Energy storage power station construction environment intelligent monitoring method and system

Abstract

The application relates to the technical field of data processing, and provides a method and system for intelligently monitoring a construction environment of an energy storage power station. The method comprises the following steps: constructing a monitoring network of a current construction environment; performing feature dimension disassembly on data monitored by the monitoring network to construct a feature analysis graph of the current construction environment; performing granularity sensitivity evaluation on feature vectors of each node in the feature analysis graph of the current construction environment to obtain recommended granularity configurations of each feature; performing enhancement processing on matched multi-modal monitoring data collection according to the recommended granularity configurations corresponding to each feature; and performing construction environment risk index evaluation, and triggering a safety alarm when the risk index exceeds a threshold value. The application breaks the data island among various construction stages, realizes cross-stage risk linkage analysis, predicts the current construction environment risk, changes from passive response to active intervention, and improves the safety management capability of the whole construction process of the energy storage power station.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an intelligent monitoring method and system for construction environment of energy storage power station. BACKGROUND

[0002] Energy storage power station plays a crucial role in energy storage and grid stability, and its site selection tends to be in remote areas, such as near wind power plants or solar power stations, in order to facilitate the on-site consumption of generated electricity. These areas have harsh natural environments and fast risk transmission, making the construction of energy storage power stations face many technical challenges. The construction of energy storage power station mainly consists of civil engineering, battery cabin installation and electrical installation stages. The traditional construction monitoring data is in a serious island state, and the construction stages cannot form an effective collaborative mechanism. For example, in the civil engineering stage, if the foundation settlement or soil deformation is not monitored in real time, the construction team will have difficulty in adjusting the installation strategy in time, which may directly cause the angle deviation of the battery cabin installation, and then may lead to insufficient electrical wiring allowance, or even connection failure, making the construction in a "passive remediation" lagging state. Since the existing monitoring system is mostly limited to static data monitoring of a single construction stage, such as the installation stage, only the attitude of installing the battery cabin is concerned, and a cross-stage risk linkage analysis mechanism is not established, making it difficult to quantify and warn the transmission path of construction hazards.

[0003] Therefore, it is urgent to propose an intelligent monitoring method that can break the data island between construction stages, realize cross-stage risk linkage analysis, and have construction environment risk prediction capability, to improve the safety management of the whole construction process of energy storage power station and improve the construction speed. SUMMARY

[0004] The present application provides an intelligent monitoring method and system for construction environment of energy storage power station, aiming to solve the technical problems of data island between construction stages in current construction, and the difficulty in realizing cross-stage and multi-dimensional dynamic linkage analysis of construction risks, to change from passive response to active intervention, and to improve the safety management capability of the whole construction process of energy storage power station.

[0005] The first aspect of the present application provides an intelligent monitoring method for construction environment of energy storage power station, which comprises:

[0006] obtaining the current construction stage of the energy storage power station from the BIM system, matching the multi-modal monitoring data required for monitoring from the construction environment description of the current construction stage, starting the sensors on the relevant monitoring points, and forming a monitoring network of the current construction environment;

[0007] Based on the data monitored by the monitoring network, spatial, temporal and physical three-dimensional features of the monitoring points in the monitoring network are extracted, spatial adjacency relationship and temporal causality of each monitoring point are analyzed, spatial-time correlation relationship between each monitoring point is obtained, and a feature analysis graph of the current construction stage is constructed with the monitoring points as nodes and the correlation relationship as edges, wherein the spatial-time correlation relationship includes composite correlation strength and causality direction;

[0008] A plurality of sampling granularity levels and their sampling frequencies are preset, each type of monitoring data obtained for each monitoring point is respectively down-sampled, an information loss function under each granularity is calculated, the information loss function under each granularity is compared with a preset feature importance level threshold, and a recommended sampling granularity of each monitoring mode is determined according to a priority progressive decision process;

[0009] According to the recommended sampling granularity of each monitoring mode, new multi-modal monitoring data is continuously collected in the monitoring network of the current construction environment, the obtained data is enhanced, the three-dimensional features of the nodes of the feature analysis graph and the composite correlation strength of the edges are updated, and dynamic evolution of the graph is realized.

[0010] Based on the evolved feature analysis graph, global topology features, key path features and key node features are extracted, input into a pre-trained risk assessment model, and a continuous risk index of the current construction environment is output, and when the risk index exceeds a preset safety threshold, a hierarchical early warning response is triggered.

[0011] Another aspect of the present application provides an intelligent monitoring system for a construction environment of an energy storage power station, which comprises:

[0012] A monitoring network construction module 11 is configured to obtain a current construction stage of the energy storage power station from a BIM system, match multi-modal monitoring data required for monitoring from a construction environment description of the current construction stage, start sensors on relevant monitoring points, and form a monitoring network of the current construction environment.

[0013] A graph construction module 12 is configured to extract spatial, temporal and physical three-dimensional features of monitoring points in the monitoring network based on data monitored by the monitoring network, analyze spatial adjacency relationship and temporal causality of each monitoring point, obtain spatial-time correlation relationship between each monitoring point, and construct a feature analysis graph of the current construction stage with the monitoring points as nodes and the correlation relationship as edges, wherein the spatial-time correlation relationship includes composite correlation strength and causality direction.

[0014] The granularity configuration module 13 is configured to preset a plurality of sampling granularity levels and their sampling frequencies, to respectively perform down-sampling on each kind of monitoring data acquired by each monitoring point in each monitoring mode, to calculate an information loss function at each granularity, to compare the information loss function at each granularity with a preset characteristic importance level threshold, and to determine a recommended sampling granularity of each monitoring mode according to a decision-making process in which priorities are progressive.

[0015] The atlas dynamic evolution module 14 is configured to continue to collect new multi-modal monitoring data in the monitoring network of the current construction environment according to the recommended sampling granularity of each monitoring mode, to enhance the acquired data, to update the composite association strength of the three-dimensional features and edges of the feature analysis atlas nodes, and to realize dynamic evolution of the atlas.

[0016] The construction environment risk module 15 is configured to extract global topological features, key path features and key node features based on the evolved feature analysis atlas, to input the pre-trained risk assessment model, and to output a continuous risk index of the current construction environment. When the risk index exceeds a preset safety threshold, a hierarchical early warning response is triggered.

[0017] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0018] 1) The stage semantic understanding ability of BIM is combined with dynamic networking of multi-modal sensors to realize intelligent mapping of "construction environment -> monitoring target -> data collection";

[0019] 2) The space-time features are fused through the atlas nodes to construct a computable construction environment risk knowledge network, to realize dynamic quantification of risk transmission paths, and to solve the problem of fragmented monitoring data that cannot form a global perspective;

[0020] 3) According to the recommended granularity configuration, the feature information of the nodes in the atlas is updated in real time to ensure that the atlas is synchronized with the actual state of construction and adapts to construction changes;

[0021] 4) The dynamic evolution of the feature analysis atlas can identify the propagation path of the risk area and support accurate early warning;

[0022] 5) The comprehensive risk of the construction environment is evaluated in real time, and a hierarchical alarm is triggered through a threshold to solve the problem that the traditional construction is in a "passive remediation" state and is difficult to respond to risks in a timely manner.

