, based on the description of the five-dimensional model structure of the twin data, based on the multi-level information space mapping relationship from the physical level of process, equipment, etc. to the comprehensive indicators, operation indicators, and control indicators, effective data mapping from the twin space to the physical space can be achieved.

[0044] Step S200: Correcting the data model based on real-time field data of the tank farm.

[0045] As a twin model, the data model is required to present consistent data changes with the tank farm itself. Therefore, the data model needs to be corrected based on the real-time data from the tank farm. This correction can include modifying the data represented in the data model to adapt to the real-time data from the tank farm.

[0046] However, in embodiments of the present invention, this correction preferably includes enabling the data model to process the real-time on-site data of the tank farm using a deep learning algorithm to predict changes in the real-time on-site data of the tank farm in future periods. In other words, embodiments of the present invention aim to enable the data model to predict possible changes in the tank farm in advance.

[0047] Preferably, the use of a deep learning algorithm to process the real-time on-site data of the tank area may include: configuring a time series neural network, a time series neural network whose current output is affected by the result of the previous moment; and using the time series neural network to process the real-time on-site data of the tank area to predict changes in the real-time on-site data of the tank area in future time periods.

[0048] For example, based on massive twin data of the entire process of the tank farm, combined with the process mechanism (the process mechanism refers to the internal working mode of each element in a certain system structure and the operating rules and principles of the elements' mutual connection and interaction under a certain environment in order to achieve a specific function during the operation process; it can be used to describe chemical processes and is easy for relevant operators to obtain), a temporal neural network modeling method is proposed with the input and output of the device or unit in the tank farm as the guide. Among them, the general structure of the temporal neural network is as follows Figure 2 As shown in Figure 2, the hidden layer input at each moment includes the current input and the output of the hidden layer at the previous moment. Therefore, the information of the hidden layer and the current output are affected by the results of the current and previous moments, and this influence is transmitted step by step. Furthermore, the forward propagation process of a temporal neural network can be expressed by the following formula:

[0049]

[0050] Among them, α represents the result of weighted calculation, b represents the value calculated by activation function, and w hk Represents the weight between the hidden layer and the output layer, w ih Represents the weight between the input layer and the hidden layer, w hh' Represents the weight of the hidden layer state at the previous moment and the current hidden layer state.

[0051] In this way, the data-driven time series neural network modeling method effectively combines the structural characteristics of the neural network with the characteristics of twin data (the characteristic is "able to change in real time"), so that based on the data change patterns in the previous period, it can effectively learn the changes in data in the future period and realize the early prediction of the twin model.

[0052] Step S300: Load the physical three-dimensional model corresponding to the tank area into the corrected data model to form a digital twin model of the tank area.

[0053] Preferably, the physical 3D model is a high-fidelity virtual space model obtained by processing a known tank farm mechanism model and the tank farm entity data using 3D visualization technology. For example, based on system operation data, analytical data, and empirical data, the tank farm mechanism model and 3D visualization technology are applied to create a high-fidelity virtual space model, achieving a virtual-physical mapping of the real.

[0054] It should be noted that for the solid 3D model of the tank farm, it is easy to see that the tank farm is a system consisting of storage tanks and pipelines. The pipeline transmission process is a typical pure hysteresis process. If the model is based on pure hysteresis, the computational complexity of the simulation model will increase. Therefore, in the simulation modeling process, the distributed parameter model of the pipeline is simplified by using lumped parameters to reduce the modeling difficulty and improve the calculation speed.

[0055] In summary, the first embodiment of the present invention provides a highly real-time, multi-temporal, multi-scale, dynamic digital twin modeling method for a tank area. This method, from a system perspective, constructs a multi-dimensional, multi-temporal, multi-scale, and mutually integrated tank area twin model, and enables the tank area twin model to have a high-confidence model prediction and analysis function, thus realizing real-time, dynamic, and full-process twin system modeling. Specifically, the first embodiment of the present invention establishes a digital twin model for a tank area that can simulate the operating status of the tank area and various abnormal and accident scenarios in the tank area, which is of great significance for studying the dangers of fires in petrochemical enterprises' storage tank areas, understanding the development process of accidents, and taking effective emergency response measures in accident states.

