Water conservancy project risk assessment method and system based on artificial intelligence

Abstract

The invention discloses a water conservancy project risk assessment method and system based on artificial intelligence, and relates to the technical field of risk assessment management, and the method comprises the steps: obtaining a feature parameter set; calling an information value quantification function to quantitatively analyze the prediction capability of the water conservancy project risk coefficient to obtain a first value coefficient; if the first value coefficient reaches a preset threshold value, adding the first feature to a candidate feature factor list; and activating the accident cause Bayesian network model to perform cause analysis on the candidate feature factor list to obtain a target cause chain and a target key feature factor list, performing dynamic feature collection on the target water conservancy project to obtain a real-time feature parameter set, and performing normalization calculation analysis to obtain a target risk coefficient of the target water conservancy project. The technical problem of low risk assessment precision caused by lack of systematic analysis of water conservancy project multi-source features in the prior art is solved, and the technical effect of improving the water conservancy project risk assessment precision is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of risk assessment management, and in particular to a water conservancy project risk assessment method and system based on artificial intelligence. Background Art

[0002] Water conservancy projects involve a variety of geological, hydrological, structural and environmental factors, and their safety is affected by multi-source heterogeneous data, such as geological survey data, construction quality data, operation monitoring data, and meteorological and hydrological data. Existing water conservancy project risk assessment methods are mostly based on traditional statistical analysis, empirical reasoning or simple mathematical modeling, which often have problems such as insufficient accuracy in complex environments and are difficult to meet the modern water conservancy project's requirements for efficient, accurate and intelligent safety monitoring and risk assessment. Summary of the invention

[0003] The present application provides an artificial intelligence-based water conservancy project risk assessment method and system, which is used to solve the technical problem that the existing technology lacks systematic analysis of the multi-source characteristics of water conservancy projects, resulting in low risk assessment accuracy.

[0004] In view of the above problems, this application provides a water conservancy project risk assessment method and system based on artificial intelligence.

[0005] In a first aspect of the present application, a water conservancy project risk assessment method based on artificial intelligence is provided, the method comprising: Traverse the multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to a first feature; call the information value quantification function to perform a quantitative analysis of the predictive ability of the first parameter on the water conservancy project risk coefficient to obtain a first value coefficient of the first feature; if the first value coefficient reaches a predetermined threshold, add the first feature to a candidate feature factor list; activate the accident causal Bayesian network model to perform causal analysis on the candidate feature factor list to obtain a target causal chain, and analyze the target causal chain to obtain a target key feature factor list; based on the target key feature factor list, dynamically collect features of the target water conservancy project to obtain a real-time feature parameter set; perform normalization calculation and analysis on the real-time feature parameter set, and perform arithmetic average calculation on the normalized feature values ​​to obtain a target risk coefficient of the target water conservancy project.

[0006] The second aspect of the present application provides a water conservancy project risk assessment system based on artificial intelligence, the system comprising: A feature parameter set acquisition module is used to traverse the multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to a first feature; a quantitative analysis module is used to call the information value quantification function to perform a quantitative analysis of the predictive ability of the first parameter on the water conservancy project risk coefficient to obtain a first value coefficient of the first feature; a threshold judgment module is used to add the first feature to the candidate feature factor list if the first value coefficient reaches a predetermined threshold; a causal analysis module is used to activate the accident causal Bayesian network model to perform causal analysis on the candidate feature factor list to obtain a target causal chain, and analyze the target causal chain to obtain a target key feature factor list; a dynamic feature collection module is used to perform dynamic feature collection on the target water conservancy project based on the target key feature factor list to obtain a real-time feature parameter set; a target risk coefficient acquisition module is used to perform normalized calculation and analysis on the real-time feature parameter set, and perform arithmetic average calculation on the normalized feature values ​​to obtain the target risk coefficient of the target water conservancy project.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: The present application traverses multi-source engineering features in a water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to a first feature; calls an information value quantification function to perform a quantitative analysis of the first parameter's predictive ability for a water conservancy project risk factor to obtain a first value coefficient of the first feature; if the first value coefficient reaches a predetermined threshold, the first feature is added to a candidate feature factor list; activates an accident causal Bayesian network model to perform a causal analysis on the candidate feature factor list to obtain a target causal chain, and analyzes the target causal chain to obtain a target key feature factor list; dynamically collects features of a target water conservancy project based on the target key feature factor list to obtain a real-time feature parameter set; performs normalization calculation and analysis on the real-time feature parameter set, and performs arithmetic mean calculation on the normalized feature values ​​to obtain a target risk coefficient for the target water conservancy project. The present invention solves the technical problem that the prior art lacks a systematic analysis of the multi-source characteristics of water conservancy projects, resulting in low risk assessment accuracy. The present invention achieves the technical effect of improving the accuracy of risk assessment of water conservancy projects by constructing a multi-source characteristic parameter set, using an information value quantification function to screen key characteristics, analyzing the causal chain in combination with a Bayesian network model of accident causation, and obtaining the risk coefficient of the water conservancy project based on dynamic feature collection and normalized calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A schematic diagram of a water conservancy project risk assessment method based on artificial intelligence provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a water conservancy project risk assessment system based on artificial intelligence provided in an embodiment of the present application.

