BIM-based pressure storage power station underground chamber construction monitoring method and system

Abstract

The invention is suitable for the field of pressure storage power station construction monitoring, and particularly relates to a BIM-based pressure storage power station underground chamber construction monitoring method and system, and the method comprises the steps: obtaining monitoring data used for representing the surrounding rock state of an underground chamber in a construction model; extracting feature distribution of the monitoring data in a time dimension by using a long-short term memory neural network; extracting feature distribution of the monitoring data in a spatial dimension based on a graph attention network; fusing the feature distribution of the monitoring data in the time dimension and the space dimension by adopting a cross attention fusion network structure, and constructing fusion features of the monitoring data; and taking the fusion features as input of a pre-constructed prediction model, and outputting a surrounding rock state prediction result. According to the invention, the BIM model and the monitoring data are combined, so that the construction state of the underground chamber can be comprehensively understood; and on the basis of feature extraction of monitoring data, monitoring and early warning of the surrounding rock state are realized, and the technical problem of poor accuracy of a monitoring result of a traditional monitoring method is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of construction monitoring of a pressure storage power station, and in particular relates to a BIM-based construction monitoring method and system for an underground chamber of a pressure storage power station. Background Art

[0002] The core of BIM is to establish a virtual three-dimensional model of the construction project and use digital technology to provide the model with a complete construction project information database that is consistent with the actual situation; the information database not only contains geometric information, professional attributes and status information describing the building components, but also includes status information of non-component objects; with the help of this three-dimensional model containing construction project information, the degree of information integration of the construction project is greatly improved, thereby providing a platform for engineering information exchange and sharing for the relevant stakeholders of the construction project.

[0003] At present, compressed air energy storage is a technology that can store electric energy on a large scale. It mainly includes key components such as generators, compressors, combustion chambers, gas storage chambers, expanders and electric motors, and is divided into two processes: energy storage and energy release. In the energy storage process, electricity is used to drive the compressor to compress air and store high-pressure air in the gas storage chamber; in the energy release process, the high-pressure air in the gas storage chamber drives the expander to work and generate electricity; the underground chamber of the compressed air storage power station is an underground structure used for compressed air energy storage systems, which is used to store high-pressure air; underground gas storage devices can mainly use mined geological caves, naturally formed water-bearing rock formations, mined or specially excavated salt caves, underground caves artificially excavated in hard rocks and other structures.

[0004] The construction of the underground chamber of the pressure storage power station is an important part of the compressed air energy storage system. The construction of the underground chamber of the pressure storage power station is difficult and has high safety risks. Safety monitoring and early warning are important means to ensure construction safety. In related technologies, the monitoring of the underground chamber of the pressure storage power station generally focuses on the state of the surrounding rock. Specifically, when using BIM technology for construction monitoring, traditional monitoring technology cannot effectively extract the features of the measured data of the surrounding rock state collected by the sensor, import the measured data into the model, and realize the construction monitoring process. Since it is impossible to effectively extract data feature information, it is impossible to provide accurate early warning, and the monitoring results obtained are less accurate and cannot truly reflect the current state of the surrounding rock of the underground chamber of the pressure storage power station. Summary of the invention

[0005] The purpose of the embodiment of the present invention is to provide a BIM-based underground chamber construction monitoring method and system for a pressure storage power station. By adopting a cross-attention fusion network structure to fuse the characteristic distribution of surrounding rock monitoring data in the time dimension and the space dimension, it is beneficial to the prediction of the surrounding rock state, and more accurate surrounding rock state data can be obtained, thus solving the technical problem of poor accuracy of monitoring results of traditional monitoring methods.

[0006] To achieve the above object, the present invention provides the following technical solutions.

[0007] An embodiment of the present invention provides a BIM-based underground chamber construction monitoring method for a pressure storage power station, the monitoring method comprising:

[0008] Construct a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station;

[0009] Acquire monitoring data used to characterize the state of the surrounding rock of the underground chamber in the construction model, wherein the monitoring data includes deformation data of the underground chamber vault settlement and horizontal convergence and the time point at which the deformation data is acquired;

[0010] The long short-term memory neural network is used to extract the characteristic distribution of monitoring data in the time dimension;

[0011] Extract the feature distribution of monitoring data in the spatial dimension based on the graph attention network;

[0012] The cross-attention fusion network structure is used to fuse the characteristic distribution of monitoring data in time and space dimensions to construct the fusion features of monitoring data;

[0013] The fusion features are used as input of a pre-constructed prediction model, and a surrounding rock state prediction result is output, and the surrounding rock state in the construction model of the underground chamber of the pressure storage power station is updated using the surrounding rock state prediction result.

[0014] Furthermore, the step of constructing a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station includes:

[0015] Constructing a BIM basic model of an underground chamber of a pressure storage power station including structural components and sensor components, wherein the BIM basic model is identical to the underground chamber structure at the construction site;

[0016] Acquire construction data of sensor components set in the construction site, establish a mapping relationship between the construction data and the sensor components in the BIM basic model, add the construction data to the BIM basic model based on the mapping relationship, and obtain a construction model.