[0056] Example 2

[0057] According to Example 1, it is clear that the full-process twin data of the tank farm is the basis for building the corresponding twin model. This data is obtained based on the tank farm's main body data. However, the tank farm's main body data is massive, and its acquisition process seriously affects the speed of model construction. In response to this, Example 2 of the present invention, based on Example 1, proposes a scheme for acquiring full-process twin data of the tank farm suitable for realizing digital twins and a corresponding preprocessing scheme for the tank farm's main body data.

[0058] In the second embodiment of the present invention, Figure 3 As shown, in Figure 1 The steps of the construction method shown also include:

[0059] Step S400: Acquire the tank area ontology data corresponding to the full-process twin data of the tank area, and perform multiple preprocessings for the digital twin on the tank area ontology data to obtain the full-process twin data of the tank area.

[0060] Preferably, the tank farm body data includes: historical data of tank farm equipment from various control systems of the tank farm; and process system information from various system analysis processes of the tank farm, wherein the process system information includes any one or more of design data, experience data, online data, fault data, panel data, analysis data and expert knowledge data; and real-time field data of the tank farm.

[0061] Among them, various control systems in the tank area, such as DCS (Distributed Control System), FCS (Fieldbus Control System), MES (Manufacturing Execution System), LIMS (Laboratory Information Management System), etc., store historical data of tank area equipment; process system information can be understood as procedural information related to the historical data of tank area equipment.

[0062] In a preferred embodiment, a tank farm entity database may be configured to store the above-mentioned tank farm entity data, and historical data and real-time data may be stored separately to facilitate calling according to circumstances.

[0063] Regarding the various tank area data involved here, Figure 4 It is a schematic diagram of the principle of the method for acquiring and preprocessing tank area data in the second embodiment of the present invention.

[0064] refer to Figure 4 , historical data of tank farm equipment is obtained from control systems such as DCS, FCS, MES, LIMS, PLC (Programmable Logic Controller), and ERP (Enterprise Resource Planning). This data is combined with process system information such as design data, experience data, online data, fault data, panel data, and real-time field data of the tank farm to form the tank farm data. The multiple pre-processing steps for the digital twin of the tank farm data can include any one or more of the following:

[0065] 1) Analyze the dimension and relevance (also called data association) of the tank farm ontology data.

[0066] In terms of correlation, it is preferred to use the time series data feature vector of the tank area body data as the input of the partial correlation coefficient or the extension correlation function, so as to measure the correlation of the corresponding data by using the partial correlation coefficient method or the extension correlation function method. For example, the system analyzes the characteristics of the data flow, and based on the system functions, characteristics, correlation, etc., determines the length of the observation time series window in the observation time interval of each variable, proposes the multivariate system partial correlation coefficient method and the extension correlation function method to measure the correlation of the sequence data, and establishes the complex connection between the time series fluctuation data; due to the dynamic changes of the time series data, the correlation obtained by using the sampling points of the fluctuation data sequence cannot reflect the overall characteristics of the time series data, and the calculation complexity is very large, so it is proposed to use the feature vector (mean, variance, amplitude, rate of change) based on the sequence data as the input of the partial correlation coefficient and the correlation function, obtain the partial correlation coefficient and the correlation matrix of the multivariate system, and analyze the correlation between the data.

[0067] In other embodiments, principal component analysis or angle vector method may be used to analyze the data correlation.

[0068] Furthermore, the purpose of dimensional analysis is to reduce the dimensionality of high-dimensional data using data reduction and dimensionality reduction methods while preserving the original data as much as possible, thereby streamlining the data volume. Furthermore, based on known correlations, multi-sensor data optimization and fusion methods can be used to achieve data decorrelation and dimensionality reduction.

[0069] 2) Detect and filter out errors and noise in the tank farm data.

[0070] For example, in order to detect and filter out errors and noise in the data, offline and online data sliding mean methods, data filtering methods (such as wavelet filtering methods), and outlier detection methods based on data clustering (such as high cluster point detection) are used.

[0071] 3) Detecting missing data in the tank farm data, and correcting the missing data based on the material and energy balance of the tank farm process and / or supplementing the missing data using time series data feature vectors.

