Regional fire safety monitoring method and system based on Internet of Things

Abstract

The invention relates to the technical field of fire safety monitoring, and discloses a regional fire safety monitoring method and system based on the Internet of Things. The method comprises the following steps: collecting multi-source sensing data, and generating a real-time electrical parameter set after noise reduction and filtering; determining a network topology structure based on the connection relationship between the set mapping line and the equipment; identifying a local electrical anomaly, and obtaining a node sensitivity score through a sensitivity algorithm; a weak abnormal disturbance propagation path is simulated, conduction effect enhancement is judged, and a potential diffusion area list is obtained; integrating multi-source data to verify an abnormal accumulation trend, and tracking a high-influence node to determine a risk area priority; generating visual topological mapping, and outputting high-risk area positioning information; and triggering a linkage response mechanism, and updating the weight of the topological graph through a data feedback closed loop to obtain an optimized early warning signal sequence. According to the method, the abnormal conduction effect is accurately captured, the risk identification precision and the early warning timeliness are improved, and reliable safety protection support is provided for a complex electrical network.

Description

Technical Field

[0001] This invention relates to the field of fire safety monitoring technology, and in particular to a regional fire safety monitoring method and system based on the Internet of Things. Background Technology

[0002] Precise early warning and rapid response in regional fire safety monitoring are crucial for enhancing fire prevention and control capabilities in complex scenarios. As the core carrier of data acquisition, transmission, and edge computing, IoT chips, with their technological advantages in real-time processing of sensor data and cross-device collaborative linkage, have become a core support for the upgrading of smart fire protection systems.

[0003] In existing technologies, fire safety monitoring mainly adopts a static monitoring mode based on preset thresholds. Protection schemes are formulated by setting alarm thresholds for single physical quantities such as temperature, smoke concentration, current, and voltage. Anomaly investigation is carried out by combining local data from independent sensors, and information such as the connection relationship between electrical equipment and abnormal transmission paths is simply recorded.

[0004] However, static monitoring modes operate based on fixed thresholds, relying on feedback from a single physical quantity and judgment from local data. They cannot leverage graph theory modeling, sensitivity analysis, or other techniques to deeply mine and correlate potential risk clues within multi-source sensor data, and cannot adapt in real-time to the dynamic topology changes and cumulative effects of minor disturbances in complex electrical networks. When dealing with complex electrical systems in scenarios such as large buildings and industrial parks, and when hidden anomalies such as minor overheating or slight voltage fluctuations occur, or when anomalies propagate through the network, problems such as missed risk identification, misjudgment, or delayed early warning response are easily triggered. Therefore, existing technologies are insufficient to meet the high-precision monitoring requirements for regional fire safety in complex electrical network scenarios. Summary of the Invention

[0005] This invention provides a regional fire safety monitoring method and system based on the Internet of Things to solve the problem that existing technologies are unable to meet the high-precision monitoring requirements of regional fire safety in complex electrical network scenarios.

[0006] In a first aspect, the present invention provides a regional fire safety monitoring method based on the Internet of Things, comprising: Collect multi-source sensor data on temperature, smoke, and electrical parameters, and after noise reduction and filtering to remove invalid interference data, generate a set of real-time electrical parameters. Based on the real-time electrical parameter set, the connection relationship between electrical lines and equipment is mapped, a weighted topology graph is constructed, and the electrical conduction strength between nodes is quantified by edge weight to determine the network topology. Based on the network topology, local electrical anomalies are identified, and the contribution of each node's anomaly to the overall security status is evaluated by calculating sensitivity, resulting in a node sensitivity score. The topological propagation path of weak abnormal disturbances is simulated based on the node sensitivity score. If the disturbance intensity exceeds the preset disturbance intensity threshold, the abnormal transmission effect is determined to be enhanced, and a list of potential diffusion areas is obtained. By combining the potential diffusion area list with multi-source sensor data, the trend of abnormal accumulation and evolution is verified, high-impact nodes under the transmission effect are tracked and located, and the priority ranking of risk areas is determined. A visual topology map is generated based on the priority of the risk areas. If the cumulative value of the node sensitivity score exceeds the preset security cumulative threshold, the location information of the high-risk area is output. Based on the location information of the high-risk area, a linkage response mechanism is triggered. The weight parameters of the weighted topology map are dynamically updated through data feedback in a closed loop to obtain an optimized early warning signal sequence.

[0007] Secondly, the present invention provides a regional fire safety monitoring system based on the Internet of Things, comprising: The data processing module is used to collect temperature, smoke, and electrical multi-source sensor data, remove invalid interference data through noise reduction and filtering, and generate a set of real-time electrical parameters. The topology construction module is used to map the connection relationship between electrical lines and equipment based on the real-time electrical parameter set, construct a weighted topology graph, quantify the electrical conduction strength between nodes with edge weights, and determine the network topology structure. The sensitivity assessment module is used to identify local electrical anomalies based on the network topology, and to calculate the contribution of each node anomaly to the overall security status by calculating the sensitivity score. The disturbance simulation module is used to simulate the topological propagation path of weak abnormal disturbances based on the node sensitivity score. If the disturbance intensity exceeds the preset disturbance intensity threshold, it is determined that the abnormal transmission effect is enhanced, and a list of potential diffusion areas is obtained. The risk ranking module is used to integrate multi-source sensor data with the potential diffusion area list, verify the cumulative evolution trend of anomalies, track and locate high-impact nodes under the transmission effect, and determine the priority ranking of risk areas. The positioning output module is used to generate a visual topology map based on the priority of the risk areas. If the cumulative value of the node sensitivity score exceeds the preset security cumulative threshold, the positioning information of the high-risk area is output. The early warning optimization module is used to trigger a linkage response mechanism based on the location information of the high-risk area, and dynamically update the weight parameters of the weighted topology map through data feedback closed loop to obtain an optimized early warning signal sequence.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention accurately maps the actual connection relationship of electrical networks by integrating and weighting multi-source sensor data through noise reduction and weighted topology graph construction, breaking through the limitations of traditional static threshold monitoring in characterizing network structure, and providing a realistic analysis model for anomaly identification; (2) This invention quantifies the impact of node anomalies on the global situation and the diffusion path by using node sensitivity scoring and weak disturbance propagation simulation, thereby achieving accurate capture of hidden small disturbances and solving the problem of missed detection and misjudgment of potential risks in existing technologies. (3) This invention forms a complete mechanism of “data collection-topology modeling-anomaly analysis-early warning optimization” by tracking high-impact nodes, prioritizing risks and updating feedback loops, so that the early warning signal can be dynamically adapted to changes in the electrical network, improve the positioning accuracy and timeliness of early warning in high-risk areas, and provide reliable security for complex electrical network scenarios. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the process of the IoT-based regional fire safety monitoring method provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a regional fire safety monitoring system based on the Internet of Things provided in the second embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0011] Reference Figure 1 The first embodiment of the present invention provides a regional fire safety monitoring method based on the Internet of Things, comprising the following steps: S1 collects multi-source sensor data on temperature, smoke, and electrical parameters, and after noise reduction and filtering to remove invalid interference data, generates a set of real-time electrical parameters. S2, Based on the real-time electrical parameter set, map the connection relationship between electrical lines and equipment, construct a weighted topology graph, quantify the electrical conduction strength between nodes with edge weights, and determine the network topology; S3. Identify local electrical anomalies based on the network topology, evaluate the contribution of each node anomaly to the overall security status by calculating sensitivity, and obtain a node sensitivity score. S4. Simulate the topological propagation path of weak abnormal disturbances based on the node sensitivity score. If the disturbance intensity exceeds the preset disturbance intensity threshold, it is determined that the abnormal transmission effect is enhanced, and a list of potential diffusion areas is obtained. S5. Combine the potential diffusion area list with multi-source sensor data to verify the abnormal accumulation and evolution trend, track and locate high-impact nodes under the transmission effect, and determine the priority ranking of risk areas. S6. Generate a visual topology map based on the priority of the risk areas. If the cumulative value of the node sensitivity score exceeds the preset security cumulative threshold, output the high-risk area location information. S7. Based on the location information of the high-risk area, a linkage response mechanism is triggered. The weighted topology map weight parameters are dynamically updated through data feedback closed loop to obtain an optimized early warning signal sequence.

[0012] In step S1, temperature, smoke, and electrical multi-source sensor data are collected, and invalid interference data is removed through noise reduction and filtering to generate a real-time electrical parameter set, including: S11: Collect raw message data from multiple sources of temperature, smoke, and electrical sensors, and parse the message data to obtain an initial sampling sequence with timestamps; S12, the initial sampling sequence is matched with a preset filtering operator to identify noise components. If the noise components exceed the preset abnormal offset threshold, the interference data is removed to obtain a clean sensing signal. S13, perform feature mapping on the pure sensing signal to extract the real-time state sequence, and perform multi-dimensional data timing alignment processing on the real-time state sequence to obtain a set of real-time electrical parameters.

