Construction site environment dynamic regulation and control method and system fused with AIoT

Abstract

The invention relates to a construction site environment dynamic regulation and control method and system fused with AIoT. According to the method, a multi-source environment sensor network is deployed to collect original data of a construction site environment in real time, a data set with aligned timestamps is generated, an abnormal event is detected by using an isolated forest algorithm after standardized denoising processing, and a specific parameter type and a time window are marked; a causal intensity matrix between parameters is constructed through Granger causal test on the basis of environment data intercepted in an abnormal time period, a directed weighted network adjacency matrix representing a pollution propagation path is generated accordingly, and an abnormal source is accurately positioned through node influence propagation calculation and timestamp verification; and finally, a regulation and control instruction sequence is dynamically generated according to the shortest influence path of the source node in the propagation network, so that autonomous analysis of implicit association among construction site environment parameters, accurate positioning of a pollution source and dynamic distribution and instant deviation correction of a multi-stage regulation and control strategy are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of construction project management, and in particular relates to a method and system for dynamically controlling a construction site environment by integrating AIoT. Background Art

[0002] Current environmental monitoring systems at construction sites generally rely on discretely deployed sensor networks to collect data on parameters such as temperature, dust, and noise. While these systems provide basic monitoring, they suffer from three major bottlenecks: First, sensors operate independently, providing only single-point measurements of local parameters and failing to identify implicit correlations between multiple parameters. When an abnormal condition occurs (such as a sudden increase in dust concentration), the system only issues an alarm but cannot distinguish between increased pollution at the source and the spread of meteorological factors, making it difficult to attribute the abnormality. Second, existing control strategies rely on fixed threshold rules and the experience of environmental management personnel. Once a parameter exceeds the limit, preset equipment control commands are activated. This rigid response mechanism neither considers the causal links between parameters (such as whether changes in wind speed will lead to increased dust dispersion) nor dynamically assesses the priority of intervention for different equipment actions. Third, verification of control effectiveness is severely delayed. After executing dust suppression or noise reduction operations, manual re-inspection or waiting for the next round of testing to confirm the results is required, making it difficult to correct control deviations in a timely manner.

[0003] Therefore, there is an urgent need to build a technical solution that can autonomously analyze the dynamic causal network between environmental parameters, accurately locate the source of anomalies, and intelligently generate multi-level control paths, so as to achieve a transition from passive alarm to active predictive intervention. Summary of the Invention

[0004] Based on this, it is necessary to provide a method and system for dynamic control of construction site environment integrating AIoT to address the above technical problems.

[0005] In the first aspect, the present application provides a method for dynamically controlling a construction site environment by integrating AIoT, including:

[0006] S1. Deploy multi-source sensors in the target area of ​​the construction site to form a distributed IoT monitoring network. Based on the distributed IoT monitoring network, collect the original data of the construction site environment and generate a raw environmental data set with aligned timestamps.

[0007] S2. Preprocessing the original environmental data set to generate a standardized environmental parameter data set; wherein the preprocessing includes wavelet threshold denoising, sliding window median filtering, and maximum and minimum value normalization;

[0008] S3. Based on the standardized environmental parameter data set and the preset dynamic threshold, the isolation forest algorithm is used to detect abnormal fluctuations and generate an abnormal event marker sequence; wherein the abnormal event marker sequence includes the abnormal parameter ID, abnormal value, and start and end timestamps;

[0009] S4. Based on the time window of the abnormal event marker sequence, the corresponding period data of the standardized environmental parameter data set is intercepted, and the dependency relationship between the parameters in the period data is calculated through the Granger causality test to generate the environmental parameter causal intensity matrix;

[0010] S5. Based on the environmental parameter causal intensity matrix, a directed weighted network is constructed with environmental parameters as nodes, causal directions as directed edges, and causal intensity as weights to obtain the environmental impact propagation path adjacency matrix;

[0011] S6. Determine the abnormal source node identifier based on the matching verification of the weighted out-degree of each node in the adjacency matrix of the environmental impact propagation path and the timestamp of the abnormal event marker sequence;

[0012] S7. Based on the abnormal source node identifier and the environmental impact propagation path adjacency matrix, the impact propagation priority is calculated using the shortest path algorithm, and an environmental control instruction sequence is generated based on the impact propagation priority.

[0013] Secondly, the present application also provides a construction site environment dynamic control system integrated with AIoT, including: a distributed Internet of Things monitoring network, a data preprocessing module, an abnormal fluctuation detection module, a causal relationship analysis module, a directed weighted network construction module, an abnormal source determination module, and a control instruction generation module; wherein the Internet of Things monitoring network is composed of multi-source sensors deployed in the target area of ​​the construction site;

[0014] A distributed IoT monitoring network is used to collect raw data about the construction site environment and generate a raw environmental dataset with aligned timestamps.

[0015] The data preprocessing module is used to preprocess the original environmental data set to generate a standardized environmental parameter data set; wherein the preprocessing includes wavelet threshold denoising, sliding window median filtering, and maximum and minimum value normalization;

[0016] The abnormal fluctuation detection module is used to detect abnormal fluctuations based on the standardized environmental parameter data set and preset dynamic thresholds using the isolation forest algorithm to generate an abnormal event marker sequence; the abnormal event marker sequence includes the abnormal parameter ID, abnormal value, and start and end timestamps;

[0017] The causal relationship analysis module is used to intercept the corresponding period data of the standardized environmental parameter data set based on the time window of the abnormal event marker sequence, calculate the dependency between the parameters in the period data through the Granger causality test, and generate the environmental parameter causal strength matrix;

[0018] A directed weighted network construction module is used to construct a directed weighted network based on the environmental parameter causal intensity matrix, with environmental parameters as nodes, causal directions as directed edges, and causal intensity as weights, to obtain the adjacency matrix of the environmental impact propagation path;

[0019] The abnormal source determination module is used to determine the abnormal source node identifier based on the matching verification of the weighted out-degree of each node in the adjacency matrix of the environmental impact propagation path and the timestamp of the abnormal event marker sequence;

[0020] The control instruction generation module is used to calculate the impact propagation priority through the shortest path algorithm based on the abnormal source node identifier and the environmental impact propagation path adjacency matrix, and generate an environmental control instruction sequence based on the impact propagation priority.

