Pumped storage power station construction anomaly detection method and system based on unmanned aerial vehicle image analysis

Abstract

The invention discloses a pumped storage power station construction anomaly detection method and system based on unmanned aerial vehicle image analysis, and relates to the technical field of engineering construction monitoring. Planning a construction area acquisition path, obtaining a multi-view image, processing multi-modal sensor data, and generating a construction site feature tensor; inputting the construction site feature tensor into a dense connection convolutional network for processing, and establishing a four-dimensional construction dynamic model in combination with a heterogeneous graph attention module with a causal reasoning capability and a gating circulation unit; calculating a characteristic distance between the mixed knowledge model and the four-dimensional construction dynamic model by adopting a double-layer contrast learning network, and calibrating an abnormal region and uncertainty thereof; and performing causal reasoning traceability analysis on the detected abnormal region, and screening an optimal compensation scheme through a case-enhanced Bayesian analyzer. According to the method, effective fusion of CAD drawings and expert knowledge is realized through the deep cross-modal knowledge distillation network, and a more comprehensive and normative knowledge basis is provided for anomaly detection.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering construction monitoring, and in particular to a method for detecting abnormalities in the construction of a pumped storage power station based on unmanned aerial vehicle image analysis. Background Art

[0002] Pumped-storage power stations are an important part of the clean energy system, and their construction quality is directly related to the safe operation of the power station. Currently, the commonly used construction anomaly detection systems mainly adopt rule-based expert systems or simple machine learning models. These methods have obvious limitations when processing high-dimensional, multimodal construction scene data. Especially when faced with problems such as uncertainty at the construction site, complexity of the construction process, and diversity of safety risks, existing technologies are difficult to achieve accurate anomaly identification and effective compensation plan generation. At the same time, due to the lack of intelligent decision-making support mechanisms, anomaly handling often relies on manual experience judgment, which can easily lead to processing delays and secondary risks. Summary of the invention

[0003] In view of the problems of insufficient knowledge fusion, limited data collection, lack of dynamic analysis and weak decision support in the existing abnormality detection system for pumped storage power station construction, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is how to construct a pumped storage power station construction anomaly detection method that integrates expert knowledge, multi-source data acquisition, dynamic modeling analysis and intelligent decision support to achieve intelligent supervision and anomaly handling of the construction process.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides a method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis, which includes using a deep cross-modal knowledge distillation network to map the CAD drawings of the pumped storage power station and the expert knowledge base into graph-structured training samples, performing semantic segmentation and annotation steps, and outputting a hybrid knowledge model with normative constraints; using a terrain-adaptive multi-drone collaborative strategy to plan the acquisition path of the construction area, acquiring multi-perspective images based on a multimodal sensor array, processing the multimodal sensor data through a hierarchical spatiotemporal attention network, and generating a construction site feature tensor; inputting the construction site feature tensor into a densely connected convolutional network for processing, combining a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities, and establishing a four-dimensional construction dynamic model; using a two-layer contrastive learning network to calculate the feature distance between the hybrid knowledge model and the four-dimensional construction dynamic model, and calibrating the abnormal area and its uncertainty through a safety level adaptive anomaly detector; performing causal reasoning and source tracing analysis on the detected abnormal area, generating multiple compensation schemes in combination with multi-agent reinforcement learning, and screening the optimal compensation scheme through a case-enhanced Bayesian analyzer.

[0006] As a preferred solution of the pumped storage power station construction anomaly detection method based on drone image analysis described in the present invention, the hybrid knowledge model is constructed as follows: the pumped storage power station CAD drawings are standardized through an image preprocessing network, and the geometric features, spatial relationships and engineering parameters in the drawings are extracted to form a drawing feature vector; the text knowledge in the expert knowledge base is parsed using a natural language processing model to extract construction specifications, anomaly judgment standards and processing methods, and generate a knowledge feature vector; a deep cross-modal knowledge distillation network is constructed, the drawing feature vector is used as the teacher network input, the knowledge feature vector is used as the student network input, and the feature representations of the two modalities are aligned through a cross-attention mechanism; the aligned features are mapped to a graph structure space based on a graph neural network, wherein the graph nodes represent the key components of the construction area, and the graph edges represent the spatial and functional relationships between the components to form an initial graph structure; a semantic segmentation network is used to perform fine-grained annotation on the initial graph structure, a corresponding relationship is established between the construction specification requirements and the physical components, and an annotation graph structure is generated; the annotation graph structure is input into a specification constraint encoder to extract the temporal dependency and quality control standards of the construction process, and a hybrid knowledge model of specification constraints is output.

[0007] As a preferred solution of the pumped storage power station construction anomaly detection method based on drone image analysis described in the present invention, wherein: generating a construction site feature tensor includes the following steps: using a laser radar to scan the terrain of the construction area to generate a three-dimensional point cloud map, dividing the construction area into multiple sub-areas through a point cloud segmentation algorithm, and calculating the terrain complexity index of each sub-area; constructing a multi-drone task allocation model based on the terrain complexity index, using an adaptive particle swarm algorithm to allocate drone formations and flight altitudes to each sub-area, and generating an initial path planning scheme; on the basis of the initial path planning scheme, combined with the obstacle distribution and weather conditions in the construction area, a collaborative flight path considering safety spacing is generated through a trajectory optimization algorithm; a multimodal sensor array equipped with a visible light camera, an infrared camera and a multispectral camera collects a construction site image sequence along the collaborative flight path, and records the corresponding position and posture information; performing multi-view image registration on the construction site image sequence, extracting image features using a deep learning network, and constructing a four-dimensional feature map containing time series information; inputting the four-dimensional feature map into a hierarchical spatiotemporal attention network, and generating a construction site feature tensor through multi-level feature extraction and time series modeling.

[0008] As a preferred solution of the pumped storage power station construction anomaly detection method based on drone image analysis described in the present invention, the invention comprises the following steps: establishing a four-dimensional construction dynamic model includes the following steps: inputting the construction site feature tensor into a densely connected convolutional network, generating a hierarchical feature map through a self-calibrated multi-scale feature extraction module; constructing a heterogeneous graph structure of the construction scene, taking construction equipment, building components and construction personnel as different types of graph nodes, establishing edge connections based on spatial position relationships and construction process dependencies, and generating a scene relationship graph; performing feature fusion on the hierarchical feature representation and the scene relationship graph, calculating the dynamic association weights between nodes through a hierarchical adaptive graph attention mechanism, and generating a time series scene graph; using a causal reasoning module to analyze the causal relationship between nodes in the time series scene graph, constructing a causal dependency network of the construction scene, and identifying key influencing factors and propagation paths; inputting the causal dependency network into a gated recurrent unit, performing time series modeling on key indicators, and generating a state evolution sequence; based on the state evolution sequence, combining information on spatial dimensions, time dimensions, construction element dimensions, and quality dimensions, establishing a four-dimensional construction dynamic model that describes the entire construction process.