[0072] Here, by correcting and / or supplementing the missing data, it is helpful to ensure a complete data space. In addition, soft measurement estimation and transformation methods can also be used to correct or supplement the missing data, which is a conventional optional method for those skilled in the art.

[0073] 4) Maintain the dynamic and real-time performance of the tank farm's real-time field data.

[0074] For example, in view of the dynamic and real-time nature of data, real-time data acquisition interface technologies such as OPC or DDE, as well as data compression, transmission and storage technologies are used to ensure real-time acquisition of device operation data (or field data).

[0075] 5) Study the nonlinear characteristics of the tank area body data to ensure the consistency of the formed tank area full-process twin data with the tank area body data.

[0076] For example, in view of the nonlinear time series relationship of data, the Gaussian / non-Gaussian characteristic expression of dynamic process data can be studied. Kernel learning methods or other nonlinear feature analysis methods can be used to extract dynamic nonlinear data features to ensure the accuracy, consistency and reliability of process data.

[0077] The above content involves the time series data feature vector. In a preferred embodiment, the time series data feature vector includes qualitative trend features and quantitative trend features, and can be obtained by the following steps: establishing trend primitives for the tank area data in a triangular time mode and extracting the qualitative trend features of single variables from the trend primitives using the interval half-division method or the fuzzy similarity method; and extracting the frequency domain sequence statistical features of the tank area data as quantitative trend features using the multi-scale wavelet analysis method. Figure 5 The acquisition of the time series data feature vector is described in more detail.

[0078] Figure 5 This is a schematic diagram of the time series data feature vector extraction method. Figure 5 , the extraction method mainly includes the following three parts:

[0079] First, for key process variables, based on the data moving window method and the single variable time series trend analysis method, it is proposed to establish trend primitives based on piecewise polynomial sequence approximation and triangular time mode in different time periods. The time series can be approximately expressed as a trend primitive composed of a binary group (sequence change rate, speed of change rate), namely the first-order and second-order differential symbols, to qualitatively describe the trend characteristics of the variable. Generally, seven trend primitives of variable sequences are defined, namely A(0,0), B(+,0), C(+,-), D(+,+), E(-,0), F(-,+), G(-,-), and then it is proposed to use the interval half-interpretation method or fuzzy similarity method to extract the qualitative trend characteristics of the single variable.

[0080] Second, for key process variables, multi-scale wavelet analysis method is used to extract frequency domain feature information, such as sequence amplitude, mean, variance, etc., to establish a quantitative description of system variable behavior, namely quantitative trend characteristics.

[0081] Third, the qualitative trend features and quantitative trend features are combined to form the process state / behavior, which is the time series data feature vector used to map features to the twin space.

[0082] In summary, the second embodiment of the present invention obtains the full-process twin data of the tank area based on multiple preprocessing schemes for the tank area main body data, which provides data support for the construction of the digital twin model of the tank area. Moreover, through the cooperation of multiple preprocessing schemes, the complexity of massive tank area data is simplified, which is conducive to improving the efficiency of constructing the digital twin model of the tank area.

[0083] Example 3

[0084] Based on the above embodiments, the method for constructing a digital twin model of a tank area in embodiment three of the present invention also includes: when the real-time on-site data of the tank area is disaster site data, constructing a disaster simulation model based on the predicted changes of the disaster site data in future time periods; and integrating the disaster simulation model into the digital twin model of the tank area.

[0085] This third embodiment uses disaster simulation as an important component of building a twin model. Common disasters in tank areas include leaks, explosions, and fires. Computational fluid dynamics software (e.g., CFD software) can be used to simulate these three disaster scenarios to generate corresponding disaster models suitable for the twin model, such as pool fire models and explosion models. However, the grid calculation method used by CFD software is not suitable for real-time simulation. Therefore, in actual calculations, a simplified modeling method can be used, applying empirical formulas, segmented calculations, and other methods to simulate disasters.

[0086] The third embodiment of the present invention further enriches the functions of the digital twin model of the tank area by adding a disaster simulation model, thereby realizing disaster prediction and improving the practicality of the digital twin model of the tank area.

