Reservoir construction comprehensive evaluation method based on DEMATEL method and dynamic Bayesian network model

Abstract

The invention discloses a reservoir construction comprehensive evaluation method based on a DEMATEL method and a dynamic Bayesian network model, and the method comprises the steps: S1, recognizing the safety, progress and investment risk factors of a construction period through consulting literatures, and constructing a comprehensive risk evaluation index system; s2, on the basis of the comprehensive risk evaluation indexes, analyzing the mutual influence relationship between risk factors under each risk event by using a DEMATEL method; and S3, establishing a Bayesian network structure based on a DEMATEL method result, constructing a dynamic Bayesian network model by introducing a time factor, and performing dynamic risk reasoning by applying GeNIe4.0 software. According to the method, a DEMATEL-DBN-based safety-progress-investment risk comprehensive evaluation method is adopted, the relation among safety, progress and investment risk events is reflected by considering the influence relation of underlying factors, and the overall construction period risk is dynamically evaluated by introducing a time factor to construct a dynamic Bayesian method. And a scientific basis is provided for evaluating the construction period risk of the project.

Description

Technical Field

[0001] The invention belongs to the technical field of water conservancy project investment risk evaluation, and specifically relates to a comprehensive evaluation method for reservoir construction based on a DEMATEL method and a dynamic Bayesian network model. Background Art

[0002] As a major infrastructure, reservoir projects play an important role in flood control, irrigation, water supply and power generation. However, due to the complexity of the project itself and the uncertainty of the construction environment, reservoir construction faces many risks, which can be generally summarized as safety risks, progress risks and investment risks. Comprehensive management of these risks is the key to ensuring the success of the project. For this reason, it is particularly important to carry out systematic risk analysis and evaluation during the construction period of reservoir projects. This not only helps project managers fully understand the overall risk level faced by the project, but also provides a scientific basis for formulating effective risk control countermeasures and safeguards.

[0003] In the risk management process, risk identification, risk analysis, risk assessment and risk control are important links in the risk management of water conservancy project construction, among which risk assessment is the basis and key to risk control. Existing research methods for risk assessment during construction, including analytic hierarchy process, fuzzy comprehensive evaluation method and grey correlation method, are mostly based on static assessment, and these methods usually assume that risk factors are independent of each other, lack analysis of the interactive relationship of risks during construction, and are difficult to reflect the impact relationship between risk factors and their dynamic changes with construction. Summary of the invention

[0004] The purpose of the present invention is to provide a comprehensive evaluation method for reservoir construction based on the DEMATEL method and the dynamic Bayesian network model, so as to solve the technical problem that the comprehensive evaluation method of safety-progress-investment risk during reservoir construction in the prior art is difficult to reflect the interaction of safety, progress and investment risk factors and the dynamic changes of risks in the construction process.

[0005] In order to achieve the above object, the present invention adopts the following scheme:

[0006] The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model includes the following steps:

[0007] Step S1, identify the risk factors of safety, progress and investment during the construction period by consulting literature, and build a comprehensive risk assessment index system;

[0008] Step S2: Based on the comprehensive risk assessment index, the DEMATEL method is used to analyze the mutual influence relationship between risk factors under each risk event;

[0009] Step S3: Based on the results of the DEMATEL method, a Bayesian network structure is established. By introducing the time factor, a dynamic Bayesian network model is constructed, and dynamic risk reasoning is performed using the GeNIe4.0 software.

[0010] In the past, most studies on risk assessment during the construction period of a project focused on static assessment of a single risk, and did not involve the relationship between risk events. In view of the shortcomings of the existing technology, the present invention is based on a comprehensive evaluation method for safety-progress investment risk during the reservoir construction period based on the combination of the DEMATEL method and the dynamic Bayesian network model. The method can reflect the mutual coupling relationship between various risk factors, dynamically capture and evaluate the changes of various risk factors during the construction period, and provide a scientific basis for risk management during the construction period.

[0011] Further optimization, in step S1, the risks during the reservoir construction period are divided into three dimensions: safety risk, progress risk and investment risk. By consulting and analyzing relevant literature and screening by multiple experts, risk factors with low recognition and possible repetitiveness are removed, and a comprehensive risk assessment index system for the reservoir construction period is established. Among them, the safety risks include risk factors such as hydrological and meteorological conditions, material quality, improper construction technology, equipment failure, personnel operating errors and inadequate safety protection.

[0012] The risk factors of progress risk include insufficient resource allocation, inadequate construction quality, unreasonable construction plan, hydrological and meteorological conditions, land acquisition and resettlement risks, changes in policies and regulations, and cross-construction of sub-projects; the risk factors of investment risk include cost overruns of manpower, machinery, and materials, changes in the market economic environment, construction changes and additional projects, inaccurate project information, budget execution deviations, unclear contract terms, and land acquisition and resettlement costs.