[0023] In summary, the technical solution provided in the present application connects BIM throughout the whole cycle of data flow and constructs a computable feature analysis atlas, which completely breaks down the data silos between construction stages, realizes risk linkage analysis, supports dynamic risk prediction, and realizes the upgrade from "after-the-fact rectification" to "pre-emptive intervention".

[0024] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating an intelligent monitoring method for the construction environment of an energy storage power station in one embodiment.

[0027] Figure 2 This is an architecture diagram of an intelligent monitoring system for the construction environment of an energy storage power station, as shown in one embodiment.

[0028] Figure labeling: Monitoring network construction module 11, map construction module 12, granularity configuration module 13, map dynamic evolution module 14, construction environment risk module 15. Detailed Implementation

[0029] This application provides an intelligent monitoring method and system for the construction environment of energy storage power stations, which solves the technical problems of low scene-data matching and low recognition efficiency caused by insufficient feature enhancement granularity in the multimodal data processing of rail transit.

[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0031] It should be noted that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0032] Example 1, as Figure 1 As shown, this application provides an intelligent monitoring method for the construction environment of an energy storage power station, including:

[0033] Obtain the current construction stage of the energy storage power station from the BIM system, match the multi-modal monitoring data required for monitoring from the construction environment description of the current construction stage, start the sensors on the relevant monitoring points, and form a monitoring network of the current construction environment.

[0034] Specifically, before the intelligent monitoring method is executed, the BIM first needs to establish a full-professional three-dimensional parameterized model, independently model key components such as battery compartments, foundation piles, and cable trenches, and assign unique ID identifiers. At the same time, the construction monitoring special attribute set is extended in the IFC standard, including component sensitivity coefficients, required sensor types, monitoring threshold values, and other key parameters. The purpose is to convert these key parameters obtained from engineering experience into a standardized format readable by machines, thereby realizing the digital expression of engineering semantics. Secondly, the model components are deeply bound with the key process requirements of the construction stage and the construction progress plan to form a BIM model with an accurate time axis, clearly defining the construction stage time window of each component and the corresponding monitoring requirements, as shown in Table 1. Finally, the construction site environment baseline data (including geological exploration reports, 50-year wind speed, and other climate parameters) need to be entered, and the normal range and early warning threshold of each monitoring parameter (such as the battery compartment inclination design threshold <0.5°) are set. In addition, the intelligent monitoring system needs to establish a sensor capability matrix and a monitoring point information database. The sensor capability matrix is a structured data table that records the technical specifications and capability parameters of all available sensor devices in the monitoring system. The matrix includes detailed information such as sensor ID, monitoring parameter type (such as temperature, humidity, displacement, stress, etc.), measurement range, accuracy level, response time, environmental conditions (temperature range, humidity range, protection level), power consumption parameters, communication protocol, installation requirements, etc. For example, the matrix record of a certain temperature sensor is: [ID: T001, Parameter: Temperature, Range: -40~85℃, Accuracy: ±0.1℃, Response Time: 1s, Protection Level: IP67]. The monitoring point information database records the spatial coordinates, coverage range, and device access capabilities of all potential monitoring points. At the same time, a construction environment-monitoring demand knowledge base is constructed to establish the mapping relationship between construction environment keywords and monitoring parameter types, providing semantic support for NLP technology to analyze construction environment descriptions.

[0035] Table 1

[0036]

[0037] Further, from the construction environment description of the current construction stage, the multi-modal monitoring data required for monitoring is matched, including:

[0038] Obtain the construction environment of the current construction stage of the energy storage power station described in words and / or language, use NLP technology to analyze the current environment description, and identify the key monitoring requirements of the current construction environment;

[0039] Retrieving a pre-constructed sensor capability matrix and a monitoring point information database based on key monitoring requirements of the current construction environment, matching the optimal sensor combination through a greedy algorithm, starting the sensors on the relevant monitoring points, forming a monitoring network of the current construction environment, and obtaining multi-modal monitoring data required for monitoring, wherein the sensor capability matrix is a structured data table for providing sensor technical specifications and performance parameters, and the monitoring point information database is for providing monitoring point spatial position, coverage range and device access information.

[0040] Specifically, scene description information of the current construction stage is obtained through various channels, including textual descriptions input by construction managers, on-site voice reports, construction log records, etc. For voice input, speech recognition technology is used to convert it into text format. Then the obtained text is preprocessed, including removing noise words, standardizing punctuation marks, and unifying synonyms, to ensure text quality. Next, natural language processing technology is used to deeply analyze the preprocessed text, using named entity recognition technology to extract key entities such as construction activities, equipment names, and environmental conditions, using dependency syntax analysis to identify semantic relationships between entities, and using word vector models to calculate semantic feature vectors of the text. Subsequently, the extracted keywords and semantic features are matched with the pre-constructed construction environment-monitoring demand knowledge base, through similarity calculation and fuzzy matching algorithms, to identify the most relevant monitoring parameter types for the current scene description. Finally, based on the matching results and confidence scores, a list of key monitoring demands for the current construction environment is output, including monitoring parameter types, priority levels, and expected monitoring intensity, providing clear demand guidance for subsequent sensor selection and configuration. The construction environment-monitoring demand knowledge base is constructed using a combination of expert knowledge extraction and historical data mining. First, professional literature, construction specifications, safety standards, and historical construction reports in the energy storage power station construction field are collected, and typical scene description texts for each construction stage are extracted. Then, experts in the fields of energy storage power station construction, monitoring, and safety are invited to label the collected scene descriptions, identifying key semantic elements such as construction activities, environmental conditions, and equipment states, and labeling the corresponding monitoring demand types. Next, natural language processing technology is used to perform word frequency analysis and semantic clustering on the labeled data, extracting high-frequency keywords and semantic patterns, such as "battery compartment hoisting" corresponding to "displacement monitoring, stress monitoring", "cable laying" corresponding to "temperature monitoring, gas monitoring", etc. Finally, the extracted keywords and monitoring parameter types are mapped to each other in a many-to-many relationship, and are continuously optimized and improved through expert verification and actual project feedback, forming a scene-demand mapping knowledge base covering the entire construction cycle of energy storage power stations.Based on the identified key monitoring requirements, the system first filters out a candidate set of sensors that can meet the requirements of each monitoring parameter type from the sensor capability matrix, and queries the monitoring point information database to obtain available monitoring point location information that meets the spatial coverage requirements. Then, a sensor-monitoring point matching matrix is constructed, and the comprehensive score of each sensor at each monitoring point is calculated, taking into account factors such as sensor accuracy matching degree, environmental adaptability, power efficiency, monitoring point spatial coverage ability, device access capacity, etc. Next, a greedy algorithm is used for optimal combination matching. The algorithm processes each monitoring requirement in order of priority. For each monitoring requirement, the sensor-monitoring point combination with the highest comprehensive score that has not been assigned is selected from the candidate set, and it is checked whether there is a spatial conflict or resource competition with the selected combination. If there is a conflict, a suboptimal solution is selected. This process is repeated until all key monitoring requirements are assigned to appropriate sensor-monitoring point combinations. Finally, the system generates a sensor start instruction sequence based on the matching results, sends configuration parameters and start commands to the selected sensors of the monitoring points through the communication protocol, and the sensors start data collection according to the specified sampling frequency and accuracy requirements after receiving the instructions. The collected multi-modal monitoring data is transmitted back to the intelligent monitoring system in real time, forming a dynamic monitoring network for the current construction environment. Each monitoring point in the network corresponds to one or more activated sensors, and these monitoring points are the monitoring points of the current construction stage.

