Dike breach emergency rescue scheme intelligent generation and optimization control method

Abstract

The invention relates to the field of hydraulic engineering intelligent emergency rescue, and particularly discloses a dike breach emergency rescue scheme intelligent generation and optimal control method, which comprises the following steps: S1, constructing a breach plugging dynamic data bottom plate; s2, establishing a breach plugging basic data access standard; s3, accessing and storing breach dynamic sensing data; s4, constructing a breach plugging knowledge graph; s5, establishing a breach plugging special knowledge base and forming a decision model; s6, an intelligent generation method of the breach plugging scheme is researched and developed; s7, performing case analysis and prediction on the breach plugging scheme; s8, constructing an intelligent evaluation scheme of the breach plugging scheme; and S9, the breach plugging scheme is rapidly corrected. According to the method, scientificity, timeliness and reliability of breach plugging decision are effectively improved, and intelligent scheme support is provided for emergency disposal of major flood disasters.

Description

Technical Field

[0001] This invention relates to the field of intelligent emergency rescue technology for water conservancy projects, specifically a method for intelligent generation and optimization control of emergency rescue plans for dike breaches. Background Technology

[0002] China has built 328,000 kilometers of river levees, protecting the lives and property of 650 million people. However, levee breaches have always posed a serious threat to flood control. In recent years, the frequent occurrence of extreme rainstorms and floods has further increased the risk of levee breaches. Due to the deep and rapid flow of water in the breach area, the poor environment for emergency rescue construction, the complexity of sealing techniques, and the difficulty in dispatching emergency supplies, traditional manual decision-making models are no longer sufficient to meet modern flood control needs. Developing intelligent methods for generating and rapidly correcting breach sealing solutions has become a key breakthrough in solving the problem of rapidly sealing large levee breaches.

[0003] Traditional containment schemes rely on human experience and have significant limitations:

[0004] (1) The data collection methods at the breach site are limited and it is difficult to integrate multi-dimensional data such as spatial geographic information, hydrological parameters (such as flow velocity and water level), geological conditions and distribution of emergency rescue resources in real time, resulting in a lack of dynamism and comprehensiveness in the formulation of the plan.

[0005] (2) Existing sealing schemes rely on historical case analogies or simplified physical models, which cannot quickly respond to complex working conditions (such as different breach widths, water flow impact force, and surrounding environmental constraints). Furthermore, the scheme adjustments lag behind on-site changes, making it easy to miss the best sealing opportunity, resulting in low rescue efficiency and high costs.

[0006] Therefore, there is an urgent need for a technical system that can integrate diverse data in real time, intelligently generate the optimal solution, and dynamically correct it, so as to improve the scientific nature and timeliness of sealing large breaches. Summary of the Invention

[0007] This invention addresses the aforementioned problems in the prior art by providing an intelligent generation and optimization control method for emergency repair plans of dike breaches. This method effectively solves the problem of intelligent generation and correction of rapid sealing plans for large dike breaches, and effectively improves the scientificity, timeliness, and reliability of breach sealing decisions.

[0008] To achieve the above objectives, this invention proposes an intelligent generation and optimization control method for emergency rescue plans in the event of a dike breach, comprising:

[0009] S1. Construct a dynamic data base for breach sealing;

[0010] S2. Establish basic data access standards for breach sealing;

[0011] S3. Access and store dynamic data of the breach;

[0012] S4. Construct a knowledge graph for breach sealing;

[0013] S5. Establish a specialized knowledge base for breach sealing and develop a decision-making model;

[0014] S6. Develop an intelligent generation method for breach sealing solutions;

[0015] S7. Case Analysis and Prediction of Breach Sealing Schemes;

[0016] S8. Construct an intelligent evaluation scheme for breach sealing solutions;

[0017] S9. Rapid revision of the breach sealing plan.

[0018] Preferably, in S1, the specific steps of the dynamic data base plate for breach sealing are to collect and organize multi-source heterogeneous data, which includes basic water conservancy data, monitoring data, geospatial data, and cross-industry shared data.

[0019] Preferably, in S2, the specific steps for establishing the basic data access standard for breach sealing are as follows:

[0020] S21. Establish a standardized framework covering the entire process, including spatiotemporal reference, data format, classification coding, and quality control.

[0021] S22. Conduct full lifecycle management of data quality control in S21, including quality verification process, data cleaning and standardization, and quality monitoring and traceability.

[0022] S23. Implement a three-level data security classification and set up a two-dimensional access control system based on roles and organizations.

[0023] Preferably, in S3, the access methods for dynamic perception data of the breach are network interface access, file import, and database synchronization; the data modalities are structured data, semi-structured data, and unstructured data; the data storage includes the data that should be stored in the full data layer, core data layer, and theme data layer.

[0024] Preferably, in S4, the specific steps for constructing the breach sealing knowledge graph are as follows:

[0025] S41. Complete the feature values ​​of case data on dike breaches and construct a reference case library;

[0026] S42. Recommend similar cases of dike breach blocking based on collaborative weights;

[0027] S43. Construct a vertical knowledge graph for the field of dike breaching through knowledge extraction and knowledge fusion.

[0028] Preferably, in S5, the resulting decision-making model is a "case reference + rule constraint" decision-making model that integrates historical scenario library, business rule library, expert experience library, blocking scheme library and emergency plan library.

