Event-driven reservoir dam emergency situation deduction method

Abstract

The invention discloses an event-driven reservoir dam emergency situation deduction method. The method comprises the following steps: constructing a reservoir dam emergency situation ontology model based on an event ontology theory; generating a multi-entity Bayesian network model according to the reservoir dam emergency situation ontology model and the mapping rule of the multi-entity Bayesian fragments; deleting irrelevant branches to construct a specific scene Bayesian network topological structure, selecting a probability assignment method according to a logic relationship between nodes and an action mechanism, initializing probability parameters, and completing construction of an emergency scene deduction model; collecting current scene information as evidence information, completing probability assignment of a root node, taking a driving event as an input variable, and determining a state variable of a subsequent scene node and a probability value of an output variable according to an inference rule; and probability deduction and emergency decision evaluation of the scene situation of the emergency event of the reservoir dam are realized. According to the invention, accurate prediction of the scene situation of the emergency of the reservoir dam is realized.

Description

Technical Field

[0001] The invention relates to a method for deducing a reservoir dam emergency scenario, in particular to an event-driven reservoir dam emergency scenario deduction method, belonging to the technical field of dam risk management of water conservancy engineering technology. Background Art

[0002] Reservoir dams play a variety of functions and benefits, such as flood control, power generation, navigation and irrigation. They are the basic guarantee for economic and social development and the implementation of major national strategies, and are also an important carrier for the construction of ecological civilization. At the same time, during the operation and service of reservoir dams, they are affected by uncertain disaster factors such as congenital defects, improper management, extreme climate, geological disasters, etc., and the "prediction-response" emergency management model has insufficient emergency response technology capabilities. Reservoir dams and even dam breaches and other emergencies have long existed, seriously threatening the safety of life, property and infrastructure downstream. With global climate change and the frequent occurrence of extreme natural disasters, the importance of emergency management of reservoir dam emergencies has been further highlighted.

[0003] The emergency management of reservoir dam emergencies is the last line of defense to ensure the safety of life and property downstream, and is the key work of dam risk management. At present, a relatively comprehensive "one plan and three systems" emergency management system has been initially established, namely, emergency plan, emergency mechanism, emergency system and emergency legal system. The emergency management system is based on the static emergency plan, selects emergencies with a high probability of occurrence for forecast analysis and plans corresponding emergency management measures, clarifies the composition of the emergency organization system and the responsibilities and tasks of each unit, formulates emergency plans in advance, and activates emergency plans to respond after the emergency occurs. It belongs to the "prediction-response" emergency management model. As an action guide for guiding reservoir operating units, competent units and downstream public to respond to emergencies, the emergency plan still has many shortcomings based on its management and application. First, the emergency plan is stored in text form, which is not conducive to the reading and communication of rescue personnel. When an emergency occurs, the command department needs to check the emergency details according to the level of danger before issuing specific action instructions to the participants, which affects the timeliness of emergency management. Second, reservoir dam emergencies are complex nonlinear systems involving heaven, earth, and people. The interaction between the evolution of events and emergency measures has significant uncertainty, and the emergency plan cannot quickly and efficiently adjust the emergency plan for uncertain factors in dynamic scenarios. These shortcomings make the "prediction-response" emergency management model not highly intelligent, the emergency plan has poor adaptability, and the efficiency of emergency decision-making is low, which restricts the improvement of emergency management capabilities and levels. Reservoir dam emergencies have the characteristics of evolution uncertainty, time urgency, and element complexity. The emergency department needs to make decisions and judgments based on limited scenario information in a very short time, formulate scientific and reasonable emergency plans at the fastest speed, carry out emergency rescue, personnel transfer and rescue, resource dispatch, etc., and dynamically adjust emergency measures according to the development of the situation to control the spread of danger and reduce disaster losses to the greatest extent. However, the traditional "prediction-response" emergency management model is difficult to meet the scientific, dynamic, and effective emergency requirements. Summary of the invention

[0004] Purpose of the invention: In view of the problems and shortcomings in the prior art, the present invention provides an event-driven emergency scenario simulation method for reservoir dam emergencies, so as to achieve quantitative evaluation of disaster-causing factors and emergency measures driving effects in emergency management, and provide a reliable decision-making basis for emergency management.

[0005] Technical solution: An event-driven emergency scenario simulation method for reservoir dam emergencies, comprising the following steps:

[0006] Step 1: Constructing the network structure of reservoir dam emergency scenario

[0007] Through the analysis of reservoir dam emergency case data, it is found that the reservoir dam emergency scenario S consists of four elements: event chain, state elements, disaster factors and emergency activities, and the reservoir dam emergency network structure of disaster factor induction-emergency activity intervention-state element mapping-event chain development is determined. The methods used in the reservoir dam emergency case data analysis include text extraction and machine learning.

[0008] Step 2: Constructing a reservoir dam emergency scenario ontology model

[0009] Based on the network structure of reservoir dam emergency scenario, the event ontology representation model composed of five elements is used to formally express the concept of reservoir dam emergency scenario knowledge. According to the semantic structure of event ontology, the logical relationship between elements is defined, and a data storage framework containing semantic relationships in the reservoir dam emergency scenario field is constructed to construct the reservoir dam emergency scenario ontology model.

[0010] Step 3: Generate a multi-entity Bayesian network model

[0011] According to the mapping rules between the reservoir dam emergency scenario ontology model and the multi-entity Bayesian fragment, the mutual mapping of concepts, attributes and relationships in the reservoir dam emergency scenario ontology model is realized, the key elements in the reservoir dam emergency scenario ontology model are extracted as network event nodes, and the state elements that affect the evolution of the event are selected as node variable parameters to generate a multi-entity Bayesian network model;

[0012] Step 4: Build an emergency scenario simulation model

[0013] For the multi-entity Bayesian network model, the model structure and scale are dynamically changed according to the actual emergency scenario, and irrelevant branches such as boundary events, conditional independent events and irrelevant events are deleted to construct a specific scenario Bayesian network topology structure. The probability assignment method is selected according to the logical relationship and action mechanism between nodes, and the probability parameters are initialized to complete the construction of the emergency scenario simulation model.

[0014] Step 5: Event-driven emergency scenario probability simulation

[0015] The current scenario information is collected as evidence information and the probability assignment of the root node is completed. The driving event is used as the input variable, and the probability values ​​of the state variables and output variables of the subsequent scenario nodes are determined according to the inference rules to realize the probability deduction of the emergency scenario situation of reservoir dam emergencies.

[0016] Step 6: Emergency decision-making based on scenario simulation

[0017] Based on the results of scenario simulation, an emergency decision-making model consisting of decision objective function, boundary conditions and constraints is constructed. The response measures are evaluated according to the probability of key scenario nodes and disaster consequences during the development of the emergency to determine the optimal emergency management measures.

[0018] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the event-driven emergency scenario simulation method for sudden incidents of reservoir dams as described above are implemented.

[0019] A computer-readable storage medium stores a computer program for executing the event-driven emergency scenario simulation method for sudden incidents of reservoir dams as described above.

