A heat station intelligent security alarm monitoring system
By integrating multimodal data fusion and unified security protection, the monitoring limitations and security blind spots of traditional heating station security systems have been solved, enabling efficient security monitoring and emergency response.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING KAICHENG ENG TECH CO LTD
- Filing Date
- 2026-05-18
- Publication Date
- 2026-07-31
AI Technical Summary
Traditional heating station security systems rely on single sensors and manual inspections, which have limitations in monitoring, and the separation of physical security from network security leads to security blind spots and high accident risks.
A multimodal recognition module is used for data collection and fusion, combined with deep learning algorithms and knowledge graphs for security monitoring, and a 3D visualization module is constructed for visualization display, realizing integrated protection of physical security and network security.
It reduces the false alarm rate and false negative rate of the system, enables early detection of potential equipment problems, quickly locates the fault point and root cause, improves the scientific nature and effectiveness of emergency response, and eliminates safety blind spots.
Smart Images
Figure CN122486718A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of heating station safety technology, specifically, it relates to an intelligent security alarm monitoring system for heating stations. Background Technology
[0002] As a core node in urban centralized heating systems, heating stations undertake the critical functions of heat conversion, distribution, and regulation. They contain specialized equipment such as high-temperature, high-pressure pipelines, valves, heat exchangers, and water pumps, and also involve electrical equipment, flammable and explosive media, and confined space working environments, posing extremely high safety risks.
[0003] Traditional heating station security systems mainly rely on single sensors and manual inspections, using independent video surveillance, temperature monitoring, or pressure monitoring. Each data point operates in isolation, resulting in limitations in monitoring. Furthermore, with the advancement of intelligent transformation of heating stations, a large number of industrial control devices are connected to the network. However, existing security systems generally manage physical security and network security separately. Network intrusions may lead to loss of control of industrial control devices, thereby causing physical security incidents. Damage to physical equipment may also affect the normal operation of the network system, creating security blind spots. Summary of the Invention
[0004] To address the aforementioned problems and technical deficiencies, this application adopts the following technical solution: an intelligent security alarm monitoring system for heating stations, comprising:
[0005] The multimodal recognition module performs multimodal data recognition and acquisition, and performs multimodal fusion on the acquired data;
[0006] The multi-source data analysis module is used for preprocessing and analyzing the data after multimodal fusion;
[0007] The security monitoring module establishes physical and network security protection models to perform physical security detection and network security monitoring.
[0008] The 3D visualization module constructs a 3D geometric model of the heating station and displays the monitoring results on the 3D geometric model of the heating station.
[0009] Preferably, the multimodal recognition module includes a visual recognition module and a multimodal fusion recognition module;
[0010] The visual recognition module includes:
[0011] Pipeline leak identification based on water stains on the pipe surface using deep learning algorithms;
[0012] The valve status is determined based on the identified instrument readings, thereby enabling equipment status identification;
[0013] It can determine whether personnel are wearing protective equipment, identify violations of operating procedures, and identify unauthorized entry into dangerous areas, thereby achieving personnel behavior identification;
[0014] Environmental anomalies are identified based on detected smoke, water accumulation, and debris buildup.
[0015] The multimodal fusion recognition module includes:
[0016] By fusing infrared thermal imaging data with visible light, the potential for equipment overheating can be detected.
[0017] By fusing sound and video, abnormal vibration faults in equipment can be identified;
[0018] By combining vibration and pressure, it can identify pipe blockages or internal leaks in valves.
[0019] Furthermore, the personnel behavior recognition is designed to meet the safety requirements of confined space operations at heating stations, enabling the identification of personnel wearing protective equipment, violations of operating procedures, and unauthorized entry into hazardous areas. It employs a method combining human key point detection and behavior sequence analysis to extract the coordinates of key skeletal points, construct a behavior feature vector, and identify personnel behavior sequences using a long short-term memory network. The formula for determining violations is as follows:
[0020] in, For the first The weight of each violation feature For the first The detection result for each violation characteristic is represented by 1 indicating its presence and 0 indicating its absence. For the sequence of key points of human skeleton, when When the value exceeds the threshold, it is determined to be a violation and an alarm of the corresponding level is triggered.