[0029] Preferably, in S6, the specific steps of the intelligent generation method for developing a breach sealing solution are as follows:

[0030] S61. Construct a named entity extraction framework based on Bert-GRU-CRF and establish a semantic mapping mechanism for materials used in flood control and emergency rescue.

[0031] S62. Assemble the generated construction method according to the four parts: breach overview, head wrapping plan, advance plan, and closure and sealing plan.

[0032] Preferably, in S7, the specific steps for case analysis and prediction of breach sealing schemes are as follows:

[0033] S71. Analyze the timing and duration of closure in historical breach cases, and the mainstream closure techniques;

[0034] S72. Establish a one-dimensional mechanized vertical blocking / hybrid blocking advance model, predict the blocking duration under different conditions, and verify it.

[0035] Preferably, in S8, the specific steps for constructing an intelligent evaluation scheme for breach sealing are as follows:

[0036] S81. Construct a two-tiered evaluation index system that covers both technical indicators and effectiveness indicators of the blockage project;

[0037] S82. Use the analytic hierarchy process (AHP) to determine the relative importance of each indicator;

[0038] S83. Develop a system indicator quantification method to quantify the characteristics of different types of indicators;

[0039] S84. An integrated intelligent evaluation algorithm is constructed by adopting the FAHP-TOPSIS coupling model, improving the fuzzy neural network and the genetic algorithm.

[0040] Preferably, in S9, the specific steps for quickly modifying the breach sealing scheme are as follows:

[0041] S91. Establish a three-level early warning mechanism of mild, moderate and severe warning, and assess the deviation between the results and the target value;

[0042] S92. Design a multi-objective optimization function and use an improved genetic algorithm to optimize the scheme; wherein, the multi-objectives include block safety, time efficiency, economy and economic applicability; the scheme includes real number encoding, fitness function design, genetic operation and convergence criterion;

[0043] S93. Perform a convergence check. If convergence is achieved, output the optimal solution. Otherwise, return to the evaluation stage of S8.

[0044] Therefore, this invention proposes an intelligent generation and optimization control method for emergency repair plans of dike breaches, the beneficial effects of which are as follows:

[0045] (1) In terms of data fusion and processing, a multi-source heterogeneous breach sealing dynamic data base system based on a hierarchical architecture was constructed. A data standardization processing method for flood control and disaster relief scenarios was proposed. An efficient data preprocessing module and service interface were developed. The unified access, fusion processing and real-time updating of multi-source data such as structured monitoring data, semi-structured documents and unstructured images were realized, providing comprehensive data support for breach disaster relief decision-making.

[0046] (2) In terms of intelligent decision-making algorithms, a vertical domain knowledge base for dike breach blocking containing historical cases was constructed based on knowledge graph technology. A similar case recommendation algorithm driven by collaborative weight was proposed. Combined with rule reasoning and deep learning technology Bert-GRU-CRF model, automatic extraction of breach features, intelligent generation of solutions and dynamic optimization were realized, which significantly improved the efficiency of emergency response.

[0047] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0048] Figure 1 This invention relates to a knowledge graph ontology model in the field of dike breach blocking, which is a method for intelligent generation and optimization control of emergency rescue plans for dike breaches.

[0049] Figure 2 This invention provides an emergency rescue plan generation template for an intelligent generation and optimization control method for emergency rescue plans in the event of a dike breach. Detailed Implementation

[0050] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of this application.

[0051] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0052] This invention provides an intelligent generation and optimization control method for emergency repair plans in the event of a dike breach, comprising:

[0053] S1. Construct a dynamic data base for breach sealing; the specific steps are to collect and organize multi-source heterogeneous data, including basic water conservancy data, monitoring data, geospatial data and cross-industry shared data.

[0054] S2. Establish basic data access standards for breach sealing;

[0055] The specific steps for establishing a basic data access standard for breach sealing are as follows:

[0056] S21. Establish a standardized framework covering the entire process, including spatiotemporal reference, data format, classification coding, and quality control.

[0057] S22. Conduct full lifecycle management of data quality control in S21, including quality verification process, data cleaning and standardization, and quality monitoring and traceability.

[0058] S23. Implement a three-level data security classification and set up a two-dimensional access control system based on roles and organizations.

[0059] S3. Access and store dynamic data of the breach;

[0060] In S3, the access methods for dynamic perception data of breaches are network interface access, file import, and database synchronization; the data modalities are structured data, semi-structured data, and unstructured data; and the data storage includes the data that should be stored in the full data layer, core data layer, and theme data layer.

[0061] S4. Construct a knowledge graph for breach sealing;

[0062] In S4, the specific steps for constructing the breach sealing knowledge graph are as follows:

[0063] S41. Complete the feature values ​​of case data on dike breaches and construct a reference case library;

[0064] S42. Recommend similar cases of dike breach blocking based on collaborative weights;

[0065] S43. Construct a vertical knowledge graph for the field of dike breaching through knowledge extraction and knowledge fusion.

[0066] S5. Establish a specialized knowledge base for breach sealing and develop a decision-making model;

[0067] In S5, the resulting decision-making model is a "case reference + rule constraint" model that integrates historical scenario databases, business rule databases, expert experience databases, blocking solution databases, and emergency plan databases.