[0021] In a third aspect, the present application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements a method for dynamically controlling a construction site environment integrating AIoT as described in the first aspect.

[0022] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements a method for dynamically controlling a construction site environment integrating AIoT as described in the first aspect.

[0023] The above-mentioned method and system for dynamic control of construction site environment integrated with AIoT deploys a multi-source environmental sensor network to collect the original data of the construction site environment in real time and generate a timestamp-aligned data set. After standardized denoising processing, the isolation forest algorithm is used to detect abnormal events and mark the specific parameter types and time windows. The Granger causality test is then used to construct a causal strength matrix between parameters based on the environmental data intercepted during the abnormal period. Based on this, a directed weighted network adjacency matrix representing the pollution propagation path is generated. The source of the abnormality is then accurately located through node influence propagation calculation and timestamp verification. Finally, a control instruction sequence is dynamically generated based on the shortest influence path of the source node in the propagation network, thereby realizing autonomous analysis of the implicit correlation between construction site environmental parameters, accurate positioning of the pollution source, and dynamic allocation and immediate correction of multi-level control strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1A schematic diagram of a process flow of a method for dynamically controlling a construction site environment by integrating AIoT provided by the present invention;

[0026] Figure 2 A schematic diagram of a flow chart for determining whether Granger causality exists between parameters in an optional embodiment of the present invention;

[0027] Figure 3 This is a structural schematic diagram of a construction site environment dynamic control system integrated with AIoT provided by the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0029] refer to Figure 1 , which presents a flow chart of a method for dynamically controlling a construction site environment by integrating AIoT provided by the present application, the method comprising the following steps:

[0030] S1. Deploy multi-source sensors in the target area of ​​the construction site to form a distributed IoT monitoring network. Based on the distributed IoT monitoring network, collect the original data of the construction site environment and generate a timestamp-aligned original environment data set.

[0031] Specifically, within the target area of ​​the construction site, multi-source sensors are rationally deployed based on the scale of the construction project and the complexity of the environment to build a distributed IoT monitoring network. The target area of ​​the construction site can include the work area, material storage area, and boundary fences. Multi-source sensors include temperature, humidity, dust concentration, noise intensity, wind speed and direction, and air pressure sensors. These sensors can be installed in key locations such as tower cranes, construction elevators, fences, scaffolding, and material storage areas to provide comprehensive and multi-dimensional perception of the construction site environment. These sensors can be connected to data collection nodes using wireless communication technologies (such as ZigBee, LoRa, or NB-IoT) to ensure real-time transmission of environmental data. The data collection nodes collect raw data from each sensor at set intervals (e.g., every 1 to 5 minutes), add timestamps to the raw data, and process it using a time alignment algorithm to generate a timestamp-aligned raw environmental data set.

[0032] S2. Preprocess the original environmental data set to generate a standardized environmental parameter data set; wherein the preprocessing includes wavelet threshold denoising, sliding window median filtering, and maximum and minimum value normalization.

[0033] Specifically, wavelet threshold denoising is first applied to the collected raw environmental data set. Based on the wavelet threshold denoising method, an appropriate wavelet basis (e.g., db4) and decomposition scale (e.g., 3 layers) are selected, and either a soft or hard threshold function is determined. This decomposition and reconstruction of the noisy environmental parameter data is performed, filtering out high-frequency noise interference while preserving the key features and trends of the original data. Next, a sliding window median filter is applied. A reasonable window length (e.g., 5 or 7 data points) is set, and the window is slid along the time series. The median value of the windowed data is used to replace the central point data value, further smoothing the data curve and eliminating occasional spikes and pulse interference. Finally, maximum and minimum value normalization is performed to linearly map each environmental parameter data to the [0, 1] interval, eliminating the influence of different dimensions and dimensional ranges, and ensuring a uniform numerical scale for the data, thus generating a standardized environmental parameter data set.

[0034] S3. Based on the standardized environmental parameter data set and the preset dynamic threshold, abnormal fluctuation detection is performed using the isolation forest algorithm to generate an abnormal event marker sequence; wherein the abnormal event marker sequence includes the abnormal parameter ID, abnormal value, and start and end timestamps.

[0035] Specifically, based on the standardized environmental parameter data set and combined with preset dynamic thresholds, the isolation forest algorithm is used to detect abnormal fluctuations. The dynamic threshold is dynamically determined based on the statistical characteristics of historical environmental data. For example, for dust concentration, the upper and lower thresholds of its normal fluctuation range are set as:

[0036] T lower =μ-kσ;

[0037] T upper =μ+kσ;

[0038] Among them, μ is the mean of historical dust concentration data, σ is the standard deviation, and k is the empirical coefficient (which can be 2 to 3). In the isolation forest algorithm, multiple isolation forest models are constructed, each of which consists of several isolated trees. For each isolated tree, a sample subset and a feature subset are randomly selected, and the tree structure is constructed by recursively dividing the hyperplane. The division principle is to separate the samples in the node as much as possible. For a given environmental parameter data point, its anomaly score in all isolation forest models is calculated. Based on the comparison result of the score with the preset dynamic threshold, it is determined whether the data point is an anomaly point. If it is abnormal, the abnormal parameter ID (such as temperature, noise and other identifiers), the abnormal value and the corresponding start and end timestamps are recorded, and finally an abnormal event marker sequence containing complete abnormal information is generated.