[0010] Explanation of the reference numerals: feature parameter set acquisition module 11 , quantitative analysis module 12 , threshold judgment module 13 , cause analysis module 14 , dynamic feature collection module 15 , target risk coefficient acquisition module 16 . DETAILED DESCRIPTION

[0011] This application provides an artificial intelligence-based water conservancy project risk assessment method and system to solve the technical problem that the existing technology lacks a systematic analysis of the multi-source characteristics of water conservancy projects, resulting in low risk assessment accuracy. By constructing a multi-source feature parameter set, using information value quantification functions to screen key features, combining the accident causal Bayesian network model to analyze the causal chain, and obtaining the risk coefficient of the water conservancy project based on dynamic feature collection and normalized calculation, the application achieves the technical effect of improving the accuracy of water conservancy project risk assessment.

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0013] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0014] Embodiment 1, as Figure 1 As shown, the present application provides a water conservancy project risk assessment method based on artificial intelligence, the method comprising: Step S100: traverse the multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to a first feature.

[0015] In the embodiment of the present application, multi-source engineering features cover multiple aspects of water conservancy projects, including engineering design feature set, engineering construction feature set, engineering material feature set, engineering entity feature set and behavior management feature set. Among them, the engineering design feature set involves key parameters in the engineering design stage, such as dam height, cross-sectional structure, flood discharge capacity, seismic fortification intensity, etc.; the engineering construction feature set involves key data in the construction process, such as construction technology, vibration frequency, concrete pouring quality, construction temperature and humidity, etc.; the engineering material feature set involves the physical and chemical properties of materials used in the project, such as concrete strength grade, geotechnical material permeability coefficient, steel yield strength, etc.; the engineering entity feature set covers structural monitoring and environmental parameters during the operation of water conservancy projects, such as dam settlement, seepage pressure, deformation monitoring data, etc.; the behavior management feature set involves engineering maintenance, management and personnel operation specifications, such as safety inspection records, maintenance execution status, accident handling response time, etc.

[0016] In the process of traversing the water conservancy project risk database, the data in each feature set is retrieved one by one to extract the core parameters related to the water conservancy project risk. Based on the data traversal, key parameters are extracted from different engineering feature sets to form feature parameter sets. Each feature parameter set consists of a feature name and the corresponding specific parameter value. The feature parameter set includes the first parameter corresponding to the first feature. As shown in Table 1.

[0017]

[0018] Furthermore, in the method provided in the embodiment of the application, before traversing the multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes the first parameter corresponding to the first feature, it also includes: Establish an engineering construction feature set; establish an engineering material feature set; establish an engineering entity feature set; establish a behavior management feature set; and obtain the multi-source engineering feature based on the engineering design feature set, the engineering construction feature set, the engineering material feature set, the engineering entity feature set and the behavior management feature set.

[0019] In the embodiment of the present application, a construction feature set is first formed, which includes core factors that affect the quality and stability of water conservancy project construction, such as construction methods, vibration technology, environmental factors, etc. Specific features include vibration frequency, ambient temperature during construction, rainfall during construction, and construction load changes. Among them, the vibration frequency (unit: Hz) affects the density and structural stability of the concrete; the ambient temperature during the construction period (unit: ℃), too high or too low temperature may affect the curing effect of the concrete; the rainfall during the construction period (unit: mm) affects the construction progress and concrete strength; the construction load change (unit: kN) refers to the impact of temporary loads during the construction process on the engineering structure.

[0020] Then, a feature set of engineering materials is formed, which focuses on the physical and chemical properties of materials used in water conservancy projects, which directly determine the durability and stability of the project. Specific features include concrete compressive strength, steel yield strength, geotechnical material permeability coefficient and concrete carbonization depth. Among them, concrete compressive strength (unit: MPa) measures the ability of concrete to withstand pressure; steel yield strength (unit: MPa) determines the tensile properties of the structure; geotechnical material permeability coefficient (unit: cm / s) affects the permeability stability of the dam or slope protection; concrete carbonization depth (unit: mm) affects long-term durability, and carbonization may cause steel corrosion.

[0021] Then, a feature set of engineering entities is formed. This feature set mainly includes the structural health and environmental characteristics monitored during the operation of the water conservancy project, which is used to evaluate the safety status of the project. Specific features include dam settlement, seepage pressure, structural strain and crack width. Among them, dam settlement (unit: mm / year) reflects the displacement of the dam due to foundation changes or deadweight; seepage pressure (unit: kPa) reflects the water flow pressure inside the dam or slope protection; structural strain measures the deformation state of the dam or supporting structure; crack width (unit: mm) is used to evaluate the degree of structural damage.

