Emergency plan simulation method and system based on digital twinning

Abstract

This invention relates to the field of industrial big data technology, specifically to an emergency response plan simulation method and system based on digital twins. The invention first acquires the node runtime sequence data, system load time sequence data, and historical emergency response plan simulation data for each node in the industrial system. Then, based on the time sequence mutual information between nodes and between nodes and the system load, node correlation coefficients are obtained, and the importance coefficient of each node is comprehensively evaluated. Furthermore, based on historical emergency response plan simulation data, the accident transmission index of each node is obtained, and then the data synchronization frequency of each node is comprehensively determined and synchronized to the digital twin system for emergency response plan simulation. This invention derives the data synchronization frequency based on node importance, reducing the synchronous transmission of redundant data while meeting the basic requirements of emergency simulation, thus improving the operational efficiency and stability of emergency response plan simulation.

Description

Technical Field

[0001] This invention relates to the field of industrial big data technology, specifically to an emergency response plan simulation method and system based on digital twins. Background Technology

[0002] Digital twin technology, by constructing high-fidelity virtual mirrors of physical entities such as industrial systems and integrating real-time operational data from each node, can accurately reproduce the operational status of complex industrial systems and various emergencies in a zero-risk digital space. This is of great significance for the simulation and deduction of emergency plans. Through repeated and realistic simulations of emergency procedures, operators can not only master the handling steps, but also proactively identify potential risk points. Furthermore, by quantitatively evaluating the simulation effects of different handling schemes, the feasibility and effectiveness of emergency plans can be continuously optimized.

[0003] When performing emergency response simulations on a digital twin platform, the synchronization frequency of different nodes (devices or data sources) in the simulation system is usually set uniformly. However, the importance of different nodes (devices or data sources) to emergency decision-making varies significantly. For critical core nodes, even a small delay in their state changes can lead to distortion of simulation results, thus affecting the accuracy of emergency decisions. For non-critical or auxiliary nodes, using the same high synchronization frequency as core nodes will result in a surge of redundant data, wasting communication bandwidth, computing resources, and storage space. A uniform synchronization frequency seriously affects the simulation effect of the emergency response plan. Summary of the Invention

[0004] To address the technical problem of poor simulation results in emergency response plans due to unreasonable synchronization frequency settings for different nodes, the present invention aims to provide an emergency response plan simulation method and system based on digital twins. The specific technical solution adopted is as follows: An emergency response plan simulation method based on digital twins, the method comprising: Acquire the node runtime sequence data and system load time sequence data of each node in the industrial system, and perform synchronous partitioning using a sliding time window; acquire the historical runtime sequence data of each node in the accident scenario from the historical emergency plan simulation data; Within each sliding window, based on the time-series mutual information of the node runtime timing data between different nodes under different time delays, and the time-series mutual information between the node runtime timing data of each node and the system load timing data, the node correlation coefficient between different nodes is obtained. By combining the node correlation coefficients between different nodes within all sliding time windows, the importance coefficient of each node in the industrial system is obtained; based on the fluctuation deviation between the historical runtime sequence data of different nodes in the historical emergency plan simulation data, the accident transmission index of each node is obtained. For each node, the data synchronization frequency of each node is determined according to the importance coefficient and the accident transmission index, and the runtime sequence data of the corresponding node is synchronized to the digital twin system for emergency plan simulation.

[0005] Furthermore, the method for obtaining the node runtime sequence data includes: Obtain the time series parameters of each node under each operating metric; for each node, obtain the variance contribution rate of each operating metric based on the principal component analysis algorithm, and use the variance contribution rate to perform weighted fusion of the time series parameters of the corresponding operating metric to obtain the node operating time series data of each node.

[0006] Furthermore, synchronous partitioning is performed using a sliding time window, including: Spectral analysis is performed on the system load time-series data to obtain the non-zero frequency with the largest power spectral density; the reciprocal of the non-zero frequency is used as the window length of the sliding window, and the sliding step size is set to half of the window length.

[0007] Furthermore, the method for obtaining the time delay includes: For each node, perform spectral analysis on the time series parameters of the index under each operating index to obtain the non-zero frequency with the largest power spectral density, and take the reciprocal of the non-zero frequency as the fluctuation period of the corresponding operating index. The maximum fluctuation period among all operating indicators of all nodes is taken as the time delay upper limit, and the minimum sampling interval among all operating indicators of all nodes is taken as the time delay interval; all time delays are determined based on the time delay upper limit and the time delay interval.

[0008] Furthermore, the method for obtaining the node correlation coefficient includes: Calculate the time-series mutual information of the node runtime sequence data between different nodes within the sliding window under each time delay, take the time delay corresponding to the maximum time-series mutual information between different nodes as the node association time delay, and take the negative correlation normalization result of the node association time delay as the time-related association parameter between different nodes. For each node, the load impact weight is determined based on the time-series mutual information between the node's runtime timing data and the system load timing data within the sliding time window; the load impact weights of different nodes are fused to obtain the comprehensive impact weight between the corresponding different nodes; The temporal mutual information between the runtime sequence data of different nodes within the sliding window is used as the basic association parameter; the basic association parameter, the time-related association parameter, and the comprehensive influence weight between different nodes are fused to obtain the node association coefficient between different nodes.

