Power distribution room monitoring method and device based on power distribution multi-element data

Abstract

The application relates to the technical field of power distribution room monitoring, and discloses a power distribution room monitoring method and device based on power distribution multi-element data. The method comprises the following steps: performing time sequence synchronization marking on multi-source data in a power distribution room to establish an association table, generating an environment-equipment correlation degree index through covariance calculation, cross-analyzing personnel safety characteristic parameters to construct a multi-element risk association graph, performing weight distribution on risk nodes to generate a comprehensive risk assessment grade, and starting corresponding environmental regulation equipment or sound-light alarm equipment protection measures according to the assessment grade for hierarchical response control. The application solves the technical problems of multi-element data isolated analysis and missing linkage control in the existing power distribution room monitoring technology, and improves the intelligent degree and abnormal response efficiency of power distribution room operation and maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution room monitoring, and in particular to a power distribution room monitoring method and device based on power distribution multi-element data. BACKGROUND

[0002] The existing power distribution room monitoring technology is mainly based on single-dimensional data collection and analysis methods, including environmental monitoring systems, equipment state detection systems, and video monitoring devices, and other independent monitoring schemes. The environmental monitoring system collects environmental parameters of the power distribution room through the deployment of temperature and humidity sensors, water immersion sensors, smoke alarms, and other sensing devices, and transmits data to the monitoring center through ZigBee wireless communication or wired network, and triggers an alarm when the environmental parameters exceed the preset threshold. In terms of equipment state detection, it mainly relies on a patrol robot system equipped with an infrared thermal imager and a high-definition camera to periodically patrol the power distribution equipment, and reads instrument values and judges the equipment operating state through image recognition technology. Personnel safety monitoring uses a behavior recognition system based on computer vision to monitor the operating behavior and safety protection state of workers.

[0003] However, various monitoring devices operate independently, and environmental data, equipment data, and personnel data are in an information island state, lacking effective data fusion mechanisms and being unable to establish a correlation between environmental changes and equipment abnormalities and personnel behavior. Secondly, the intelligent analysis capability of the existing system is weak, most of which only have simple threshold alarm functions, lack deep learning and prediction analysis capabilities, and still require manual recording of instrument readings. The personnel behavior recognition accuracy is generally low, and it is unable to perform fault prediction and trend analysis. In addition, the monitoring, analysis, and control execution links are mutually fragmented, and when an anomaly is detected, manual intervention is required for decision-making and operation, resulting in long response delays and easy expansion of faults.

[0004] Due to the complex interrelationship between environmental factors, equipment states, and personnel behavior, single-dimensional monitoring methods cannot capture these potential correlation patterns, resulting in incomplete perception of the overall operation state of the power distribution room. Further, the lack of effective multi-element data space-time synchronization mechanisms makes it impossible to correlate and analyze data from different sensors and monitoring devices under a unified time reference, thereby failing to identify abnormal patterns and risk propagation paths across data sources. More critically, the existing technology lacks an intelligent decision-making mechanism based on multi-element data fusion, and is unable to automatically generate control strategies based on the comprehensive state of the environment-equipment-personnel, which directly affects the intelligent level of power distribution room operation and maintenance and the efficiency of emergency response. SUMMARY

[0005] The application provides a power distribution room monitoring method and device based on power distribution multi-element data, and aims to solve the technical problem of missing multi-element data isolated analysis and linkage control in the existing power distribution room monitoring technology, and improve the intelligent degree and abnormal response efficiency of power distribution room operation and maintenance.

[0006] In a first aspect, the application provides a power distribution room monitoring method based on power distribution multi-element data, which comprises the following steps:

[0007] Step S1: synchronously marking the environment sensor data, equipment image data and personnel behavior data in the power distribution room according to the time sequence correlation, and establishing a multi-source data space-time correlation table;

[0008] Step S2: based on the time overlap relationship between different data types in the multi-source data space-time correlation table, generating an environment-equipment correlation degree index by calculating the covariance of the environment parameter change trend and the equipment state fluctuation trend, which is used to reflect the influence degree of environmental factors on the equipment;

[0009] Step S3: cross analyzing the environment-equipment correlation degree index and the safety feature parameters extracted from the personnel behavior image, and constructing a multi-element risk correlation graph;

[0010] Step S4: weight distribution of the risk nodes in the multi-element risk correlation graph, when the environment anomaly and the equipment failure occur at the same time, the composite risk level is increased by one level, when the personnel irregular behavior and the environment anomaly occur at the same time, the safety early warning state is triggered, and the comprehensive risk assessment level is generated;

[0011] Step S5: hierarchical response control according to the comprehensive risk assessment level, when the assessment level is mild risk, the environment adjusting equipment is started, when the assessment level reaches severe risk, the sound and light alarm and the equipment protection measures are activated at the same time.

[0012] In a second aspect, the application provides a power distribution room monitoring device based on power distribution multi-element data, which comprises the following steps:

[0013] The marking module is used for synchronously marking the environment sensor data, equipment image data and personnel behavior data in the power distribution room according to the time sequence correlation, and establishing a multi-source data space-time correlation table;

[0014] The generating module is used for generating an environment-equipment correlation degree index by calculating the covariance of the environment parameter change trend and the equipment state fluctuation trend based on the time overlap relationship between different data types in the multi-source data space-time correlation table, which is used to reflect the influence degree of environmental factors on the equipment;

[0015] an analysis module configured to cross-analyze the environment-equipment correlation index and a safety feature parameter extracted from a personnel behavior image to construct a multi-element risk correlation graph;

[0016] a distribution module configured to distribute weights to risk nodes in the multi-element risk correlation graph, to increase a composite risk level by one level when an environment anomaly and a device failure occur simultaneously, and to trigger a safety early warning state when a personnel rule violation and an environment anomaly occur concurrently to generate a comprehensive risk assessment level;

[0017] a start module configured to perform hierarchical response control according to the comprehensive risk assessment level, to start an environment adjustment device when the assessment level is a mild risk, and to simultaneously activate an audible and visual alarm and a device protection measure when the assessment level reaches a severe risk.

[0018] In a third aspect, a power distribution room monitoring device based on power distribution multi-element data is provided, comprising a memory and at least one processor, the memory storing instructions; the at least one processor calling the instructions in the memory to cause the power distribution room monitoring device based on power distribution multi-element data to perform the power distribution room monitoring method based on power distribution multi-element data described above.

[0019] In a fourth aspect, a computer readable storage medium is provided, the computer readable storage medium storing instructions, when executed on a computer, causing the computer to perform the power distribution room monitoring method based on power distribution multi-element data described above.