[0020] Beneficial effects: Compared with the prior art, the event-driven emergency scenario simulation method for reservoir dam emergencies provided by the present invention adopts modular entity fragments to characterize the uncertainty of emergency scenarios of emergencies, and quantitatively evaluates the driving effects of disaster-causing factors and emergency measures of reservoir dam emergencies based on scenario situation reasoning rules, which solves the problem that the feedback between scenario situation and disposal countermeasures is difficult to characterize, realizes the accurate prediction of scenario situation of reservoir dam emergencies, and determines emergency rescue measures according to the scenario simulation results, providing a reliable decision-making basis for emergency management. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of a network structure of a reservoir dam emergency scenario in an embodiment of the present invention;

[0022] Figure 2 Schematic diagram of entity fragment expression of flood overtopping event in an embodiment of the present invention;

[0023] Figure 3 It is the overtopping and dam-breaking event chain in the scenario ontology model in the embodiment of the present invention;

[0024] Figure 4 Schematic diagram of a multi-entity Bayesian network model for rainstorm and flood emergencies in an embodiment of the present invention;

[0025] Figure 5 is a typical cross-sectional view of a slope in an embodiment of the present invention;

[0026] Figure 6 yes Figure 5 Schematic diagram of the model unit grid;

[0027] Figure 7 It is a schematic diagram of a rainstorm and flood emergency scenario simulation model for a PB hydropower station in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The present invention is further explained below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0029] An event-driven emergency scenario simulation method for reservoir dam emergencies includes the following steps:

[0030] Step 1: Constructing the network structure of reservoir dam emergency scenario

[0031] By analyzing the case data of reservoir dam emergencies through text extraction or machine learning methods, it is found that the reservoir dam emergency scenario S consists of four elements: event chain, state elements, disaster factors and emergency activities. The scenario network structure and the logical relationship between the elements are shown in Figure 1 The network hierarchy consisting of event chain-state elements-driving elements was clarified, and the scenario evolution mechanism of reservoir dam emergencies was depicted, including disaster factor induction-emergency activity intervention-state element mapping-event chain development.

[0032]

[0033] In the formula, S i It is a fragment of the reservoir dam emergency scenario i; It is the set of event chain elements in scenario segment i, which refers to the scenario network composed of the events in the scenario segment, such as the linear event chain composed of super-standard flood, gate failure, reservoir water level rise, flood overtopping and other elements; It is the set of state elements in scenario segment i. The causal relationship between events is the external representation of the interaction between scenario state elements, that is, the essence of scenario evolution is the internal state transition, including the reservoir water level change speed, reservoir water level elevation, etc. It is the set of emergency activity elements in scenario segment i, which plays a leading role in the development of events and scenario transitions, including emergency measures such as launching emergency plans, emergency rescue measures, and evacuating downstream; It is a collection of disaster-causing factor elements in scenario segment i, which refers to the abnormal factors in the disaster-prone environment that may aggravate the hidden dangers of dam failure or even cause dam failure.

[0034] Step 2: Constructing a reservoir dam emergency scenario ontology model

[0035] 2.1 Based on the network structure of reservoir dam emergency event scenarios, an event ontology representation model consisting of five elements is used to formally express the concept of reservoir dam emergency event scenario knowledge.

[0036] e=<A,O,T,P,S>

[0037] In the formula, A is the action element, O is the object set, T is the time when the event occurs, P is the environment in which the event occurs, and S is the key attributes and characteristics of the event.

[0038] 2.2 Define the logical relationship between elements based on the semantic structure of the event ontology, build a data storage framework that includes the semantic relationships in the reservoir dam emergency scenario field, and construct a reservoir dam emergency scenario ontology model.

[0039] Step 3: Generate a multi-entity Bayesian network model

[0040] According to the mapping rules of the reservoir dam emergency scenario ontology model and the multi-entity Bayesian fragment, the mutual mapping of concepts, attributes and relationships in the model is realized. The knowledge concepts and state parameters in the reservoir dam emergency scenario ontology are mapped into entity fragments and nodes within the fragments, and the logical relationships between the elements are mapped into the topological structure of the multi-entity Bayesian network model, so as to generate the multi-entity Bayesian network model.

[0041] The mapping rules are as follows:

[0042] According to the scenario ontology model structure and multi-entity fragment expression, the mapping rules between the reservoir dam emergency scenario ontology model and the multi-entity Bayesian fragment are proposed, and the conversion channel between the two models is constructed to achieve the mutual mapping of concepts, attributes and relationships between the models. The mapping rules are as follows: Figure 2 shown.

[0043] (1) Mapping of situational concepts to multi-entity fragments

[0044] The first task to realize the mapping of reservoir dam emergency scenario ontology model to multi-entity Bayesian network model is to complete the association of scenario element concepts to entity fragments. Entity fragments are composed of key scenario nodes of reservoir dam emergency, while the scenario ontology model is constructed based on scenario element concepts and their relationships, and there is consistency between the two models. Therefore, the relevant scenario element concepts defined in the emergency ontology are mapped to entity fragments, and the scenario concepts in the scenario ontology are used to name the entity fragments. First, the top-level event of the reservoir dam emergency is determined, and the relevant scenario concepts are obtained from the scenario ontology model. Then, all scenario elements in the reservoir dam emergency scenario ontology model are traversed, and the scenario element concepts with mutual relationships are mapped to entity fragments in the multi-entity Bayesian network.

[0045] (2) Mapping of scenario element state parameters to nodes within entity fragments

[0046] In the reservoir dam emergency scenario ontology model, state parameters are used to describe the characteristics, properties and values ​​of scenario elements. In the multi-entity Bayesian network model, each entity fragment has three types of random variable nodes: context nodes, input nodes and inherent nodes, to characterize the activation conditions, states and characteristics of the entity fragment. Therefore, the node set in the multi-entity Bayesian fragment can be generated by mapping the state parameters of the scenario elements of the reservoir dam emergency scenario ontology model. The context node is used as the judgment condition for the activation of the entity fragment. The preceding event of the entity fragment event in the scenario ontology is used as the context node and represented by a Boolean variable to ensure that the entity fragment can be activated when the scenario is deduced to this node. The inherent node is a random variable defined in the entity fragment, and the state parameters of the scenario elements corresponding to the entity fragment can be directly mapped to the inherent node. In order to facilitate calculation, the state parameters are discretized. The input node, as the parent node of the inherent node, comes from the inherent node of the preceding fragment, so the state parameters of the preceding event in the reservoir dam emergency scenario ontology model are mapped to the input node.

[0047] (3) Mapping of logical relationships between scenario elements to topological structures

[0048] In the reservoir dam emergency scenario ontology, the logical relationships between scenario elements include "following relationship", "causal relationship", "exclusive relationship" and "concurrent relationship". Among them, "following relationship" and "causal relationship" can be regarded as hierarchical relationship, and the parent node can be sorted out. Such relationship between scenarios can be connected with the upper and lower levels in the Bayesian network using directed edges to construct the node logical relationship. However, "exclusive relationship" and "concurrent relationship" are regarded as non-hierarchical relationships, and such relationship between scenarios has no direct effect on the probability deduction of emergency scenarios. Therefore, only the hierarchical relationship between scenario elements is considered in the construction of the topological network to simplify the mapping rules of the topological structure. The topological structure is the key part of the multi-entity Bayesian network model, which includes two parts: directed edges between entity fragments and directed acyclic graphs of nodes within entity fragments. According to the logical relationship between scenario elements in the reservoir dam emergency scenario ontology model, the action relationship between entity fragments in the multi-entity Bayesian network is determined and represented by directed edges; the relationship between the state parameters of the scenario elements is mapped to the relationship between the nodes within the entity fragment, so as to construct the topological structure of the nodes within the entity fragment.