[0009] Furthermore, the method for obtaining the importance coefficient includes: Within each sliding time window, the global correlation coefficient of each node is obtained by combining the node correlation coefficients between each node and all other nodes; based on the temporal distribution characteristics of the global correlation coefficients of each node, the importance coefficient of each node in the industrial system is obtained.

[0010] Furthermore, the method for obtaining the accident transmission index includes: Based on the node association coefficient between each node and each of the other nodes in all sliding windows, filter out all associated nodes of each node from the remaining nodes; Based on the fluctuation deviation of the historical runtime sequence data of each node relative to the historical runtime sequence data of each associated node in the historical emergency plan simulation data, the accident transmission index of each node is obtained.

[0011] Furthermore, the method for obtaining the data synchronization frequency includes: For each node, the synchronization weight is obtained by fusing the accident transmission index and the importance coefficient; Obtain the basic synchronization frequency, lower limit of synchronization frequency, and upper limit of synchronization frequency of the digital twin system, and determine the adjustment weight of synchronization frequency based on the lower limit of synchronization frequency, the upper limit of synchronization frequency, and the synchronization weight of each node; use the adjustment weight to weight the basic synchronization frequency to obtain the data synchronization frequency of each node.

[0012] Furthermore, the method for obtaining the adjustment weights includes: The ratio of the lower limit of the synchronization frequency to the basic synchronization frequency is used as the minimum adjustment coefficient, and the ratio of the upper limit of the synchronization frequency to the basic synchronization frequency is used as the maximum adjustment coefficient. The dynamic adjustment amplitude is determined based on the minimum adjustment coefficient and the maximum adjustment coefficient. The dynamic adjustment amplitude is weighted using the synchronization weight and then added to the minimum adjustment coefficient to obtain the adjustment weight.

[0013] The emergency response plan simulation system based on digital twins includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the emergency response plan simulation method based on digital twins.

[0014] The present invention has the following beneficial effects: This invention acquires the node runtime sequence data and system load time-series data of each node in an industrial system, and uses a sliding time window for synchronous division to analyze from a local time domain perspective, avoiding obscuring key instantaneous node correlations. It further acquires the historical runtime sequence data of each node in accident scenarios from historical emergency plan simulation data, preparing for subsequent evaluation of the node's accident propagation characteristics. Then, within each sliding time window, it analyzes the operational correlation between nodes based on the time-series mutual information of the node runtime sequence data under different time delays, and analyzes the correlation influence between nodes and the system based on the time-series mutual information between each node's runtime sequence data and system load time-series data, thereby obtaining the node correlation coefficient between different nodes. Finally, it integrates the node correlation coefficients between different nodes within all sliding time windows to obtain an importance coefficient reflecting the correlation influence between each node and other nodes in the industrial system. Furthermore, it assesses the accident propagation index of each node from accident scenarios based on the fluctuation deviations between the historical runtime sequence data of different nodes in historical emergency plan simulation data. Finally, it determines the data synchronization frequency of each node based on the importance coefficient and the accident propagation index, and synchronizes the corresponding node runtime sequence data to a digital twin system for emergency plan simulation. This invention derives the data synchronization frequency based on node importance. While meeting the basic requirements of emergency simulation, it reduces the synchronous transmission of redundant data, enabling digital twins to accurately capture equipment status in both routine operation and emergency response, improving the timeliness of fault prediction and response, and thus enhancing the operational efficiency and stability of emergency plan simulation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating an emergency response plan simulation method based on digital twins, provided as an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for obtaining node correlation coefficients according to an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an emergency response plan simulation method and system based on digital twins proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of an emergency response plan simulation method and system based on digital twins provided by this invention.

[0020] Please see Figure 1 The diagram illustrates a flowchart of an emergency response plan simulation method based on digital twins, provided by an embodiment of the present invention, specifically including: Step S1: Obtain the node runtime sequence data and system load time sequence data of each node in the industrial system, and perform synchronous division using a sliding time window; obtain the historical runtime sequence data of each node in the accident scenario from the historical emergency plan simulation data.

[0021] The implementation scenarios targeted by this invention are: in remote monitoring scenarios or storage scenarios with limited storage resources, by implementing differentiated synchronization strategies for the operating data of different nodes in an industrial system, each node can maintain a reasonable status record during daily operation and maintenance. As a result, when using digital twin technology to perform emergency simulation of the industrial system, it can reduce the synchronous transmission of redundant data and avoid excessive occupation of communication bandwidth while meeting the basic requirements of emergency simulation, thereby improving the operating efficiency and stability of emergency plan simulation.

[0022] One embodiment of the present invention takes a coal-fired power plant as an example for analysis and description, regarding the coal-fired power plant as an industrial system, and each piece of equipment in the coal-fired power plant (such as boiler, steam turbine, generator, feedwater pump, induced draft fan, etc.) as a node in the industrial system; in another embodiment, a factory production line, etc., can also be regarded as an industrial system for emergency plan simulation.

[0023] In one embodiment of the present invention, a connection is first established with the distributed control system (DCS) or plant-level monitoring information system (SIS) of the power plant through an industry standard protocol (such as OPC UA, Modbus TCP); then, the operating parameters of each device in the coal-fired power plant and the output power of the power plant units are extracted within a preset historical period, such as a historical week; the acquisition frequency of all devices is set to 1Hz, thereby obtaining a series of time-series data; the implementer may also adjust the acquisition frequency according to the actual application situation, but the time-series data of different devices need to be resampled to ensure time alignment.

