Multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method and system

Abstract

The present invention discloses a method, system, computer device and storage medium for scenario behavior deduction and reliability prediction analysis of a multi-source data fusion dynamic system. The method, based on the Markov / CCMT dynamic reliability prediction analysis method, combines multi-source data fusion and assimilation methods, and uses Monte Carlo probability model random sampling to simulate and statistically analyze the complex dynamic behavior characteristics of digital process control with strong interactive coupling, nonlinearity and high uncertainty. Then, through dynamic search and analysis of the system state transition probability matrix model, forward deduction analysis and reliability prediction of the system operation state are achieved. The present invention can realize the adaptive update construction of the state transition probability matrix of large-scale complex digital process control systems and the dynamic deduction and analysis of scenario behavior, avoid the problem of high-dimensional system state space search explosion, and can accurately simulate and map the system dynamic behavior characteristics to realize system dynamic reliability prediction analysis.

Description

Technical Field

[0001] The present invention belongs to the field of complex digital process control system simulation and dynamic reliability safety analysis, and particularly relates to a multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method, system, computer equipment and storage medium. Background Art

[0002] In recent years, all industries have been undergoing digital transformation. Even the nuclear industry, which is less sensitive to emerging technologies, is undergoing digital technology upgrades and comprehensive application. With the rapid development of digital intelligent sensing technology, distributed communication network technology, and computer technology, nuclear power has ushered in the era of big data. Digital instrumentation and control systems, as the nerve center of nuclear power plants, are crucial to their operational safety. However, while these advanced digital system upgrades offer significant advantages, they also present challenges such as the highly interactive dynamic characteristics of the systems, the embedded software control algorithms, and the strong coupling between system components. These challenges complicate system verification and validation, as well as dynamic reliability and safety analysis. Nuclear power regulatory agencies, such as the U.S. Nuclear Regulatory Commission (NRC), have pioneered research on reliability prediction and analysis methods for digital I&C systems and their software in nuclear power plants. Comparative analysis with benchmark systems indicates that traditional static fault tree / event tree analysis methods are ineffective for dynamic reliability modeling and risk analysis of digital I&C systems. New dynamic reliability prediction and analysis methods or integrated approaches are needed to address the complex dynamic interactions, multi-state, nonlinear, time-series, and high uncertainty characteristics of digital I&C systems, enabling reliability modeling and comprehensive safety analysis.

[0003] At present, research on new dynamic reliability and risk assessment methods mainly focuses on methods such as dynamic fault tree / event tree, Markov / CCMT, dynamic flow graph method (DFM), Bayesian belief network, Petri net, Monte Carlo simulation, etc. However, due to the high complexity of digital process control systems, the research on related technologies has progressed slowly. Many methods generally have computational complexity bottlenecks, and the theoretical research on related algorithms is still at the stage of simple case system demonstration and explanation, making it difficult to expand its application to the dynamic reliability and safety evaluation of large and complex systems. Summary of the Invention

[0004] In response to practical problems such as the difficulty in accurately modeling and efficiently calculating the dynamic interaction processes of large-scale, complex nonlinear digital control systems, the present invention provides a method, device, computer equipment and storage medium for dynamic system scenario behavior deduction and reliability prediction analysis using multi-source data fusion. By self-updating the system state transition probability mapping matrix model after data assimilation and merging and reducing the dimensionality of system state points, the present invention can effectively eliminate the uncertainty in the system dynamic behavior simulation process, improve modeling accuracy and matrix space search efficiency, and achieve advanced prediction of system scenario behavior and dynamic reliability prediction, which can be used to guide the safe operation of nuclear power plants.

[0005] The first object of the present invention is to provide a method for scenario behavior deduction and reliability prediction analysis of a dynamic system using multi-source data fusion.

[0006] The second object of the present invention is to provide a multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis system.

[0007] A third object of the present invention is to provide a computer device.

[0008] A fourth object of the present invention is to provide a storage medium.

[0009] The first object of the present invention can be achieved by adopting the following technical solutions:

[0010] A method for scenario behavior deduction and reliability prediction analysis of a dynamic system using multi-source data fusion, the method comprising:

[0011] According to the initial value input or assumption of the system state, the particle swarm distribution of the system assimilation initial state is obtained through Gaussian sampling, which is imported into the system simulation model to simulate the trajectory change of the particle swarm. Combined with the real-time monitoring input of the system state, the posterior distribution of the particle swarm of the current system state is obtained through data assimilation, and used as the particle swarm distribution of the system assimilation initial state at the next moment;

[0012] According to the posterior distribution range of the particle swarm of the current system state obtained through the data assimilation process, the upper and lower boundary values ​​of the system state originating grid element are determined, and this is used as the new grid element scale to discretize the system state space. Through digital coding, the reconstruction of the system state space model is completed.

[0013] Uniform sampling is performed on the system state originating grid element to obtain the distribution of the initial state particle swarm of the system dynamic behavior deduction. The initial state particle swarm of the system dynamic behavior deduction is imported into the system simulation model, and the trajectory change of the obtained particle swarm is simulated. The conditional transition probability matrix of the system state under the current system configuration is estimated based on the number of particle swarms falling into different grid elements; the conditional transition probability matrix is ​​combined with the state transition probability matrix of the system physical components to obtain the Markov / CCMT system state transition probability mapping matrix model;

[0014] Based on the constructed Markov / CCMT system state transition probability mapping matrix model, the system state scenario behavior dynamic deduction algorithm is applied to perform system state advance prediction analysis. The same system states are further merged, and the merged and reduced unique system states are pressed into the search sequence structure. The Markov / CCMT system state transition probability mapping matrix model is repeated to construct an update and iterative search process until the specified search depth is reached.

[0015] According to the probability of the system state after the merger, the dynamic evolution of the system state scenario sequence is sorted and predicted, and the safe operation of the system is guided through the graphical display of the interface.

[0016] Further, the system configuration structure is configured according to the real-time status monitoring input of the system equipment, and the search sequence structure is constructed according to the initialization and expansion system state after the configuration;

[0017] The extended system state includes the system state and its accompanying parameters. The accompanying parameter state is the associated system parameter of the system state change. The system state and its accompanying parameters and the state of the system's physical components are coupled with each other, but do not directly affect the dynamic behavior characteristics of the system. They react indirectly through the impact on the system state change.

[0018] Furthermore, the system state scenario behavior dynamic deduction algorithm includes:

[0019] The assimilated current system state and its accompanying parameters are used as the starting point of the system state and added to the search sequence structure. At the same time, the system initial state probability, search depth, and time step parameters are initialized.

[0020] Each system state in the search sequence structure is extracted one by one, and the next transition point of the system state is obtained by searching for non-zero elements in the Markov / CCMT system state transition probability mapping matrix model after assimilation and update. The probability of occurrence of each system state transition sequence path is obtained through probability calculation during the search process; according to needs, a prior truncation criterion is set in each step of the system state transition process;

[0021] After each traversal of the system state in the search sequence structure is completed, the system state transition sequence path sets with the same final state are merged, and the merged system state is added to the search sequence structure as a new parent node to enter the next iterative search;

[0022] For each new parent node, before entering the next system state transition sequence branch search, it is necessary to reconstruct the Markov / CCMT system state transition probability mapping matrix model through Monte Carlo system state grid element representative point random sampling, simulation and statistical analysis to respond to changes in accompanying parameters.