[0087] Example 4

[0088] The first embodiment of the present invention provides a digital twin model of a tank area mainly composed of a physical three-dimensional model and a data model. In the fourth embodiment of the present invention, the digital twin model of the tank area is further optimized so that it integrates more types of models.

[0089] The method for constructing the digital twin model of the tank area in the fourth embodiment of the present invention is Figure 1 The corresponding construction method of embodiment 1 may further include:

[0090] Step S500 (not shown in the figure) uses a deep learning algorithm to build a tank area abnormality warning model for the variable trend changes and disturbances in the real-time field data of the tank area, and / or uses a graph deep neural network to build a tank area abnormality tracing model for the abnormal operating condition data in the real-time field data of the tank area.

[0091] For example, tank farms are typical slow-changing systems. By applying a contribution graph approach to anomaly identification based on canonical variable analysis, and leveraging process big data, we can track and identify variable trend changes and disturbances through methods such as slow feature analysis, thereby establishing a deep learning-based tank farm anomaly early warning model. We are also researching anomaly monitoring and tracing methods based on slow feature analysis and Bayesian causal analysis, and establishing anomaly tracing models for pipeline blockages, leaks, sensor anomalies, and other issues based on graph deep networks.

[0092] Step S600 (not shown in the figure) integrates the tank area abnormality warning model and / or the tank area abnormality tracing model into the tank area digital twin model.

[0093] Specifically, through the above steps, the tank area abnormality warning model is configured to calculate the variable trend changes and disturbances in the real-time field data of the tank area, and issue warnings based on this; the tank area abnormality tracing model is configured to determine the abnormal operating condition data in the real-time field data of the tank area to trace the abnormal operating condition.

[0094] The fourth embodiment of the present invention further integrates a tank area abnormality warning model and a tank area abnormality tracing model into the tank area digital twin model. These two models essentially utilize the characteristics of the corrected data model that can "predict the changes in the real-time field data of the tank area in future time periods", specifically predict abnormal working conditions, and guide early warning processing and tracing work.

[0095] Example 5

[0096] Implementations three and four of the present invention improve the solution of implementation one from the perspective of expanding the prediction function of the tank area digital twin model, while implementation five of the present invention further optimizes the tank area digital twin model of implementation one from the perspective of the characteristics of the tank area itself, so that it can integrate more types of models.

[0097] The method for constructing the digital twin model of the tank area in the fifth embodiment of the present invention is Figure 1 The construction method of the corresponding embodiment may further include:

[0098] In step S700 (not shown in the figure), a corresponding process mechanism model is established based on the physical properties of the storage tanks in the tank farm, the material properties of the stored materials, and the chemical process characteristics.

[0099] The material characteristics of stored materials include physical and chemical properties. For example, the material stored in a storage tank may be in a liquid, gaseous, or gas-liquid mixed state. The types of materials are diverse. Using a complex material system characterization method using benchmark pseudo-components, based on data such as distillation fraction and distillation temperature from distillation tests, a series of pseudo-components or continuous component distributions are segmented to establish a calculation model for their physical and chemical properties. This calculation then yields process data for the storage tank under different operating conditions, which is then used to establish a process mechanism model.

[0100] Among them, the chemical process characteristics mainly refer to the three transfer and one reaction principle of chemical process, namely heat transfer, mass transfer, momentum transfer and chemical reaction process.

[0101] In step S800 (not shown in the figure), a data-driven model or a soft-sensing model of key process variables of the tank farm is established by using a logistics analysis method and an information flow analysis method.

[0102] For example, the system analyzes the process flow diagram (PFD diagram) and the process piping and instrumentation flow diagram (P&ID diagram), and combines the process mechanism model and system hierarchy, series and parallel production operation mode, based on the system logistics analysis method and information flow analysis method, to establish a data-driven model or soft measurement model of key process variables. The model can be established using methods such as neural network, support vector machine regression, principal component analysis (PCA), and partial least squares (PLS).

[0103] Step S900 (not shown in the figure) integrates the process mechanism model, the data-driven model and / or the soft measurement model into the tank farm digital twin model.