[0013] Further optimization, in step S2, the DEMATEL method is used to analyze the interaction relationship among all risk factors in safety, progress and investment risks. The specific calculation steps are as follows:

[0014] Step S2.1: Construct the initial direct impact matrix B of risk factors:

[0015] B=[b ij ] n×n (10)

[0016] In the above formula: b ij represents the impact of risk factor i on risk factor j. Each factor cannot affect itself, that is, the main diagonal elements of matrix B are all 0; n is the number of risk factors, i,j∈[1,n];

[0017] Step S2.2, normalize the initial direct influence matrix to obtain the matrix H:

[0018]

[0019] Step S2.3, calculate the comprehensive impact matrix G:

[0020] G=H(EH) -1 (12)

[0021] In the above formula: E is the unit matrix, G = [g ij ] n×n .

[0022] Step S2.4: For the dth risk factor, calculate the impact R d 、Influence C d , centrality e d And the cause degree f d , d∈[1,n]:

[0023]

[0024] In the above formula: R d is the influence degree, which is the sum of the elements in the dth row of the comprehensive influence matrix G, describing the comprehensive impact of the dth risk factor on other risk factors; C d is the degree of influence, which is the sum of the elements in the dth column of the comprehensive influence matrix G, describing the comprehensive influence of other risk factors on the dth risk factor; d is the centrality, indicating the importance of the indicator; f d If its value is greater than 0, it is a cause factor; if its value is less than 0, it is a result factor.

[0025] Step S2.5: Construct an importance-causality diagram based on the calculation results of step S2.4.

[0026] The Decision Experiment and Evaluation Laboratory DEMATEL method is a structured method for the mutual coupling relationship between multiple factors in complex systems by combining target causal relationship diagrams and matrix tools. In risk analysis, the DEMATEL method can visualize and structure the attributes of each factor in a complex system and their relationships. It can not only help decision makers identify causal factors and result factors, but also intuitively display the dynamic interactions and key factors between risk factors through causal relationship diagrams.

[0027] Further optimization, the step S3 specifically includes the following steps:

[0028] Step S3.1, set a threshold R, delete the elements with influence relationship strength less than R in the comprehensive influence matrix G, and retain the elements with influence relationship strength greater than or equal to R;

[0029] Step S3.2: Based on the elements retained in step S3.1, combined with the importance-causality diagram, identify the risk transmission path with influencing relationship factors, and convert the elements greater than the threshold R in the comprehensive impact matrix G into the DBN initial network B 0 Valid nodes and arcs in ;

[0030] Step S3.3, based on the risk transfer path analysis, according to the principle of identifying and eliminating cycles, eliminate the possible cycles, and then establish the safety-progress-investment comprehensive risk DBN network structure;

[0031] Step S3.4, divide the reservoir project construction period into several time slices; use expert experience to determine the prior probability of each parent node risk factor of the DBN network model, the conditional probability of the parent node risk factor to the child node risk factor, and the original transfer probability of the parent node risk factor in adjacent time slices;

[0032] Step S3.5: Input the state probability of each node obtained in step S3.4 into the DBN network structure model of safety-progress-investment comprehensive risk obtained in step S3.3, and update the prior probability of the key time slice according to the actual project situation; use GeNIe4.0 software to dynamically infer the risk events of each time slice to obtain the dynamic probability change curve of the occurrence of safety-progress-investment comprehensive risk during the reservoir construction period.

[0033] Further optimization, in step S3.3, in view of the fact that the influence relationship in the comprehensive influence matrix may have loops, in order to convert it into the DBN initial network, the loops are identified and eliminated according to the following principles: 1) The DBN network not only reflects the causal relationship between risk factors, but also reflects its direction, eliminating bidirectional arcs; for example, there is α affecting β (α→β), but there cannot be β→α; 2) DBN mainly represents direct causal relationships, eliminating loops in the initial network diagram caused by indirect causal relationships; 3) Eliminate loops in the initial network diagram caused by the influence of time or other variables on the causal relationship between risk factors. According to the above principles, the bidirectional arcs and closed-loop arcs in the initial network are checked, and the loop arcs are identified and eliminated to obtain the dynamic Bayesian initial network diagram.

[0034] Further optimization, in the step S3.4, considering the characteristics of reservoir project construction such as strong cycle, complex and intertwined projects and dynamic changes in risks, the reservoir project construction period is divided into time slices according to the following principles: 1) Multi-time slice division: Through multi-time slice division, the dynamic changes in risks in the construction process are carefully displayed; 2) Phase correspondence: The time slice division should correspond to the key project stages, which include the project preparation period, construction diversion project, main construction project and project completion period; 3) Refine time slices: Considering the long construction time span and the complex cross-construction of sub-projects in the project stage, identify the key sub-projects to refine the time slices; 4) Span between time slices: Determine the span of the time slice according to the construction duration of the sub-projects and the connection between the sub-projects in each project stage.