[0022] In addition, a behavioral management feature set is established, which involves data on construction and operation management, reflecting the effectiveness of management decisions and execution. Specific features include safety inspection execution rate and maintenance plan execution rate. Among them, the safety inspection execution rate (unit: %) measures the completion of the inspection work; the accident response time (unit: h) reflects the efficiency of handling emergencies; the maintenance plan execution rate (unit: %) measures the implementation of regular maintenance work of the project.

[0023] Finally, a complete multi-source engineering feature is formed based on the engineering design feature set, engineering construction feature set, engineering material feature set, engineering entity feature set and behavior management feature set.

[0024] Step S200: Retrieve the information value quantification function to perform a quantitative analysis of the predictive capability of the first parameter on the risk factor of the water conservancy project, and obtain the first value coefficient of the first feature.

[0025] In the embodiment of the present application, the information value quantification function is called to perform quantitative analysis on the first parameter (i.e., the specific parameter value corresponding to a certain feature) to evaluate its predictive ability for the water conservancy project risk coefficient, and finally calculate the first value coefficient of the first feature. This process analyzes the data in the water conservancy project risk database and uses information entropy, conditional entropy, and information gain to measure the contribution of the characteristic parameters to the uncertainty of the water conservancy project risk.

[0026] Specifically, firstly, the first parameter data stored in the water conservancy project risk database is extracted, and the distribution of the parameter under different water conservancy project conditions is statistically analyzed. Subsequently, the influence of the parameter on the water conservancy project risk coefficient is calculated, and its conditional probability under different risk levels is analyzed. Next, the information value quantification function is called to calculate the information gain based on the probability distribution of the feature parameter, and quantify the contribution of the feature to the water conservancy project risk prediction. Finally, the first value coefficient of the first feature is obtained.

[0027] Furthermore, in the method provided in the embodiment of the application, the expression of the information value quantification function is: ; in, The first feature The first value coefficient, Refers to the risk factor of the water conservancy project The entropy of Refers to the risk factor of the water conservancy project Take the first The probability of a value, Refers to the risk factor of the water conservancy project The total number of all values ​​of The first feature The conditional entropy of The first feature Value The probability of It means in Risk coefficient of water conservancy project under the conditions The entropy of .

[0028] In the embodiment of the present application, data is first extracted from the water conservancy project risk database, and the water conservancy project risk coefficient is calculated. Entropy , which measures the uncertainty of the risk factor ;in, Refers to the risk factor of water conservancy projects Take the first The probability of a value, Refers to the risk factor of water conservancy projects The total number of all values ​​of .

[0029] Next, calculate the first feature The conditional entropy , that is, when the characteristic parameters are known, the uncertainty of the risk coefficient of the water conservancy project still exists ,in, The first feature Value The probability of It means in Risk coefficient of water conservancy projects under The entropy of .

[0030] Then calculate the information gain ,Right now The information gain indicates that when this feature is included After that, the uncertainty of water conservancy project risk is reduced. Therefore, information gain It is the first value coefficient of the first feature, which is used to measure the role of this feature in the risk assessment of water conservancy projects.

[0031] Step S300: If the first value coefficient reaches a predetermined threshold, the first feature is added to a candidate feature factor list.

[0032] In the embodiment of the present application, the calculated first value coefficient is compared with a predetermined threshold value, wherein the predetermined threshold value is pre-set by a technical expert. When the first value coefficient is greater than or equal to the predetermined threshold value, it means that the feature corresponding to the value coefficient has a significant impact on the risk assessment of the water conservancy project and is added to the candidate feature factor list.

[0033] Furthermore, in the method provided in the embodiment of the application, before activating the accident causation Bayesian network model to perform causal analysis on the candidate feature factor list to obtain a target causal chain, and analyzing the target causal chain to obtain a target key feature factor list, it also includes: Each characteristic factor in the candidate characteristic factor list and the water conservancy project risk coefficient are used as network structure nodes; the first node and the second node in the network structure node are extracted respectively; a first directed edge is generated according to the first causal data of the first node and the second node, and a network structure directed edge is formed; a Bayesian network topology structure is constructed according to the network structure nodes and the network structure directed edges; a first predetermined probability constraint of the first node is obtained, and the first predetermined probability constraint is rendered to the Bayesian network topology structure to obtain the accident causation Bayesian network model.

[0034] Furthermore, the method provided in the application embodiment also includes: The first predetermined probability constraint includes a first priori probability and a first conditional probability table, wherein the first conditional probability table refers to the conditional probability that the target node is in any state under a predetermined parent node state.

[0035] In the embodiment of the present application, in order to construct a Bayesian network model of accident causation, a causal relationship is established between the characteristic factors and the risk coefficient of the water conservancy project based on a list of candidate characteristic factors, and a topological structure that can be used for reasoning and prediction is constructed. This method makes the risk assessment of water conservancy projects more accurate and intelligent through methods such as network structure construction, causal analysis, and probability constraint calculation.