[0049] Step 4: Construct a specific scenario Bayesian network model, namely the emergency scenario deduction model

[0050] The model structure and scale are dynamically changed according to the actual emergency scenario, and irrelevant branches such as boundary events, conditional independent events and irrelevant events of the multi-entity Bayesian network model are deleted to construct a specific scenario Bayesian network topology structure. The probability assignment method is selected according to the logical relationship and action mechanism between nodes, and the probability parameters are initialized to complete the construction of the emergency scenario simulation model. The specific implementation steps are as follows:

[0051] Step 41, dynamically changing the model structure and scale according to the actual emergency scenario, deleting irrelevant branches such as boundary events, conditional independent events and irrelevant events of the multi-entity Bayesian network model to construct a specific scenario Bayesian network topology structure.

[0052] Step 42, selecting a probability assignment method according to the logical relationship and action mechanism between nodes.

[0053] The assignment methods are:

[0054] (1) Expert experience assignment method based on DS evidence theory

[0055] Experts from different fields are invited to judge the probability of scenario state nodes based on the probability estimation table from their respective professional perspectives. Experts from different professions have different knowledge fields, and the evaluation results of each expert are not uniform. DS evidence theory has been widely used in expert systems, information fusion, multi-attribute decision-making, etc., and has good objectivity and multi-source information comprehensive fusion capabilities. Therefore, this embodiment uses DS evidence theory to integrate the opinions of different experts to increase the objectivity and accuracy of probability assignment.

[0056] (2) Parameter learning method based on sample data

[0057] The sample data analysis method uses historical case data as the sample library to estimate the conditional probability distribution parameters. However, the consequences of reservoir dam emergencies are serious, and the information retained by historical cases is extremely limited, resulting in a small amount of sample data and some missing data. Therefore, this embodiment selects the expected maximum (EM) algorithm suitable for data missing samples. The EM algorithm is an iterative optimization algorithm that estimates missing data variables by alternating expectation steps and maximization steps to construct a complete training sample set. Assume that the reservoir dam emergency data sample set is X = {X1, X2,…, X n}, the missing data set is Y={Y1,Y2,…,Y n}, the parameter to be estimated is θ, and P(X,Y|θ) is the joint probability distribution of X and Y after θ is determined.

[0058] (3) Structural reliability analysis method

[0059] For scenario nodes with clear logical relationships and clear action mechanisms in emergencies, the structural reliability analysis method is used to determine the node probability parameters. According to the logical relationship and action mechanism between events, the uncertainty influencing factors are sorted out, and the functional function of scenario events is established through finite element calculation, response surface equation and other methods:

[0060] Z=g(X1,X2,...,X n )

[0061] Where Z is the function function of a certain situation event; X1, X2, ..., X n is a random variable that affects the state of scenario events. Structural reliability refers to the probability of achieving a given function under specified conditions, and the functional function describes the functional relationship between the structural state and the random variable. When the functional function Z is less than 0, it means that the structure is in a failure state, that is, P(Z<0) is the failure probability of the structure. Therefore, by solving the functional function through the JC method, Monte Carlo method, first-order second moment method, etc., the scenario node probability parameters can be determined.

[0062] Step 43, according to the assignment result, the probability parameters are initialized to complete the construction of the emergency scenario simulation model.

[0063] Step 5: Event-driven emergency scenario probability simulation

[0064] Collect the current scenario information as evidence information and complete the probability assignment of the root node, take the driving event as the input variable, determine the probability value of the state variable and output variable of the subsequent scenario node according to the inference rules, and realize the probability deduction of the emergency scenario situation of the reservoir dam emergency. The specific deduction steps are as follows:

[0065] Step 51, when the initial scenario of the emergency occurs, collect the current scenario information as evidence information and complete the probability assignment of the root node, then predict and reason the emergency situation according to the inference rules, determine the probability values ​​of the state variables and output variables of the subsequent scenario nodes, and reason and calculate layer by layer until the terminal scenario node, so as to determine the state probability of each event on the evolution path. Among them, the driving event node in the emergency scenario deduction model is the input variable, the disaster event node is in an uncertain state, and the emergency response event node with a variable type of Boolean variable is in an unresponsive state.

[0066] Step 52: As the real scenario evolves, new disaster events or emergency decision responses continue to occur, which in turn affects the state probability of scenario events. When a driving event occurs, the corresponding event node becomes an evidence node, that is, it becomes a deterministic state. Then, the scenario node probability in the model is updated according to the inference rules, thereby driving the dynamic probability deduction of the emergency scenario.

[0067] In this step, as the scenario develops, the parameters and state values ​​of the deduction model are constantly changing. When there is a new disaster-causing event or an emergency decision is made, the node state in the model will change from the original uncertain node to a deterministic node. For example, the original state of building a sub-dam has two states, yes and no. When the measures of building a sub-dam are taken, the node becomes a deterministic node with a state value of yes and a probability of 100%.

[0068] The rules of situational reasoning are as follows:

[0069] The scenario deduction model expresses the relationship between sub-events through the Bayesian network. Assume that the sub-event EN i and subevent EN j There is a correlation relationship, let I(i j1 ,i j2 ,…,i jn ) is EN j The input variable, S(s j1 ,s j2 ,…,s jm ) is EN j The state variable, O(o j1 ,o j2 ,…,o jt ) is EN j The inference rules of the scenario deduction model are as follows:

[0070] (1) Input variables and state variables

[0071] P(s je )=∑P(s je |i j1 ,i j2 ,...,i jn )P(i j1 ,i j2 ,...,i jn ) (5-1)

[0072] P(s je )=∑P(s je |i j1 ,i j2 ,...,i jn ,s i1 ,s i2 ,...,s im )P(i ji1 ,i j2 ,...,i jn ,s i1 ,s i2 ,...,s im ) (5-2)

[0073] Formula (5-1) indicates that the state variable is only related to EN j The input variable is related to EN. Formula (5-2) shows that the state variable is related to EN j Input variables and EN i The state variables of je ) is the state variable s je The probability of occurrence; n, m are I j and S i The number of variables.

[0074] (2) State variables and state variables

[0075] P(s je )=∑P(s je |s j1 ,s j2 ,...,s jm )P(s j1 ,s j2 ,...,s jm ) (5-3)

[0076] Formula (5-3) indicates that the state variable is only related to EN j Other state variables are related. In the formula, P(s je ) is the state variable s je The probability of occurrence; m is S j The number of variables.