[0024] Further construct the node runtime sequence data and system load time sequence data of each node (equipment) in the industrial system (coal-fired power plant); among them, the output power of the power plant units in the coal-fired power plant within a preset historical period is standardized by Z-score to obtain a series of system load time sequence data.

[0025] Preferably, in one embodiment of the present invention, considering that each node (such as a boiler) may have multiple operating indicators (such as main steam pressure, main steam temperature, steam drum water level, furnace negative pressure, etc.), in order to comprehensively evaluate the operating status of each node, it is necessary to comprehensively evaluate all operating indicators to obtain node runtime sequence data characterizing the node's operating status; and since the fluctuations of each operating indicator vary greatly, principal component analysis can identify the nodes carrying the main operating variation information, thereby enabling the comprehensively evaluated node runtime sequence data to retain the true fluctuation information of the node's operating status to the greatest extent; therefore, the method for obtaining node runtime sequence data includes: Obtain the time series parameters of each node under each operating metric; for each node, obtain the variance contribution rate of each operating metric based on the principal component analysis algorithm, and use the variance contribution rate to perform weighted fusion of the time series parameters of the corresponding operating metric to obtain the node operating time series data of each node.

[0026] As an example, we first obtain the time-series parameters of each node under each type of operational metric. The time-series parameters refer to the parameters of each operational metric at each collection time within a preset historical period. Taking any node as an example, for each operational metric, after cleaning the series of time-series parameters, we further use the Z-score standardization method to standardize them to eliminate the dimensions, thereby obtaining a series of standardized metric parameters. We then sort the series of standardized index parameters in chronological order to construct a feature column vector. Then, the feature column vectors under all indicators are randomly sorted to construct a feature matrix. The Principal Components Analysis (PCA) algorithm is then used to calculate the variance contribution rate of each feature (each type of operating indicator). The variance contribution rate is used as the feature weight. The standardized indicator parameters of different operating indicators at each collection time are weighted. The weighted sum of all operating indicators is used as the node operating data at the corresponding collection time, thus obtaining a series of node operating sequence data at different collection times.

[0027] It should be noted that data cleaning, Z-score standardization, and the PCA algorithm are all well-known technologies familiar to those skilled in the art, and will not be elaborated upon further.

[0028] In highly coupled and complex industrial systems, there may be certain fault propagation characteristics between nodes. For example, an abnormal failure of one node may trigger a series of parameter anomalies, which in turn trigger a chain reaction in related nodes, eventually leading to a partial shutdown or a system-wide accident. The purpose of emergency plan simulation is to simulate the dynamically evolving crisis process, thereby blocking the propagation of faults along the fault propagation path with minimal emergency costs (such as emergency switching or activating backup). Therefore, analyzing the correlation between different nodes in an industrial system can help analyze the fault propagation path and assess the importance of nodes, providing a basis for subsequently adjusting the synchronization frequency of differentiated operational data to improve the simulation effect of emergency plans.

[0029] Furthermore, since the relationships between different nodes in an industrial system typically change with operating conditions (such as start-up, shutdown, and load variation), using global time-series data within a preset historical period to analyze these relationships may obscure crucial instantaneous strong relationships (such as the moment of fault propagation), thus failing to reflect the true dynamic relationships. Therefore, this embodiment of the invention further utilizes a sliding time window to synchronously divide the node runtime time-series data and system load time-series data of each node, providing a data foundation for subsequent analysis of node relationship information.

[0030] Preferably, in one embodiment of the present invention, considering that the fluctuations in the operating data of equipment in a power plant are usually related to the inherent oscillation mode or control cycle of the system, using this cycle as the window length can ensure that each time window contains at least one complete dominant oscillation cycle, thereby capturing the complete dynamic mode within that cycle; furthermore, considering that the system load time series data can comprehensively reflect the operating status of the entire power plant, its fluctuation cycle often represents the inherent cycle of the overall operation of the power plant (such as the load adjustment cycle, grid dispatch cycle, etc.); and the peak frequency of the power spectrum corresponds to the strongest and most stable oscillation component in the signal, which usually reflects the inherent physical cycle of the system, thus determining the window length; also considering that setting the sliding step size of the window to half the window length can avoid the loss of boundary information, that is, prevent key transient correlations from occurring at the junction of two windows and being ignored; therefore, synchronous division using a sliding time window includes: Perform spectral analysis on the system load time-series data to obtain the non-zero frequency with the highest power spectral density; use the reciprocal of the non-zero frequency as the window length of the sliding window, and set the sliding step size to half of the window length.

[0031] As an example, the system load time-series data is sorted in chronological order to construct the system load time-series curve. Fast Fourier Transform (FFT) spectrum analysis is performed on the system load time-series curve to identify the non-zero frequency with the largest power spectral density in the frequency domain spectrum. The reciprocal of this non-zero frequency is used as the window length of the sliding window, and the sliding step size is set to half of the window length. That is, starting from the beginning of the preset historical period, the first sliding window is determined, and then the window is slid along the time direction by half a length to obtain the second sliding window, and so on, to obtain all the sliding windows.