[0009] As a preferred solution of the abnormality detection method for pumped storage power station construction based on drone image analysis described in the present invention, the abnormal area and its uncertainty are calibrated by a safety level adaptive abnormality detector, including the following steps: constructing a feature-adapted two-layer contrast learning network, dimensional alignment and feature standardization of the hybrid knowledge model and the four-dimensional construction dynamic model through a preprocessing module, and generating a normalized feature vector by using a hierarchical feature extraction and a residual retention mechanism; calculating a semantic similarity matrix for the normalized feature vector, using a multi-granularity cross-attention mechanism to highlight key feature differences, and generating a feature deviation map; constructing a hierarchical safety assessment system based on engineering specifications and historical cases, mapping the feature deviation map to different safety levels, and generating a safety assessment vector; using an adaptive threshold mechanism to dynamically adjust the sensitivity of the anomaly detector according to the safety assessment vector, and determining a multi-scale detection threshold; using a multi-scale detection threshold to perform hierarchical screening on the feature deviation map, locate potential abnormal areas, and output a set of abnormal areas; combining the Monte Carlo sampling method to evaluate the detection confidence of each area in the abnormal area set, and generating an abnormal calibration result including abnormal position, type and uncertainty.

[0010] As a preferred solution of the abnormality detection method for pumped storage power station construction based on drone image analysis described in the present invention, wherein: screening the optimal compensation plan through a case-enhanced Bayesian analyzer includes the following steps: using a dynamic graph causal attention network to process the abnormality calibration results, establishing a spatiotemporal causal graph of abnormal events, extracting the correlation measurement values ​​between abnormal features and construction elements, and outputting an abnormal propagation matrix; initializing the parameters of the multi-agent interaction model through a hierarchical knowledge distillation architecture, constructing a teacher network containing construction expert rules, setting hierarchical reward indicators, and forming an agent training environment; deploying an adversarial reinforcement learning model to train professional agents, and according to the construction field The domain divides the roles of intelligent agents, executes the competitive game training process, and generates differentiated compensation strategy combinations; builds a graph neural network case retrieval model, encodes historical compensation plans into a time-series graph structure, calculates the case graph matching metric, and extracts empirical feature mapping; configures a two-stream Bayesian neural network to process cost and duration indicators and risk reliability indicators respectively, applies attention weight fusion, and outputs a plan evaluation vector; performs combinatorial optimization calculations under uncertainty constraints, jointly considers evaluation scores and resource constraints, and generates a compensation plan set with emergency plans; sorts the compensation plan set based on the plan evaluation vector, and selects the plan with the highest comprehensive score as the optimal compensation plan.

[0011] As a preferred solution of the method for detecting abnormalities in the construction of pumped storage power stations based on drone image analysis of the present invention, the method comprises the following steps: constructing a deep cross-modal knowledge distillation network, comprising the following steps: constructing a multi-head self-attention encoder, respectively enhancing the drawing feature vector and the knowledge feature vector, and generating drawing attention features and knowledge attention features; designing a dual-stream cross-attention module, taking the drawing attention features as the query vectors of the teacher network, and the knowledge attention features as the key-value pairs, and calculating the cross-modal attention score matrix; establishing a feature distillation loss function, comprising a knowledge distillation loss term and a feature consistency loss term, and adopting an adaptive temperature adjustment strategy to adjust the knowledge distillation process; constructing a feature fusion module, fusing the features output by the teacher network and the student network through a residual connection method, adopting a multi-layer feature pyramid network to extract feature representations of different scales, and adaptively selecting the optimal feature combination through the attention mechanism; normalizing the fused features, eliminating the scale differences of features of different modalities through a batch normalization layer, and outputting the aligned feature representations.

[0012] In the second aspect, an embodiment of the present invention provides a pumped storage power station construction anomaly detection system based on drone image analysis, which includes a knowledge modeling module, which is used to map the pumped storage power station CAD drawings and the expert knowledge base into graph structured training samples using a deep cross-modal knowledge distillation network, perform semantic segmentation and annotation steps, and output a hybrid knowledge model with normative constraints; a data acquisition module, which is used to plan the construction area acquisition path using a terrain-adaptive multi-drone collaborative strategy, acquire multi-perspective images based on a multi-modal sensor array, and process the multi-modal sensor data through a hierarchical spatiotemporal attention network to generate a construction site feature tensor; a dynamic modeling module, which is used to input the construction site feature tensor into a densely connected convolutional network for processing, and establish a four-dimensional construction dynamic model in combination with a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities; an anomaly detection module, which is used to calculate the feature distance between the hybrid knowledge model and the four-dimensional construction dynamic model using a two-layer contrastive learning network, and calibrate the abnormal area and its uncertainty through a safety level adaptive anomaly detector; a solution generation module, which is used to perform causal reasoning and source tracing analysis on the detected abnormal area, generate multiple compensation solutions in combination with multi-agent reinforcement learning, and screen the optimal compensation solution through a case-enhanced Bayesian analyzer.

[0013] The beneficial effects of the present invention are as follows: the present invention realizes the effective fusion of CAD drawings and expert knowledge through a deep cross-modal knowledge distillation network, providing a more comprehensive and standardized knowledge basis for anomaly detection; based on terrain-adaptive multi-UAV collaborative strategy and multimodal sensor data fusion technology, the problems of difficulty and low efficiency in information collection at construction sites under complex terrain environments are solved; a four-dimensional construction dynamic model is constructed by using a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities, which realizes the precise positioning and traceability analysis of construction anomalies; through the combination of multi-agent reinforcement learning and case-enhanced Bayesian analysis, a more feasible and adaptable compensation scheme is provided, which significantly improves the intelligence level of construction quality management. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 The framework flow chart of the anomaly detection method for pumped storage power station construction based on UAV image analysis.