[0020] In the technical solution provided in the present application, through the establishment of the multi-source data space-time correlation table, the technical defects of independent environment data, device data and personnel data in the traditional monitoring technology are fundamentally solved, the unified time reference synchronization and space location correlation of heterogeneous data sources are realized, and a solid data foundation is laid for subsequent in-depth analysis. The generation of the environment-equipment correlation index establishes a quantitative correlation between the changes of environment parameters and the fluctuations of device states through covariance calculation processing, breaks through the limitation of the prior art that can only perform single-dimensional monitoring, can accurately identify the influence degree of environmental factors on device operation, and significantly improves the accuracy and timeliness of power distribution room anomaly warning. The construction of the multi-element risk correlation graph adopts a graph structure data model to systematically model environment nodes, device nodes and personnel nodes and their mutual connection relationships, realizes visual analysis of risk propagation paths and quantitative evaluation of influence ranges, and compared with the traditional independent alarm mode, can more comprehensively master the overall risk state of the power distribution room. The generation of the comprehensive risk assessment level establishes an associated early warning mechanism of environment anomalies, device failures and personnel rule violations through weighted fusion and concurrent detection analysis of multi-dimensional risk factors, solves the technical blind spot that the prior art cannot identify composite risks and cross-influences, and greatly improves the intelligent level and risk identification ability of power distribution room safety monitoring.

[0021] The application of the covariance calculation algorithm in the environment-device correlation analysis realizes the technical leap from qualitative observation to quantitative evaluation through statistical correlation analysis of the trend of environmental parameter changes and the trend of device state fluctuations, provides scientific data support for power distribution room operation and maintenance decision, and has higher objectivity and accuracy compared with the traditional experience judgment method. The application of the graph attention network in the construction of a multi-element risk correlation graph realizes intelligent modeling of a complex risk network and adaptive analysis of a propagation path through dynamic calculation of the correlation weight between different risk nodes by the attention mechanism, significantly enhances the perception ability and prediction accuracy of the power distribution room monitoring device for potential risks. The threshold matching algorithm and parallel decomposition processing mechanism in the hierarchical response control strategy convert the traditional artificial decision-making process into an automated intelligent control process, accurately generates and executes the environmental regulation control instruction and the composite emergency control instruction, realizes the full-automation closed-loop control of "monitoring-analysis-decision-execution", and fundamentally solves the technical problems of the existing technology, such as the disconnection between monitoring and control, and long response delay, shortens the abnormal response time from the traditional minutes level to the second level, and significantly improves the emergency disposal ability and system reliability of the power distribution room operation and maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on these drawings.

[0023] Figure 1 An embodiment schematic diagram of the power distribution room monitoring method based on power distribution multi-element data in the embodiment of the present application;

[0024] Figure 2 An embodiment schematic diagram of the power distribution room monitoring device based on power distribution multi-element data in the embodiment of the present application;

[0025] Figure 3 An embodiment schematic diagram of the power distribution room monitoring device based on power distribution multi-element data in the embodiment of the present application; DETAILED DESCRIPTION

[0026] The embodiment of the present application provides a power distribution room monitoring method and device based on power distribution multi-element data. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0027] For ease of understanding, the specific process of the embodiment of the present application is described below. Please refer to Figure 1 One embodiment of the power distribution room monitoring method based on power distribution multi-element data in the embodiment of the present application comprises the following steps.

[0028] Step S1: synchronously mark the environment sensor data, equipment image data and personnel behavior data in the power distribution room according to the time sequence correlation, and establish a multi-source data space-time correlation table;

[0029] Step S2: based on the time overlap relationship between different data types in the multi-source data space-time correlation table, the covariance of the environment parameter change trend and the equipment state fluctuation trend is calculated, the environment-equipment correlation degree index is generated, and the influence degree of the environment factor on the equipment is reflected;

[0030] Step S3: cross-analyzing the environment-equipment correlation degree index and the safety feature parameters extracted from the personnel behavior image, and constructing a multi-element risk correlation graph;

[0031] Step S4: weight distribution is performed on the risk nodes in the multi-element risk correlation graph, the composite risk level is increased by one level when the environment anomaly and the equipment failure occur at the same time, and the safety early warning state is triggered when the personnel irregular behavior and the environment anomaly occur at the same time, and the comprehensive risk assessment level is generated;

[0032] Step S5: hierarchical response control is performed according to the comprehensive risk assessment level, the environment adjusting equipment is started when the assessment level is mild risk, and the sound and light alarm and the equipment protection measures are activated at the same time when the assessment level reaches severe risk.

[0033] It can be understood that the execution subject of the present application can be a power distribution room monitoring device based on power distribution multi-element data, and can also be a terminal or a server, and the specific place is not limited. The embodiment of the present application takes the server as the execution subject for example.

[0034] Specifically, the multi-source heterogeneous data in the power distribution room is uniformly marked in space and time. The temperature and humidity sensor SHT31 collects environmental data every 30 seconds and adds a millisecond timestamp. The water immersion sensor JXBS-3001 records the state transition time immediately when detecting a change in the on-off signal. The 400 million pixel camera shoots images every 5 minutes in H.265 encoding format and embeds accurate time information. Through a unified time reference, the three types of data are integrated to form a multi-source data space-time correlation table, which contains data source identifiers, collection time stamps, and sensor spatial coordinate information.

[0035] Based on the established space-time correlation table, the correlation analysis of the environment and the device state is carried out. The change rate parameter is calculated by extracting the temperature and humidity time series changes in the environmental sensor data stream. The pointer angle change of the pointer instrument in the device image data is detected by Hough transform, the reading fluctuation of the digital instrument is extracted by OCR character recognition, and the state switching frequency of the indicator light is obtained by HSV color space analysis. The covariance calculation is performed between the environmental parameter change trend curve and the device state fluctuation trend curve. The Pearson correlation coefficient method is used for covariance calculation. When the covariance value exceeds the preset correlation threshold of 0.6, it is confirmed that there is a significant causal correlation between the environmental factors and the device state, and a quantitative environment-device correlation index is generated.

[0036] The 17 key points of the personnel behavior image are detected by the OpenPose algorithm, and the feature coordinates of the head region and the torso region are extracted to form personnel body structure feature data. The safety hat wearing state is judged based on the head region features through the CNN network and the wearing probability value is calculated. The workwear wearing standard is recognized by analyzing the color consistency and texture features of the torso region, and a standard degree score is generated. The environment-device correlation index is decomposed into three-dimensional components according to the temperature influence factor, the humidity influence factor and the device failure factor. The safety feature parameters are constructed into a two-dimensional vector containing the safety hat wearing probability and the workwear standard degree. The two vectors are subjected to tensor product operation to generate an environment-personnel correlation matrix. The elements in the matrix that exceed the risk judgment threshold are marked as risk correlation pairs. Through graph theory connectivity analysis, the strongly connected risk nodes are clustered to form a risk area. Finally, a multi-element risk correlation graph is constructed, which includes environmental nodes, device nodes, personnel nodes and their mutual connection relationships.