[0023] Furthermore, the system state scenario behavior dynamic deduction algorithm will lump together the same system state points after each iterative search step, avoiding the exponential growth of the number of system state sequence branches during each iteration, so that the number of system state points will never exceed the size of the entire system state space, effectively reducing the system state search space and significantly improving the search efficiency.

[0024] Furthermore, during the dynamic deduction of the system state scenario behavior, the potential transition of the system state after each assimilation is evolved, and the probability of occurrence of different system states after evolution is obtained by summing the cumulative probabilities during each matrix iterative search process, thereby realizing dynamic system reliability prediction analysis, including:

[0025] The estimated system states obtained from each iterative search are sorted by their probability of occurrence through a list. The larger the probability value, the greater the possibility of the system state occurring.

[0026] The digitally coded system state vector after lumped sorting at each search step is converted into a common language description that can be understood by system operators, and the dynamic evolution process of the system state is displayed through the human-machine interface; at the same time, the obtained system state failure probability is compared with the acceptable criteria of the system operation safety limit to guide the safe operation of the system.

[0027] Furthermore, for complex nonlinear process control systems, it is difficult to obtain an accurate analytical solution for the conditional transfer probability matrix. The system state transfer probability mapping matrix model adopts an accurate sampling method of the system state sampling point movement distribution based on Monte Carlo simulation, and is obtained through statistical analysis of system state migration trajectory simulation and tracing.

[0028] Furthermore, the method for accurately sampling the movement distribution of the system state sampling points includes:

[0029] Combined with the data assimilation process based on particle filtering, the initial state position of the system is determined by monitoring the input or hypothesis of the system state at the initial moment, Gaussian sampling is performed on the initial state position point of the system, and N particles are generated around each variable in the system state and its accompanying parameters. The initial state position point of the system includes its accompanying parameters.

[0030] Import the system state of the sampled particles and their accompanying parameters into the system simulation model to simulate and predict the particle trajectory changes and the final value distribution of the particles;

[0031] Reading in the real-time state observation data of the system, and using the data assimilation algorithm based on particle filtering to complete the weight value calculation and resampling of the simulated prediction particles;

[0032] According to the movement distribution of the particle swarm after resampling, the system state space is reconstructed to more accurately simulate and reflect the dynamic behavior characteristics of the system.

[0033] Furthermore, the method also includes a self-updating construction process of the Markov / CCMT system state transition probability mapping matrix model. By searching for matching elements in the potential transition mapping relationship submatrix of the current system state and lumping the same system state, rapid deduction of system state scenario behavior and dynamic reliability prediction analysis are achieved, including updating the state transition probability matrix of the system physical components and the conditional transition probability matrix, wherein:

[0034] The updating of the state transition probability matrix of the system physical components takes the actual state monitoring of the system equipment as input, and performs zeroing and sparse processing on the irrelevant system configuration configuration and state transition mapping relationship of the state transition probability matrix of the system physical components;

[0035] The update of the conditional transition probability matrix takes the actual state of the system equipment and the state monitoring of the system process variables as input, uniformly samples the particle swarm in the grid element space where the initial state of the system is located after assimilation and update, simulates the trajectory changes of the particle swarm on the system simulation model, and statistically calculates the distribution of the particle final value position in different grid element spaces. The conditional transition probability of the system state under the current system configuration is approximately estimated and updated by the equal-weighted dot product method;

[0036] While completing the self-updated construction of the system state transition probability mapping matrix of this round, the system state accompanying parameters of different representative points falling into the gate element obtained by random sampling and simulation statistical analysis in this round of iterative search steps are averaged to adapt to the reconstruction of the Markov / CCMT system state transition probability mapping matrix model in the next round of iterative search.

[0037] Furthermore, the dynamic search range of the Markov / CCMT system state transition probability mapping matrix model Q is limited to the potential transition mapping relationship submatrix of the current system state, that is, in the self-update construction process of the Markov / CCMT system state transition probability mapping matrix model Q, only the conditional transition probability of the system state under the current system configuration is considered, and the conditional transition probabilities between the remaining system states are automatically assigned to 0, thereby reducing the complexity of the construction of the conditional transition probability matrix G, and only traversing and storing the non-zero elements in the search process, thereby improving the search speed.

[0038] The second object of the present invention can be achieved by adopting the following technical solutions:

[0039] A multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis system, the system comprising:

[0040] The particle swarm distribution acquisition module is used to obtain the particle swarm distribution of the system assimilation initial state through Gaussian sampling based on the system state spatiotemporal coupling model and the initial value input or assumption of the system state. The module is then imported into the system simulation model to simulate the trajectory changes of the particle swarm. Combined with the real-time monitoring input of the system state, the posterior distribution of the particle swarm of the system state at the current moment is obtained through data assimilation, and used as the particle swarm distribution of the system assimilation initial state at the next moment.

[0041] The system state space model reconstruction module is used to determine the upper and lower boundary values ​​of the system state originating grid element based on the posterior distribution range of the current system state particle swarm obtained through the data assimilation process, and use this as the new grid element scale to discretize the system state space. Through digital coding, the reconstruction of the system state space model is completed;

[0042] The Markov / CCMT system state transition probability mapping matrix model generation module is used to uniformly sample the system state originating grid element to obtain the distribution of the initial state particle swarm for system dynamic behavior deduction, import the initial state particle swarm for system dynamic behavior deduction into the system simulation model, simulate the trajectory changes of the taken particle swarm, and obtain the conditional transition probability matrix of the system state under the current system configuration based on the statistical estimation of the number of particle swarms falling into different grid elements; the conditional transition probability matrix is ​​combined with the state transition probability matrix of the system physical components to obtain the Markov / CCMT system state transition probability mapping matrix model;

[0043] The system state scenario behavior dynamic deduction algorithm and reliability prediction analysis module is used to perform advanced system state prediction analysis based on the constructed Markov / CCMT system state transition probability mapping matrix model. The system state scenario behavior dynamic deduction algorithm is further combined to merge the same system states, and the combined and reduced unique system states are pressed into the search sequence structure. The Markov / CCMT system state transition probability mapping matrix model is repeated to construct an update and iterative search process until the specified search depth is reached.

[0044] The system analysis result display module is used to sort and predict the dynamic evolution of the system state scenario sequence according to the probability of the system state after the merger, and guide the safe operation of the system through a graphical display on the interface.

[0045] The third object of the present invention can be achieved by adopting the following technical solutions:

[0046] A computer device includes a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, the above-mentioned multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method is implemented.

[0047] The fourth object of the present invention can be achieved by adopting the following technical solutions:

[0048] A storage medium stores a program, which, when executed by a processor, implements the above-mentioned multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method.