[0104] Among them, the process mechanism model is, for example, the model of the oil-slurry system. By generating a special twin model for it, big data and artificial intelligence technology can be integrated to improve the simulation, analysis, prediction and optimization capabilities of the oil-slurry system, and realize virtual control of the real; key process variables are, for example, the coking procedure, ambient temperature, flow rate, etc. of the oil-slurry system. By establishing data-driven models and / or soft measurement models for these variables, the changes in the corresponding variables can be predicted to provide early warning and traceability guidance.

[0105] In a preferred embodiment, for step S900, the construction method may further include: for example, using fuzzy hierarchical analysis method to evaluate the model results of the process mechanism model, the data-driven model or the soft measurement model, and correcting and coordinating the process mechanism model according to the evaluation results.

[0106] The fifth embodiment of the present invention improves the simulation, analysis, prediction and optimization capabilities of the process mechanism and key process variables by configuring the process mechanism model, the data-driven model of the key process variables and / or the soft measurement model, thereby realizing virtual control of the real, which is conducive to the virtual-reality mapping and feedback guidance of the digital twin system of the tank area.

[0107] Example 6

[0108] Example 6 provides the structure of the digital twin model of the tank area that integrates Example 1, Example 3, Example 4 and Example 5. Figure 6 This is a schematic diagram of the principle of forming a digital twin model of a tank farm in the sixth embodiment of the present invention. Figure 6 The digital twin model of the tank farm of the sixth embodiment of the present invention may include: a physical three-dimensional model of the tank farm, which serves as a support for the digital twin model of the tank farm; a data model constructed based on the full-process twin data of the tank farm, and the data model is configured to process the real-time field data of the tank farm using a deep learning algorithm to predict changes in the real-time field data of the tank farm in future time periods; and any one or more of the following models:

[0109] 1) A process mechanism model is constructed based on the physical characteristics of the tanks in the tank farm, the material characteristics of the materials stored in the tanks, and the characteristics of the chemical process.

[0110] 2) Data-driven model or soft measurement model of key process variables of the tank area constructed using logistics analysis methods and information flow analysis methods.

[0111] 3) A tank area abnormality warning model is constructed using a deep learning algorithm based on the variable trend changes and disturbances in the real-time field data of the tank area.

[0112] 4) A tank area abnormality tracing model is constructed using a graph deep neural network for abnormal operating condition data in the tank area's real-time field data.

[0113] 5) When the real-time on-site data of the tank farm is disaster on-site data, a disaster simulation model is constructed based on the predicted changes of the disaster on-site data in future time periods.

[0114] For the specific implementation details of each model, please refer to the above embodiments and will not be repeated here.

[0115] The digital twin model of the tank area in embodiment six of the present invention mirrors the physical space tank area scene through virtual space, realizing information interaction and virtual monitoring between the digital twin and the physical entity, so that the digital twin model of the tank area can be used to perform visual intelligent monitoring of the physical space of the tank area.

[0116] Example 7

[0117] Based on the digital twin model of the tank area that has been constructed in the above embodiment, the seventh embodiment of the present invention further optimizes the construction method from the application perspective of the digital twin model of the tank area.

[0118] Embodiment 7 of the present invention provides a method for constructing a digital twin model of a tank area. Compared with the construction method of the above embodiment, it also includes: simulating the entire life cycle of the tank area based on the formed digital twin model of the tank area, and performing abnormality analysis, parameter optimization and / or operation control of the tank area based on the simulation results.

[0119] That is, the seventh embodiment of the present invention is equivalent to adding a twin application layer to the digital twin model of the tank area to form a digital twin application system of the tank area. Figure 7 The structure of the digital twin application system of the tank area of ​​the seventh embodiment of the present invention is shown, including: a data twin layer, which constructs the digital twin model of the tank area of ​​the above embodiment, and the digital twin model of the tank area may include a process mechanism model, a data model, etc. as in the fifth embodiment; a twin application layer, including an abnormality analysis module, a parameter optimization module and / or an operation control module to perform corresponding abnormality analysis, parameter optimization and / or operation control.

[0120] Among them, the data twin layer can be specifically understood as including a physical layer, a data layer and a model layer. The physical layer corresponds to the tank area body, the data layer, for example, includes a database for storing the above-mentioned tank area body data, and the model layer, for example, includes the tank area digital twin model of the above-mentioned embodiment.