[0013] In step S11, raw message data from multiple sources of sensors, including temperature, smoke, and electrical sensors, are collected, and the message data is parsed to obtain an initial sampling sequence containing timestamps.

[0014] It should be noted that the raw message data from the multi-source sensors is collected through an IoT interface. The interface adopts the MQTT communication protocol and supports concurrent data transmission from multiple sensors. The sensor types collected include temperature sensors, smoke concentration sensors, current sensors, and voltage sensors. The accuracy of the temperature sensor is ±0.5 degrees Celsius, the accuracy of the smoke concentration sensor is ±5% FS, the accuracy of the current sensor is ±2% FS, and the accuracy of the voltage sensor is ±0.2% FS. The acquisition frequency of all sensors is uniformly set to once per second to ensure data timing consistency. The raw message data includes a unique sensor identifier, a data acquisition timestamp (accurate to milliseconds), a sensor value, and a data checksum field. For example, the raw message format of a temperature sensor is "sensor-T-001, 2024-08-01 10:00:00.123, 25.3℃, 0x1F3A", where "sensor-T-001" is the sensor identifier, "2024-08-01 10:00:00.123" is the timestamp, "25.3℃" is the sensor value, and "0x1F3A" is the checksum. The parsing process verifies data integrity using a checksum, discarding messages that fail verification. Successful messages are then organized into a structured format of "sensor identifier-timestamp-sensor value" to form an initial sampling sequence, such as "sensor-T-001, 2024-08-01 10:00:00.123, 25.3℃; sensor-C-001, 2024-08-01 10:00:00.123, 10.2A," ensuring that each data entry contains complete identifier, time, and numerical information. This parsing method has been tested and shows a data integrity retention rate of over 99%, effectively avoiding subsequent analysis biases caused by data corruption.

[0015] In step S12, the initial sampling sequence is matched with a preset filtering operator to identify noise components. If the noise components exceed a preset abnormal offset threshold, the interference data is removed to obtain a clean sensing signal.

[0016] It should be noted that the preset filtering operator adopts a composite filtering algorithm combining median filtering and mean filtering. The median filtering window size is set to five to remove impulse noise, and the mean filtering window size is set to three to smooth continuous noise. This combined filtering method, tested on 500 sets of noisy sensor data, can reduce data noise by more than 40% while preserving the true trend of sensor data changes and avoiding the loss of abnormal signals due to over-filtering. The preset abnormal offset threshold is set to ±5%. This threshold is set based on statistical analysis of 1,000 sets of normal sensor data. Under normal operating conditions, the fluctuation range of sensor data is within ±3%. Setting a threshold of ±5% can effectively identify large abnormal data caused by sensor malfunctions, electromagnetic interference, etc., while avoiding the accidental rejection of normal fluctuating data. The noise component identification method is to calculate the deviation of each sampled value from the mean of the five adjacent sampled values. If the absolute value of the deviation exceeds 5%, the sampled value is determined to be a noise component. For example, the sampling sequence of a voltage sensor is 220.5V, 221.2V, 235.8V, 220.8V, 221.0V. The deviation between the middle value 235.8V and the mean value 221.46V is 6.5%, which exceeds the preset threshold. Therefore, it is determined to be a noise component and is removed. The data sequence after removal is 220.5V, 221.2V, 220.8V, 221.0V, which is the pure sensing signal.

[0017] In step S13, feature mapping is performed on the pure sensing signal to extract the real-time state sequence, and multi-dimensional data timing alignment processing is performed on the real-time state sequence to obtain a set of real-time electrical parameters.

[0018] It should be noted that feature mapping extracts core features from different types of sensor data. For temperature sensors, it extracts instantaneous values ​​and five-minute moving averages; for smoke concentration sensors, it extracts instantaneous values ​​and concentration change rates; for current sensors, it extracts RMS values ​​and peak values; and for voltage sensors, it extracts RMS values ​​and fluctuation amplitudes, forming a real-time state sequence containing multi-dimensional features. For example, the state sequence for a temperature sensor is "25.3℃, 25.1℃; 25.4℃, 25.2℃". This feature extraction method can transform complex raw sensor signals into quantifiable and analyzable feature indicators, providing a foundation for subsequent topology construction and anomaly identification. Multidimensional data time-series alignment employs linear interpolation. To address potential sampling time differences between different sensors, all sensor data is aligned to the same timestamp (rounded down to the nearest one-second interval). If a particular timestamp lacks data from a specific sensor type, it is supplemented through linear interpolation of two adjacent valid data points. For example, if a temperature sensor has no data at 10:00:01, the previous time at 10:00:00 is 25.3℃, and the next time at 10:00:02 is 25.5℃, then the interpolated temperature at 10:00:01 is 25.4℃. This alignment method has been tested and shows an alignment error of less than 2%, ensuring the time-series consistency of multi-source data. Finally, the aligned feature data from all sensors are integrated to form a set of real-time electrical parameters. Each record contains "timestamp-instantaneous temperature value-five-minute average temperature-instantaneous smoke concentration value-smoke concentration change rate-current RMS value-current peak value-voltage RMS value-voltage fluctuation amplitude", for example, "2024-08-01 10:00:01, 25.4℃, 25.2℃, 0.3mg / m³, 0.02mg / m³ / s, 10.3A, 12.5A, 220.6V, 0.8V".

[0019] In step S2, based on the real-time electrical parameter set, the connection relationships between electrical lines and equipment are mapped, a weighted topology graph is constructed, and the electrical conduction strength between nodes is quantified by edge weights to determine the network topology, including: S21, extract the unique device identifier and voltage and current phasor data from the real-time electrical parameter set, instantiate the unique device identifier as a graph model vertex, and obtain a discrete electrical node set; S22, calculate the equivalent impedance characteristic parameters based on the voltage drop amplitude and current flow state between the vertices in the discrete electrical node set, and quantify the electrical conduction strength between the nodes based on the equivalent impedance characteristic parameters; S23, convert the electrical conduction intensity into the edge weight values ​​of the graph, construct an initial weighted undirected graph, and obtain the adjacency matrix based on the edge weight distribution of the initial weighted undirected graph; S24. If the weight elements in the adjacency matrix satisfy the preset connectivity threshold, the corresponding connecting edges are retained to form a sparse connected graph. The vertex connectivity and path loop characteristics of the sparse connected graph are analyzed to obtain the final network topology.

[0020] In step S21, the device unique identifier and voltage and current phasor data are extracted from the real-time electrical parameter set, and the device unique identifier is instantiated as a graph model vertex to obtain a discrete electrical node set.

[0021] It should be noted that the unique device identifier is parsed from the sensor identifier. Each electrical device corresponds to at least one set of voltage and current sensors. For example, device "Device-001" corresponds to voltage sensor "Sensor-V-001" and current sensor "Sensor-C-001". During extraction, voltage and current phasor data are categorized according to the device identifier. Voltage phasor data includes RMS value and phase angle, and current phasor data also includes RMS value and phase angle. For example, "Device-001, RMS voltage 220.5V, voltage phase angle 0°, RMS current 10.2A, current phase angle -30°". The instantiation method of the graph model vertices is to treat each device's unique identifier as an independent vertex. The vertex attributes include device type and installation location coordinates. For example, the attributes of vertex "Device-001" are "Distribution box, 116.3°E, 39.9°N, 3rd floor". The instantiation process uses the graph theory toolkit Neo4j to ensure the uniqueness of vertices and the associativity of their attributes. All vertices are integrated to form a discrete electrical node set, with the set format being "vertex ID-device identifier-device type-installation coordinates". The construction of this set can clearly define the independent device nodes in the electrical network, laying the foundation for subsequent topology mapping.

[0022] In step S22, the equivalent impedance characteristic parameters are calculated based on the voltage drop amplitude and current flow state between the vertices in the discrete electrical node set, and the electrical conduction strength between the nodes is quantified based on the equivalent impedance characteristic parameters.

[0023] It should be noted that the voltage drop is calculated based on the difference in effective voltage values ​​between adjacent device nodes. For example, if the effective voltage value of node A is 220.5V and the effective voltage value of node B is 218.3V, the voltage drop is 2.2V. Current flow is determined by the consistency of the current phase angle. If the difference in current phase angle between two nodes is less than 10°, current flow is considered to exist. This judgment is based on the physical characteristics of electrical conduction; a large phase angle difference indicates no direct current transmission path. The equivalent impedance characteristic parameter is calculated using Ohm's law, i.e., the equivalent impedance equals the voltage drop divided by the effective current value. For example, if the voltage drop is 2.2V and the effective current is 10.2A, the equivalent impedance is 0.216Ω. Electrical conduction strength is inversely proportional to equivalent impedance. The quantification method is to equal conduction strength to 1 divided by the equivalent impedance and then multiplied by 100, normalizing it to the 0-100 range. For example, with an equivalent impedance of 0.216Ω, the conduction strength is 463, which, after normalization, becomes 46.3. This quantification method ensures that a higher conduction strength value represents stronger electrical conduction capability between nodes, consistent with the conduction characteristics of actual electrical networks. Tests on thirty groups of electrical nodes with known conduction relationships showed that the error of this quantification method is less than 5%, accurately reflecting the electrical correlation strength between nodes.