[0049] The present invention has the following beneficial effects compared to the prior art:

[0050] 1. The method provided by the present invention designs an accurate sampling method for the moving distribution of system state sampling points based on data assimilation on the basis of the existing equal-weighted dot product method based on Monte Carlo simulation. According to the assimilated system state and the posterior distribution of the particle swarm of its accompanying parameters, the grid cell structure of the system state space can be reshaped, avoiding the sampling point deviation problem caused by the division of the fixed grid cell structure, thereby more accurately describing and predicting the dynamic behavior characteristics of the system.

[0051] 2. The method provided by the present invention integrates multi-source data such as system simulation data, system real-time observation data, and equipment reliability characteristic data, and realizes a more accurate mapping of the system state and its accompanying parameters through a data assimilation algorithm based on particle filtering. Combined with the system state probability mapping matrix model generation and self-update construction process, it accurately predicts the dynamic evolution process of the system.

[0052] 3. The matrix-coding-based system state scenario behavior dynamic deduction algorithm and reliability prediction analysis method provided by the present invention realizes the low-dimensional sparse improvement of the high-dimensional system state space matrix model during the model construction process, and effectively suppresses the system state space explosion problem by aggregating and normalizing the same system state points in the subsequent search and analysis process. The search efficiency is greatly improved, meeting the needs of advanced / ultra-real-time prediction and analysis, and can provide proactive safety guidance for the intelligent operation and maintenance of large-scale, complex, critical safety systems such as nuclear power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0054] Figure 1 This is a schematic diagram of the multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method of Example 1 of the present invention.

[0055] Figure 2 This is a flowchart of the multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method of Example 1 of the present invention.

[0056] Figure 3 This is a structural block diagram of a water level control system for a nuclear power plant steam generator according to embodiment 2 of the present invention.

[0057] Figure 4 This is a schematic diagram of the control process of the water level control system of a nuclear power plant steam generator according to Example 2 of the present invention.

[0058] Figure 5 This is a flow chart of an accurate sampling algorithm for system state sampling point movement distribution based on data assimilation according to embodiment 2 of the present invention.

[0059] Figure 6 This is a diagram showing the system state space matrix encoding process of Example 2 of the present invention.

[0060] Figure 7 These are the system state evolution results at different search depths in Example 2 of the present invention.

[0061] Figure 8 This is a structural block diagram of the multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis system of Example 3 of the present invention.

[0062] Figure 9 This is a structural block diagram of a computer device according to embodiment 4 of the present invention. DETAILED DESCRIPTION

[0063] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. It should be understood that the specific embodiments described are only used to explain this application and are not used to limit this application.

[0064] This paper presents a method for scenario-based behavior deduction and reliability prediction analysis of dynamic systems using multi-source data fusion. It primarily simulates trajectory generation in complex process control systems, enabling dynamic reliability and scenario sequence deduction and analysis. The algorithm and software application platforms involved are implemented in MATLAB / Simulink 2021 and Eclipse 4.18, respectively.

[0065] Example 1:

[0066] like Figure 1 、 2 As shown, the multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method provided in this embodiment includes the following steps:

[0067] (1) According to the system operation design requirements and equipment failure mode and impact analysis, the state of the system's controlled continuous process variables and the state of the system's control equipment are discretized and represented by digital matrix coding to construct a spatiotemporal coupling model of the initialization system state.

[0068] The system state spatiotemporal coupling model is the coupling of the state space of the process variables of the discretized controlled system and the state space of the physical control components of the system at different time points, where:

[0069] The state space of the system's physical control components is determined based on the system equipment failure mode and impact analysis, including a finite number of states: normal state, fault state 1, fault state 2, etc.

[0070] The discretization process of the system's continuous process variable state space needs to comprehensively consider the system operation design requirements and the accuracy requirements of modeling and analysis applications, and divide the system process variables in each dimension into multiple continuous but non-overlapping discrete grid element spaces;

[0071] The discretization of the time dimension is reflected in the setting of the time step. In principle, the system time step should be determined based on the specific accuracy requirements of the system modeling and analysis application, combined with the cell division of the system state space, to meet the principle of maximizing the probability of system state transitions between adjacent cells within a unit time step, thereby accurately describing and characterizing the dynamic behavior characteristics of the system. The matrix search time step used in the multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method provided by this invention is consistent with the system simulation time step.

[0072] The system states at different time points are related through the Chapman-Kolmogorov equation to realize the interactive coupling between the discretized grid element space of the system physical component states and the discretized grid element space of the system process variable states, and the system state spatiotemporal coupling model is obtained through digital coding.

[0073] (2) Establish a mathematical simulation model of the system based on the system dynamic operation equations and control laws, build an equipment fault injection model based on the failure modes and failure mechanisms of the system components, and embed the equipment fault injection model into the system simulation model.

[0074] The system simulation model includes a normal simulation model and a device fault injection model, where:

[0075] Normal simulation model construction uses the system operation dynamic partial differential equations and control laws as the basic modeling input, and is implemented through MATLAB functions and Simulink simulation modules;

[0076] The construction of the equipment fault injection model is based on the analysis of equipment failure modes and failure mechanisms. The equipment failure mode is reflected as the explicit mapping relationship between the input and output of the equipment parameters, and its internal mapping function is determined by the equipment failure mechanism, which is also implemented in the MATLAB / Simulink simulation environment.

[0077] (3) Based on the initialized spatiotemporal coupling model of the system state, according to the initial value input or assumption of the system state, the initial state of the system at the initial moment and its accompanying parameters (S0(t0), S'0(t0)) are determined and Gaussian sampling with a variance of σ is performed N times to obtain the particle swarm distribution of the initial state of the system at the initial moment (S ss (t0), S s ' s (t0)), and simulate the expected sampled particles at different times t on the system simulation model. i The running trajectory and terminal position point distribution on (S se (t i ), S' se (t i )).

[0078] (4) Pairing the actual observed values ​​of the system state and its accompanying parameters at the next moment (S m (t i ), S' m (t i )) input, the predicted value of the particle simulation sampled by the system state (S se (t i ), S' se (t i )) and the actual observed value (S m (t i ), S' m (t i The weight value w of the sampled particles in each system initial state is calculated by the distance between n and w′ n .

[0079] The discrete degree distribution (expressed as distance) of the simulated predicted value of the sampled particle state relative to the actual observed value is calculated by the following formula:

[0080]

[0081] Resampling is performed based on the particle weights (importance) of the system state to obtain a more accurate posterior estimated distribution of the system state and its accompanying parameters after assimilation (S r (t i ), S' r (t i )) and re-enter the system as the initial state particle swarm input to the system simulation model for further simulation and prediction of the next moment t i+1 The system state and its accompanying parameter particle posterior estimation distribution (S se (t i+1 ), S' se (t i+1 )).

[0082] (5) According to the assimilated system state (S r (t i )) Particle swarm distribution, reconstructing the system state space grid element structure; at the same time, the assimilated system state is used as the initial state of the system assimilation at the current moment Used for system state scenario behavior deduction analysis, the system state accompanying parameter S' after resampling r (t i ) to perform average initialization processing to adapt to the initial state of the system after assimilation

[0083] The reconstruction of the grid cell structure in the system state space is to use the two particles with the greatest distance between them in the system state particle group as the upper and lower boundary values ​​of the initial state grid cell of the assimilated system, and use this as a new scale to redefine the system state space. After the other grid cells are determined, the grid cells at the boundary of the normal range of the system are adaptively adjusted according to the remaining space in the normal range, so as to finally complete the reconstruction of the grid cell structure in the system state space.