[0077] (3) Output variables and state variables

[0078] P(o je )=∑P(o je |s j1 ,s j2 ,...,s jm )P(s j1 ,s j2 ,...,s jm ) (5-4)

[0079] Formula (5-4) shows that the output variable is related to EN j State variables are related. In the formula, P(o je ) is the output variable o je The probability of occurrence; m is S j The number of variables.

[0080] (4) Relationship between variables among events

[0081]

[0082] Formula (5-5) represents EN i The output variable I j and EN j The input variable O i The intersection is not an empty set. j and O i The number of variables.

[0083] Step 6: Build an emergency decision-making model based on scenario simulation

[0084] According to the probability of key scenario nodes and disaster consequences during the development of emergencies, response measures are evaluated to determine the optimal emergency management measures. The emergency decision-making model based on scenario simulation consists of decision objective function, boundary conditions and constraints. The expected loss of life in the downstream risk area is used as the objective function of the emergency decision-making model; the boundary condition is an external factor that controls the decision-making target, which is the downstream risk map element, including the scope of the downstream risk area, the distribution of the risk population and the traffic network; the constraint condition is the key scenario variable that affects the disposal effect, including emergency rescue measures, emergency relief measures, etc.

[0085] A "scenario-response" emergency decision-making model was constructed with the expected downstream life and economic losses as the objective function:

[0086]

[0087] Constraints:

[0088]

[0089]

[0090] Where Z1 is the expected loss of life and economy downstream; P i is the probability of the occurrence of the i-th scenario; L i is the number of lives lost due to the i-th scenario; F1 represents the calculation function of scenario probability; D1 represents the calculation function of life loss; x i is a constraint condition, which represents the random variable of the scenario factors that affect the treatment effect in the i-th scenario; c i is the boundary condition, which represents the flood risk index affecting life in the i-th scenario.

[0091] 6.1 Determine flood risk indicators based on scenario simulation, conduct flood risk analysis, and then estimate disaster losses

[0092] 6.2 Input the scenario probability simulation results and disaster loss estimation results into the emergency decision-making model, determine the emergency rescue measures and generate specific emergency plans.

[0093] On the one hand, the emergency scenario deduction model realizes the probability prediction of scenario development trends and grasps the impact of driving events on scenario evolution in real time; on the other hand, it quantitatively analyzes the effects of emergency measures by inferring the probability states of scenarios driven by different emergency response measures. By coupling the emergency scenario probability deduction and the emergency decision-making model of disaster loss consequences, more scientific and effective emergency measures can be selected to provide a reliable decision-making basis for emergency decision-making.

[0094] Contents to be supplemented (existing technology)

[0095] Multi-Entity Bayesian Networks (MEBN)

[0096] The multi-entity Bayesian network is an extension of the ordinary Bayesian network. Its core idea is to use modular entity fragments (MEBN Fragments, MFrags) to describe event attributes and associations. Since the MEBN model expresses the knowledge of the emergency domain based on rule fragments, the multi-entity rules are used to integrate the fragments for logical reasoning, and the network structure is clearer and easier to update. Its essence is a knowledge base expression logic based on rule fragments. The core entity fragment consists of five tuples:

[0097] F=<C,I,R,G,D>

[0098] Where C is the context node, which is a logical (true or false) random variable, indicating the intrinsic node enabling condition defined in the entity fragment; R is the intrinsic node, which is a random variable defined in the entity fragment; I is the input node, which is referenced to the intrinsic node of other entity fragments; G is the topological structure composed of I and R nodes in the fragment; and D is the probability distribution between nodes.

[0099] Figure 2 The figure is a schematic diagram of the entity fragment expression of flood overtopping events. This entity fragment describes the probabilistic influence relationship between rainstorm flood events and reservoir overtopping events. The rounded rectangular box at the top of the entity fragment is the context node, which clarifies the activation condition of the entity fragment, that is, when a flood event occurs, the overtopping event entity fragment will be activated. Among them, Flood_Type (flood type) represents the input node I; Overtopping (overtopping dam breach) represents the inherent node R; the topological structure of the event composition is a directed acyclic graph G, and the arrows represent the logical relationship between events, indicating that the possibility of flood overtopping events is related to the flood magnitude; the table is the probability distribution relationship D between events.

[0100] DS Evidence Theory

[0101] Assume that a scenario node in the Bayesian network has (S1, S2, ...S m )m states, n experts E were invited n (n=1,2,3…) for evaluation. The expert decision results are shown in Table 1.

[0102] Table 1 Expert evaluation results

[0103]

[0104] Based on DS evidence theory, the node state is used as the identification framework, P(A=S) is used as the mass function, and the synthesis rule is used to comprehensively integrate the expert decision opinions. The fusion formula is as follows:

[0105]

[0106] Where P j (A=S i ) is A=S given by experts i The probability of K being the normalization factor is used to measure the degree of conflict between different evidences. The smaller the value, the more contradictory the evidence is.

[0107] Example:

[0108] In order to verify the scientificity and applicability of the reservoir dam emergency scenario simulation method, the PB hydropower station project encountered a rainstorm and flood emergency event as an example for application.

[0109] 1. Use machine learning to extract and identify professional terms for reservoir dam emergencies from relevant specifications, standards and works such as "Guidelines for Construction of Production Safety Accident Scenarios", "Construction of Major Emergency Scenarios - Theory and Practice", and "Guidelines for Preparation of Reservoir Dam Safety Management Emergency Plans", and complete the construction of the reservoir dam emergency scenario ontology model based on the reservoir dam emergency scenario network structure;

[0110] 2. Construct a multi-entity Bayesian network model for rainstorm and flood emergencies in PB hydropower stations.

[0111] 2.1 Extract the knowledge network of the rainstorm and flood event chain in the reservoir dam emergency event ontology model, see Figure 3 .

[0112] 2.2 Mapping and generation of PB rainstorm and flood emergency multi-entity Bayesian network model.

[0113] Based on the overtopping and dam-breaking event chain in the reservoir dam emergency scenario ontology model, a multi-entity Bayesian network model of PB hydropower station rainstorm and flood emergency events is constructed according to the scenario ontology-multi-entity Bayesian fragment mapping rules. For example, extreme rainfall is mapped into a rainstorm and flood fragment (scenario element concept-multi-entity fragment), the causal relationship between the reservoir bank landslide event and the surge is mapped into the parent node relationship between the reservoir bank landslide fragment near the dam and the surge fragment in front of the dam, and the logical relationship between the state elements of the reservoir bank landslide event and the surge event is mapped into the landslide volume volume and sliding surface nodes of the reservoir bank landslide fragment as the input nodes of the surge height node in the surge fragment in front of the dam. Figure 4 As shown in the figure, the multi-entity Bayesian network model generated by mapping consists of 11 segments, including gate failure, landslide near the dam bank, spillway structure damage, dam surge, and dam overtopping.

[0114] 3. Constructing a Scenario-Specific Bayesian Network Model

[0115] The multi-entity Bayesian network model clarifies the Bayesian network topology of the PB hydropower station rainstorm flood emergency event and gives the causal relationship between the nodes in the entity segment. In order to realize the probability deduction of the scenario model, it is also necessary to assign values ​​to the node probability parameters, clarify the element node status and threshold interval, and determine the probability distribution table, including the root node probability and the conditional probability of the interaction between nodes. The node probability distribution in the entity segment needs to be unique to meet the consistency constraint of the Bayesian network.