[0039] S4. Based on the time window of the abnormal event marker sequence positioning, the corresponding period data of the standardized environmental parameter data set is intercepted, and the dependency relationship between the parameters in the period data is calculated through the Granger causality test to generate the environmental parameter causal intensity matrix.

[0040] Specifically, based on the time window of the abnormal event marker sequence, the data of the corresponding period is accurately intercepted from the standardized environmental parameter data set. Using the Granger causality test method, the following autoregressive model can be constructed for each two environmental parameter time series:

[0041]

[0042] in, and They represent the values ​​of two environmental parameters at time t, p is the autoregressive order, and q is the lag order of Granger causality test. j Models with their own lags and models containing X i The F statistic of the model with the lag term is used to determine whether X i Is it X j Granger cause. Based on the test results, the causal strength value between each parameter is calculated, which reflects the closeness of the causal relationship between the parameters, and finally generates the environmental parameter causal strength matrix. The matrix element C ij It represents the causal strength of environmental parameter i on environmental parameter j, providing a quantitative basis for the subsequent construction of the environmental impact transmission path network.

[0043] S5. Based on the environmental parameter causal intensity matrix, a directed weighted network is constructed with environmental parameters as nodes, causal directions as directed edges, and causal intensity as weights to obtain the environmental impact propagation path adjacency matrix.

[0044] Specifically, the directed edges between nodes represent the causal direction, that is, from the cause parameter to the result parameter; the weight of the edge represents the causal strength value. ij >0, then draw a line with weight C from node i to node j ij By traversing all parameter nodes and their causal relationships, a complete directed weighted network structure is generated, and then the adjacency matrix of the environmental impact propagation path is obtained. The element A of the adjacency matrix ij Indicates whether there is a directed edge from node i to node j and the corresponding weight value. If there is a directed edge, then A ij =C ij , otherwise A ij = 0. This adjacency matrix comprehensively and intuitively presents the influence propagation path and intensity relationship between various parameters in the construction site environment, providing a key network structure foundation for the subsequent determination of abnormal source nodes and influence propagation priorities.

[0045] S6. Based on the matching verification of the weighted out-degree of each node in the adjacency matrix of the environmental impact propagation path and the timestamp of the abnormal event marker sequence, the abnormal source node identifier is determined.

[0046] Specifically, based on the adjacency matrix of the environmental impact propagation path, the weighted out-degree of each node is first calculated. The weighted out-degree of node i is defined as the sum of the weights of all directed edges starting from the node, that is:

[0047]

[0048] Where n is the total number of environmental parameter nodes. Simultaneously, the weighted out-degree of each node is verified against the anomaly time series by combining the timestamp information in the anomaly event marker sequence. Through methods such as cross-correlation analysis of time series, we determine which nodes' weighted out-degree changes have a high temporal correlation with the occurrence of the anomaly within the time window of the anomaly. By comprehensively considering the weighted out-degree and the degree of temporal matching, we can apply methods such as cluster analysis or threshold judgment to ultimately identify the source node that caused the environmental anomaly fluctuation and accurately locate the origin of the anomaly.

[0049] S7. Based on the abnormal source node identifier and the environmental impact propagation path adjacency matrix, the impact propagation priority is calculated using the shortest path algorithm, and an environmental control instruction sequence is generated based on the impact propagation priority.

[0050] Specifically, after identifying the anomaly source node, the impact propagation path from the anomaly source node to all other nodes is calculated using a shortest path algorithm (such as the Dijkstra algorithm or the Floyd-Warshall algorithm) in conjunction with the environmental impact propagation path adjacency matrix. The priority of the impact propagation is determined based on the sum of the path weights (i.e., the sum of the weights of each edge along the path). A smaller sum indicates faster and more direct impact propagation, and therefore a higher priority. Based on the ranking of impact propagation priorities, a corresponding environmental control instruction sequence is generated for environmental parameter nodes of different priorities, combined with a preset control strategy library. This control instruction sequence specifies the start-up time, operating parameters (such as spray intensity and ventilation volume), and control duration for each environmental control device (such as spray dust suppression equipment, ventilation equipment, and noise barriers). This ensures that control instructions are accurately and efficiently implemented according to the order of environmental impact propagation, enabling dynamic and intelligent control of the construction site environment, effectively improving the site's environmental quality and reducing the adverse impact of environmental fluctuations on the construction process and the surrounding environment.

[0051] The above-mentioned method for dynamic control of construction site environment integrated with AIoT deploys a multi-source environmental sensor network to collect the original data of the construction site environment in real time and generate a timestamp-aligned data set. After standardized denoising processing, the isolation forest algorithm is used to detect abnormal events and mark the specific parameter types and time windows. The Granger causality test is then used to construct a causal strength matrix between parameters based on the environmental data intercepted during the abnormal period. Based on this, a directed weighted network adjacency matrix representing the pollution propagation path is generated. The source of the abnormality is then accurately located through node influence propagation calculation and timestamp verification. Finally, a control instruction sequence is dynamically generated based on the shortest influence path of the source node in the propagation network, thereby realizing autonomous analysis of the implicit correlation between construction site environmental parameters, accurate positioning of the pollution source, and dynamic allocation and immediate correction of multi-level control strategies.

[0052] In an optional embodiment, S3 includes the following steps:

[0053] S31. Based on the standardized environmental parameter data set, construct an isolation forest decision tree set containing multiple parameters; wherein each decision tree in the isolation forest decision tree set is generated by randomly selecting features and split points.

[0054] Specifically, the process of constructing an isolation forest decision tree set can be as follows:

[0055] 1) Data preprocessing: interpolate or delete missing data to ensure that the input data set is complete and has no missing values.

[0056] 2) Parameter initialization: Set the parameters of the isolation forest model, such as the number of trees (can be 100 to 200 trees) and the sample subset size of each decision tree (can be set to the square root of the dataset size).