[0084] Re-encode the reconstructed system state space grid element digitally to obtain a new system state space matrix model;

[0085] System state accompanying parameter S' r (t i ) is to average the predicted values ​​of all N system state accompanying parameter sampling particles, which is calculated by the following formula:

[0086]

[0087] Where S' r (t i ) represents the particle value after the system state and parameter resampling at the current moment, and N is the total number of particles sampled in the initial state of the system.

[0088] (6) The assimilated updated system state and its average accompanying parameters As the initial state of the system at the current moment S0(t i ) and its accompanying parameter S'0(t i ) and adds it to the system state search sequence structure, and performs initial settings on the occurrence probability P0 of the current system state, the search depth K, and the system configuration m′.

[0089] The system configuration is performed by monitoring the status of the system physical devices in real time, and the configured extended system status is constructed into an initial search sequence structure.

[0090] (7) The extended system state in the search sequence structure is extracted one by one, and the starting point of the current extended system state is uniformly sampled according to the gate element position of the current extended system state, and the dynamic trajectory migration changes of the random sampling points under the specific system configuration between the gate elements of different system state spaces are simulated through the system mathematical simulation model; further, uniform sampling refers to uniformly randomly extracting U representative point particles from the gate element space where the current extended system state is located, and simulating U times on the system simulation model.

[0091] (8) Statistically analyze the distribution of the points where the system state trajectory falls in different grid element spaces, calculate the conditional transition probability of the system state, couple the state transition probability of the system physical components, and generate the system state transition probability mapping matrix model Q; at the same time, use the following formula (3) to perform local averaging on the final values ​​of the system state accompanying parameters falling into different space grid elements, and use them as the system state accompanying parameters corresponding to the initial grid element in the next iterative search process;

[0092]

[0093] Where s′ s are different accompanying parameter particle values ​​under the same system state set, ∑ s s′ s It represents the sum of all the particle values ​​of the accompanying parameters under the same system state, n s is the number of particles in the set with the same system state, Represents the local average value of the system state parameters falling into different gate cells.

[0094] The system state transition probability mapping matrix model includes the system physical component state transition probability matrix (H matrix) and the system state conditional transition probability matrix (G matrix), where:

[0095] The construction of the state transition probability matrix of the system's physical components is based on failure mode and impact analysis of the system components and finite state machine modeling. Under the assumption that component failures are independent of each other and the operating time span is very small, the state transition probability of the system's physical components can be simply treated as the product of the failure probabilities of individual components. The model's self-update construction process can timely update the system's physical component state transition probability matrix H based on feedback from the system's real-time operation monitoring data on the state of the system's physical equipment, zeroing out state transition probabilities that are unrelated to the current system configuration, thereby achieving low-dimensional sparse processing of the H matrix.

[0096] The self-updating construction of the system state conditional transition probability matrix is ​​achieved through the equal-weighted dot product method based on Monte Carlo simulation. Based on the construction of the aforementioned system simulation model, accurate sampling of system state grid element representative points, simulation of system state migration trajectory, tracing and statistical analysis, the system state conditional transition probability is calculated using the following equal-weighted dot product method:

[0097]

[0098] Where u(j|j') is the number of system states transferred from cell j' to cell j under a given system configuration m', and U is the total number of sampled particles.

[0099] The transition probability of the system state in different gate element spaces can be approximately estimated by the ratio of the number of points where a specific gate element falls into to the total number of sampling points;

[0100] Finally, multiply the matrix G by the matrix H to obtain the updated system state transition probability matrix Q.

[0101] (9) Through the forward search analysis of the synchronous system state transition probability matrix model, the potential transition path sequence of the system state is identified, and the occurrence probability of the system state transition sequence path is calculated.

[0102] The core of the system state space search analysis is to identify non-zero or conforming elements (applicable to the application of truncation criterion ε) in the system state transition probability mapping matrix Q; the occurrence probability of a sequence path is the product of the transition probabilities of different branch segments on the sequence path.

[0103] (10) After each search step (traversing all extended system states in the search sequence structure) is completed, the system state transition branches with the same final state are merged, and the probability of the merged system state is equal to the sum of the probabilities of all system state transition branches under the set to which the same system state belongs.

[0104] (11) The merged unique system state is re-added to the search sequence structure as a new parent node, and for each extended system state in the search sequence structure, its system state transition probability mapping probability submatrix Q is reconstructed through Monte Carlo random sampling simulation and statistical analysis, and the next iterative search is entered until the specified deduction search depth K or a specific exit condition is reached, and the system state assimilation value at the next moment is updated and input.

[0105] (12) The dynamic evolution of the system state scenario sequence is graphically displayed in the form of a vector matrix, and the possibility of the development of the system state scenario sequence is sorted according to the probabilistic risk, which is used to proactively guide the safety of nuclear power plant operation.

[0106] The method provided in this embodiment is based on the Markov / CCMT dynamic reliability prediction and analysis method, combined with the fusion and assimilation method of multiple source data (real-time system status monitoring data stream input, simulation data, system equipment reliability characteristic data, etc.), and uses the Monte Carlo probability model random sampling idea to simulate and statistically analyze the complex dynamic behavior characteristics of digital process control under strong interactive coupling, nonlinearity and high uncertainty. On this basis, through dynamic search analysis of the system state transition probability matrix model, forward deduction analysis and reliability prediction of the system operating state are realized, solving key technical problems such as the difficulty in obtaining analytical solutions for nonlinear dynamic process control systems and alleviating the system state space search explosion, thereby laying the foundation for accurate modeling, efficient analysis and calculation, and intelligent operation and maintenance management of digital instrumentation and control systems in large and complex nuclear power plants.

[0107] Example 2:

[0108] like Figure 3 、 4 As shown, this embodiment takes a simplified nuclear power plant steam generator water level control system as an example to carry out dynamic reliability modeling and scenario sequence deduction analysis of the digital process control system to verify the effectiveness of the method of the present invention. The implementation steps specifically include:

[0109] Step 1: Perform a failure mode and impact analysis on the digital control unit in the example system and define the state of the system's physical control unit. The nuclear power plant steam generator water level control system in this embodiment consists of a water level sensor, a steam flow sensor, a feedwater flow sensor, a PI controller, a feedwater flow control valve, and a steam flow control valve. Assume that the sensor unit has four states: operation, stuck, constant gain change, and constant deviation failure. The PI control unit and feedwater flow control unit have four states: operation, stuck, high output, and low output. Since the steam generator water level is primarily controlled by the feedwater flow control valve in actual applications, this embodiment does not consider the operation and failure modes of the steam flow control valve.

[0110] The states of the system's physical control components are represented by the following vector matrix [S1, S2, S3, S4, S5]. S1 represents the state of the water level sensor, S2 represents the state of the steam flow sensor, S3 represents the state of the feedwater flow sensor, S4 represents the state of the PI controller, and S5 represents the state of the feedwater flow control valve. The digital codes corresponding to the states of different system devices are shown in Table 1.