[0017] Furthermore, the step of extracting the characteristic distribution of monitoring data in the time dimension by using the long short-term memory neural network includes the following steps:

[0018] The monitoring data of the surrounding rock state are normalized and sequence segmented to obtain the monitoring sequence data of the surrounding rock state;

[0019] The input gate, forget gate and output gate of the long short-term memory neural network are used to extract the characteristic representation of the monitoring sequence data of the surrounding rock state in the time dimension; among which:

[0020] The input gate is represented as: i t =σ(W i ·[h t-1 , x t ]+b i );

[0021] The forget gate is represented by: f t =σ(W f ·[h t-1 , x t ]+b f );

[0022] The output gate is represented as: o t =σ(W o ·[h t-1 , x t ]+b o );

[0023] In the formula, σ represents the activation function; i t represents the input gate; o t represents the output gate;

[0024] f t represents the forget gate; h t-1 Indicates the hidden layer state at the previous moment; x t represents the monitoring sequence data at time t; W i 、b i Respectively represent the weight matrix and bias of the input gate; W f 、b f Respectively represent the weight matrix and bias of the forget gate; W o 、b o Represent the weight matrix and bias of the output gate respectively;

[0025] The characteristic of the monitoring sequence data of the surrounding rock state in the time dimension is expressed as F t ,in:

[0026]

[0027] In the formula, C t-1 and Represent the cell state at the previous moment and the candidate cell state respectively.

[0028] Furthermore, the feature distribution of the monitoring data in the spatial dimension is extracted based on the graph attention network, including:

[0029] Construct a spatiotemporal graph of surrounding rock monitoring data, taking each monitoring point in the surrounding rock monitoring area as a node of the spatiotemporal graph, and the characteristic matrix of the node is the data collected by the corresponding monitoring point;

[0030] The self-attention mechanism is used to focus on the information of each node in the spatiotemporal graph, and the characteristic distribution of monitoring data in the spatial dimension is extracted based on the attention correlation between nodes.

[0031] Furthermore, the step of using each monitoring point in the surrounding rock monitoring area as a node of the space-time graph includes:

[0032] Determine the edge relationship between two monitoring points, expressed as:

[0033]

[0034] Where: A ij represents the neighboring relationship between monitoring point j and monitoring point j, 1 means that the two monitoring points are connected, and 0 means that they are not connected; d(i,j) represents the Euclidean distance between the two monitoring points; D represents the threshold used to determine whether there is an edge relationship between the monitoring points;

[0035] According to the neighbor relationship and time sequence, a space-time graph sequence is constructed:

[0036]

[0037] Where Z represents the number of monitoring points in the target surrounding rock area; n represents the length of the time series; V represents the number of features at each monitoring point; represents the characteristic matrix of the i-th monitoring point at time t; S i Represents the characteristic matrix of all monitoring points in the target surrounding rock area at time t.

[0038] Furthermore, the step of extracting the feature distribution of the monitoring data in the spatial dimension based on the attention correlation between the nodes includes:

[0039] Calculate the attention correlation between monitoring point i and monitoring point j. The calculation formula is expressed as:

[0040] e ij =a tt (Ws i ,Ws j );

[0041] In the formula, e ij Indicates the importance of monitoring point i to monitoring point j; a tt (·) represents a single-layer feedforward neural network; s i represents the feature vector of monitoring point i; s j represents the feature vector of monitoring point j; W represents the model training parameters;

[0042] The attention coefficient a between monitoring point i and monitoring point j ij The calculation is done using the following formula:

[0043]

[0044] In the formula, Q i represents the set of all monitoring points that have an edge relationship with monitoring point i, and LeakyRelu represents a nonlinear activation function;

[0045] After obtaining the attention coefficient, the linear combination of the feature vectors of monitoring point i and its connected monitoring points is calculated to obtain the aggregate representation of each monitoring point i Then the feature distribution at time t in the spatial dimension is expressed as:

[0046]

[0047] Where Z represents the number of monitoring points in the target surrounding rock area.

[0048] Furthermore, the step of adopting a cross-attention fusion network structure to fuse the feature distribution of the monitoring data in the time dimension and the space dimension to construct the fusion feature of the monitoring data includes:

[0049] Given a time dimension feature set

[0050] And, given the spatial dimension feature set

[0051] Use the dimension conversion function to unify the time dimension feature set T and the space dimension feature set P across dimensions into feature representations of the same size;

[0052] Calculate the correlation matrix of the time dimension feature set T and the space dimension feature set P respectively, use the correlation matrix as the weight of the time dimension feature set T, fuse the time dimension feature set T and the space dimension feature set P, and get the fusion feature F TP ,in:

[0053]

[0054] Where LN(·) represents a linear transformation function, which is used to transform the time dimension feature set T into a feature map with the same dimension as the spatial dimension feature set P; d represents the size of the feature dimension.