[0041] Based on the data monitored by the monitoring network, spatial, temporal, and physical features of the monitoring points in the monitoring network are extracted, spatial adjacency relationships and temporal causal relationships of the monitoring points are analyzed, and spatial-temporal correlation relationships between the monitoring points are obtained. A feature analysis graph of the current construction stage is constructed with the monitoring points as nodes and the correlation relationships as edges, wherein the spatial-temporal correlation relationships include composite correlation strength and causal direction.

[0042] Specifically, in the construction environment of energy storage power station, there is a complex mutual influence relationship between each monitoring point. Firstly, the multi-modal monitoring data such as temperature, vibration, displacement and the like continuously collected are pre-processed, including data cleaning, outlier rejection and missing value interpolation. Then, through GPS time synchronization and BIM coordinate mapping, the data collected by different sensors are aligned in time and space to form multi-modal monitoring data time series with accurate time stamp. Based on these time series data, a three-dimensional feature vector is constructed for each monitoring point: the spatial feature is obtained by extracting the three-dimensional coordinates (x, y, z) of the monitoring point and calculating the topological connection relationship with the adjacent monitoring points using the Delaunay triangulation algorithm, to generate a feature vector containing position information, number of adjacent points, average distance and spatial density, such as the mechanical conduction triangle formed by the battery cabin monitoring point A and three foundation points; the time feature is obtained by statistical analysis and frequency domain transformation of the monitoring data time series, to extract the mean, variance, trend slope, main frequency or periodicity strength and the like; and the physical feature is constructed according to the physical properties of the monitoring parameters, for example, for temperature monitoring data, the physical feature includes the current temperature value, temperature rise rate or thermal conductivity and the like; and for vibration monitoring data, the physical feature includes amplitude, frequency, acceleration peak value or vibration energy density and the like. This three-dimensional feature design realizes comprehensive representation of the spatial position, time evolution and physical nature of the monitoring point, and provides a rich feature information basis for subsequent correlation analysis.

[0043] Further, a feature analysis atlas of the current construction stage is constructed, including:

[0044] The monitoring data of the monitoring network are continuously collected, and the multi-modal monitoring data of each monitoring point are pre-processed and aligned in time and space to obtain multi-modal monitoring data time series with time stamp;

[0045] Based on the multi-modal monitoring data time series and the corresponding monitoring points, the spatial, time and physical three-dimensional features of each monitoring point are constructed, wherein:

[0046] The spatial feature is obtained by Delaunay triangulation to generate the adjacent relationship and spatial influence weight between the monitoring points;

[0047] The time feature is used to represent the law of change of the monitoring data with time, and at least includes any one of the features of mean value, variance and mutation rate extracted by sliding window;

[0048] The physical feature is used to represent the physical properties of the monitoring data, and at least includes dimension, unit and safety threshold;

[0049] The time series of multi-modal monitoring data of each monitoring point is subjected to Granger causality test, the driving and response relationship of the time series of the same physical quantity or physically related physical quantity between different monitoring points is analyzed, the time causal relationship between the monitoring points is extracted, and a space-time joint correlation relationship is generated by fusing the space adjacency relationship between the monitoring points;

[0050] Based on the graph database, the monitoring points are taken as nodes, the attributes thereof include spatial coordinates, three-dimensional features and time stamps, the space-time joint correlation relationship is taken as an edge, and a feature analysis graph of the current construction environment is constructed.

[0051] Specifically, the spatial adjacency relationship between the monitoring points is analyzed by using a Delaunay triangulation method, the method can automatically identify the optimal adjacent connection in space, avoids the subjectivity of manually setting an adjacent threshold, ensures that each monitoring point is connected with the most relevant adjacent point in space, and the generated triangular mesh can accurately reflect the spatial topological structure of the construction site of the energy storage power station. Meanwhile, the time causal relationship between the monitoring data is analyzed by using the Granger causality test, the method can identify whether the historical data of one monitoring point has a prediction capability for the current data of another monitoring point, so as to find the time causal conduction path between the monitoring points, for example, the spatial mechanical conduction network of the monitoring points is established by using the Delaunay triangulation, for example, in the battery cabin hoisting stage, 32 monitoring points distributed at four corners of the cabin and the foundation are subjected to triangular mesh division, the effective conduction path between the adjacent monitoring points is automatically identified, and the spatial conduction strength of each edge is calculated. Meanwhile, the time series correlation is analyzed by using the Granger causality test, for example, it is found that the settlement data (p<0.01) of the foundation monitoring point A can predict the inclination change of the cabin monitoring point B after 2 hours, so that the time causal edge with directionality is established. Finally, the three-dimensional feature fusion is realized through the node-edge structure of the graph database: each node stores structured data such as “node ID: VIB_003, spatial coordinates [35.2, 18.7, 2.1], time feature {trend: 0.12 mm / h, period: 24 h}, physical feature {vibration energy: 1.2 g}”, and the edge relationship records quantitative correlation indexes such as “spatial conduction strength 0.83, time causal strength 0.76”. When the feature analysis graph is constructed based on the graph database, each node stores complete information of a monitoring point, including a spatial feature vector, a time feature vector, a physical feature vector and a time stamp, the edges between the nodes represent the correlation relationship and carry a composite correlation strength as a weight, the graph structure can directly express the complex correlation network between the monitoring points in the construction environment of the energy storage power station, provides a structured knowledge representation for subsequent risk propagation analysis and intelligent decision-making, compared with the traditional table or matrix storage mode, the graph database can efficiently perform complex query operations such as graph traversal, path searching and subgraph matching, and is more suitable for processing correlation analysis tasks in the monitoring network.

[0052] Further, the space-time joint correlation relationship is generated, including:

[0053] The spatial coordinates of the monitoring points in the monitoring network are obtained, a spatial coordinate set is generated, and the spatial coordinate set is divided by using a Delaunay triangulation method to generate a triangular element set;

[0054] The process characteristics of the current construction stage are obtained, the process characteristics include the main operation content performed in the construction stage and the dominant physical disturbance mode caused thereby, and the corresponding maximum effective conduction distance is queried from a preset process-space attenuation parameter table according to the process characteristics;

[0055] All triangular elements in the triangular element set are traversed, the Euclidean distance between the spatial coordinates of the two monitoring points connected by each edge is calculated to obtain the length of each edge, and the spatial influence weight of each edge is calculated in combination with the maximum effective conduction distance, and the edge with a spatial influence weight greater than 0 is defined as an effective conduction path, wherein the spatial influence weight calculation formula is as follows:

[0056] ;

[0057] The key risk mode preset based on the main operation content of the current construction stage is obtained, the target mode to be analyzed is determined according to the key risk mode, the monitoring data time series of the target mode is extracted for a pair of monitoring points connected by each effective conduction path, and a two-by-two Granger causality test is performed;

[0058] According to the F statistic obtained by the Granger causality test, under the condition of meeting the preset significant level, the time causality strength of each effective path is obtained after normalization processing, and the causality direction is recorded;

[0059] The composite correlation strength of each effective conduction path is calculated, and the composite correlation strength = spatial influence weight x time causality strength. The composite correlation strength and the causality direction are jointly used as the space-time joint correlation relationship of each effective conduction path.