[0016] Figure 2 A flow chart is constructed for the hybrid knowledge model of anomaly detection method for pumped storage power station construction based on UAV image analysis.

[0017] Figure 3 Flowchart of anomaly detector calibration for anomaly detection method of pumped storage power station construction based on UAV image analysis. DETAILED DESCRIPTION

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0019] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0020] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0021] Example 1, reference Figure 1~Figure 3 , which is the first embodiment of the present invention, and provides a method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis. The framework flow chart is as follows Figure 1 As shown, including, S1: Use the deep cross-modal knowledge distillation network to map the pumped storage power station CAD drawings and the expert knowledge base into graph-structured training samples, perform semantic segmentation and annotation steps, and output a hybrid knowledge model with normative constraints; Specifically, the hybrid knowledge model construction flow chart is as follows: Figure 2 As shown, the following steps are included: S1.1: The CAD drawings of the pumped storage power station are standardized through the image preprocessing network, and the geometric features, spatial relationships and engineering parameters in the drawings are extracted to form the drawing feature vector.

[0022] Among them, the image preprocessing network adopts a multi-scale convolutional neural network structure, including four key modules: image denoising, geometric correction, edge enhancement and feature extraction. The preprocessing parameters are adaptively optimized through the validation set.

[0023] S1.2: The text knowledge in the expert knowledge base is parsed using a natural language processing model to extract construction specifications, anomaly determination criteria and processing methods, and generate knowledge feature vectors.

[0024] Among them, the natural language processing model adopts a BERT-based bidirectional encoder architecture and combines domain adaptive pre-training to improve the ability to understand professional terms.

[0025] S1.3: Construct a deep cross-modal knowledge distillation network, take the drawing feature vector as the teacher network input, take the knowledge feature vector as the student network input, and align the feature representations of the two modalities through the cross-attention mechanism.

[0026] Specifically, the construction process of the deep cross-modal knowledge distillation network is as follows: construct a multi-head self-attention encoder to enhance the features of the drawing feature vector and the knowledge feature vector respectively, and generate the drawing attention feature and the knowledge attention feature; design a two-stream cross-attention module, use the drawing attention feature as the query vector of the teacher network, and the knowledge attention feature as the key-value pair, and calculate the cross-modal attention score matrix; establish a feature distillation loss function, which includes the knowledge distillation loss term and the feature consistency loss term, where the knowledge distillation loss guides the student network learning through the teacher network, and the feature consistency loss constrains the semantic alignment of the two modal features; adopt an adaptive temperature adjustment strategy to adjust the knowledge distillation process, and the temperature parameter is dynamically adjusted according to the KL divergence of the teacher network and the student network. When the divergence is greater than the preset threshold, the temperature parameter value is increased to strengthen the knowledge transfer, otherwise the temperature parameter value is reduced to maintain the uniqueness of the feature; construct a feature fusion module, fuse the features output by the teacher network and the student network through the residual connection method, use a multi-layer feature pyramid network to extract feature representations of different scales, and adaptively select the optimal feature combination through the attention mechanism; normalize the fused features, eliminate the scale differences of different modal features through the batch normalization layer, and output the aligned feature representation.

[0027] Among them, the specific formula of the feature distillation loss function is as follows: in, is the knowledge distillation loss term (which measures the similarity between the output distributions of the teacher model and the student model, based on KL divergence), is the feature consistency loss (through Fusion features of norm-constrained teacher model and student model and distance, ensuring the alignment of intermediate features), is a weight hyperparameter (used to balance the impact of the two losses).

[0028] Knowledge Distillation Loss The formula is as follows: in, They are the multi-head cross-attention features of the teacher model and the student model, is the Softmax function (used to generate probability distribution), is the KL divergence (used to measure the difference in the distribution of attention features between the teacher and student models), and T is the dynamic temperature parameter.

[0029] Feature consistency loss The formula is as follows: They are the fusion features of the teacher model and the student model (after FPN and residual connection), for The square of the norm.

[0030] The adaptive temperature adjustment formula is as follows: in, is the base temperature value, is the temperature regulation coefficient, is the KL divergence between the probability distributions output by the teacher model and the student model, is the KL divergence threshold (only when The temperature will be adjusted only when are the probability distributions of the teacher model and the student model respectively.

[0031] The present invention has been verified through experiments that, compared with the existing one-way knowledge distillation methods (such as KD and FitNet), the present method improves the cross-modal feature alignment accuracy by 25.3% (based on cosine similarity evaluation), the discriminability of feature representation by 18.7% (based on t-SNE visualization clustering effect), and the anomaly detection accuracy by 15.6% in actual engineering applications.

[0032] S1.4: Based on the graph neural network, the aligned features are mapped to the graph structure space, where the graph nodes represent the key components of the construction area, and the graph edges represent the spatial and functional relationships between the components, forming the initial graph structure.

[0033] S1.5: Use the semantic segmentation network to perform fine-grained annotation on the initial graph structure, establish a correspondence between the construction specification requirements and the physical components, and generate an annotated graph structure.

[0034] S1.6: Input the annotated graph structure into the specification constraint encoder, extract the temporal dependencies and quality control standards of the construction process, and output the hybrid knowledge model of the specification constraints.

[0035] Specifically, the process of encoding specification constraints is as follows: construct a timing dependency graph, establish the predecessor and successor relationship between construction process nodes based on the annotated graph structure, and use a directed acyclic graph to represent the timing constraints between processes; design a quality feature tensor, map the quality control indicators of each process into a multidimensional feature space, and define the qualified range of the indicators through the interval estimation method; construct a specification constraint encoder, input the timing dependency graph and the quality feature tensor into a two-branch deep neural network, and extract the process constraint and quality constraint characteristics; use the constraint association module to calculate the interaction matrix between the process constraints and the quality constraints, and establish the dynamic change rules of the quality indicators as the process progresses; construct a three-layer constraint feature pool, and hierarchically store the constraint features of different levels, including single component constraints (such as concrete strength), regional constraints (such as construction area temperature) and global constraints (such as project progress); merge the constraint features with the annotated graph structure to generate a hybrid knowledge model with construction specification constraints.

[0036] Preferably, the present invention provides a more standardized and structured knowledge basis for anomaly detection tasks by designing a deep cross-modal knowledge distillation network and a normative constraint encoder, using graph structured representation and semantic segmentation annotation. Compared with traditional rule-based anomaly detection methods, the present invention realizes the organic combination of drawing information and expert experience, improves detection accuracy, reduces false alarm rate, and can provide explainable anomaly cause analysis.

