Hydropower hub dam break path analysis and verification method driven by multi-source data

Abstract

The invention discloses a multi-source data-driven hydropower hub dam break path analysis and verification method, which belongs to the field of hydropower hub dam break path analysis, and comprises the following steps: performing entity extraction and relationship extraction based on a geological data set, a meteorological data set, a biological data set and an engineering data set, and constructing a hydropower hub break knowledge graph; respectively utilizing a path sorting algorithm and a tensor decomposition algorithm to supplement and correct an error relation and a missing relation in the hydroelectric hub outburst knowledge graph; verifying the entity relationship in the corrected hydroelectric hub outburst knowledge graph to obtain a verified hydroelectric hub outburst knowledge graph; according to the verified hydropower hub break knowledge graph, all paths from all disaster-inducing factors to dam break nodes are determined through a breadth-first search algorithm, all the paths are visualized, and hydropower hub dam break event node path analysis is completed. The method solves the problem that an existing method depends too much on artificial experience, so that the reliability of an evaluation result is low.

Description

Technical Field

[0001] This invention belongs to the field of hydropower dam failure path analysis, and particularly relates to a verification method for hydropower dam failure path analysis driven by multi-source data. Background Technology

[0002] Accurately identifying the causative factors and impact pathways of hydropower dam failures is crucial for ensuring the safe operation of these projects. Existing analytical methods have several shortcomings: traditional methods are often based on single-discipline knowledge, making it difficult to comprehensively consider the combined effects of geological, meteorological, engineering, and human factors; they have weak integration capabilities for multi-source heterogeneous data, failing to fully explore the potential relationships behind the data; and in threshold selection and impact path analysis, they often rely on human experience, lacking objectivity and accuracy, resulting in low reliability of risk assessment results. Therefore, a more scientific and comprehensive method is urgently needed to analyze the causative factors and impact pathways of hydropower dam failures. Summary of the Invention

[0003] To address the aforementioned shortcomings in existing technologies, this invention provides a multi-source data-driven method for analyzing and verifying the failure path of hydropower projects, which solves the problem of low reliability of evaluation results due to excessive reliance on human experience in existing methods.

[0004] To achieve the aforementioned objectives, the technical solution adopted by this invention is: a multi-source data-driven method for analyzing and verifying the dam failure path of a hydropower project, comprising: Obtain multi-source data related to dam failures at hydropower projects; The multi-source data is divided into geological datasets, meteorological datasets, biological datasets, and engineering datasets; Entity extraction and relation extraction are performed based on geological datasets, meteorological datasets, biological datasets and engineering datasets to obtain the final entity set and the relation information between entities. The final entity set and the relation information between entities are then transformed into graph data to construct a knowledge graph of hydropower hub failure. The erroneous and missing relationships in the hydropower hub failure knowledge graph were supplemented and corrected using the path sorting algorithm and the tensor decomposition algorithm, respectively, resulting in a corrected hydropower hub failure knowledge graph. The entity relationships in the revised hydropower hub failure knowledge graph are verified to obtain the verified hydropower hub failure knowledge graph. Based on the validated hydropower dam failure knowledge graph, a breadth-first search algorithm is used to determine all paths from each disaster-causing factor to the dam failure node, and each path is visualized to complete the node path analysis of hydropower dam failure events.

[0005] The beneficial effects of this invention are as follows: This invention breaks through the limitations of traditional hydropower dam failure analysis methods. Starting from the perspective of multi-source data integration and comprehensive analysis, it assesses the failure risk of hydropower dams, greatly expanding the breadth and depth of analysis. It abandons the traditional model that relies solely on knowledge from a single discipline, comprehensively covering geological, meteorological, biological, and human (engineering) factors, providing a more comprehensive analytical perspective. By constructing a knowledge graph to mine potential relationships in the data, it is no longer limited by the difficulty of integrating multi-source heterogeneous data, enabling a more systematic grasp of failure risk and making it applicable to various complex hydropower dam scenarios. The dam failure path quantum topology verification model (DPRQTVM), utilizing quantum entanglement encoding by a quantum encoder, can deeply understand entity-relationship interactions, explicitly modeling the interactions between entities and relationships in the path, more accurately capturing semantic relationships, and deepening the understanding of the path. The spatiotemporal decomposer can effectively process spatiotemporal information, decomposing and processing the path in both time and space dimensions, fully mining the value of spatiotemporal information, and better adapting to the characteristics of different types of paths, especially suitable for dam failure paths involving spatiotemporal changes. The topology verification network combines graph neural networks and attention mechanisms to verify path topology, effectively identifying unreasonable path structures and improving the robustness of verification results. Furthermore, the hybrid decision module makes decisions based on credibility, further enhancing the reliability of the verification results. In addition, the system architecture design possesses good adaptability and scalability, handling paths of different lengths and types. Its modular design and independent modules also facilitate its application to knowledge graph verification in other fields.

[0006] Furthermore, the entity extraction and relation extraction are specifically performed as follows: Identify the entity types related to dam failure, including: hazard-causing entity, dam failure event entity, spatiotemporal entity, attribute entity, and engineering component entity; Based on various entity types, several regular expressions of different complexities were designed, and entity extraction and relation extraction were performed on geological datasets, meteorological datasets, biological datasets and engineering datasets based on each regular expression to obtain entity sets extracted from different data and relation information between entities in each entity set. Based on the entity sets and the relationship information between entities within each entity set, the information of the same entity is integrated to obtain the final entity set and the relationship information between entities.

[0007] The beneficial effects of the above-mentioned further solutions are as follows: by identifying entity types such as disaster-causing factors and dam-break events related to dam failure, information extraction becomes more targeted; by designing regular expressions of varying complexity, data from different sources can be flexibly processed, improving the comprehensiveness of information extraction; and by integrating information on the same entity to form a complete final entity set and relational information, a comprehensive and accurate data foundation can be provided for subsequent in-depth analysis of dam-break information and decision support.

[0008] Furthermore, the verified hydropower dam failure knowledge graph is specifically as follows: Construct a quantum topological verification model for dam-break paths to verify entity relationships; The modified hydropower hub failure knowledge graph was verified using a quantum topological verification model of dam failure paths, resulting in a verified hydropower hub failure knowledge graph.

[0009] The beneficial effects of the aforementioned further approach are as follows: By constructing a quantum topological verification model of dam-break paths, the entity relationships in the revised hydropower dam failure knowledge graph can be verified from a quantum topological perspective. Utilizing the properties of quantum mechanics and topological analysis, a novel and accurate method for verifying relationships in the knowledge graph is provided. This verification method can more deeply and comprehensively examine the accuracy and completeness of relationships in the knowledge graph, helping to discover potential errors or inconsistencies, thereby improving the quality and reliability of the hydropower dam failure knowledge graph and providing more accurate and effective knowledge support for the safety assessment and risk prediction of hydropower projects.

[0010] Furthermore, the quantum topology verification model for dam break paths includes a path extraction module, a quantum encoder, a spacetime decomposer, a dynamic gating fusion module, a topology verification network, a credibility assessment module, and a hybrid decision-making module; The path extraction module is used to extract dam failure paths from the corrected hydropower hub failure knowledge graph; The quantum encoder is used to convert the dam-break path into a quantum state, encode the path through quantum entanglement, and output the quantum-encoded path. The spatiotemporal decomposer is used to decompose the quantum-encoded path into a temporal convolution branch and a spatial relation branch, thereby obtaining temporal features and spatial features, respectively. The dynamic gating fusion module is used to fuse temporal and spatial features and output spatiotemporal fusion features; The topology verification network is used to construct a topology graph based on spatiotemporal fusion features, calculate the topology vulnerability index of the topology graph, and process and verify the topology graph through graph neural networks and attention mechanisms to output the credibility assessment results of the dam failure path. The credibility assessment module is used to quantify the credibility assessment results of the topological vulnerability index and the dam failure path, and obtain the quantification results. The hybrid decision-making module is used to make decisions based on the quantification results and output verification results.