[0068] S6. Develop an intelligent generation method for breach sealing solutions;

[0069] In S6, the specific steps for developing an intelligent generation method for breach sealing solutions are as follows:

[0070] S61. Construct a named entity extraction framework based on Bert-GRU-CRF and establish a semantic mapping mechanism for materials used in flood control and emergency rescue.

[0071] S62. Assemble the generated construction method according to the four parts: breach overview, head wrapping plan, advance plan, and closure and sealing plan.

[0072] S7. Case Analysis and Prediction of Breach Sealing Schemes;

[0073] In S7, the specific steps for case analysis and prediction of breach sealing schemes are as follows:

[0074] S71. Analyze the timing and duration of closure in historical breach cases, and the mainstream closure techniques;

[0075] S72. Establish a one-dimensional mechanized vertical blocking / hybrid blocking advance model, predict the blocking duration under different conditions, and verify it.

[0076] S8. Construct an intelligent evaluation scheme for breach sealing solutions;

[0077] In S8, the specific steps for constructing an intelligent evaluation scheme for breach sealing are as follows:

[0078] S81. Construct a two-tiered evaluation index system that covers both technical indicators and effectiveness indicators of the blockage project;

[0079] S82. Use the analytic hierarchy process (AHP) to determine the relative importance of each indicator;

[0080] S83. Develop a system indicator quantification method to quantify the characteristics of different types of indicators;

[0081] S84. An integrated intelligent evaluation algorithm is constructed by adopting the FAHP-TOPSIS coupling model, improving the fuzzy neural network and the genetic algorithm.

[0082] S9. Rapid revision of the breach sealing plan.

[0083] In S9, the specific steps for quickly modifying the breach sealing scheme are as follows:

[0084] S91. Establish a three-level early warning mechanism of mild, moderate and severe warning, and assess the deviation between the results and the target value;

[0085] S92. Design a multi-objective optimization function and use an improved genetic algorithm to optimize the scheme; wherein, the multi-objectives include block safety, time efficiency, economy and economic applicability; the scheme includes real number encoding, fitness function design, genetic operation and convergence criterion;

[0086] S93. Perform a convergence check. If convergence is achieved, output the optimal solution. Otherwise, return to the evaluation stage of S8.

[0087] Example 1

[0088] The breach sealing data baseboard stores basic water conservancy data, monitoring data, business management data, geospatial data, and cross-industry shared data.

[0089] The basic water resources data should originate from the national flood risk map, the flood and drought disaster risk survey, water conservancy project design documents, the integrated water resources map, and other shared data. Monitoring data should come from hydrological and meteorological monitoring stations, flash flood disaster monitoring and early warning platforms, related models, engineering operation management systems, and cross-industry sharing, and should be obtained through remote sensing satellites and image monitoring stations to acquire timed / real-time images and videos of breaches and surrounding areas. Business management data should come from flood control and drought relief command systems, flash flood disaster early warning systems, other flood control-related business systems, decision-making and command systems, etc. Geospatial data is divided into three levels: L1, L2, and L3, according to data accuracy and construction scope. L1 data comes from the Ministry of Water Resources' shared access and is updated in sync with the data source. L2 and L3 data come from the integrated water resources map, cross-industry sharing, and newly built data, and are updated dynamically as needed. Cross-industry shared data comes from data from other departments or sources, such as data from emergency rescue teams, emergency material warehouses, emergency material reserves, and road traffic.

[0090] To ensure that various types of data can be effectively integrated, stored, queried, and applied, data models are classified into water conservancy data models and water conservancy grid models.

[0091] The water conservancy data model is based on the operational needs of dike breach sealing. It constructs a classification system for water conservancy objects, enabling data on various water conservancy objects such as rivers, lakes, water conservancy projects, and dikes to be classified and organized according to a unified framework, facilitating subsequent data management and application. The water conservancy grid model constructs a grid-based management model based on the administrative divisions, natural watersheds, water resource functional zones, and numerical calculation requirements of the breach occurrence area. This enables grid-based linkage of water conservancy operations such as dike breach control, water resource management and allocation, and water conservancy project operation management.

[0092] Establish a data connection and aggregation channel between the data source and the data platform, and sort out the "one data, one source" principle, clarifying the data source storage type, data access method, data scope, data content, data update method and frequency, etc.

[0093] For the aggregation of basic and operational data, data from existing databases or operational systems are extracted and stored in batches using methods such as database synchronization and interface integration. Unstructured operational data from water conservancy departments at the provincial, municipal, and county levels are collected in file format, with metadata also entered. For the aggregation of monitoring data, real-time monitoring data from existing and newly built monitoring systems (rainfall, water levels, video flow measurement, engineering conditions, etc.) are aggregated using real-time data synchronization. For the aggregation of geospatial data, geospatial data such as remote sensing images, high-precision topographic maps, and 3D models are collected in file format, with spatial extent and metadata information also entered. Vector data such as underlying surface extraction results (including land use, residential areas, etc.), transportation roads, soil texture, and administrative divisions are aggregated using full data extraction. For the aggregation of cross-industry data, external data such as meteorological data and socio-economic data are downloaded and transmitted through data networks, accessed through data services, and their metadata is entered. The data aggregation process is executed by the data aggregation module of the data platform.