[0057] 3) Decision tree generation: For each decision tree, a sample subset is randomly extracted from the standardized environmental parameter data set. Starting from the root node, a feature (environmental parameter) and a split point on the feature are randomly selected. The split point is selected between the minimum and maximum values ​​of the feature. According to the split point, the data of the current node is divided into two parts. One part with a feature value less than the split point enters the left subtree, and the other part enters the right subtree. Repeat the above process until the number of data points in all nodes is less than the preset stopping threshold (such as 1 or 2), at which time the node is set as a leaf node. Through recursive splitting, a complete isolation forest decision tree is finally constructed.

[0058] S32. Count the average path length from the root node to the data point corresponding to each parameter in the isolation forest decision tree set.

[0059] Specifically, for each data point, starting from the root node, we follow the decision tree's split path down layer by layer until we reach a leaf node, and record the path length of that data point in each decision tree. Path length is defined as the number of edges from the root node to the leaf node.

[0060] Sum the path lengths of the data point in all decision trees and then divide it by the total number of decision trees to get the average path length of the data point. The mathematical expression is:

[0061]

[0062] Among them, n represents the total number of decision trees, Path Length i Represents the path length of the data point in the i-th decision tree.

[0063] S33. Calculate an anomaly score based on the average path length. When the anomaly score exceeds a preset anomaly threshold, mark the corresponding data point as an anomaly point, and generate an anomaly event marker sequence containing an anomaly parameter ID, an anomaly value, and a start and end timestamp corresponding to the anomaly point.

[0064] Specifically, the anomaly score is calculated based on the average path length obtained in step S32. The calculation formula is:

[0065]

[0066] Here, c(n) is a constant related to the size of the sample subset and is used to adjust the distribution of the anomaly score. The calculated anomaly score is compared with the preset anomaly threshold. The preset anomaly threshold can be determined based on historical data and anomalies in actual application scenarios. If the anomaly score exceeds the preset anomaly threshold, the corresponding data point is marked as an anomaly. For each point marked as an anomaly, an anomaly event marker sequence is generated, which includes the anomaly parameter ID (identifying the specific environmental parameter anomaly), the anomaly value (the specific value of the anomaly point), and the start and end timestamps (recording the time period when the anomaly occurred).

[0067] refer to Figure 2 In an optional embodiment, S4 includes the following steps:

[0068] S41. Based on the start and end timestamps in the abnormal event mark sequence, extract the time series of each parameter in the corresponding time period from the standardized environmental parameter data set.

[0069] Specifically, this process ensures that the selected data is highly temporally correlated with the abnormal event, providing a temporal focus for subsequent causal analysis. Time series data can clearly demonstrate the dynamic trends of various environmental parameters during the abnormal event, laying the foundation for subsequent Granger causality testing.

[0070] S42. Based on the time series, perform a Granger causality test on each pair of parameters (A, B) to determine whether there is a Granger causal effect from parameter A to parameter B, including:

[0071] S421. Construct a first autoregressive model based on the time series of parameter B, and calculate the residual sum of squares of the first autoregressive model; the expression of the first autoregressive model is:

[0072]

[0073] Among them, B t is the observed value of parameter B at time t, B t-i Represents the observed value of parameter B at time ti, i is the lag order of parameter B, p is the number of lag terms of parameter B, α i is the autoregressive coefficient of the first autoregressive model, ∈ t is the residual term of the first autoregressive model.

[0074] Specifically, this model captures the time series characteristics of parameter B and the impact of its historical values ​​on its current value. The residual sum of squares (RSS1) calculated based on this model reflects the goodness of fit of the model to the data when only the historical information of parameter B is considered.

[0075] S422: Construct a second autoregressive model based on the time series of parameter A and parameter B, and calculate the residual sum of squares of the second autoregressive model; the expression of the second autoregressive model is:

[0076]

[0077] Among them, A t-j represents the observed value of parameter A at time tj, j is the lag order, q is the number of lag terms of parameter A, β i is the autoregressive coefficient of the second autoregressive model, γ j It is the coefficient that measures the effect of the lag term of parameter A on the current value of parameter B, μ t is the residual term of the second autoregressive model.

[0078] Specifically, this model investigates whether the inclusion of historical information about parameter A leads to a more accurate prediction of the current value of parameter B. Similarly, the model calculates the residual sum of squares (RSS2). RSS2 reflects the forecast error after combining historical information about parameters A and B. By comparing RSS1 and RSS2, we can preliminarily determine whether the inclusion of parameter A significantly improves the prediction of parameter B.

[0079] S423. Calculate the F statistic of the significance of the causal effect of parameter A on parameter B based on the residual sum of squares of the first autoregressive model and the residual sum of squares of the second autoregressive model. The calculation formula of the F statistic is:

[0080]

[0081] Where RSS1 is the residual sum of squares of the first autoregressive model, RSS2 is the residual sum of squares of the second autoregressive model, and T is the sample size.

[0082] Specifically, the F statistic is used to measure whether the model's goodness of fit improves significantly after introducing historical information about parameter A. The larger the F value, the more significant the causal effect of parameter A on parameter B.

[0083] S424. Based on the F statistic and the degrees of freedom (q, Tpq), the corresponding significance probability is obtained by querying the F distribution table. When the F statistic exceeds the preset critical value and the significance probability is less than the preset probability value, it is determined that there is a Granger causal effect from parameter A to parameter B.

[0084] Specifically, the judgment standard is based on the hypothesis testing principle in statistics, and determines whether a causal relationship exists by comparing whether the improvement in goodness of fit before and after the introduction of parameter A into the model is statistically significant.

[0085] S43. When Granger causality exists, based on the coefficient γ in the second autoregressive model j The causal strength value is calculated by the standardized absolute value of , and all the causal strength values ​​are filled into an N×N matrix to generate the environmental parameter causal strength matrix; where N is the number of parameter types of the original data of the construction site environment.