[0011] The beneficial effects of the above-mentioned further scheme are as follows: The quantum topology verification model of dam failure path consists of multiple modules such as path extraction and quantum encoding. Each module works in concert. First, the dam failure path is extracted and the information processing efficiency is improved by quantum encoder with quantum state and entanglement encoding. Then, the spatiotemporal features are accurately extracted and fused by spatiotemporal decomposer and dynamic gating fusion module. Next, the topology verification network combines graph neural network and attention mechanism to deeply analyze the topology structure. Finally, the credibility assessment module and hybrid decision module perform quantitative assessment and decision-making, thereby comprehensively and accurately verifying the dam failure path in the hydropower hub failure knowledge graph and ensuring the quality and reliability of the knowledge graph.

[0012] Furthermore, the expression for the quantum-encoded path is:

[0013]

[0014]

[0015]

[0016] in, The path after quantum encoding; for The path states of the layered quantum circuit are connected to each layer through tensor products; It is an exponential function with the natural constant as its base; The imaginary unit; For the first Hermitian matrix dynamically generated by layered quantum circuits; To correct the linear unit; For the first Trainable qubit parameters of layered quantum circuits; For the first Normalized dam-break path of layered quantum circuits; It is the tensor product; It is the hyperbolic tangent function; For the first Trainable qubit parameters of +1 layer quantum circuit; For the first Normalized dam-break path of +1 layer quantum circuit; For layer index; It is a diagonal phase matrix; diagonal elements; For transpose; For the first Learnable entanglement gate parameters for layered quantum circuits; For the first Layered quantum circuits are projected onto classical features activated by ReLU; For the first Layered quantum circuits project classical features through tanh activation; For the first Layered quantum circuit path position encoding unitary matrix; For the first The dam-break path of layered quantum circuits; It is an L2 norm; It is a multilayer perceptron; This is a constructor for diagonal matrices.

[0017] The beneficial effects of the above-mentioned further scheme are as follows: the quantum-encoded path expression achieves efficient information encoding by means of quantum states, tensor products, etc., flexibly adapts to data by utilizing numerous trainable parameters, extracts nonlinear features through activation functions and multilayer perceptrons, and provides support for topology construction by combining special matrix operations. It can effectively improve the information processing capability of dam break paths, optimize the encoding effect, capture complex relationships, and assist in topology analysis.

[0018] Furthermore, the expression for decomposing the quantum-encoded path into a temporal convolution branch and a spatial relation branch is as follows:

[0019]

[0020] in, Spatiotemporal characteristics; For Long Short-Term Memory networks, it handles temporal dependencies; The real part of the quantum path encoding reflects the causal strength; The path after quantum encoding; This is an outer product operation; For graph attention networks, spatial relationships between nodes are handled; The imaginary part, which encodes the quantum path, reflects the phase relationship; The Sigmoid function compresses the weights to the (0,1) interval. These are trainable vectors; To capture global time series patterns by averaging temporal features; As a time characteristic, input the real part of the quantum state; To maximize spatial features and highlight key areas; For spatial characteristics, input the imaginary part of the quantum state; For dynamic coupling weights; This is the normalized cross-correlation function; For transpose; It is an L2 norm.

[0021] The beneficial effects of the above-mentioned further scheme are as follows: by integrating long short-term memory networks, graph attention networks, etc., this expression can effectively decompose and extract the spatiotemporal features of the quantum-encoded path, and use special functions and operations to adaptively adjust weights and measure correlations. It can capture the time series dependence and spatial location correlation during the dam failure process, and comprehensively and deeply show the spatiotemporal dynamic characteristics of the dam failure path, providing key information support for the verification of the hydropower hub failure knowledge graph and the dam failure risk analysis.

[0022] Furthermore, the expression for the spatiotemporal fusion feature is:

[0023] in, It features spatiotemporal fusion; The Sigmoid function compresses the weights to the (0,1) interval. It is a time-related feature; Spatial features; These are trainable weights; for and The splicing result; To pass Calculated weights; To smooth the ReLU function and ensure the output is positive; These are trainable weights; This is an outer product operation; For Hadema; To pass Calculate the nonlinear coupling coefficients.

[0024] The beneficial effects of the above-mentioned further scheme are as follows: the comprehensive feature expression can adaptively fuse temporal and spatial features through trainable weights, special activation functions and operations. The Sigmoid function and softplus function respectively ensure that the weights have a reasonable range of values ​​and that the output is positive. Operations such as outer product and Hadamard product deeply explore the correlation between features and comprehensively integrate the spatiotemporal information of dam break path, providing more accurate and richer feature expressions for subsequent topology verification, and improving the accuracy of understanding and analyzing the complex process of dam break.

[0025] Furthermore, the expressions for the topological vulnerability index and the credibility assessment results of the dam failure path are as follows:

[0026]

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033] in, This is a topological vulnerability indicator. It is a topological graph; For connected components The diameter; For branches; The total number of vertices; It is a first-order Betti number; For vertex set; It is an edge set; For the first Features of the vertices of a topological graph; For the first Spatiotemporal fusion features corresponding to vertices of a topological graph; For the first The topological graph vertex and the first A topological graph with vertices and edges; This is an indicator function; its value is 1 if the condition is met, and 0 otherwise. For the first Spatiotemporal fusion features corresponding to vertices of a topological graph; It is an L2 norm; Distance threshold; To take the median of the upper triangular portion of the distance matrix, excluding the diagonal; The results of the credibility assessment of the dam failure path; Normalization factor; This represents the total number of local subgraphs. For the first The number of Betti numbers corresponding to each local subgraph; This is a multi-head attention mechanism; For the first Feature representation of a local subgraph; A feature representation of the global graph; It is a natural constant; The Betti number calculation function is used for the homology group matrix; For the first The homology group matrix corresponding to each local subgraph.

[0034] The beneficial effects of the above-mentioned further scheme are as follows: By constructing a topological graph and combining features such as connectivity and ring structure, the above expression accurately generates topological edges using distance thresholds. It then leverages a multi-head attention mechanism to integrate local and global features, obtaining a dam failure path reliability assessment result. This allows for a systematic and quantitative evaluation of topological stability and dam failure path reliability from a graph theory perspective, providing a scientific basis for the risk analysis of hydropower project failures and enhancing the understanding and control of complex dam failure scenarios.

[0035] Furthermore, the expression for the quantization result is:

[0036] in, To quantify the results; It is a function type; This is a topological vulnerability indicator. It is a topological graph; It is the hyperbolic tangent function; The results represent the credibility assessment of the dam failure path.

[0037] The beneficial effects of the above-mentioned further scheme are as follows: The quantitative result expression integrates the topological vulnerability index and the dam failure path reliability assessment results, and uses the S-shaped function and hyperbolic tangent function to process and normalize the two, so as to comprehensively quantify the information on topological stability and dam failure path reliability, provide a concise and unified numerical result for dam failure risk assessment, facilitate intuitive comparison and analysis of the risk level of different dam failure scenarios, and provide clearer data support for the safety decision of hydropower projects.