[0111] Table 1 Definition of digital codes for different equipment states

[0112]

[0113] Step 2: According to Figure 1and Figure 2 The structure, control law and operating characteristics of the steam generator water level control system of the nuclear power plant shown in the figure are analyzed, and the system dynamic equation is established as follows:

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129] in:

[0130]

[0131]

[0132]

[0133] W so =C so ·P s (twenty three)

[0134]

[0135]

[0136]

[0137] det=trc11·trc3+trc12·trc2 (27)

[0138]

[0139]

[0140]

[0141]

[0142]

[0143]

[0144] trc2=V s ·(ρ s ·trc7+h s ·trc9)+h e ·cfdr4 (34)

[0145] trc3=V s ·(ρ s ·trc8+h s ·trc10)-h e ·cfdr5 (35)

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152]

[0153] trc11=V s ·trc9+cfdr4 (43)

[0154] trc12=V s ·trc10+cfdr5 (44)

[0155]

[0156]

[0157]

[0158]

[0159]

[0160] h e =h f +x e ·h fg (50)

[0161]

[0162] L=L sb +L b (52)

[0163] Based on the above system dynamic equations, a mathematical simulation model of the system was established in the MATLAB / Simulink simulation environment. A fault injection model was also developed for the digital components and embedded into the system simulation model. The symbols and meanings of the relevant parameters in the model are shown in Table 2.

[0164] Table 2 Related variables and symbol meanings in the examples

[0165]

[0166]

[0167]

[0168] Step 3: The state transition probability matrix H(m|m',j'→j,Δt) of the system physical components is calculated based on the reliability characteristic parameters such as the failure rate and repair rate of the system components.

[0169] Step 4: Determine the system initial state and its accompanying parameters based on the system operation status input.

[0170] Assume that the initial water level of the system is L dw =3.299m, the system initial state accompanying parameters include the average temperature of the fluid in the primary rising section T p1 =302.415℃, average temperature of fluid in the primary downflow section T p2 =295.470℃, average temperature of the primary side rising section pipe wall T m1 =296.707℃, average temperature of the primary side descending section pipe wall T m2 =291.053℃, working pressure of gas-liquid mixing zone P s =5.699Mpa, gas content at the outlet of the U-tube heating zone X e=0.2253, the average temperature of the fluid in the water supply chamber T dw =261.236℃, the average temperature of the fluid in the descending channel T d =261.230℃, filter output signal V = 2.542e-5, PI controller 1 output signal U = 1.189e-5, PI controller 2 output signal W = -0.264, steam flow and feed water flow difference m = -26.209 kg / s, feed water valve output signal r = -3.027e-10, feed water flow W fi =457.218kg / s.

[0171] Step 5: Obtain the particle swarm distribution of the system initial state and its accompanying parameter values ​​through random sampling of Gaussian distribution (N=100). The variance of random sampling of Gaussian distribution is σ=diag(0.0001, 0.001, 0.001, 0.001, 0.001, 0.0001, 0.025, 0.001). Set the system simulation step length Δt=0.1s. Substitute the particles sampled from the system initial state and its accompanying parameters into the system mathematical simulation model to simulate the trajectory change of the system state, and statistically analyze to obtain the simulated predicted movement distribution of the particle swarm sampled from the system initial state and its accompanying parameters at the initial moment; read in the actual observation value of the system state at the current moment, and realize the precise value of the system initial water level and its accompanying parameters at the current moment through the data assimilation algorithm based on particle filtering. The implementation process of the precise sampling distribution method of the system state sampling point movement distribution based on data assimilation is shown in the attached figure. Figure 5 .

[0172] The exact values ​​of the system initial water level and accompanying parameters after particle filter data assimilation are listed as follows:

[0173] L dw =3.2403m, T p1 =302.4184℃、T p2 =295.4658℃、T m1 =296.7055℃、T m2 =291.0511℃、P s =5.6992Mpa, X e =0.2253, T dw =261.2309℃、T d =261.2289℃, V=2.542e-5, U=1.189e-5, W=-0.264, m=-26.209kg / s, r=-3.027e-10, W fi =457.218kg / s.

[0174] Step 6: Reconstruct the system state cell space based on the posterior estimated distribution of the system water level state and its accompanying parameters after resampling and assimilation, and use the two particles with the farthest distance between them in the resampled system state particle group as the upper and lower boundary values ​​of the starting cell of the system initial state at the current moment, and use this as a ruler to redefine and divide the system state space cell structure.

[0175] Assuming that the effective control range of the system water level is: 3m≤x≤5m, the distribution range of the system state particle swarm after assimilation is (3.144, 3.324], and the system water level state is reconstructed as a new scale and defined as follows:

[0176] x1<3m, 3m≤x2≤3.144m, 3.144m <x3≤3.324m,3.324m<x4≤3.504m,3.504m<x5≤3.684m,3.684m<x6≤3.864m,3.864m<x7≤4.044m,4.044m<x8≤4.224m,4.224m<x9≤4.404m,4.404m<x 10 ≤4.584m, 4.584m <x 11 ≤4.764m, 4.764m <x 12 ≤4.944m, 4.944m <x 13 ≤5.0m, x 14 >5m.

[0177] Step 7: Add the assimilated system state as the system initial state to the system state search sequence structure for iterative deduction and analysis of system state scenario behavior. The implementation process of the system state scenario behavior dynamic deduction method based on matrix coding data assimilation is shown in the attached Figure 2 , where the matrix encoding process is shown in Figure 6 .

[0178] Set the system simulation step Δt = 0.1s, the deduction depth K = 5, the system initial state probability P0 = 1, and the truncation probability ε = 10 -5 The system state elements in the system state search sequence structure are extracted one by one, and for the extracted current system state, U=100 particles are uniformly and randomly sampled in the grid element space where it is located. The sampled current system state grid element representative points are injected into the system simulation model to simulate and predict the trajectory changes of the sampled particles of the current system state, and the number of points FP where the sampled particles fall in each grid element is counted. jThe conditional transition probability of the current system state is approximated by the equal-weight dot product method, and the conditional transition probabilities of the system state between the remaining gate elements are set to 0, thereby obtaining the system state conditional transition probability matrix G(j|j',m',Δt). The system state transition probability matrix Q is obtained by multiplying the system state conditional transition probability matrix G with the state transition probability matrix H of the system physical control components.

[0179] Step 8: Based on the self-updating construction and forward search of the system state probability mapping matrix model, the dynamic evolution deduction analysis of the system state is realized. The results of the system state deduction analysis are shown in Figure 7 .

[0180] The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solution and invention concept of the present invention within the scope disclosed by the present invention, which falls within the scope of protection of the present invention.

[0181] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above embodiments may be completed by instructing related hardware through a program, and the corresponding program may be stored in a computer-readable storage medium.

[0182] It should be noted that although the method operations of the above embodiments are described in a particular order in the accompanying drawings, this does not require or imply that the operations must be performed in this particular order, or that all of the illustrated operations must be performed to achieve the desired results. Rather, the depicted steps may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into a single step, and / or a single step may be broken down into multiple steps.