[0116] Define the local probability distribution of node variables in the multi-real Bayesian network model according to the analysis method given in step 4.

[0117] The following is an introduction using the landslide segment near the dam bank and the gate failure segment as examples.

[0118] Landslide fragment near the dam bank

[0119] The risk of landslide and dam overflow caused by the instability of the bank slope near the dam is common in dam overflow accidents of water conservancy and hydropower projects, and dam failures caused by this also occur from time to time. In southwest my country, most of the existing reservoir dam projects are located in high mountain valleys. The terrain and geological conditions in the reservoir area are complex, the bank slope is steep, and there are many potential dangerous landslides, which have become one of the important risk sources threatening the safety of dam operation. The event of landslide and dam overflow near the dam bank includes two nodes: the instability of the bank slope and the landslide surge after the instability.

[0120] (1) Root node of landslide near dam bank

[0121] In the numerical analysis of rock and soil slope stability, the finite element method is used to analyze the risk of landslides near the dam bank. A finite element model is constructed based on the typical section of the tensile deformation body of the right bank slope of the PB hydropower station. The bottom elevation of the model is 600m. Figure 5 and 6 As shown, there are 4071 units in total. Horizontal constraints and fixed constraints are applied to the left, right and bottom boundaries of the model, and there is no constraint on the top.

[0122] For slope reliability, the shear strength parameters of rock mass have the most significant impact on slope stability. The rock mass cohesion and internal friction angle are regarded as random variables obeying the normal distribution, and the physical and mechanical parameters of various rock masses of the tensile deformation body are determined.

[0123] The strength reduction method is used to search for potential failure channels of the landslide body. The sliding arc is located at the interface between the loose deformation body and the lower layer of strong unloading rock mass, and part of the sliding arc passes through the inside of the loose deformation body. The stability function of the failure channel is constructed based on the response surface method, and its construction form is as follows:

[0124]

[0125]

[0126]

[0127] In the formula, G L (X) is the instability failure function; n is the total number of elements on the slip surface; b i is the length of element i along the slip surface direction; is the normal stress of element i on the slip surface; is the shear stress of element i on the slip surface; f i ′ is the friction coefficient of element i on the slip surface; c i is the cohesion of element i on the slip surface; X is the random variable x 1 , x 2 , ···, x n ; a0, b i and c i are the response surface equation parameters.

[0128] According to the distribution characteristics of random parameters, a total of 25 groups of schemes were designed using the orthogonal test method. The undetermined coefficients of the response surface equation were determined through the finite element simulation results and statistical analysis. The multiple correlation coefficient R 2 of the response surface equation reached 0.98, indicating a good fitting effect. Based on the slope instability function and the response surface equation, Monte-Carlo simulation was carried out on the most dangerous slip surface. The sampling results showed that the slope instability probability was 3.37×10 -4 . Thus, the state probability of the root node of the landslide of the reservoir bank slope was determined as P(Slope = YES) = 3.37×10 -4 , P(Slope = NO) = 0.999663.

[0129] (2) Node of the surge height in front of the dam

[0130] The Pan's algorithm was used to calculate the landslide surge. It was assumed that the river channel water body was a semi-infinite long water body with parallel sides and a width of B. The cross-section of the reservoir bank within the range L where the landslide body was located was consistent. The formula for the surge height ξ (h″ w ) at the dam site occurring within the time period 0 < t < T was:

[0131]

[0132] In the formula, k r is the reflection coefficient of the wave propagating to the front of the dam; ξ0 is the initial wave height when the sliding body enters the water, which is directly affected by the average water depth of the reservoir , and can be expressed as ξ0(h″ w ) and calculated according to formulas (3-5) to (3-7); x0 is the distance from the farther end of the landslide body to the dam site; L is the width of the landslide body along the reservoir bank; k is the reflection coefficient of the wave propagating to the opposite bank; θ n ​​is the angle between the nth incident ray of the wave and the normal line of the bank slope; n is the number of waves superimposed when the maximum wave height in front of the dam is generated, which is determined by the ratio of the landslide duration to the time required for the wave to propagate to the other side, T / Δt, as shown in Table 2, where Δt = B / v c , v c is the propagation wave velocity of surge waves in water, calculated according to formula (3-7).

[0133] Table 2 Correspondence between landslide duration and wave superposition quantity

[0134]

[0135] ξ0(h″ w )=ξ h (h″ w )cos 2 α+ξ v (h″ w )sin 2 α (3-5)

[0136]

[0137]

[0138] Where λ is the average thickness of the landslide block; ξ h (h″ w ), v (h″ w ) is divided into horizontal and vertical velocities of the landslide body (v h 、v v ) and follows the following formula:

[0139] v h =v·cosα;v v =v·sinα (3-8)

[0140]

[0141]

[0142]

[0143] Where a is the overall slope angle of the landslide; W i 、a i is the weight of the ith slider and its average slope angle; c and f are the cohesion and friction coefficient of the landslide; a 0i is the slope angle of the line connecting the midpoints of the i-1th and i-th sliders; W is the total weight of the landslide.

[0144] The most dangerous sliding surface determined by finite element calculation is used as the sliding surface. The range of collapse and sliding is based on the boundary of the tensile deformation body on the plane geological map, which is about 360m wide. The riverbed is 80m wide and the slope angle of the opposite bank is 40°. Since the landslide body is located above the water surface, it is assumed that during the sliding of the slope, only friction f exists on the contact surface, and there is no pore water pressure. The dynamic strength parameters are used in the calculation. Since the friction coefficient of the landslide should be the dynamic friction coefficient after the start of the landslide, according to the review report on the slope stability research, the internal friction angle is determined to be 20.96° and the cohesion is 0.11Mpa. The landslide body is divided into 18 strips in the horizontal direction, each with a width of △L of 10m. Since the landslide body is located above the water surface, it is assumed that during the sliding of the slope, only friction f exists on the contact surface, and there is no pore water pressure. According to formulas (3-5) to (3-11), it can be obtained that the velocity reaches the maximum value v at t = 55.73s max =5.96m / s. Taking the normal water level of 850.00m and the dam top elevation of 856.00m as an example to calculate the surge height, the corresponding surge height in front of the dam is 5.92m and 6.14m.