[0038] Furthermore, the expression for the verification result is:

[0039] in, To verify the results; The path is valid; The lower limit of the quantization result used to determine the validity of the path; The path portion is valid; The upper limit of the quantization result for determining an invalid path; The path is invalid. To quantify the results.

[0040] The beneficial effects of the above-mentioned further scheme are as follows: By setting different threshold ranges for the quantification results, the verification result expression presents the complex dam failure path verification situation in three clear categories: "effective", "partially effective" and "ineffective", making the verification results intuitive and easy to understand. This facilitates relevant personnel to quickly judge the reliability of the dam failure path of the hydropower project, providing a clear and direct basis for subsequent decisions (such as engineering reinforcement, risk warning, etc.), and improving the efficiency and accuracy of decision-making. Attached Figure Description

[0041] Figure 1 This is a flowchart of the method of the present invention.

[0042] Figure 2 This is a system architecture diagram of the DPRQTVM quantum topology verification model for dam break paths in an embodiment of the present invention.

[0043] Figure 3 This is a knowledge graph path diagram showing how meteorological factors lead to dam failure in an embodiment of the present invention.

[0044] Figure 4 This is a knowledge graph path diagram showing how geological factors lead to dam failure in an embodiment of the present invention.

[0045] Figure 5 This is a knowledge graph path diagram illustrating how biological factors lead to dam failure in an embodiment of the present invention.

[0046] Figure 6 This is a knowledge graph path diagram showing how human (engineering) factors can cause dam failures in an embodiment of the present invention. Detailed Implementation

[0047] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0048] like Figure 1 As shown, in one embodiment of the present invention, a multi-source data-driven method for verifying the failure path analysis of hydropower projects includes: Obtain multi-source data related to dam failures at hydropower projects; The multi-source data is divided into geological datasets, meteorological datasets, biological datasets, and engineering datasets; Entity extraction and relation extraction are performed based on geological datasets, meteorological datasets, biological datasets and engineering datasets to obtain the final entity set and the relation information between entities. The final entity set and the relation information between entities are then transformed into graph data to construct a knowledge graph of hydropower hub failure. The erroneous and missing relationships in the hydropower hub failure knowledge graph were supplemented and corrected using the path sorting algorithm and the tensor decomposition algorithm, respectively, resulting in a corrected hydropower hub failure knowledge graph. The entity relationships in the revised hydropower hub failure knowledge graph are verified to obtain the verified hydropower hub failure knowledge graph. Based on the validated hydropower dam failure knowledge graph, a breadth-first search algorithm is used to determine all paths from each disaster-causing factor to the dam failure node, and each path is visualized to complete the node path analysis of hydropower dam failure events.

[0049] In this embodiment, web crawling technology is used to obtain multi-source data related to the hydropower dam failure from various professional websites, academic databases, government report websites and other data sources. At the same time, various academic documents, engineering books and other materials are collected using document management software. Then, the collected data is manually classified into four categories: geological data, meteorological data, biological data and human (engineering) data.

[0050] Data Acquisition: A simple web crawler was written using Scrapy to analyze the webpage structure of professional hydropower engineering websites and locate HTML elements containing data related to hydropower projects. Historical meteorological data, including rainfall and temperature, for the areas where hydropower projects are located were obtained from water conservancy department websites, and data on geological structures and seismic activity were obtained from geological survey agency websites. Simultaneously, literature management software was used to connect to academic databases (CNKI, Web of Science, Google Scholar) and search for relevant academic literature using keywords such as "hydropower project failure," "disaster-causing factors," "hydropower project safety," "dam failure mechanism," "geological disasters and hydropower projects," "impact of meteorological factors on hydropower projects," "human factors leading to hydropower project risks," "operational and management errors of hydropower projects," "causes of dam structure failure," "impact of floods on hydropower projects," and "response of hydropower projects under seismic action," which was then imported into the software for management.

[0051] Data Classification: The collected data is classified. Volcanic activity, soil and rock properties, geological structure, and dam foundation stability are classified as geological data; rainfall, temperature changes, extreme weather events, floods, and other meteorological variables are classified as meteorological data; data on the impact of plant activity, animal activity, microbial activity, algal influence, and insect erosion on the project are classified as biological data; and data on design defects, construction problems, management oversights, insufficient maintenance, operational errors, third-party sabotage, lack of monitoring, inadequate emergency response, economic factors, and policy factors are classified as human (engineering) data.

[0052] The entity extraction and relation extraction are performed as follows: Identify the entity types related to dam failure, including: hazard-causing entity, dam failure event entity, spatiotemporal entity, attribute entity, and engineering component entity; Based on various entity types, several regular expressions of different complexities were designed, and entity extraction and relation extraction were performed on geological datasets, meteorological datasets, biological datasets and engineering datasets based on each regular expression to obtain entity sets extracted from different data and relation information between entities in each entity set. Based on the entity sets and the relationship information between entities within each entity set, the information of the same entity is integrated to obtain the final entity set and the relationship information between entities.

[0053] In this embodiment, based on the data collected and classified in the previous step, the entity types related to hydropower dam failure are first determined, including disaster-causing entity, dam failure event entity, spatiotemporal entity, attribute entity, and engineering component entity. Then, a set of regular expressions is used to extract various entities and relationships from the preprocessed data. The extracted entities and relationships are then stored in the form of a graph to construct a hydropower dam failure knowledge graph. Finally, a knowledge graph completion algorithm is used to modify, predict, and supplement erroneous and missing relationships in the graph.

[0054] Construct a set of regular expressions for multi-source data: By designing regular expression patterns of varying complexity, we can handle large amounts of complex text structures and semantic relationships to achieve entity name extraction. Table 1 shows examples of regular expressions.

[0055] Table 1

[0056] ([Disaster-causing factor].+?): Matches disaster-causing factors such as geology, meteorology, biology, and man-made engineering (e.g., "rainstorm" or "abnormal permeability coefficient").

[0057] ([dam failure event].+?): Matches dam failure events (such as "overburden dam failure" or "piping failure").

[0058] ([Time point / time period].+?): Matches time descriptions (e.g., "July 2023" "lasts 72 hours").

[0059] ([facilities].+?): Matches attribute components (such as "dam foundation" or "spillway").

[0060] Entity extraction and relation extraction: Based on regular expressions, various entities (hazardous factors, dam failure events, spatiotemporal factors, attributes) related to dam failure are extracted from geological, meteorological, biological, and anthropogenic (engineering) data texts, along with the influence relationships between entities. After entity extraction, matching rules based on entity name and attributes are used to determine whether entities from different sources belong to the same entity. If two entities have completely identical names, they are considered the same entity. For the same entity, relevant information from different texts is integrated. For entities with similar but different names, TF-IDF (Term Frequency-Inverse Document Frequency) combined with cosine similarity is used to calculate the similarity between entities. First, the text containing entities is converted into vector form, and the weight of each word is calculated using TF-IDF to reflect its importance in the text. Then, cosine similarity is used to calculate the cosine value of the angle between two vectors; the closer this value is to 1, the higher the similarity between the two entities. A threshold (0.8) is set; when the similarity is higher than the threshold, further manual review is conducted to determine whether they are the same entity; if the similarity is lower than the threshold, they are determined to be different entities.

[0061] Knowledge Graph Fusion and Optimization: Errors and omissions in identifying relationships between entities (hazards, dam failure events, spatiotemporal factors, attributes).