[0183] Example 3:

[0184] like Figure 8 As shown, this embodiment provides a multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis system, which includes a particle swarm distribution acquisition module 801, a system state space model reconstruction module 802, a system state transition probability mapping matrix model generation module 803, a system state scenario behavior dynamic deduction algorithm and reliability prediction analysis module 804 and a system analysis result display module 805, wherein:

[0185] The particle swarm distribution acquisition module 801 is used to obtain the particle swarm distribution of the system assimilation initial state through Gaussian sampling based on the system state spatiotemporal coupling model and the system state initial value input or assumption, import it into the system simulation model to simulate the trajectory change of the particle swarm, and combine it with the real-time monitoring input of the system state to obtain the posterior distribution of the particle swarm of the current system state through data assimilation, and use it as the particle swarm distribution of the system assimilation initial state at the next moment;

[0186] The system state space model reconstruction module 802 is used to determine the upper and lower boundary values ​​of the system state origin cell based on the posterior distribution range of the current system state particle swarm obtained during the data assimilation process, and use this as the new cell scale to discretize the system state space and complete the reconstruction of the system state space model through digital coding;

[0187] The system state transition probability mapping matrix model generation module 803 is used to uniformly sample the system state originating grid element to obtain the distribution of the initial state particle swarm for system dynamic behavior deduction, import the initial state particle swarm for system dynamic behavior deduction into the system simulation model, simulate the trajectory changes of the obtained particle swarm, and obtain the conditional transition probability matrix of the system state under the current system configuration based on the statistical estimation of the number of particle swarms falling into different grid elements; the conditional transition probability matrix is ​​combined with the state transition probability matrix of the system physical components to obtain the Markov / CCMT system state transition probability mapping matrix model;

[0188] The system state scenario behavior dynamic deduction algorithm and reliability prediction analysis module 804 is used to apply the system state scenario behavior dynamic deduction algorithm based on the constructed Markov / CCMT system state transition probability mapping matrix model to perform system state advance prediction analysis, further merge identical system states, and compress the merged and reduced unique system states into the search sequence structure, repeating the Markov / CCMT system state transition probability mapping matrix model construction update and iterative search process until a specified search depth is reached;

[0189] The system analysis result display module 805 is used to sort and predict the dynamic evolution of the system state scenario sequence according to the probability of the merged system state, and guide the safe operation of the system through a graphical interface display.

[0190] The specific implementation of each module in this embodiment can be found in the above-mentioned embodiment 1, and will not be described one by one here; it should be noted that the system provided in this embodiment is only illustrated by the division of the above-mentioned functional modules. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure can be divided into different functional modules to complete all or part of the functions described above.

[0191] Example 4:

[0192] This embodiment provides a computer device, which can be a computer, such as Figure 9 As shown, a processor 902, a memory, an input device 903, a display 904, and a network interface 905 are connected via a system bus 901. The processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium 906 and an internal memory 907. The non-volatile storage medium 906 stores an operating system, a computer program, and a database. The internal memory 907 provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. When the processor 902 executes the computer program stored in the memory, the multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method of the above-mentioned embodiment 1 is implemented as follows:

[0193] According to the initial value input or assumption of the system state, the particle swarm distribution of the system assimilation initial state is obtained through Gaussian sampling, which is imported into the system simulation model to simulate the trajectory change of the particle swarm. Combined with the real-time monitoring input of the system state, the posterior distribution of the particle swarm of the current system state is obtained through data assimilation, and used as the particle swarm distribution of the system assimilation initial state at the next moment;

[0194] According to the posterior distribution range of the particle swarm of the current system state obtained through the data assimilation process, the upper and lower boundary values ​​of the system state originating grid element are determined, and this is used as the new grid element scale to discretize the system state space. Through digital coding, the reconstruction of the system state space model is completed.

[0195] Uniform sampling is performed on the system state originating grid element to obtain the distribution of the initial state particle swarm of the system dynamic behavior deduction. The initial state particle swarm of the system dynamic behavior deduction is imported into the system simulation model, and the trajectory change of the obtained particle swarm is simulated. The conditional transition probability matrix of the system state under the current system configuration is estimated based on the number of particle swarms falling into different grid elements; the conditional transition probability matrix is ​​combined with the state transition probability matrix of the system physical components to obtain the Markov / CCMT system state transition probability mapping matrix model;

[0196] Based on the constructed Markov / CCMT system state transition probability mapping matrix model, the system state scenario behavior dynamic deduction algorithm is applied to perform system state advance prediction analysis. The same system states are further merged, and the merged and reduced unique system states are pressed into the search sequence structure. The Markov / CCMT system state transition probability mapping matrix model is repeated to construct an update and iterative search process until the specified search depth is reached.

[0197] According to the probability of the system state after the merger, the dynamic evolution of the system state scenario sequence is sorted and predicted, and the safe operation of the system is guided through the graphical display of the interface.

[0198] Example 5:

[0199] This embodiment provides a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, the method for scenario behavior deduction and reliability prediction analysis of a multi-source data fusion dynamic system of the above-mentioned embodiment 1 is implemented as follows:

[0200] According to the initial value input or assumption of the system state, the particle swarm distribution of the system assimilation initial state is obtained through Gaussian sampling, which is imported into the system simulation model to simulate the trajectory change of the particle swarm. Combined with the real-time monitoring input of the system state, the posterior distribution of the particle swarm of the current system state is obtained through data assimilation, and used as the particle swarm distribution of the system assimilation initial state at the next moment;

[0201] According to the posterior distribution range of the particle swarm of the current system state obtained through the data assimilation process, the upper and lower boundary values ​​of the system state originating grid element are determined, and this is used as the new grid element scale to discretize the system state space. Through digital coding, the reconstruction of the system state space model is completed.

[0202] Uniform sampling is performed on the system state originating grid element to obtain the distribution of the initial state particle swarm of the system dynamic behavior deduction. The initial state particle swarm of the system dynamic behavior deduction is imported into the system simulation model, and the trajectory change of the obtained particle swarm is simulated. The conditional transition probability matrix of the system state under the current system configuration is estimated based on the number of particle swarms falling into different grid elements; the conditional transition probability matrix is ​​combined with the state transition probability matrix of the system physical components to obtain the Markov / CCMT system state transition probability mapping matrix model;

[0203] Based on the constructed Markov / CCMT system state transition probability mapping matrix model, the system state scenario behavior dynamic deduction algorithm is applied to perform system state advance prediction analysis. The same system states are further merged, and the merged and reduced unique system states are pressed into the search sequence structure. The Markov / CCMT system state transition probability mapping matrix model is repeated to construct an update and iterative search process until the specified search depth is reached.

[0204] According to the probability of the system state after the merger, the dynamic evolution of the system state scenario sequence is sorted and predicted, and the safe operation of the system is guided through the graphical display of the interface.