[0037] S2: A terrain-adaptive multi-UAV collaborative strategy is used to plan the collection path in the construction area, multi-view images are acquired based on a multimodal sensor array, and multimodal sensor data are processed through a hierarchical spatiotemporal attention network to generate a construction site feature tensor.

[0038] Specifically, the steps include: S2.1: Use LiDAR to scan the terrain of the construction area and generate a three-dimensional point cloud map. Use the point cloud segmentation algorithm to divide the construction area into multiple sub-areas and calculate the terrain complexity index of each sub-area.

[0039] It should be noted that the terrain in the construction area of ​​a pumped-storage power station is usually complex, including multiple functional areas such as upper and lower reservoirs, water transmission systems, underground plant buildings, etc. The terrain characteristics of each area are significantly different. Therefore, it is necessary to calculate the terrain complexity index based on point cloud data, including terrain undulation, surface roughness, obstacle density, etc. These indicators reflect the terrain characteristics of different areas and provide an important basis for the subsequent reasonable planning of drone inspection routes.

[0040] S2.2: A multi-UAV task allocation model is constructed based on the terrain complexity index, and an adaptive particle swarm algorithm is used to allocate UAV formations and flight altitudes to each sub-area to generate an initial path planning solution.

[0041] Specifically, based on the terrain complexity index, an environmental constraint matrix considering visibility, wind speed and obstacle distribution is constructed, and the flight altitude range and safety distance threshold are set according to the UAV performance parameters to generate a mission environment assessment model; the minimum coverage waypoint set is calculated for each sub-area, and the waypoint spacing threshold is set according to the terrain undulation. The waypoint groups are divided using the convex hull algorithm, and the number of UAVs required for each group is determined based on the waypoint density distribution to form an initial task allocation plan; a multi-objective optimization function is constructed, and the task completion time, energy efficiency and data collection quality are taken as optimization goals, and the UAV payload capacity, communication range and flight time are taken as constraints to establish a task optimization model; an adaptive particle physics optimization model is designed. The particle encoding scheme of the subswarm algorithm, in which the particle position represents the UAV formation parameters and flight altitude parameters, and the particle speed represents the parameter update strategy, dynamically adjusts the inertia weight and learning factor through the terrain complexity index; designs the fitness function based on the task optimization model, and iteratively optimizes the particle swarm in combination with the task environment assessment model, and uses the elite retention strategy and mutation operator to improve the population diversity until the convergence condition is met or the maximum number of iterations is reached; decodes the optimized optimal particles into the UAV formation and flight altitude parameters, generates the waypoint sequence of each UAV in combination with the initial task allocation plan, generates a smooth flight trajectory through the cubic spline interpolation algorithm, and outputs the initial path planning plan.

[0042] Among them, the convergence condition is set as the relative change rate of the optimal solution for 50 consecutive generations is less than 0.1% or the maximum number of iterations reaches 1000 generations; in addition, an adaptive weight adjustment strategy based on terrain complexity is adopted for the multi-objective optimization function. The weight of data acquisition quality is increased in complex terrain areas, and the weight of task completion time is increased in flat areas.

[0043] Experiments show that the adaptive particle swarm algorithm achieves adaptive optimization of terrain complexity by dynamically adjusting the inertia weight and learning factor. Compared with the traditional fixed-parameter path planning method, the multi-objective optimization strategy of the present invention can take into account the task completion time, energy efficiency and data collection quality at the same time, significantly improving the operating efficiency of the UAV formation, and the optimization effect of path planning is improved by about 20%.

[0044] S2.3: Based on the initial path planning scheme, combined with the obstacle distribution and weather conditions in the construction area, a collaborative flight path that takes safety distance into consideration is generated through a trajectory optimization algorithm.

[0045] In this embodiment, the trajectory optimization process is as follows: construct dynamic obstacle distribution based on three-dimensional point cloud map, and establish environmental constraints in combination with meteorological data; perform spatiotemporal conflict detection on the initial path to identify potential risk areas; use trajectory smoothing algorithm to ensure flight stability, and generate a flight path with safe spacing through multi-agent collaborative optimization method; consider meteorological influence to correct the trajectory, and output the final collaborative flight plan after simulation verification. Compared with the traditional single trajectory planning method, this method has stronger environmental adaptability and collaborative operation capabilities.

[0046] S2.4: A multimodal sensor array equipped with visible light cameras, infrared cameras, and multispectral cameras collects construction site image sequences along the collaborative flight path and records the corresponding position and posture information.

[0047] S2.5: Perform multi-view image registration on the construction site image sequence, use a deep learning network to extract image features, and construct a four-dimensional feature map containing temporal information.

[0048] S2.6: Input the four-dimensional feature map into the hierarchical spatiotemporal attention network to generate the construction site feature tensor through multi-level feature extraction and temporal modeling.

[0049] Preferably, the present invention solves the technical problems of difficulty and low efficiency in construction site information collection in complex terrain environments of pumped storage power stations through a technical solution combining terrain-adaptive multi-UAV collaborative strategy planning with multi-modal sensor data fusion processing. Compared with the prior art, this solution realizes the refined partitioning of the construction area and quantification of terrain features based on lidar point cloud data, uses an adaptive particle swarm algorithm to optimize the formation and flight parameters of the UAV fleet, ensures flight safety through multi-agent collaborative optimization, and achieves an improvement of about 20% in path planning efficiency; at the same time, the multi-source heterogeneous data obtained by the multi-modal sensor array is combined with the hierarchical spatiotemporal attention network for feature extraction and fusion modeling, which improves the comprehensiveness and spatiotemporal consistency of construction site information collection and provides reliable data support for engineering construction quality control and safety management.

[0050] S3: The construction site feature tensor is input into a densely connected convolutional network for processing, and a four-dimensional construction dynamic model is established by combining a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities.

[0051] Specifically, the steps include: S3.1: The construction site feature tensor is input into a densely connected convolutional network, and a hierarchical feature map is generated through a self-calibrated multi-scale feature extraction module while maintaining the correlation of features at different levels.

[0052] S3.2: Construct a heterogeneous graph structure of the construction scene, treat construction equipment, building components, and construction personnel as different types of graph nodes, establish edge connections based on spatial position relationships and construction process dependencies, and generate a scene relationship graph.