[0086] Specifically, the coefficient γ corresponding to the lag term of parameter A is extracted from the second autoregressive model j These coefficients reflect the degree to which the change of parameter A at different lag orders affects the current value of parameter B.

[0087] The standardized absolute values ​​of these coefficients γj are calculated. The purpose of standardization is to eliminate the influence of different dimensions and make the causal strength between different parameters comparable. The calculation formula is as follows:

[0088]

[0089] Among them, |γ j ∣ is the coefficient γ j The absolute values ​​of min|γ| and max|γ| are respectively the absolute values ​​of all coefficients γ j The minimum and maximum values ​​in absolute value.

[0090] The causal strength value can be calculated in two ways:

[0091] 1) Sum method: The coefficient γ corresponding to the standardized absolute value of all lag terms of parameter A j Add them together to get the comprehensive causal strength value of parameter A on parameter B. This method can emphasize the cumulative impact of parameter A on parameter B at different lag orders.

[0092] 2) Maximum method: Take the coefficient γ after the standardized absolute value of all lag terms of parameter A j The maximum value among them is taken as the causal strength value. This method can highlight the effect of the lag order where parameter A has the strongest influence on parameter B.

[0093] Based on the corresponding numbers of parameter A and parameter B, find the corresponding position in the N×N causal strength matrix and fill it with the calculated causal strength value. Each element in the matrix represents the causal strength between the corresponding parameter pair, thus forming a complete environmental parameter causal strength matrix.

[0094] In an optional embodiment, S5 includes the following steps:

[0095] S51. Based on the positions of elements greater than 0 in the environmental parameter causal strength matrix, determine the existence of directed edges between nodes to obtain a node set and a directed edge set.

[0096] Specifically, for the element C in the causal strength matrix C ij , if C ij If >0, it means that there is a directed edge from node i (corresponding to parameter i) to node j (corresponding to parameter j). By traversing all elements in the matrix, all node pairs with directed edges are determined, thus obtaining the node set V and the directed edge set E.

[0097] S52. Map the element values ​​in the environmental parameter causal intensity matrix to the weight values ​​of the corresponding directed edges to obtain a weight set.

[0098] Specifically, for a node pair (i, j) with a directed edge, the corresponding weight value w ij That is the matrix element C ij The weight value reflects the strength of the influence of node i on node j. The larger the weight value, the stronger the influence.

[0099] S53. Integrate the node set, the directed edge set, and the weight set, construct an adjacency matrix data structure, and obtain the environmental impact propagation path adjacency matrix.

[0100] Specifically, the node set V, the directed edge set E and the weight set W are integrated to construct an adjacency matrix data structure. The adjacency matrix is ​​an N×N matrix (N is the total number of nodes), where the element A ijRepresents the weight value from node i to node j. If there is no directed edge from node i to node j, then A ij =0; otherwise, A ij =w ij .

[0101] In an optional embodiment, S6 includes the following steps:

[0102] S61. Sum the adjacency matrix of the environmental impact propagation path row by row to obtain the weighted out-degree value of each node.

[0103] S62: Filter nodes whose weighted out-degree values ​​exceed a preset source influence threshold to obtain a set of candidate source nodes.

[0104] S63: Compare the abnormal start timestamp of each node in the candidate source node set with the timestamps of adjacent nodes in the abnormal event mark sequence. When the node start time is earlier than all adjacent nodes, generate an abnormal source node identifier.

[0105] Specifically, the preset source impact threshold can be determined based on historical data or actual application scenarios. The setting of this threshold should take into account the weighted out-degree value distribution of each node under normal conditions. The value of this threshold can be a part greater than the upper limit of the normal value, for example: T source =μ OD +kσ OD ; where μ OD is the mean of the weighted out-degree values, σ OD is the standard deviation of the weighted out-degree values, and k is an empirical coefficient (which can be 2 to 3).

[0106] In an optional embodiment, S7 includes the following steps:

[0107] S71. Using the abnormal source node as the path starting point, in the network defined by the environmental impact propagation path adjacency matrix, calculate the shortest path length from the starting point to other nodes and the path cumulative weight using the Dijkstra algorithm.

[0108] Specifically, the shortest path length of the starting node is set to 0, and the shortest path lengths of all other nodes are initialized to infinity. Create an unvisited node set that initially contains all nodes.

[0109] Select the node with the shortest path length from the set of unvisited nodes as the current node. For each neighbor node of the current node, calculate the path length from the starting node to the neighbor node. The path length calculation formula is: new path length = shortest path length of the current node + weight value from the current node to the neighbor node. If the new path length is less than the current shortest path length of the neighbor node, update the shortest path length of the neighbor node to the new path length, and record the cumulative weight of the path (that is, the sum of the weight values ​​of each edge on the path). Mark the current node as visited and remove it from the set of unvisited nodes. Repeat the above process until all nodes have been visited.

[0110] The shortest path length and cumulative path weight of each node are stored in the corresponding data structure for subsequent processing.

[0111] S72. Sort in ascending order by the shortest path length. When the shortest paths have the same length, sort in descending order by the cumulative weight of the paths to generate an environment parameter impact priority queue.

[0112] Specifically, all nodes are sorted from smallest to largest according to the shortest path length. If two nodes have the same shortest path length, they are sorted from largest to smallest according to the cumulative weight of the path.

[0113] Based on the sorting results, a priority queue for environmental parameter impact is generated. The order of nodes in the queue reflects the urgency and intensity of the impact of the anomaly. Nodes with higher priorities require more prioritization.

[0114] S73. Traverse the environmental parameter impact priority queue, and when the cumulative weight of the path of the target parameter exceeds the emergency processing threshold corresponding to the type of the target parameter, generate an environmental control instruction sequence.

[0115] Specifically, corresponding emergency processing thresholds are set according to the type of target parameter (such as temperature, dust, noise, etc.). These thresholds can be determined based on historical data, industry standards or expert experience.