[0094] To support the unified management and efficient retrieval of breach sealing data, a standardized framework covering the entire process, including spatiotemporal benchmarks, data formats, classification and coding, and quality control, is established. The spatial benchmark should adopt the CGCS2000 National Geodetic Coordinate System 2000, the elevation benchmark should adopt the 1985 National Elevation Benchmark, and the time benchmark should be uniformly set to Beijing time. Non-standard time fields need to be automatically converted using scripts, and the original time zone information should be marked in the metadata. The data timestamp format should be YYYY-MM-DDHH:mm:ss to ensure time alignment of multi-source data. Based on SL / T213 "General Rules for Classification and Coding of Water Conservancy Objects," water conservancy objects such as dikes, reservoirs, and monitoring stations should be uniformly classified and coded to form a standardized classification catalog.

[0095] Analyzing different types of data, the ways to connect this data to the data platform can be summarized into three methods: network interface access, file import, and database synchronization.

[0096] Data is transmitted via communication networks, using different network communication protocols or interfaces, such as TCP / UDP protocols, data message buses, or streaming media servers. Predictive models for rainfall, water levels, and engineering conditions are accessed through dedicated interfaces. The data is already connected to a computer and stored in a database. It is then synchronized with the breach sealing data baseboard. Data is stored in data files in various formats, such as Word, Excel, PDF, CSV, or JSON, and needs to be imported into the breach sealing data baseboard.

[0097] Based on different data modalities, data is categorized into structured data, semi-structured data, and unstructured data, each using different storage schemes. Structured data, such as basic data, monitoring data, and forecasting data, is presented in tabular form and is suitable for storage in relational database systems. Semi-structured data, such as various contingency plans, emergency response plans, and other breach sealing summary reports, is presented in text form and is suitable for storage in non-relational database systems for subsequent analysis. Unstructured data, such as images and videos collected by remote sensing satellites and video points, and BIM / DEM data of dike projects, should be stored as objects in the computer system and indexed to retain retrieval methods.

[0098] Example 2

[0099] Based on the on-site information of the dike breach and the hydraulic characteristics of the breach, and integrating the information on the reserve of plugging materials provided by the logistics department, the process of generating a plugging plan was initiated. The process is as follows: First, a named entity recognition algorithm is used to extract material entities from the text information, and an entity alignment algorithm is used to achieve accurate matching with knowledge graph nodes; then, applicable construction methods are ranked based on similar case recommendation technology, and the selection of construction methods is optimized based on the plugging material data; finally, relying on knowledge graph reasoning and a pre-set template strategy, an emergency response plan that meets the construction requirements is generated.

[0100] Constructing a knowledge graph for the vertical domain of dike breach closure is a crucial step in achieving intelligent solution generation. By mining and reconstructing relevant knowledge in the dike breach closure domain, and introducing knowledge graph technology, core knowledge elements and their internal connections are efficiently managed and structured, improving knowledge utilization efficiency and providing knowledge support for enhancing the efficiency and accuracy of on-site decision-making during breach rescue. The construction process is as follows:

[0101] ① Collect textual data from various sources such as academic papers, emergency rescue manuals, construction plans, news reports, and web page materials, and refine the raw data through methods such as text correction, format standardization, and rule extraction;

[0102] ②Based on the characteristic analysis of typical breach cases, a multi-level ontology model including breach type, emergency rescue technology, and material parameters is established;

[0103] ③ Based on the characteristics of text data, design and apply corresponding entity recognition and relation extraction methods to achieve structured expression of knowledge;

[0104] ④ By using word vector space mapping and similarity calculation, entity alignment across data sources can be achieved, ensuring the relevance and consistency of heterogeneous entities;

[0105] ⑤ Choose an appropriate storage method to store the extracted entities and relationships in a structured knowledge base for efficient retrieval and visualization.

[0106] An ontology model constructs a rigorous semantic framework by defining key concepts and describing the relationships between them, thereby guiding the construction of the graph structure and enabling the graph to possess semantic correctness and reasoning capabilities. For raw texts with rich textual organization, the ontology structure needs to be designed specifically for different semantic entities. Through the ontology modeling of a dike breach closure case study, diverse and heterogeneous data such as breach morphology parameters, performance indicators of closure materials and equipment, and closure method information are structured and organized to support quantitative decision-making in the generation of closure solutions.

[0107] The knowledge of dike breach sealing is abstracted into the following entity and attribute concepts, as shown in Table 1:

[0108] Table 1 Abstract Concepts of the Graph Ontology Model

[0109]

[0110] Simultaneously, defining the relationships between concepts structures domain knowledge, thus forming a domain knowledge graph ontology model for the "dike-blocking" project. The graph construction rules are as follows: Figure 1 As shown.

[0111] The collected text data is primarily unstructured, making it impossible to directly construct triples from the text. Furthermore, the data size is relatively small, insufficient to meet the large training data requirements of deep learning models for extracting entities and relationships. Considering that concepts in the dam-blocking domain typically rely on specific contextual patterns, regular expressions are used to extract content in a specific format, supplemented by manual extraction.

[0112] The regular expression returned a list of concepts containing some relational structures, but many entity attribute relationships did not have a fixed normal form, so manual matching was performed, resulting in a total of 418 entities and 3212 triple relationships.