[0060] Specifically, first, the three-dimensional spatial coordinates of all activated monitoring points in the current construction stage are extracted to form a coordinate set, for example, the coordinate set P of 32 monitoring points in the battery cabin installation stage {(x1, y1, z1), (x2, y2, z2),..., (x 32 , y 32 , z 32 )} is formed. The coordinate set is projected and divided by using a Delaunay triangulation algorithm, the three-dimensional coordinates are projected onto a two-dimensional plane of a construction operation surface, the triangular mesh is divided by maximizing the minimum internal angle of the triangle, and a triangular element set A = {A1, A2,..., An} composed of a plurality of triangular elements is generated, wherein each triangular element Ai It consists of three monitoring points, such as A1=(P1, P5, P... 12 ) indicates that the monitoring points are P1, P5, and P 12 The resulting triangular mesh is composed of triangles. This triangulation process ensures that the generated triangular mesh has the following characteristics: no four monitoring points are concyclic, that is, the circumcircle of each triangle does not contain other monitoring points, thus guaranteeing the uniqueness and optimality of the triangulation. The generated set of triangular elements covers the convex hull of the entire monitoring area. Each triangle edge represents a potential spatial association path between two monitoring points. For example, the three edges of triangular element A1 are (P1, P5), (P5, P... 12 ), (P 12 These edges (P1) form the basic path network for subsequent calculations of spatial mechanical transmission intensity, ensuring that each monitoring point is connected to its most relevant neighboring monitoring point in space through the shortest path. The maximum effective transmission distance is determined based on the technological characteristics of the current construction stage. These characteristics include parameters such as the type of construction operation, the properties of the components involved, and the scope of the construction impact. Construction operation types include, but are not limited to: foundation excavation, concrete pouring, steel structure hoisting, electromechanical equipment installation, battery module assembly, high-voltage cable laying, and system commissioning. The corresponding dominant physical disturbance types include soil vibration, structural impact, heat conduction, mechanical vibration, structural stress, or electromagnetic interference. The maximum effective transmission distance is determined based on this environmental information because the propagation range of physical impacts varies significantly across different construction stages. For example, the battery compartment installation stage uses precision hoisting technology, and the resulting vibrations and stresses are mainly transmitted through structural connections, resulting in a relatively limited impact range. Based on engineering experience and mechanical analysis, the maximum effective transmission distance is determined to be 15 meters. In contrast, the foundation construction stage uses large-scale mechanical excavation and heavy-duty pouring technology, and the resulting dynamic loads and deformations are propagated through the soil and structural foundation, resulting in a wider impact range. Therefore, the maximum effective transmission distance is set at 25 meters. These conduction distance parameters are pre-set in the system's process-space attenuation parameter table based on engineering practice experience and mechanical propagation theory during energy storage power station construction. The system automatically calls the corresponding conduction distance parameters based on the current construction stage identifier obtained from BIM. Next, the system iterates through all triangular elements generated by Delaunay triangulation, calculating the side length and spatial influence weight for each triangular element's three sides. Taking triangular element A1 as an example, its three sides (P1, P5), (P5, P... 12 ), (P 12 The side lengths of P1 and P2 are calculated as follows: d 512 and d 121Then the spatial influence weight of each edge is calculated by using a linear decay function. When the calculation result is greater than 0, the edge is defined as an effective conduction path and the weight value is recorded. When the calculation result is equal to 0, it indicates that the length of the edge exceeds the effective range of physical influence and does not constitute an effective spatial correlation relationship. In this way, the system filters and quantizes all the edges of the triangular elements, and finally forms a spatial correlation network composed of effective conduction paths, each path carrying a corresponding spatial influence weight, providing a spatial constraint condition for subsequent time causal analysis. Traverse all the effective conduction paths, extract a pair of monitoring points connected by each path. Since each monitoring point may collect multiple modal data such as temperature, vibration and pressure at the same time, and Granger causality test requires input of time series data of the same type, it is necessary to determine the target modal for analysis according to the key risk type of the current construction stage. For example, in the battery installation stage, the focus is on the risk of thermal runaway, and the temperature modal is selected for causal analysis. In the structural construction stage, the focus is on the deformation risk, and the displacement modal is selected for causal analysis. After determining the target modal, the system extracts the data of the same modal from the multi-modal time series of each pair of monitoring points, and then implements the specific process of Granger causality test:

[0061] Obtain a pair of monitoring points connected by each effective conduction path, P q and P g , the target modal monitoring data time series of the P q monitoring point is y t , the corresponding time series of the P g monitoring point is x t , q, g=(1, 2,...,N), q≠g;

[0062] Use Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC) to establish models with different lag orders (usually from 1 to 10) and calculate information criterion values. Select the lag order that minimizes AIC or BIC as the optimal k value. Then establish two autoregressive models. The restricted model is:

[0063] y t-i , i=1, 2,...,k,

[0064] where y t is the target modal monitoring value of the P q monitoring point at time t, c is a constant term, is the autoregressive coefficient of the i-th lag term, y t-i is the i-th lag term of the P q monitoring point, is the residual term, and k is the optimal lag order.

[0065] The unrestricted model is: , wherein y t is the target modal monitoring value of the monitoring point at time t, and c is a constant term, q is the target modal monitoring value of the monitoring point at time t, and c is a constant term, is the autoregressive coefficient of the i-th lag term, y t-i is the target modal monitoring value of the monitoring point at time t, and c is a constant term, q is the i-th lag term of the monitoring point, is the cross-regressive coefficient of the j-th lag term, is the target modal monitoring value of the monitoring point at time t, and c is a constant term, g is the target modal monitoring value of the monitoring point at time t, and c is a constant term, is the residual term, and k is the optimal lag order,

[0066] The model parameters are estimated by the least square method, and the residual sum of squares RSS1 and RSS2 of the two models are calculated, and then the F statistic F = [(RSS1-RSS2) / RSS2]x[(n-2k) / k] is calculated, wherein n is the sample number of the target modal monitoring data, and the significance level is 0.05, and the F distribution critical value F 临界 is searched for the degrees of freedom (k, n-2k), when the calculated F statistic is greater than F 临界 , the null hypothesis is rejected, and it is considered that P q has a Granger causality relationship with P g , otherwise the null hypothesis is accepted, and the null hypothesis is that P q has no Granger causality relationship with P g . The time causality strength is calculated based on the F statistic, and the C = F / (F+n-2k) formula is used to map the F value to the [0, 1] interval as the causality strength index, wherein n-2k is the denominator degree of freedom of the F statistic distribution, and when the F statistic is significant, the causality direction is recorded as P q → P g , and the causality of P g to P q is tested in the reverse direction, and finally the bidirectional time causality strength and direction of each effective conduction path are obtained, which provides a quantitative index in the time dimension for subsequent composite association strength calculation.

[0067] A plurality of sampling granularity levels and their sampling frequencies are preset, and each modal monitoring data obtained for each monitoring point is respectively down-sampled, and an information loss function under each granularity is calculated, and the information loss function under each granularity is compared with a preset feature importance level threshold, and a priority progressive decision process is used to determine the recommended sampling granularity of each monitoring modal.