[0145] According to the entity fragment, the factors affecting the surge height in front of the dam include the volume of the landslide body, the sliding surface, the reservoir water level and the location of the landslide. The sliding surface, the volume of the landslide body and the location parameters of the tensile body were determined by the finite element analysis of slope stability. Therefore, the uncertainty factors of the surge height event in front of the dam are the landslide event and the reservoir water level. The reservoir water level state is a continuous variable. In order to facilitate the application of the inference rules of the Bayesian network, the reservoir water level variable is discretized. According to the characteristic value of the reservoir water level, it is discretized into five state values, namely (790,850], (850,853.78], (853.78,855], (855,856], (856,∞]. Since the rupture body is located above the reservoir water level, the two are regarded as independent events. According to formula (3-4), the functional relationship between the surge height and the reservoir water level is established. Therefore, according to the reservoir water level state, the surge height is also discretized into five state values, namely 0, (0,5.92], (5.92,6.06], (6.06,6.10], (6.10,6.14]. According to the logical relationship between the scenario elements, the surge height state probability calculation formula is given as follows:

[0146]

[0147] P(H i )=P(S YES )·P(L i-1 ) i=2,3,4 (3-13)

[0148] P(H5)=P(S YES )·(P(L4)+P(L5)) (3-14)

[0149] Gate failure segment 2

[0150] Discharge capacity is one of the core indicators to ensure the safety of dams during flood season. Under deterministic parameter conditions, the dam discharge can be accurately calculated through hydraulic calculations, but sudden disasters such as gate failure and spillway structure damage have caused uncertainty in the actual discharge capacity of the dam. There are many factors that affect the normal opening and closing of the gate, including gate strength and stiffness, stability, backup power supply, management system and regulations, structural aging, etc., and the relationship between the influencing factors is complex, making it difficult to accurately evaluate the failure probability of the gate through numerical models and logical reasoning. Therefore, the expert experience assignment method based on DS evidence theory is used to evaluate the gate failure probability.

[0151] According to the survey and statistical results of the gate opening and closing conditions when floods occur, as well as the evaluation basis for gate failure events in relevant literature, a single gate failure probability estimation table is given based on the principles of operability and quantifiability as a basis for experts to judge the gate failure probability, see Table 3.

[0152] Table 3 Single gate failure probability estimation table

[0153]

[0154] Five senior experts in the field of water conservancy engineering with a background in metal structures and more than ten years of experience in reservoir dam safety appraisal were invited to evaluate the state probability of a single gate based on the gate failure probability evaluation table, with reference to the engineering characteristics, design, construction, and operation management of the PB hydropower station. The expert evaluation results were collected and evaluated based on the evidence theory fusion formula. The state probability of a single gate failure node was finally determined to be P(Gate_failure=YES)=0.27×10 -2 , P(Gate_failure=NO)=0.9973.

[0155] 4. Application of probability simulation of reservoir dam emergency scenarios

[0156] The states and thresholds of the scenario network nodes are shown in Table 4, where event element nodes are described by Boolean variables, such as the node states of gate failure are "occurrence" and "non-occurrence"; state element nodes are determined according to engineering characteristics and probability distribution, such as rainstorm flood nodes are determined according to the flood design standards of PB hydropower stations. After the node states and probability distributions are determined, based on the assumed PB hydropower station scenario information, the corresponding entity fragments are selected from the multi-entity Bayesian network model to synthesize the scenario event node chain, and irrelevant events, isolated events and edge events are removed, forming a specific scenario Bayesian network model extending from the evidence event to the target reasoning event. Figure 7 As shown, this is the PB hydropower station rainstorm and flood emergency scenario simulation model.

[0157] Table 4 Network node status and thresholds for rainstorm and flood emergency scenarios of PB hydropower stations

[0158]

[0159] During the evolution of reservoir dam emergencies, multiple driving factors change dynamically over time. Scenario simulation is essentially a dynamic recursive process. In order to ensure the accuracy of the simulation results, the node status of the scenario model is updated in a timely manner according to the driving factors. In the actual scenario simulation process, the uncertainty of disaster factors and emergency measures will cause different scenario evolution directions. In order to clarify complex problems for model construction and verification, the evolution process of the emergency scenario is simplified, and the evolution path only considers the core event chain and key influencing factors. The simulation process of the initial scenario of heavy rain and flood at PB Hydropower Station and the subsequent encounter with different disaster factors is as follows:

[0160] Scenario 1: During the flood season, the reservoir water level of PB Hydropower Station is at the flood limit water level of 841.00m, and it encounters a once-in-a-millennium rainstorm and flood. No abnormalities are found in other disaster factors. The once-in-a-millennium rainstorm and flood are used as the initial disaster factor to update the state probability of the rainstorm and flood root node in the deduction model. In the deduction results, the probability that the reservoir water level peak is in the L3 (850,853) interval is 97.5%, and the probability of the dam overflowing and breaking is 0.54%. The results show that when there are no other disaster-causing events and encounter a once-in-a-millennium rainstorm and flood event, the safe operation of the hydropower station can be guaranteed by normal flood dispatching, and there is no need to initiate other emergency measures. The verification flood standard of PB Hydropower Station is the possible maximum flood (PMF), and its flood control capacity is higher than the flood standard of once in ten thousand years. In addition, the hydropower station has a sound management and maintenance system, the discharge capacity of the flood discharge structure and the energy dissipation facilities meet the requirements of the specification, the gate system is safe and reliable, and the slope near the dam is reinforced with anchor support, so the probability of disaster-causing events such as gate failure, spillway structure damage and landslide near the dam is low. Combined with the current status of the hydropower station project, in the scenario of once in ten thousand years heavy rain and flood, the model simulation results are consistent with the actual risk status of the hydropower station.

[0161] Scenario 1.1: During the evolution of scenario 1, the slope safety monitoring system sounded an alarm. Under the influence of heavy rainfall, the displacement of the slope showed obvious abnormalities, and the risk of landslides increased significantly. In order to evaluate the impact of landslide hazards near the dam bank on the overtopping event chain during the flood period, it was input into the deduction model as a deterministic event. In the deduction results, when the PB hydropower station encountered a once-in-a-millennium rainstorm flood, the probability of the surge height in front of the dam caused by the landslide near the dam bank was H4 (6.06, 6.10] was 78.4%, and the probability of the overtopping and dam failure of the barrage was 61.5%. Through the flood evolution analysis, it was determined that the loss of life under this condition was 1,453. Considering the emergency measure of building a sub-dam on the dam top and inputting it as a deterministic event, the probability of the overtopping event was reduced to 12.4%, and the loss of life under this condition was 163 5 people. The simulation results show that the successful implementation of the emergency measures of building a sub-dam on the dam top can significantly reduce the probability of dam overtopping and dam failure caused by landslides near the dam bank. Once an overtopping and dam failure event occurs, although the construction of a sub-dam on the dam top will increase the peak flow of the dam failure, the expected loss of life is reduced from Z1=894 to Z2=202 due to the significant reduction in the probability of dam failure. Through the scenario simulation model, it can be concluded that the construction of a sub-dam on the dam top can significantly reduce the risk of overtopping and dam failure in the current scenario, providing a quantitative basis for emergency decision-making.

[0162] In summary, the results of the PB hydropower station rainstorm flood emergency scenario simulation are basically consistent with the actual situation of the project. The event-driven emergency scenario simulation model can infer the probability changes of the scenario node status and accurately predict the scenario development trend. The results of the engineering example application are consistent with the actual situation, verifying the scientificity and effectiveness of the method, realizing the quantitative evaluation of the disaster-causing factors and the driving effects of emergency measures, and providing a reliable decision-making basis for emergency management.