[0113] In the process of integrating multi-source texts, different data sources may use different expressions for the same thing. For example, "sand" and "sand soil" both refer to transported soil, and "Henan Province" and "Henan" refer to the same place name. This leads to redundant data in the knowledge graph, increasing storage overhead, reducing retrieval efficiency, and increasing the difficulty of reasoning. Therefore, it is necessary to identify and unify these duplicate entities through semantic entity fusion to avoid redundant modeling. Two models, word2vec and BERT, are used to embed semantic entities, and coreference resolution is achieved through similarity calculation.

[0114] After the graph is established, to further organize and manage the generated knowledge effectively, a suitable storage method is selected to save entities, attributes, and relationships. Neo4j graph database, which combines the advantages of an intuitive and flexible modeling process, efficient relationship querying, and broad support from downstream task ecosystems, is chosen as the storage database and will continuously provide data support for the system in subsequent solution generation processes.

[0115] In the emergency repair of dike breaches, considering that the historical scenario database is a necessary reference for the subsequent breach-closing plan, special consideration was given to parameters that could influence the decision-making process. The information included was: breach information (breach time, breach location, breach width, breach depth, breach velocity, breach flow rate, head difference, dike type, and dike construction materials); breach-closing material information (locally sourced materials, precast blocks, frame assemblies, floating caissons, and geotextiles); breach-closing construction methods (heading method, encroachment method, closure method, and airtight method); and breach-closing process information (breach-closing equipment, number of participants, timing of breach closing, and time for dike restoration).

[0116] Based on the collected and summarized characteristics of the sealing methods, the methods are divided into traditional methods and anchor array rooting-blocking collaborative construction methods according to whether they are applicable to extreme scenarios. According to different tactics, they can be divided into three categories: vertical sealing, horizontal sealing, and mixed sealing. According to the construction and emergency rescue stage (dike head protection, dike embankment advance, breach closure, sealing and airtightening), the methods are sorted and classified, and the sealing materials and construction equipment required for each method are recorded.

[0117] After a breach occurs, a list of feasible alternative construction methods is generated based on the breach conditions and material reserves. To further generate a more executable solution document, it is improved by referring to the business rules for breach sealing methods.

[0118] The construction steps of various breach closure schemes are summarized and recorded, forming a reusable, verifiable, and optimizable business rule base. The emergency response plan database system integrates standardized handling procedures for different breach levels (e.g., general, major, and extremely serious emergencies), forming a tiered response system. The database includes emergency response template schemes based on parameters such as breach size and hydraulic characteristics of the breach entrance, including dike head stabilization schemes, dike reinforcement schemes, breach closure schemes, and sealing and airtightness schemes. When generating a breach closure scheme, a complete text is generated based on the plan template.

[0119] The expert experience database system comprehensively summarizes professional knowledge in the field of flood prevention and control, transforming tacit experience into distinctive conceptual explanations. It stores granular knowledge summarized by technical personnel in practical situations, such as dike construction techniques (e.g., "moon dike," "grid dike," "dike pillow," "inverted ridge"), emergency response measures (e.g., "hanging corner," "roller curtain," "running dike"), and practical construction methods (e.g., "willow stone revetment," "pile pillow submersion"). At the upper level, a tagging system is used to associate the database with breach scenario parameters, matching the experience entries in the construction method text during solution generation, forming a data-driven collaborative decision-making model that supplements joint knowledge.

[0120] Analyzing the emergency rescue plan texts for dike breach closure in the database, the generated construction methods were assembled according to four parts: breach overview, breach sealing plan, advance plan, and closure and sealing plan. A specific emergency rescue plan template is shown below. Figure 2 As shown.

[0121] Example 3

[0122] The process of evaluating and revising dike breach sealing schemes includes four main steps: First, based on a two-layer evaluation index system, data on various indicators are collected and processed, and different indicators are converted into a unified scale using a standardization method; second, the weight of each indicator is determined using the analytic hierarchy process (AHP) to construct an index weight system; third, an evaluation model is established by improving a fuzzy neural network, and the model parameters are optimized using a genetic algorithm; finally, evaluation results are generated, including a comprehensive score and a graded determination.

[0123] The comprehensive evaluation index system covers the technical indicators of the breach engineering and the effectiveness indicators of the solution, and comprehensively evaluates the feasibility, safety and economy of the sealing solution.

[0124] The engineering and technical indicators layer focuses on evaluating the technical feasibility and applicability of the plugging scheme, and mainly includes three types of primary indicators: hydraulic condition indicators, engineering condition indicators, and implementation condition indicators.

[0125] Hydraulic condition indicators include four secondary indicators: breach flow rate, head difference, flow velocity, and water depth. These indicators directly affect the technical difficulty and method selection for closure. Engineering condition indicators include four secondary indicators: breach width, breach depth, geological conditions, and transportation conditions. These indicators determine the scale and complexity of the closure project. Implementation condition indicators include four secondary indicators: operating space, construction technology complexity, equipment requirements, and material requirements. These indicators relate to the practical feasibility of the closure plan.

[0126] The scheme effectiveness indicator layer focuses on evaluating the implementation effect and impact of the blockade scheme, mainly including four types of primary indicators: blockade effectiveness indicators, time efficiency indicators, economic indicators, and applicability indicators.

[0127] The effectiveness indicators for plugging include three secondary indicators: plugging integrity, seepage control, and stability, reflecting the technical effectiveness of the plugging scheme. The time efficiency indicators include three secondary indicators: preparation time, implementation time, and critical path, reflecting the time efficiency of the plugging scheme. The economic indicators include three secondary indicators: equipment cost, material cost, and labor cost, reflecting the economic rationality of the plugging scheme. The applicability indicators include three secondary indicators: environmental adaptability, operability, and risk controllability, reflecting the practical applicability of the plugging scheme.

