Construction power utilization and special equipment safety monitoring method and system

Abstract

The application provides a construction power and special equipment safety monitoring method and system, relates to the field of construction safety management, and comprises the following steps: performing time series analysis, spatial analysis and logical reasoning on multi-source heterogeneous data containing power parameters and special equipment operation state parameters; processing the collected multi-source data based on a time-space sequence network to obtain a current health state score and a remaining service life prediction value of the equipment; generating the severity of an abnormal situation of the monitored data, the cause of the abnormal situation and the affected upstream and downstream equipment as early warning information according to the analysis results of the time series dimension, the spatial dimension and the reasoning dimension; and generating predictive maintenance early warning information according to the health state score and the remaining service life prediction value. Early identification and predictive maintenance of the equipment performance degradation trend are realized, and the intelligent level of the construction power and special equipment safety monitoring is significantly improved.

Description

Methods and systems for safety monitoring of construction power supply and special equipment Technical Field

[0001] This invention relates to the field of construction safety management technology, and more specifically, to a method and system for monitoring the safety of construction electricity and special equipment. Background Technology

[0002] In the current construction industry, temporary power safety and special equipment safety monitoring are typically two independent subsystems: 1. Power monitoring: Primarily monitors parameters such as voltage, current, leakage current, and cable temperature in distribution boxes, lacking linkage analysis with equipment operating status. 2. Equipment monitoring: For example, tower crane black boxes mainly monitor torque, amplitude, and weight; construction elevators monitor load, speed, floor position, and door lock status; large bridge erecting equipment monitors stress, displacement, and hydraulic pressure. The separation of power data and equipment operating data makes it impossible to assess the overall safety status. Most systems only alarm when parameters exceed limits, which is a reactive alarm and cannot provide pre-emptive prediction or trend warnings. Management personnel need to monitor multiple independent platforms simultaneously, lacking unified command and decision support.

[0003] Therefore, how to achieve accurate perception and intelligent control of the overall status of equipment safety and power consumption has become a technical problem that needs to be solved. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art or related technologies, and discloses a method and system for monitoring the safety of construction power supply and special equipment. It realizes early identification and predictive maintenance of equipment performance degradation trends, significantly improves the intelligent level of construction power supply and special equipment safety monitoring, and has important engineering application value.

[0005] The first aspect of this invention discloses a method for monitoring the safety of construction power supply and special equipment, comprising: multi-dimensional collaborative correlation analysis: performing time-series analysis, spatial analysis, and logical reasoning on multi-source heterogeneous data containing power parameters and special equipment operating status parameters, wherein the time-series analysis includes: using a statistical method based on Granger causality test to calculate time-series correlation, quantifying and judging the causal direction and strength between power supply events and equipment status, and obtaining the analysis results of the time-series dimension; the spatial analysis includes: assigning spatial attributes to the monitoring data based on geographic information systems and real-time positioning technology, calculating the electrical and physical distances between special equipment, and obtaining the analysis results of the spatial dimension; the logical reasoning includes: performing logical reasoning on the analysis results of the time-series dimension and the spatial dimension based on rule engines and knowledge graphs to deduce potentially affected upstream and downstream equipment. The system prepares data for analysis based on logical reasoning. It then performs model-based intelligent diagnosis by processing multi-source data collected using a spatiotemporal sequence network (FSLM), which integrates CNN and LSTM network structures, including sequentially connected convolutional layers, pooling layers, LSTM layers, and fully connected layers. The LSTM layers also incorporate an attention mechanism to weight time steps. Real-time collected device operation data is input into the trained FLM to obtain the device's current health status score and predicted remaining lifespan. Finally, it provides tiered early warning: based on the analysis results from the temporal, spatial, and logical reasoning dimensions, it generates the severity of anomalies in the monitored data, the causes of these anomalies, and the affected upstream and downstream devices, serving as early warning information. Predictive maintenance early warning information is also generated based on the health status score and predicted remaining lifespan.