[0062] Prioritize paths with three or fewer entity relationships, as these concise paths are more prone to missing or incorrect relationships. Simultaneously, manually review each relationship path using domain knowledge, actual data, and logical rules to identify all problematic paths. For erroneous relationship paths, use the Path Ranking Algorithm (PRA); for missing relationship paths, use the Tensor Decomposition Algorithm (RESCAL model). Combining these two algorithms will modify and supplement the paths.

[0063] Determine the optimal weights between the PRA and RESCAL algorithms.

[0064] There are two quantification standards for weight allocation in Path Ranking Algorithm (PRA) and Tensor Decomposition Algorithm (RESCAL). One is based on dataset characteristics, such as datasets rich in path features and datasets with complex relationships and implicit semantics. The other is based on experimental validation. Taking the FB15k-237 dataset as an example, the dataset is divided into training, validation, and test sets. Nine weight combinations are set, and the Mean Reciprocal Ranking (MRR) and Hits@K are used as evaluation metrics. PRA and RESCAL models are trained separately, and the performance of each combination is evaluated on the validation set. The optimal combination is then evaluated on the test set to determine the final weights. After determining the weights, the PRA and RESCAL models are used to comprehensively process the remaining erroneous and missing relationships in the knowledge graph. First, PRA is used to correct the marked erroneous relationships in the knowledge graph one by one, finding alternative relationships based on the calculated path feature similarity and updating the knowledge graph. Then, the RESCAL model is used to scan the knowledge graph, predict and supplement potential missing relationships, and add the predicted relationships to the knowledge graph to improve it.

[0065] The weight selection method is as follows: ① Weight allocation based on dataset characteristics: Datasets rich in path features: If there are many obvious and traceable paths between entities in the knowledge graph, PRA is more effective. Because PRA can uncover potential relationships between entities through path search, it can be assigned a higher weight, 0.7, while RESCAL is assigned 0.3.

[0066] Datasets with complex relationships and implicit semantics: When the relationships in a knowledge graph are complex and contain a lot of implicit semantics, making them difficult to describe using simple paths, the RESCAL algorithm, which can learn the latent representations of entities and relationships, shows a clear advantage. In this case, RESCAL is assigned a weight of 0.7, and PRA is assigned a weight of 0.3.

[0067] ② Weight allocation based on experimental verification: Prepare a knowledge graph dataset and divide it into a training set, a validation set, and a test set.

[0068] Set different weight combinations: (0.1,0.9), (0.2,0.8), (0.3,0.7), (0.4,0.6), (0.5,0.5), (0.6,0.4), (0.7,0.3), (0.8,0.2), (0.9,0.1).

[0069] ③ Determine the weights: The dataset used is the common knowledge graph dataset FB15k-237, which contains 14,541 entities, 237 relations, and over 175,000 triples. The mean reciprocal rank (MRR) and Hits@K were used as evaluation metrics. MRR is the average of the reciprocal rank of all query triples, and Hits@K represents the proportion of query triples whose correct answer ranks within the top K positions of the prediction list. In terms of experimental procedures, the dataset was first divided into training, validation, and test sets in an 8:1:1 ratio. Then, the PRA and RESCAL models were trained separately. Different weight combinations were then tried, and the performance of each combination was evaluated on the validation set. Finally, the weight combination with the best performance on the validation set was selected and evaluated on the test set to determine the final weights. Examples of the weights are shown in Table 2.

[0070] Table 2

[0071] The experimental results show that when the PRA weight is 0.4 and the RESCAL weight is 0.6, the model achieves optimal performance on both the validation and test sets. This indicates that, under this dataset and task scenario, this weight combination provides the best knowledge graph completion performance for the fusion model.

[0072] The verified hydropower dam failure knowledge graph is as follows: Construct a quantum topological verification model for dam-break paths to verify entity relationships; The modified hydropower hub failure knowledge graph was verified using a quantum topological verification model of dam failure paths, resulting in a verified hydropower hub failure knowledge graph.

[0073] like Figure 2 As shown, the quantum topology verification model for dam break paths includes a path extraction module, a quantum encoder, a spacetime decomposer, a dynamic gating fusion module, a topology verification network, a credibility assessment module, and a hybrid decision module. The path extraction module is used to extract dam failure paths from the corrected hydropower hub failure knowledge graph; The quantum encoder is used to convert the dam-break path into a quantum state, encode the path through quantum entanglement, and output the quantum-encoded path. The spatiotemporal decomposer is used to decompose the quantum-encoded path into a temporal convolution branch and a spatial relation branch, thereby obtaining temporal features and spatial features, respectively. The dynamic gating fusion module is used to fuse temporal and spatial features and output spatiotemporal fusion features; The topology verification network is used to construct a topology graph based on spatiotemporal fusion features, calculate the topology vulnerability index of the topology graph, and process and verify the topology graph through graph neural networks and attention mechanisms to output the credibility assessment results of the dam failure path. The credibility assessment module is used to quantify the credibility assessment results of the topological vulnerability index and the dam failure path, and obtain the quantification results. The hybrid decision-making module is used to make decisions based on the quantification results and output verification results.

[0074] In this embodiment, based on the constructed knowledge graph, the DPRQTVM quantum topological verification model for dam break paths is introduced to conduct in-depth analysis and verify the information on entities and relationships between entities in the text.

[0075] A quantum topological verification model (DPRQTVM) for dam-break paths was constructed to verify the aforementioned knowledge graph. By introducing techniques such as quantum encoding, spacetime decomposition, and topological verification, the accuracy, robustness, and adaptability of knowledge graph verification are improved. This model can more effectively handle different types of paths, playing a particularly important role in the verification of critical areas such as dam-break paths.

[0076] The expression for the quantum-encoded path is:

[0077]

[0078]

[0079]

[0080] in, The path after quantum encoding; for The path states of the layered quantum circuit are connected to each layer through tensor products; It is an exponential function with the natural constant as its base; The imaginary unit; For the first Hermitian matrix dynamically generated by layered quantum circuits; To correct the linear unit; For the first Trainable qubit parameters of layered quantum circuits; For the first Normalized dam-break path of layered quantum circuits; It is the tensor product; It is the hyperbolic tangent function; For the first Trainable qubit parameters of +1 layer quantum circuit; For the first Normalized dam-break path of +1 layer quantum circuit; For layer index; It is a diagonal phase matrix; diagonal elements; For transpose; For the first Learnable entanglement gate parameters for layered quantum circuits; For the first Layered quantum circuits are projected onto classical features activated by ReLU; For the first Layered quantum circuits project classical features through tanh activation; For the first Layered quantum circuit path position encoding unitary matrix; For the first The dam-break path of layered quantum circuits; It is an L2 norm; It is a multilayer perceptron; This is a constructor for diagonal matrices.

[0081] The expression for decomposing the quantum-encoded path into temporal convolutional branches and spatial relational branches is as follows:

[0082]

[0083] in, Spatiotemporal characteristics; For Long Short-Term Memory networks, it handles temporal dependencies; The real part of the quantum path encoding reflects the causal strength; The path after quantum encoding; This is an outer product operation; For graph attention networks, spatial relationships between nodes are handled; The imaginary part, which encodes the quantum path, reflects the phase relationship; The Sigmoid function compresses the weights to the (0,1) interval. These are trainable vectors; To capture global time series patterns by averaging temporal features; As a time characteristic, input the real part of the quantum state; To maximize spatial features and highlight key areas; For spatial characteristics, input the imaginary part of the quantum state; For dynamic coupling weights; This is the normalized cross-correlation function; For transpose; It is an L2 norm.