[0055] Further, in a preferred embodiment of the present disclosure, the pre-constructed prediction model is an LSTM neural network prediction model constructed based on the TensorFlow.js framework;

[0056] Among them, in the hyperparameters of the prediction model, the number of input layer neurons is set to 6, the number of hidden layer neurons is set to 12, the number of output layer neurons, i.e., the number of fully connected layer input layer neurons is set to 10, the number of fully connected layer hidden layer neurons is set to 10, and the number of output layer neurons is set to 1; in addition, the initial step size of the model is set to 4, the number of iterations is set to 3000 times, and the learning rate is set to 0.005.

[0057] Furthermore, the pre-constructed prediction model uses a particle swarm optimization algorithm to perform parameter optimization;

[0058] In the initial stage, the particle swarm optimization algorithm adopts a concave function decreasing strategy to reduce the inertia weight coefficient nonlinearly; wherein the concave function decreasing strategy is expressed as:

[0059]

[0060] In the formula, ω max and ω min are the maximum and minimum values ​​of the inertia weight respectively; t represents the current iteration number; t max represents the maximum number of iterations; k is a control parameter used to control the smoothness of the ω curve.

[0061] Another embodiment of the present invention provides a BIM-based underground chamber construction monitoring system for a pressure storage power station, the monitoring system comprising:

[0062] A model building unit, used to build a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station;

[0063] A data acquisition unit, used to acquire monitoring data for characterizing the state of the surrounding rock of the underground chamber in the construction model, wherein the monitoring data includes deformation data of the underground chamber vault settlement and horizontal convergence and the time point at which the deformation data is acquired;

[0064] A first feature extraction unit, used to extract feature distribution of monitoring data in a time dimension using a long short-term memory neural network;

[0065] A second feature extraction unit is used to extract the feature distribution of monitoring data in the spatial dimension based on the graph attention network;

[0066] A feature fusion unit is used to fuse the feature distribution of monitoring data in the time dimension and the space dimension by using a cross-attention fusion network structure to construct a fusion feature of the monitoring data;

[0067] The prediction updating unit is used to use the fusion feature as the input of the pre-constructed prediction model, output the surrounding rock state prediction result, and use the surrounding rock state prediction result to update the surrounding rock state in the construction model of the underground chamber of the pressure storage power station.

[0068] In summary, the implementation of the present invention has the following beneficial effects:

[0069] First, the present invention constructs a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station. By combining the BIM model and the monitoring data, the construction status of the underground chamber can be more comprehensively understood, and the analysis efficiency and accuracy of the monitoring data can be improved;

[0070] Second, the present invention uses the input gate, forget gate and output gate of the long short-term memory neural network to extract the characteristic representation of the monitoring sequence data of the surrounding rock state in the time dimension, which can effectively capture the time dependence of the surrounding rock state change and identify the characteristics of the time dimension in the surrounding rock state change; based on the graph attention network, the characteristic distribution of the monitoring data in the spatial dimension is extracted, so that in the monitoring of the surrounding rock state of the underground chamber, different weights are assigned to the neighbors of each node through the attention mechanism, and the spatial correlation and interaction between the surrounding rock monitoring points are identified and captured, so as to more accurately reflect the state of the surrounding rock;

[0071] Fourth, the present invention fuses the characteristic distribution of surrounding rock monitoring data in time and space dimensions by adopting a cross-attention fusion network structure, which is beneficial to the prediction of surrounding rock status, can obtain more accurate surrounding rock status data, and solves the technical problem of poor accuracy of monitoring results of traditional monitoring methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the background technology, the drawings required for use in the embodiments of the present application or the background technology will be described below.

[0073] Figure 1 This is a flow chart of a BIM-based construction monitoring method for an underground chamber of a pressure storage power station according to the present invention;

[0074] Figure 2 It is a sub-flow chart of the construction monitoring method of the underground chamber of the pressure storage power station based on BIM of the present invention;

[0075] Figure 3 The present invention is a structural block diagram of a BIM-based underground chamber construction monitoring system for a pressure storage power station. DETAILED DESCRIPTION

[0076] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0077] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0078] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0079] At present, in the compressed air energy storage system, the underground chamber is the main place for storing compressed air, and its stability directly affects the operating efficiency and safety of the entire system. During the construction of the underground chamber, the monitoring results of the surrounding rock area of ​​the underground chamber are also an important factor in judging the structural stability and safety of the chamber during the construction process; the present invention aims to monitor the state of the surrounding rock area of ​​the underground chamber to realize the construction monitoring process of the underground chamber of the compressed air energy storage power station.

[0080] The monitoring methods of related technologies are unable to effectively extract features from the measured data of surrounding rock conditions collected by sensors, resulting in the inability to provide accurate early warnings. This makes the monitoring results of surrounding rock conditions less accurate and unable to truly reflect the current state of surrounding rock in the underground chamber of the pressure storage power station.

[0081] Therefore, in order to achieve effective monitoring of the state of the surrounding rock area of ​​the underground chamber, the present invention fuses the characteristic distribution of the surrounding rock monitoring data in the time dimension and the space dimension, which is beneficial to the prediction of the surrounding rock state and can obtain more accurate surrounding rock state data.