[0037] The weight distribution of each risk node in the multi-element risk correlation graph is performed, the contribution of each node to the overall risk is calculated by analyzing the risk propagation path of the environment node, the device node and the personnel node, the node importance weight coefficient is generated, the weight coefficient is multiplied with the current risk value of the corresponding node, when the environment abnormal node and the device failure node are activated at the same time, the risk multiplication result is upgraded according to the risk amplification factor, when the personnel violation behavior node and the environment abnormal event are concurrent in the preset time window, the safety warning state flag is triggered, the risk type, severity and response priority are comprehensively considered to generate a comprehensive risk assessment level.

[0038] According to the comprehensive risk assessment level, the hierarchical response control is executed, when the risk level is mild, the environment adjustment control instruction is generated and transmitted to the exhaust equipment control module, the exhaust power and duration parameters are calculated according to the environment abnormality degree, when the risk level reaches severe, the composite emergency control instruction is generated, the 110dB alarm activation signal is sent to the sound and light alarm module and the power-off protection signal is sent to the device protection module through parallel decomposition processing, and the action state is recorded to form the linkage control execution record after the control execution is completed.

[0039] In a specific embodiment, step S1 comprises:

[0040] The temperature data and humidity data collected by the temperature and humidity sensor are time-stamped and processed to obtain environment sensor data stream;

[0041] The water immersion state conversion event is detected based on the on-off signal change of the water immersion sensor, the water immersion state conversion event is time-synchronized and aligned with the environment sensor data stream, and the environment state synchronous data set is obtained;

[0042] The image collected by the camera is time-stamped and embedded according to the encoding format, combined with the time reference in the environment state synchronous data set, a multi-source data space-time correlation table containing data source identification, collection time and sensor coordinates is generated.

[0043] Specifically, the unified time reference mark of various sensor data is performed through the built-in clock module of the Rui-Kai micro RK3568 chip, the temperature and humidity sensor SHT31 is connected with the chip through the I2C bus, and the real-time clock RTC module in the chip immediately generates a complete time stamp containing year, month, day, hour, minute, second and millisecond when receiving the sensor data. The time stamp format is 64-bit binary data structure, including 32-bit second-level time and 32-bit millisecond-level precision, each temperature data point and humidity data point is encapsulated into a data packet containing numerical value, time stamp and sensor ID, forming a continuous environment sensor data stream. The data stream is stored in the DDR4 memory buffer area of the chip in chronological order, and the buffer area uses a circular queue structure to manage data writing and reading operations.

[0044] The switch signal change detection of the water immersion sensor JXBS-3001 is triggered by the level change of the GPIO pin. When the sensor detects water immersion, it outputs a low-level signal. When the chip detects that the GPIO5 pin jumps from high level to low level, it triggers an edge detection interrupt. The interrupt service program immediately records the current time as the occurrence time of the water immersion state transition event. Time synchronization alignment processing is to match and find the timestamp in the environmental sensor data stream. Through the binary search algorithm, the closest data point is located in the environmental sensor data stream. The water immersion state transition event is inserted into the corresponding time position. The environmental state synchronization data set containing temperature data, humidity data and water immersion state is formed. Each data record in the synchronization data set contains a unified time reference, data type identification and value information.

[0045] The timestamp embedding processing of the camera image is to write the time information directly into the metadata area of the video frame in the H.265 encoding process. The chip reads the current system timestamp every time the camera takes an image. The timestamp is converted into the time code format defined in the H.265 standard. The time code contains frame number and absolute time information. The encoder embeds the time code into the header field of each I frame and P frame when generating the compressed video stream. The accurate shooting time can be directly extracted from the video stream during decoding. The time reference combination is to calibrate and compare the image timestamp with the unified time reference established in the environmental state synchronization data set. Through the calculation of time deviation and the application of linear interpolation algorithm, the clock drift between different data sources is eliminated. The generation of multi-source data space correlation table is to merge and sort the calibrated image data and environmental state synchronization data set according to the time dimension. The time index structure is established to facilitate the fast search of all data types at a specific time.

[0046] The data source identification is to assign a unique device ID number to each type of sensor. The temperature and humidity sensor is identified as ENV_01, the water immersion sensor is identified as FLOOD_01, and the camera is identified as CAM_01. The collection time records the absolute time information of data generation. The sensor coordinates record the three-dimensional spatial position information of each sensor in the power room. The coordinate system is established with the southwest corner of the power room as the origin. The temperature and humidity sensor is installed on the wall surface 2 meters high from the ground. The water immersion sensor is installed at the lowest point of the ground. The camera is installed at a height of 3 meters from the ground in the northeast corner of the room. The combination of spatial coordinates and timestamps forms a four-dimensional space-time coordinate system. The space-time correlation table is stored in a relational database structure. The primary key is the timestamp, and the foreign key is the device ID. The data field includes sensor value, data type, quality identification and other attribute information.

[0047] In a specific embodiment, step S2 comprises:

[0048] The temperature change rate and the humidity change rate are calculated based on the time sequence change of the environmental sensor data stream in the multi-source data space-time association table, and an environmental parameter change trend curve is obtained.

[0049] The instrument reading and the indicator light state in the equipment image data are subjected to time sequence tracking analysis, the instrument value fluctuation amplitude and the indicator light state switching frequency are calculated, and an equipment state fluctuation trend curve is obtained.

[0050] The environmental parameter change trend curve and the equipment state fluctuation trend curve are subjected to covariance calculation processing, and when the covariance value exceeds a preset association threshold, it is determined that there is a causal association relationship, and an environment-equipment association degree index is generated.

[0051] Specifically, the temperature change rate is calculated by using the temperature difference between adjacent time points divided by the time interval, and the specific process is to extract the temperature sequence from the association table in chronological order, calculate the difference between the current time temperature value and the previous time temperature value, and then divide the difference by the time interval between the two time points to obtain the temperature change per second. The humidity change rate is calculated by using the same difference calculation method, and the continuous change rate values are arranged in chronological order to form a change rate time sequence array. The moving average filter algorithm is applied to the change rate array to remove noise interference, and the moving average window is set to 5 data points. The filtered value of each data point is equal to the arithmetic mean of 5 values including the current point and the previous and next 2 points. The filtered change rate data constitutes the environmental parameter change trend curve, which is represented in a two-dimensional coordinate system with time as the horizontal axis and change rate as the vertical axis.