[0006] This technical solution combines a rule engine with a spatiotemporal sequence network algorithm to achieve deep cross-system correlation analysis, health status assessment, and intelligent early warning. On one hand, the rule engine performs real-time logical judgments based on a pre-built expert knowledge base; on the other hand, the spatiotemporal sequence network algorithm is trained based on historical health data from long-term equipment operation to construct a dynamic health baseline for the equipment, enabling the assessment of its health status. The outputs of both aspects complement each other, jointly describing electrical safety and equipment safety status. For example, the spatiotemporal sequence network detects an early abnormal trend in the pressure of the main lifting cylinder of a bridge-building machine (yellow warning), while the rule engine monitors an abnormal increase in the filter element pressure difference in the circuit containing that cylinder (yellow warning). After knowledge graph fusion analysis, it is determined that "increased hydraulic oil contamination leads to early symptoms of internal leakage in the cylinder," and the overall risk level is raised to orange (high risk). The rule engine outputs deterministic and causally clear judgments. For example, "If the main circuit current > 200% of the rated value and the duration > 2 seconds, then trigger an overload trip warning (emergency level)"; the spatiotemporal sequence network outputs probabilistic and correlational judgments. For example, the conclusion of the spatiotemporal sequence network is based on statistical significance: "The current vibration pattern has an anomaly probability of 92% compared to the historical baseline, and its deterioration trend predicts that it may exceed the limit within the next 72 hours." If the conclusions of the two systems conflict, the conclusion of the rule engine shall prevail.

[0007] According to the construction power supply and special equipment safety monitoring method disclosed in this invention, preferably, before the step of multi-dimensional collaborative correlation analysis, it further includes:

[0008] Collect multi-source heterogeneous data: Through sensor units deployed on various monitoring objects at the construction site, collect real-time operating status data of the construction power system and special equipment; transmit the collected multi-source heterogeneous data to the cloud platform or central server via the network;

[0009] Data cleaning and spatiotemporal alignment: For the collected multi-source heterogeneous data, obvious outliers and invalid data caused by sensor failure or transmission interference are removed; linear interpolation, spline interpolation, or time series-based prediction and imputation algorithms are used to repair the data and ensure data continuity; Kalman filtering is used for real-time signal smoothing, and wavelet transform is used for multi-scale noise separation; a time synchronization protocol is used to assign a unified millisecond-level precision timestamp to all heterogeneous data sources and establish a spatiotemporal correlation mapping table to bind the special equipment operation data with the equipment spatial coordinates, forming a multi-source heterogeneous data set with spatiotemporal labels.

[0010] According to the construction power supply and special equipment safety monitoring method disclosed in this invention, preferably, the training process of the spatiotemporal sequence network includes:

[0011] The spatiotemporal sequence network was trained using special equipment operating status data and corresponding historical health status data. Mean squared error was used as the loss function, and the network parameters were optimized using backpropagation. An early stopping strategy was employed to prevent overfitting, and cross-validation was used to evaluate the model performance. The special equipment operating status data and corresponding historical health status data included: unique equipment ID, model, manufacturer, manufacturing date, current, voltage, pressure, flow rate, and speed.

[0012] According to the safety monitoring method for construction power supply and special equipment disclosed in this invention, preferably, the time-series analysis specifically includes:

[0013] When a characteristic surge curve is detected in the starting current of a special equipment motor, the change in the bus voltage of the main distribution box within the same time window is analyzed simultaneously. By calculating the time-delay cross-correlation function and transfer entropy value of the two time series, it is determined whether there is a statistically significant causal relationship, thereby identifying the cause of the current surge curve.

[0014] According to the safety monitoring method for construction power supply and special equipment disclosed in this invention, preferably, the spatial analysis specifically includes:

[0015] Each distribution box and each special equipment is assigned a unique spatial coordinate; the electrical and physical distances between equipment are calculated; when a power grid disturbance occurs, not only is time-series analysis performed, but also verification is carried out in combination with the principle of spatial proximity: the equipment with the closest electrical distance and the highest power is given priority as a potential source; at the same time, the spatial interference risk between equipment is predicted by tracking the movement trajectory of mobile equipment in real time.

[0016] According to the construction power supply and special equipment safety monitoring method disclosed in this invention, preferably, the logical reasoning specifically includes:

[0017] A reasoning model based on a rule engine and a knowledge graph is constructed. The rule engine includes expert knowledge and safety procedures. The knowledge graph is used to model the power supply and consumption relationship network at the construction site, representing the topological connection relationship and electrical characteristics between "distribution box-power supply line-electrical equipment". After identifying potential causal relationships through time sequence analysis and spatial analysis, the impact path of the event is traced in the knowledge graph to deduce the upstream and downstream equipment that may be affected, thereby achieving a comprehensive assessment from single-point events to system-level impacts.

[0018] According to the construction power supply and special equipment safety monitoring method disclosed in this invention, preferably, the convolutional layer, pooling layer, and LSTM layer of the spatiotemporal sequence network are configured for multi-channel parallel processing. The outputs of multiple channels are fused using a concatenate operation and then fed into a fully connected layer. The fully connected layer has a linear activation function for outputting the prediction result. The computation process of the spatiotemporal sequence network specifically includes:

[0019] The input data first passes through a convolutional layer to extract features;

[0020] The output of the convolutional layer is further processed by the pooling layer, which reduces the dimensionality of the data by extracting the maximum value in each region;

[0021] The dimensionality-reduced feature sequence is input into an LSTM network for temporal modeling;

[0022] The output of the LSTM layer is fed into a fully connected layer, where the signal features are classified using the Softmax activation function to output the prediction result.