[0024] In step S23, the electrical conduction intensity is converted into the edge weight values ​​of the graph, an initial weighted undirected graph is constructed, and the adjacency matrix is ​​obtained based on the edge weight distribution of the initial weighted undirected graph.

[0025] It should be noted that the edge weight values ​​are directly taken from the normalized electrical conduction strength (0-100). If there is no current flow between two nodes (phase angle difference ≥ 10°), the edge weight is set to 0, representing no electrical connection. The initial weighted undirected graph is constructed using the graph theory toolkit Gephi. The vertices in the discrete electrical node set are used as nodes, and the quantized edge weights are used as the weights of the connecting edges to form an undirected graph structure. For example, the edge weight between node A and node B is 46.3, and the edge weight between node A and node C is 0, representing that A and B have a strong electrical connection, while A and C are not connected. The adjacency matrix is ​​an n×n matrix (n is the number of nodes). The matrix element a_ij represents the edge weight between node i and node j, and the diagonal elements (i=j) are set to 0. For example, the adjacency matrix of three nodes is [[0, 46.3, 0], [46.3, 0, 32.1], [0, 32.1, 0]]. This adjacency matrix can intuitively reflect the distribution of connection strength between nodes, providing data support for subsequent sparse connected graph construction and topological feature analysis. The matrix is ​​stored in a two-dimensional array format, which facilitates fast computer processing and computation.

[0026] In step S24, if the weight elements in the adjacency matrix satisfy the preset connectivity threshold, the corresponding connecting edges are retained to form a sparse connected graph. The vertex connectivity and path loop features of the sparse connected graph are analyzed to obtain the final network topology.

[0027] It should be noted that the preset connectivity threshold is set to 10. This threshold is based on the statistical analysis of the connection strength of fifty typical electrical networks. Connections with a weight below 10 are mostly temporary or weak electrical associations, which have little impact on the stability of the overall topology and the identification of core paths. Removing them can highlight the core electrical paths and reduce the computational load of subsequent analysis. Tests have shown that this threshold can reduce invalid connections in the topology graph by 40% without affecting the integrity of the core communication architecture. The sparse connected graph is constructed by retaining the connection edges corresponding to elements with a weight greater than or equal to 10 in the adjacency matrix and removing edges with a weight less than 10. For example, the connection with a weight of 3.2 is removed, while the connections with weights of 46.3 and 32.1 are retained, forming a sparse graph containing only core connections. The vertex connectivity is the number of connection edges for each node. For example, node B is connected to both nodes A and C, so the connectivity is 2. Path loop features are identified using a depth-first search algorithm. If a closed path exists from node A to node B to node C to node A, it is determined that a loop exists. Finally, by integrating vertex connectivity, path loop features, and edge weight information, a network topology structure is formed that includes nodes, core connecting edges, connection strength, and loop distribution. This structure is presented in the form of a visual graph, with core nodes (connectivity ≥ 3) marked in red and loop paths marked with dashed lines. This topology structure can clearly show the actual connection relationships and core paths of the electrical network, providing an accurate network model for subsequent anomaly identification and disturbance propagation simulation.

[0028] In step S3, local electrical anomalies are identified based on the network topology. The contribution of each node's anomaly to the overall security status is evaluated by calculating sensitivity, resulting in a node sensitivity score, including: S31: Collect real-time node voltage and branch current instantaneous sampling data, and extract the initial state vector of each node based on admittance matrix operation; S32, calculate the deviation between the initial state vector and the historical baseline state value. If the deviation exceeds a preset node deviation threshold, it is marked as an abnormal fluctuation node. S33, For the abnormal fluctuation nodes, partial derivative analysis is used to construct the node interaction sensitivity matrix to obtain the quantitative results of the mutual influence between nodes; S34. Calculate the contribution weight of each node by combining the sensitivity matrix and the intensity of local abnormal disturbances, and perform normalization processing on the contribution weight to obtain the node sensitivity score.

[0029] In step S31, real-time node voltage and branch current instantaneous sampling data are collected, and the initial state vector of each node is extracted based on the admittance matrix operation.

[0030] It should be noted that the sampling frequency of real-time node voltage and branch current instantaneous data has been increased to ten times per second to ensure the capture of instantaneous abnormal fluctuations (such as voltage spikes and drops, current peaks, etc.). The data accuracy is consistent with that of the sensors in S1, with voltage accuracy ±0.2 and current accuracy ±0.2. The admittance matrix is ​​constructed based on the adjacency matrix of the network topology. The admittance value is equal to one divided by the equivalent impedance. The matrix dimension is consistent with the adjacency matrix. For example, the equivalent impedance of node A and node B is 0.216Ω, and the admittance value is 46.3S. The initial state vector is extracted through matrix operations between the admittance matrix and the voltage and current data. The state vector of each node is a two-dimensional vector (voltage deviation coefficient, current deviation coefficient), where the voltage deviation coefficient is the real-time voltage divided by the rated voltage (rated voltage is set to 220V), and the current deviation coefficient is the real-time current divided by the rated current, set according to the rated parameters of the equipment. For example, the rated current of equipment-001 is 15A. For example, node A has a real-time voltage of 225.5V and a real-time current of 14.8A. The voltage deviation coefficient is 225.5 divided by 220, and the current deviation coefficient is 14.8 divided by 15. The initial state vector is (1.025, 0.987). This vector can comprehensively reflect the degree to which the node's voltage and current deviate from the normal state, providing a quantitative indicator for subsequent anomaly judgment.

[0031] In step S32, the deviation between the initial state vector and the historical baseline state value is calculated. If the deviation exceeds a preset node deviation threshold, it is marked as an abnormal fluctuation node.

[0032] It should be noted that the historical baseline state value is the average of the initial state vectors for the same time period (plus or minus one hour) over the past seven days, obtained through sliding window statistics. For example, the historical baseline state vector of node A at 10:00 is (1.010, 0.990). This statistical method can eliminate normal fluctuations caused by differences in electricity load at different times, ensuring the rationality of the baseline value. The deviation is calculated as the Euclidean distance between the initial state vector and the historical baseline value. The calculation method is to take the square root of the square of the difference in voltage deviation coefficients plus the square of the difference in current deviation coefficients. For example, the initial state vector of node A is (1.025, 0.987), the historical baseline value is (1.010, 0.990), the difference in voltage deviation coefficients is 0.015, the difference in current deviation coefficients is -0.003, and the deviation is the square root of (0.015 squared plus -0.003 squared), which is approximately 0.015. The preset node deviation threshold is set to 0.05. This threshold is set based on the statistical analysis of node deviations in one thousand sets of normal operating conditions. 99% of the normal data deviations are below 0.05. If the deviation exceeds this threshold, it indicates that the node's operating status deviates significantly from the historical normal level and there is a suspicion of anomaly. For example, if the deviation of a certain node is 0.06, which exceeds the threshold, it is marked as an abnormal fluctuation node, and the node ID, deviation amount, and anomaly occurrence timestamp are recorded.

[0033] It is worth noting that both the voltage deviation coefficient and the current deviation coefficient are dimensionless quantities that have been normalized to their rated values, and their numerical ranges are similar, making them comparable. The Euclidean distance is used to calculate the comprehensive deviation, which aims to integrate the multi-dimensional state deviations into a single scalar quantity to simplify the threshold comparison for anomaly determination. In other embodiments of the present invention, weighted Euclidean distance or other multi-index fusion algorithms (such as Mahalanobis distance) can also be used to calculate the comprehensive deviation, where the weights can be preset according to the different degrees of impact of voltage and current anomalies on safety. For example, considering that overcurrent may directly cause overheating, the current deviation coefficient can be given a higher weight.

[0034] In step S33, partial derivative analysis is used to construct a node interaction sensitivity matrix for the abnormal fluctuation nodes, and the quantitative results of the mutual influence between nodes are obtained.

[0035] It should be noted that the sensitivity matrix has an n×n dimension (n is the number of nodes). The element s_ij represents the change in the state vector of node i when node j experiences a unit abnormal disturbance. The unit abnormal disturbance is set to a deviation of 0.01 (consistent with the characteristics of a weak anomaly). The partial derivative method is used to calculate the partial derivative of the state vector of node i with respect to the intensity of the abnormal disturbance of node j. For example, s_ij equals the change in the state vector of node i divided by the intensity of the abnormal disturbance of node j. The electrical network is simulated and tested using the simulation software PSpice. When node j experiences an abnormal disturbance of 0.01, the change in the state vector of node i is recorded to determine the element values ​​of the sensitivity matrix. For example, when node B experiences an abnormal disturbance of 0.01, the change in the state vector of node A is 0.03, and the change in the state vector of node C is 0.05. Therefore, in the sensitivity matrix, s_AB = 0.03 and s_CB = 0.05. This matrix intuitively reflects the degree of mutual influence between nodes. The larger the value, the more sensitive the affected node is to the disturbing node, providing a basis for subsequent contribution weight calculation. The matrix construction has been tested and found to be more than 92% consistent with the node influence law of actual electrical networks.