[0052] The equipment state fluctuation trend analysis detects the pointer angle of the instrument image by using the Hough transform algorithm. The Hough transform converts the straight line detection problem in the image space into the peak detection problem in the parameter space. The algorithm starts from the edge pixel points of the instrument image, calculates all the straight line parameters that can be formed by each pixel point, including the distance and angle of the straight line, divides the parameter space into grids, and accumulates the votes for each grid. The grid with the highest number of votes corresponds to the most likely pointer straight line. The angle value of the pointer in the image coordinate system is obtained by inverse transformation. The angle change rate is obtained by angle difference calculation on the pointer angle sequence of continuous multiple frames. The angle change rate reflects the fluctuation degree of the instrument reading. The indicator light state detection is realized by HSV color space analysis. The HSV color space decomposes color information into hue H, saturation S, and brightness V. The algorithm extracts the average HSV value of the indicator light region, judges the red and green states of the indicator light through the preset color threshold, compares the states between continuous frames to obtain the state switching event sequence, and calculates the state switching frequency by counting the number of switching times per unit time. The instrument value fluctuation amplitude and the indicator light state switching frequency are combined to form the equipment state fluctuation trend curve.

[0053] The covariance calculation process is a statistical method for measuring the linear correlation degree between the change trend of the environmental parameter and the fluctuation trend of the device state. The covariance calculation requires that the two variable sequences have the same time length and sampling interval. The algorithm first aligns the environmental parameter change trend curve and the device state fluctuation trend curve in time, and unifies the two curves to the same time sampling point through linear interpolation method. The covariance calculation formula involves the mean value calculation and deviation product summation of the two sequences. The specific calculation process is to calculate the arithmetic mean value of the environmental change rate sequence and the device fluctuation sequence respectively, then calculate the deviation of the values of the two sequences at each time point from the respective mean values, multiply the deviations at the corresponding time points and sum them up, and finally divide the covariance value by the sequence length minus one. The positive and negative and size of the covariance value reflect the correlation strength and direction of the two sequences. The preset correlation threshold is determined according to the statistical analysis of the historical data of the power distribution room. When the calculated covariance value exceeds the threshold, it indicates that there is a significant causal relationship between the environmental change and the device state. The environment-device correlation degree index is the standardized value of the covariance value after normalization processing. The normalization processing maps the covariance value to the interval range of zero to one, which is convenient for subsequent correlation analysis and risk assessment calculation.

[0054] In a specific embodiment, step S3 comprises:

[0055] The personnel behavior image is subjected to key point detection processing to extract feature coordinates of the head region and the torso region of the personnel, and to obtain personnel body structure feature data.

[0056] Based on the personnel body structure feature data, safety hat wearing state recognition and work clothes wearing state recognition are performed, a safety hat wearing probability value and a work clothes standard degree score are calculated, and safety feature parameters are obtained.

[0057] The environment-device correlation degree index is subjected to matrix cross operation with the safety feature parameters as an environmental influence factor. When the environmental influence factor exceeds the correlation threshold and the safety feature parameters are lower than the safety standard, a high-risk node is marked, and a multi-element risk correlation graph containing environmental nodes, device nodes, personnel nodes and their mutual connection relationships is generated.

[0058] Specifically, the OpenPose algorithm is used to locate the body key points of the personnel image captured by the camera. The OpenPose algorithm extracts multi-scale features of the image through a convolutional neural network. The network structure includes a feature extraction layer and a key point detection layer. The feature extraction layer uses VGG-19 as the backbone network to extract deep feature maps of the image. The key point detection layer locates and connects the human body key points through two branches of partial affinity field and key point heat map. The algorithm first detects 17 human body key points in the image, including the top of the head, the neck, the left and right shoulders, the left and right elbows, the left and right wrists, the left and right hips, the left and right knees, and the left and right ankles. Each key point is represented by a pixel coordinate in the image coordinate system. The head region feature coordinates are determined by the boundary box formed by the top of the head and the neck. The left upper corner coordinate of the boundary box is offset by 30 pixels to the left and up from the top of the head coordinate, and the right lower corner coordinate is offset by 20 pixels to the right and down from the neck coordinate. The torso region feature coordinates are determined by the polygon region formed by the neck, the left and right shoulders, and the left and right hips. The polygon vertex coordinates are sequentially connected to form the torso contour. The personnel body structure feature data includes the two-dimensional coordinates of each key point, the confidence score, and the connection relationship between the key points. The data structure is stored in JSON format for subsequent processing.

[0059] The safety helmet wearing state recognition is based on local image analysis of head region feature coordinates. The algorithm crops the sub-image of the head region from the original image. The size of the sub-image is dynamically adjusted according to the size of the head boundary box. The contrast of the sub-image is enhanced through histogram equalization. The texture features of the head region are extracted using Haar feature descriptors. Haar features describe local texture patterns through the weighted sum and difference of pixel values within a rectangular box. The extracted Haar feature vector is input into a support vector machine classifier for safety helmet detection. The support vector machine learns the feature boundary between safety helmet and non-safety helmet images through training samples. The classifier outputs the probability distribution of the two categories of wearing and not wearing. The safety helmet wearing probability value is the probability value of the wearing category. The workwear wearing state recognition is achieved by analyzing the color features and texture features of the torso region. The algorithm extracts the RGB color histogram of the torso region image, calculates the color saturation and color consistency of the dominant color, and calculates the texture contrast and homogeneity of the torso region through the gray level co-occurrence matrix. The contrast reflects the clarity of the texture, and the homogeneity reflects the uniformity of the texture. The workwear standardization score is calculated by the weighted average of the color matching degree and the texture features. The safety feature parameters include the safety helmet wearing probability value and the workwear standardization score.