[0023] During model training, the Adam optimization algorithm is used for error backpropagation to update network parameters layer by layer.

[0024] According to the construction power supply and special equipment safety monitoring method disclosed in this invention, preferably, the power parameters include: three-phase voltage, current, power, leakage current and cable temperature of each level of distribution box.

[0025] According to the construction power supply and special equipment safety monitoring method disclosed in this invention, preferably, the special equipment includes: tower cranes, construction hoists, bridge building machines, cantilever beam building machines, and mobile formwork; the operating status data of the special equipment includes: lifting capacity, radius, height, torque, wind speed, tilt angle, running speed, and door lock status of tower cranes and construction hoists; stress and strain of key structural parts of bridge building machines, cantilever beam building machines, and mobile formwork, overall posture and deflection deformation, hydraulic system oil pressure and oil temperature, synchronization of traveling mechanism, and outrigger pressure.

[0026] The second aspect of the present invention discloses a construction power supply and special equipment safety monitoring system, comprising: a memory for storing program instructions; and a processor for calling the program instructions stored in the memory to implement the construction power supply and special equipment safety monitoring method as described in any of the above technical solutions.

[0027] The beneficial effects of this invention include at least the following: It constructs a multi-dimensional, deeply integrated causal correlation analysis engine, achieving intelligent identification and dynamic modeling of complex correlations between power consumption events and equipment operating conditions in a construction environment through the synergistic effect of three technical dimensions: time series analysis, spatial computation, and logical reasoning. By constructing a spatiotemporal sequence network, it fully utilizes the advantages of CNN networks in spatial feature extraction and the strengths of LSTM networks in time series modeling, achieving deep fusion and mining of spatiotemporal features of multi-source monitoring data of equipment, overcoming the limitations of traditional methods in effectively handling multi-sensor, multi-dimensional data. The predictive capability based on the spatiotemporal sequence network model can establish a dynamic health status baseline based on historical equipment operating data, improving monitoring accuracy and intelligence. Attached Figure Description

[0028] Figure 1 shows a schematic diagram of the spatiotemporal sequence network of a construction power supply and special equipment safety monitoring method according to an embodiment of the present invention.

[0029] Figure 2 shows a schematic block diagram of a construction power supply and special equipment safety monitoring system according to an embodiment of the present invention.

[0030] Figure 3 shows a flowchart of a method for monitoring the safety of construction power supply and special equipment according to an embodiment of the present invention. Detailed Implementation

[0031] To better understand the above-described objects, features, and advantages of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Many specific details are set forth in the following description to provide a thorough understanding of the invention; however, the invention may be practiced in other ways different from those described herein, and therefore, the invention is not limited to the specific embodiments disclosed below.

[0032] As shown in Figure 3, an embodiment of the present invention discloses a method for monitoring the safety of construction power supply and special equipment, including:

[0033] Step 1, Collect multi-source heterogeneous data: Collect real-time operating status data of construction power system and special equipment by sensor units deployed on various monitoring objects at the construction site; transmit the collected multi-source heterogeneous data to cloud platform or central server through network;

[0034] Step 2, Data Cleaning and Spatiotemporal Alignment: For the collected multi-source heterogeneous data, obvious outliers and invalid data caused by sensor failure or transmission interference are removed; linear interpolation, spline interpolation, or time series-based prediction imputation algorithms are used for data repair to ensure data continuity; Kalman filtering is used for real-time signal smoothing, and wavelet transform is used for multi-scale noise separation; a time synchronization protocol is used to assign a unified millisecond-level precision timestamp to all heterogeneous data sources and establish a spatiotemporal correlation mapping table to bind the special equipment operation data with the equipment spatial coordinates, forming a multi-source heterogeneous data set with spatiotemporal labels.

[0035] Step 3, Multi-dimensional Collaborative Correlation Analysis: For multi-source heterogeneous data containing power parameters and special equipment operating status parameters, time-series analysis, spatial analysis, and logical reasoning are performed. Time-series analysis includes: using statistical methods based on Granger causality tests to calculate time-series correlations, quantifying the causal direction and strength between power consumption events and equipment status, and obtaining time-series analysis results; spatial analysis includes: assigning spatial attributes to monitoring data based on geographic information systems and real-time positioning technology, calculating the electrical and physical distances between special equipment, prioritizing the equipment with the closest electrical distance and highest power as potential root causes, and obtaining spatial analysis results; logical reasoning includes: using rule engines and knowledge graphs to perform logical reasoning on the time-series and spatial analysis results to deduce potentially affected upstream and downstream equipment, and obtaining logical reasoning analysis results.