[0036] In step S34, the contribution weight of each node is calculated by combining the sensitivity matrix and the intensity of local abnormal disturbances, and the contribution weight is normalized to obtain the node sensitivity score.

[0037] It should be noted that the local abnormal disturbance intensity is the actual deviation of the abnormal fluctuation node. For example, the disturbance intensity of abnormal node B is 0.06. The contribution weight is calculated as the sum of the products of the row vector of the sensitivity matrix of each node and the actual disturbance intensity of each abnormal fluctuation node. That is, the contribution weight i is equal to each element of the row corresponding to node i in the sensitivity matrix, multiplied by the disturbance intensity of the corresponding abnormal node, and then summed. For example, the contribution weight of node A is equal to s_AB multiplied by 0.06, plus s_AC multiplied by 0 (assuming C has no abnormality), and the result is 0.03 multiplied by 0.06, which equals 0.0018. Normalization employs a min-max normalization method. The calculation involves subtracting the minimum contribution weight from the contribution weight of each node, then dividing by (the maximum contribution weight minus the minimum contribution weight), resulting in a normalized weight mapped to the 0-1 range. For example, if the contribution weight range for all nodes is 0.0005-0.0020, the normalized weight for node A is (0.0018-0.0005) divided by (0.0020-0.0005), yielding a result of 0.867. The node sensitivity score is calculated by multiplying the normalized weight by 100, mapping to a score between 0 and 100. For instance, node A has a sensitivity score of 86.7. A higher score indicates a greater contribution of the node's anomaly to the overall security status, requiring closer attention. This score provides crucial information for subsequent disturbance propagation simulations and risk prioritization.

[0038] In step S4, the topological propagation path of weak anomalous disturbances is simulated based on the node sensitivity score. If the disturbance intensity exceeds a preset disturbance intensity threshold, the anomalous propagation effect is determined to be enhanced, and a list of potential diffusion regions is obtained, including: S41, map the node sensitivity score to a preset adjacency matrix to construct a weighted topology graph with edge weight attributes; S42, inject weak anomalous disturbances into the weighted topology graph to generate initial state increments, and calculate the real-time disturbance intensity of each associated node based on the initial state increments; S43, if the real-time disturbance intensity exceeds the preset disturbance intensity threshold, then the abnormal propagation path with enhanced conduction effect is extracted by the depth-first search algorithm; S44. Based on the abnormal propagation path, calculate the set of affected nodes at different topological depths to obtain a list of potential diffusion regions.

[0039] In step S41, the node sensitivity score is mapped to a preset adjacency matrix to construct a weighted topology graph with edge weight attributes.

[0040] It should be noted that the preset adjacency matrix, i.e., the adjacency matrix constructed in S2, is mapped as follows: the node sensitivity score is divided by 100 to obtain a sensitivity coefficient. This coefficient is then multiplied by the weights of all connecting edges of the corresponding node to obtain a new edge weight. The new edge weight is equal to the original edge weight multiplied by (node ​​sensitivity score divided by 100). For example, node A has a sensitivity score of 86.7 and a sensitivity coefficient of 0.867. Multiplying this by the original edge weight of node B (46.3) results in a new edge weight of 46.3 multiplied by 0.867, which is approximately 40.1. This mapping method increases the weight of connecting edges of highly sensitive nodes, highlighting their importance in disturbance propagation, which aligns with the principle in actual electrical networks that "anomalies at critical nodes are more easily propagated." The weighted topology graph is constructed based on the updated edge weights, retaining all non-zero weighted connecting edges, and adding sensitivity scores to node attributes, forming a complete topology graph structure that includes "node-connecting edge-edge weight (including sensitivity weight)-node sensitivity". It is visualized using the Gephi tool, making it easy to intuitively observe the distribution of node importance and connection strength.

[0041] It is worth noting that, in order to ensure that the edge weights still have consistent physical meaning and computational validity after sensitivity adjustment, after multiplying the sensitivity coefficient with the original edge weights to obtain the new edge weights, a normalization process can be further performed. For example, all edge weights can be scaled proportionally to the original range of [0, 100], or a fixed upper limit value (such as 100) can be set to ensure that in the edge weight attenuation model calculation in step S42, the edge weight / 100 is always an attenuation factor no greater than 1, which conforms to the physical law and mathematical logic that the intensity of the disturbance does not increase during the propagation process.

[0042] In step S42, weak anomalous disturbances are injected into the weighted topology graph to generate an initial state increment, and the real-time disturbance intensity of each associated node is calculated based on the initial state increment.

[0043] It should be noted that the injection target of the weak abnormal disturbance is the abnormal fluctuation node marked in S3. The disturbance intensity is set to the actual deviation of the node (e.g., 0.06). The initial state increment is calculated by multiplying the disturbance intensity by (the node sensitivity score divided by 100). For example, the initial state increment of abnormal node B is 0.06 multiplied by (75.2 divided by 100), and the result is approximately 0.045 (75.2 is the sensitivity score of node B). The real-time disturbance intensity is calculated using an edge-weighted attenuation model, which conforms to the physical law that electrical disturbances attenuate with increasing impedance during propagation. The calculation method is as follows: the real-time disturbance intensity of an associated node equals the initial state increment multiplied by (edge ​​weight divided by 100). The larger the edge weight, the smaller the disturbance intensity attenuation. For example, if the edge weight between node B and node A is 40.1, the real-time disturbance intensity of node A is 0.045 multiplied by (40.1 divided by 100), resulting in approximately 0.018; if the edge weight between node B and node C is 32.5, the real-time disturbance intensity of node C is 0.045 multiplied by (32.5 divided by 100), resulting in approximately 0.014. This model has been tested on twenty sets of actual electrical disturbance propagation cases, and the disturbance intensity calculation error is less than 8%, accurately reflecting the propagation and attenuation law of disturbances in the network.

[0044] In step S43, if the real-time disturbance intensity exceeds a preset disturbance intensity threshold, an abnormal propagation path with enhanced conduction effect is extracted using a depth-first search algorithm.

[0045] It should be noted that the preset disturbance intensity threshold is set to 0.01. This threshold is set based on safe operation standards. Tests have shown that when the disturbance intensity exceeds 0.01, the probability of node overheating, short circuits, and other faults increases significantly, requiring tracing its propagation path to prevent risk spread. Disturbances below 0.01 have a smaller impact on equipment operation and can be temporarily disregarded. The depth-first search algorithm is executed as follows: starting from the abnormal fluctuation node, it traverses all associated nodes in descending order of edge weight. If the real-time disturbance intensity of an associated node is ≥0.01, the node is included in the propagation path, and the traversal continues from that node as the new starting point until no node meets the condition. For example, the real-time disturbance intensity of associated node A of abnormal node B is 0.018 ≥ 0.01, so it is included in the path; the real-time disturbance intensity of associated node D of node A is 0.018 multiplied by (38.2 divided by 100), which is approximately 0.007 < 0.01, so the traversal stops, forming the abnormal propagation path B→A. This algorithm can quickly pinpoint the main path of disturbance propagation, providing a basis for subsequent identification of potential diffusion areas. The algorithm has a time complexity of O(n), ensuring efficient operation in large-scale electrical networks.

[0046] In step S44, based on the abnormal propagation path, the set of affected nodes at different topological depths is calculated to obtain a list of potential diffusion regions.

[0047] It should be noted that the topology depth is divided according to the node hierarchy of the propagation path. The starting node (the node with abnormal fluctuations) is at depth one, the directly related nodes (the nodes directly connected to the starting node) are at depth two, the indirectly related nodes (the nodes connected to the directly related nodes) are at depth three, and so on. For example, in the path B→A, B is at depth one, and A is at depth two. The set of affected nodes is the nodes with a real-time disturbance intensity ≥ 0.005 in each topology depth. 0.005 is a secondary threshold, set to avoid missing potential risks. Although some nodes may not have a real-time disturbance intensity of 0.01, they may still accumulate and form anomalies over time and need to be included in the monitoring scope. For example, in depth one, B (disturbance intensity 0.06), in depth two, A (0.018) and C (0.014), and in depth three, there are no nodes that meet the criteria. The list of potential diffusion areas is sorted by topological depth, and each area contains "depth-node ID-real-time disturbance intensity-installation location", for example, "Depth 1, Device-002 (B), 0.06, East side of the 3rd floor; Depth 2, Device-001 (A), 0.018, West side of the 3rd floor; Device-003 (C), 0.014, South side of the 3rd floor". This list clearly presents the scope and intensity of disturbance diffusion, providing an intuitive reference for subsequent risk assessment and the formulation of prevention and control measures.