[0068] Specifically, in the construction environment of energy storage power stations, there are significant differences in the importance of different monitoring features to risk identification, for example, the prediction ability of battery temperature features on thermal runaway risk is much higher than that of environmental humidity features, so different monitoring granularity is needed for data collection and processing, and a sensitivity score model is constructed, which can quantify the influence of each feature on the accuracy of risk identification at different granularity levels; Finally, select the granularity level that maximizes the sensitivity score for each feature as the recommended configuration, ensure that important features have high-precision monitoring resources, and secondary features use moderate monitoring granularity, so that under the constraints of limited computing and storage resources, the performance of the subsequent risk assessment model is maximized. This granularity optimization strategy avoids the "one-size-fits-all" resource allocation method in traditional methods, and realizes the precise placement of monitoring resources, laying a data quality foundation for accurate risk index assessment.

[0069] Further, determining the recommended sampling granularity of each monitoring modality includes:

[0070] presetting three granularity levels of fine, standard and coarse sampling hertz, for each monitoring point, the acquired monitoring data of each modality is respectively down-sampled according to the three sampling hertz, to obtain down-sampled data at three granularities, calculate the variance of each down-sampled data and the variance of the corresponding collected monitoring data, based on the two variances, calculate the information loss rate at each granularity;

[0071] setting a first threshold and a second threshold for each importance level, comparing the information loss rate of each modality monitoring data with the corresponding importance level threshold to determine its recommended sampling granularity:

[0072] If the information loss rate at the coarse granularity is less than the first threshold, the coarse granularity is recommended;

[0073] Otherwise, if the information loss rate at the standard granularity is less than the second threshold, the standard granularity is recommended;

[0074] Otherwise, the fine granularity is recommended, wherein the first threshold is less than the second threshold.

[0075] Specifically, the step is a priority progressive decision process, under the premise of meeting the accuracy, the low sampling frequency is used as much as possible to save resources, starting from the most resource-saving "coarse granularity", if the information loss rate of coarse granularity is small enough (< first threshold) → it means that coarse can be used, and direct selection is made; otherwise, "standard granularity" is tried; if the information loss rate of standard granularity is in the acceptable interval (< second threshold) → standard is selected; otherwise, "fine granularity" must be used. The specific sampling frequency parameters of the three granularity levels are preset, the fine granularity corresponds to 1 Hz sampling frequency, the standard granularity corresponds to 0.5 Hz sampling frequency, and the coarse granularity corresponds to 0.1 Hz sampling frequency. Then, the granularity sensitivity of each modality monitoring data on each monitoring point in the feature analysis graph is tested. The specific process is as follows: according to the preset sampling frequency, the original monitoring data is down-sampled to different degrees, for example, the standard granularity retains every 2 data points to simulate 0.5 Hz effect, and the coarse granularity retains every 10 data points to simulate 0.1 Hz effect. Based on the down-sampled data, the variance value of each feature under different granularity is recalculated. The information loss is evaluated by comparing the change degree of the coarse feature variance and the original feature variance. The calculation formula is information loss rate equals to 1 minus the ratio of coarse feature variance divided by original feature variance. The system sets the threshold according to the feature importance level. The first threshold is 20%, and the second threshold is 50%. For medium importance features, the first threshold is 10%, and the second threshold is 30%. For low importance features, the first threshold is 20%, and the second threshold is 50%. The calculated information loss rate is compared with the preset threshold of importance level. When the loss rate is less than the first threshold, it means that the feature is not sensitive to granularity change and can select coarse granularity to save computing resources. When the loss rate is between the two thresholds, standard granularity is selected to balance the effect and efficiency. When the loss rate is greater than the second threshold, it means that the feature is sensitive to granularity change and must select fine granularity to ensure information integrity. Finally, the individualized granularity configuration scheme of each monitoring modality is formed. The individualized granularity configuration scheme of each monitoring modality includes feature identification, monitoring point position, importance level, recommended granularity level (fine / standard / coarse), corresponding sampling frequency parameter and decision basis information, forming a complete monitoring data-granularity configuration mapping relationship.

[0076] According to the recommended sampling granularity of each monitoring modality, new multi-modal monitoring data is continuously collected in the monitoring network of the current construction environment, the acquired data is enhanced, the three-dimensional feature and edge complex association strength of the feature analysis graph nodes are updated, and the dynamic evolution of the graph is realized.

[0077] Specifically, according to the obtained personalized granularity configuration scheme of each feature, the multi-modal monitoring data collection of the energy storage power station construction site is accurately enhanced, such as the system reading the feature identifier, recommended granularity level and corresponding sampling frequency parameters in the granularity configuration scheme, and then sending configuration instructions to the related sensors in the monitoring network to adjust the sampling frequency, data accuracy and storage strategy, for example, adjusting the recommended fine granularity temperature feature to 1Hz high frequency sampling and 16-bit precision processing, and adjusting the recommended coarse granularity illumination feature to 0.1Hz low frequency sampling and 8-bit precision processing, to realize differentiated data collection strategy. The system continuously collects multi-modal monitoring data optimized in granularity.

[0078] Further, it is characterized in that the three-dimensional feature and edge complex association strength of the feature analysis graph node is updated to realize dynamic evolution of the graph, including:

[0079] For the collected new multi-modal monitoring data, a differentiated data enhancement strategy is adopted to improve its representation quality, wherein,

[0080] For the recommended fine granularity sampling, a fine-grained local enhancement strategy is adopted, based on the sampling data, the local dynamic trend is extracted using a sliding window, and the possible short-time missing is filled by combining linear or spline interpolation algorithm, to further improve the time resolution and signal continuity;

[0081] For the recommended standard granularity sampling, a mixed granularity hierarchical enhancement strategy is adopted, the collected data is processed by time domain filtering, and the trend changes of adjacent monitoring points are combined to identify and correct the obvious deviating abnormal values;

[0082] For the recommended coarse granularity sampling, a coarse-grained statistical enhancement strategy is adopted, the low-frequency sampling data is subjected to mean or variance statistics, and the exponential smoothing or Kalman filter is applied to suppress noise fluctuations, and the long-term trend features are retained;

[0083] The new multi-modal monitoring data processed by the data enhancement strategy is subjected to spatio-temporal synchronization processing, based on the enhanced data after spatio-temporal synchronization, the three-dimensional features of each node are re-extracted, and the spatial transmission strength and temporal causal strength between nodes are re-calculated;

[0084] The enhanced monitoring data after spatio-temporal synchronization is fed back to the feature analysis graph, the three-dimensional features of each node are re-extracted, and the complex association strength of the edges between nodes is re-calculated;

[0085] The updated three-dimensional features of the nodes and the complex association strength of the edges between the nodes are injected into the feature analysis graph, completing the dynamic evolution of the graph, representing the state evolution trend of the current construction environment.