[0036] Step 4, Model-Based Intelligent Diagnosis: The collected multi-source data is processed using a spatiotemporal sequence network. This network integrates CNN and LSTM (Long Short-Term Memory) structures, including sequentially connected convolutional layers, pooling layers, LSTM layers, and fully connected layers. The LSTM layers also incorporate an attention mechanism to weight time steps. The convolutional, pooling, and LSTM layers are configured for multi-channel parallel processing. The outputs from multiple channels are fused using a concatenation operation before being fed into the fully connected layer, which has a linear activation function to output the prediction results. Real-time collected device operation data is input into the trained spatiotemporal sequence network to obtain the device's current health status score and predicted remaining lifespan.

[0037] Step 5, Tiered Early Warning: Based on the analysis results of the time-series dimension, spatial dimension, and inference dimension, generate the severity of the abnormal situation of the monitored data, the cause of the abnormal situation, and the affected upstream and downstream equipment as early warning information; and generate predictive maintenance early warning information based on the health status score and the remaining service life prediction value.

[0038] According to the above embodiments, the training process of the spatiotemporal sequence network further includes:

[0039] The spatiotemporal sequence network was trained using special equipment operation status data and corresponding historical health status data. Mean squared error was used as the loss function, the network parameters were optimized by backpropagation algorithm, an early stopping strategy was adopted to prevent overfitting, and cross-validation method was used to evaluate the model performance.

[0040] According to the above embodiments, the timing analysis further includes:

[0041] When a characteristic surge curve is detected in the starting current of a special equipment motor, the change in the bus voltage of the main distribution box within the same time window is analyzed simultaneously. By calculating the time-delay cross-correlation function and transfer entropy value of the two time series, it is determined whether there is a statistically significant causal relationship, thereby identifying the cause of the current surge curve.

[0042] According to the above embodiments, the spatial analysis further includes:

[0043] Each distribution box and each special equipment is assigned a unique spatial coordinate; the electrical and physical distances between equipment are calculated; when a power grid disturbance occurs, not only is time-series analysis performed, but also verification is carried out in combination with the principle of spatial proximity: the equipment with the closest electrical distance and the highest power is given priority as a potential source; at the same time, the spatial interference risk between equipment is predicted by tracking the movement trajectory of mobile equipment in real time.

[0044] According to the above embodiments, the logical reasoning further includes:

[0045] A reasoning model based on a rule engine and a knowledge graph is constructed. The rule engine includes expert knowledge and safety procedures. The knowledge graph is used to model the power supply and consumption relationship network at the construction site, representing the topological connection relationship and electrical characteristics between "distribution box-power supply line-electrical equipment". After identifying potential causal relationships through time sequence analysis and spatial analysis, the impact path of the event is traced in the knowledge graph to deduce the upstream and downstream equipment that may be affected, thereby achieving a comprehensive assessment from single-point events to system-level impacts.

[0046] According to the above embodiments, the computation process of the spatiotemporal sequence network further includes:

[0047] The input data first passes through a convolutional layer to extract features;

[0048] The output of the convolutional layer is further processed by the pooling layer, which reduces the dimensionality of the data by extracting the maximum value in each region;

[0049] The dimensionality-reduced feature sequence is input into an LSTM network for temporal modeling;

[0050] The output of the LSTM layer is fed into a fully connected layer, where the signal features are classified using the Softmax activation function to output the prediction result.

[0051] During model training, the Adam optimization algorithm is used for error backpropagation to update network parameters layer by layer.

[0052] According to the above embodiments, the power parameters further include: the three-phase voltage, current, power, leakage current and cable temperature of each level of distribution box.

[0053] According to the above embodiments, the special equipment further includes: tower cranes, construction hoists, bridge building machines, cantilever beam building machines, and mobile formwork; the operating status data of the special equipment includes: the lifting capacity, radius, height, torque, wind speed, tilt angle, running speed, and door lock status of the tower cranes and construction hoists; the stress and strain of key structural parts, overall posture and deflection deformation, hydraulic system oil pressure and temperature, synchronization of the traveling mechanism, and outrigger pressure of the bridge building machines, cantilever beam building machines, and mobile formwork.