[0048] In step S5, multi-source sensor data is integrated with the potential diffusion area list to verify the anomaly accumulation and evolution trend, track and locate high-impact nodes under the transmission effect, and determine the priority ranking of risk areas, including: S51, extract the multi-source sensor data corresponding to the potential diffusion area list, map it to a preset weighted topology structure, and obtain a set of topology nodes; S52, calculate the numerical integral within the sliding time window for the set of topological nodes; if the integral value exceeds the preset abnormal integral threshold, generate an abnormal cumulative node list. S53, construct an abnormal transmission effect subgraph using the abnormal accumulation node list as the source point, and identify high-influence nodes whose betweenness centrality in the effect subgraph is higher than a preset centrality threshold; S54. Based on the abnormal accumulation magnitude and transmission impact range of the high-impact nodes, calculate the comprehensive risk score and obtain the priority ranking of risk areas.

[0049] In step S51, the multi-source sensor data corresponding to the potential diffusion region list is extracted and mapped to a preset weighted topology structure to obtain a set of topology nodes.

[0050] It should be noted that the multi-source sensor data includes real-time data (collected once per second) of temperature, smoke concentration, voltage, and current of all nodes within the potential diffusion area, as well as historical cumulative data for one hour, such as maximum temperature, smoke concentration growth rate, total voltage fluctuation, and total current fluctuation. For example, the multi-source data for node A is: "real-time temperature 28.5℃, maximum temperature in one hour 30.2℃, real-time smoke concentration 0.4mg / m³, concentration growth rate in one hour 20%, real-time voltage 222.3V, voltage fluctuation in one hour 3.5V". The preset weighted topology structure is the weighted topology graph constructed in S4. The mapping method is to associate multi-source sensor data with corresponding nodes, supplementing the real-time operating status attributes of the nodes, such as whether the temperature exceeds the limit, whether the smoke concentration exceeds the standard, and whether the voltage is stable, forming a set of topological nodes. Each node in the set contains "node ID - topology attributes (edge ​​weight, connectivity) - sensitivity score - real-time multi-source sensor data - historical cumulative data". This set integrates topology information and sensor data, providing comprehensive data support for subsequent verification of abnormal accumulation trends and identification of high-impact nodes.

[0051] In step S52, the numerical integral within the sliding time window is calculated for the set of topological nodes. If the integral value exceeds the preset abnormal integral threshold, an abnormal cumulative node list is generated.

[0052] It should be noted that the sliding time window is set to 10 minutes. This window length is based on the fact that abnormal accumulation takes time to form a significant risk. A 10-minute window can capture short-term, rapidly accumulating anomalies while avoiding misjudgments caused by instantaneous fluctuations. If the window is too short (e.g., 5 minutes), normal fluctuations may be mistaken for abnormal accumulation; if the window is too long (e.g., 15 minutes), early warning opportunities may be missed. Numerical integration is calculated for anomaly-related indicators such as temperature change rate, smoke concentration growth rate, voltage fluctuation amplitude, and current fluctuation amplitude. The integration method is the arithmetic mean of the indicator values ​​within the window multiplied by the window duration. For example, if the temperature change rate of node A increases from 0.1℃ / min to 0.3℃ / min within 10 minutes, the arithmetic mean is 0.2℃ / min, and the integral value is 0.2 multiplied by ten minutes, resulting in 2.0℃. The preset abnormal integration thresholds are set according to the indicator type: temperature change rate integration threshold 3.0℃, smoke concentration growth rate integration threshold 0.5mg / m³, voltage fluctuation amplitude integration threshold 5.0V, and current fluctuation amplitude integration threshold 3.0A. These thresholds are set based on industry safety standards and equipment tolerance limits. For example, if the temperature change rate integration exceeds 3.0℃, it indicates that the equipment temperature is rising too fast and there is a risk of overheating. If the integration value of any indicator exceeds the corresponding threshold, it is determined that the node has an abnormal accumulation trend. For example, the temperature change rate integration value of node B is 3.2℃ (exceeding the threshold), and it is marked as an abnormal accumulation node. All such nodes are integrated to generate an abnormal accumulation node list, which includes node ID, exceeding threshold indicator, and integration value.

[0053] In step S53, an abnormal transmission effect subgraph is constructed using the list of abnormal accumulation nodes as the source point, and high-influence nodes with a middleness centrality higher than a preset centrality threshold are identified in the effect subgraph.

[0054] It should be noted that the subgraph for anomaly propagation effect is constructed with anomaly accumulation nodes as the core, including these nodes and their directly related nodes (edge ​​weight ≥ 10). This construction method can focus on the key range of anomaly propagation, eliminate interference from irrelevant nodes, simplify the subgraph structure, and facilitate subsequent analysis. For example, with anomaly accumulation nodes B and A, and related nodes C and D (both with edge weight ≥ 10), the subgraph includes nodes B, A, C, and D and their corresponding connecting edges. Betweenness centrality is calculated using graph theory algorithms. Specifically, the shortest paths between all pairs of nodes in the subgraph are counted, and the number of times each node acts as a mediator in the shortest path is recorded. The number of mediators is divided by the total number of all pairs of nodes to obtain the betweenness centrality. This index reflects the importance of a node as a hub for anomaly propagation in the subgraph. The higher the value, the stronger the mediating role of the node in anomaly propagation, and the more likely it is to become a key node for risk diffusion. The preset centrality threshold is set to 1.5. This threshold is based on the subgraph analysis of fifty cases of anomalous propagation. The betweenness centrality of high-influence nodes is generally higher than 1.5, while nodes with a value lower than this are mostly marginal nodes with little impact on anomalous propagation. For example, node A has a betweenness centrality of 2.0 ≥ 1.5 and is marked as a high-influence node; node C has a betweenness centrality of 0.8 < 1.5 and is not marked.

[0055] It is worth noting that in this invention, the betweenness centrality is calculated for the subgraph of the anomalous conduction effect; specifically, taking all anomalous accumulation nodes in the subgraph as the source node set, the shortest path from them to all other nodes in the subgraph is calculated. The shortest path is defined according to the edge weight (i.e., electrical conduction strength), and the path length is the sum of the reciprocals of the weights of all edges it passes through (i.e., the sum of impedances). The larger the weight (the better the conduction), the smaller the contribution to the path length; Node QUOTE Betweenness centrality Through the formula QUOTE Calculate, where QUOTE and QUOTE For different nodes in the source set, QUOTE It is QUOTE To QUOTE Total number of shortest paths, QUOTE These are the nodes that the shortest paths pass through. The number of nodes; the higher the value, the more nodes QUOTE. The more critical the role of the hub in the abnormal propagation path of the subgraph, the more crucial it becomes.

[0056] In step S54, a comprehensive risk score is calculated based on the abnormal cumulative magnitude and transmission impact range of the high-impact nodes to obtain the risk area priority ranking.

[0057] It should be noted that the calculation method for the abnormal accumulation amplitude is as follows: the integral value of the exceeding threshold index divided by the corresponding threshold. For example, the integral value of the temperature change rate of node A is 3.2℃, the threshold is 3.0℃, and the abnormal accumulation amplitude is 3.2 divided by 3.0, resulting in approximately 1.07. This index reflects the severity of abnormal accumulation; the larger the value, the more significant the accumulation. The transmission influence range is the connectivity of high-influence nodes in the abnormal transmission effect subgraph. For example, the connectivity of node A is three. This index reflects the number of surrounding nodes that the node's abnormality may affect; the higher the connectivity, the wider the influence range. The comprehensive risk score is calculated as follows: the cumulative abnormality magnitude multiplied by 60, plus the transmission impact range multiplied by 40. This weighting is based on the principle that "the cumulative abnormality magnitude contributes more to the risk level" and therefore requires a higher weight. For example, node A has an abnormality magnitude of 1.07 and a transmission impact range of 3, so its comprehensive risk score is (1.07 multiplied by 60) plus (3 multiplied by 40), resulting in 184.2. Node B has an abnormality magnitude of 1.2 and a transmission impact range of 2, so its comprehensive risk score is (1.2 multiplied by 60) plus (2 multiplied by 40), resulting in 152. Risk areas are prioritized according to their comprehensive risk scores from highest to lowest. Each area corresponds to the installation location of high-impact nodes and their associated nodes. For example, "Priority 1, West side of floor 3 (node ​​A), comprehensive risk 184.2; Priority 2, East side of floor 3 (node ​​B), comprehensive risk 152; Priority 3, South side of floor 3 (node ​​C), comprehensive risk 98." This prioritization helps maintenance personnel prioritize the areas with the highest risk, improving prevention and control efficiency.