[0084] The expression for the spatiotemporal fusion feature is:

[0085] in, It features spatiotemporal fusion; The Sigmoid function compresses the weights to the (0,1) interval. It is a time-related feature; Spatial features; These are trainable weights; for and The splicing result; To pass Calculated weights; To smooth the ReLU function and ensure the output is positive; These are trainable weights; This is an outer product operation; For Hadema; To pass Calculate the nonlinear coupling coefficients.

[0086] The expressions for the topological vulnerability index and the credibility assessment results of the dam failure path are as follows:

[0087]

[0088]

[0089]

[0090]

[0091]

[0092]

[0093]

[0094] in, This is a topological vulnerability indicator. It is a topological graph; For connected components The diameter; For branches; The total number of vertices; It is a first-order Betti number; For vertex set; It is an edge set; For the first Features of the vertices of a topological graph; For the first Spatiotemporal fusion features corresponding to vertices of a topological graph; For the first The topological graph vertex and the first A topological graph with vertices and edges; This is an indicator function; its value is 1 if the condition is met, and 0 otherwise. For the first Spatiotemporal fusion features corresponding to vertices of a topological graph; It is an L2 norm; Distance threshold; To take the median of the upper triangular portion of the distance matrix, excluding the diagonal; The results of the credibility assessment of the dam failure path; Normalization factor; This represents the total number of local subgraphs. For the first The number of Betti numbers corresponding to each local subgraph; This is a multi-head attention mechanism; For the first Feature representation of a local subgraph; A feature representation of the global graph; It is a natural constant; The Betti number calculation function is used for the homology group matrix; For the first The homology group matrix corresponding to each local subgraph.

[0095] The expression for the quantization result is:

[0096] in, To quantify the results; It is a function type; This is a topological vulnerability indicator. It is a topological graph; It is the hyperbolic tangent function; The results represent the credibility assessment of the dam failure path.

[0097] The expression for the verification result is:

[0098] in, To verify the results; The path is valid; The lower limit of the quantization result used to determine the validity of the path; The path portion is valid; The upper limit of the quantization result for determining an invalid path; The path is invalid. To quantify the results.

[0099] In this embodiment, the verification model operation steps are as follows: The input path is fed into a quantum encoder for quantum encoding to obtain the quantum-encoded path. The quantum-encoded path is then fed into a spatiotemporal decomposer, where temporal and spatial features are obtained through temporal convolution and spatial relation branches, respectively. These temporal and spatial features are then fed into a dynamic gating fusion module for fusion to obtain comprehensive features. The comprehensive features are then fed into a topology verification network to construct a topology graph, which is then processed and verified to obtain the path's credibility assessment result. The credibility assessment result is then quantified using a credibility assessment module. Based on the assessed credibility, a hybrid decision module makes a decision and outputs the verification result.

[0100] Specific examples are as follows: Case 1. Verification of Dam Failure Path Triggered by Rainfall Input path: Extreme rainfall → Water level exceeds limit → Overtopping → Dam weakening → Dam failure; Quantum encoding process:

[0101] Run output: Quantum encoding time: 12.3ms; Spatiotemporal feature decomposition: Time component: [0.82, 0.76, 0.91] (L2-norm=1.32); Spatial components: [0.68, 0.73, 0.85] (L2-norm=1.24); Dynamic coupling weights: =0.79; Topology verification: Betti number: =3, =1; Persistent coherence score: 0.87; Final credibility: 0.83 → Quantum decision VALID.

[0102] Case 2. Verification of Earthquake-Induced Dam Failure Path Input path: High earthquake intensity → Liquefaction → Landslide → Impact → Dam failure; Spatiotemporal decomposition process: ; Run output: Quantum encoding time: 9.7ms; Spatiotemporal feature decomposition: Time component: [0.91, 0.88, 0.95] (L2-norm=1.58); Spatial components: [0.72, 0.81, 0.89] (L2-norm=1.42); Dynamic coupling weights: =0.85; Topology verification: Betti number: =2, =2; Persistent coherence score: 0.78; Final credibility: 0.81 → Quantum decision VALID.

[0103] Case 3. Validation of dam failure pathways triggered by biological factors Input path: termite nest → piping → structural erosion → collapse → dam failure; Topology verification process: ; Run output: Quantum encoding time: 11.2ms; Spatiotemporal feature decomposition: Time component: [0.63, 0.59, 0.72] (L2-norm=1.13); Spatial components: [0.81, 0.77, 0.83] (L2-norm=1.39); Dynamic coupling weights: =0.62; Topology verification: Betti number: =1, =1; Persistent cohomology score: 0.64; Final confidence level: 0.68 → Classical revision PARTIAL.

[0104] All path results (summary table, as shown in Table 3).

[0105] Table 3

[0106] in conclusion: The model achieves an accuracy rate of over 90% in validating meteorological / geological pathways, demonstrating that quantum encoding can effectively capture physical regularities; biological factor pathways often require classical correction (confidence level 0.4-0.7), reflecting their nonlinear characteristics.

[0107] 38% of the human (engineering) factor paths were classified as INVALID, mainly due to: Low quantum coherence (average <0.3); Spatiotemporal coupling coefficient anomaly ( >1.2 or <0.4); Low-reliability paths are manually reviewed.

[0108] In this embodiment, the dam failure event entity is used as the target node. A breadth-first search algorithm is employed in the knowledge graph to determine all paths from each hazard-causing factor to the dam failure node. Visualization technology is then used to display all paths from the hazard-causing factors to the dam failure event node, as well as the relationships between them, in an intuitive graph format, completing the entire analysis process.

[0109] like Figure 3 The diagram shown is a knowledge graph path diagram of meteorological factors leading to dam failure, including: 1. Rainfall characteristics trigger dam break paths High extreme rainfall threshold → Reservoir water level exceeds limit → Overtopping situation → Water flow continuously erodes the dam crest, weakening the dam structure → Dam body cannot withstand water pressure → Dam failure; High rainfall intensity → surge in surface runoff → large volume of water rapidly impacts the dam body, increasing dam pressure → local structural damage to the dam body, reduced stability → dam failure; Prolonged rainfall duration → continuous infiltration → increased permeability coefficient → piping formation → continuous expansion of piping, eroding the interior of the dam body → dam structure instability → dam failure.

[0110] 2. Temperature changes trigger dam failure pathways Multiple freeze-thaw cycles → cracking of dam materials → increased permeability coefficient → piping formation → piping damages the dam foundation, leading to structural imbalance → dam failure; Extreme high temperature → thermal expansion and cracking of concrete → failure of anti-seepage layer → concentrated leakage → long-term water erosion inside the dam body, resulting in reduced strength → dam failure.

[0111] 3. Pathways for dam failures triggered by extreme weather events Strong typhoon winds → storm surge → abnormal rise in reservoir water level → overtopping situation → water flow erodes the dam crest and foundation, causing erosion → damage to dam structure → dam failure. Thunderstorms and heavy rainfall → flash floods → localized erosion of the dam body → localized structural damage to the dam body, affecting overall stability → dam failure; Extreme snowfall → snowmelt floods → sudden rise in reservoir water level → overtopping situation → water flow exerts excessive pressure on the dam body, exceeding the bearing limit → dam failure. Large hailstone diameter → spalling of concrete protective layer → accelerated corrosion of steel bars → weakening of dam structure strength → dam unable to withstand water pressure and external forces → dam failure; High wind speeds during sandstorms → wear and tear on surface materials → increased porosity → accelerated seepage → abnormal seepage inside the dam body, structural damage → dam failure; Freezing rain increases the self-weight of the structure → increases the stress on the dam foundation → uneven settlement → deformation of the dam structure, resulting in cracks → dam failure; High wind speed in a straight line → control system failure → inability to discharge floodwater → continuous rise in reservoir water level → overtopping situation → water flow erosion damages dam structure → dam failure. Prolonged high temperatures in the thermal dome → a surge in reservoir water evaporation → exposure and cracking of the anti-seepage layer → formation of seepage channels → severe internal leakage of the dam body, reduced structural strength → dam failure.