[0054] According to another embodiment of the present invention, the specific application of the construction power supply and special equipment safety monitoring method of the above embodiments is also disclosed:

[0055] S1 Multi-Source Heterogeneous Data Acquisition and Upload:

[0056] By deploying sensor units on various monitoring objects at the construction site, real-time operational status data of the construction power system and special equipment are collected. The construction power system data includes three-phase voltage, current, power, leakage current, and cable temperature of each level of distribution box. The special equipment includes tower cranes, construction hoists, bridge building machines, cantilever beam building machines, and mobile formwork. Their operational status data includes: lifting capacity, amplitude, height, torque, wind speed, tilt angle, running speed, and door lock status of tower cranes and construction hoists; stress and strain of key structural parts, overall posture and deflection deformation, hydraulic system oil pressure and temperature, synchronization of the traveling mechanism, and outrigger pressure of bridge building machines, cantilever beam building machines, and mobile formwork. The collected multi-source heterogeneous data is transmitted to a cloud platform or central server via wired or wireless networks.

[0057] S2 data cleaning, spatiotemporal alignment, and feature extraction:

[0058] The cloud platform or central server performs standardized preprocessing and deep processing on the multi-source heterogeneous streaming data received by S1. By establishing a comprehensive data quality control pipeline, high-quality, structured time-series datasets are prepared for subsequent multi-source data fusion analysis. During the data cleaning and denoising stage, the system employs multiple techniques to ensure data reliability: First, anomaly detection is performed based on physical thresholds and statistical rules, using a sliding window algorithm to identify and remove obvious outliers and invalid data caused by sensor malfunctions or transmission interference. For transient missing data in the time-series data, linear interpolation, spline interpolation, or time-series-based prediction imputation algorithms are used for data repair to ensure data continuity. To further eliminate high-frequency noise and signal jitter, the system applies advanced digital signal processing techniques, including Kalman filtering for real-time signal smoothing and wavelet transform for multi-scale noise separation, effectively preserving the true trend characteristics of the signal. Simultaneously, the system establishes a data quality assessment mechanism to quantitatively evaluate the integrity, consistency, and accuracy of the cleaned data, providing a reliable data foundation for subsequent analysis.

[0059] After the data preprocessing is completed, the system performs refined spatio-temporal alignment and feature extraction operations. In the spatio-temporal alignment stage, a high-precision time synchronization protocol is adopted to assign a unified millisecond-level accurate timestamp to all heterogeneous data sources, and a spatio-temporal association mapping table is established to accurately bind the device operation data with its spatial coordinates, forming a multi-dimensional data set with spatio-temporal tags.

[0060] S3 Rule- and Model-Based Multi-Source Data Fusion Health Analysis and Linkage Warning:

[0061] S3.1 Causal Association Analysis Engine with Multi-Dimensional Deep Fusion:

[0062] Through the collaborative action of three technical dimensions: statistical causal analysis (Granger) in time series, real-time relationship calculation (GIS) in space, and impact chain deduction in logic (rule engine + knowledge graph), this engine establishes a three-dimensional and dynamic association model. This model completely changes the limitation of isolated alarms in traditional monitoring systems and realizes the intelligent identification and dynamic modeling of the internal association between power consumption events and equipment operating conditions in complex construction environments.

[0063] At the time series analysis level, the system adopts a statistical method based on Granger causality test to deeply mine the high-frequency synchronous data of the power consumption system and equipment operation. By calculating the time series correlation between power parameters such as voltage and current and equipment operation state parameters, the system can quantitatively judge the causal direction and intensity between power consumption events and equipment states. When a characteristic soaring curve appears in the starting current of the tower crane motor, the system will simultaneously analyze the change of the bus voltage of the main distribution box within the same time window. By calculating the time-delay cross-correlation function and transfer entropy value of the two time series, the system can scientifically determine whether there is a statistically significant causal relationship, so as to accurately identify that "equipment startup causes power grid disturbance" rather than a simple time series coincidence.

[0064] In the spatial calculation dimension, the system introduces a geographic information system (GIS) and real-time positioning technology to endow all monitoring data with spatial attributes. Each distribution box and each special equipment have unique spatial coordinates, and the system dynamically calculates the electrical distance and physical distance between devices through a spatial database. When a power grid disturbance occurs, the system not only analyzes the time series causal relationship but also verifies it in combination with the principle of spatial proximity: preferentially examines the device with the closest electrical distance and the largest power as the potential root cause. At the same time, by real-time tracking the movement trajectories of mobile devices (such as tower crane hooks and bridge erector booms), the system can predict the spatial interference risk between devices and provide spatial relationship support for realizing active anti-collision warning.