[0032] It should be noted that during the sliding partitioning process, there may be sliding time windows that are shorter than the window length, which should be regarded as a single sliding time window. Implementers can also adjust the sliding step size themselves, but it should not exceed the window length to ensure that there is overlap between time windows. Fast Fourier Transform and determination of the non-zero frequency with the largest power spectral density are well-known techniques and will not be elaborated further.

[0033] It should be noted that when no significant non-zero frequencies are identified, the window length should be set to a preset value, such as 1 hour.

[0034] Because the operating status of some key nodes (such as safety valves) may remain almost unchanged in daily operation, it is difficult to assess their decisive role in an accident. This may lead to misjudgment of node importance in subsequent node correlation analysis, that is, stable nodes may be misjudged as unimportant nodes, resulting in low data synchronization frequency and affecting the simulation effect of emergency plan. However, in accident scenarios (simulation data), the changes in the operating status of nodes can reflect their decisive role in the accident, and thus their node importance can be comprehensively assessed from normal operation scenarios and accident scenarios.

[0035] Based on this, the embodiments of the present invention further obtain the historical runtime sequence data of each node in the accident scenario in the historical emergency plan simulation data; specifically, extracting the historical node runtime sequence data of each node in the accident scenario from the digital twin platform or the preset fault case library is a well-known technical means, and will not be described in detail here.

[0036] Step S2: Within each sliding time window, based on the time-series mutual information of the node runtime sequence data between different nodes under different time delays, and the time-series mutual information between the node runtime sequence data of each node and the system load time-series data, obtain the node correlation coefficient between different nodes.

[0037] Given that time-series mutual information, based on information theory principles, can help quantify the nonlinear correlations in fluctuations of power plant equipment parameters (node ​​timing parameters), and that fluctuation (or fault) propagation exhibits time propagation characteristics, the shorter the propagation time between different devices (nodes), the stronger the correlation, and the greater the reference value for fault prediction; however, time-series mutual information can only help assess the correlation between fluctuations between devices, but cannot help determine the propagation path of fluctuations, thus failing to accurately assess the importance of devices; furthermore, devices closely related to the power of power plant units directly affect system operation, and strengthening such correlations can accurately pinpoint key devices; Based on this, in each sliding time window, the present invention will obtain the node correlation coefficient between different nodes according to the time-series mutual information of the node runtime sequence data between different nodes under different time delays, and the time-series mutual information between the node runtime sequence data of each node and the system load time-series data. The node correlation coefficient comprehensively evaluates the local dynamic correlation between different nodes from the perspective of fluctuation transmission and system correlation, and prepares for subsequent analysis and evaluation of the importance of nodes in the industrial system.

[0038] Preferably, in one embodiment of the present invention, considering that time-series mutual information is sensitive to time alignment, in order to accurately capture the fluctuation overlap of different nodes, the smallest sampling interval among all operating indicators of all nodes can be used as the minimum time delay interval, i.e., the minimum misalignment, to analyze the fluctuation overlap under different alignment conditions; the maximum fluctuation period among all operating indicators of all nodes is used as the upper limit of the time delay, thereby avoiding repeated fluctuation overlap analysis while achieving comprehensive coverage analysis; and the peak frequency of the power spectrum can help evaluate the fluctuation period of the node operating data; therefore, the method for obtaining the time delay includes: For each node, perform spectral analysis on the time series parameters of the index under each operating index to obtain the non-zero frequency with the largest power spectral density, and take the reciprocal of the non-zero frequency as the fluctuation period of the corresponding operating index. The maximum fluctuation period among all operating indicators of all nodes is taken as the time delay upper limit, and the minimum sampling interval among all operating indicators of all nodes is taken as the time delay interval; all time delays are determined based on the time delay upper limit and the time delay interval.

[0039] As an example, taking any node as an example, firstly, the time series parameters of each operating indicator are sorted in time sequence to construct the indicator time series curve; then, fast Fourier transform spectrum analysis is performed on the time series curve of each indicator to identify the non-zero frequency with the largest power spectral density in the frequency domain spectrum, and the reciprocal of the non-zero frequency is taken as the fluctuation period of the operating indicator.

[0040] It should be noted that when no significant non-zero frequency can be identified, the fluctuation period should be set to a preset value such as 1 hour.

[0041] Since the sampling frequency of all operating indicators of all nodes is set to 1Hz, the time delay interval is set to 1 in this example. Then, the maximum value N is selected from the fluctuation period of all operating indicators of all nodes as the fluctuation period, thereby determining a series of time delays such as 1, 2, 3, ..., N.

[0042] Once the time lag is determined, the node correlation coefficients between different nodes within each sliding window can be further analyzed and obtained.

[0043] Preferably, in one embodiment of the present invention, please refer to Figure 2 The flowchart illustrates a method for obtaining node correlation coefficients according to an embodiment of the present invention, specifically including: It should be noted that the method for analyzing and obtaining the node correlation coefficients between different nodes within each sliding window is consistent. Here, we will take any sliding window as an example for analysis and description, and will not go into detail again.

[0044] Step S201: Calculate the time-series mutual information of node runtime sequence data between different nodes within the sliding time window under each time delay, take the time delay corresponding to the maximum time-series mutual information between different nodes as the node association time delay, and take the negative correlation normalization result of the node association time delay as the time-series association parameter between different nodes.