[0112] 4. Flood-induced dam breach pathways Large flood peak flow → exceeding the designed flood discharge capacity → overtopping situation → continuous water flow eroding the dam body, leading to structural damage → dam failure; The flood lasted for a long time → prolonged high water level soaking → aging of the seepage prevention layer → increased seepage → piping formation → piping damage to the dam structure, leading to dam failure.

[0113] 5. Other meteorological variables triggering dam break paths High air humidity → steel gate corrosion → reduced flood discharge capacity → rising reservoir water level → increased seepage → piping formation → piping damages dam structure, leading to dam failure; High wind speed → waves eroding the dam slope → damage to the slope protection structure → increased local seepage → damage to the internal structure of the dam body, reduced stability → dam failure.

[0114] like Figure 4 The diagram shown is a knowledge graph of the path leading to dam failure due to geological factors, including: 1. Magmatic activity Thick weak interlayer → insufficient bearing capacity of dam foundation → foundation failure; Lava flow impact → Damage to dam structure → Structural instability → Dam failure; Wide fault fracture zone → seepage channel formation → piping failure and dam collapse; Volcanic ash accumulation → alteration of seepage channels → crack expansion and dam failure; Karst development → concentrated seepage → seepage dam failure.

[0115] 2. Volcanoes and earthquakes Folded structure → uneven settlement of dam body → dam body tilting → shear failure and dam collapse; Acidic gas corrosion → foundation deformation → crack expansion and dam failure; High joint density → dam body crack development → shear failure → dam collapse; Anomaly in ground stress → stress concentration failure → dam collapse.

[0116] 3. Dam break paths related to soil and rock properties Low soil shear strength → soil sliding → dam sliding failure; Low rock compressive strength → soil sliding → dam sliding failure; High permeability coefficient → increased seepage → seepage dam failure.

[0117] 4. Dam break paths related to earthquake activity High earthquake intensity → earthquake liquefaction → liquefaction dam failure; High earthquake intensity → landslide → landslide impact → impact dam failure; High earthquake intensity → abnormal ground stress → stress concentration failure and dam collapse; Large peak ground acceleration → wide fault fracture zone → seepage channel → piping failure; High peak ground acceleration (PGA) → high joint density → crack development → crack propagation and dam failure; Landslide → Landslide impact → Impact dam failure; Debris flow → Siltation and blockage → Reduced flood discharge capacity → Dam overtopping and collapse; Deep subsidence → foundation instability → foundation dam failure; Large reservoir capacity of landslide dammed lake → rising water level of landslide dammed lake → seepage channels → piping failure of dam.

[0118] like Figure 5 The diagram shown is a knowledge graph path diagram of biological factors leading to dam failure, including: 1. Dam break pathways related to plant activity Deep root penetration → dam cracks → concentrated leakage → local deformation → structural instability → dam failure; Low vegetation cover → weak soil shear resistance → increased risk of soil sliding → dam landslide → dam failure; Humus accumulation thickness → decreased permeability coefficient → altered seepage field → localized deformation → structural instability → dam failure.

[0119] 2. Animal activity-related dam breach pathways High termite nest density → piping channels → piping → hollowing out of the dam's internal structure → dam collapse → dam failure. Large rodent burrow diameter → localized cavities → structural instability → dam tilting or cracking → dam failure; Increased bird nesting volume leads to increased structural load, increased additional stress, localized deformation of the dam body, structural instability, and ultimately, dam failure.

[0120] 3. Microbial-related dam-break pathways Large microbial film thickness → increased seepage resistance → abnormal seepage → local water pressure imbalance in the dam body → rupture at weak points in the dam body → dam failure. High rate of biological decomposition → decrease in material strength → structural instability → insufficient bearing capacity of dam body → dam collapse → dam failure. Gas generation (high methane content) → increased internal pressure → risk of explosion → instantaneous dam failure → dam collapse.

[0121] 4. Algae influences related dam break pathways Algal bloom (excessive biomass) → Eutrophication of water body → Deterioration of water quality → Increased corrosivity of water body → Corrosion of dam materials → Instability of dam structure → Dam failure; Thick biological sedimentation leads to reduced flood discharge capacity, risk of overtopping, dam overtopping, dam erosion and damage, and ultimately, dam failure.

[0122] 5. Dam breakage pathways related to insect erosion High insect density → surface erosion → damage to protective layer → exposure of main dam material → accelerated material damage → dam structure failure → dam collapse; Corrosion by secretions (abnormal pH value) → Accelerated carbonation of concrete → Corrosion of steel bars → Reduced strength of steel bars → Decreased load-bearing capacity of dam structure → Dam failure.

[0123] like Figure 6 The diagram shown is a knowledge graph path diagram of dam failure caused by human (engineering) factors, including: 1. Dam break paths related to design flaws Low flood control standards (the flood control level was set too low during the dam design, without fully considering the historical flood data and future development trends in the region) → Insufficient flood control capacity (the dam cannot withstand the impact of a large flood) → Flood overtopping (the flood level exceeds the dam crest elevation, and water overflows from the dam crest) → Dam failure (the overtopping water continuously erodes the dam body, causing structural damage and dam failure). Improper material selection (failure to select suitable building materials based on dam design requirements and operating environment) → Material failure (premature performance degradation and damage of materials under the influence of water, pressure, etc.) → Structural damage (material failure leads to the dam structure losing support and protection, gradually deteriorating) → Dam failure (severe structural damage makes it impossible to maintain the overall stability of the dam, resulting in dam failure). Structural design error (dam structural design defects) → Uneven stress distribution (during dam operation, the stress distribution in various parts does not meet design expectations, and stress concentration occurs in some areas) → Crack formation (cracks appear in areas of stress concentration due to excessive pressure) → Problem retention (cracks are not detected and treated in time, and continue to expand) → Decision-making error (based on incorrect design concepts or failure to detect crack problems, incorrect maintenance and management decisions are made) → Dam failure (the development of cracks eventually leads to the collapse of the dam structure and dam failure).

[0124] 2. Construction-related dam break paths Poor construction quality (using inferior materials, rough construction techniques, and failure to strictly follow specifications) → Increased seepage (weak materials and loose joints make it easier for water to seep into the dam body) → Dam leakage (after a large amount of water seeps in, seepage occurs on the dam surface) → Risk out of control (seepage continuously erodes the dam structure, and related risk control measures fail) → Dam failure (the dam structure is severely damaged and cannot withstand water pressure, leading to dam failure). Construction process violations (failure to fill in layers according to the design plan, failure to vibrate and compact concrete pouring, etc.) → Unresolved hidden dangers (vacuum holes, cracks, and other hidden dangers caused by illegal construction were not detected and repaired) → Dam leakage (the hidden danger area becomes a seepage channel, and water seeps out of the dam body) → Risk out of control (leakage triggers a chain reaction, and the stability of the dam body continues to decline and cannot be controlled) → Dam failure (ultimately, the dam structure collapses, and a dam failure accident occurs). Cutting corners (during dam construction, the construction party reduces the amount of materials used and simplifies construction procedures to seek profits) → Weak structure (cutting corners leads to insufficient strength and stability of the dam structure, with weak links) → Decreased load-bearing capacity (the actual load-bearing capacity of the dam body cannot meet the design requirements and is unable to withstand external forces such as water pressure) → Dam failure (the weak structure gradually deteriorates under stress, eventually leading to dam failure).