[0065] At the logical reasoning level, the system constructs an intelligent reasoning module based on a rule engine and a knowledge graph. The rule engine incorporates rich expert knowledge and safety procedures, defining various "if-then" logical rules. The knowledge graph models the complex power supply and consumption network at the construction site, clearly representing the topological connections and electrical characteristics between "distribution boxes, power lines, and electrical equipment." After time-series analysis and spatial computation initially identify potential causal relationships, the system activates the logical reasoning engine to trace the impact path of events in the knowledge graph, deducing potentially affected upstream and downstream equipment, thus achieving a comprehensive assessment from single-point events to system-level impact. These three dimensions of analysis are not conducted in isolation but are collaboratively optimized through a deeply integrated algorithmic framework. Time-series analysis discovers statistical correlations, spatial computation provides geographical location verification, and logical reasoning completes the influence chain deduction, ultimately forming a complete causal judgment. For example, when the system detects a sudden drop in grid voltage, it first uses Granger causality analysis to identify a tower crane that started at the same time; then, it uses the GIS system to confirm that the tower crane is indeed powered by the distribution box where the voltage drop occurred; finally, it uses a knowledge graph to deduce other potentially affected precision equipment on the same line (such as the CNC system of a bridge-building machine) and generates tiered early warning information. Tiered early warning refers to classifying early warning information into different levels based on the causal certainty, potential impact range, and severity of the grid disturbance event, and taking corresponding response measures. For example, Level 1: Attention Level (blue / information alert), time series analysis shows weak statistical correlation; Level 2: Warning Level (yellow / warning), time series analysis shows strong statistical significance; Level 3: Alarm Level (orange / high risk), time series analysis shows extremely strong statistical significance.

[0066] S3.2 A predictive maintenance method for special equipment based on spatiotemporal sequence network algorithms and health status baselines:

[0067] To address the shortcomings of existing technologies that only provide passive alarms after exceeding limits, a spatiotemporal sequence network is proposed as an innovative method for predictive maintenance. The health status baseline refers to the pattern, range, or mathematical characteristics that key performance parameters, operating data, and environmental responses should follow when equipment is in normal, good working condition.

[0068] As shown in Figure 1, the spatiotemporal sequence network proposed in this invention is a hybrid network structure that integrates LSTM and an attention mechanism within the overall framework of a CNN network. It is used for modeling and predicting sequential data. In prediction, the main role of the spatiotemporal sequence network is feature extraction and temporal modeling. Convolutional layers can filter and extract features from the input time-series data, extracting key features from the sequence. LSTM can perform temporal modeling on these features, capturing long-term dependencies in the sequence and performing excellently in prediction tasks. This spatiotemporal sequence network also integrates an attention mechanism within the LSTM layer: To construct the spatiotemporal sequence network model with an attention mechanism, firstly, the input data undergoes spatial feature extraction via a CNN (convolutional layer), with the kernel size, stride, and number of filters set to 12, 3, and 128, respectively, followed by a pooling layer to enhance the features. Subsequently, the extracted feature sequence is fed into an LSTM to capture temporal dependencies. Finally, the output vector of the LSTM hidden layer is input to the attention layer (attention mechanism) to weight the time steps and highlight key information. Repeat the above steps to construct a spatiotemporal sequence network model channel with an attention mechanism, connect the outputs of multiple channels, perform feature fusion using the Concatenate operation, input the fusion result into a fully connected layer with a linear activation function, and finally output the classification (prediction) result.

[0069] The specific computational process of the spatiotemporal sequence network model is as follows: (1) The input data is first standardized to unify the data scale, and then the features are adaptively extracted through the sliding window operation. (2) The output of the convolutional layer is further processed by the max pooling layer to achieve data dimensionality reduction by extracting the maximum value in each region, while retaining key features. (3) The dimensionality-reduced feature sequence is input into the LSTM model for temporal modeling. (4) During the model training process, the Adam optimization algorithm is used for backpropagation of error to update the network parameters layer by layer. (5) The signal features are classified through the Softmax activation function to achieve the classification task of multi-feature input sequences. (6) The final result is given by the output layer.

[0070] The core of the LSTM layer is the gating unit, which includes the input gate, forget gate, and output gate. These gating units are used to control the reading, writing, and forgetting of input data, thereby retaining and selecting important information in long sequences. Each LSTM unit calculates the input according to formulas (1)-(6):

[0071] (1)

[0072] (2)

[0073] (3)

[0074] (4)

[0075] (5)

[0076] (6)

[0077] in, It is the hidden state at time t. It is the hidden state at time t-1. This represents taking the sigmoid function. It is the tuple state at time t. It is the tuple state at time t-1; It is the input at time t; , , , These represent the input gate, forget gate, selection gate, and output gate, respectively. The weight matrix representing the input gate, The weight matrix representing the forget gate, The weight matrix represents the selection gate. The weight matrix represents the output gate. This represents the bias vector of the input gate. The bias vector representing the forget gate. This represents the bias vector of the selection gate. This represents the bias vector of the output gate.