[0086] Specifically, according to the obtained feature granularity configuration result, different data enhancement processing strategies are adopted for features of different importance. For high importance features of fine granularity configuration, the system adopts a fine-grained local enhancement strategy. The specific process is to set a small sliding window (such as 5-10 data points) to perform local scanning on the original monitoring data, generate denser interpolation points between adjacent data points through cubic spline interpolation or Lagrange interpolation algorithm, enhance the original 1Hz sampling data to 2-5Hz high-density time series, and improve the positioning accuracy of spatial coordinates by using bilinear interpolation method to ensure that important features obtain the most complete information expression. For medium importance features of standard granularity configuration, the system adopts a mixed granularity hierarchical enhancement strategy. First, a larger sliding window (such as 20-30 data points) is used to calculate local statistical features including mean, variance and trend, then local interpolation technology is used for fine processing in key time periods, and statistical smoothing method is used for noise reduction processing in smooth time periods, realizing the organic combination of fine enhancement and statistical enhancement, balancing data quality and computing efficiency. For low importance features of coarse granularity configuration, the system adopts a coarse-grained statistical enhancement strategy. By setting a large window (such as 50-100 data points) to segment and aggregate the original data, the statistical representative values such as median and quartile in each time period are calculated, and moving average or Gaussian smoothing filter is used to eliminate high-frequency noise, compressing the data to a lower time resolution but keeping the main trend, saving computing resources while ensuring that secondary features can still provide effective monitoring information. The monitoring data after granularity enhancement processing has undergone significant changes in quality and accuracy. Important features have higher time resolution and spatial accuracy, and secondary features have reduced noise interference through statistical smoothing. These improved data must be fed back to the feature analysis atlas to play a role, otherwise the atlas based on the original rough data cannot reflect the optimization effect.

[0087] After the differential enhancement and spatio-temporal synchronization of multi-modal monitoring data are completed, the system starts the dynamic evolution of the feature analysis graph. First, the spatial coordinates of all monitoring points are obtained, and the spatial topology structure is constructed by using Delaunay triangulation to generate a set of triangular elements, ensuring the rationality of geometric distribution. Combined with the process characteristics of the current construction stage (such as concrete pouring, prestressed tensioning, etc.) and the dominant physical disturbance mode it causes, the system queries the corresponding maximum effective conduction distance from the pre-set "process-space attenuation parameter table". Traverse all edges of the triangular element, calculate the Euclidean distance between the connected nodes, and evaluate the spatial influence weight combined with the maximum effective conduction distance to select the edges with a spatial influence weight greater than zero as the effective conduction path. According to the pre-set key risk mode of the current construction stage, the target physical mode (such as strain, temperature, etc.) to be analyzed is determined, and for each node pair on the effective path, the target mode monitoring data time series in the latest time window is extracted for Granger causality test. Under the condition of meeting the pre-set significance level, the test results are normalized to obtain the time causality strength of each path, and the causality direction is recorded. Subsequently, the composite correlation strength of each effective conduction path is calculated, which is the product of the spatial influence weight and the time causality strength, comprehensively reflecting the joint correlation characteristics of the path in spatial accessibility and temporal drivability. At the same time, based on the synchronized enhanced data, the spatial position, time state and physical quantity value of each node are re-extracted to form a three-dimensional feature vector. The updated node features and the composite correlation strength and causality direction of the effective path are injected into the feature analysis graph to replace the original attributes, completing the state refresh of the graph. By periodically executing this process, the graph continuously evolves, dynamically representing the disturbance propagation path and risk evolution trend in the construction environment, achieving precise perception and interpretable modeling of the structure state.

[0088] Based on the evolved feature analysis graph, the global topology feature, key path feature and key node feature are extracted, input into the pre-trained risk assessment model, and the continuous risk index of the current construction environment is output. When the risk index exceeds the pre-set safety threshold, a graded warning response is triggered.

[0089] Specifically, the construction environment of energy storage power station is complex and changeable, and risk factors interact with each other. The traditional single data source evaluation method is easy to produce misjudgment or omission. By fusing the accuracy of real-time monitoring data and the relevance of feature graph, the accuracy of risk identification and the reliability of risk index calculation can be significantly improved, realizing precise quantitative evaluation of the safety state of the construction environment.

[0090] Further, the continuous risk index of the current construction environment is output, including:

[0091] Periodically acquire the evolved feature analysis graph state, extract global topology features, key path features and key node features based on graph structure analysis method, wherein the global topology features include node risk density, correlation edge strength mean and variance, network aggregation coefficient, the key path features are the first K paths with the highest composite correlation strength in the graph, and the key node features are the node feature value change trend on the key path;

[0092] The global topology features, the key path features and the key node features are combined into an input vector, which is input into a pre-trained gradient boosting tree model to output a continuous risk index;

[0093] The risk index is compared with a preset safety threshold in real time, and a hierarchical alarm protocol is triggered when the risk index exceeds the threshold.

[0094] Specifically, this step quantitatively calculates the risk level of the current construction environment. The specific steps are as follows: Dynamic feature extraction of the graph, periodically (triggered by a pre-set time interval or event-driven) generate snapshots for the dynamically evolving feature analysis graph, extract the following dynamic evolution feature sequences: Global topology features, including node risk density: calculate the proportion of the number of nodes in the graph whose eigenvalues exceed their physical security threshold to the total number of nodes, correlation edge strength mean and variance: calculate the average value of "space-time compound correlation strength" on all edges and its fluctuation. A decrease in the mean or an increase in the variance may indicate system instability, network clustering coefficient: measures the degree of clustering between monitoring points, an abnormal increase in the clustering coefficient may indicate that risk is accumulating locally; Key path features, identify the top K paths (such as TOP5) in the graph with compound correlation strength sorted from high to low as key risk transmission paths, calculate the trend of correlation strength on these paths (such as the slope obtained by linear fitting), K>0. A negative slope indicates that the transmission relationship is weakening or failing, a positive slope and a value that is too large indicate that risk is accelerating; Node feature evolution, for nodes on the key path, track the first-order difference (change) of their eigenvalues over time to form a change sequence. Calculate the statistics (such as mean, standard deviation) of this sequence to determine whether the change is smooth. Node eigenvalues refer to specific attributes or metrics that each node has, which can be physical quantities directly monitored from the construction environment or derived indicators calculated based on these physical quantities. In different application scenarios, the specific meaning of node eigenvalues may vary. Physical quantity monitoring values: such as temperature, humidity, stress, strain, etc. can be directly obtained from sensors. Derived indicators: some indicators calculated based on raw monitoring data, such as change rate (such as stress change rate), cumulative damage indicator (calculated based on fatigue damage theory), etc. The trend of key node eigenvalues can be obtained by calculating the first-order difference of these node eigenvalues (i.e. the difference between adjacent time points), and further analyzing the statistical properties (such as mean, standard deviation) of these change sequences to determine whether they are smooth or have abnormal fluctuations. This method helps to timely discover potential risk concentration areas or impending failure points. Input the above dynamic features into a trained risk rating model (such as Gradient Boosting Tree GBDT), and output a continuous risk index R (range 0-100). An example of the input feature vector of the model: Input feature vector = [node risk density (0.35), edge strength mean (0.72), edge strength variance (0.18), clustering coefficient (0.41), key path 1 strength trend (-0.05),..., standard deviation of key node A settlement change rate (0.12)].The risk rating model is a pre-trained gradient boosting tree model. A plurality of feature analysis graph snapshots are constructed based on historical construction project data. Dynamic evolution features corresponding to each snapshot are extracted as input. A risk index associated with the time distance of a risk event is used as a label. A gradient boosting tree algorithm is used for supervised learning to obtain a pre-trained model for online evaluation of the current construction environment risk level. The specific process is as follows: historical construction data is obtained. A plurality of historical engineering project data of completed construction is collected. Each data includes: original time series observation values (such as temperature, stress, displacement, vibration, and other physical quantities) of each monitoring point in the entire construction process, risk events (such as structure cracking, equipment overheating, support instability, etc.) recorded in the construction log and their occurrence time and severity level, and corresponding construction stage information; a historical feature analysis graph containing node three-dimensional feature vectors (space, time, and physics) and dynamic associated edges is constructed; the model input historical feature vectors are extracted. According to the occurrence time of the historical risk event, a corresponding risk index label Rt is generated for each graph snapshot moment, and the value range is [0, 100]. The calculation formula is: where Δt represents the time interval from the current time t to the occurrence of the next risk event, λ is a decay coefficient, and the value range is (0.1, 1.0). The decay coefficient is used to control the risk accumulation rate, and is preferably 0.5. The input feature vectors and corresponding risk index labels of all historical projects are combined to form a training sample set, which is randomly divided into a training set and a test set. A gradient boosting tree algorithm (such as XGBoost or LightGBM) is used for supervised learning training. The goal is to minimize the mean square error between the predicted risk index and the true label. The model hyperparameters are optimized through cross-validation, and the model performance is evaluated on the test set to ensure that the determination coefficient R 2 >0.85, and the mean absolute error (MAE) is less than 8. The GBDT model parameters that meet the verification criteria are serialized and saved (such as saved as a.pkl or.json format file) as a "pre-trained risk rating model" integrated into the online monitoring system. The overall risk index of the current construction environment of the energy storage power station is calculated. The index is a continuous value between 0 and 100. The higher the value, the greater the construction environment risk. The calculated risk index is compared with the safety threshold preset based on the construction safety specifications of the energy storage power station in real time. Multiple threshold values are usually set, such as a low risk threshold of 25, a medium risk threshold of 50, and a high risk threshold of 75. When the calculated risk index exceeds the corresponding threshold value, the system immediately triggers a graded alarm of the corresponding level. Finally, alarm information including specific risk location, risk level, etc. is output to the construction management personnel through various ways such as audible and visual alarms, SMS notifications, and system interface pop-ups, to ensure that the construction safety risk can be responded and disposed in a timely and effective manner.