[0125] 3. Management oversights related to dam failure pathways Improper reservoir management (failure to properly control reservoir water level, exceeding reservoir capacity or excessively rapid water storage) → Excessive water storage (reservoir water level exceeds safe range, exerting excessive pressure on dam body) → Uncontrolled risk (dam body bears excessive pressure, relevant risk control measures cannot effectively alleviate pressure) → Dam failure (dam body cannot withstand excessive pressure and dam fails). Unreasonable scheduling (when floods arrive, the reservoir's flood discharge scheduling plan is unscientific, floods are not discharged in time or the discharge volume is not properly controlled) → Water level exceeds limit (the combination of floods and unreasonable scheduling causes the reservoir water level to rise sharply and exceed the safe value) → Uncontrolled water flow (excessively high water level causes the water flow to have uncontrolled impact and scouring force on the dam body) → Dam failure (the dam body is damaged and collapses under the impact of uncontrolled water flow). Lack of inspection → No hidden dangers discovered → Problems continue to worsen → Dam damage → Dam collapse.

[0126] 4. Inadequate maintenance related to dam failure pathways Unrepaired aging (long-term operation of the dam leads to aging of dam materials and damage to structural components without timely maintenance and repair) → Reduced stability (aging causes a decrease in the strength of dam materials and loosening of structural connections, resulting in poor overall stability) → Flood inundation (under the action of floods, dams with poor stability are more easily inundated) → Dam failure (overwhelming floods further exacerbate dam damage, ultimately leading to dam failure). Equipment failure (dam spillway equipment, monitoring equipment, etc. cannot operate normally due to aging, malfunction, etc.) → Flood discharge obstruction (flood discharge equipment failure leads to the inability of floodwater to be discharged in time, and the reservoir water level continues to rise) → Uncontrolled water flow (the rise in water level causes the pressure and scouring force of water flow on the dam body to become uncontrolled) → Dam failure (under the action of uncontrolled water flow, the dam body is structurally damaged and eventually fails). Damage to the anti-seepage layer → Increased leakage → Damage to the dam foundation → Dam failure.

[0127] 5. Dam failure pathways related to operational errors Misoperation (during dam operation and management, staff may open the wrong valve or adjust the water level incorrectly due to operational errors) → Water level exceeding limit (misoperation causes the reservoir water level to rise abnormally, exceeding the safe water level range) → Uncontrolled water flow (excessive water level causes the water flow to lose control of its impact and scouring force on the dam body) → Dam failure (the dam body is damaged and collapses under the impact of uncontrolled water flow).

[0128] 6. Third-party disruption of relevant dam failure pathways Tree felling (a large number of trees were felled in the area around the dam, destroying the original vegetation protection system) → slope instability (the stability of the dam slope soil decreased due to the loss of the soil stabilization effect of tree roots) → dam collapse (after the slope becomes unstable, it cannot effectively protect the dam body, resulting in damage to the dam structure and dam collapse). Human sabotage (deliberately damaging the dam structure, such as drilling holes in the dam body or damaging flood discharge facilities) → Structural damage (human sabotage leads to the destruction of the dam structure's integrity) → Dam failure (the structure is severely damaged and cannot withstand external forces such as water pressure, resulting in dam failure). Illegal sand mining (illegal sand mining in the downstream river channel or near the dam body, damaging the riverbed and dam foundation structure) → Damage to the anti-seepage layer (sand mining activities damage the anti-seepage layer of the dam, making it easier for water to penetrate into the dam body) → Uncontrolled water flow (damage to the anti-seepage layer causes water flow turbulence, resulting in abnormal scouring and pressure on the dam body) → Dam failure (the dam body structure is damaged and the dam fails under the action of uncontrolled water flow).

[0129] 7. Monitoring missing relevant dam failure paths Lack of monitoring equipment (no installation of key monitoring equipment such as water level monitors and seepage monitors) → Monitoring blind spots (unable to obtain important data such as dam water level and seepage in real time, resulting in monitoring gaps) → Crack formation (cracks appear unnoticed due to the inability to detect changes in internal stress of the dam in a timely manner) → Problem retention (cracks are not detected and dealt with in a timely manner, and continue to develop) → Decision-making errors (decisions are made based on incomplete or incorrect information, and effective measures are not taken) → Dam failure (cracks develop severely, eventually leading to dam failure). Data error (errors, incorrect data entry, or transmission failures in dam monitoring data) → Decision error (making incorrect dam operation, management, and maintenance decisions based on incorrect data) → Risk out of control (incorrect decisions fail to effectively address the actual risks of the dam, leading to the continuous escalation and loss of control of the risks) → Dam failure (the loss of control of the risks ultimately leads to the collapse of the dam).

[0130] 8. Dam failure pathways related to inadequate emergency response Delayed emergency response (failure to activate the emergency plan and organize rescue work in a timely manner after discovering the abnormal situation of the dam) → Risk out of control (the abnormal situation is not dealt with in a timely manner, and the risk continues to expand and deteriorate during the delay, exceeding the controllable range) → Dam failure (ultimately, due to the risk out of control, the dam structure is destroyed and the dam fails). Insufficient rescue capacity → Delayed maintenance → Exacerbated disaster impact after dam failure.

[0131] 9. Economic Factors Related to Dam Failure Pathways Funding shortage (insufficient investment in dam maintenance, making normal repair and upkeep impossible) → Maintenance delay (due to lack of funds, problems such as aging and damage to the dam body cannot be repaired in a timely manner, delaying maintenance work) → Quality decline (long-term lack of maintenance of the dam body leads to a gradual decrease in structural performance and a deterioration in quality) → Worsening of hidden dangers (existing hidden dangers continue to worsen due to failure to address them in a timely manner, developing into serious problems) → Dam failure (when hidden dangers worsen to a certain extent, the dam body cannot operate normally, resulting in a dam failure). Cost overruns → Quality decline → Dam construction quality suffers → Increased risk of dam failure.

[0132] 10. Policy-related dam failure pathways Policy deficiencies (inadequate policies related to dam construction and management, such as a lack of strict quality supervision policies and insufficient maintenance funding policies) → Reduced stability (due to policy deficiencies, the quality of the dam cannot be guaranteed during construction and operation, and its stability gradually declines) → Flood inundation (the dam body with insufficient stability cannot withstand the pressure of floods and is inundated) → Dam failure (the continuous damage to the dam body after flood inundation eventually leads to dam failure). Lack of supervision (relevant departments lack effective supervision and inspection of the dam construction and operation process) → Exceeding water level limit (no supervision leads to the reservoir water level exceeding the safety limit, and no timely flood discharge or other measures are taken) → Uncontrolled water flow (excessive water level causes water flow turbulence, resulting in abnormal scouring and pressure on the dam body) → Dam failure (under the action of uncontrolled water flow, the dam body is structurally damaged and eventually fails).