[0053] S3.3: Perform feature fusion on the hierarchical feature representation and scene relationship graph, calculate the dynamic association weights between nodes through the hierarchical adaptive graph attention mechanism, and generate a temporal scene graph.

[0054] S3.4: Use the causal reasoning module to analyze the causal relationship between nodes in the time-series scene graph, build a causal dependency network of the construction scene, and identify key influencing factors and propagation paths.

[0055] Specifically, the time-series scene graph is decomposed into multi-scale time windows, the graph sequence is divided into overlapping sub-sequences based on the sliding window mechanism, and the node state change features are extracted by the time-series attention network to generate a time-series feature matrix; a bidirectional causal graph neural network is constructed, the time-series feature matrix is ​​input into the encoder, the long-term and short-term dependencies between nodes are captured through the multi-head self-attention mechanism, and the node association measurement matrix is ​​output; a causal hypothesis set for the construction scene is established based on the node association measurement matrix, and the structural equation model is used to calculate the direct and indirect effects between node pairs to generate an initial causal relationship graph; the initial causal relationship graph is input into the graph convolutional network, local and global causal patterns are extracted through multi-layer feature aggregation, and the causal discovery algorithm is used to identify false associations to generate a revised causal relationship graph; a probabilistic graph model is constructed based on the revised causal relationship graph, and the variational inference algorithm is used to calculate the conditional independence and causal strength between nodes, identify key influencing factors, and generate a causal influence score; a causal path tracing algorithm is used to prune and optimize the causal chain based on the causal influence score, and the feasibility of the path is verified in combination with the expert rule base to construct a causal dependency network containing key influencing factors and propagation paths.

[0056] Preferably, the causal reasoning module realizes the extraction of dynamic change characteristics of the construction scene through the combination of multi-scale time window decomposition and bidirectional causal graph neural network, and constructs the causal dependency network of the construction scene based on the structural equation model and variational inference algorithm. Compared with the traditional correlation analysis method, the present invention can effectively capture the long-term and short-term dependencies between nodes, identify and eliminate false associations, and realize the reliable identification of key influencing factors and propagation paths in combination with the expert rule base, providing theoretical support for the traceability analysis and risk warning of construction anomalies.

[0057] S3.5: Input the causal dependency network into the gated recurrent unit to perform time series modeling on key indicators and generate a state evolution sequence.

[0058] Among them, key indicators include construction progress, quality status and resource utilization.

[0059] S3.6: Based on the state evolution sequence, combined with the information of spatial dimension, time dimension, construction element dimension and quality dimension, a four-dimensional construction dynamic model describing the entire construction process is established.

[0060] Specifically, the establishment of a four-dimensional construction dynamic model that describes the entire construction process includes the following steps: inputting the state evolution sequence into a multi-branch feature fusion network, constructing a spatial dimension branch, a temporal dimension branch, a construction element branch, and a quality dimension branch respectively, and extracting the feature representation of each dimension through an independent encoder; using a three-dimensional convolutional neural network to process the spatial dimension branch, extracting the spatial layout characteristics of the construction site, establishing the component position relationship and spatial constraints, and generating a spatial feature vector; combining the long short-term memory network and the self-attention mechanism in the temporal dimension branch to model the construction progress and process connection, capture the critical path of the construction period, and output the time series feature vector; constructing a heterogeneous graph embedding network for the construction element branch, mapping the element nodes such as personnel, equipment, and materials to a unified feature space, and generating an element feature vector based on the element interaction relationship; deploying a multi-layer perceptron through the quality dimension branch, processing quality inspection data and specification constraint indicators, establishing a quality assessment model, and outputting a quality feature vector; designing a cross-dimensional attention fusion module to adaptively weight the feature vectors of the four dimensions, maintaining the original feature information through residual connections, and constructing a four-dimensional construction dynamic model.

[0061] S4: A two-layer contrastive learning network is used to calculate the feature distance between the hybrid knowledge model and the four-dimensional construction dynamic model, and the abnormal area and its uncertainty are calibrated through an anomaly detector with adaptive safety level.

[0062] Specifically, the anomaly detector calibration flow chart is as follows: Figure 3 As shown, the following steps are included: S4.1: Construct a feature-adapted two-layer contrastive learning network, align the dimensions and standardize the features of the hybrid knowledge model and the four-dimensional construction dynamic model through the preprocessing module, and generate a normalized feature vector using hierarchical feature extraction and residual preservation mechanism.

[0063] S4.2: Calculate the semantic similarity matrix for the normalized feature vector, use a multi-granularity cross-attention mechanism to highlight the key feature differences, and generate a feature deviation map.

[0064] S4.3: Construct a hierarchical safety assessment system based on engineering specifications and historical cases, map the characteristic deviation graph to different safety levels, and generate a safety assessment vector.

[0065] Specifically, the construction of a hierarchical safety assessment system includes the following steps: using knowledge graph technology to build an engineering specification library, converting the safety standards, quality requirements and technical indicators in the construction specifications into structured knowledge nodes, establishing hierarchical associations between nodes through a relationship extraction algorithm, and generating a specification knowledge graph; conducting in-depth mining of the historical case database, using a text classification model to label cases with safety levels, building a case association network based on a similarity measurement method, and outputting a case feature matrix; inputting the specification knowledge graph and the case feature matrix into a dual-channel deep neural network, extracting specification constraint features and experience features respectively, fusing the two types of features through an attention mechanism, and generating a safety assessment benchmark; constructing a multi-level safety level division model, setting key indicator thresholds based on the safety assessment benchmark, using a fuzzy comprehensive evaluation method to establish a mapping relationship between indicators and safety levels, and outputting a grading standard set; inputting a feature deviation map into the grading standard set, calculating the weight coefficients of each indicator through a hierarchical analysis method, and quantitatively evaluating the deviation features in combination with an expert scoring mechanism to generate a preliminary safety level; using a deep confidence network to perform probabilistic reasoning on the preliminary safety level, dynamically correcting it in combination with the real-time status of the engineering site, and outputting a safety assessment vector containing safety assessment scores for each dimension.