[0095] In the second embodiment, the same inventive concept as in the foregoing embodiments is used to construct an energy storage power station construction environment intelligent monitoring method, which is as follows: Figure 2As shown, the application provides an intelligent monitoring system for construction environment of energy storage power station, comprising:

[0096] A monitoring network construction module is configured to obtain the current construction stage of the energy storage power station from the BIM system, match the multi-modal monitoring data required for monitoring from the construction environment description of the current construction stage, start the sensors on the relevant monitoring points, and form a monitoring network of the current construction environment;

[0097] A graph construction module is configured to extract spatial, temporal and physical three-dimensional features for the monitoring points in the monitoring network based on the data monitored by the monitoring network, analyze the spatial adjacency relationship and temporal causal relationship of the monitoring points, obtain the spatial-temporal correlation relationship between the monitoring points, and construct a feature analysis graph of the current construction stage with the monitoring points as nodes and the correlation relationship as edges, wherein the spatial-temporal correlation relationship includes composite correlation strength and causal direction.

[0098] A granularity configuration module is configured to preset a plurality of sampling granularity levels and their sampling frequencies, respectively perform down-sampling on each kind of modal monitoring data obtained for each monitoring point, calculate the information loss function under each granularity, compare the information loss function under each granularity with the preset feature importance level threshold, and determine the recommended sampling granularity of each monitoring mode according to a priority progressive decision process.

[0099] A graph dynamic evolution module is configured to continue collecting new multi-modal monitoring data in the monitoring network of the current construction environment according to the recommended sampling granularity of each monitoring mode, enhance the obtained data, update the three-dimensional features of the feature analysis graph nodes and the composite correlation strength of the edges, and realize dynamic evolution of the graph.

[0100] A construction environment risk module is configured to extract global topology features, key path features and key node features based on the evolved feature analysis graph, input a pre-trained risk assessment model, and output a continuous risk index of the current construction environment. When the risk index exceeds a preset safety threshold, a hierarchical early warning response is triggered.

[0101] It should be noted that the above sequence of the embodiments of the application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes a specific embodiment of the present application. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0102] The above only describes the preferred embodiments of the application and does not limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the application shall be included in the protection scope of the application.

[0103] The specification and drawings are, of course, subject to various interpretations and should not be viewed in any limiting sense. It will be understood that various modifications and changes can be made to the application disclosed without departing from the scope of the application. It is therefore intended that the application be limited only by the scope of the appended claims, including any amendments thereof, and their equivalents.

Claims

1. A method for intelligent monitoring of construction environment of energy storage power station, characterized in that, Comprise: Obtain the current construction stage of the energy storage power station from the BIM system, match the multi-modal monitoring data required for monitoring from the construction environment description of the current construction stage, start the sensors on the relevant monitoring points, and form the monitoring network of the current construction environment; Based on the data monitored by the monitoring network, extract the spatial, temporal and physical three-dimensional features of the monitoring points in the monitoring network, analyze the spatial adjacency relationship and temporal causality of each monitoring point, obtain the spatial-temporal correlation relationship between each monitoring point, and construct the feature analysis graph of the current construction stage with the monitoring points as nodes and the correlation relationship as edges, wherein the spatial-temporal correlation relationship includes composite correlation strength and causal direction; Pre-set multiple sampling granularity levels and their sampling frequencies, respectively down-sample each type of monitoring data obtained for each monitoring point, calculate the information loss function at each granularity, compare the information loss function at each granularity with the pre-set feature importance level threshold, and determine the recommended sampling granularity of each monitoring modality according to the priority progressive decision process; According to the recommended sampling granularity of each monitoring modality, continue to collect new multi-modal monitoring data in the monitoring network of the current construction environment, enhance the obtained data, update the three-dimensional features of the nodes and the composite correlation strength of the edges of the feature analysis graph, and realize the dynamic evolution of the graph; Based on the evolved feature analysis graph, extract the global topology feature, key path feature and key node feature, input them into the pre-trained risk assessment model, and output the continuous risk index of the current construction environment. When the risk index exceeds the pre-set safety threshold, trigger the hierarchical early warning response.

2. The intelligent monitoring method for construction environment of energy storage power station according to claim 1, characterized in that, Match the multi-modal monitoring data required for monitoring from the construction environment description of the current construction stage, comprising: Obtain the construction environment of the current construction stage of the energy storage power station described by words and / or language, analyze the current environment description using NLP technology, and identify the key monitoring requirements of the current construction environment; Based on the key monitoring requirements of the current construction environment, retrieve the pre-constructed sensor capability matrix and monitoring point information database, match the optimal sensor combination through the greedy algorithm, start the sensors on the relevant monitoring points, form the monitoring network of the current construction environment, and obtain the multi-modal monitoring data required for monitoring, wherein the sensor capability matrix is a structured data table for providing sensor technical specifications and performance parameters, and the monitoring point information database is used to provide monitoring point spatial position, coverage range and device access information.