[0078] Model Training: The spatiotemporal sequence network hybrid model is trained using historical data, with mean squared error (MSE) as the loss function. Backpropagation is used to optimize network parameters. An early stopping strategy is employed to prevent overfitting, and cross-validation is used to evaluate model performance. Real-time collected equipment operation data is input into the trained spatiotemporal sequence network model to obtain the current health status score and predicted remaining service life. When the health status score falls below a preset threshold or the performance degradation trend exceeds the allowable range, the system generates a predictive maintenance warning, indicating maintenance needs and recommended measures.

[0079] According to another embodiment of the present invention, the following situation is disclosed regarding the triggering of subsequent linkage response events in the actual application of the construction power supply and special equipment safety monitoring method provided in the above embodiments: the spatiotemporal sequence network detects an early abnormal trend in the pressure of the main lifting cylinder of a bridge-building machine (yellow warning), and at the same time, the rule engine monitors an abnormal increase in the filter element pressure difference of the circuit where the cylinder is located (yellow warning). After knowledge graph fusion analysis, it is determined to be "early symptoms of internal leakage in the cylinder caused by increased hydraulic oil contamination", and the comprehensive risk level is raised to orange (high risk).

[0080] As shown in FIG. 2, according to another embodiment of the present invention, a construction electricity and special equipment safety monitoring system 200 is further disclosed, including: a memory 201 for storing program instructions; a processor 202 for calling the program instructions stored in the memory to implement the construction electricity and special equipment safety monitoring method as described in the above embodiments.

[0081] In summary, the present invention performs rule-based logical judgment and model-based intelligent diagnosis in parallel. On the one hand, the rule engine performs real-time logical judgment based on the preset expert knowledge base. On the other hand, the spatio-temporal sequence network algorithm is used to train based on the historical health data of the long-term operation of the equipment to construct its dynamic health baseline. The system compares the real-time data with the baseline for trend prediction and anomaly detection. The analysis results trigger a hierarchical early warning mechanism to initiate corresponding linkage response processes according to the event type, risk level and influence range. The present invention realizes in-depth correlation analysis, health status evaluation and intelligent early warning across systems by combining the rule engine and the spatio-temporal sequence network algorithm. It realizes the early identification of the equipment performance degradation trend and predictive maintenance, significantly improves the intelligent level of construction electricity and special equipment safety monitoring, and has important engineering application value.

[0082] All or part of the steps in the various methods of the above embodiments can be completed by controlling related hardware through a program, and the program can be stored in a readable storage medium. The storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disk memories, tape memories, or any other readable medium capable of carrying or storing data.

[0083] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for safety monitoring of construction electricity and special equipment, characterized in that, include: Multi-dimensional collaborative correlation analysis: This involves performing time-series analysis, spatial analysis, and logical reasoning on multi-source heterogeneous data containing power parameters and special equipment operating status parameters. The time-series analysis includes: using a statistical method based on Granger causality testing to calculate time-series correlations, quantifying the causal direction and strength between power consumption events and equipment status, and obtaining time-series analysis results. Specifically, the time-series analysis includes: when a characteristic surge curve is detected in the motor starting current of special equipment, simultaneously analyzing the changes in the main distribution box bus voltage within the same time window, and calculating the two time series... The time-delay cross-correlation function and transfer entropy value of the series are used to determine whether there is a statistically significant causal relationship, thereby identifying the cause of the current surge curve; the spatial analysis includes: assigning spatial attributes to monitoring data based on geographic information systems and real-time positioning technology, calculating the electrical and physical distances between special equipment, and obtaining spatial dimension analysis results; the spatial analysis specifically includes: assigning unique spatial coordinates to each distribution box and each special equipment; calculating the electrical and physical distances between equipment, and when power grid disturbances occur, not only time series analysis is performed, but also verification is carried out in combination with the principle of spatial proximity: priority is given to examining the electrical distance closest to the target. The nearest and most powerful device is identified as a potential source of interference. Simultaneously, the spatial interference risk between devices is predicted by real-time tracking of the movement trajectory of mobile devices. The logical reasoning includes: performing logical reasoning based on the analysis results of the temporal and spatial dimensions using a rule engine and knowledge graph to deduce potentially affected upstream and downstream devices, thus obtaining the analysis results for the logical reasoning dimension. Model-based intelligent diagnosis involves processing the collected multi-source data using a spatiotemporal sequence network, which integrates CNN and LSTM network structures, including sequentially connected convolutional layers, pooling layers, and LSTM. The system consists of a LSTM layer and a fully connected layer; the LSTM layer also incorporates an attention mechanism to weight time steps; real-time collected device operation data is input into the trained spatiotemporal sequence network to obtain the device's current health status score and remaining service life prediction value; tiered early warning: based on the analysis results of the temporal dimension, spatial dimension, and logical reasoning dimension, the severity of abnormalities in the monitored data, the causes of the abnormalities, and the affected upstream and downstream devices are generated as early warning information; and predictive maintenance early warning information is generated based on the health status score and remaining service life prediction value.