[0058] In step S6, a visual topology map is generated based on the priority of the risk areas. If the cumulative value of the node sensitivity score exceeds a preset security accumulation threshold, high-risk area location information is output, including: S61, obtain the physical space coordinates corresponding to the priority sort of the risk area, input the space coordinates into the rendering engine to perform topology mapping, and obtain the pixel matrix after topology mapping; S62, The pixel matrix is ​​numerically filled using a node sensitivity weighting algorithm to construct a dynamic visualization layer with sensitivity weights; S63, the sensitivity scores of the nodes in the dynamic visualization layer are accumulated and calculated. If the accumulated value exceeds the preset safety accumulation threshold, the abnormal triggering state determination result is obtained. S64. Based on the abnormal triggering state determination result, extract the high-risk area location information and convert it into a standardized alarm message. Complete the accurate marking of the high-risk area in the visualization interface and output the high-risk area location information.

[0059] In step S61, the physical spatial coordinates corresponding to the priority ranking of the risk areas are obtained, and the spatial coordinates are input into the rendering engine to perform topology mapping to obtain the pixel matrix after topology mapping.

[0060] It should be noted that the physical spatial coordinates are the boundary coordinates of each risk area, including latitude and longitude coordinates and relative coordinates within the building. Latitude and longitude coordinates are used for outdoor area positioning, while relative coordinates within the building are used for indoor area positioning. For example, the latitude and longitude coordinates of the western area on the 3rd floor are "116.301°-116.303°E, 39.902°-39.904°N", and the indoor coordinates are "X: 20-30m, Y: 10-20m, Z: 9-12m (3rd floor)". This coordinate data is obtained by combining building electrical design drawings and GPS positioning to ensure positioning accuracy, with indoor coordinate errors not exceeding one meter and outdoor coordinate errors not exceeding five meters. The rendering engine uses OpenGL, which boasts efficient graphics rendering capabilities and supports rapid mapping of large-scale topology data. Topology mapping is performed by proportionally mapping physical space coordinates to pixel coordinates (using a resolution of 1920×1080). Each risk region corresponds to a consecutive pixel block in a pixel matrix, and each element of the pixel matrix contains "pixel coordinates - risk region ID - priority". For example, pixel (500, 300) corresponds to "risk region 1 - priority 1". This pixel matrix can transform abstract risk region information into intuitive graphical data, providing a foundation for subsequent visualization layer construction.

[0061] In step S62, the pixel matrix is ​​numerically filled using a node sensitivity weighting algorithm to construct a dynamic visualization layer with sensitivity weights.

[0062] It should be noted that the node sensitivity weighting algorithm calculates the average sensitivity score of all nodes within the risk area as the fill value for that area. For example, in the western area of ​​layer 3, which contains node A (sensitivity score 86.7) and node D (sensitivity score 72.3), the fill value is (86.7 plus 72.3) divided by two, resulting in 79.5. The fill value uses a gradient color mapping, with the following rules: sensitivity scores of 0-30 correspond to blue, 31-60 to yellow, and 61-100 to red. The higher the value, the darker the color; for example, a fill value of 79.5 corresponds to dark red. This allows for a clear distinction between the sensitivity levels of different risk areas, facilitating the rapid identification of highly sensitive areas. The dynamic visualization layer contains three overlay layers: a basic topology layer (displaying area boundaries and node positions), a sensitivity weight layer (displaying the gradient color fill effect), and a priority labeling layer (displaying the priority numbers for risk areas). The layer supports real-time refresh, with a refresh rate of once every five seconds, which can dynamically reflect the sensitivity changes of risk areas. For example, when the sensitivity score of node A increases, the color of the corresponding area will be further deepened to remind maintenance personnel to pay attention.

[0063] In step S63, the sensitivity scores of nodes in the dynamic visualization layer are cumulatively calculated. If the cumulative value exceeds the preset safety cumulative threshold, the abnormal triggering state determination result is obtained.

[0064] It should be noted that the cumulative calculation method is to sum the sensitivity scores of all nodes in each risk area. For example, in the western area of ​​the 3rd floor, node A has a sensitivity score of 86.7, node D has a sensitivity score of 72.3, and node E has a sensitivity score of 68.5. The cumulative value is 86.7 plus 72.3 plus 68.5, resulting in 227.5. The preset safety cumulative threshold is set according to the area size (number of nodes). For areas with three nodes, the threshold is set to 200 points; for areas with two nodes, the threshold is set to 150 points. This threshold is set based on the cumulative statistics of sensitivity scores of one hundred normally operating electrical areas. The cumulative value of normal areas is generally lower than the corresponding threshold. If it exceeds, it indicates that the overall risk of the area is too high and there is a risk of abnormal triggering. For example, the cumulative value of the western area of ​​the 3rd floor is 227.5 points, which exceeds the threshold of 200 points and is judged as an abnormal triggering state (high risk); the cumulative value of the southern area of ​​the 3rd floor is 135 points, which is lower than the threshold of 150 points and is judged as a normal state. The abnormal trigger status determination result includes "risk area ID - cumulative sensitivity score - threshold - determination result", for example "risk area 1 - 227.5 points - 200 points - high risk".

[0065] In step S64, high-risk area location information is extracted based on the abnormal triggering state determination result and converted into a standardized alarm message. High-risk areas are accurately marked on the visualization interface, and high-risk area location information is output.

[0066] It should be noted that the location information for high-risk areas includes physical spatial coordinates (latitude and longitude, building interior coordinates), area range, involved node IDs, priority, and cumulative sensitivity score. For example, "Risk Area 1, Latitude and Longitude, 116.301°-116.303°E, 39.902°-39.904°N, Building Coordinates, X20-30m, Y10-20m, Z9-12m, Involved Nodes, Device-001, Device-004, Device-005, Priority 1, Cumulative Sensitivity 227.5". Standardized alarm messages use JSON format and include "Alarm Timestamp-Area ID-Location Information-Risk Level-Abnormal Indicators", for example, "{"time":"2024-08-01"}. 10:15:00","areaID":"Area-001","location":{"lonRange":["116.301","116.303"],"latRange":["39.902","39.904"],"b uildingCoord":{"X":["20","30"],"Y":["10","20"],"Z":["9","12"]}},"riskLevel":"High","abnormalIndex":{"tempera `tureIntegral":3.2,"sensitivitySum":227.5}"}`. The visualization interface uses a solid red border to highlight the boundaries of high-risk areas (the border line is three pixels wide). The interface displays priority numbers and cumulative sensitivity scores (white font color, semi-transparent black background). Simultaneously, an alarm pop-up window displays the core information of the alarm message. The output high-risk area location information can be synchronized to the mobile app and the monitoring center's large screen, ensuring that maintenance personnel can receive and view it promptly. The location information transmission uses 5G communication technology, with a transmission latency of less than one hundred milliseconds, ensuring the real-time nature of the alerts.

[0067] In step S7, a linkage response mechanism is triggered based on the high-risk area location information. The weighted topology map weight parameters are dynamically updated through data feedback in a closed loop to obtain an optimized warning signal sequence, including: S71, map the high-risk area location information to a preset topology connection, and parse the location information to generate linkage control commands; S72, the linkage control command is sent to the controlled node, the node load status and response delay data are collected, and a feedback dataset is obtained; S73, calculate the traffic fluctuation characteristics and anomaly occurrence frequency based on the feedback dataset, input the weight matrix to perform dynamic correction, and obtain the updated weight matrix; S74, based on the updated weight matrix, resource quotas are reallocated and sensor signal strength is adjusted. The warning sequence is then prioritized according to signal strength to obtain an optimized warning signal sequence.

[0068] In step S71, the high-risk area location information is mapped to a preset topology connection, and the location information is parsed to generate linkage control commands.

[0069] It should be noted that the preset topology connection refers to the node control link in the network topology structure, which includes the correspondence between device nodes and control modules. Control modules include circuit breakers, cooling fans, sprinkler systems, smoke detectors, etc. The mapping method is to query the corresponding control module from the control link database based on the node ID of the high-risk area. For example, node A corresponds to circuit breaker "control-001", cooling fan "fan-001", and sprinkler system "sprinkler-001". The linkage control command is generated according to the risk level. High-risk areas trigger the "circuit breaker ready to trip + fan start + sprinkler ready to start" command. The circuit breaker ready to trip is set with a 30-second delay (to facilitate emergency response by maintenance personnel), the fan start speed is set to 3,000 rpm (to ensure rapid cooling), and the sprinkler ready to start is in a pressurized standby state (it will automatically start if the temperature continues to rise). Medium-risk areas trigger the "fan start + enhanced monitoring" command, with the fan speed set to 2,000 rpm and enhanced monitoring increasing the sensor sampling frequency to 20 times per second. The instruction format includes "Control Module ID - Instruction Type - Execution Parameters - Trigger Condition", for example, "Control-001, Prepare for Trip, 30-second Delay, Cumulative Sensitivity > 200; Fan-001, Start, Speed ​​3000 RPM, No Delay; Spray-001, Prepare for Start, Pressure 0.8MPa, Temperature > 50℃". This instruction can accurately respond to risk management needs, ensuring the pertinence and effectiveness of control measures.