[0035] Further optimization, in step S3.4, the transfer probability of the risk factor of the parent node of any event in the DBN network between two adjacent time slices is shown in formula (5):

[0036]

[0037] In the above formula: is the kth node risk factor on time slice t; For Node The parent node risk factor set; m is the number of transfer nodes, that is, the number of parent node risk factors.

[0038] Further optimization, in step S3.4, on the basis of fully understanding the actual situation of the project, the risk probabilities are qualitatively expressed using the IPCC probability qualitative expression in Table 1. In order to reduce the subjective differences in scoring by different experts, experts are organized to re-evaluate the risk factors with large differences in language variables until the evaluation values ​​are in the same interval or adjacent intervals in Table 1;

[0039] Table 1 IPCC probability qualitative expression

[0040] Probability Range % Language (0，1] Almost impossible (1，10] Very unlikely to happen (10,33] Less likely to happen (33,66] Moderately likely to occur (66,90] More likely to happen (90,99] It is very likely to happen (99,100] It will happen .

[0041] Further optimization is performed by using the Leaky Noisy-or gate extended model to correct the conditional probability and further improve the accuracy of the model. In the Leaky Noisy-or Gate extended model, it is assumed that the parent node of risk event X is X k and X all , and their connection probabilities are P k and P all , where X all is divided by X k The sum of all factors other than the identified risk factors, that is, the sum of all unidentified risks except the identified risk factors, then:

[0042] P(X / X k )=P k +P all -P k P all (15)

[0043]

[0044] In the above formula: Yes X k opposing events.

[0045] Combining equations (6) and (7), we can obtain the connection probability P of all parent nodes of risk event X: k , see formula (9).

[0046]

[0047] Combine all the unconsidered risk factors in the parent nodes of each risk event into an unknown factor X l , let its connection probability be P l , then the conditional probability of risk event X is:

[0048]

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention analyzes the relationship between risk factors under safety, progress and investment based on decision-making experiments and laboratory evaluation method DEMATEL, reflects the interaction between safety, progress and investment from the bottom-level factor influence relationship, and analyzes its influence path; finally, based on the analysis of the influence path, Bayesian is horizontally expanded into a multi-stage dynamic Bayesian network DBN according to the actual project, and the individual risks and comprehensive risks during the reservoir construction period are dynamically evaluated in combination with the actual project detection data, which provides a scientific basis for evaluating the risks during the construction period of such projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 To invent the construction period safety-progress-investment risk comprehensive evaluation model framework;

[0052] Figure 2 for importance-causality diagram;

[0053] Figure 3 It is the DBN network topology diagram;

[0054] Figure 4 This is the DBN model diagram for the comprehensive risk of safety-progress-investment;

[0055] Figure 5This is a bar chart of the reservoir project construction progress;

[0056] Figure 6 This is a schematic diagram of the DBN inference model for reservoir engineering;

[0057] Figure 7 It is the time series change curve of each risk probability of reservoir project. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. 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 creative work are within the scope of protection of the present invention.

[0059] The method of the present invention is further described below in conjunction with the Yuanwan Reservoir project. The total storage capacity of the Yuanwan Reservoir is 251 million m 3 , controlling a watershed area of ​​480km 2 The main structures of the reservoir include dam, overflow surface hole, flood discharge bottom hole, water transfer tunnel and power station. The total construction period of the project is 48 months.

[0060] like Figure 1 As shown in FIG. 1 , the comprehensive evaluation method for reservoir construction based on the DEMATEL method and the dynamic Bayesian network model includes the following steps:

[0061] Step S1: Identify the risk factors of safety, progress and investment during the construction period by consulting literature, and build a comprehensive risk assessment indicator system.

[0062] In this embodiment, the risks during the reservoir construction period are divided into three dimensions: safety risk, schedule risk and investment risk. By consulting and analyzing relevant literature and screening by multiple experts, the risk factors with low recognition and possible repetitiveness are removed, and a comprehensive risk assessment index system for the reservoir construction period is established as shown in Table 1.

[0063] Table 1 Comprehensive risk assessment index system during reservoir construction period

[0064]

[0065] Step S2: Based on the comprehensive risk assessment index, the DEMATEL method is used to analyze the mutual influence relationship between risk factors under each risk event. The specific calculation steps are as follows:

[0066] Step S2.1: Construct the initial direct impact matrix B of risk factors:

[0067] B=[b ij ]n×n (19)

[0068] In the above formula: b ij It represents the influence of risk factor i on risk factor j. Each factor cannot affect itself, that is, the main diagonal elements of matrix B are all 0; n is the number of risk factors, i,j∈[1,n].

[0069] In this embodiment, according to the evaluation index system determined in step S1, 10 water conservancy engineering experts are organized, including 3 water conservancy engineering technical experts, 2 project managers, and 5 university professors. Through questionnaires, the impact degree of each event risk factor is scored. The questionnaires of each expert are combined and the average value is taken. The 1-5 scale method is used to divide the impact degree from large to small into 5 levels, where 0, 1, 2, 3, and 4 represent no impact, very low impact, low impact, high impact, and very high impact, respectively. The initial direct impact matrix B is obtained as [b ij ] 20×20 , as shown in Table 2.