2. The method for safety monitoring of construction power supply and special equipment according to claim 1, characterized in that, Before the multi-dimensional collaborative correlation analysis step, the following steps are also included: collecting multi-source heterogeneous data: real-time collection of operational status data of construction power systems and special equipment through sensor units deployed on various monitoring objects at the construction site; transmitting the collected multi-source heterogeneous data to a cloud platform or central server via network; data cleaning and spatiotemporal alignment: for the collected multi-source heterogeneous data, removing obvious outliers and invalid data caused by sensor failures or transmission interference; using linear interpolation, spline interpolation, or time series-based prediction and imputation algorithms to repair the data and ensure data continuity; using Kalman filtering for real-time signal smoothing and wavelet transform for multi-scale noise separation; using a time synchronization protocol to assign a unified millisecond-level precision timestamp to all heterogeneous data sources and establish a spatiotemporal correlation mapping table, binding the special equipment operation data with the equipment spatial coordinates to form a multi-source heterogeneous data set with spatiotemporal labels.

3. The method for monitoring the safety of construction power supply and special equipment according to claim 1, characterized in that, The training process of the spatiotemporal sequence network includes: training the spatiotemporal sequence network using special equipment operation status data and corresponding historical health status data, using mean squared error as the loss function, optimizing network parameters through backpropagation algorithm, using early stopping strategy to prevent overfitting, and using cross-validation method to evaluate model performance.

4. The method for monitoring the safety of construction power supply and special equipment according to claim 1, characterized in that, The logical reasoning specifically includes: constructing a reasoning model based on a rule engine and a knowledge graph, wherein the rule engine includes expert knowledge and safety procedures; using the knowledge graph to model the power supply and consumption relationship network at the construction site, representing the topological connection relationship and electrical characteristics between "distribution box-power supply line-electrical equipment"; after identifying potential causal relationships through time-series analysis and spatial analysis, tracing the impact path of events in the knowledge graph, and deducing the affected upstream and downstream equipment, so as to achieve a comprehensive assessment from single-point events to system-level impacts.

5. The method for safety monitoring of construction power supply and special equipment according to claim 1, characterized in that, The convolutional layer, pooling layer, and LSTM layer of the spatiotemporal sequence network are configured for multi-channel parallel processing. The outputs of multiple channels are fused using a concatenate operation and then fed into the fully connected layer. The fully connected layer has a linear activation function for outputting the prediction result. The computation process of the spatiotemporal sequence network specifically includes: the input data first passes through a convolutional layer to extract features; the output of the convolutional layer is further processed by a pooling layer to reduce the dimensionality of the data by extracting the maximum value in each region; the dimensionality-reduced feature sequence is input into an LSTM network for temporal modeling; the output of the LSTM layer is sent to a fully connected layer, and the signal features are classified by the Softmax activation function to output the prediction result; during the model training process, the Adam optimization algorithm is used for error backpropagation to update the network parameters layer by layer.

6. The method for monitoring the safety of construction power supply and special equipment according to any one of claims 1 to 5, characterized in that, The electrical parameters include: three-phase voltage, current, power, leakage current, and cable temperature of each level of distribution box.

7. The method for safety monitoring of construction power supply and special equipment according to any one of claims 1 to 5, characterized in that, The special equipment includes: tower cranes, construction hoists, bridge building machines, cantilever beam building machines, and mobile formwork; the operating status data of the special equipment includes: lifting capacity, radius, height, torque, wind speed, tilt angle, running speed, and door lock status of tower cranes and construction hoists; stress and strain of key structural parts, overall posture and deflection deformation, hydraulic system oil pressure and temperature, synchronization of traveling mechanism, and outrigger pressure of bridge building machines, cantilever beam building machines, and mobile formwork.

8. A construction power supply and special equipment safety monitoring system, characterized in that, include: Memory, used to store program instructions; A processor is configured to invoke the program instructions stored in the memory to implement the construction power supply and special equipment safety monitoring method as described in any one of claims 1 to 7.

Images