[0082] The following is a detailed description of the specific implementation of the BIM-based pressure storage power station underground chamber construction monitoring method and system of the present invention in conjunction with specific embodiments.

[0083] In a disclosed embodiment, if Figure 1 As shown, a BIM-based underground chamber construction monitoring method for a pressure storage power station is provided, comprising the following steps:

[0084] S101: Build a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station;

[0085] Specifically, the present invention arranges several monitoring sections in the gas storage chamber to monitor the stress and strain of the support anchor rods and the concrete support structure during the construction stage; arranges multi-point displacement meters and earth pressure boxes on the support anchor rods near the monitoring sections to monitor the stress changes of the rock mass, so as to obtain the monitoring data of the surrounding rock of the underground chamber of the pressure storage power station based on the monitoring data of the multi-point displacement meters and the earth pressure boxes;

[0086] Therefore, the present invention arranges a measuring section on the gas storage chamber during the construction of the chamber, and sets an earth pressure box and a multi-point displacement meter on the anchor rod near the measuring section to monitor the rock stress changes of the section during the rock construction, so as to judge the rock deformation of the section through the data monitored by the earth pressure box and the multi-point displacement meter.

[0087] Further, such as Figure 2 As shown, in a preferred embodiment of the present disclosure, the step of constructing a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station includes the following steps:

[0088] S201: constructing a BIM basic model of an underground chamber of a pressure storage power station including structural components and sensor components, wherein the BIM basic model is the same as the underground chamber structure at the construction site;

[0089] S202: Acquire construction data of sensor components set in the construction site, establish a mapping relationship between the construction data and the sensor components in the BIM basic model, and add the construction data to the BIM basic model based on the mapping relationship to obtain a construction model.

[0090] Therefore, the present invention constructs a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station. By combining the BIM model and the monitoring data, the construction status of the underground chamber can be more comprehensively understood, and the analysis efficiency and accuracy of the monitoring data can be improved.

[0091] Please continue to refer to Figure 1 The BIM-based construction monitoring method for underground chambers of a pressure storage power station of the embodiment of the present disclosure further includes:

[0092] S102: Acquire monitoring data for characterizing the state of the surrounding rock of the underground chamber in the construction model, wherein the monitoring data includes deformation data of the underground chamber vault settlement and horizontal convergence and the time point at which the deformation data is acquired;

[0093] Among them, in the disclosed embodiment, the monitoring items of the underground chamber of the pressure storage power station mainly include surrounding rock monitoring, among which the deformation measurement points of the surrounding rock include the arch sinking monitoring point arranged at the center of the arch, and the horizontal convergence monitoring point includes horizontal and oblique convergence.

[0094] S103: extracting characteristic distribution of monitoring data in the time dimension using a long short-term memory neural network;

[0095] Among them, in the implementation method of extracting the characteristic distribution of monitoring data in the time dimension by using the long short-term memory neural network, the monitoring data of the surrounding rock state is first normalized and sequence segmented to obtain the monitoring sequence data of the surrounding rock state; optionally, a linear interpolation method is also used to fill in missing data and replace variant data;

[0096] Preferably, in the embodiment of the present disclosure, the formula for normalizing the monitoring data of the surrounding rock state is expressed as:

[0097]

[0098] In the formula, x ′ represents the normalized data; x represents the original data; x max Indicates the maximum monitoring value within the target time period; x min Indicates the minimum monitoring value within the target time period.

[0099] Furthermore, the step of extracting the characteristic distribution of monitoring data in the time dimension by using the long short-term memory neural network also includes the following steps:

[0100] The input gate, forget gate and output gate of the long short-term memory neural network are used to extract the characteristic representation of the monitoring sequence data of the surrounding rock state in the time dimension; among which:

[0101] The input gate is represented as: i t =σ(W i ·[h t-1 , x t ]+b i );

[0102] The forget gate is represented by: f t =σ(W f ·[h t-1 , x t ]+b f );

[0103] The output gate is represented as: o t =σ(W o ·[h t-1 , x t ]+b o );

[0104] In the formula, σ represents the activation function; i t represents the input gate; o t represents the output gate;

[0105] f trepresents the forget gate; h t-1 Indicates the hidden layer state at the previous moment; x t represents the monitoring sequence data at time t; W i 、b i Respectively represent the weight matrix and bias of the input gate; W f 、b f Respectively represent the weight matrix and bias of the forget gate; W o 、b o Represent the weight matrix and bias of the output gate respectively;

[0106] The characteristic of the monitoring sequence data of the surrounding rock state in the time dimension is expressed as F t ,in:

[0107]

[0108] In the formula, C t-1 and Represent the cell state at the previous moment and the candidate cell state respectively

[0109] S104: Extracting the feature distribution of monitoring data in the spatial dimension based on the graph attention network;

[0110] The feature distribution of the monitoring data in the spatial dimension is extracted based on the graph attention network, including:

[0111] Construct a spatiotemporal graph of surrounding rock monitoring data, taking each monitoring point in the surrounding rock monitoring area as a node of the spatiotemporal graph, and the characteristic matrix of the node is the data collected by the corresponding monitoring point;

[0112] The self-attention mechanism is used to focus on the information of each node in the spatiotemporal graph, and the characteristic distribution of monitoring data in the spatial dimension is extracted based on the attention correlation between nodes.