[0066] Significantly, the hierarchical safety assessment system combines knowledge graph technology with deep learning methods to achieve effective integration of engineering specification knowledge and historical case experience, uses a dual-channel deep neural network to extract specification constraint features and experience features, and uses a deep belief network for probabilistic reasoning and dynamic correction. Compared with the existing fixed threshold assessment method, the adaptive assessment mechanism constructed by the present invention can dynamically adjust the safety level according to the real-time status of the engineering site, improves the scientificity and reliability of safety assessment, and provides more comprehensive technical support for construction management decisions.

[0067] S4.4: Adopt an adaptive threshold mechanism to dynamically adjust the sensitivity of the anomaly detector according to the security assessment vector and determine the multi-scale detection threshold.

[0068] S4.5: Use multi-scale detection thresholds to perform hierarchical screening on the feature deviation map, locate potential abnormal areas, and output a set of abnormal areas.

[0069] S4.6: Combine the Monte Carlo sampling method to evaluate the detection confidence of each area in the abnormal area set, and generate an anomaly calibration result including the anomaly location, type and uncertainty.

[0070] S5: Perform causal reasoning and source tracing analysis on the detected abnormal areas, generate multiple compensation schemes by combining multi-agent reinforcement learning, and screen the optimal compensation scheme through a case-enhanced Bayesian analyzer.

[0071] Specifically, the steps include: S5.1: Use the dynamic graph causal attention network to process the anomaly calibration results, establish the spatiotemporal causal graph of abnormal events, extract the correlation metrics between abnormal features and construction elements, and output the anomaly propagation matrix.

[0072] S5.2: Initialize the parameters of the multi-agent interaction model through the hierarchical knowledge distillation architecture, build a teacher network containing construction expert rules, set hierarchical reward indicators, and form an agent training environment.

[0073] S5.3: Deploy adversarial reinforcement learning models to train professional agents, divide agent roles according to construction fields, execute competitive game training processes, and generate differentiated compensation strategy combinations.

[0074] Specifically, training professional agents and generating differentiated compensation strategy combinations include the following steps: constructing an agent role system based on construction domain knowledge, dividing the construction process participants into progress management agents, quality control agents, resource allocation agents and safety supervision agents, and establishing an agent decision space; constructing a hierarchical adversarial network architecture, designing the competitive game relationship between agents, embedding the abnormal propagation matrix into the state representation space, using a graph encoder to extract state features, and generating a game state vector; deploying a deep deterministic policy gradient algorithm, training the agent policy network based on the game state vector, accumulating adversarial samples through the experience replay mechanism, and outputting the initial compensation strategy set; adopting a multi-agent collaborative training framework, setting up communication channels between agents, introducing an attention mechanism to extract key information, optimizing the agent decision model through policy gradient descent, and generating a collaborative compensation strategy; constructing a compensation effect evaluation module, inputting the collaborative compensation strategy into the simulation environment, calculating the strategy execution cost and effect improvement, dynamically adjusting the agent reward function based on the evaluation results, and optimizing the compensation strategy; adopting a strategy integration method to fuse the compensation strategies of multiple rounds of training, screening complementary strategy combinations through diversity metrics, and generating a differentiated compensation strategy combination.

[0075] S5.4: Build a graph neural network case retrieval model, encode the historical compensation scheme into a temporal graph structure, calculate the case graph matching metric, and extract the empirical feature map.

[0076] S5.5: Configure a two-stream Bayesian neural network to process the cost and duration indicators and the risk reliability indicators respectively, apply attention weight fusion, and output the solution evaluation vector.

[0077] S5.6: Perform combinatorial optimization calculations under uncertainty constraints, jointly consider evaluation scores and resource constraints, and generate a set of compensation schemes with contingency plans.

[0078] S5.7: Sort the compensation scheme set based on the scheme evaluation vector and select the scheme with the highest comprehensive score as the optimal compensation scheme.

[0079] Furthermore, the present embodiment also provides a pumped storage power station construction anomaly detection system based on drone image analysis, including a knowledge modeling module, which is used to map the pumped storage power station CAD drawings and the expert knowledge base into graph structured training samples using a deep cross-modal knowledge distillation network, perform semantic segmentation and annotation steps, and output a hybrid knowledge model with normative constraints; a data acquisition module, which is used to plan the construction area acquisition path using a terrain-adaptive multi-drone collaborative strategy, acquire multi-view images based on a multi-modal sensor array, and process the multi-modal sensor data through a hierarchical spatiotemporal attention network to generate a construction site feature tensor; a dynamic modeling module, which is used to input the construction site feature tensor into a densely connected convolutional network for processing, and establish a four-dimensional construction dynamic model in combination with a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities; an anomaly detection module, which is used to calculate the feature distance between the hybrid knowledge model and the four-dimensional construction dynamic model using a two-layer contrastive learning network, and calibrate the abnormal area and its uncertainty through a safety level adaptive anomaly detector; a solution generation module, which is used to perform causal reasoning and source tracing analysis on the detected abnormal area, generate multiple compensation solutions in combination with multi-agent reinforcement learning, and screen the optimal compensation solution through a case-enhanced Bayesian analyzer.

[0080] In summary, the present invention realizes the effective fusion of CAD drawings and expert knowledge through a deep cross-modal knowledge distillation network, providing a more comprehensive and standardized knowledge basis for anomaly detection; based on terrain-adaptive multi-UAV collaborative strategy and multimodal sensor data fusion technology, it solves the problems of difficulty and low efficiency in information collection at construction sites under complex terrain environments; a four-dimensional construction dynamic model is constructed by using a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities, which realizes the precise positioning and traceability analysis of construction anomalies; through the combination of multi-agent reinforcement learning and case-enhanced Bayesian analysis, a more feasible and adaptable compensation scheme is provided, which significantly improves the intelligent level of construction quality management.

[0081] Example 2, reference Figure 1~Figure 3 , which is the second embodiment of the present invention, provides a method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0082] To verify the effectiveness of the method of the present invention, the research team selected a pumped storage power station under construction as the experimental object. The power station is located in a mountainous and hilly area, with a total installed capacity of 1200MW, a height difference of about 500 meters between the upper and lower reservoirs, and a total construction area of ​​about 12 square kilometers. The experimental environment uses an Intel Xeon dual-core server (64 cores, 256GB memory) equipped with an NVIDIA A100 GPU for deep learning model training.