[0036] Specifically, the feature classification and data discretization methods are first used to take the candidate feature factors and the water conservancy project risk coefficient as network structure nodes. All features are extracted from the candidate feature factor list and defined as network structure nodes together with the water conservancy project risk coefficient. The candidate feature factor list is the features screened out based on the first value coefficient calculated in the previous step, which represent the key factors that may affect the risk of water conservancy projects. In Bayesian network modeling, network structure nodes are used to represent the states of different feature variables. In order to improve the computational efficiency, the data discretization method is used to convert continuous feature variables (such as "dam body settlement") into discrete states (such as "low, medium, high settlement") to adapt to the Bayesian network calculation model.

[0037] Then, the mutual information analysis method is used to extract the first node and the second node in the network structure. When constructing a Bayesian network, it is necessary to identify causal relationships, that is, which features may affect other features or risk coefficients. For this purpose, the first node and the second node are extracted from the network structure nodes. The first node represents the antecedent factor of a certain risk factor or feature, such as "rainfall during the construction period" may affect "seepage pressure". The second node represents the consequence factor that changes due to the first node, such as "seepage pressure" may affect "dam settlement" or "water conservancy project risk coefficient". In order to quantitatively analyze the correlation between features, the mutual information analysis method is used to calculate the mutual information value between feature variables, thereby quantifying the degree of information sharing between variables and determining whether there is a significant causal relationship between the first node and the second node.

[0038] Then, the Granger causality test method is used to generate the first directed edge based on the first causal data, and the network structure directed edge is formed. In order to describe the causal relationship between the characteristic factors, the first directed edge is generated based on the first causal data, that is, the causal connection between the nodes is defined. The Granger causality test is used to analyze the time series data to determine whether the past value of a characteristic variable can effectively predict the current or future value of another characteristic variable. For example, by analyzing historical data, it can be found that "precipitation during the construction period" has a significant predictive ability for the change of "seepage pressure", so a directed edge "precipitation during the construction period → seepage pressure" is generated and included in the network structure directed edge set, and finally a complete causal network structure is formed.

[0039] Subsequently, the K2 structural learning algorithm is used to construct the Bayesian network topology based on the network structure nodes and directed edges. After determining all network nodes and the directed edges between them, the Bayesian network topology is formally constructed based on the Bayesian network modeling method. The topology of the Bayesian network is a directed acyclic graph used to represent the causal relationship between feature variables. By using the K2 structural learning algorithm, all possible network structures are traversed in a greedy search manner, and the Bayesian score of the network structure is calculated to find the optimal Bayesian network topology to ensure the correctness and efficiency of causal reasoning.

[0040] Next, the maximum likelihood estimation method is used to obtain the first predetermined probability constraint of the first node, which includes the first prior probability and the first conditional probability table. The first conditional probability table refers to the conditional probability that the target node is in any state under the predetermined parent node state. For example, if the parent nodes of a certain engineering entity quality node are "raw material quality" and "construction management level", then its first conditional probability table needs to clearly define the probability values ​​of the entity quality being in the three states of "excellent", "medium" and "poor" under different combinations of the two. In order to obtain these probabilities, the maximum likelihood estimation method is used to estimate based on the frequency of existing observed samples, thereby completing the probability definition of the node. In a Bayesian network, the state of each node is not only determined by its parent node, but also needs to meet the predetermined probability constraints, that is, the prior probability and conditional probability set based on historical data or expert experience. The first predetermined probability is any pre-set probability value, which is used to constrain the probability distribution of network nodes. In order to estimate these probabilities, the maximum likelihood estimation method is used to calculate the probability distribution of each feature variable under different conditions. For example, if historical data show that "when the rainfall during the construction period of the parent node is high, the probability that the seepage pressure of the child node is increased is 75%", the corresponding conditional probability is calculated by the MLE method and stored in the conditional probability table of the Bayesian network.

[0041] Finally, the expectation maximization algorithm is used to render the first predetermined probability constraint to the Bayesian network topology to obtain the accident causation Bayesian network model. The calculated first predetermined probability constraint is applied to the Bayesian network topology to complete the parameter learning of the Bayesian network. This process uses the expectation maximization algorithm to estimate missing data and optimize the conditional probability table. Specifically, the EM algorithm gradually optimizes the parameters of the Bayesian network by alternating the E step (Expectation) and the M step (Maximization) to make it more consistent with the actual data distribution of the risk assessment of water conservancy projects. Finally, a complete accident causation Bayesian network model is generated, which can infer and predict the risk change trend of water conservancy projects based on the input new observation data.

[0042] Step S400: activating the accident causal Bayesian network model to perform causal analysis on the candidate characteristic factor list to obtain a target causal chain, and analyzing the target causal chain to obtain a target key characteristic factor list.