[0113] Furthermore, the step of using each monitoring point in the surrounding rock monitoring area as a node of the space-time graph includes:

[0114] Determine the edge relationship between two monitoring points, expressed as:

[0115]

[0116] Where: A ij represents the neighboring relationship between monitoring point j and monitoring point j, 1 means that the two monitoring points are connected, and 0 means that they are not connected; d(i,j) represents the Euclidean distance between the two monitoring points; D represents the threshold used to determine whether there is an edge relationship between the monitoring points;

[0117] Furthermore, based on the neighbor relationship and time sequence, a spatiotemporal graph sequence is constructed:

[0118]

[0119] Where Z represents the number of monitoring points in the target surrounding rock area; n represents the length of the time series; V represents the number of features at each monitoring point; represents the characteristic matrix of the i-th monitoring point at time t; S i Represents the characteristic matrix of all monitoring points in the target surrounding rock area at time t.

[0120] Furthermore, the step of extracting the feature distribution of the monitoring data in the spatial dimension based on the attention correlation between the nodes includes:

[0121] Calculate the attention correlation between monitoring point i and monitoring point j. The calculation formula is expressed as:

[0122] e ij =a tt (Ws i ,Ws j );

[0123] In the formula, e ij Indicates the importance of monitoring point i to monitoring point j; a tt (·) represents a single-layer feedforward neural network; s i represents the feature vector of monitoring point i; s j represents the feature vector of monitoring point j; W represents the model training parameters;

[0124] The attention coefficient a between monitoring point i and monitoring point j ij The calculation is done using the following formula:

[0125]

[0126] In the formula, Q i represents the set of all monitoring points that have an edge relationship with monitoring point i, and LeakyRelu represents a nonlinear activation function;

[0127] Furthermore, in the embodiment of the present disclosure, after obtaining the attention coefficient, the linear combination of the feature vectors of the monitoring point i and its connected monitoring points is calculated to obtain the aggregate representation of each monitoring point i: Then the feature distribution at time t in the spatial dimension is expressed as:

[0128]

[0129] Where Z represents the number of monitoring points in the target surrounding rock area.

[0130] Therefore, the present invention utilizes the input gate, forget gate and output gate of the long short-term memory neural network to extract the characteristic representation of the monitoring sequence data of the surrounding rock state in the time dimension, which can effectively capture the time dependence of the surrounding rock state change and identify the characteristics of the time dimension in the surrounding rock state change;

[0131] In addition, the present invention extracts the characteristic distribution of monitoring data in the spatial dimension based on the graph attention network, so that in the monitoring of the surrounding rock status of underground chambers, different weights are assigned to the neighbors of each node through the attention mechanism, and the spatial correlation and interaction between the surrounding rock monitoring points are identified and captured, which can more accurately reflect the status of the surrounding rock of the underground chamber.

[0132] S105: Using a cross-attention fusion network structure to fuse the feature distribution of the monitoring data in the time dimension and the spatial dimension, and construct the fusion feature of the monitoring data;

[0133] Specifically, the step of using a cross-attention fusion network structure to fuse the feature distribution of monitoring data in the time dimension and the space dimension to construct the fusion feature of the monitoring data includes:

[0134] Given the time dimension feature set of monitoring data

[0135] And, given the spatial dimension feature set of monitoring data

[0136] Furthermore, the present invention uses a dimensionality conversion function to unify the time dimension feature set T and the space dimension feature set P across dimensions into feature representations of the same size; the correlation matrix of the time dimension feature set T and the space dimension feature set P is calculated respectively, and the correlation matrix is ​​used as the weight of the time dimension feature set T to fuse the time dimension feature set T and the space dimension feature set P to obtain the fused feature F TP The fused features contain both spatial dimension information and temporal dimension information, which effectively enhances the robustness of the feature expression of underground chamber surrounding rock monitoring data; among them:

[0137]

[0138] Where LN(·) represents a linear transformation function, which is used to transform the time dimension feature set T into a feature map with the same dimension as the spatial dimension feature set P; d represents the size of the feature dimension.

[0139] Therefore, the present invention adopts a cross-attention fusion network structure to fuse the characteristic distribution of surrounding rock monitoring data in the time dimension and the space dimension, which is beneficial to the prediction of the surrounding rock state, can obtain more accurate surrounding rock state data, and solves the technical problem of poor accuracy of monitoring results of traditional monitoring methods.