[0163] Obviously, those skilled in the art should understand that the various steps of the event-driven emergency scenario simulation method for reservoir dam emergencies of the above-mentioned embodiment of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiment of the present invention is not limited to any specific combination of hardware and software.

Claims

1. An event-driven emergency scenario simulation method for reservoir dam emergencies, characterized in that: The steps include: Step 1: Constructing the network structure of reservoir dam emergency scenario Through the analysis of reservoir dam emergency case data, it is found that the reservoir dam emergency scenario S is composed of four elements: event chain, state elements, disaster factors and emergency activities, and the reservoir dam emergency network structure of disaster factor induction-emergency activity intervention-state element mapping-event chain development is determined; the methods used in the reservoir dam emergency case data analysis include text extraction and machine learning; Step 2: Constructing a reservoir dam emergency scenario ontology model Based on the network structure of reservoir dam emergency scenario, the event ontology representation model composed of five elements is used to formally express the concept of reservoir dam emergency scenario knowledge; the logical relationship between elements is defined according to the semantic structure of event ontology, and a data storage framework containing semantic relationships in the field of reservoir dam emergency scenario is constructed, and the reservoir dam emergency scenario ontology model is constructed based on this data storage framework; Step 3: Generate a multi-entity Bayesian network model According to the mapping rules between the reservoir dam emergency scenario ontology model and the multi-entity Bayesian fragment, the mutual mapping of concepts, attributes and relationships in the reservoir dam emergency scenario ontology model is realized, the key elements in the reservoir dam emergency scenario ontology model are extracted as network event nodes, and the state elements that affect the evolution of the event are selected as node variable parameters to generate a multi-entity Bayesian network model; Step 4: Build an emergency scenario simulation model For the multi-entity Bayesian network model, the model structure and scale are dynamically changed according to the actual emergency scenario, irrelevant branches are deleted to construct the scenario Bayesian network topology structure, and the probability assignment method is selected according to the logical relationship and action mechanism between nodes, the probability parameters are initialized, and the construction of the emergency scenario deduction model is completed; Step 5: Event-driven emergency scenario probability simulation Collect current scenario information as evidence information and complete the probability assignment of the root node, use the driving event as the input variable of the emergency scenario deduction model, determine the probability value of the state variable and output variable of the subsequent scenario node according to the inference rules, and realize the probability deduction of the emergency scenario situation of reservoir dam emergencies; Step 6: Emergency decision-making based on scenario simulation Based on the results of scenario simulation, an emergency decision-making model consisting of decision objective function, boundary conditions and constraints is constructed; the response measures are evaluated according to the probability of key scenario nodes and disaster consequences during the development of the emergency to determine the optimal emergency management measures.

2. The event-driven emergency scenario simulation method for reservoir dam emergencies according to claim 1 is characterized in that: In the step 1, the network structure of the reservoir dam emergency scenario is: In the formula, S i It is a fragment of the reservoir dam emergency scenario i; is the set of event chain elements in scene segment i, which refers to the scene network composed of the events in the scene segment; is the set of state elements in scene segment i. The causal relationship between events is the external representation of the interaction between the state elements of the scene, that is, the essence of scene evolution is the internal state change; is the set of emergency activity elements in scenario i; It is the set of disaster-causing factor elements in scenario segment i.

3. The event-driven emergency scenario simulation method for reservoir dam emergencies according to claim 1 is characterized in that: Step 2: Constructing the reservoir dam emergency scenario ontology model, including the following steps: 2.1 Based on the network structure of reservoir dam emergency event scenario, the event ontology representation model consisting of five elements is used to formally express the concept of reservoir dam emergency event scenario knowledge: e=<A,O,T,P,S> In the formula, A is the action element, O is the object set, T is the time when the event occurs, P is the environment in which the event occurs, and S is the key attributes and characteristics of the event; 2.2 Define the logical relationship between elements based on the semantic structure of the event ontology, build a data storage framework that includes the semantic relationships in the reservoir dam emergency scenario field, and use this data storage framework to build a reservoir dam emergency scenario ontology model.

4. The event-driven emergency scenario simulation method for reservoir dam emergencies according to claim 1 is characterized in that: In step 3, based on the scenario ontology model structure and multi-entity fragment expressions, the mapping rules between the reservoir dam emergency scenario ontology model and the multi-entity Bayesian fragment were proposed, and the conversion channel between the two models was constructed to achieve the mutual mapping of concepts, attributes and relationships between the models. (1) Mapping of situational concepts to multi-entity fragments The first task of mapping the reservoir dam emergency scenario ontology model to the multi-entity Bayesian network model is to complete the association of scenario element concepts to entity fragments; the entity fragments are composed of reservoir dam emergency scenario nodes, and the scenario ontology model is built based on scenario element concepts and their relationships, and the two models are consistent; therefore, the relevant scenario element concepts defined in the emergency ontology are mapped to entity fragments, and the scenario concepts in the scenario ontology are used to name the entity fragments; first, the top-level event of the reservoir dam emergency is determined, and the relevant scenario concepts are obtained from the scenario ontology model, and then all scenario elements in the reservoir dam emergency scenario ontology model are traversed, and the scenario element concepts with mutual relationships are mapped to entity fragments in the multi-entity Bayesian network; (2) Mapping of scenario element state parameters to nodes within entity fragments In the reservoir dam emergency scenario ontology model, state parameters are used to describe the characteristics, properties and values ​​of scenario elements; in the multi-entity Bayesian network model, each entity fragment has three types of random variable nodes: context nodes, input nodes and inherent nodes, to characterize the enabling conditions, states and characteristics of the entity fragment; Therefore, the node set in the multi-entity Bayesian fragment can be generated by mapping the state parameters of the scenario elements of the reservoir dam emergency scenario ontology model; the context node is used as the judgment condition for the activation of the entity fragment, and the preceding event of the entity fragment event in the scenario ontology is used as the context node, and is represented by a Boolean variable to ensure that the entity fragment is activated when the scenario is deduced to this node; the inherent node is used as a random variable defined in the entity fragment, and the state parameters of the scenario elements corresponding to the entity fragment are mapped to the inherent node, and the state parameters are discretized; the input node is used as the parent node of the inherent node, and all come from the inherent node of the preceding fragment, so the state parameters of the preceding event in the reservoir dam emergency scenario ontology model are mapped to the input node; (3) Mapping of logical relationships between scenario elements to topological structures In the reservoir dam emergency scenario ontology, the logical relationships between scenario elements include "following relationship", "causal relationship", "exclusion relationship" and "concurrency relationship"; among them, "following relationship" and "causal relationship" are regarded as hierarchical relationships, and the parent node is sorted out. The directed edge is used to connect the superior and subordinate nodes in the Bayesian network to construct the node logical relationship; in the construction of the topological network, only the hierarchical relationship between scenario elements is considered to simplify the mapping rules of the topological structure; the topological structure includes two parts: the directed edge between entity fragments and the directed acyclic graph of the nodes in the entity fragment; according to the logical relationship between scenario elements in the reservoir dam emergency scenario ontology model, the action relationship between entity fragments in the multi-entity Bayesian network is determined and represented by directed edges; the relationship between the state parameters of the scenario elements is mapped to the relationship between the nodes in the entity fragment, and the topological structure of the nodes in the entity fragment is constructed.