3. The intelligent monitoring method for construction environment of energy storage power station according to claim 1, characterized in that, Construct the feature analysis graph of the current construction stage, comprising: Continuously collect monitoring data of the monitoring network, pre-process and time-space align the multi-modal monitoring data of each monitoring point, and obtain the multi-modal monitoring data time series with time stamp; Based on the multi-modal monitoring data time series and the corresponding monitoring points, construct the spatial, temporal and physical three-dimensional features of each monitoring point, wherein: The spatial feature generates the adjacency relationship and spatial influence weight between the monitoring points through Delaunay triangulation; The time feature is used to represent the law of monitoring data changing with time, and at least includes any one of the mean value, variance and mutation rate extracted through the sliding window; The physical features are used to represent physical properties of the monitoring data, and at least include dimensions, units, and safety thresholds; The Granger causality test is performed on the multi-modal monitoring data time series of each monitoring point to analyze the driving and response relationship of the time series of the same physical quantity or physically related physical quantity between different monitoring points, extract the time causality relationship between the monitoring points, and fuse the space-time joint correlation relationship with the spatial adjacency relationship between the monitoring points; Based on the graph database, the monitoring points are taken as nodes, the attributes of which include spatial coordinates, three-dimensional features, and time stamps, and the space-time joint correlation relationship is taken as an edge to construct a feature analysis graph of the current construction environment.

4. The intelligent monitoring method for construction environment of energy storage power station according to claim 3, characterized in that, The space-time joint correlation relationship is fused, including: The spatial coordinates of the monitoring points in the monitoring network are obtained to generate a spatial coordinate set, and the spatial coordinate set is divided by using a Delaunay triangulation method to generate a triangular element set; The process characteristics of the current construction stage are obtained, the process characteristics including the main operation content and the dominant physical disturbance mode caused by the main operation content in the construction stage, and the corresponding maximum effective transmission distance is queried from a preset process-space attenuation parameter table according to the process characteristics; All triangular elements in the triangular element set are traversed, the Euclidean distance between the spatial coordinates of the two monitoring points connected by each edge is calculated to obtain the length of each edge, and the spatial influence weight of each edge is calculated in combination with the maximum effective transmission distance, and the edges with a spatial influence weight greater than 0 are defined as effective transmission paths, wherein the spatial influence weight calculation formula is as follows: ； The key risk mode preset based on the main operation content of the current construction stage is obtained, the target modality to be analyzed is determined according to the key risk mode, and the monitoring data time series of the target modality is extracted for a pair of monitoring points connected by each effective transmission path, and a Granger causality test is performed therebetween; According to the F statistic obtained by the Granger causality test, the time causality strength of each effective path is obtained after normalization under the condition of meeting the preset significant level, and the causality direction is recorded; The composite correlation strength of each effective transmission path is calculated, and the composite correlation strength = spatial influence weight × time causality strength. The composite correlation strength and the causality direction are taken as the space-time joint correlation relationship of each effective transmission path.

5. The intelligent monitoring method for construction environment of energy storage power station according to claim 1, characterized in that, The recommended sampling granularity of each monitoring modality is determined, including: Three sampling hertz of fine, standard, and coarse granularity levels are preset, the monitoring data of each modality obtained for each monitoring point is respectively down-sampled according to the three sampling hertz to obtain down-sampled data at three granularities, the variance of each down-sampled data and the variance of the collected monitoring data corresponding thereto are calculated, and the information loss rate at each granularity is calculated based on the two variances; The first threshold and the second threshold are set for each importance level, the information loss rate of each modality monitoring data is compared with the threshold of the corresponding importance level to determine the recommended sampling granularity: If the information loss rate at the coarse granularity is less than the first threshold, the coarse granularity is recommended; Otherwise, if the information loss rate at the standard granularity is less than the second threshold, the standard granularity is recommended; Otherwise, recommend fine granularity, wherein the first threshold is less than the second threshold.

6. The intelligent monitoring method for construction environment of energy storage power station according to claim 5, characterized in that, Update the three-dimensional feature of the feature analysis graph node and the composite association strength of the edge to realize the dynamic evolution of the graph, including: Collect new multi-modal monitoring data, and use differentiated data enhancement strategies to improve its representation quality, wherein, For fine-grained sampling recommendation, use fine-grained local enhancement strategy, based on sampling data, use sliding window to extract local dynamic trend, and combine linear or spline interpolation algorithm to fill possible short-term missing, further improve time resolution and signal continuity; For standard granularity sampling recommendation, use mixed granularity hierarchical enhancement strategy, time domain filtering processing is performed on the collected data, and the trend change of the adjacent monitoring points is combined to identify and correct the obvious abnormal values; For rough grain sampling recommendation, use coarse-grained statistical enhancement strategy, mean or variance statistics is performed on low-frequency sampling data, and exponential smoothing or Kalman filter is applied to suppress noise fluctuations and retain long-term trend characteristics; Perform spatio-temporal synchronization processing on the new multi-modal monitoring data processed by the data enhancement strategy, and based on the enhanced data after spatio-temporal synchronization, re-extract the three-dimensional features of each node and re-calculate the spatial transmission strength and temporal causality strength between nodes; The enhanced monitoring data after spatio-temporal synchronization is fed back to the feature analysis graph, the three-dimensional features of each node are re-extracted, and the composite association strength of the edge between nodes is re-calculated. The updated three-dimensional features of the nodes and the composite association strength of the edges between the nodes are injected into the feature analysis graph to complete the dynamic evolution of the graph, representing the state evolution trend of the current construction environment.

7. The intelligent monitoring method for construction environment of energy storage power station according to claim 6, characterized in that, Perform construction environment risk index assessment, including: Periodically obtain the evolved feature analysis graph state, extract global topology features, key path features, and key node features based on graph structure analysis methods, wherein the global topology features include node risk density, associated edge strength mean and variance, network clustering coefficient, the key path features are the top K paths in the graph with the highest composite association strength, and the key node features are the node feature value trend on the key path; Input the global topology features, key path features, and key node features into the pre-trained gradient boosting tree model to output a continuous risk index. Real-time comparison of risk index and preset safety threshold, triggering hierarchical alarm protocol when risk index exceeds threshold.

8. The intelligent monitoring system for construction environment of energy storage power station, characterized in that, Including: A monitoring network construction module for obtaining the current construction stage of the energy storage power station from the BIM system, matching the multi-modal monitoring data required for monitoring from the construction environment description of the current construction stage, starting the sensors on the relevant monitoring points, and forming the monitoring network of the current construction environment; A graph construction module for extracting spatial, temporal, and physical three-dimensional features for monitoring points in the monitoring network based on the data monitored by the monitoring network, analyzing the spatial adjacency relationship and temporal causality relationship of each monitoring point, obtaining the spatial-temporal association relationship between each monitoring point, and constructing the feature analysis graph of the current construction stage, wherein the spatial-temporal association relationship includes composite association strength and causality direction; A granularity configuration module is configured to preset multiple sampling granularity levels and their sampling frequencies, to respectively perform down-sampling on each type of modal monitoring data obtained for each monitoring point, to calculate information loss functions at each granularity, to compare the information loss functions at each granularity with preset characteristic importance level thresholds, and to determine a recommended sampling granularity for each monitoring modality according to a decision-making process in which priorities are progressively increased. A graph dynamic evolution module is configured to continue collecting new multi-modal monitoring data in the monitoring network of the current construction environment according to the recommended sampling granularity for each monitoring modality, to enhance the obtained data, to update the composite association strength of the three-dimensional features and edges of the feature analysis graph nodes, and to realize dynamic evolution of the graph. A construction environment risk module is configured to extract global topology features, key path features and key node features based on the evolved feature analysis graph, to input the pre-trained risk assessment model, and to output a continuous risk index of the current construction environment. When the risk index exceeds a preset safety threshold, a graded early warning response is triggered.

Images