[0140] Finally, please continue to refer to Figure 1 A BIM-based construction monitoring method for an underground chamber of a pressure storage power station in an embodiment of the present disclosure further includes the following steps:

[0141] S106: using the fusion feature as an input of a pre-constructed prediction model, outputting a surrounding rock state prediction result, and using the surrounding rock state prediction result to update the surrounding rock state in the construction model of the underground chamber of the pressure storage power station;

[0142] Preferably, in the disclosed embodiment, the step of using the fused features as the input of the pre-constructed prediction model and outputting the surrounding rock state prediction result includes: allocating the fused features to the training set and the test set, the allocation ratio being 80% of the training set and 20% of the test set; wherein, the fused feature data is brought into the PSO algorithm, the fused feature data is initialized, and the individual fitness of the particle swarm is calculated; the optimization goal is to select individuals with high fitness; if the highest fitness is not reached, the iteration is continued until the fitness reaches the highest and no longer changes, and the optimization parameters are obtained; the optimization parameters belong to the LSTM prediction model for learning and memory, and the surrounding rock state prediction result is output when the set maximum number of iterations is reached.

[0143] Therefore, the surrounding rock state prediction results of the present invention can be used to update the construction model to reflect the actual state and potential risks of the surrounding rock area of ​​the underground chamber; it is convenient for construction personnel to timely adjust the support design, optimize the construction steps or take preventive measures to avoid potential risks.

[0144] Optionally, in a preferred embodiment of the present disclosure, the pre-constructed prediction model is an LSTM neural network prediction model built based on the TensorFlow.js framework, and in the hyperparameters of the prediction model, the number of input layer neurons is 6, the number of hidden layer neurons is 12, the number of output layer neurons, i.e., the number of fully connected layer input layer neurons is 10, the number of fully connected layer hidden layer neurons is 10, and the number of output layer neurons is 1;

[0145] Furthermore, the initial step size of the pre-constructed prediction model is set to 4, the number of iterations is 3000, and the learning rate is 0.005.

[0146] Further, in the embodiment of the present disclosure, the pre-constructed prediction model uses a particle swarm optimization algorithm to optimize parameters; in the related art, in the traditional particle swarm optimization algorithm (PSO), the inertia weight decreases linearly with the increase of the number of iterations, which will cause the algorithm to have a better global search capability in the early stage of iteration, and only have a better local search capability in the later stage, so there is a problem of premature convergence of the algorithm; based on this, the particle swarm optimization algorithm of the embodiment of the present disclosure adopts a concave function decreasing strategy in the initial stage to reduce the inertia weight coefficient nonlinearly to avoid premature convergence; wherein the concave function decreasing strategy is expressed as:

[0147]

[0148] In the formula, ω max and ω min are the maximum and minimum values ​​of the inertia weight respectively; t represents the current iteration number; t max represents the maximum number of iterations; k is a control parameter used to control the smoothness of the ω curve.

[0149] Furthermore, in order to obtain the diversity of particles in the initial search phase of the PSO, an intervention operation is performed on the particles in the PSO of the embodiment of the present disclosure, and the position of the particles is adjusted with a certain probability so that the particles enter other areas for search, which can effectively expand the search range of the particles and quickly optimize the parameters of the prediction model; wherein the intervention operation is represented as:

[0150] p(i,k)=5×rand,x>0.95;

[0151] Where p(i,k) represents the k-dimensional transformation operation of particle i in the particle group, rand represents the random number of the transformation in the range [0,1], and the particle transformation occurs when the random number x>0.95.

[0152] Furthermore, in the embodiment of the present disclosure, the following evaluation indicators are used to evaluate the prediction results of the prediction model, and the prediction model is further tuned according to the evaluation results to make the prediction results of the prediction model more accurate; specifically, the evaluation indicators include root mean square error RMSE, mean absolute error MAE, mean square error MAE, mean square error MSE, and mean absolute percentage error MAPE; wherein:

[0153] The mean absolute error MAE is expressed as:

[0154] The mean square error MAE is expressed as:

[0155] The root mean square error RMSE is expressed as:

[0156] The mean absolute percentage error MAPE is expressed as:

[0157] Where, T n Indicates the number of days for the prediction of surrounding rock status; s i Indicates the actual value of surrounding rock monitoring; Represents the predicted value of the surrounding rock state.

[0158] Please refer to Figure 3 In another embodiment of the present disclosure, a BIM-based underground chamber construction monitoring system for a pressure storage power station is provided by the present invention, comprising:

[0159] The model building unit 301 is used to build a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station;

[0160] The data acquisition unit 302 is used to acquire monitoring data used to characterize the surrounding rock state of the underground chamber in the construction model, wherein the monitoring data includes deformation data of the underground chamber vault settlement and horizontal convergence and the time point at which the deformation data is acquired;

[0161] A first feature extraction unit 303 is used to extract the feature distribution of the monitoring data in the time dimension by using a long short-term memory neural network;

[0162] A second feature extraction unit 304 is used to extract feature distribution of monitoring data in a spatial dimension based on a graph attention network;

[0163] The feature fusion unit 305 is used to fuse the feature distribution of the monitoring data in the time dimension and the space dimension by using a cross attention fusion network structure to construct a fusion feature of the monitoring data;

[0164] The prediction updating unit 306 is used to use the fusion feature as the input of the pre-constructed prediction model, output the surrounding rock state prediction result, and use the surrounding rock state prediction result to update the surrounding rock state in the construction model of the underground chamber of the pressure storage power station.