[0070] Table 2 Initial direct impact matrix B

[0071] <![CDATA[S 1 ]]> <![CDATA[S 2 ]]> <![CDATA[S 3 ]]> <![CDATA[S 4 ]]> <![CDATA[S 5 ]]> <![CDATA[S 6 ]]> <![CDATA[P 1 ]]> <![CDATA[P 2 ]]> <![CDATA[P 3 ]]> <![CDATA[P 4 ]]> <![CDATA[P 5 ]]> <![CDATA[P 6 ]]> <![CDATA[P 7 ]]> <![CDATA[I 1 ]]> <![CDATA[I 2 ]]> <![CDATA[I 3 ]]> <![CDATA[I 4 ]]> <![CDATA[I 5 ]]> <![CDATA[I 6 ]]> <![CDATA[I 7 ]]> <![CDATA[S 1 ]]> 0 1.3 0.1 1.1 0.3 0.5 0.6 0.6 1.3 0 0 0.5 0.7 0.4 1.2 0.3 0 0.1 0 0 <![CDATA[S 2 ]]> 0 0 0.3 0.6 0.1 0 1.5 2.8 0.2 0 0 0 1.5 1.2 0 0.2 0 0.1 0 0 <![CDATA[S 3 ]]> 0 0 0 0.9 0.6 0.1 0.6 2.7 0.5 0 0 0 1.1 1.2 0 0.7 0 0.3 0 0 <![CDATA[S 4 ]]> 0 0 0.5 0 0 0 2.6 0.9 0.8 0 0 0 1.3 2.9 0 0.3 0 0 0 0 <![CDATA[S 5 ]]> 0 0 0.1 1.5 0 0 0 0.6 0 0 0 0 0.3 0.5 0 0 0 0 0 0 <![CDATA[S 6 ]]> 0 0 0 0 0 0 0.1 0 0.4 0 0 0 0.4 1 0 0.1 0 0.1 0 0 <![CDATA[P 1 ]]> 0 0.1 0.7 0 0.1 0.3 0 0.6 0.5 0 0 0 0.6 0 0 0 0 0 0 0 <![CDATA[P 2 ]]> 0 0 0 0 0 0.4 0.1 0 0.3 0 0 0 1 2.9 0 1.1 0 0.2 0 0 <![CDATA[P 3 ]]> 0 0 0.6 0.2 0.3 0.7 0.7 0.5 0 0 0 0 1.1 0.7 0 1.2 0 0.3 0 0 <![CDATA[P 4 ]]> 0 1.3 0.1 1.1 0.3 0..5 0.6 0.6 1.3 0 0 0.5 0.7 0.4 1.2 0.3 0 0.1 0 0 <![CDATA[P 5 ]]> 0 0 0 0 0 0.1 0.2 0 0.1 0 0 0 0 0 0 0 0 0 0 0 <![CDATA[P 6 ]]> 0 0.1 0.1 0 0 0.2 0.3 0 0.1 0 1.5 0 0.1 0.4 2.7 0.2 0 0.1 0.4 1.5 <![CDATA[P 7 ]]> 0 0 1 0.4 0.4 1.4 0.8 0.7 0.8 0 0 0 0 0.6 0 0.3 0 0.2 0 0 <![CDATA[I 1 ]]> 0 0.3 0 0 0 0.9 0.9 0.3 0.2 0 0 0 0.4 0 0 0.1 0 0.2 0 0 <![CDATA[I 2 ]]> 0 0.3 0 0 0 0.1 0.5 0 0.1 0 0.2 0 0.1 0.2 0 0 0 0 0 0.4 <![CDATA[I 3 ]]> 0 0 1 0.2 0.6 0.8 2.8 0.7 1 0 0.1 0 1 0.9 0 0 0 0.6 0.1 0 <![CDATA[I 4 ]]> 0 0 0.6 0 2.7 0.1 0 0 3.1 0 0.1 0 0.4 0 0 0.4 0 0 0 0 <![CDATA[I 5 ]]> 0 0.6 0.3 0 0 2.7 0.8 0.4 0.3 0 0 0 0.6 0.1 0 0 0 0 0 0 <![CDATA[I 6 ]]> 0 0 0.5 0 0 0.1 0.1 0 0.8 0 0.9 0 0.3 0.1 0 0.8 0 0.1 0 0.3 <![CDATA[I 7 ]]> 0 0.2 0.1 0 0 0.4 0.5 0 0 0 0 0 1 0.1 0 0 0 0 0 0

[0072] Step S2.2, normalize the initial direct influence matrix to obtain the matrix H:

[0073]

[0074] Step S2.3, calculate the comprehensive influence matrix G:

[0075] G=H(EH) -1 (twenty one)

[0076] In the above formula: E is the unit matrix, G = [g ij ] n×n .