[0116] Traverse each node in the environmental parameter impact priority queue in turn and check whether its path cumulative weight exceeds the corresponding emergency processing threshold.

[0117] When the cumulative weight of a node's path exceeds the emergency processing threshold, corresponding environmental control instructions are generated based on the type of target parameter and the abnormality. These instructions include the operating parameters of the control equipment (such as spray intensity for dust suppression equipment and ventilation volume for ventilation equipment), as well as the start time and operating duration. These instructions are organized into an environmental control instruction sequence based on priority, guiding the operation of on-site environmental control equipment and achieving effective control of the site environment.

[0118] The above-mentioned method for dynamic control of construction site environment integrated with AIoT deploys a multi-source environmental sensor network to collect the original data of the construction site environment in real time and generate a timestamp-aligned data set. After standardized denoising processing, the isolation forest algorithm is used to detect abnormal events and mark the specific parameter types and time windows. The Granger causality test is then used to construct a causal strength matrix between parameters based on the environmental data intercepted during the abnormal period. Based on this, a directed weighted network adjacency matrix representing the pollution propagation path is generated. The source of the abnormality is then accurately located through node influence propagation calculation and timestamp verification. Finally, a control instruction sequence is dynamically generated based on the shortest influence path of the source node in the propagation network, thereby realizing autonomous analysis of the implicit correlation between construction site environmental parameters, accurate positioning of the pollution source, and dynamic allocation and immediate correction of multi-level control strategies.

[0119] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0120] Based on the same inventive concept, the embodiment of the present application also provides a system for implementing the above-mentioned method for dynamically controlling the construction site environment by integrating AIoT. The implementation scheme for solving the problem provided by this system is similar to the implementation scheme described in the above-mentioned method. Therefore, the specific limitations in the embodiments of one or more systems for dynamically controlling the construction site environment by integrating AIoT provided below can be found in the above-mentioned limitations on the method for dynamically controlling the construction site environment by integrating AIoT, and will not be repeated here.

[0121] In an exemplary embodiment, Figure 3 As shown, a construction site environment dynamic control system 30 integrated with AIoT is provided, including a distributed Internet of Things monitoring network 31, a data preprocessing module 32, an abnormal fluctuation detection module 33, a causal relationship analysis module 34, a directed weighted network construction module 35, an abnormal source determination module 36 and a control instruction generation module 37; wherein, the Internet of Things monitoring network is composed of multi-source sensors deployed in the target area of ​​the construction site.

[0122] The distributed Internet of Things monitoring network 31 is used to collect raw data of the construction site environment and generate a raw environment data set with time stamp alignment.

[0123] The data preprocessing module 32 is used to preprocess the original environmental data set to generate a standardized environmental parameter data set; wherein the preprocessing includes wavelet threshold denoising processing, sliding window median filtering processing and maximum and minimum value normalization processing.

[0124] The abnormal fluctuation detection module 33 is used to detect abnormal fluctuations based on the standardized environmental parameter data set and the preset dynamic threshold through the isolation forest algorithm to generate an abnormal event marker sequence; wherein the abnormal event marker sequence includes the abnormal parameter ID, abnormal value and start and end timestamps.

[0125] The causal relationship analysis module 34 is used to intercept the time period data corresponding to the standardized environmental parameter data set based on the time window located by the abnormal event marker sequence, calculate the dependency between the parameters in the time period data through the Granger causality test, and generate the environmental parameter causal strength matrix.

[0126] The directed weighted network construction module 35 is used to construct a directed weighted network with environmental parameters as nodes, causal directions as directed edges, and causal strengths as weights based on the environmental parameter causal strength matrix, and obtain an environmental impact propagation path adjacency matrix.

[0127] The abnormality source determination module 36 is used to determine the abnormality source node identifier based on the matching verification of the weighted out-degree of each node in the environmental impact propagation path adjacency matrix and the timestamp of the abnormal event marker sequence.

[0128] The control instruction generation module 37 is used to calculate the impact propagation priority through the shortest path algorithm based on the abnormal source node identifier and the environmental impact propagation path adjacency matrix, and generate an environmental control instruction sequence based on the impact propagation priority.

[0129] Optionally, the abnormal fluctuation detection module includes:

[0130] The isolation forest construction unit is used to construct an isolation forest decision tree set containing multiple parameters based on a standardized environmental parameter data set; wherein each decision tree in the isolation forest decision tree set is generated by randomly selecting features and split points.

[0131] The path length statistics unit is used to count the average path length from the root node to the data point corresponding to each parameter in the isolation forest decision tree set.

[0132] The anomaly determination and marking unit is used to calculate the anomaly score based on the average path length. When the anomaly score exceeds the preset anomaly threshold, the corresponding data point is marked as an anomaly point, and an anomaly event marking sequence containing the anomaly parameter ID, anomaly value, and start and end timestamps corresponding to the anomaly point is generated.

[0133] Optional causal relationship analysis module includes:

[0134] The time series extraction unit is used to extract the time series of each parameter of the corresponding time period from the standardized environmental parameter data set based on the start and end timestamps in the abnormal event mark sequence.

[0135] The Granger causality test unit is used to perform a Granger causality test on each pair of parameters (A, B) based on the time series to determine whether there is a Granger causal effect from parameter A to parameter B. The Granger causality test unit includes:

[0136] The first autoregressive model construction and calculation subunit is used to construct the first autoregressive model based on the time series of parameter B and calculate the residual sum of squares of the first autoregressive model; the expression of the first autoregressive model is:

[0137]

[0138] Among them, B t is the observed value of parameter B at time t, B t-i Represents the observed value of parameter B at time ti, i is the lag order of parameter B, p is the number of lag terms of parameter B, ɑ i is the autoregressive coefficient of the first autoregressive model, ∈ t is the residual term of the first autoregressive model.