[0060] The matrix cross operation of the environmental influence factor and the safety feature parameter first decomposes the environment-equipment correlation index into three components of a temperature influence factor, a humidity influence factor and a device failure factor. The decomposition process is realized by a principal component analysis method. The principal component analysis projects the correlation index onto three principal component directions. Each principal component corresponds to an environmental influence dimension. The temperature influence factor reflects the influence degree of temperature change on the equipment and personnel. The humidity influence factor reflects the influence degree of humidity change. The device failure factor reflects the severity of the abnormal state of the equipment. The three influence factors form a three-dimensional vector. The safety feature parameter forms a two-dimensional vector. The matrix cross operation multiplies the three-dimensional vector and the two-dimensional vector through a tensor product operation to generate a six-dimensional environment-personnel correlation matrix. Each element in the matrix represents the correlation strength of a specific environmental factor and a specific safety feature. The high-risk node marking is realized through a double threshold value judgment. When any component of the environmental influence factor exceeds a preset correlation threshold value and the corresponding safety feature parameter is lower than a safety standard, the combination of the environmental factor and the safety feature is marked as a high-risk node. The multi-element risk correlation graph is represented by a directed graph data structure. The nodes in the graph are divided into three categories of environment nodes, device nodes and personnel nodes. The environment nodes store environmental parameters such as temperature and humidity and change trends. The device nodes store device information such as instrument reading and indicator light state. The personnel nodes store safety feature parameters and behavior risk assessment. The connection edges between the nodes represent the risk propagation path. The weight value of the edge represents the correlation strength. The graph structure supports risk tracing and influence range analysis.

[0061] In a specific embodiment, the matrix cross operation of the environment-equipment correlation index as an environmental influence factor and a safety feature parameter includes:

[0062] The environment-equipment correlation index is decomposed into components according to the temperature influence factor, the humidity influence factor and the device failure factor to obtain a three-dimensional environmental influence factor vector.

[0063] A two-dimensional safety state vector is constructed based on the safety hat wearing probability value and the work clothes specification score in the safety feature parameter. The three-dimensional environmental influence factor vector and the two-dimensional safety state vector are subjected to tensor product operation to obtain an environment-personnel correlation matrix.

[0064] Each matrix element in the environment-personnel correlation matrix is subjected to threshold value judgment processing. When the matrix element value exceeds a risk judgment threshold value, the corresponding environmental factor and personnel factor combination is marked as a risk correlation pair to obtain a primary risk node set.

[0065] The mutual influence strength between the nodes is calculated based on the spatial proximity of the risk correlation pairs in the primary risk node set. The risk nodes with strong connectivity are clustered to form a risk area through graph theory connectivity analysis to obtain a risk node grouping result.

[0066] According to the risk node grouping result, identifiers of environment nodes, device nodes and personnel nodes are assigned to each risk area, directed connection edges and their weight values between nodes are established, and a multi-element risk association graph containing node attributes, connection relationships and propagation paths is generated.

[0067] Specifically, the composite correlation degree index is decomposed into independent influence factor components using a principal component analysis algorithm. The principal component analysis projects the original correlation degree data onto three main variance directions through eigenvalue decomposition. The temperature influence factor corresponds to the principal component direction of the correlation between temperature change and device state, the humidity influence factor corresponds to the principal component direction of the correlation between humidity change and device abnormality, and the device failure factor corresponds to the principal component direction of the device internal fault propagation. The decomposition process first calculates the covariance matrix of the correlation degree index. The weight coefficients of the three principal components are determined by solving the eigenvalues and eigenvectors of the covariance matrix. The original correlation degree values are multiplied by the corresponding weight coefficients to obtain the values of the three components. The three-dimensional environmental influence factor vector stores the values of the three components in array form. Each element of the vector represents the contribution of a specific environmental factor to the overall risk of the power distribution room.

[0068] The two-dimensional safety state vector is constructed by combining the safety hat wearing probability value and the work clothes standardization score into a two-element vector. The safety hat wearing probability value is obtained by analyzing the head region features using a support vector machine classifier. The value ranges from zero to one, with a value close to one indicating that the safety hat is confirmed to be worn, and a value close to zero indicating that the safety hat is not worn. The work clothes standardization score is calculated by analyzing the color matching degree and texture features of the torso region. The score is also standardized to a range of zero to one, with a higher value indicating that the work clothes are more standardized. The tensor product operation cross-multiplies each element of the three-dimensional environmental influence factor vector with each element of the two-dimensional safety state vector. The specific calculation process is that the temperature influence factor is multiplied by the safety hat wearing probability and the work clothes standardization, respectively. The humidity influence factor is multiplied by the two safety parameters, respectively. The device failure factor is also multiplied by the two safety parameters, respectively. A total of six product results are generated. These six values are arranged in the order of temperature-safety hat, temperature-work clothes, humidity-safety hat, humidity-work clothes, device failure-safety hat, and device failure-work clothes to form a six-element environmental-personnel correlation matrix. Each element in the matrix represents the interaction strength between a specific environmental factor and a specific personnel safety factor.

[0069] The threshold judgment process performs risk assessment on the six elements of the environment-personnel association matrix one by one, and the risk judgment threshold is determined according to historical statistical data and safety standards of the safe operation of the power distribution room. When the value of the matrix element exceeds the preset threshold, it indicates that the combination of the corresponding environmental factor and personnel factor constitutes a significant safety risk. The risk association pair is formed by combining and pairing the environmental factor and the personnel factor corresponding to the matrix element exceeding the threshold, such as a high-temperature environment and not wearing a safety helmet, a high-humidity environment and non-standard work clothes, etc. The primary risk node set contains all the marked risk association pairs and their corresponding risk intensity values. The set is stored in a hash table data structure for fast lookup and update operations.

[0070] The spatial proximity calculation is based on the spatial position information of the sensors and personnel in the power distribution room. The spatial distance between different risk nodes is calculated by the Euclidean distance formula. The spatial distance reflects the physical path and influence range of risk propagation. The mutual influence strength between nodes is quantified by the reciprocal of the distance. The closer the distance between risk nodes, the stronger the mutual influence relationship. The graph theory connectivity analysis uses a depth-first search algorithm to traverse the connection relationship between risk nodes. The algorithm starts from any risk node and recursively searches along the connection edges with an influence strength exceeding the threshold. All nodes visited during the search process are classified into the same connected component. Strongly connected relationship indicates that there is a bidirectional strong influence path between nodes. The risk area formed by clustering contains multiple risk nodes with close geographical location and close risk propagation relationship. The risk node grouping result is stored in a graph structure, and each group corresponds to a connected subgraph in the graph.

[0071] The risk node grouping result processing assigns a unique area identifier to each risk area. The environmental node identifier is prefixed with ENV and the area number. The device node identifier is prefixed with DEV. The personnel node identifier is prefixed with PER. The directed connection edge establishes a causal relationship based on risk propagation. The direction of the edge represents the direction of risk propagation, and the weight value represents the propagation strength. The weight calculation considers three factors: spatial distance, temporal correlation, and influence degree. The multi-element risk association graph is stored in an adjacency list data structure. The graph contains a node attribute table that records the type, location, state, and risk level of each node. The connection relationship table records the directed edges between nodes and their weights. The propagation path table records the propagation sequence and propagation delay of risk between different nodes.