[0205] It should be noted that the computer-readable storage medium of the present embodiment may be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0206] In summary, the present invention constructs a mathematical simulation model of the system, and grid-cells the system state space; determines the system state and its accompanying parameters at the initial moment according to the initial value input or assumption of the system operation state, randomly samples the initial state of the system and its accompanying parameters at the initial moment based on Gaussian distribution, and uses the system simulation model to simulate and predict the trajectory changes and final value position point movement distribution of the sampled particles of the initial state of the system at the initial moment; combines the actual observation data input of the system operation state, and realizes the weight calculation and resampling of the particle group of the system state simulation prediction point through the data assimilation algorithm to obtain the assimilated system state and its accompanying parameters. The posterior estimation distribution of the accompanying parameters; according to the distribution of the system state particle swarm after resampling and assimilation, the system state space grid element structure is reconstructed, and the assimilated system state is used as the system initial state for subsequent system state scenario behavior deduction analysis; at the same time, the accompanying parameters of the assimilated system state are initially averaged to adapt to the normalization and initialization of the assimilated system state particle swarm; the probability P0 of the system assimilated initial state at the current moment, the deduction depth K, and the system configuration m′ are initialized, the system assimilated initial state at the current moment is added to the search sequence structure, and the system state is extracted from the search sequence structure one by one. The representative points of the grid element space where the current system state is located are randomly sampled by uniform distribution, and the running trajectory simulation prediction and statistical analysis of the particle swarm sampled by the current system state are also carried out on the system simulation model; based on the statistical distribution of the landing points of the system state sampling particle swarm in different grid element spaces, the equal-weighted dot product method is used to estimate the conditional transfer probability matrix G of the system state at the current moment, and through coupling with the system physical device state transfer probability matrix H, the self-updating construction of the system state transfer probability matrix model Q is realized; based on the system state probability mapping matrix model, the system state scenario behavior dynamic evolution analysis algorithm is used The method searches for non-zero matching elements in the matrix model and obtains the probability of occurrence of the potential transfer path sequence of the current system state through probability accumulation calculation; after completing each iterative deduction search, the same system state point and its transfer path sequence are collected and merged, and the probability of the evolution of the system state is estimated by summing the probabilities of all transfer path sequence branches under the corresponding set; the unique system state after the collection and simplification is added as a new parent node to the search sequence structure for the next iterative deduction search until the specified deduction depth K is specified, and the system state assimilation update input is waited for at the next moment, so as to dynamically deduce and predict the evolution of the system state. The present invention can realize the adaptive update construction of the state transition probability matrix of large-scale complex digital process control systems and the dynamic deduction analysis of scenario behavior. By low-dimensional sparsification of the matrix and assimilation and intensive processing of system state data, the explosion problem of high-dimensional system state space search is avoided. At the same time, combined with the accurate sampling of the movement distribution of system state sampling points based on data assimilation, the dynamic behavior characteristics of the system are accurately simulated and mapped, and the dynamic reliability prediction analysis of the system is realized.

[0207] The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solution and inventive concept of the present invention within the scope disclosed by the present invention, which falls within the scope of protection of the present invention.

Claims

1. A method for scenario behavior deduction and reliability prediction analysis of dynamic systems using multi-source data fusion, characterized by: The method comprises: According to the initial value input or assumption of the system state, the particle swarm distribution of the system assimilation initial state is obtained through Gaussian sampling, which is imported into the system simulation model to simulate the trajectory change of the particle swarm. Combined with the real-time monitoring input of the system state, the posterior distribution of the particle swarm of the current system state is obtained through data assimilation, and used as the particle swarm distribution of the system assimilation initial state at the next moment; According to the posterior distribution range of the particle swarm of the current system state obtained through the data assimilation process, the upper and lower boundary values ​​of the system state originating grid element are determined, and this is used as the new grid element scale to discretize the system state space. Through digital coding, the reconstruction of the system state space model is completed. Uniform sampling is performed on the system state originating grid element to obtain the distribution of the initial state particle swarm for system dynamic behavior deduction. The initial state particle swarm for system dynamic behavior deduction is imported into the system simulation model, and the trajectory changes of the obtained particle swarm are simulated. Based on the statistical estimation of the number of particle swarms falling into different grid elements, the conditional transition probability matrix of the system state under the current system configuration is obtained; the conditional transition probability matrix is ​​multiplied by the state transition probability matrix of the system physical components to obtain the Markov / CCMT system state transition probability mapping matrix model; under the condition that the components are independent of each other, the state transition probability matrix of the system physical components is the product of the failure rates of the components within time Δt; Based on the constructed Markov / CCMT system state transition probability mapping matrix model, the system state scenario behavior dynamic deduction algorithm is applied to perform system state advance prediction analysis. The same system states are further merged, and the merged and reduced unique system states are pressed into the search sequence structure. The Markov / CCMT system state transition probability mapping matrix model is repeated to construct an update and iterative search process until the specified search depth is reached. The system status scenarios are sorted by probability after merging, and the dynamic evolution of the system status scenario sequence is predicted. The system status evolution process is graphically displayed on the interface. When the probability of system status failure exceeds the safety limit, an early warning is triggered to guide the safe operation of the system. wherein, the system configuration structure is configured according to the real-time status monitoring input of the system equipment, and the search sequence structure is constructed according to the initialization and expansion system state after the configuration; The extended system state includes the system state and its accompanying parameters. The accompanying parameter state is the system parameter associated with the system state change. The system state and its accompanying parameters and the state of the system's physical components are coupled with each other, but do not directly affect the dynamic behavior characteristics of the system. They react indirectly through the impact on the system state change. The system state scenario behavior dynamic deduction algorithm includes: The assimilated current system state and its accompanying parameters are used as the starting point of the system state and added to the search sequence structure. At the same time, the system initial state probability, search depth, and time step parameters are initialized. Each system state in the search sequence structure is extracted one by one, and the next transition point of the system state is obtained by searching for non-zero elements in the Markov / CCMT system state transition probability mapping matrix model after assimilation and update. The probability of occurrence of each system state transition sequence path is obtained by probability calculation during the search process; and a prior truncation criterion is set in each step of the system state transition process as needed; After each traversal of the system state in the search sequence structure is completed, the system state transition sequence path sets with the same final state are merged, and the merged system state is added to the search sequence structure as a new parent node to enter the next iterative search; For each new parent node, before entering the next system state transition sequence branch search, it is necessary to reconstruct the Markov / CCMT system state transition probability mapping matrix model through Monte Carlo system state grid element representative point random sampling, simulation and statistical analysis to respond to changes in accompanying parameters.

2. The multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method according to claim 1 is characterized in that: The system state scenario behavior dynamic deduction algorithm will aggregate and merge the same system state points after each iterative search step, avoiding the exponential growth of the number of system state sequence branches during each iteration, so that the number of system state points will never exceed the size of the entire system state space, effectively reducing the system state search space and significantly improving the search efficiency.

3. The multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method according to claim 1 is characterized in that: During the dynamic deduction of the system state scenario behavior, the potential transition of the system state after each assimilation is evolved, and the probability of occurrence of different system states after evolution is obtained by summing the cumulative probabilities during each matrix iterative search process, thereby realizing dynamic system reliability prediction analysis, including: The estimated system states obtained from each iterative search are sorted by their probability of occurrence through a list. The larger the probability value, the greater the possibility of the system state occurring. The digitally encoded system state vector after lumped sorting at each search step is converted into a common language description that can be understood by system operators, and the dynamic evolution process of the system state is displayed through the human-computer interface.