[0070] In step S72, the linkage control command is sent to the controlled node, and the node load status and response delay data are collected to obtain the feedback dataset.

[0071] It should be noted that the command issuance uses Industrial Ethernet (Profinet protocol), which features high real-time performance and reliability with a transmission latency of less than 100 milliseconds, meeting the real-time requirements of electrical control. After receiving the command, the controlled node immediately returns an acknowledgment signal, which includes the command ID, reception time, and node status, ensuring successful command transmission. Node load status includes the current open / closed state and load rate of the circuit breaker, the operating power and status of the fan, and the pressure and standby status of the sprinkler system. For example, "Control-001, Closed, Load Rate 80%; Fan-001, Not Running, Load Rate 0%; Sprinkler-001, Standby, Pressure 0.6MPa". The response latency is the difference between the command issuance timestamp and the acknowledgment signal reception timestamp. For example, if the command issuance time is 10:15:00.000 and the acknowledgment signal reception time is 10:15:00.050, the response latency is 50 milliseconds. The feedback dataset is stored in a structured format: "Control Module ID - Command Reception Status - Load Status - Response Delay". For example, "Control-001, Received, Closed / 80%, 50ms; Fan-001, Received, Not Running / 0%, 45ms; Sprinkler-001, Received, Standby / 0.6MPa, 48ms". This dataset provides real-time feedback data for weight matrix correction, reflecting the execution of control commands and the operating status of nodes.

[0072] In step S73, the traffic fluctuation characteristics and anomaly frequency are calculated based on the feedback dataset, and the weight matrix is ​​input to perform dynamic correction to obtain the updated weight matrix.

[0073] It should be noted that the flow fluctuation characteristics are calculated based on the current and voltage data of electrical nodes, including fluctuation amplitude and fluctuation frequency. The fluctuation amplitude is calculated by subtracting the minimum value from the maximum value of current or voltage within one hour. The fluctuation frequency is calculated by the number of times the current or voltage exceeds the normal range (±5% of the rated value) within one hour. For example, if the current fluctuation amplitude of node A decreases from 3.5A to 2.1A, the fluctuation frequency decreases from five times per minute to twice per minute. The anomaly occurrence frequency is the number of times abnormal indicators are triggered at nodes in high-risk areas within ten minutes. For example, if node A triggers temperature anomalies twice and voltage anomalies once, the anomaly occurrence frequency is three times. The weight matrix correction uses a proportional correction method. The correction coefficient is calculated as follows: 1 + (frequency of anomalies multiplied by 0.1) - (rate of decrease in traffic fluctuation multiplied by 0.2). The rate of decrease in traffic fluctuation is (original fluctuation rate minus current fluctuation rate) divided by the original fluctuation rate. For example, if the frequency of anomalies is three, the rate of decrease in traffic fluctuation is (3.5 - 2.1) divided by 3.5, resulting in 40%. The correction coefficient is 1 + (three times multiplied by 0.1) - (40% multiplied by 0.2), resulting in 1.22. The updated edge weights are calculated as the original edge weight multiplied by the correction coefficient. For example, the original edge weight between nodes A and B is 40.1; after the update, it is 40.1 multiplied by 1.22, resulting in approximately 48.9. This correction method allows the weight matrix to dynamically reflect the current operating status and anomalies of the nodes, improving the accuracy of subsequent analysis.

[0074] It is worth noting that after triggering the linkage response, in order to evaluate the effectiveness of control measures and dynamically optimize the monitoring model, the system introduces a feedback learning mechanism. In addition to recording the direct response status, it also monitors the dynamic changes in the electrical parameters of the controlled nodes and their associated lines, i.e., the flow fluctuation characteristics. This characteristic reflects the transient response of the electrical network under early warning and intervention actions. A stable network should tend to have reduced fluctuations and lower frequencies after intervention. Therefore, using the flow fluctuation characteristics and the frequency of anomalies together as the basis for adjusting the weight matrix can make the connection weights of the topology graph not only reflect the static physical connection strength and topological importance, but also dynamically adapt to the real-time operational health and stability of the network. Specifically, for connections where flow fluctuations are significantly reduced after the linkage response, their weights can be appropriately increased (meaning that the connection tends to be more stable and reliable); for connections where anomalies occur frequently and fluctuations intensify, their weights need to be weakened or they need to be kept under close scrutiny.

[0075] In step S74, resource quotas are reallocated and sensor signal strength is adjusted based on the updated weight matrix. The warning sequence is then prioritized according to signal strength to obtain an optimized warning signal sequence.

[0076] It should be noted that resource quotas include monitoring resources and control resources. Monitoring resources are the sensor acquisition frequency and data processing computing power, while control resources are the circuit breaker tripping priority and the start-up priority of fans and sprinklers. Nodes with higher weight connections are allocated more resources. For example, node A's sampling frequency is increased to 20 times per second, its data processing computing power allocation is increased to 35%, its circuit breaker tripping priority is set to 1 (the highest), and its fan start-up priority is set to 1. The sensor signal strength is adjusted by multiplying the signal transmission power by a correction factor (correction factor is 1.2) to ensure the stability of signal transmission in high-risk areas and avoid data loss due to signal attenuation. For example, node A's signal strength is increased by 20%. The priority of the warning sequence is sorted by multiplying the signal strength by the edge weight of the weight matrix to obtain the priority value. The higher the value, the higher the priority. For example, node A's warning priority value is 1.2 multiplied by 48.9, resulting in approximately 58.7; node B's warning priority value is 1.1 multiplied by 42.4, resulting in approximately 46.6. The optimized warning signal sequence is arranged from highest to lowest priority, for example, "Node A (priority 58.7) → Node B (priority 46.6) → Node C (priority 35.1)". This sequence ensures that warning signals from high-risk nodes are processed first, improving the timeliness and relevance of the warning response and providing strong protection for the safe operation of electrical networks.

[0077] Reference Figure 2 The second embodiment of the present invention provides an Internet of Things-based regional fire safety monitoring system, comprising: The data processing module is used to collect temperature, smoke, and electrical multi-source sensor data, remove invalid interference data through noise reduction and filtering, and generate a set of real-time electrical parameters. The topology construction module is used to map the connection relationship between electrical lines and equipment based on the real-time electrical parameter set, construct a weighted topology graph, quantify the electrical conduction strength between nodes with edge weights, and determine the network topology structure. The sensitivity assessment module is used to identify local electrical anomalies based on the network topology, and to calculate the contribution of each node anomaly to the overall security status by calculating the sensitivity score. The disturbance simulation module is used to simulate the topological propagation path of weak abnormal disturbances based on the node sensitivity score. If the disturbance intensity exceeds the preset disturbance intensity threshold, it is determined that the abnormal transmission effect is enhanced, and a list of potential diffusion areas is obtained. The risk ranking module is used to integrate multi-source sensor data with the potential diffusion area list, verify the cumulative evolution trend of anomalies, track and locate high-impact nodes under the transmission effect, and determine the priority ranking of risk areas. The positioning output module is used to generate a visual topology map based on the priority of the risk areas. If the cumulative value of the node sensitivity score exceeds the preset security cumulative threshold, the positioning information of the high-risk area is output. The early warning optimization module is used to trigger a linkage response mechanism based on the location information of the high-risk area, and dynamically update the weight parameters of the weighted topology map through data feedback closed loop to obtain an optimized early warning signal sequence.

[0078] It should be noted that the IoT-based regional fire safety monitoring system provided in this embodiment of the invention is used to execute all the process steps of the IoT-based regional fire safety monitoring method in the above embodiments. The two correspond one-to-one in terms of module function, working logic and technical effect, and have the same risk identification accuracy and prevention and control adaptability. Therefore, the specific working details and beneficial effects of the system can be referred to the foregoing method embodiments, and will not be repeated here.

[0079] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the electronic device. For example, the computer program can be divided into traffic processing instruction segments, graph construction instruction segments, cluster analysis instruction segments, etc., corresponding to functions such as traffic simplification, graph construction, and cluster determination, respectively.

[0080] The electronic device may be an edge computing gateway, an industrial control server, a desktop computer, or other device with data processing capabilities. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or use different components. For example, the electronic device may also include a data interface, a network module, an alarm module, etc. The data interface is used to connect the traffic acquisition component to the IoT device, the network module is used to realize data interaction between devices, and the alarm module is used to trigger audible, visual, or SMS alarms when abnormal communication patterns are detected.

[0081] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the electronic device through various interfaces and lines. It implements the network security protection data processing and decision generation functions of the electronic device by running or executing computer programs and / or modules stored in the memory, and by calling data stored in the memory.

[0082] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0083] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that 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 for those skilled in the art.