[0139] The second autoregressive model construction and calculation subunit is used to construct the second autoregressive model based on the time series of parameters A and B, and calculate the residual sum of squares of the second autoregressive model; the expression of the second autoregressive model is:

[0140]

[0141] Among them, A t-j represents the observed value of parameter A at time tj, j is the lag order, q is the number of lag terms of parameter A, β i is the autoregressive coefficient of the second autoregressive model, γ j It is the coefficient that measures the effect of the lag term of parameter A on the current value of parameter B, μ t is the residual term of the second autoregressive model.

[0142] The F-statistic calculation subunit is used to calculate the F-statistic of the significance of the causal effect of parameter A on parameter B based on the residual sum of squares of the first autoregressive model and the residual sum of squares of the second autoregressive model. The calculation formula of the F-statistic is:

[0143]

[0144] Where RSS1 is the residual sum of squares of the first autoregressive model, RSS2 is the residual sum of squares of the second autoregressive model, and T is the sample size.

[0145] The causal influence determination subunit is used to obtain the corresponding significance probability by querying the F distribution table based on the F statistic and degrees of freedom (q, Tpq). When the F statistic exceeds the preset critical value and the significance probability is less than the preset probability value, it is determined that there is a Granger causal influence from parameter A to parameter B.

[0146] Causal strength matrix generation unit, used when there is Granger causality, based on the coefficient γ in the second autoregressive model j The causal strength value is calculated by the standardized absolute value of , and all the causal strength values ​​are filled into an N×N matrix to generate the environmental parameter causal strength matrix; where N is the number of parameter types of the original data of the construction site environment.

[0147] Optional, directed weighted network building blocks include:

[0148] The directed edge determination unit is used to determine the existence of directed edges between nodes based on the positions of elements greater than 0 in the environmental parameter causal strength matrix, and obtain a node set and a directed edge set.

[0149] The weight mapping unit is used to map the element values ​​in the environmental parameter causal intensity matrix to the weight values ​​of the corresponding directed edges to obtain a weight set.

[0150] The adjacency matrix construction unit is used to integrate the node set, the directed edge set and the weight set, construct the adjacency matrix data structure, and obtain the environmental impact propagation path adjacency matrix.

[0151] Optionally, the abnormality source determination module includes:

[0152] The weighted out-degree calculation unit is used to sum the adjacency matrix of the environmental impact propagation path row by row to obtain the weighted out-degree value of each node.

[0153] The candidate source screening unit is used to screen nodes whose weighted out-degree values ​​exceed a preset source influence threshold to obtain a set of candidate source nodes.

[0154] The abnormal source identification generating unit is used to compare the abnormal start timestamp of each node in the candidate source node set with the timestamps of adjacent nodes in the abnormal event mark sequence. When the node start time is earlier than all adjacent nodes, the abnormal source node identification is generated.

[0155] Optionally, the control instruction generation module includes:

[0156] The impact propagation path calculation unit is used to use the abnormal source node as the path starting point, and calculate the shortest path length and path cumulative weight from the starting point to other nodes in the network defined by the environmental impact propagation path adjacency matrix through the Dijkstra algorithm.

[0157] The impact priority sorting unit is used to sort in ascending order by the shortest path length. When the shortest path lengths are the same, they are sorted in descending order by the cumulative weight of the path to generate an environment parameter impact priority queue.

[0158] The control instruction generation unit is used to traverse the environmental parameter impact priority queue and generate an environmental control instruction sequence when the path cumulative weight of the target parameter exceeds the emergency processing threshold corresponding to the type of the target parameter.

[0159] An embodiment of the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0160] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0161] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts 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 can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0162] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A method for dynamically controlling a construction site environment by integrating AIoT, characterized in that: The method comprises: S1. Deploy multi-source sensors in the target area of ​​the construction site to form a distributed Internet of Things monitoring network; based on the distributed Internet of Things monitoring network, collect raw data of the construction site environment and generate a raw environment data set with aligned timestamps; S2. Preprocessing the original environmental data set to generate a standardized environmental parameter data set; wherein the preprocessing includes wavelet threshold denoising, sliding window median filtering, and maximum and minimum value normalization; S3. Based on the standardized environmental parameter data set and the preset dynamic threshold, abnormal fluctuation detection is performed using the isolation forest algorithm to generate an abnormal event marker sequence; wherein the abnormal event marker sequence includes an abnormal parameter ID, an abnormal value, and a start and end timestamp; S4. Based on the time window of the abnormal event marker sequence, intercept the time period data corresponding to the standardized environmental parameter data set, calculate the dependency relationship between the parameters in the time period data through Granger causality test, and generate an environmental parameter causal strength matrix; S5. Based on the environmental parameter causal intensity matrix, construct a directed weighted network with environmental parameters as nodes, causal directions as directed edges, and causal intensity as weights to obtain an environmental impact propagation path adjacency matrix; S6. Determine the abnormal source node identifier based on the matching verification of the weighted out-degree of each node in the environmental impact propagation path adjacency matrix and the timestamp of the abnormal event marker sequence; S7. Based on the abnormal source node identifier and the environmental impact propagation path adjacency matrix, calculate the impact propagation priority through the shortest path algorithm, and generate an environmental control instruction sequence based on the impact propagation priority.

2. The method according to claim 1, characterized in that The S3 includes: S31. Based on the standardized environmental parameter data set, construct an isolation forest decision tree set containing multiple parameters; wherein each decision tree in the isolation forest decision tree set is generated by randomly selecting features and split points; S32, counting the average path length from the root node to the data point corresponding to each parameter in the isolation forest decision tree set; S33. Calculate an anomaly score based on the average path length. When the anomaly score exceeds a preset anomaly threshold, mark the corresponding data point as an anomaly point, and generate the anomaly event marker sequence including the anomaly parameter ID, an anomaly value, and a start and end timestamp corresponding to the anomaly point.