[0043] In an embodiment of the present application, when activating the accident causal Bayesian network model and performing causal analysis on the list of candidate characteristic factors, firstly, based on the sensitivity analysis method, the influence of each candidate characteristic factor on the risk coefficient of the water conservancy project is evaluated, and the sensitive characteristic factors that are most sensitive to risk changes are screened out. Then, in combination with these sensitive characteristic factors, the causal path analysis of the network structure nodes is performed to identify multiple possible causal chains, that is, the causal relationship paths between different risk factors. These causal chains are then identified and analyzed to screen out the largest causal chain with the strongest influence, that is, the most important risk propagation path. Finally, the target causal chain is obtained, and its corresponding characteristic factors are included in the target key characteristic factor list.

[0044] Furthermore, in the method provided in the embodiment of the application, the accident causation Bayesian network model is activated to perform causal analysis on the candidate feature factor list to obtain a target causal chain, and the target causal chain is analyzed to obtain a target key feature factor list, and further includes: A sensitivity analysis is performed on the candidate characteristic factor list through the accident causation Bayesian network model to obtain sensitive characteristic factors; a causal path analysis of the network structure nodes is performed in combination with the sensitive characteristic factors to obtain multiple causal chains; the multiple causal chains are identified and analyzed to obtain the largest causal chain, and the characteristic factors corresponding to the largest causal chain are added to the target key characteristic factor list.

[0045] In the embodiment of the present application, the local sensitivity analysis method is first used to evaluate the list of candidate characteristic factors in the accident causation Bayesian network model to determine the degree of influence of each characteristic factor on the risk coefficient of the water conservancy project. The local sensitivity analysis is performed by univariate perturbation, that is, while keeping other variables unchanged, gradually adjusting the value of a certain characteristic factor, and observing the amplitude of change in the risk coefficient of the water conservancy project, calculating the sensitivity index of each characteristic factor, and then screening out the features with the greatest impact on the risk, and forming a list of sensitive characteristic factors. In the calculation process, the conditional probability partial derivative calculation method is adopted, that is, the contribution of the value change of each characteristic factor to the change of the risk coefficient of the water conservancy project is calculated, so as to quantify the importance of each feature. For example, if the probability of an increase in the risk coefficient of the water conservancy project exceeds 50% when the seepage pressure changes from "normal" to "high", it indicates that the seepage pressure is a highly sensitive feature, so it is included in the list of sensitive characteristic factors.

[0046] Then, combined with sensitive characteristic factors, the Markov boundary search method is used to analyze the causal path of the Bayesian network structure nodes and identify multiple possible causal chains. The causal chain refers to the risk chain that starts from a certain characteristic factor, propagates along the causal path, and ultimately affects the risk coefficient of the water conservancy project in the Bayesian network. In order to identify these paths, the direct parent node, direct child node and adjacent nodes of each sensitive characteristic factor are determined by the Markov boundary search method, so as to construct possible causal paths. For example, if the dam body settlement is a sensitive characteristic factor, its possible causal chain includes: "Construction period rainfall → Seepage pressure → Dam body settlement → Water conservancy project risk coefficient". By establishing the Markov boundary, the propagation path that leads to the change of water conservancy project risk is identified, and the causal relationship analyzed is ensured to be logical and accurate.

[0047] Then, after identifying multiple causal chains, we further analyze which causal path has the greatest impact on the risk coefficient of water conservancy projects. To this end, the Bayesian belief propagation method is used to perform probabilistic impact analysis on each causal chain, calculate the contribution of different causal chains to the risk coefficient, and select the largest causal chain with the most significant impact. Bayesian belief propagation is an inference method for probabilistic graphical models. It calculates the conditional probability of each causal chain by transmitting information to each node in the network. For example, suppose there are two causal chains: "Construction period rainfall → seepage pressure → dam body settlement → water conservancy project risk coefficient", and its impact probability is 85%; while the impact probability of "construction process deviation → dam body settlement → water conservancy project risk coefficient" is 65%. Since the former has a greater impact, it is selected as the largest causal chain for subsequent screening of key characteristic factors.

[0048] Finally, the characteristic factors corresponding to the largest causal chain are added to the target key characteristic factor list, and the conditional independence test method is used to conduct independence analysis on the characteristic factors to confirm which factors still have independent contributions to the risk coefficient of water conservancy projects after controlling other variables. The conditional independence test is used to evaluate whether a certain characteristic factor still has a significant impact on the risk coefficient of water conservancy projects after removing other interfering factors. For example, after removing the influence of "rainfall during construction period", if "seepage pressure" still significantly affects "dam settlement" and "water conservancy project risk coefficient", "seepage pressure" is confirmed as an independent key feature. Therefore, by calculating the conditional independence probability of each characteristic factor, the target key characteristic factor list is finally determined for water conservancy project risk prediction and decision support.