Claims

1. A regional fire safety monitoring method based on the Internet of Things, characterized in that, include: Collect multi-source sensor data on temperature, smoke, and electrical parameters, and after noise reduction and filtering to remove invalid interference data, generate a set of real-time electrical parameters. Based on the real-time electrical parameter set, the connection relationship between electrical lines and equipment is mapped, a weighted topology graph is constructed, and the electrical conduction strength between nodes is quantified by edge weight to determine the network topology. Based on the network topology, local electrical anomalies are identified, and the contribution of each node's anomaly to the overall security status is evaluated by calculating sensitivity, resulting in a node sensitivity score. The topological propagation path of weak abnormal disturbances is simulated based on the node sensitivity score. If the disturbance intensity exceeds the preset disturbance intensity threshold, the abnormal transmission effect is determined to be enhanced, and a list of potential diffusion areas is obtained. By combining the potential diffusion area list with multi-source sensor data, the trend of abnormal accumulation and evolution is verified, high-impact nodes under the transmission effect are tracked and located, and the priority ranking of risk areas is determined. A visual topology map is generated based on the priority of the risk areas. If the cumulative value of the node sensitivity score exceeds the preset security cumulative threshold, the location information of the high-risk area is output. Based on the location information of the high-risk area, a linkage response mechanism is triggered. The weight parameters of the weighted topology map are dynamically updated through data feedback in a closed loop to obtain an optimized early warning signal sequence.

2. The regional fire safety monitoring method based on the Internet of Things according to claim 1, characterized in that, The collected temperature, smoke, and electrical multi-source sensor data are processed through noise reduction and filtering to remove invalid interference data, generating a set of real-time electrical parameters, including: Collect raw message data from multiple sources including temperature, smoke, and electrical sensors, and parse the message data to obtain an initial sampling sequence with timestamps; The initial sampling sequence is matched with a preset filtering operator to identify noise components. If the noise component exceeds a preset abnormal offset threshold, the interference data is removed to obtain a clean sensing signal. The real-time state sequence is extracted by feature mapping of the pure sensing signal, and multi-dimensional data timing alignment processing is performed on the real-time state sequence to obtain a set of real-time electrical parameters.

3. The regional fire safety monitoring method based on the Internet of Things according to claim 1, characterized in that, The process of mapping the connection relationships between electrical lines and equipment based on the real-time electrical parameter set, constructing a weighted topology graph, quantifying the electrical conduction strength between nodes with edge weights, and determining the network topology includes: Extract the unique device identifier and voltage and current phasor data from the real-time electrical parameter set, and instantiate the unique device identifier as a graph model vertex to obtain a discrete electrical node set. Based on the voltage drop amplitude and current flow state between vertices in the discrete electrical node set, the equivalent impedance characteristic parameters are calculated, and the electrical conduction strength between nodes is quantified based on the equivalent impedance characteristic parameters. The electrical conduction intensity is converted into the edge weight values ​​of the graph, an initial weighted undirected graph is constructed, and the adjacency matrix is ​​obtained based on the edge weight distribution of the initial weighted undirected graph. If the weight elements in the adjacency matrix satisfy a preset connectivity threshold, the corresponding connecting edges are retained to form a sparse connected graph. The vertex connectivity and path loop characteristics of the sparse connected graph are analyzed to obtain the final network topology.

4. The IoT-based regional fire safety monitoring method according to claim 1, characterized in that, The process of identifying local electrical anomalies based on the network topology, evaluating the contribution of each node's anomaly to the overall security status by calculating sensitivity, and obtaining a node sensitivity score includes: Real-time node voltage and branch current instantaneous sampling data are collected, and the initial state vector of each node is extracted based on the admittance matrix operation. Calculate the deviation between the initial state vector and the historical baseline state value. If the deviation exceeds a preset node deviation threshold, it is marked as an abnormal fluctuation node. For the abnormal fluctuation nodes, partial derivative analysis is used to construct the node interaction sensitivity matrix to obtain the quantitative results of the mutual influence between nodes; The contribution weight of each node is calculated by combining the sensitivity matrix and the intensity of local anomalies, and the contribution weight is normalized to obtain the node sensitivity score.

5. The IoT-based regional fire safety monitoring method according to claim 1, characterized in that, The topological propagation path of weak anomalous disturbances is simulated based on the node sensitivity score. If the disturbance intensity exceeds a preset disturbance intensity threshold, the anomalous propagation effect is determined to be enhanced, and a list of potential diffusion regions is obtained, including: The node sensitivity scores are mapped to a preset adjacency matrix to construct a weighted topology graph with edge weight attributes; Weak anomalous perturbations are injected into the weighted topology graph to generate an initial state increment, and the real-time perturbation intensity of each associated node is calculated based on the initial state increment; If the real-time disturbance intensity exceeds a preset disturbance intensity threshold, an abnormal propagation path with enhanced conduction effect is extracted using a depth-first search algorithm; Based on the abnormal propagation path, the set of affected nodes at different topological depths is calculated to obtain a list of potential diffusion regions.

6. The regional fire safety monitoring method based on the Internet of Things according to claim 1, characterized in that, The process of integrating multi-source sensor data with the potential diffusion area list to verify the cumulative evolution trend of anomalies, track and locate high-impact nodes under the transmission effect, and determine the priority ranking of risk areas includes: Extract the multi-source sensor data corresponding to the potential diffusion region list and map it to a preset weighted topology structure to obtain a set of topology nodes; Calculate the numerical integral within the sliding time window for the set of topological nodes. If the integral value exceeds the preset abnormal integral threshold, generate an abnormal cumulative node list. An abnormal transmission effect subgraph is constructed using the list of abnormal accumulation nodes as the source point, and high-impact nodes with a middleness centrality higher than a preset centrality threshold are identified in the effect subgraph. Based on the abnormal accumulation magnitude and transmission impact range of the high-impact nodes, a comprehensive risk score is calculated to obtain the priority ranking of risk areas.

7. The IoT-based regional fire safety monitoring method according to claim 1, characterized in that, The process involves generating a visual topology map based on the priority of risk areas. If the cumulative value of a node's sensitivity score exceeds a preset security accumulation threshold, high-risk area location information is output, including: Obtain the physical spatial coordinates corresponding to the priority ranking of the risk areas, input the spatial coordinates into the rendering engine to perform topology mapping, and obtain the pixel matrix after topology mapping; The pixel matrix is ​​numerically filled using a node sensitivity weighting algorithm to construct a dynamic visualization layer with sensitivity weights. The sensitivity scores of nodes in the dynamic visualization layer are accumulated and calculated. If the accumulated value exceeds the preset safety accumulation threshold, the abnormal triggering state determination result is obtained. Based on the abnormal triggering state determination result, the high-risk area location information is extracted and converted into a standardized alarm message. The high-risk area is accurately marked on the visualization interface, and the high-risk area location information is output.

8. The IoT-based regional fire safety monitoring method according to claim 1, characterized in that, The linkage response mechanism triggered based on the high-risk area location information dynamically updates the weighted topology map weight parameters through data feedback closed-loop to obtain an optimized early warning signal sequence, including: The high-risk area location information is mapped to a preset topology connection, and the location information is parsed to generate linkage control commands. The linkage control command is sent to the controlled node, and the node load status and response delay data are collected to obtain the feedback dataset. Based on the feedback dataset, the characteristics of traffic fluctuations and the frequency of anomalies are calculated, and the weight matrix is ​​input to perform dynamic correction to obtain the updated weight matrix. Based on the updated weight matrix, resource quotas are redistributed and sensor signal strengths are adjusted. The warning sequences are then prioritized according to signal strength to obtain optimized warning signal sequences.

9. A regional fire safety monitoring system based on the Internet of Things, characterized in that, include: The data processing module is used to collect temperature, smoke, and electrical multi-source sensor data, remove invalid interference data through noise reduction and filtering, and generate a set of real-time electrical parameters. The topology construction module is used to map the connection relationship between electrical lines and equipment based on the real-time electrical parameter set, construct a weighted topology graph, quantify the electrical conduction strength between nodes with edge weights, and determine the network topology structure. The sensitivity assessment module is used to identify local electrical anomalies based on the network topology, and to calculate the contribution of each node anomaly to the overall security status by calculating the sensitivity score. The disturbance simulation module is used to simulate the topological propagation path of weak abnormal disturbances based on the node sensitivity score. If the disturbance intensity exceeds the preset disturbance intensity threshold, it is determined that the abnormal transmission effect is enhanced, and a list of potential diffusion areas is obtained. The risk ranking module is used to integrate multi-source sensor data with the potential diffusion area list, verify the cumulative evolution trend of anomalies, track and locate high-impact nodes under the transmission effect, and determine the priority ranking of risk areas. The positioning output module is used to generate a visual topology map based on the priority of the risk areas. If the cumulative value of the node sensitivity score exceeds the preset security cumulative threshold, the positioning information of the high-risk area is output. The early warning optimization module is used to trigger a linkage response mechanism based on the location information of the high-risk area, and dynamically update the weight parameters of the weighted topology map through data feedback closed loop to obtain an optimized early warning signal sequence.

Images