[0083] During the data collection phase, the research team deployed a multi-UAV collaborative operation system, including 6 UAVs equipped with multimodal sensor arrays. The sensor configuration includes: a visible light camera with a resolution of 40 million pixels, an infrared thermal imager with a resolution of 640×512, and a 6-band multispectral camera. At the same time, a lidar was used for terrain scanning, and the point cloud density was set to 50 points / square meter. The meteorological conditions during the experiment were: temperature 15~25℃, wind speed 3~5m / s, and good visibility.

[0084] During the knowledge model construction phase, the researchers collected 458 CAD drawings of the project, including civil engineering, electromechanical installation and other disciplines, and sorted out more than 2,000 expert knowledge items. Through deep cross-modal knowledge distillation network training, the model's feature alignment accuracy on the cross-validation set reached 92.5%, and the knowledge transfer loss converged to 0.086. The network training used a batch size of 64, the initial value of the learning rate was set to 0.001, and the cosine annealing strategy was used for adjustment, with a total of 100 rounds of training.

[0085] In terms of drone path planning, the construction area was divided into 28 sub-areas based on the lidar point cloud data, and the terrain complexity index ranged from 0.35 to 0.89 (normalized value). The population size of the adaptive particle swarm algorithm was set to 200, the maximum number of iterations was 1000, and the convergence threshold was 0.1%. The total length of the optimized flight path was shortened by 23.6% compared with the initial plan, and the task completion time was reduced by 18.9%. Through trajectory optimization with environmental constraints, the minimum safe distance between drones was ensured to be maintained above 50 meters.

[0086] As shown in Table 1, the method of the present invention is compared and verified with the prior art solution on multiple key indicators: Table 1 Performance comparison between the method of the present invention and the traditional method Judging from the experimental results, the method of the present invention has significant advantages in anomaly detection performance. During the three-month field verification, the system performed a total of 312 inspection tasks, and the cumulative collected image data volume reached 2.1TB. After processing by the hierarchical spatiotemporal attention network, 147 potential anomalies were successfully identified, of which 142 were confirmed by on-site experts, and the real anomaly detection rate reached 96.5%. The main types of anomalies include concrete cracks (accounting for 32%), steel bar spacing deviation (accounting for 28%), water seepage points (accounting for 21%) and other quality defects (accounting for 19%).

[0087] In terms of compensation plan generation, after 500,000 training iterations, the reward value of the policy network converged to a stable state. For detected abnormal situations, the system generates an average of 4-6 differentiated compensation plans, and the Pareto optimal solution set of the execution cost and expected effect of the plan contains an average of 3 optional plans. Through case-enhanced Bayesian analyzer screening, the implementation success rate of the final recommended plan reached 91.3%, an increase of 15.6 percentage points over the traditional expert experience method.

[0088] Particularly noteworthy is the outstanding performance of the present invention in causal reasoning. Through the dynamic graph causal attention network, the system can accurately track the root cause of the anomaly and predict its potential impact. On the validation set, the accuracy of causal relationship identification reached 89.6%, which provides more valuable decision support for construction managers. At the same time, the average response time of the system is only 0.8 seconds, which meets the requirements of real-time monitoring.

[0089] From the perspective of cost-effectiveness, although the method of the present invention has high costs in the initial hardware investment and algorithm development, it significantly reduces the cost of manual inspections in actual applications, improves the timeliness of abnormal discovery, and reduces the expenditure on construction rework and quality rectification. According to project statistical data, the commissioning of the system has saved about 2.8% of the total construction cost of the project.

[0090] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A pumped storage power station construction anomaly detection method based on drone image analysis, characterized by: include, The deep cross-modal knowledge distillation network is used to map the pumped storage power station CAD drawings and the expert knowledge base into graph-structured training samples, perform semantic segmentation and annotation steps, and output a hybrid knowledge model with normative constraints. A terrain-adaptive multi-UAV collaborative strategy is used to plan the collection path in the construction area, and multi-view images are acquired based on a multi-modal sensor array. The multi-modal sensor data is processed through a hierarchical spatiotemporal attention network to generate a construction site feature tensor. The construction site feature tensor is input into a densely connected convolutional network for processing, and a four-dimensional construction dynamic model is established by combining a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities; A double-layer contrastive learning network is used to calculate the characteristic distance between the hybrid knowledge model and the four-dimensional construction dynamic model, and an abnormal area and its uncertainty are calibrated by an abnormal detector with adaptive safety level; Causal reasoning and tracing analysis are performed on the detected abnormal areas, multiple compensation schemes are generated by combining multi-agent reinforcement learning, and the optimal compensation scheme is screened through a case-enhanced Bayesian analyzer.

2. The method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis according to claim 1, characterized in that: The construction process of the hybrid knowledge model is as follows: The CAD drawings of the pumped storage power station are standardized through the image preprocessing network, and the geometric features, spatial relationships and engineering parameters in the drawings are extracted to form the drawing feature vectors. The text knowledge in the expert knowledge base is parsed using a natural language processing model to extract construction specifications, abnormality judgment criteria and processing methods, and generate knowledge feature vectors; Constructing a deep cross-modal knowledge distillation network, taking the drawing feature vector as the teacher network input, taking the knowledge feature vector as the student network input, and aligning the feature representations of the two modalities through a cross-attention mechanism; Based on the graph neural network, the aligned features are mapped to the graph structure space, where the graph nodes represent the key components of the construction area, and the graph edges represent the spatial and functional relationships between the components, forming the initial graph structure. A semantic segmentation network is used to perform fine-grained annotation on the initial graph structure, a corresponding relationship is established between the construction specification requirements and the physical components, and an annotated graph structure is generated; The annotated graph structure is input into a specification constraint encoder to extract the temporal dependency and quality control standards of the construction process, and output a hybrid knowledge model of the specification constraint.

3. The method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis according to claim 1, characterized in that: Generating the construction site feature tensor comprises the following steps: Use laser radar to scan the terrain of the construction area, generate a three-dimensional point cloud map, divide the construction area into multiple sub-areas through the point cloud segmentation algorithm, and calculate the terrain complexity index of each sub-area; A multi-UAV task allocation model is constructed based on the terrain complexity index, and an adaptive particle swarm algorithm is used to allocate UAV formations and flight altitudes to each sub-area to generate an initial path planning solution; Based on the initial path planning scheme, combined with the obstacle distribution and weather conditions in the construction area, a collaborative flight path that takes into account the safety distance is generated through a trajectory optimization algorithm; A multimodal sensor array equipped with a visible light camera, an infrared camera, and a multispectral camera collects a sequence of images of the construction site along the collaborative flight path and records corresponding position and posture information; Performing multi-view image registration on the construction site image sequence, extracting image features using a deep learning network, and constructing a four-dimensional feature map containing time series information; The four-dimensional feature map is input into a hierarchical spatiotemporal attention network, and a construction site feature tensor is generated through multi-level feature extraction and temporal modeling.