[0049] Step S500: Dynamically collect characteristics of the target water conservancy project based on the target key characteristic factor list to obtain a real-time characteristic parameter set.

[0050] In the embodiment of the present application, when the dynamic characteristics of the target water conservancy project are collected based on the target key characteristic factor list, the wireless sensor network (WSN) is first combined to deploy monitoring equipment such as fiber grating sensors, pressure sensors, and GNSS displacement monitoring equipment to collect real-time data on key characteristics such as dam settlement, seepage pressure, foundation soil deformation, and rainfall. Then, the monitoring data is processed by denoising, anomaly removal, and data completion using a data preprocessing algorithm to improve data quality and stability. Next, a multi-source data fusion method is used to integrate heterogeneous data such as sensor data, remote sensing images, and manual inspection data, and time synchronization, error correction, and feature matching are performed to ensure data consistency. Finally, the processed feature data is stored in the risk assessment database to form a complete set of real-time feature parameters.

[0051] Step S600: performing normalized calculation and analysis on the real-time characteristic parameter set to obtain a target risk coefficient of the target water conservancy project.

[0052] In the embodiment of the present application, when normalizing and calculating the real-time feature parameter set, the minimum-maximum normalization method is used to normalize and calculate each feature data in the real-time feature parameter set to ensure that the data scales of different features are consistent and eliminate the influence of data dimensions. All normalized feature values ​​are compressed to the interval [0, 1] so that they are calculated at the same scale to avoid affecting the calculation results due to different data units. For example, the range of dam body settlement is 0-50mm, while the range of seepage pressure is 0-100kPa. If normalization is not performed, the influence of settlement will be masked by the high value of pressure. Therefore, the normalization process makes the two have the same influence during calculation, improving the comparability of data.

[0053] Then, the target risk coefficient of the water conservancy project is calculated by using the arithmetic mean method, that is, taking the average of all normalized eigenvalues ​​to obtain the target risk coefficient of the target water conservancy project.

[0054] Furthermore, in the method provided in the embodiment of the application, after normalizing and calculating the real-time characteristic parameter set to obtain the target risk coefficient of the target water conservancy project, the method further includes: The real-time feature parameter set is normalized and weighted to obtain a first key feature weight distribution; a second key feature weight distribution is obtained based on the best-worst principle; and the target risk coefficient is corrected according to an allocation adjustment made to the first key feature weight distribution based on the second key feature weight distribution.

[0055] In the embodiment of the present application, the minimum-maximum normalization method is first used to normalize the real-time feature parameter set to eliminate the influence of different feature dimensions and ensure that the data is calculated on the same scale. Then, the coefficient of variation method is used to calculate the weight distribution of the first key feature. The coefficient of variation is used to measure the degree of fluctuation of the data and reflect the discreteness and uncertainty of the feature variable. Its calculation formula is: ,in, It is a feature The standard deviation of is its mean, No. The coefficient of variation of the features. Then the weight of the first key feature is calculated based on the calculated coefficient of variation. The calculation formula is: , the first key feature weight distribution is obtained by this method. Among them, No. The initial weight of the feature.

[0056] Next, the best and worst principle is used to calculate the weight distribution of the second key feature. This method first defines the best and worst reference values ​​of each feature in all observation sample dimensions. The feature in The normalized value of the sample is , then the optimal reference value and the worst reference value Respectively represent the best and worst normalized values ​​in the feature dimension, where , . Indicates The optimal reference value of a feature, Indicates The worst reference value for a feature.

[0057] Then, the distance of each feature relative to the best and worst features is calculated based on the Euclidean distance. The specific calculation formula is as follows: ,in, Indicates The distance between a feature and its corresponding optimal reference value, Indicates The distance between a feature and its corresponding worst reference value. Represents the total number of samples. According to the best and worst weight allocation method, the final feature weight is calculated ,in, is the optimized second key feature weight distribution. The second key feature weight distribution is obtained through this process.

[0058] Then, the weight distribution of the first key feature is optimized and adjusted based on the weight distribution of the second key feature to complete the correction of the target risk coefficient. The final feature weight calculation formula is: ,in, is the balance coefficient, which is preset by technical experts, for example, 0.6.

[0059] Finally, based on the corrected characteristic weights, the corrected target risk coefficient of the target water conservancy project is calculated. ,in, is the final target risk factor, is the final weight for each feature, is the normalized eigenvalue, is the total number of features. Through this process, the correction of the target risk coefficient is completed and the final target risk coefficient is obtained.