5. The event-driven emergency scenario simulation method for reservoir dam emergencies according to claim 1 is characterized in that: The specific implementation steps of step 4 are as follows: Step 41, dynamically changing the model structure and scale according to the actual emergency scenario, deleting the irrelevant branches of the boundary events, conditional independent events and irrelevant events of the multi-entity Bayesian network model to construct a specific scenario Bayesian network topology structure; Step 42, selecting a probability assignment method according to the logical relationship and action mechanism between nodes; the assignment method includes an expert experience assignment method based on DS evidence theory, a parameter learning method based on sample data, and a structural reliability analysis method; (1) Expert experience assignment method based on DS evidence theory Experts in different fields judge the probability of scenario state nodes based on the probability estimation table; (2) Parameter learning method based on sample data Select the expected maximum algorithm suitable for data missing samples. The expected maximum algorithm is an iterative optimization algorithm that estimates the missing data variables by alternating the expectation step and the maximization step to construct a complete training sample set. Suppose the reservoir dam emergency event data sample set is X = {X1, X2, …, X n }, the missing data set is Y={Y1,Y2,…,Y n }, the parameter to be estimated is θ, and P(X,Y|θ) is the joint probability distribution of X and Y after θ is determined; (3) Structural reliability analysis method For scenario nodes with clear logical relationships and clear action mechanisms in emergencies, the structural reliability analysis method is used to determine the node probability parameters; according to the logical relationships and action mechanisms between events, the uncertainty influencing factors are sorted out, and the functional functions of scenario events are established through finite element calculation and response surface equation methods: Z=g(X1,X2,…,X n ) Where Z is the function function of a certain situation event; X1, X2, ..., X n is a random variable that affects the state of scenario events; structural reliability refers to the probability of achieving a given function under specified conditions, and the functional function describes the functional relationship between the structural state and the random variable. When the functional function Z is less than 0, it means that the structure is in a failure state, that is, P(Z<0) is the failure probability of the structure; the functional function is solved by the JC method, Monte Carlo method or first-order second moment method to determine the scenario node probability parameter; Step 43, according to the assignment result, the probability parameters are initialized to complete the construction of the emergency scenario simulation model.

6. The event-driven emergency scenario simulation method for reservoir dam emergencies according to claim 1 is characterized in that: In step 5, the current scenario information is collected as evidence information and the probability assignment of the root node is completed. The driving event is used as the input variable, and the probability values ​​of the state variables and output variables of the subsequent scenario nodes are determined according to the inference rules to realize the probability deduction of the emergency scenario situation of the reservoir dam emergency event. The specific deduction steps are as follows: Step 51, when the initial scenario of the emergency occurs, the current scenario information is collected as evidence information and the probability assignment of the root node is completed, and then the emergency situation is predicted and inferred according to the inference rules to determine the probability values ​​of the state variables and output variables of the subsequent scenario nodes, and the inference calculation is performed layer by layer until the terminal scenario node, so as to determine the state probability of each event on the evolution path; wherein, the driving event node in the emergency scenario deduction model is the input variable, wherein the disaster event node is in an uncertain state, and the emergency response event node whose variable type is a Boolean variable is in an unresponsive state; Step 52, as the real scenario evolves and develops, new disaster events or emergency decision responses continue to occur, which in turn affects the state probability of the scenario event; when the driving event occurs, the corresponding event node becomes an evidence node, that is, it becomes a deterministic state; then the scenario node probability in the model is updated according to the inference rules, thereby driving the dynamic probability deduction of the emergency scenario.

7. The event-driven emergency scenario simulation method for reservoir dam emergencies according to claim 6 is characterized in that: The scenario deduction model expresses the correlation between sub-events through Bayesian network; suppose the sub-event EN in the scenario i and subevent EN j There is a correlation relationship, let I(i j1 ,i j2 ,…,i jn ) is EN j The input variable, S(s j1 ,s j2 ,…,s jm ) is EN j The state variable, O(o j1 ,o j2 ,…,o jt ) is EN j The output variables of the scenario deduction model are as follows: (1) Input variables and state variables P(s je )=∑P(s je |i j1 ,i j2 ,...,i jn )P(i j1 ,i j2 ,...,i jn ) (5-1) P(s je )=∑P(s je |i j1 ,i j2 ,...,i jn ,s i1 ,s i2 ,...,s im )P(i ji1 ,i j2 ,...,i jn ,s i1 ,s i2 ,...,s im )(5-2) Formula (5-1) indicates that the state variable is only related to EN j The input variables are related to the state variables. Formula (5-2) shows that the state variables are related to EN j Input variables and EN i The state variables of je ) is the state variable S je The probability of occurrence; n, m are I j and S i The number of variables; (2) State variables and state variables P(s je )=∑P(s je |s j1 ,s j2 ,...,s jm )P(s j1 ,s j2 ,...,s jm ) (5-3) Formula (5-3) indicates that the state variable is only related to EN j Other state variables are related; where P(s je ) is the state variable s je The probability of occurrence; m is S j The number of variables; (3) Output variables and state variables P(o je )=∑P(o je |s j1 ,s j2 ,...,s jm )P(s j1 ,s j2 ,...,s jm ) (5-4) Formula (5-4) represents the relationship between the output variable and EN j State variable related; where P(o je ) is the output variable o je The probability of occurrence; m is S j The number of variables; (4) Relationship between variables between events Formula (5-5) represents EN i The output variable I j and EN j The input variable O i The intersection is not an empty set; where t and n are I j and O i The number of variables.

8. The event-driven emergency scenario simulation method for reservoir dam emergencies according to claim 1 is characterized in that: In the step six, the response measures are evaluated according to the scenario node probabilities and disaster consequences during the development of the emergency event to determine the optimal emergency management measures; the emergency decision-making model based on scenario deduction is composed of a decision objective function, boundary conditions and constraint conditions, wherein the expected loss of life in the downstream risk area is used as the objective function of the emergency decision-making model; the boundary conditions are external factors that control the decision-making objectives, which are the elements of the downstream risk map, including the scope of the downstream risk area, the distribution of the risk population and the traffic network information; the constraint conditions are key scenario variables that affect the disposal effect, including emergency rescue measures and emergency relief measures; A "scenario-response" emergency decision-making model was constructed with the expected downstream life and economic losses as the objective function: Constraints: Where Z1 is the expected loss of life and economy downstream; P i is the probability of the occurrence of the i-th scenario; L i is the number of lives lost due to the i-th scenario; F1 represents the calculation function of scenario probability; D1 represents the calculation function of life loss; x i is a constraint condition, which represents the random variable of the scenario factors that affect the treatment effect in the i-th scenario; c i is the boundary condition, which represents the flood risk index affecting life in the i-th scenario.

9. A computer device, characterized in that: The computer device 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, the steps of the event-driven emergency scenario simulation method for reservoir dam emergencies as described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the event-driven emergency scenario simulation method for a reservoir dam emergency as described in any one of claims 1-8.

Images