[0045] Considering that the maximum temporal mutual information between different nodes can characterize the strongest correlation, the smaller the time delay corresponding to the strongest correlation, the shorter the fluctuation propagation time between nodes and the stronger the correlation; based on this, the time correlation parameters between different nodes can be obtained from the perspective of time delay.

[0046] As an example, taking any two different nodes (i and j) in an industrial system as an example, firstly, determine the node runtime sequence data corresponding to node i and node j within the sliding time window, and then construct local time series such as node i: {A1, A2, ..., Am}, node j: {B1, B2, ..., Bm}, where m refers to the window length of the sliding time window; Then, the local time series sequence of node i is extended along the time series direction by the corresponding time delay data points. For example, when the time delay is 2, the time delay result of the local time series sequence of node i is {empty, empty, A1, A2, ..., Am}. At the same time, in order to ensure that the lengths of the local time series sequences of node i and node j are consistent for subsequent analysis of time series mutual information, they are synchronously truncated and aligned. Then, node i: {A1, A2, ..., Am-2}, that is, the last two bits are truncated, and node j: {B3, B4, ..., Bm}, that is, the first two bits are truncated. This makes A1 and B3 correspond to two time delay units. Then, the time series mutual information between the two truncated and aligned sequences is analyzed. By changing the time lag, the time-series mutual information of nodes i and j under different time lags is obtained. Then, the maximum time-series mutual information is selected and the time lag corresponding to the maximum time-series mutual information is used as the node association time lag. The node association time lag is used to characterize the time for the wave propagation from node i to node j. The node association time lag is further negatively correlated and normalized to obtain the time-series correlation parameter Tij between nodes i and j.

[0047] It should be noted that the calculation of time-series mutual information is a well-known technique and will not be elaborated further. The negative correlation normalization method used is as follows: first, the node association delay is divided by a fixed value, such as the upper limit of the delay, for normalization; then, negative correlation normalization is performed by subtracting the normalization value from 1. Implementers may also use other negative correlation normalization methods, such as reciprocal calculation.

[0048] It should be noted that the time-related parameter Tij between node i and node j represents a directed correlation, that is, the influence of the fluctuation of node i on the fluctuation of node j. In this embodiment, the time-related parameter between node i and node j also includes Tji (which has a different meaning from Tij) which represents the influence of the fluctuation of node j on the fluctuation of node i. Specifically, the time-related parameter Tji is analyzed by extending the local time series sequence of node j along the time series direction by the corresponding time delay data points, and then truncating and aligning it with the local time series sequence of node i. (This will not be elaborated further.)

[0049] By changing different nodes, the time-related parameters between all different nodes can be obtained.

[0050] Step S202: For each node, determine the load impact weight based on the time-series mutual information between the corresponding node's runtime timing data and the system load timing data within the sliding time window; merge the load impact weights of different nodes to obtain the comprehensive impact weight between the corresponding different nodes.

[0051] Considering that the greater the time-series mutual information between the node's runtime timing data and the system load timing data, the more directly the node's operating status is related to the operating status of the industrial system, and the higher its influence or importance in the industrial system; therefore, this embodiment of the invention further determines the load influence weight of each node from the perspective of system correlation; and further comprehensively evaluates the comprehensive influence weight between two different nodes, in order to prepare for the subsequent evaluation of the node correlation between different nodes.

[0052] As an example, taking node i as an example, firstly, determine the node runtime timing data corresponding to node i within the sliding time window, as well as the system load timing data within the sliding time window; then further analyze the timing mutual information, normalize the timing mutual information, and thus obtain the load impact weight of node i; change the node and obtain the load impact weight of each node. The normalization method used is to divide the time-series mutual information by a fixed value (such as the maximum value of the time-series mutual information between the node's runtime time-series data and the system load time-series data); implementers may also use other normalization methods. Taking node i and node j as examples, the load impact weight of node i and the load impact weight of node j are averaged to obtain the comprehensive impact weight between node i and node j; implementers can also use other methods such as weighted summation to integrate the weights.

[0053] Step S203: Use the temporal mutual information between the runtime sequence data of different nodes within the sliding time window as the basic association parameter; integrate the basic association parameter, the time-related association parameter, and the comprehensive influence weight between different nodes to obtain the node association coefficient between different nodes.

[0054] Considering the temporal mutual information between the runtime sequence data of different nodes within the sliding window, an objective statistical basis for node association can be provided. Based on this, the time-dependent association parameters obtained from the time delay perspective analysis and the comprehensive influence weight obtained from the system association perspective analysis are introduced to comprehensively evaluate the node association coefficient Wij between different nodes.

[0055] As an example, taking node i and node j as an example, we first obtain the basic association parameters between node i and node j, and then multiply and fuse the basic association parameters, the time-related association parameters, and the comprehensive influence weight, and normalize them to obtain the node association coefficient between node i and node j.

[0056] It should be noted that the normalization method used is to divide the product by a fixed value (such as the maximum product value corresponding to different nodes); implementers may also use other normalization methods.

[0057] It should be noted that the node correlation coefficient Wij is also a directed correlation, referring to the influence of the fluctuation of node i on the fluctuation of node j.

[0058] Step S3: Combine the node correlation coefficients between different nodes within all sliding time windows to obtain the importance coefficient of each node in the industrial system; based on the fluctuation deviation between the historical runtime sequence data of different nodes in the historical emergency plan simulation data, obtain the accident transmission index of each node.