[0060] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects: The present application traverses multi-source engineering features in a water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to a first feature; calls an information value quantification function to perform a quantitative analysis of the first parameter's predictive ability for a water conservancy project risk factor to obtain a first value coefficient of the first feature; if the first value coefficient reaches a predetermined threshold, the first feature is added to a candidate feature factor list; activates an accident causal Bayesian network model to perform a causal analysis on the candidate feature factor list to obtain a target causal chain, and analyzes the target causal chain to obtain a target key feature factor list; dynamically collects features of a target water conservancy project based on the target key feature factor list to obtain a real-time feature parameter set; performs normalization calculation and analysis on the real-time feature parameter set, and performs arithmetic mean calculation on the normalized feature values ​​to obtain a target risk coefficient for the target water conservancy project. The present invention solves the technical problem that the prior art lacks a systematic analysis of the multi-source characteristics of water conservancy projects, resulting in low risk assessment accuracy. The present invention achieves the technical effect of improving the accuracy of risk assessment of water conservancy projects by constructing a multi-source characteristic parameter set, using an information value quantification function to screen key characteristics, analyzing the causal chain in combination with a Bayesian network model of accident causation, and obtaining the risk coefficient of the water conservancy project based on dynamic feature collection and normalized calculation.

[0061] Embodiment 2 is based on the same inventive concept as the water conservancy project risk assessment method based on artificial intelligence in the above embodiment. Figure 2 As shown, the present application provides a water conservancy project risk assessment system based on artificial intelligence, and the system and method embodiments in the embodiments of the present application are based on the same inventive concept. Among them, the system includes: The feature parameter set acquisition module 11 is used to traverse the multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to the first feature; the quantitative analysis module 12 is used to call the information value quantification function to perform a quantitative analysis of the predictive ability of the first parameter on the water conservancy project risk coefficient to obtain the first value coefficient of the first feature; the threshold judgment module 13 is used to add the first feature to the candidate feature factor list if the first value coefficient reaches a predetermined threshold; the causal analysis module 14 is used to activate the accident causal Bayesian network model to perform causal analysis on the candidate feature factor list to obtain a target causal chain, and analyze the target causal chain to obtain a target key feature factor list; the dynamic feature collection module 15 is used to perform dynamic feature collection on the target water conservancy project based on the target key feature factor list to obtain a real-time feature parameter set; the target risk coefficient acquisition module 16 is used to perform normalized calculation and analysis on the real-time feature parameter set, and perform arithmetic average calculation on the normalized feature values ​​to obtain the target risk coefficient of the target water conservancy project.

[0062] Furthermore, the system is also used to implement the following functions: Establish an engineering design feature set; establish an engineering construction feature set; establish an engineering material feature set; establish an engineering entity feature set; establish a behavior management feature set; and obtain the multi-source engineering features based on the engineering design feature set, the engineering construction feature set, the engineering material feature set, the engineering entity feature set, and the behavior management feature set.

[0063] Furthermore, the system is also used to implement the following functions: ; in, The first feature The first value coefficient, Refers to the risk factor of the water conservancy project The entropy of Refers to the risk factor of the water conservancy project Take the first The probability of a value, Refers to the risk factor of the water conservancy project The total number of all values ​​of The first feature The conditional entropy of The first feature Value The probability of It means in Risk coefficient of water conservancy project under the conditions The entropy of .

[0064] Furthermore, the system is also used to implement the following functions: Each characteristic factor in the candidate characteristic factor list and the water conservancy project risk coefficient are used as network structure nodes; the first node and the second node in the network structure node are extracted respectively; a first directed edge is generated according to the first causal data of the first node and the second node, and a network structure directed edge is formed; a Bayesian network topology structure is constructed according to the network structure nodes and the network structure directed edges; a first predetermined probability constraint of the first node is obtained, and the first predetermined probability constraint is rendered to the Bayesian network topology structure to obtain the accident causation Bayesian network model.

[0065] Furthermore, the system is also used to implement the following functions: The first predetermined probability constraint includes a first priori probability and a first conditional probability table, wherein the first conditional probability table refers to the conditional probability that the target node is in any state under a predetermined parent node state.

[0066] Furthermore, the system is also used to implement the following functions: A sensitivity analysis is performed on the candidate characteristic factor list through the accident causation Bayesian network model to obtain sensitive characteristic factors; a causal path analysis of the network structure nodes is performed in combination with the sensitive characteristic factors to obtain multiple causal chains; the multiple causal chains are identified and analyzed to obtain the largest causal chain, and the characteristic factors corresponding to the largest causal chain are added to the target key characteristic factor list.

[0067] Furthermore, the system is also used to implement the following functions: The real-time feature parameter set is normalized and weighted to obtain a first key feature weight distribution; a second key feature weight distribution is obtained based on the best-worst principle; and the target risk coefficient is corrected according to an allocation adjustment made to the first key feature weight distribution based on the second key feature weight distribution.