[0128] By defining a quantification method for the system's indicators, the objectivity and consistency of the quantification results can be ensured. For directly quantifiable indicators, such as breach flow, head difference, and plug integrity, this embodiment uses actual measured values ​​and converts them into values ​​within the range of 0-1 through standardization.

[0129] The standardized formula is X′=(XX) min ) / (X max -X min ), where X is the original measurement value, X min and X max These are the minimum and maximum values ​​of the indicator, respectively.

[0130] For indicators of different nature, corresponding transformation functions were established to ensure the consistency between the quantitative results and the importance of the indicators. For semi-quantitative indicators, such as geological conditions and construction technology complexity, a five-level quantitative standard was established, including excellent (0.9-1.0), good (0.7-0.9), average (0.5-0.7), poor (0.3-0.5) and bad (0.1-0.3).

[0131] For each semi-quantitative indicator, detailed grading standards and scoring rules were developed, and a correction coefficient λ was introduced for adjustment to adapt to the special requirements of different engineering environments. For example, the five-level standard for geological conditions includes: excellent (hard rock, no leakage), good (firm soil layer, slight leakage), average (medium soil layer, limited leakage), poor (loose soil layer, significant leakage), and bad (silt or sandy soil, severe leakage). For qualitative indicators, such as environmental adaptability, operability, and risk controllability, expert scoring is used, and fuzzy mathematics is employed for quantification.

[0132] The specific steps include: first, inviting 7-9 experienced experts to conduct independent scoring; then, eliminating the highest and lowest scores and calculating the average score; finally, converting the qualitative evaluation into a quantitative result using a fuzzy membership function. To ensure the consistency of the evaluation, a Kendall concordance coefficient W is introduced. When W ≥ 0.7, it is considered that the experts have reached a consensus; otherwise, a second round of evaluation is required until the consensus requirement is met.

[0133] In the evaluation of dike breach sealing schemes, scientifically and rationally determining the weights of each indicator is a crucial step in ensuring the accuracy of the evaluation. The Analytic Hierarchy Process (AHP) is adopted as the core method for determining indicator weights, following these four steps: establishing a hierarchical structure model → constructing the judgment matrix corresponding to each level → hierarchical single ranking and consistency check → ranking and consistency check. Santy's 1.9 scaling method is used for quantification and assignment to determine importance, and the sum-product method is selected as the solution method.

[0134] Through the analysis of multiple engineering cases, the weight distribution of each indicator in the two-level evaluation index system was obtained.

[0135] Among the primary indicators, the weight ratio of engineering and technical indicators to scheme effectiveness indicators is 0.55:0.45, reflecting the balanced importance of the two in the comprehensive evaluation. Among the three secondary indicators of engineering and technical indicators, the weights of hydraulic conditions, engineering conditions, and implementation conditions are 0.5, 0.3, and 0.2, respectively. Among the four secondary indicators of scheme effectiveness indicators, the weights of plugging effect, time efficiency, economy, and applicability are 0.4, 0.3, 0.1, and 0.2, respectively.

[0136] A consistency check is performed on the judgment matrix to ensure the logical consistency of expert judgments. A consistency index and a random consistency ratio (CR) are calculated. When CR < 0.1, the judgment matrix is ​​considered to have satisfactory consistency; otherwise, the judgment matrix needs to be reconstructed. In practical applications, by adjusting inconsistent judgment results, the consistency ratio of each level of index is ultimately made below 0.08, ensuring the reliability of the weight calculation results.

[0137] An evaluation method based on an improved fuzzy neural network is adopted, which integrates the advantages of fuzzy mathematics and neural networks to construct an intelligent evaluation model with self-learning capabilities.

[0138] An improved fuzzy neural network was established to evaluate the effectiveness of dike breach sealing, and multiple sealing schemes were compared and evaluated.

[0139] An improved fuzzy neural network evaluation method was used to design a five-layer network structure, including an input layer, a fuzzification layer, a rule layer, a defuzzification layer, and an output layer, which realizes intelligent processing of the entire process from index input to evaluation result output.

[0140] The input layer receives standardized indicator values, with the number of nodes equal to the total number of indicators; the fuzzification layer performs fuzzification on the input indicators, converting precise values ​​into fuzzy membership degrees; the rule layer processes fuzzy rules, with the number of nodes equal to the number of fuzzy rules; the defuzzification layer converts the fuzzy results back into precise values; and the output layer generates the final evaluation results and classification.

[0141] Different membership functions are designed for different types of indicators. For quantitative indicators, a trapezoidal membership function is used, with the following form: The parameters {a,b,c,d} are automatically generated based on the statistical distribution of historical data. This membership function can effectively reflect the hierarchical characteristics and transformation process of quantitative indicators, and is suitable for fuzzy processing of indicators such as breach flow and head difference.

[0142] For qualitative indicators, Gaussian membership functions are used, with the following functional form: The parameters c and σ are determined by fitting expert evaluation data using the least squares method. The smoothing properties of the Gaussian membership function are more suitable for handling the fuzziness and uncertainty of qualitative indicators, such as environmental adaptability and operability.