[0059] After obtaining the node correlation coefficients between different nodes within each sliding window, the importance coefficient of each node in the industrial system can be further obtained. The importance coefficients analyze the decisive role of each node in the wave (or fault) propagation path from the perspective of node wave correlation.

[0060] Preferably, in one embodiment of the present invention, considering that the node correlation coefficient between each node and the other nodes can be used to assess the comprehensive impact of the fluctuation of each node on the fluctuation of all other nodes, the global correlation coefficient of each node is obtained. Then, based on the temporal distribution characteristics of the global correlation coefficient of each node, relatively stable high-correlation relationships are comprehensively screened to assess the importance of each node in the industrial system; therefore, the method for obtaining the importance coefficient includes: Within each sliding time window, the global correlation coefficient of each node is obtained by combining the node correlation coefficients between each node and all other nodes; based on the temporal distribution characteristics of the global correlation coefficient of each node, the importance coefficient of each node in the industrial system is obtained.

[0061] As an example, taking node i as an example, within each sliding window, the sum of the node correlation coefficients Wij, Wik, Wil, etc. between node i and all other nodes (such as j, k, l, etc.) is used as the global correlation coefficient of each node; further, the standard deviation is used to measure the distribution characteristics, and the negative correlation normalization result of the standard deviation of the global correlation coefficient of node i in all sliding windows is used as the importance coefficient of node i in the industrial system.

[0062] The negative correlation normalization method used is as follows: the standard deviation corresponding to node i is divided by the maximum value among the standard deviations of all nodes for normalization. The larger the standard deviation, the higher the volatility of the global correlation coefficient in time series, the more unstable its correlation with other nodes, and the lower the confidence of its importance. Therefore, negative correlation normalization is further performed by subtracting the normalization value from the standard deviation. Implementers may also use other negative correlation normalization methods, such as mapping to an exponential function with the natural constant e as the base, which will not be elaborated here.

[0063] It should be noted that the calculation of standard deviation is a well-known technique and will not be elaborated further; when the maximum value of the standard deviation is 0, the importance coefficient is directly set to the preset value of 1.

[0064] Considering that the operating status of some nodes may remain almost unchanged in daily operation and their correlation with other nodes in industrial systems is relatively weak, their importance score is relatively low; however, in accident scenarios, they may quickly propagate and trigger a chain reaction, even affecting the operation of the entire system. Therefore, in order to accurately assess the overall importance or influence of each node, this embodiment of the invention further obtains the accident transmission index of each node based on the fluctuation deviation between the historical runtime sequence data of different nodes in the historical emergency plan simulation data. The accident transmission index provides an importance reference for each node from the accident scenario, and prepares for the subsequent comprehensive adjustment of the node operation data synchronization frequency in combination with the importance coefficients obtained from the assessment under the daily operation scenario.

[0065] Preferably, in one embodiment of the present invention, in order to analyze the fault propagation of each node in an accident scenario, firstly, associated nodes with strong correlation to each node are screened in the industrial system. Further analysis is then performed on the fluctuation deviation between each node and associated nodes in historical emergency response simulation data. The more significant the fluctuation of a node causes fluctuations in its associated nodes, the greater the chain reaction of the accident caused by that node, and the greater its accident propagation index. Therefore, the method for obtaining the accident propagation index includes: Based on the node association coefficient between each node and every other node in all sliding windows, filter out all associated nodes of each node from the remaining nodes; Based on the fluctuation deviation of the historical runtime sequence data of each node relative to the historical runtime sequence data of each associated node in the historical emergency plan simulation data, the accident transmission index of each node is obtained.

[0066] As an example, taking any node i as an example, within each sliding window, the remaining nodes whose node association coefficient with node i is greater than a preset threshold (such as the average node association coefficient) are taken as the initial associated nodes of node i; then the union of all initial associated nodes in all sliding windows is taken as the associated node set; when it is an empty set, the accident propagation index can be directly set to 0; in other examples, the implementer can also select downstream nodes that have physical connections with node i in the industrial system according to the process flow, and take them as the associated nodes of node i. Furthermore, in the historical emergency response simulation data, for node i, the absolute value of the difference between each historical runtime sequence data (after Z-score standardization) and the mean is taken as the fluctuation amount. The average of all fluctuation amounts is used to obtain the node fluctuation parameter of node i. Similarly, the node fluctuation parameters of all its associated nodes are calculated. The mean of the node fluctuation parameters of all associated nodes is added to a minimum positive parameter such as 0.01 as the numerator, and the node fluctuation parameter of node i is added to a minimum positive parameter such as 0.01 as the denominator. The maximum and minimum normalized results of the ratio of the fractions are used as the accident transmission index of node i.

[0067] Step S4: For each node, determine the data synchronization frequency of each node based on the importance coefficient and the accident transmission index, and synchronize the runtime sequence data of the corresponding node to the digital twin system for emergency plan simulation.

[0068] After obtaining the accident transmission index of each node, its accident impact in the industrial system can be comprehensively evaluated by combining the importance coefficient. Then, the data synchronization frequency for synchronizing the node operation data of each node to the digital twin platform for emergency plan simulation can be analyzed and determined. An appropriate data synchronization frequency can reduce the synchronous transmission of redundant data while meeting the basic requirements of emergency simulation, thereby improving the operational efficiency and stability of emergency plan simulation.