[0068] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0069] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0070] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. The risk assessment method of water conservancy projects based on artificial intelligence is characterized by: include: Traversing the multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to the first feature; Retrieving the information value quantification function to perform a quantitative analysis of the predictive ability of the first parameter on the risk factor of the water conservancy project, and obtaining a first value coefficient of the first feature; If the first value coefficient reaches a predetermined threshold, adding the first feature to a list of candidate feature factors; Activate the accident causation Bayesian network model to perform causal analysis on the candidate characteristic factor list to obtain a target causal chain, and analyze the target causal chain to obtain a target key characteristic factor list; Based on the target key characteristic factor list, dynamic characteristics of the target water conservancy project are collected to obtain a real-time characteristic parameter set; The real-time characteristic parameter set is subjected to normalized calculation and analysis, and the normalized characteristic values ​​are subjected to arithmetic mean calculation to obtain the target risk coefficient of the target water conservancy project.

2. According to the artificial intelligence-based water conservancy project risk assessment method of claim 1, it is characterized in that: Before traversing the multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to the first feature, the method includes: Build an engineering design feature set; Establishing a set of engineering construction features; Build a feature set of engineering materials; Establishing engineering entity feature sets; Build a behavior management feature set; The multi-source engineering features are obtained based on the engineering design feature set, the engineering construction feature set, the engineering material feature set, the engineering entity feature set and the behavior management feature set.

3. The water conservancy project risk assessment method based on artificial intelligence according to claim 1 is characterized in that: The expression of the information value quantification function is: ； in, The first feature The first value coefficient, Refers to the risk factor of the water conservancy project The entropy of Refers to the risk factor of the water conservancy project Take the first The probability of a value, Refers to the risk factor of the water conservancy project The total number of all values ​​of The first feature The conditional entropy of The first feature Value The probability of It means in Risk coefficient of water conservancy project under the conditions The entropy of .

4. The artificial intelligence-based water conservancy project risk assessment method according to claim 1 is characterized in that: Before activating the accident causation Bayesian network model to perform causation analysis on the candidate feature factor list to obtain a target causation chain, and analyzing the target causation chain to obtain a target key feature factor list, the following steps are included: Taking each characteristic factor in the candidate characteristic factor list and the water conservancy project risk coefficient as a network structure node; Respectively extracting a first node and a second node from the network structure nodes; Generate a first directed edge according to the first causal data of the first node and the second node, and form a network structure directed edge; Constructing a Bayesian network topology structure according to the network structure nodes and the network structure directed edges; A first predetermined probability constraint of the first node is obtained, and the first predetermined probability constraint is rendered to the Bayesian network topology structure to obtain the accident causation Bayesian network model.

5. The artificial intelligence-based water conservancy project risk assessment method according to claim 4 is characterized in that: The first predetermined probability constraint includes a first priori probability and a first conditional probability table, wherein the first conditional probability table refers to the conditional probability that the target node is in any state under a predetermined parent node state.

6. The artificial intelligence-based water conservancy project risk assessment method according to claim 5 is characterized in that: Activate the accident causation Bayesian network model to perform causation analysis on the candidate feature factor list to obtain a target causation chain, and analyze the target causation chain to obtain a target key feature factor list, including: Performing sensitivity analysis on the candidate feature factor list by using the accident causation Bayesian network model to obtain sensitive feature factors; Performing causal path analysis of the network structure nodes in combination with the sensitive characteristic factors to obtain multiple causal chains; The plurality of causal chains are identified and analyzed to obtain a maximum causal chain, and characteristic factors corresponding to the maximum causal chain are added to the target key characteristic factor list.

7. The artificial intelligence-based water conservancy project risk assessment method according to claim 1 is characterized in that: After normalizing and calculating the real-time characteristic parameter set to obtain the target risk coefficient of the target water conservancy project, the method further includes: Performing normalized variation weighting on the real-time feature parameter set to obtain a first key feature weight distribution; Obtaining the weight distribution of the second key feature based on the best and worst principle; The target risk factor is corrected according to the allocation adjustment of the first key feature weight allocation based on the second key feature weight allocation.

8. The artificial intelligence-based water conservancy project risk assessment system is characterized by: The system comprises: A feature parameter set acquisition module, used for traversing multi-source engineering features in the water conservancy project risk database to obtain a feature parameter set, wherein the feature parameter set includes a first parameter corresponding to a first feature; A quantitative analysis module, used to call an information value quantification function to perform a quantitative analysis on the predictive ability of the first parameter on the risk factor of the water conservancy project, and obtain a first value coefficient of the first feature; a threshold judgment module, configured to add the first feature to a candidate feature factor list if the first value coefficient reaches a predetermined threshold; A cause analysis module is used to activate the accident cause Bayesian network model to perform cause analysis on the candidate characteristic factor list to obtain a target cause chain, and analyze the target cause chain to obtain a target key characteristic factor list; A dynamic feature collection module, used to collect dynamic features of a target water conservancy project based on the target key feature factor list to obtain a real-time feature parameter set; The target risk coefficient acquisition module is used to perform normalized calculation and analysis on the real-time characteristic parameter set, and perform arithmetic mean calculation on the normalized characteristic values ​​to obtain the target risk coefficient of the target water conservancy project.

Images