[0072] In a specific embodiment, step S4 includes:

[0073] Based on the risk propagation path of the environmental nodes, device nodes, and personnel nodes in the multi-element risk association graph, the influence range of each node is analyzed, the contribution of each node to the overall risk of the system is calculated, and the node importance weight coefficient is obtained.

[0074] The node importance weight coefficient is multiplied by the current risk value of the corresponding node, and when the environmental abnormal node and the equipment failure node are activated at the same time, the risk product result is upgraded according to a risk amplification factor to obtain a composite risk level score.

[0075] The composite risk level score and the risk value of the personnel violation behavior node are concurrently detected and analyzed, and when it is detected that the personnel violation behavior and the environmental abnormal event occur at the same time within a preset time window, a safety warning state flag is triggered, and a comprehensive risk assessment level including a risk type, a severity and a response priority is generated.

[0076] Specifically, a graph traversal algorithm is used to perform a depth-first search on the multi-element risk correlation graph, the algorithm starts from each node and recursively traverses along the directed connection edges, records all downstream nodes that can be affected by the node, and the influence range is calculated by counting the number of reachable nodes and the propagation path length of each node. The number of reachable nodes reflects the direct and indirect influence breadth of the node, and the propagation path length reflects the time delay and attenuation degree of risk propagation. The contribution of the node to the overall risk of the system is calculated by using the PageRank algorithm. The PageRank algorithm regards the risk correlation graph as a directed graph, and calculates the importance score of each node by iteration. The core idea of the algorithm is that important nodes are pointed to by more important nodes. During the calculation process, the importance score of each node is equal to the weighted sum of the importance scores of all upstream nodes pointing to the node and the corresponding edge weight. After multiple iterations until the importance score converges, the final convergence value is taken as the node importance weight coefficient. The weight coefficient reflects the core position and influence of the node in the entire risk propagation network.

[0077] The weighted product operation multiplies the importance weight coefficient of each node with the current risk value of the node. The current risk value comes from the real-time monitoring data and state evaluation results of the node. The risk value of the environmental node is calculated based on the degree of temperature and humidity exceeding the standard, the risk value of the device node is calculated based on the degree of instrument abnormality and indicator light failure, and the risk value of the personnel node is calculated based on the degree of non-standard safety protection. The product result reflects the actual contribution of the node to the system risk in the current state. The risk amplification factor is a correction parameter designed for the nonlinear risk growth phenomenon when multiple key nodes are abnormal at the same time. When the activation state of the environmental abnormal node and the equipment failure node is true at the same time, the risk amplification mechanism is triggered. The amplification factor is dynamically calculated according to the type combination and severity of the abnormal node. The combination of high temperature environment and equipment overload has a higher amplification factor. The combination of high humidity environment and electrical equipment failure also has a significant amplification effect. The level upgrading process multiplies the original risk product result by the corresponding amplification factor, and then maps the continuous risk value to discrete risk levels according to the preset level division threshold. The composite risk level score adopts a five-level standard, and different numerical intervals correspond to different risk levels from low risk to extremely high risk.

[0078] The concurrent detection analysis detects the time overlap relationship between the personnel violation behavior and the environmental abnormal event through a time window matching algorithm. The time window is a preset time interval, usually set to several minutes to tens of minutes. The algorithm maintains a sliding time window queue to record all risk events occurring in the recent time period. When a new personnel violation behavior event arrives, the algorithm searches whether there is an environmental abnormal event in the current time window. The search process is matched by comparing the time stamp and event type of the event. When it is found that the personnel violation behavior event and the environmental abnormal event occur in the same time window, the concurrent detection flag is triggered. The safety warning state flag is a Boolean state variable. When the concurrent detection condition is met, the flag is set to true, indicating that the current power distribution room is in a high-risk state and needs to be responded immediately. The comprehensive risk assessment level generation process considers the composite risk level score and the concurrent detection result. The risk type identification is based on the node type combination that triggers the warning. The environmental-equipment composite risk, the environmental-personnel composite risk, the equipment-personnel composite risk, and the ternary composite risk correspond to different risk type identifiers. The severity assessment maps the composite risk level score to four severity levels: mild, moderate, severe, and extremely severe. The response priority is determined according to the combination of risk type and severity. The extremely severe ternary composite risk has the highest response priority and needs to start the emergency plan immediately. The moderate-severe dual composite risk has a high priority. The mild-moderate single risk has a medium-low priority.

[0079] In a specific embodiment, step S5 comprises:

[0080] Based on the risk type and severity in the comprehensive risk assessment level, the grading threshold matching process is performed. When the risk level is in the mild risk range, the environmental adjustment control instruction is generated. When the risk level is in the severe risk range, the composite emergency control instruction is generated, and the graded response control strategy is obtained.

[0081] The environmental adjustment control instruction in the graded response control strategy is input into the exhaust equipment control module for wind speed and running time adjustment processing. According to the environmental abnormality degree, the exhaust power parameter and the duration parameter are calculated, and the environmental equipment execution instruction is obtained.

[0082] The composite emergency control instruction in the graded response control strategy is processed in parallel, and the alarm activation signal is sent to the sound and light alarm module and the power-off protection signal is sent to the device protection module. When the device execution confirmation feedback is received, the control action completion state is recorded, and the linkage control execution record is generated.

[0083] Specifically, the multi-dimensional threshold judgment matrix is used to classify the comprehensive risk assessment level, the threshold judgment matrix is constructed according to the risk type and severity, the risk type dimension includes environmental abnormalities, equipment failures, personnel violations and various compound risks, the severity dimension includes four levels of mild, moderate, severe and extremely severe, and the matching process determines the corresponding control strategy type through the lookup table index. When the risk type of the comprehensive risk assessment level is single environmental abnormality and the severity is in the mild range, the threshold matching algorithm extracts the corresponding environmental regulation control instruction template from the lookup table, which includes the exhaust air start instruction, the running time parameter and the power regulation range. When the risk type is a multi-element compound risk and the severity reaches the severe range, the matching algorithm extracts the compound emergency control instruction template, which contains multiple parallel control actions. The hierarchical response control strategy is a structured control instruction set generated based on the threshold matching result, and the strategy data structure includes control type identifier, execution priority, parameter configuration and time constraint attribute fields.