[0069] Preferably, in one embodiment of the present invention, since both the accident transmission index and the importance coefficient can characterize the accident influence of a node in an industrial system, the two are first fused to determine the synchronization weight. The larger the synchronization weight, the higher the synchronization frequency should be. Further, the adjustment of the synchronization frequency is determined based on the data synchronization frequency limit of the digital twin system, thereby adaptively determining the data synchronization frequency of each node. The method for obtaining the data synchronization frequency includes: For each node, the synchronization weight is obtained by fusing the accident transmission index and the importance coefficient; the basic synchronization frequency, the lower limit of the synchronization frequency, and the upper limit of the synchronization frequency of the digital twin system are obtained, and the adjustment weight of the synchronization frequency is determined based on the lower limit of the synchronization frequency, the upper limit of the synchronization frequency, and the synchronization weight of each node; the basic synchronization frequency is weighted using the adjustment weight to obtain the data synchronization frequency of each node.

[0070] As an example, taking any node as an example, firstly, the accident transmission index and importance coefficient are weighted and summed to obtain the synchronization weight; where, the standard deviation of the accident transmission index of all nodes is... Normalization is performed to obtain the weighted weights of the accident transmission index; the standard deviation of the accident transmission index for all nodes is then calculated. Normalization is performed to obtain the weighted weights of the importance coefficients; the normalization method is to divide by... This makes the sum of the two weights equal to 1, where when When the standard deviation is 0, both weights are set to 0.5. The larger the standard deviation, the higher the distinguishability of the accident transmission index or importance coefficient between nodes, and the higher its corresponding weight. Further review the configuration parameters of the digital twin system to determine the basic synchronization frequency, lower limit of the synchronization frequency, and upper limit of the synchronization frequency. In this example, the basic synchronization frequency is 1Hz, the lower limit of the synchronization frequency is 0.1Hz, and the upper limit of the synchronization frequency is 50Hz. Implementers can also adjust these parameters according to their actual circumstances.

[0071] In a preferred embodiment of the present invention, the method for obtaining the adjustment weight includes: The ratio of the lower limit of the synchronization frequency to the basic synchronization frequency is used as the minimum adjustment coefficient, and the ratio of the upper limit of the synchronization frequency to the basic synchronization frequency is used as the maximum adjustment coefficient. The dynamic adjustment amplitude is determined based on the minimum and maximum adjustment coefficients. The dynamic adjustment amplitude is weighted using the synchronization weight and then the minimum adjustment coefficient is added to obtain the adjustment weight.

[0072] Specifically, the lower limit of the synchronization frequency is used as the numerator, the basic synchronization frequency as the denominator, and the ratio as the minimum adjustment coefficient to ensure the basic monitoring requirements of the nodes; the upper limit of the synchronization frequency is used as the numerator, the basic synchronization frequency as the denominator, and the ratio as the maximum adjustment coefficient to avoid exceeding the hardware capabilities of the digital twin system; then, the maximum adjustment coefficient is subtracted from the minimum adjustment coefficient to obtain the dynamic adjustment amplitude representing the adjustment space; then, the synchronization weight is multiplied by the dynamic adjustment amplitude, and the product is added to the minimum adjustment coefficient to obtain the adjustment weight.

[0073] Finally, the adjustment weights are applied to the basic synchronization frequency to obtain the data synchronization frequency for each node.

[0074] In another embodiment of the present invention, the basic synchronization frequency can be weighted by adding a constant 1 to the synchronization weight of each node to obtain the data synchronization frequency of the node; however, the calculated data synchronization frequency needs to be limited between the lower limit and the upper limit of the synchronization frequency.

[0075] At this point, the data synchronization frequency of each node has been obtained. When an emergency scenario is triggered (such as when maintenance personnel activate a specific emergency plan or the system detects an actual anomaly), the digital twin platform will collect the node runtime sequence data (or the time sequence parameters of each runtime indicator) of the corresponding node according to the data synchronization frequency to simulate the emergency plan. The simulation process of the emergency plan is a well-known technology that is familiar to those skilled in the art and is not the focus of this embodiment of the invention. It will not be described in detail here.

[0076] Based on the same inventive concept, this invention also proposes an emergency response plan simulation system based on digital twins, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the emergency response plan simulation method based on digital twins described in S1-S4 above.

[0077] In summary, this invention acquires the node runtime sequence data and system load time-series data of each node in an industrial system and uses a sliding time window for synchronization. It acquires the historical runtime sequence data of each node in the accident scenario from historical emergency plan simulation data. Within each sliding time window, based on the time-series mutual information of the node runtime sequence data between different nodes under different time delays, and the time-series mutual information between the node runtime sequence data of each node and the system load time-series data, it obtains the node correlation coefficient between different nodes. Then, it comprehensively obtains the importance coefficient of each node in the industrial system. Next, based on the fluctuation deviation between the historical runtime sequence data of different nodes in the historical emergency plan simulation data, it obtains the accident transmission index of each node. Furthermore, it determines the data synchronization frequency of each node based on the importance coefficient and the accident transmission index, and synchronizes the corresponding node runtime sequence data to the digital twin system for emergency plan simulation. This invention derives the data synchronization frequency based on node importance, reducing the synchronous transmission of redundant data while meeting the basic requirements of emergency simulation. This enables the digital twin to accurately capture equipment status in both routine maintenance and emergency response, improving the timeliness of fault prediction and response, thereby enhancing the operational efficiency and stability of emergency plan simulation.