[0077] In this embodiment, the comprehensive influence matrix G is obtained by calculation according to formulas (1)-(3), as shown in Table 3.

[0078] Table 3 Comprehensive impact matrix G

[0079]

[0080]

[0081] Step S2.4: For the dth risk factor, calculate the impact R d 、Influence C d , centrality e d And the cause degree f d , d∈[1,n]:

[0082]

[0083] In the above formula: R d is the influence degree, which is the sum of the elements in the dth row of the comprehensive influence matrix G, describing the comprehensive impact of the dth risk factor on other risk factors; C d is the degree of influence, which is the sum of the elements in the dth column of the comprehensive influence matrix G, describing the comprehensive influence of other risk factors on the dth risk factor; d is the centrality, indicating the importance of the indicator; f d is the cause degree. If its value is greater than 0, it is a cause factor. If its value is less than 0, it is a result factor.

[0084] Step S2.5: Based on the calculation results of step S2.4, construct an importance-causality diagram, such as Figure 2 shown.

[0085] according to Figure 2 The cause factors and result factors as well as the importance of the factors can be known, such as insufficient resource allocation P 1 , Cost overrun of manpower and equipment 1 The cause degree is the lowest, indicating that it is affected by other factors; material quality status S 2 、Project information is inaccurate 4 The reasons are high and can significantly affect other factors. 1 , Cross-construction of sub-projects 7 , Insufficient construction quality 2 , Insufficient resource allocation 1 The centrality is high, and the surface is in a key position in the entire risk factor.

[0086] Step S3: Based on the results of the DEMATEL method, a Bayesian network structure is established, a dynamic Bayesian network model is constructed by introducing the time factor, and the GeNIe4.0 software is used for dynamic risk reasoning. Specifically, the following steps are included:

[0087] Step S3.1, set a threshold R, delete the elements with influence relationship strength less than R in the comprehensive influence matrix G, and retain the elements with influence relationship strength greater than or equal to R.

[0088] In this embodiment, the threshold R is determined to be 0.25 based on expert experience.

[0089] Step S3.2: Based on the elements retained in step S3.1, the importance-causality diagram is combined to identify the risk transmission path with influencing factors, and the elements greater than the threshold R in the comprehensive impact matrix G are converted into the DBN initial network B. 0Valid nodes and arcs in DBN initial network B 0 like Figure 3 As shown in (a) in .

[0090] In this embodiment, for example: the material quality status S 2 , Improper construction technology 3 → Inadequate construction quality 2 → Cost overrun of manpower and equipment I 1 , that is, node S 2 , S 3 There is not only a directed arc pointing to node S, but also a directed arc pointing to node P. 2 The directed arc of node P 2 Also; Equipment failure S 4 → Insufficient resource allocation 1 , that is, there is not only node S 4 There is a directed arc pointing to node S, and there is also a node S 4 Point to node P 1 Directed arcs of construction changes and additional projects I 3 → Insufficient resource allocation 1 , that is, there is not only node I 3 There is a directed arc pointing to node I, and there is also a node I 3 Point to node P 1 Directed arcs; item information is inaccurate I 4 → Human operation error S 5 , that is, there is not only node I 4 There is a directed arc pointing to node I, and there is also a node I 4 Point to node S 5 A directed arc of .

[0091] Step S3.3: Based on the risk transfer path analysis, according to the principle of identifying and eliminating cycles, eliminate the possible cycles, and then establish the dynamic network structure DBN of safety-progress-investment comprehensive risk, such as Figure 4 shown.

[0092] Given that the influence relationships in the comprehensive influence matrix may have loops, in order to convert it into the DBN initial network, the loops are identified and eliminated according to the following principles: 1) The DBN network not only reflects the causal relationship between risk factors, but also reflects its direction, eliminating bidirectional arcs; for example, there is an α influence on β (α→β), but there cannot be β→α; 2) DBN mainly represents direct causal relationships, eliminating loops in the initial network diagram caused by indirect causal relationships; 3) Eliminate loops in the initial network diagram caused by the influence of time or other variables on the causal relationship between risk factors. According to the above principles, the bidirectional arcs and closed-loop arcs in the initial network are checked, and the loop arcs are identified and eliminated to obtain a dynamic Bayesian network.

[0093] Step S3.4, divide the reservoir construction period into several time slices. Use expert experience to determine the prior probability of each parent node risk factor of the DBN network model, the conditional probability of the parent node risk factor to the child node risk factor, and the original transfer probability of the root node risk factor of the adjacent time slice.