[0084] The environmental regulation control instruction is input into the exhaust equipment control module for parameter analysis and calculation processing. The control module first extracts the target temperature and target humidity parameters in the instruction, then reads the real-time values of the current environmental sensor, calculates the deviation of the current temperature and humidity from the target values, and calculates the temperature deviation by subtracting the target temperature from the current temperature. The humidity deviation is calculated in the same way. The wind speed regulation processing is based on the PID control algorithm. The PID algorithm calculates the control output by weighted combination of proportional, integral and differential terms. The proportional term is proportional to the current deviation, the integral term is proportional to the cumulative deviation, and the differential term is proportional to the deviation rate. The weighted sum of the three terms is output as the control signal of the wind speed regulation to the frequency converter of the exhaust equipment. The running time regulation processing is dynamically calculated according to the environmental abnormality degree. The abnormality degree is quantified by the weighted average of the temperature and humidity deviation. The larger the deviation, the more serious the abnormality. The running time is positively correlated with the abnormality degree. The calculation process uses a piecewise linear function. Mild abnormalities correspond to shorter running times, and severe abnormalities correspond to longer running times. The exhaust power parameter calculation considers the power room space volume, current temperature and humidity, target temperature and humidity, and external environmental conditions. The power calculation uses the heat transfer equation of thermodynamics. The duration parameter is calculated based on the prediction model of the temperature and humidity change rate. The environmental equipment execution instruction contains specific device address, control parameter value and execution timing information.

[0085] The parallel decomposition processing of the composite emergency control instruction adopts a multi-thread task scheduling algorithm. The scheduling algorithm decomposes the composite instruction into multiple independent sub-tasks, each of which corresponds to a specific control module and an execution action. The sound and light alarm sub-task includes generation and transmission of an alarm activation signal. The signal adopts a digital pulse coding format, including alarm type coding, severity coding and duration coding. The alarm activation signal is output to the sound and light alarm module LTE-1101J through the GPIO pin. After receiving the signal, the module starts a 110 decibel alarm sound and a flashing warning light. The device protection sub-task generates a power-off protection signal and transmits it to the device protection module. The protection signal adopts a relay control protocol, including a list of devices to be powered off, a power-off sequence and a recovery condition. The protection module controls the corresponding circuit breaker according to the signal content to achieve safe power-off protection of key devices. The synchronous execution is realized through the time synchronization mechanism of the task scheduler. The scheduler maintains a unified time reference, and all sub-tasks are started and executed at the same time point. The device execution confirmation feedback is obtained through the state query mechanism. The control module periodically sends a state query instruction to the controlled device. The device returns the current running state and execution result. The control action completion state record includes the task identifier, the execution start time, the completion time, the execution result and the exception information. The linkage control execution record is stored in a time sequence database format, supporting fast query and statistical analysis based on time.

[0086] The power distribution room monitoring method based on power distribution multi-element data in the embodiments of the application is described above. The power distribution room monitoring device based on power distribution multi-element data in the embodiments of the application is described below. Please refer to Figure 2 The power distribution room monitoring device based on power distribution multi-element data in the embodiments of the application includes one embodiment:

[0087] The marking module is configured to synchronously mark the environment sensor data, device image data and personnel behavior data in the power distribution room according to the time sequence correlation, and establish a multi-source data space-time association table.

[0088] The generating module is configured to generate an environment-device correlation degree index by calculating the covariance of the environment parameter change trend and the device state fluctuation trend based on the time overlap relationship between different data types in the multi-source data space-time association table, and reflect the influence degree of environmental factors on the device.

[0089] The analysis module is configured to cross-analyze the environment-device correlation degree index and the safety feature parameters extracted from the personnel behavior image, and construct a multi-element risk association graph.

[0090] The distribution module is used for weight distribution of the risk nodes in the multi-element risk correlation graph, and when the environmental abnormality and the equipment failure occur simultaneously, the composite risk level is promoted by one level, when the personnel violation behavior and the environmental abnormality occur simultaneously, a safety early warning state is triggered, and a comprehensive risk assessment level is generated.

[0091] The starting module is used for hierarchical response control according to the comprehensive risk assessment level, and when the assessment level is a mild risk, the environmental regulation equipment is started, and when the assessment level reaches a severe risk, the sound and light alarm and the equipment protection measure are activated simultaneously.

[0092] The above Figure 2 The power distribution room monitoring device based on power distribution multi-element data in the embodiment of the application is described in detail from the perspective of modular functional entities, and the power distribution room monitoring device based on power distribution multi-element data in the embodiment of the application is described in detail from the perspective of hardware processing.

[0093] Referring Figure 3 The embodiment of the application also provides a power distribution room monitoring device based on power distribution multi-element data, which can be a server, and the internal structure thereof can be as shown in Figure 3 The power distribution room monitoring device based on power distribution multi-element data comprises a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer is used for providing calculation and control capabilities. The memory of the power distribution room monitoring device based on power distribution multi-element data comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the power distribution room monitoring device based on power distribution multi-element data is used for storing corresponding data in the embodiment. The network interface of the power distribution room monitoring device based on power distribution multi-element data is used for communicating with an external terminal through network connection. The computer program is executed by the processor to implement the above method.

[0094] Those skilled in the art can understand Figure 3 The structure shown in the embodiment of the application is only a block diagram of part of the structure related to the scheme of the application, and does not constitute a limitation on the power distribution room monitoring device based on power distribution multi-element data to which the scheme of the application is applied.

[0095] The application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium or a volatile computer readable storage medium. The computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer executes the steps of the power distribution room monitoring method based on power distribution multi-element data.