4. The method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis according to claim 1, characterized in that: The establishment of the four-dimensional construction dynamic model comprises the following steps: The construction site feature tensor is fed into a densely connected convolutional network to generate a hierarchical feature map through a self-calibrated multi-scale feature extraction module; Construct a heterogeneous graph structure of the construction scene, treat construction equipment, building components and construction personnel as different types of graph nodes, establish edge connections based on spatial position relationships and construction process dependencies, and generate a scene relationship graph; Performing feature fusion on the hierarchical feature representation and the scene relationship graph, calculating the dynamic association weights between nodes through a hierarchical adaptive graph attention mechanism, and generating a temporal scene graph; A causal reasoning module is used to analyze the causal relationship between nodes in the time sequence scene graph, to construct a causal dependency network of the construction scene, and to identify key influencing factors and propagation paths; Input the causal dependency network into a gated recurrent unit, perform time series modeling on key indicators, and generate a state evolution sequence; Based on the state evolution sequence, combined with the information of space dimension, time dimension, construction element dimension and quality dimension, a four-dimensional construction dynamic model describing the whole construction process is established.

5. The method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis according to claim 1, characterized in that: The method of calibrating the abnormal area and its uncertainty by using the safety level adaptive abnormal detector comprises the following steps: A feature-adapted two-layer contrastive learning network is constructed. The hybrid knowledge model and the four-dimensional construction dynamic model are dimensionally aligned and feature-normalized through the preprocessing module. The hierarchical feature extraction and residual preservation mechanism are used to generate a normalized feature vector. Calculating a semantic similarity matrix for the normalized feature vectors, using a multi-granularity cross-attention mechanism to highlight key feature differences, and generating a feature deviation map; Building a hierarchical safety assessment system based on engineering specifications and historical cases, mapping the characteristic deviation graph to different safety levels, and generating a safety assessment vector; Adopting an adaptive threshold mechanism to dynamically adjust the sensitivity of the anomaly detector according to the security assessment vector and determine the multi-scale detection threshold; Using the multi-scale detection threshold to perform hierarchical screening on the feature deviation map, locate potential abnormal areas, and output a set of abnormal areas; The detection confidence of each area in the abnormal area set is evaluated in combination with the Monte Carlo sampling method to generate an abnormal calibration result including the abnormal location, type and uncertainty.

6. The method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis according to claim 1, characterized in that: The method of screening the optimal compensation scheme by using the case-enhanced Bayesian analyzer includes the following steps: The dynamic graph causal attention network is used to process the anomaly calibration results, establish the spatiotemporal causal graph of abnormal events, extract the correlation measurement values ​​between abnormal features and construction elements, and output the anomaly propagation matrix; The parameters of the multi-agent interaction model are initialized through a hierarchical knowledge distillation architecture, a teacher network containing construction expert rules is constructed, hierarchical reward indicators are set, and an agent training environment is formed; Deploy adversarial reinforcement learning models to train professional agents, divide agent roles according to construction fields, execute competitive game training processes, and generate differentiated compensation strategy combinations; Build a graph neural network case retrieval model, encode historical compensation plans into a time-series graph structure, calculate case graph matching metrics, and extract empirical feature maps; Configure a two-stream Bayesian neural network to process cost and duration indicators and risk reliability indicators respectively, apply attention weight fusion, and output a solution evaluation vector; Perform combinatorial optimization calculations under uncertainty constraints, jointly consider evaluation scores and resource constraints, and generate a set of compensation plans with contingency plans; The compensation scheme set is sorted based on the scheme evaluation vector, and the scheme with the highest comprehensive score is selected as the optimal compensation scheme.

7. The method for detecting abnormalities in the construction of a pumped storage power station based on drone image analysis as claimed in claim 2, characterized in that: The construction of a deep cross-modal knowledge distillation network includes the following steps: Construct a multi-head self-attention encoder to enhance the drawing feature vector and knowledge feature vector respectively, and generate drawing attention features and knowledge attention features; Design a two-stream cross-attention module, use the drawing attention features as the query vector of the teacher network, and the knowledge attention features as key-value pairs to calculate the cross-modal attention score matrix; Establish a feature distillation loss function, which includes knowledge distillation loss term and feature consistency loss term, and use an adaptive temperature adjustment strategy to adjust the knowledge distillation process; Construct a feature fusion module to fuse the features output by the teacher network and the student network through residual connection, use a multi-layer feature pyramid network to extract feature representations of different scales, and adaptively select the optimal feature combination through the attention mechanism; The fused features are normalized, the scale differences of features of different modalities are eliminated through the batch normalization layer, and the aligned feature representation is output.

8. A pumped storage power station construction anomaly detection system based on drone image analysis, based on the pumped storage power station construction anomaly detection method based on drone image analysis according to any one of claims 1 to 7, characterized in that: Also includes, The knowledge modeling module is used to map the CAD drawings of the pumped storage power station and the expert knowledge base into graph-structured training samples using a deep cross-modal knowledge distillation network, perform semantic segmentation and annotation steps, and output a hybrid knowledge model with normative constraints; The data acquisition module is used to plan the acquisition path in the construction area using a terrain-adaptive multi-UAV collaborative strategy, acquire multi-view images based on a multi-modal sensor array, process the multi-modal sensor data through a hierarchical spatiotemporal attention network, and generate a construction site feature tensor; A dynamic modeling module, used to input the construction site feature tensor into a densely connected convolutional network for processing, and to establish a four-dimensional construction dynamic model by combining a heterogeneous graph attention module and a gated recurrent unit with causal reasoning capabilities; An anomaly detection module, used to calculate the characteristic distance between the hybrid knowledge model and the four-dimensional construction dynamic model by using a double-layer contrast learning network, and calibrate the abnormal area and its uncertainty by using an anomaly detector with adaptive safety level; The solution generation module is used to perform causal reasoning and traceability analysis on the detected abnormal areas, generate multiple compensation solutions by combining multi-agent reinforcement learning, and screen the optimal compensation solution through a case-enhanced Bayesian analyzer.

Images