[0121] Among them, the abnormality analysis refers to the corresponding processing of the abnormalities existing in the tank area by analyzing the abnormal data simulated by the digital twin model of the tank area; the parameter optimization refers to the simulation results of the tank area simulated by the digital twin model of the tank area, discovering the optimizable points of each tank area device, so as to optimize the various process flows of the entire tank area; the operation control refers to the simulation results of the tank area simulated by the digital twin model of the tank area, knowing the corresponding tank area site operation status and predicting the operation status in the future period, so as to judge whether there is a risk of disaster accidents, and accordingly carry out dynamic early warning, traceability and other work.

[0122] Based on this, the seventh embodiment of the present invention can use the digital twin model of the tank area to monitor the real-time data of the tank area site, promptly discover abnormal working conditions on site, and dynamically monitor the tank area site by performing ultra-real-time simulation on the digital twin model, predict disasters and accidents, and conduct early warning and tracing.

[0123] An embodiment of the present invention also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the method for constructing a digital twin model of a tank area described in the above embodiment.

[0124] An embodiment of the present invention provides a processor, which is used to run a program, wherein when the program is running, the method for constructing a digital twin model of a tank area described in the above embodiment is executed.

[0125] An embodiment of the present invention provides a device, which includes a processor, a memory, and a program stored in the memory and runnable on the processor. When the processor executes the program, the method for constructing a digital twin model of the tank area described in the above embodiment is implemented.

[0126] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of initializing the method for constructing a digital twin model of a tank area described in the above embodiment.

[0127] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0128] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0129] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0131] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0132] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0133] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0134] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0135] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for constructing a digital twin model of a tank farm, characterized in that: The construction method comprises: The tank farm data is obtained, and multiple preprocessing steps for the digital twin are performed on the tank farm data to obtain the full-process twin data of the tank farm, wherein the multiple preprocessing steps include: Detect missing data in the tank farm data and supplement the missing data using a time series data feature vector, wherein the time series data feature vector includes a qualitative trend feature and a quantitative trend feature and is obtained by the following steps: The trend primitives are established for the tank farm data in different time periods using a triangular time model, and Extracting qualitative trend characteristics of single variables based on trend primitives using the interval half-split method or fuzzy similarity method; and Using a multi-scale wavelet analysis method, extracting frequency domain sequence statistical features of the tank farm ontology data as quantitative trend features; Based on the full-process twin data of the tank farm, a data model suitable for the digital twin model of the tank farm is constructed; based on the real-time field data of the tank farm, the data model is corrected, wherein the correction includes enabling the data model to process the real-time field data of the tank farm using a deep learning algorithm to predict changes in the real-time field data of the tank farm in future time periods; and The physical three-dimensional model corresponding to the tank area is loaded into the corrected data model to form a digital twin model of the tank area.

2. The method for constructing a digital twin model of a tank farm according to claim 1, characterized in that: The tank farm data includes: Historical data of tank farm equipment from various control systems at the tank farm; Process system information from various system analysis processes at the tank farm, wherein the process system information includes any one or more of design data, experience data, online data, failure data, panel data, analysis data, and expert knowledge data; and The tank farm real-time on-site data.

3. The method for constructing a digital twin model of a tank farm according to claim 1, characterized in that: The multiple pre-processing steps may also include any one or more of the following: Analyzing the dimensions and relevance of the tank farm ontology data; Detecting and filtering out errors and noise in the tank farm ontology data; Correcting the missing data based on material and energy balances of the tank farm processes; Maintaining the dynamic and real-time nature of the tank farm's real-time field data; and The nonlinear characteristics of the tank area body data are studied to ensure the consistency of the formed tank area full-process twin data with the tank area body data.

4. The method for constructing a digital twin model of a tank farm according to claim 3, characterized in that: The following steps are used to analyze the relevance of the tank farm ontology data: The time series data feature vector of the tank area ontology data is used as the input of the partial correlation coefficient or the extension correlation function, so as to measure the correlation of the corresponding data by using the partial correlation coefficient method or the extension correlation function method.