[0094] The details of the time slice division and DBN network model parameter determination process for the reservoir project construction period are as follows:

[0095] Time slice division:

[0096] Taking into account the characteristics of reservoir project construction such as strong cycle, complex and intertwined projects and dynamic changes in risks, the reservoir project construction period is divided into time slices according to the following principles: 1) Multi-time slice division: Through multi-time slice division, the dynamic changes in risks in the construction process are carefully displayed; 2) Phase correspondence: The time slice division should correspond to the key project stages, which include the project preparation period, construction diversion project, main construction project and project completion period; 3) Refinement of time slices: Considering the long construction time span and complex cross-construction of sub-projects, identify key sub-projects to refine the time slices; 4) Span between time slices: Determine the span of the time slice according to the construction duration of the sub-projects and the connection between the sub-projects in each project stage.

[0097] In this embodiment, the entire construction period of the Yuanwan Reservoir Project mainly includes the project preparation period. 1 , Construction diversion project 2 , Main project construction 3 and project completion period W 4 Four parts. The construction diversion project includes three sub-projects: the first-phase cofferdam, the construction diversion open channel dam section cofferdam, and the diversion open channel. The main project construction includes seven sub-projects: auxiliary dam, earth dam, concrete dam, diversion open channel dam section, power station project, traffic project, and metal structure installation and commissioning. Since there are many sub-projects in this project and most of them are cross-construction, in order to more carefully show the dynamic changes of risks with the construction process, the entire construction process is divided into 10 time slices. The project preparation period is time slice 1, the construction diversion project and part of the auxiliary dam construction in the main dam body project are time slices 2 and 3. The main project construction sub-projects are numerous and the construction cross-time span is large, so the main dam body project and the traffic project excluding part of the auxiliary dam construction are divided into time slices 4 to 9, and the project completion period time slice is 10. For specific divisions, see Figure 5 shown.

[0098] Model parameter determination:

[0099] Dynamic Bayesian Network (DBN) extends the static Bayesian Network by adding a time dimension, which can better analyze the process of risk events changing over time. 0 and the transfer network B of the subsequent time slice → Composition, that is, first establish the initial network B 0 , and define the joint probability distribution P at the initial moment B0 (X); then construct the transfer network B → According to the characteristics of the project stage or the actual analysis requirements, the entire project stage is divided into several time periods (also called time slices), and the transition probability P(X) of the same risk event (i.e. the same parent node) in adjacent time periods is defined. t / X t-1 ). At the initial moment, the parent node of event X[1] is the prior network B 0 At time t+1, the parent node of event X[t+1] is the transition network (B → ) nodes, such as Figure 3 As shown in (b).

[0100] The transfer probability of the risk factor of any event root node in the DBN network between two adjacent time slices is shown in formula (5):

[0101]

[0102] In the above formula: is the kth node risk factor on time slice t; For Node The parent node risk factor set; m is the number of transfer nodes, that is, the number of parent node risk factors.

[0103] In order to simplify the complexity of dynamic systems, existing research results all assume that DBN satisfies the two premise assumptions of first-order Markov and constant transition probability, which requires that the directed edges between nodes must be located in the same time slice or between adjacent time slices and B → The transition probability P(X t / X t-1 ) does not change over time. Based on the above assumptions, the joint probability distribution of any node in DBN like

[0104] As shown in formula 6:

[0105]

[0106] According to the assumption that the transfer probability is constant, the transfer probability of the same event in two adjacent time slices is the same. However, considering the uncertainty of risk factors during the reservoir construction period, this probability is difficult to obtain through parameter learning or sample training. Therefore, the original transfer probability of each node is formulated by water conservancy engineering experts after fully understanding the actual project.

[0107] In the study of water conservancy project construction risks, due to the particularity and dynamic characteristics of water conservancy project construction risks, a large amount of accident-related data is difficult to obtain, so the expert survey and scoring method is used to determine the prior probability and conditional probability. In order to unify the probability expression when experts score, and make the scoring data more convenient for reference and distinction, the present invention adopts the 7-level risk probability expression method proposed by the Intergovernmental Panel on Climate Change (IPCC) of the United Nations, as shown in Table 4.

[0108] Table 4 IPCC probability qualitative expression

[0109] Probability Range % Language (0，1] Almost impossible (1，10] Very unlikely to happen (10,33] Less likely to happen (33,66] Moderately likely to occur (66,90] More likely to happen (90,99] It is very likely to happen (99,100] It will happen

[0110] There are many risk factors in engineering practice, and human identification of risk factors has certain limitations. There are some unknown or missed risks, which are not considered in the Bayesian model. In order to fully consider the impact of various factors on nodes and improve the accuracy of probability calculation, Henrion proposed the Leaky Noisy-or Gate extension model. In the Leaky Noisy-or Gate extension model, assume that the parent node of event X is X k and X all , and their connection probabilities are P k and P all , where X all is divided by X k The sum of all factors other than those identified in Table 1, that is, the sum of unidentified risks, then:

[0111] P(X / X k )=P k +P all -P k P all (25)

[0112]

[0113] Where: Yes X k opposing events.

[0114] Combining equations (7) and (8), we can get the connection probability P of all parent nodes of node X: k , see formula (9).