[0143] The establishment of fuzzy rules is a core component of fuzzy neural networks. Based on expert knowledge and historical cases, a complete fuzzy rule base was constructed, containing evaluation rules for various combinations of indicators. For example, "If the breach flow is large and the head difference is high, the closure difficulty is high," and "If the breach width is small and the geological conditions are good, the closure difficulty is low." These rules are represented using an IF-THEN structure and are matched and activated at the rule layer, forming the basis of fuzzy inference. The initial rule base contains 128 basic rules, covering various common combinations of indicators.

[0144] Network training is a crucial step for fuzzy neural networks to achieve self-learning. An improved backpropagation (BP) algorithm was employed for network training, introducing a momentum factor α and an adaptive learning rate η(t), which significantly improved convergence speed and stability. The momentum factor α ranged from 0.6 to 0.9 to accelerate convergence and avoid local optima; the initial learning rate η0 ranged from 0.01 to 0.1, adaptively adjusted gradually during training to balance training speed and accuracy. Through training on 40 historical cases and testing on 10 validation cases, the model's average error rate was reduced to 4.8%, significantly lower than the 12.3% of traditional evaluation methods, validating the effectiveness and superiority of the improved fuzzy neural network.

[0145] To further improve the model's adaptability and generalization ability, an automatic rule base optimization mechanism is designed. This mechanism can automatically correct and supplement fuzzy rules based on newly added training cases, achieving dynamic updates and optimization of the rule base. When the prediction error of a new case exceeds a preset threshold, the system analyzes the cause of the error and generates new rules or modifies the weights of existing rules, thereby continuously improving the model's prediction accuracy and adaptability. Through this mechanism, the model can continuously learn and improve, adapting to constantly changing engineering environments and needs.

[0146] Finally, early warnings are issued for deviations from the target of the blockade, and the plan is revised accordingly.

[0147] Fuzzy neural network models involve a large number of parameters, such as membership function parameters and rule weights. The optimization of these parameters directly affects the accuracy and stability of the evaluation.

[0148] To address the problem of traditional gradient descent methods easily getting trapped in local optima, a genetic algorithm is introduced for global parameter optimization, which improves model performance and prediction accuracy.

[0149] Genetic algorithms are global optimization methods based on natural selection and genetic mechanisms. They search for optimal solutions by simulating selection, crossover, and mutation operations in biological evolution. In the parameter optimization process, model parameters are first represented using real-number encoding, and the chromosome structure of each individual is X = [x1, x2, ..., x...]. n ], where x i This represents a parameter in the model. Compared to traditional binary encoding, real number encoding offers higher precision and efficiency, making it particularly suitable for handling continuous variables such as membership function parameters.

[0150] To address the discrepancy between the assessment results and the target values, a three-tiered early warning mechanism is established, including mild, moderate, and severe warnings.

[0151] (1) Mild warning (0.1≤|D|<0.3): The parameter fine-tuning method is used to adjust the weight coefficients and non-critical parameters. For example, in the optimization of a blockage case, mild warning is mainly improved by adjusting the complexity of the construction process and the implementation time.

[0152] (2) Moderate warning (0.3≤|D|<0.5): The local optimization method is adopted to recalculate the weights of sensitive indicators and adjust key parameters. For example, in the optimization of a blockage case, the blockage effect and time efficiency indicators were adjusted first during the moderate warning stage.

[0153] (3) Severe warning (|D|≥0.5): The scheme reconstruction method is adopted to adjust and optimize the objectives and constraints, and to comprehensively optimize the scheme.

[0154] The multi-objective optimization function F = w1·f1 + w2·f2 + w3·f3 + w4·f4 is designed to comprehensively consider four aspects: safety of the blockage, time efficiency, economy and applicability. Through the multi-objective optimization function, various key indicators can be balanced during the optimization process to achieve global optimum.

[0155] (1) Safety objective function of the blockade: f1=λ1·b 11 +λ2·b 12 +λ3·b 13 , where λ i The internal weights of each indicator.

[0156] (2) Time efficiency objective function: f2=T0 / (β1·b21 +β2·b 22 ), where T0 is the standardized parameter, β i The internal weights of each indicator.

[0157] (3) Economic feasibility objective function: f3=C0 / (γ1·b 31 +γ2·b 32 +γ3·b 33 ), where C0 is the standardized parameter, γ i The internal weights of each indicator.

[0158] (4) Economic objective function: f4 = δ1·b 41 +δ2·b 42 +δ3·b 43 , where δ i The internal weights of each indicator.

[0159] An improved genetic algorithm is used to optimize the scheme, including real number encoding, fitness function design, genetic operations, and convergence criteria.

[0160] (1) Real number encoding: The scheme parameters are represented using real number encoding, and the chromosome structure is X = [x1, x2, ..., x...]. n ].

[0161] (2) Fitness function: designed as Where F(X) is the value of the multi-objective optimization function, and S(X) is the constraint satisfaction.

[0162] (3) Genetic operation: crossover probability P c =0.8, mutation probability P m =0.1, population size N=100, maximum number of iterations G max =200.

[0163] (4) Convergence criterion: The change in the optimal solution over 30 consecutive generations is less than 1% or the standard deviation of population diversity σ is less than 0.01.