5. The method for constructing a digital twin model of a tank farm according to claim 1, characterized in that: The data model is configured as a five-dimensional data model for tank farm process big data, wherein the five dimensions correspond to a time domain, an object domain, a relationship domain, an attribute, and an attribute value, respectively.

6. The method for constructing a digital twin model of a tank farm according to claim 1, characterized in that: The use of a deep learning algorithm to process the real-time field data of the tank farm includes: Configuring a temporal neural network whose current output is affected by the result at the previous moment; and The time series neural network is used to process the real-time field data of the tank area to predict changes in the real-time field data of the tank area in future time periods.

7. The method for constructing a digital twin model of a tank farm according to claim 1, characterized in that: The construction method further comprises: In a case where the real-time on-site data of the tank farm is disaster on-site data, constructing a disaster simulation model based on predicted changes of the disaster on-site data in future time periods; and Integrate the disaster simulation model into the digital twin model of the tank farm.

8. The method for constructing a digital twin model of a tank farm according to any one of claims 1 to 7, characterized in that: The construction method further comprises: Based on the trend changes and disturbances of variables in the real-time field data of the tank farm, a deep learning algorithm is used to build a tank farm abnormality early warning model, and / or based on the abnormal operating condition data in the real-time field data of the tank farm, a graph deep neural network is used to build a tank farm abnormality tracing model; and The tank area abnormality warning model and / or the tank area abnormality tracing model are integrated into the tank area digital twin model.

9. The method for constructing a digital twin model of a tank farm according to any one of claims 1 to 7, characterized in that: The construction method further comprises: Establish a process mechanism model based on the physical characteristics of the tanks in the tank farm, the material characteristics of the stored materials, and the chemical process characteristics; Use logistics analysis methods and information flow analysis methods to establish data-driven models and / or soft sensor models for key process variables in the tank farm; and The process mechanism model, the data-driven model and / or the soft sensor model are integrated into the tank farm digital twin model.

10. The method for constructing a digital twin model of a tank farm according to any one of claims 1 to 7, characterized in that: The construction method further comprises: The digital twin model of the tank farm is used to simulate the entire life cycle of the tank farm, and abnormality analysis, parameter optimization and / or operation control of the tank farm are carried out based on the simulation results.

11. A digital twin application system for a tank farm, characterized in that: The tank farm digital twin application system includes: A data twin layer, on which is configured a tank farm digital twin model constructed by the construction method of any one of claims 1 to 10, and the tank farm digital twin model is used to simulate the entire life cycle process of the tank farm; and The twin application layer communicates with the data twin layer and includes an anomaly analysis module, a parameter optimization module and / or an operation control module, wherein the anomaly analysis module, the parameter optimization module and / or the operation control module are respectively used to perform corresponding anomaly analysis, parameter optimization and / or operation control on the tank area according to the simulation results of the tank area digital twin model.

12. The digital twin application system for the tank farm according to claim 11, characterized in that: The digital twin model of the tank farm includes: A physical three-dimensional model of the tank farm, which serves as a support for the digital twin model of the tank farm; A data model built based on the full-process twin data of the tank farm, and the data model is configured to process the real-time field data of the tank farm using a deep learning algorithm to predict changes in the real-time field data of the tank farm in future time periods; Any one or more of the following models: In the case where the real-time on-site data of the tank farm is disaster on-site data, a disaster simulation model is constructed based on the predicted changes of the disaster on-site data in future time periods; A process mechanism model based on the physical characteristics of the tanks in the tank farm, the material properties of the materials stored in the tanks, and the chemical process characteristics; Data-driven models or soft-sensing models of key process variables in the tank farm constructed using logistics analysis methods and information flow analysis methods; A tank farm abnormality warning model constructed using a deep learning algorithm based on variable trend changes and disturbances in the real-time field data of the tank farm; and Aiming at the abnormal operating condition data in the real-time field data of the tank area, a tank area abnormality tracing model is constructed using a graph deep neural network.

13. A machine-readable storage medium having instructions stored thereon, wherein the instructions are used to enable a machine to execute the method for constructing a digital twin model of a tank area as described in any one of claims 1 to 10.

Images