[0165] Correspondingly, the monitoring system provided by the embodiment of the present disclosure can also utilize the model building unit 301 in combination with the BIM model and the monitoring data to more comprehensively understand the construction status of the underground chamber and improve the analysis efficiency and accuracy of the monitoring data; and utilize the first feature extraction unit 302 to effectively capture the time dependence of the surrounding rock state change and identify the characteristics of the time dimension in the surrounding rock state change; utilize the second feature extraction unit 303 to identify and capture the spatial correlation and interaction between the surrounding rock monitoring points to more accurately reflect the state of the surrounding rock; further, the monitoring system also fuses the characteristic distribution of the surrounding rock monitoring data in the time dimension and the spatial dimension through the feature fusion unit 305, which is conducive to the prediction of the surrounding rock state, and further utilizes the prediction update unit 306 to obtain more accurate surrounding rock state data, solving the technical problem of poor accuracy of monitoring results of traditional monitoring methods.

[0166] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0167] Those skilled in the art can clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0168] Another preferred embodiment of the present invention provides a monitoring device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements a BIM-based pressure storage power station underground chamber construction monitoring method as described in any one of the above embodiments.

[0169] The monitoring device may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server; the identification device may include but is not limited to a processor and a memory.

[0170] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0171] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0172] Another preferred embodiment of the present invention provides a storage medium, which is a computer-readable storage medium. The computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned method embodiments can be implemented.

[0173] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0174] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A BIM-based underground chamber construction monitoring method for a pressure storage power station, characterized in that: include: Construct a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station; Acquire monitoring data used to characterize the state of the surrounding rock of the underground chamber in the construction model, wherein the monitoring data includes deformation data of the underground chamber vault settlement and horizontal convergence and the time point at which the deformation data is acquired; The long short-term memory neural network is used to extract the characteristic distribution of monitoring data in the time dimension; Extract the feature distribution of monitoring data in the spatial dimension based on the graph attention network; The cross-attention fusion network structure is used to fuse the characteristic distribution of monitoring data in time and space dimensions to construct the fusion features of monitoring data; The fusion features are used as input of a pre-constructed prediction model, and a surrounding rock state prediction result is output, and the surrounding rock state in the construction model of the underground chamber of the pressure storage power station is updated using the surrounding rock state prediction result.

2. According to a BIM-based pressure storage power station underground chamber construction monitoring method according to claim 1, it is characterized in that: The step of constructing a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station includes: Constructing a BIM basic model of an underground chamber of a pressure storage power station including structural components and sensor components, wherein the BIM basic model is identical to the underground chamber structure at the construction site; Acquire construction data of sensor components set in the construction site, establish a mapping relationship between the construction data and the sensor components in the BIM basic model, add the construction data to the BIM basic model based on the mapping relationship, and obtain a construction model.

3. According to a BIM-based pressure storage power station underground chamber construction monitoring method according to claim 2, it is characterized in that: The step of extracting the characteristic distribution of monitoring data in the time dimension by using the long short-term memory neural network includes: The monitoring data of the surrounding rock state are normalized and sequence segmented to obtain the monitoring sequence data of the surrounding rock state; The input gate, forget gate and output gate of the long short-term memory neural network are used to extract the characteristic representation of the monitoring sequence data of the surrounding rock state in the time dimension; among which: The input gate is represented as: i t =σ(W i ·[h t-1 , x t ]+b i ); The forget gate is represented by: f t =σ(W f ·[h t-1 , x t ]+b f ); The output gate is represented as: o t =σ(W o ·[h t-1 , x t ]+b o ); In the formula, σ represents the activation function; i t represents the input gate; o t represents the output gate; f t represents the forget gate; h t-1 Indicates the hidden layer state at the previous moment; x t represents the monitoring sequence data at time t; W i 、b i Respectively represent the weight matrix and bias of the input gate; W f 、b f Respectively represent the weight matrix and bias of the forget gate; W o 、b o Represent the weight matrix and bias of the output gate respectively; The characteristic of the monitoring sequence data of the surrounding rock state in the time dimension is expressed as F t ,in: In the formula, C t-1 and Represent the cell state at the previous moment and the candidate cell state respectively.

4. According to a BIM-based pressure storage power station underground chamber construction monitoring method according to claim 3, it is characterized in that: The feature distribution of the monitoring data in the spatial dimension is extracted based on the graph attention network, including: Construct a spatiotemporal graph of surrounding rock monitoring data, taking each monitoring point in the surrounding rock monitoring area as a node of the spatiotemporal graph, and the characteristic matrix of the node is the data collected by the corresponding monitoring point; The self-attention mechanism is used to focus on the information of each node in the spatiotemporal graph, and the characteristic distribution of monitoring data in the spatial dimension is extracted based on the attention correlation between nodes.