3. The method according to claim 1, characterized in that The S4 includes: S41, extracting the time series of each parameter of the corresponding time period from the standardized environmental parameter data set based on the start and end timestamps in the abnormal event mark sequence; S42. Based on the time series, perform a Granger causality test on each pair of parameters (A, B) to determine whether there is a Granger causal effect from parameter A to parameter B, including: S421: Build a first autoregressive model based on the time series of parameter B, and calculate the residual sum of squares of the first autoregressive model; the expression of the first autoregressive model is: Among them, B t is the observed value of parameter B at time t, B t-i Represents the observed value of parameter B at time ti, i is the lag order of parameter B, p is the number of lag terms of parameter B, α i is the autoregressive coefficient of the first autoregressive model, ∈ t is the residual term of the first autoregressive model; S422: Construct a second autoregressive model based on the time series of parameter A and parameter B, and calculate the residual sum of squares of the second autoregressive model; the expression of the second autoregressive model is: Among them, A t-j represents the observed value of parameter A at time tj, j is the lag order, q is the number of lag terms of parameter A, β i is the autoregressive coefficient of the second autoregressive model, γ j It is the coefficient that measures the effect of the lag term of parameter A on the current value of parameter B, μ t is the residual term of the second autoregressive model; S423. Calculate the F statistic of the significance of the causal effect of parameter A on parameter B based on the residual sum of squares of the first autoregressive model and the residual sum of squares of the second autoregressive model. The calculation formula of the F statistic is: Wherein, RSS1 is the residual sum of squares of the first autoregressive model, RSS2 is the residual sum of squares of the second autoregressive model, and T is the sample size; S424. Based on the F statistic and the degrees of freedom (q, Tpq), obtain the corresponding significance probability by querying the F distribution table. When the F statistic exceeds a preset critical value and the significance probability is less than a preset probability value, determine that there is a Granger causal effect from parameter A to parameter B. S43. When Granger causality exists, based on the coefficient γ in the second autoregressive model, j The causal strength value is calculated by the standardized absolute value of , and all the causal strength values ​​are filled into an N×N matrix to generate the environmental parameter causal strength matrix; wherein N is the number of parameter types of the original data of the construction site environment.

4. The method according to claim 1, wherein The S5 includes: S51. Based on the positions of elements greater than 0 in the environmental parameter causal strength matrix, determine the existence of directed edges between nodes to obtain a node set and a directed edge set; S52, mapping element values ​​in the environmental parameter causal intensity matrix to weight values ​​of corresponding directed edges to obtain a weight set; S53: Integrate the node set, the directed edge set, and the weight set to construct an adjacency matrix data structure to obtain the environmental impact propagation path adjacency matrix.

5. The method according to claim 1, wherein The S6 includes: S61, summing the adjacency matrix of the environmental impact propagation path row by row to obtain a weighted out-degree value of each node; S62: Filter the nodes whose weighted out-degree values ​​exceed a preset source influence threshold to obtain a set of candidate source nodes; S63: Compare the abnormal start timestamp of each node in the candidate source node set with the timestamps of adjacent nodes in the abnormal event mark sequence. When the node start time is earlier than all adjacent nodes, generate an abnormal source node identifier.

6. The method according to any one of claims 1 to 5, characterized in that The S7 includes: S71, using the abnormal source node identifier as the path starting point, in the network defined by the environmental impact propagation path adjacency matrix, calculate the shortest path length and path cumulative weight from the starting point to other nodes using the Dijkstra algorithm; S72. Sort in ascending order by the shortest path length. If the shortest paths have the same length, sort in descending order by the cumulative weight of the paths to generate an environment parameter impact priority queue. S73. Traverse the environmental parameter impact priority queue, and when the path cumulative weight of the target parameter exceeds the emergency processing threshold corresponding to the type of the target parameter, generate the environmental control instruction sequence.

7. A construction site environment dynamic control system integrated with AIoT, characterized by: The system includes: a distributed Internet of Things monitoring network, a data preprocessing module, an abnormal fluctuation detection module, a causal relationship analysis module, a directed weighted network construction module, an abnormal source determination module, and a control instruction generation module; wherein the Internet of Things monitoring network is composed of multi-source sensors deployed in the target area of ​​the construction site; The distributed Internet of Things monitoring network is used to collect raw data on the construction site environment and generate a raw environmental data set with aligned timestamps; The data preprocessing module is used to preprocess the original environmental data set to generate a standardized environmental parameter data set; wherein the preprocessing includes wavelet threshold denoising, sliding window median filtering, and maximum and minimum value normalization; The abnormal fluctuation detection module is used to detect abnormal fluctuations based on the standardized environmental parameter data set and the preset dynamic threshold value using the isolation forest algorithm to generate an abnormal event marker sequence; wherein the abnormal event marker sequence includes an abnormal parameter ID, an abnormal value, and a start and end timestamp; The causal relationship analysis module is used to intercept the time period data corresponding to the standardized environmental parameter data set based on the time window located by the abnormal event marker sequence, calculate the dependency relationship between the parameters in the time period data through Granger causality test, and generate an environmental parameter causal strength matrix; The directed weighted network construction module is used to construct a directed weighted network based on the environmental parameter causal strength matrix, with environmental parameters as nodes, causal directions as directed edges, and causal strengths as weights, to obtain an environmental impact propagation path adjacency matrix; The abnormality source determination module is used to determine the abnormality source node identifier based on the matching verification of the weighted out-degree of each node in the environmental impact propagation path adjacency matrix and the timestamp of the abnormal event marker sequence; The control instruction generation module is used to calculate the impact propagation priority through the shortest path algorithm based on the abnormal source node identifier and the environmental impact propagation path adjacency matrix, and generate an environmental control instruction sequence based on the impact propagation priority.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images