[0115]

[0116] Combine all the unconsidered risk factors in the parent nodes of each event into an unknown factor X l , let its connection probability be P l , then the conditional probability of node X is:

[0117]

[0118] Due to space limitations, in this embodiment, only the “P 1 "The probability calculation of node risk factors is used as an example. Assume that the P1 node is event X and its parent node risk factor is S 4 ,I 3 , assuming that event X 1 , X 2 . "X k =1" indicates that the event has occurred, "X k =0" indicates that the event does not occur. kt ) represents the original transition probability of event k, which can be understood as the probability of event k occurring in the current time slice given that it has occurred in the previous time slice.

[0119] In this embodiment, the root node (parent node) event X is determined by expert experience. 1 , X 2 The prior probability of X 1 and X 2 The conditional probability of the child node X under the occurrence and non-occurrence conditions and X 1 and X 2 The original transition probability is shown in Table 5.

[0120] Table 5 Probability parameters of node X

[0121]

[0122] In this embodiment, suppose the unknown factor, P l = 0.1, and substitute it into formula (10) to obtain the modified conditional probability of X, as shown in Table 6.

[0123] Table 6 Conditional Probability after X Correction

[0124] <![CDATA[X 1 ]]> <![CDATA[X 2 ]]> P(X=1) P(X=0) 1 0 0.33 0.67 0 1 0.38 0.62 1 1 0.54 0.46

[0125] The calculation of the probability of the "dam body failure" node in the DBN model is taken as an example to illustrate.

[0126] Step S3.5: Input the state probability of each node obtained in step S3.4 into the DBN network structure model of safety-progress-investment comprehensive risk obtained in step S3.3, such as Figure 6 As shown in the figure, the prior probability of the key time slice is updated according to the actual situation of the project; the risk events of each time slice are dynamically inferred using the GeNIe4.0 software to obtain the dynamic probability change curve of the comprehensive risk of safety, progress and investment during the reservoir construction period, as shown in the figure. Figure 7 As shown, Figure 7 (a) is the time series curve of safety risk probability. Figure 7 (b) is the time series change curve of schedule risk probability. Figure 7 (c) in the figure is the time series curve of investment risk probability. Figure 7 (d) in the figure is the time series change curve of comprehensive risk probability.

[0127] according to Figure 7 It can be seen that the probability of each risk event during the construction period of the reservoir changes. The probability of each time slice with more risk events is below 0.33, and the risk is less likely to occur. Some time slices are above 0.33. The main reason is that during these time periods, sub-projects are frequently cross-constructed and are in the flood season, which increases the complexity of the construction site environment and the difficulty of communication, organization and coordination. According to the DBN reasoning results, it can help project managers grasp the dynamic changes of various risk events during the construction period and make targeted risk prevention and control measures in a timely manner.

[0128] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model, characterized in that: The following steps are involved: Step S1, identify the safety, progress and investment risk factors during the construction period by consulting literature, and build a comprehensive risk assessment index system; Step S2: Based on the comprehensive risk assessment index, the DEMATEL method is used to analyze the mutual influence relationship between risk factors under each risk event; Step S3: Based on the results of the DEMATEL method, a Bayesian network structure is established. By introducing the time factor, a dynamic Bayesian network model is constructed, and dynamic risk reasoning is performed using the GeNIe4.0 software.

2. The comprehensive evaluation method for reservoir construction based on the DEMATEL method and the dynamic Bayesian network model according to claim 1 is characterized in that: In step S1, the risks during the reservoir construction period are divided into three dimensions: safety risk, schedule risk and investment risk. By consulting and analyzing relevant literature and screening by multiple experts, the risk factors with low recognition and possible duplication are removed to establish a comprehensive risk assessment index system for the reservoir construction period. Among them, safety risks include risk factors such as hydrological and meteorological conditions, material quality, improper construction technology, equipment failure, human operating errors and inadequate safety protection; The risk factors of schedule risk include insufficient resource allocation, insufficient construction quality, unreasonable construction plan, hydrological and meteorological conditions, risks of land acquisition and resettlement, changes in policies and regulations, and cross-construction of sub-projects; Investment risks include cost overruns of manpower, equipment and materials, changes in the market economic environment, construction changes and additional projects, inaccurate project information, budget execution deviations, unclear contract terms and resettlement costs.

3. The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model according to claim 2 is characterized in that: The specific calculation steps of step S2 are as follows: Step S2.1: Construct the initial direct impact matrix B of risk factors: B=[b ij ] n×n (1) In the above formula: b ij represents the impact of risk factor i on risk factor j. Each factor cannot affect itself, that is, the main diagonal elements of matrix B are all 0; n is the number of risk factors, i,j∈[1,n]; Step S2.2, normalize the initial direct influence matrix to obtain the matrix H: Step S2.3, calculate the comprehensive influence matrix G: G=H(E-H) -1 (3) In the above formula: E is the unit matrix, G = [g ij ] n×n ; Step S2.4: For the dth risk factor, calculate the impact R d 、Influence C d , centrality e d And the cause degree f d , d∈[1,n]: In the above formula: R d is the influence degree, which is the sum of the elements in the dth row of the comprehensive influence matrix G, describing the comprehensive impact of the dth risk factor on other risk factors; C d is the degree of influence, which is the sum of the elements in the dth column of the comprehensive influence matrix G, describing the comprehensive influence of other risk factors on the dth risk factor; d is the centrality, indicating the importance of the indicator; f d is the cause degree. If its value is greater than 0, it is a cause factor. If its value is less than 0, it is a result factor. Step S2.5: Construct an importance-causality diagram based on the calculation results of step S2.