[0096] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0097] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a plurality of instructions for causing a power distribution room monitoring device based on power distribution multi-element data (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0098] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring a power distribution room based on multi-source power distribution data, characterized in that, The method includes: Step S1: Synchronize and label the environmental sensor data, equipment image data, and personnel behavior data in the power distribution room according to their temporal correlation, and establish a multi-source data spatiotemporal correlation table; Step S2: Based on the temporal overlap relationship between different data types in the multi-source data spatiotemporal correlation table, the covariance between the environmental parameter change trend and the equipment status fluctuation trend is calculated to generate an environment-equipment correlation index, which is used to reflect the degree of influence of environmental factors on equipment. Step S3: Cross-analyze the environment-equipment correlation index with the safety feature parameters extracted from personnel behavior images to construct a multivariate risk correlation map; Step S4: Assign weights to the risk nodes in the multivariate risk association graph. When environmental anomalies and equipment failures occur simultaneously, the composite risk level is increased by one level. When personnel violations and environmental anomalies occur concurrently, a safety warning state is triggered, and a comprehensive risk assessment level is generated. Step S5: Implement graded response control based on the comprehensive risk assessment level. When the assessment level is mild risk, activate the environmental control equipment. When the assessment level reaches severe risk, activate the audible and visual alarm and equipment protection measures simultaneously. Step S3 includes: The environment-equipment correlation index is used as an environmental impact factor and the safety feature parameter is used for matrix cross operation. When the environmental impact factor exceeds the correlation threshold and the safety feature parameter is lower than the safety standard, it is marked as a high-risk node, and the multivariate risk correlation map containing environmental nodes, equipment nodes and personnel nodes and their interconnection relationships is generated. The step of performing matrix cross-operation between the environment-equipment correlation index as an environmental impact factor and the safety characteristic parameters includes: The environment-equipment correlation index is decomposed into components based on temperature influence factor, humidity influence factor and equipment failure factor to obtain a three-dimensional environmental influence factor vector. A two-dimensional safety state vector is constructed based on the helmet wearing probability value and work uniform standardization score in the safety feature parameters. The three-dimensional environmental impact factor vector is multiplied with the two-dimensional safety state vector to obtain the environment-personnel correlation matrix. Each element in the environment-personnel correlation matrix is ​​subjected to threshold judgment processing. When the value of a matrix element exceeds the risk judgment threshold, the corresponding combination of environmental factors and personnel factors is marked as a risk correlation pair, thus obtaining a primary risk node set. Based on the spatial proximity of risk association pairs in the primary risk node set, the mutual influence strength between nodes is calculated. Then, risk nodes with strong connectivity are clustered into risk regions through graph theory connectivity analysis to obtain the risk node grouping results. Based on the risk node grouping results, identifiers for environmental nodes, equipment nodes, and personnel nodes are assigned to each risk area. Directed connection edges between nodes and their weight values ​​are established to generate the multivariate risk association graph containing node attributes, connection relationships, and propagation paths. Step S4 includes: Based on the risk propagation path analysis of environmental nodes, equipment nodes, and personnel nodes in the multivariate risk correlation graph, the influence range of each node is analyzed, the contribution of each node to the overall system risk is calculated, and the node importance weight coefficient is obtained. The importance weight coefficient of the node is multiplied by the current risk value of the corresponding node. When an environmental anomaly node and an equipment failure node are activated at the same time, their risk product results are upgraded according to the risk amplification factor to obtain a composite risk level score.

2. The method for monitoring a power distribution room based on multi-source power distribution data according to claim 1, characterized in that, Step S1 includes: The temperature and humidity data collected by the temperature and humidity sensors are timestamped to obtain the environmental sensor data stream. Based on the detection of water immersion state transition events by the change of switch signal of the water immersion sensor, the water immersion state transition events are time-synchronized and aligned with the data stream of the environmental sensor to obtain an environmental state synchronization dataset. The images captured by the camera are timestamped according to the encoding format, and combined with the time base in the environmental state synchronization dataset, a multi-source data spatiotemporal association table containing data source identifiers, acquisition times, and sensor coordinates is generated.

3. The method for monitoring a power distribution room based on multi-source power distribution data according to claim 1, characterized in that, Step S2 includes: Based on the temporal changes of environmental sensor data streams in the multi-source data spatiotemporal correlation table, the rate of temperature change and the rate of humidity change are calculated to obtain the trend curve of environmental parameter changes. Perform time-series tracking analysis on the instrument readings and indicator light status in the equipment image data, calculate the fluctuation amplitude of the instrument values ​​and the switching frequency of the indicator light status, and obtain the equipment status fluctuation trend curve. The covariance of the environmental parameter change trend curve and the equipment status fluctuation trend curve is calculated. When the covariance value exceeds the preset correlation threshold, a causal relationship is determined, and the environment-equipment correlation index is generated.

4. The method for monitoring a power distribution room based on multi-source power distribution data according to claim 1, characterized in that, Step S3 further includes: By performing key point detection processing on images of human behavior, feature coordinates of the head and torso regions are extracted to obtain human body structure feature data. Based on the personnel's body structure feature data, the helmet wearing status and work clothes wearing status are identified, the helmet wearing probability value and work clothes standardization score are calculated, and safety feature parameters are obtained.

5. The method for monitoring a power distribution room based on multi-source power distribution data according to claim 1, characterized in that, Step S4 further includes: The composite risk level score and the risk value of the personnel violation node are concurrently detected and analyzed. When personnel violation and environmental abnormal event are detected to occur simultaneously within a preset time window, a safety warning status flag is triggered, and the comprehensive risk assessment level containing risk type, severity and response priority is generated.

6. The method for monitoring a power distribution room based on multi-source power distribution data according to claim 1, characterized in that, Step S5 includes: Based on the risk type and severity in the comprehensive risk assessment level, a graded threshold matching process is performed. When the risk level is in the low risk range, an environmental regulation control instruction is generated. When the risk level is in the high risk range, a composite emergency control instruction is generated, resulting in a graded response control strategy. The environmental regulation control command in the graded response control strategy is input into the exhaust equipment control module for wind speed and running time adjustment. The exhaust power parameter and duration parameter are calculated according to the degree of environmental anomaly to obtain the environmental equipment execution command. The composite emergency control commands in the hierarchical response control strategy are decomposed in parallel, and alarm activation signals are sent to the audible and visual alarm module and power failure protection signals are sent to the equipment protection module simultaneously. When the equipment execution confirmation feedback is received, the control action completion status is recorded, and a linkage control execution record is generated.

7. A power distribution room monitoring device based on multi-source power distribution data, characterized in that, For implementing the power distribution room monitoring method based on multi-source power distribution data as described in any one of claims 1-6, the power distribution room monitoring device based on multi-source power distribution data comprises: The tagging module is used to synchronously tag environmental sensor data, equipment image data, and personnel behavior data in the power distribution room according to their temporal correlation, and to establish a spatiotemporal correlation table for multi-source data. The generation module is used to generate an environment-equipment correlation index based on the time overlap relationship between different data types in the multi-source data spatiotemporal correlation table, by calculating the covariance between the environmental parameter change trend and the equipment status fluctuation trend, to reflect the degree of influence of environmental factors on equipment. The analysis module is used to perform cross-analysis of the environment-equipment correlation index with the safety feature parameters extracted from personnel behavior images to construct a multivariate risk correlation map; The allocation module is used to assign weights to the risk nodes in the multivariate risk association graph. When environmental anomalies and equipment failures occur simultaneously, the composite risk level is raised by one level. When personnel violations and environmental anomalies occur concurrently, a safety warning state is triggered, and a comprehensive risk assessment level is generated. The activation module is used to perform graded response control based on the comprehensive risk assessment level. When the assessment level is mild risk, the environmental control equipment is activated. When the assessment level reaches severe risk, the audible and visual alarms and equipment protection measures are activated simultaneously.

8. A power distribution room monitoring device based on multi-source power distribution data, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the power distribution room monitoring method based on power distribution multi-source data as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the power distribution room monitoring method based on power distribution multi-source data as described in any one of claims 1 to 6.

Images