Claims

1. A multi-source data-driven method for verifying the failure path analysis of hydropower projects, characterized in that, include: Obtain multi-source data related to dam failures at hydropower projects; The multi-source data is divided into geological datasets, meteorological datasets, biological datasets, and engineering datasets; Entity extraction and relation extraction are performed based on geological datasets, meteorological datasets, biological datasets and engineering datasets to obtain the final entity set and the relation information between entities. The final entity set and the relation information between entities are then transformed into graph data to construct a knowledge graph of hydropower hub failure. The erroneous and missing relationships in the hydropower hub failure knowledge graph were supplemented and corrected using the path sorting algorithm and the tensor decomposition algorithm, respectively, resulting in a corrected hydropower hub failure knowledge graph. The entity relationships in the revised hydropower hub failure knowledge graph are verified to obtain the verified hydropower hub failure knowledge graph. Based on the validated hydropower dam failure knowledge graph, a breadth-first search algorithm is used to determine all paths from each disaster-causing factor to the dam failure node, and each path is visualized to complete the node path analysis of hydropower dam failure events.

2. The multi-source data-driven hydropower dam failure path analysis and verification method according to claim 1, characterized in that, The entity extraction and relation extraction are performed as follows: Identify the entity types related to dam failure, including: hazard-causing entity, dam failure event entity, spatiotemporal entity, attribute entity, and engineering component entity; Based on various entity types, several regular expressions of different complexities were designed, and entity extraction and relation extraction were performed on geological datasets, meteorological datasets, biological datasets and engineering datasets based on each regular expression to obtain entity sets extracted from different data and relation information between entities in each entity set. Based on the entity sets and the relationship information between entities within each entity set, the information of the same entity is integrated to obtain the final entity set and the relationship information between entities.

3. The multi-source data-driven method for analyzing and verifying the dam failure path of a hydropower project according to claim 1, characterized in that, The verified hydropower dam failure knowledge graph is as follows: Construct a quantum topological verification model for dam-break paths to verify entity relationships; The modified hydropower hub failure knowledge graph was verified using a quantum topological verification model of dam failure paths, resulting in a verified hydropower hub failure knowledge graph.

4. The multi-source data-driven hydropower dam failure path analysis and verification method according to claim 3, characterized in that, The quantum topology verification model for dam break paths includes a path extraction module, a quantum encoder, a spacetime decomposer, a dynamic gating fusion module, a topology verification network, a credibility assessment module, and a hybrid decision-making module. The path extraction module is used to extract dam failure paths from the corrected hydropower hub failure knowledge graph; The quantum encoder is used to convert the dam-break path into a quantum state, encode the path through quantum entanglement, and output the quantum-encoded path. The spatiotemporal decomposer is used to decompose the quantum-encoded path into a temporal convolution branch and a spatial relation branch, thereby obtaining temporal features and spatial features, respectively. The dynamic gating fusion module is used to fuse temporal and spatial features and output spatiotemporal fusion features; The topology verification network is used to construct a topology graph based on spatiotemporal fusion features, calculate the topology vulnerability index of the topology graph, and process and verify the topology graph through graph neural networks and attention mechanisms to output the credibility assessment results of the dam failure path. The credibility assessment module is used to quantify the credibility assessment results of the topological vulnerability index and the dam failure path, and obtain the quantification results. The hybrid decision-making module is used to make decisions based on the quantification results and output verification results.

5. The multi-source data-driven hydropower dam failure path analysis and verification method according to claim 4, characterized in that, The expression for the quantum-encoded path is: in, The path after quantum encoding; for The path states of the layered quantum circuit are connected to each layer through tensor products; It is an exponential function with the natural constant as its base; The imaginary unit; For the first Hermitian matrix dynamically generated by layered quantum circuits; To correct the linear unit; For the first Trainable qubit parameters of layered quantum circuits; For the first Normalized dam-break path of layered quantum circuits; It is the tensor product; It is the hyperbolic tangent function; For the first Trainable qubit parameters of +1 layer quantum circuit; For the first Normalized dam-break path of +1 layer quantum circuit; For layer index; It is a diagonal phase matrix; diagonal elements; For transpose; For the first Learnable entanglement gate parameters for layered quantum circuits; For the first Layered quantum circuits are projected onto classical features activated by ReLU; For the first Layered quantum circuits project classical features through tanh activation; For the first Layered quantum circuit path position encoding unitary matrix; For the first The dam-break path of layered quantum circuits; It is an L2 norm; It is a multilayer perceptron; This is a constructor for diagonal matrices.

6. The multi-source data-driven hydropower dam failure path analysis and verification method according to claim 4, characterized in that, The expression for decomposing the quantum-encoded path into temporal convolutional branches and spatial relational branches is as follows: in, Spatiotemporal characteristics; For Long Short-Term Memory networks, it handles temporal dependencies; The real part of the quantum path encoding reflects the causal strength; The path after quantum encoding; This is an outer product operation; For graph attention networks, spatial relationships between nodes are handled; The imaginary part, which encodes the quantum path, reflects the phase relationship; The Sigmoid function compresses the weights to the (0,1) interval. These are trainable vectors; To capture global time series patterns by averaging temporal features; As a time characteristic, input the real part of the quantum state; To maximize spatial features and highlight key areas; For spatial characteristics, input the imaginary part of the quantum state; For dynamic coupling weights; This is the normalized cross-correlation function; For transpose; It is an L2 norm.

7. The multi-source data-driven method for analyzing and verifying the dam failure path of a hydropower project according to claim 4, characterized in that, The expression for the spatiotemporal fusion feature is: in, It features spatiotemporal fusion; The Sigmoid function compresses the weights to the (0,1) interval. It is a time-related feature; Spatial features; These are trainable weights; for and The splicing result; To pass Calculated weights; To smooth the ReLU function and ensure the output is positive; These are trainable weights; This is an outer product operation; For Hadema; To pass Calculate the nonlinear coupling coefficients.

8. The multi-source data-driven method for analyzing and verifying the dam failure path of a hydropower project according to claim 4, characterized in that, The expressions for the topological vulnerability index and the credibility assessment results of the dam failure path are as follows: in, It serves as a topological vulnerability indicator. It is a topological graph; For connected components The diameter; For branches; The total number of vertices; It is a first-order Betti number; For vertex set; It is an edge set; For the first Features of the vertices of a topological graph; For the first Spatiotemporal fusion features corresponding to vertices of a topological graph; For the first The vertex of the topological graph and the first A topological graph with vertices and edges; This is an indicator function; its value is 1 if the condition is met, and 0 otherwise. For the first Spatiotemporal fusion features corresponding to vertices of a topological graph; It is an L2 norm; Distance threshold; To take the median of the upper triangular portion of the distance matrix, excluding the diagonal; The results of the credibility assessment of the dam failure path; Normalization factor; This represents the total number of local subgraphs. For the first The number of Betti numbers corresponding to each local subgraph; This is a multi-head attention mechanism; For the first Feature representation of a local subgraph; A feature representation of the global graph; It is a natural constant; The Betti number calculation function is used for the homology group matrix; For the first The homology group matrix corresponding to each local subgraph.

9. The multi-source data-driven method for analyzing and verifying the dam failure path of a hydropower project according to claim 4, characterized in that, The expression for the quantization result is: in, To quantify the results; It is a function type; This is a topological vulnerability indicator. It is a topological graph; It is the hyperbolic tangent function; The results represent the credibility assessment of the dam failure path.

10. The multi-source data-driven method for analyzing and verifying the dam failure path of a hydropower project according to claim 4, characterized in that, The expression for the verification result is: in, To verify the results; The path is valid; The lower limit of the quantization result used to determine the validity of the path; The path portion is valid; The upper limit of the quantization result for determining an invalid path; The path is invalid. To quantify the results.

Images