4.

4. The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model according to claim 3 is characterized in that: The step S3 specifically includes the following steps: Step S3.1, set a threshold R, delete the elements with influence relationship strength less than R in the comprehensive influence matrix G, and retain the elements with influence relationship strength greater than or equal to R; Step S3.2: Based on the elements retained in step S3.1, combined with the importance-causality diagram, identify the risk transmission path with influencing relationship factors, and convert the elements greater than the threshold R in the comprehensive impact matrix G into valid nodes and arcs in the DBN initial network B0; Step S3.3, based on the risk transfer path analysis, according to the principle of identifying and eliminating cycles, eliminate the possible cycles, and then establish the safety-progress-investment comprehensive risk DBN network structure; Step S3.4, divide the reservoir project construction period into several time slices; use expert experience to determine the prior probability of each parent node risk factor of the DBN network model, the conditional probability of the parent node risk factor to the child node risk factor, and the original transfer probability of the parent node risk factor in adjacent time slices; Step S3.5, input the state probability of each node obtained in step S3.4 into the safety-progress-investment comprehensive risk DBN network structure model obtained in step S3.3, and update the prior probability of the key time slice according to the actual project situation; The GeNIe4.0 software was used to dynamically infer the risk events in each time slice, and the dynamic probability change curve of the comprehensive risk of safety, progress and investment during the reservoir construction period was obtained.

5. The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model according to claim 4 is characterized in that: In step S3.3, loops are identified and eliminated according to the following principles: 1) DBN network reflects not only the causal relationship between risk factors, but also its direction, eliminating bidirectional arcs; 2) DBN mainly represents direct causal relationships, eliminating loops existing in the initial network diagram due to indirect causal relationships; 3) Eliminate loops existing in the initial network due to the influence of time or other variables on the causal relationship between risk factors; According to the above principles, the bidirectional arcs and closed-loop arcs in the initial network are checked, the loop arcs are identified and removed, and the dynamic Bayesian initial network graph is obtained.

6. The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model according to claim 5 is characterized in that: In step S3.4, considering the characteristics of reservoir project construction such as strong cycle, complex and intertwined projects and dynamic changes in risks, the reservoir project construction period is divided into time slices according to the following principles: 1) Multi-time slice division: Through multi-time slice division, the dynamic changes in risks during the construction process are carefully displayed; 2) Phase correspondence: The time slice division should correspond to the key project phases, which include the project preparation period, construction diversion project, main construction project and project completion period; 3) Refine time slices: Consider the project phases with long construction time spans and complex cross-construction of sub-projects, identify key sub-projects to refine time slices; 4) Span between time slices: Determine the span of the time slice based on the construction duration of the sub-projects in each project phase and the connection between the sub-projects.

7. The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model according to claim 6 is characterized in that: In step S3.4, the transfer probability of the risk factor of the parent node of any event in the DBN network between two adjacent time slices is shown in formula (5): In the above formula: is the kth node risk factor on time slice t; For Node The parent node risk factor set; m is the number of transfer nodes, that is, the number of parent node risk factors.

8. The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model according to claim 7 is characterized in that: In step S3.4, based on a full understanding of the actual situation of the project, the probability of each risk is qualitatively expressed using the IPCC probability qualitative expression in Table 1; In order to reduce the subjective differences in scores from different experts, experts were organized to re-evaluate risk factors with large differences in language variables until the evaluation values ​​were in the same range or adjacent ranges in Table 1; Table 1 IPCC probability qualitative expression 。 9. The comprehensive evaluation method for reservoir construction based on DEMATEL method and dynamic Bayesian network model according to claim 8 is characterized in that: The Leaky Noisy-or gate extended model is used to correct the conditional probability. In the Leaky Noisy-or Gate extended model, it is assumed that the parent node of risk event X is X k and X all , and their connection probabilities are P k and P all , where X all is divided by X k The sum of all factors other than the risk factors identified in claim 2, that is, the sum of all unidentified risks, then: P(X / X k )=P k +P all -P k P all (6) In the above formula: Yes X k The opposing events; Combining equations (6) and (7), we can obtain the connection probability P of all parent nodes of risk event X: k , see formula (8); Combine all the unconsidered risk factors in the parent nodes of each risk event into an unknown factor X l , let its connection probability be P l , then the conditional probability of risk event X is:

Images