[0164] Therefore, this invention provides an intelligent generation and optimization control method for emergency response plans to dike breaches. It rapidly generates a breach data base, constructs a multimodal knowledge graph of dike breaches, and intelligently generates a set of breach sealing plans under different emergency response objectives based on pattern recognition, machine learning, and deep learning methods. Subsequently, it constructs a breach sealing simulation system for important dikes, establishes an evaluation index system for dike breach sealing plans, and uses machine learning and other methods to conduct comparative analysis of multiple plans for rapid breach sealing, evaluating the comprehensive effectiveness of the sealing plans and completing the intelligent optimization and rapid correction of rapid breach sealing plans. This effectively improves the scientific rigor, timeliness, and reliability of breach sealing decisions, providing intelligent solution support for emergency response to major flood disasters.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent generation and optimization control of emergency rescue plans for dike breaches, characterized in that, include: S1. Construct a dynamic data base for breach sealing; S2. Establish basic data access standards for breach sealing; S3. Access and store dynamic data of the breach; S4. Construct a knowledge graph for breach sealing; S5. Establish a specialized knowledge base for breach sealing and develop a decision-making model; S6. Develop an intelligent generation method for breach sealing solutions; S7. Case Analysis and Prediction of Breach Sealing Schemes; S8. Construct an intelligent evaluation scheme for breach sealing solutions; S9. Rapid revision of the breach sealing plan.

2. The intelligent generation and optimization control method for emergency repair plans of dike breaches according to claim 1, characterized in that, In S1, the specific steps of the dynamic data base plate for breach sealing are to collect and organize multi-source heterogeneous data, which includes basic water conservancy data, monitoring data, geospatial data and cross-industry shared data.

3. The intelligent generation and optimization control method for emergency repair plans of dike breaches according to claim 1, characterized in that, In S2, the specific steps for establishing the basic data access standard for breach sealing are as follows: S21. Establish a standardized framework covering the entire process, including spatiotemporal reference, data format, classification coding, and quality control. S22. Conduct full lifecycle management of data quality control in S21, including quality verification process, data cleaning and standardization, and quality monitoring and traceability. S23. Implement a three-level data security classification and set up a two-dimensional access control system based on roles and organizations.

4. The intelligent generation and optimization control method for emergency repair schemes of dike breaches according to claim 1, characterized in that, In S3, the access methods for dynamic perception data of the breach are network interface access, file import, and database synchronization; the data modalities are structured data, semi-structured data, and unstructured data; the data storage includes the data that should be stored in the full data layer, core data layer, and theme data layer.

5. The intelligent generation and optimization control method for emergency repair schemes of dike breaches according to claim 1, characterized in that, In S4, the specific steps for constructing the breach sealing knowledge graph are as follows: S41. Complete the feature values ​​of case data on dike breaches and construct a reference case library; S42. Recommend similar cases of dike breach blocking based on collaborative weights; S43. Construct a vertical knowledge graph for the field of dike breaching through knowledge extraction and knowledge fusion.

6. The intelligent generation and optimization control method for emergency repair schemes for dike breaches according to claim 1, characterized in that, In S5, the resulting decision-making model is a "case reference + rule constraint" model that integrates historical scenario databases, business rule databases, expert experience databases, blocking solution databases, and emergency plan databases.

7. The intelligent generation and optimization control method for emergency repair plans of dike breaches according to claim 1, characterized in that, In S6, the specific steps for developing an intelligent generation method for breach sealing solutions are as follows: S61. Construct a named entity extraction framework based on Bert-GRU-CRF and establish a semantic mapping mechanism for materials used in flood control and emergency rescue. S62. Assemble the generated construction method according to the four parts: breach overview, head wrapping plan, advance plan, and closure and sealing plan.

8. The intelligent generation and optimization control method for emergency repair plans of dike breaches according to claim 1, characterized in that, In S7, the specific steps for case analysis and prediction of breach sealing schemes are as follows: S71. Analyze the timing and duration of closure in historical breach cases, and the mainstream closure techniques; S72. Establish a one-dimensional mechanized vertical blocking / hybrid blocking advance model, predict the blocking duration under different conditions, and verify it.

9. The intelligent generation and optimization control method for emergency repair schemes of dike breaches according to claim 1, characterized in that, In S8, the specific steps for constructing an intelligent evaluation scheme for breach sealing are as follows: S81. Construct a two-tiered evaluation index system that covers both technical indicators and effectiveness indicators of the blockage project; S82. Use the analytic hierarchy process (AHP) to determine the relative importance of each indicator; S83. Develop a system indicator quantification method to quantify the characteristics of different types of indicators; S84. An integrated intelligent evaluation algorithm is constructed by adopting the FAHP-TOPSIS coupling model, improving the fuzzy neural network and the genetic algorithm.

10. The intelligent generation and optimization control method for emergency rescue plans of dike breaches according to claim 1, characterized in that, In S9, the specific steps for quickly modifying the breach sealing scheme are as follows: S91. Establish a three-level early warning mechanism of mild, moderate and severe warning, and assess the deviation between the results and the target value; S92. Design a multi-objective optimization function and use an improved genetic algorithm to optimize the scheme; Among these, the multiple objectives include the safety of the blockade, time efficiency, economy, and economic applicability; The scheme includes real number encoding, fitness function design, genetic operations, and convergence criteria; S93. Perform a convergence check. If convergence is achieved, output the optimal solution. Otherwise, return to the evaluation stage of S8.

Images