5. According to the BIM-based construction monitoring method for underground chambers of pressure storage power stations according to claim 4, it is characterized in that: The step of using each monitoring point in the surrounding rock monitoring area as a node of the space-time graph comprises: Determine the edge relationship between two monitoring points, expressed as: Where: A ij represents the neighboring relationship between monitoring point j and monitoring point j, 1 means that the two monitoring points are connected, and 0 means that they are not connected; d(i,j) represents the Euclidean distance between the two monitoring points; D represents the threshold used to determine whether there is an edge relationship between the monitoring points; According to the neighbor relationship and time sequence, a space-time graph sequence is constructed: Where Z represents the number of monitoring points in the target surrounding rock area; n represents the length of the time series; V represents the number of features at each monitoring point; represents the characteristic matrix of the i-th monitoring point at time t; S i Represents the characteristic matrix of all monitoring points in the target surrounding rock area at time t.

6. According to the BIM-based construction monitoring method for underground chambers of pressure storage power stations according to claim 5, it is characterized in that: The step of extracting the feature distribution of monitoring data in the spatial dimension based on the attention correlation between nodes includes: Calculate the attention correlation between monitoring point i and monitoring point j. The calculation formula is expressed as: e ij =a tt (Ws i ,Ws j ); In the formula, e ij Indicates the importance of monitoring point i to monitoring point j; a tt (·) represents a single-layer feedforward neural network; s i represents the feature vector of monitoring point i; s j represents the feature vector of monitoring point j; W represents the model training parameters; The attention coefficient a between monitoring point i and monitoring point j ij The calculation is done using the following formula: In the formula, Q i represents the set of all monitoring points that have an edge relationship with monitoring point i, and LeakyRelu represents a nonlinear activation function; After obtaining the attention coefficient, the linear combination of the feature vectors of monitoring point i and its connected monitoring points is calculated to obtain the aggregate representation of each monitoring point i Then the feature distribution at time t in the spatial dimension is expressed as: Where Z represents the number of monitoring points in the target surrounding rock area.

7. The BIM-based construction monitoring method for underground chambers of pressure storage power stations according to claim 6 is characterized in that: The step of adopting a cross-attention fusion network structure to fuse the feature distribution of the monitoring data in the time dimension and the space dimension to construct the fusion feature of the monitoring data includes: Given a time dimension feature set and spatial dimension feature set Use the dimension conversion function to unify the time dimension feature set T and the space dimension feature set P across dimensions into feature representations of the same size; Calculate the correlation matrix of the time dimension feature set T and the space dimension feature set P respectively, use the correlation matrix as the weight of the time dimension feature set T, fuse the time dimension feature set T and the space dimension feature set P, and get the fusion feature F TP ,in: Where LN(·) represents a linear transformation function, which is used to transform the time dimension feature set T into a feature map with the same dimension as the spatial dimension feature set P; d represents the size of the feature dimension.

8. The BIM-based construction monitoring method for underground chambers of pressure storage power stations according to claim 7 is characterized in that: The pre-constructed prediction model is an LSTM neural network prediction model built based on the TensorFlow.js framework. Among the hyperparameters of the prediction model, the number of neurons in the input layer is 6, the number of neurons in the hidden layer is 12, the number of neurons in the output layer is 10, the number of neurons in the fully connected layer is 10, and the number of neurons in the output layer is 1. The initial step size of the model is set to 4, the number of iterations is 3000, and the learning rate is 0.

005.

9. The method for monitoring construction of underground chambers of a pressure storage power station based on BIM according to claim 8 is characterized in that: The pre-constructed prediction model uses a particle swarm optimization algorithm to optimize parameters, wherein the particle swarm optimization algorithm uses a concave function decreasing strategy to reduce the inertia weight coefficient nonlinearly in the initial stage; wherein the concave function decreasing strategy is expressed as: In the formula, ω max and ω min are the maximum and minimum values ​​of the inertia weight respectively; t represents the current iteration number; t max represents the maximum number of iterations; k is a control parameter used to control the smoothness of the ω curve.

10. A system for implementing a BIM-based pressure storage power station underground chamber construction monitoring method as claimed in any one of claims 1 to 9, characterized in that: The system comprises: A model building unit, used to build a construction model based on the BIM basic model and monitoring data of the underground chamber of the pressure storage power station; A data acquisition unit, used to acquire monitoring data for characterizing the state of the surrounding rock of the underground chamber in the construction model, wherein the monitoring data includes deformation data of the underground chamber vault settlement and horizontal convergence and the time point at which the deformation data is acquired; A first feature extraction unit, used to extract feature distribution of monitoring data in a time dimension using a long short-term memory neural network; A second feature extraction unit is used to extract the feature distribution of monitoring data in the spatial dimension based on the graph attention network; A feature fusion unit is used to fuse the feature distribution of monitoring data in the time dimension and the space dimension by using a cross-attention fusion network structure to construct a fusion feature of the monitoring data; The prediction updating unit is used to use the fusion feature as the input of the pre-constructed prediction model, output the surrounding rock state prediction result, and use the surrounding rock state prediction result to update the surrounding rock state in the construction model of the underground chamber of the pressure storage power station.

Images