[0078] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0079] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An emergency response plan simulation method based on digital twins, characterized in that, The method includes: Acquire the node runtime sequence data and system load time sequence data of each node in the industrial system, and perform synchronous partitioning using a sliding time window; acquire the historical runtime sequence data of each node in the accident scenario from the historical emergency plan simulation data; Within each sliding window, based on the time-series mutual information of the node runtime timing data between different nodes under different time delays, and the time-series mutual information between the node runtime timing data of each node and the system load timing data, the node correlation coefficient between different nodes is obtained. By combining the node correlation coefficients between different nodes within all sliding time windows, the importance coefficient of each node in the industrial system is obtained; based on the fluctuation deviation between the historical runtime sequence data of different nodes in the historical emergency plan simulation data, the accident transmission index of each node is obtained. For each node, the data synchronization frequency of each node is determined according to the importance coefficient and the accident transmission index, and the runtime sequence data of the corresponding node is synchronized to the digital twin system for emergency plan simulation.

2. The emergency response plan simulation method based on digital twins according to claim 1, characterized in that, The method for obtaining the node runtime sequence data includes: Obtain the time series parameters of each node under each operating metric; for each node, obtain the variance contribution rate of each operating metric based on the principal component analysis algorithm, and use the variance contribution rate to perform weighted fusion of the time series parameters of the corresponding operating metric to obtain the node operating time series data of each node.

3. The emergency response plan simulation method based on digital twins according to claim 1, characterized in that, Synchronization is achieved using a sliding time window, including: Spectral analysis is performed on the system load time-series data to obtain the non-zero frequency with the largest power spectral density; the reciprocal of the non-zero frequency is used as the window length of the sliding window, and the sliding step size is set to half of the window length.

4. The emergency response plan simulation method based on digital twins according to claim 2, characterized in that, The method for obtaining the time delay includes: For each node, perform spectral analysis on the time series parameters of the index under each operating index to obtain the non-zero frequency with the largest power spectral density, and take the reciprocal of the non-zero frequency as the fluctuation period of the corresponding operating index. The maximum fluctuation period among all operating indicators of all nodes is taken as the time delay upper limit, and the minimum sampling interval among all operating indicators of all nodes is taken as the time delay interval; all time delays are determined based on the time delay upper limit and the time delay interval.

5. The emergency response plan simulation method based on digital twins according to claim 1, characterized in that, The method for obtaining the node correlation coefficient includes: Calculate the time-series mutual information of the node runtime sequence data between different nodes within the sliding window under each time delay, take the time delay corresponding to the maximum time-series mutual information between different nodes as the node association time delay, and take the negative correlation normalization result of the node association time delay as the time-related association parameter between different nodes. For each node, the load impact weight is determined based on the time-series mutual information between the node's runtime timing data and the system load timing data within the sliding time window; the load impact weights of different nodes are fused to obtain the comprehensive impact weight between the corresponding different nodes; The temporal mutual information between the runtime sequence data of different nodes within the sliding window is used as the basic association parameter; the basic association parameter, the time-related association parameter, and the comprehensive influence weight between different nodes are fused to obtain the node association coefficient between different nodes.

6. The emergency response plan simulation method based on digital twins according to claim 1, characterized in that, The method for obtaining the importance coefficient includes: Within each sliding time window, the global correlation coefficient of each node is obtained by combining the node correlation coefficients between each node and all other nodes; based on the temporal distribution characteristics of the global correlation coefficients of each node, the importance coefficient of each node in the industrial system is obtained.

7. The emergency response plan simulation method based on digital twins according to claim 1, characterized in that, The method for obtaining the accident transmission index includes: Based on the node association coefficient between each node and each of the other nodes in all sliding windows, filter out all associated nodes of each node from the remaining nodes; Based on the fluctuation deviation of the historical runtime sequence data of each node relative to the historical runtime sequence data of each associated node in the historical emergency plan simulation data, the accident transmission index of each node is obtained.

8. The emergency response plan simulation method based on digital twins according to claim 1, characterized in that, The method for obtaining the data synchronization frequency includes: For each node, the synchronization weight is obtained by fusing the accident transmission index and the importance coefficient; Obtain the basic synchronization frequency, lower limit of synchronization frequency, and upper limit of synchronization frequency of the digital twin system, and determine the adjustment weight of synchronization frequency based on the lower limit of synchronization frequency, the upper limit of synchronization frequency, and the synchronization weight of each node; use the adjustment weight to weight the basic synchronization frequency to obtain the data synchronization frequency of each node.

9. The emergency response plan simulation method based on digital twins according to claim 8, characterized in that, The method for obtaining the adjustment weights includes: The ratio of the lower limit of the synchronization frequency to the basic synchronization frequency is used as the minimum adjustment coefficient, and the ratio of the upper limit of the synchronization frequency to the basic synchronization frequency is used as the maximum adjustment coefficient. The dynamic adjustment amplitude is determined based on the minimum adjustment coefficient and the maximum adjustment coefficient. The dynamic adjustment amplitude is weighted using the synchronization weight and then added to the minimum adjustment coefficient to obtain the adjustment weight.

10. An emergency response plan simulation system based on digital twins, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the emergency response plan simulation method based on any one of claims 1-9.

Images