4. The method for scenario behavior deduction and reliability prediction analysis of a multi-source data fusion dynamic system according to any one of claims 1 to 3, characterized in that: For complex nonlinear process control systems, it is difficult to obtain an accurate analytical solution for the conditional transition probability matrix. The system state transition probability mapping matrix model adopts an accurate sampling method of the system state sampling point movement distribution based on Monte Carlo simulation, and is obtained through statistical analysis of system state migration trajectory simulation and tracing. The method for accurately sampling the movement distribution of the system state sampling points includes: Combined with the data assimilation process based on particle filtering, the initial state position of the system is determined by monitoring the input or hypothesis of the system state at the initial moment, and Gaussian sampling is performed on the initial state position point of the system to generate a N Particles, the system initial state position point includes its accompanying parameters; Import the system state of the sampled particles and their accompanying parameters into the system simulation model to simulate and predict the particle trajectory changes and the final value distribution of the particles; Reading in the real-time state observation data of the system, and using the data assimilation algorithm based on particle filtering to complete the weight value calculation and resampling of the simulated prediction particles; According to the movement distribution of the particle swarm after resampling, the system state space is reconstructed to more accurately simulate and reflect the dynamic behavior characteristics of the system.

5. The multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method according to claim 1 is characterized in that: The method also includes a self-updating construction process of the Markov / CCMT system state transition probability mapping matrix model, which realizes rapid deduction of system state scenario behavior and dynamic reliability prediction analysis by searching for matching elements in the potential transition mapping relationship submatrix of the current system state and lumping the same system state, including updating the state transition probability matrix of the system physical components and the conditional transition probability matrix, wherein: The updating of the state transition probability matrix of the system physical components is based on the actual state monitoring of the system equipment as input, and the irrelevant system configuration configuration of the state transition probability matrix of the system physical components and the state transition mapping relationship thereof are zeroed out; The update of the conditional transition probability matrix takes the actual state of the system equipment and the state monitoring of the system process variables as input, uniformly samples the particle swarm in the grid element space where the initial state of the system is located after assimilation and update, simulates the trajectory changes of the particle swarm on the system simulation model, and statistically calculates the distribution of the particle final value position in different grid element spaces. The conditional transition probability of the system state under the current system configuration is approximately estimated and updated by the equal-weighted dot product method; While completing the self-updated construction of the system state transition probability mapping matrix for this round, the system state accompanying parameters of different representative points falling into the grid element obtained by random sampling and simulation statistical analysis on the iterative search step of this round are averaged to adapt to the reconstruction of the Markov / CCMT system state transition probability mapping matrix model in the next round of iterative search; Wherein, the equal-weight dot product method is: Where, is the system state conditional transition probability, Configure the configuration for a given system The system state is determined by the gate element Transfer to gate cell j the number of U is the total number of sampled particles.

6. The multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis method according to claim 5 is characterized in that: The Markov / CCMT system state transition probability mapping matrix model Q The dynamic search range is limited to the potential transfer mapping relationship submatrix of the current system state, that is, in the Markov / CCMT system state transition probability mapping matrix model Q During the self-update construction process, only the conditional transition probability of the system state under the current system configuration is considered, and the conditional transition probability between the other system states is automatically assigned to 0, reducing the conditional transition probability matrix G The complexity of the construction is reduced, and only the non-zero elements in the search process are traversed and stored to improve the search speed.

7. A multi-source data fusion dynamic system scenario behavior deduction and reliability prediction analysis system, characterized by: The system comprises: The particle swarm distribution acquisition module is used to obtain the particle swarm distribution of the system assimilation initial state through Gaussian sampling based on the system state spatiotemporal coupling model and the initial value input or assumption of the system state. The module is then imported into the system simulation model to simulate the trajectory changes of the particle swarm. Combined with the real-time monitoring input of the system state, the posterior distribution of the particle swarm of the system state at the current moment is obtained through data assimilation, and used as the particle swarm distribution of the system assimilation initial state at the next moment. The system state space model reconstruction module is used to determine the upper and lower boundary values ​​of the system state originating grid element based on the posterior distribution range of the current system state particle swarm obtained through the data assimilation process, and use this as the new grid element scale to discretize the system state space. Through digital coding, the reconstruction of the system state space model is completed; The Markov / CCMT system state transition probability mapping matrix model generation module is used to uniformly sample the system state originating grid element to obtain the distribution of the initial state particle swarm for system dynamic behavior deduction, import the initial state particle swarm for system dynamic behavior deduction into the system simulation model, simulate the trajectory changes of the obtained particle swarm, and obtain the conditional transition probability matrix of the system state under the current system configuration based on the statistical estimation of the number of particle swarms falling into different grid elements; multiply the conditional transition probability matrix with the state transition probability matrix of the system physical components to obtain the Markov / CCMT system state transition probability mapping matrix model; under the condition that the components are independent of each other, the state transition probability matrix of the system physical components is the product of the failure rates of the components within time Δt; The system state scenario behavior dynamic deduction algorithm and reliability prediction analysis module is used to perform advanced system state prediction analysis based on the constructed Markov / CCMT system state transition probability mapping matrix model. The system state scenario behavior dynamic deduction algorithm is further combined to merge the same system states, and the combined and reduced unique system states are pressed into the search sequence structure. The Markov / CCMT system state transition probability mapping matrix model is repeated to construct an update and iterative search process until the specified search depth is reached. The system analysis result display module is used to sort the probabilities of the merged system states and predict the dynamic evolution of the system state scenario sequence. The system state evolution process is displayed graphically through the interface. When the probability of system state failure exceeds the safety limit, an early warning is triggered to guide the safe operation of the system. wherein, the system configuration structure is configured according to the real-time status monitoring input of the system equipment, and the search sequence structure is constructed according to the initialization and expansion system state after the configuration; The extended system state includes the system state and its accompanying parameters. The accompanying parameter state is the system parameter associated with the system state change. The system state and its accompanying parameters and the state of the system's physical components are coupled with each other, but do not directly affect the dynamic behavior characteristics of the system. They react indirectly through the impact on the system state change. The system state scenario behavior dynamic deduction algorithm includes: The assimilated current system state and its accompanying parameters are used as the starting point of the system state and added to the search sequence structure. At the same time, the system initial state probability, search depth, and time step parameters are initialized. Each system state in the search sequence structure is extracted one by one, and the next transition point of the system state is obtained by searching for non-zero elements in the Markov / CCMT system state transition probability mapping matrix model after assimilation and update. The probability of occurrence of each system state transition sequence path is obtained through probability calculation during the search process; according to needs, a prior truncation criterion is set in each step of the system state transition process; After each traversal of the system state in the search sequence structure is completed, the system state transition sequence path sets with the same final state are merged, and the merged system state is added to the search sequence structure as a new parent node to enter the next iterative search; For each new parent node, before entering the next system state transition sequence branch search, it is necessary to reconstruct the Markov / CCMT system state transition probability mapping matrix model through Monte Carlo system state grid element representative point random sampling, simulation and statistical analysis to respond to changes in accompanying parameters.

Images