[0021] Preferably, the preprocessing includes:
[0022] The collected data is synchronized in time and aligned spatially.
[0023] Features from different modalities are extracted and weighted fusion is performed using an attention mechanism;
[0024] Decision-making is based on the DS evidence theory to fuse multiple identification results;
[0025] The confidence level of abnormal events is calculated by combining multimodal data, and an alarm is triggered when the confidence level exceeds a threshold.
[0026] Furthermore, the multi-source data analysis module includes a knowledge graph module, which constructs a safety knowledge graph for the heating station by acquiring entity relationships between equipment, personnel, environment, safety hazards and accident cases, and performs abnormal event association analysis and root cause tracing based on the knowledge graph, as well as risk assessment based on rule reasoning and machine learning.
[0027] Preferably, the physical security detection includes:
[0028] Based on video recognition, it can be determined whether there is unauthorized intrusion or malfunction. If so, the damage status of vibration sensors and video analysis equipment can be further assessed.
[0029] The analysis results are matched with the association rules between security events and valuation parameters to locate equipment failure points and analyze the causes of failures.
[0030] Network security monitoring first involves in-depth analysis and anomaly detection of industrial control protocols, then real-time monitoring and auditing of network traffic based on the detection results. If abnormal communication behavior is detected, network intrusion detection is performed based on the abnormal characteristics and behaviors. If a network intrusion event occurs, the risk assessment results are correlated with the network security event to generate a security risk assessment report.
[0031] Preferably, the safety monitoring module includes a simulation and contingency plan optimization module, which simulates the development process and impact range of different types of safety hazards, conducts simulation evaluation of different emergency response plans in a digital twin space, and optimizes the emergency response plans based on the simulation results.
[0032] Furthermore, the three-dimensional geometric model of the heating station maps equipment attributes, operating parameters, and the location and status of security equipment, providing a three-dimensional visualization of the heating station's security status, while also locating abnormal events, visually labeling them, and displaying a heat map of safety risks.
[0033] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the contents of a smart security alarm monitoring system for a heating station as described above.
[0034] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the contents of a smart security alarm monitoring system for a heating station as described above.
[0035] Compared to existing technologies, the beneficial effects of this application are as follows:
[0036] (1) This application reduces the false alarm rate of the system by deep fusion of multimodal data and decision fusion of DS evidence theory. By fusion of infrared and visible light, sound and video, and vibration and pressure, it can detect early hidden dangers of equipment in advance, realize predictive maintenance, reduce the occurrence of safety accidents, and reduce the false alarm rate and missed alarm rate.
[0037] (2) This application integrates physical security and network security protection, eliminates security blind spots, builds an integrated security protection system, and then uses knowledge graph-based abnormal event correlation analysis and root cause tracing to quickly locate fault points and root causes. Through digital twin simulation, it can quantitatively evaluate and optimize the effects of different emergency response plans, thereby improving the scientificity and effectiveness of emergency response. Attached Figure Description
[0038] In the attached diagram:
[0039] Figure 1 This is a schematic diagram of the system structure according to an embodiment of this application. Detailed Implementation
[0040] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of this application, but not all embodiments. Generally, the components of the embodiments of this application described and shown in the accompanying drawings can be arranged and designed in various different configurations.
[0041] Example 1, as Figure 1 As shown, a smart security alarm monitoring system for a heating station includes:
[0042] The multimodal recognition module performs multimodal data recognition and acquisition, and performs multimodal fusion on the acquired data;
[0043] The multimodal recognition module includes a visual recognition module and a multimodal fusion recognition module;
[0044] The visual recognition module includes:
[0045] Pipeline leak identification based on water stains on the pipe surface using deep learning algorithms;
[0046] The valve status is determined based on the identified instrument readings, thereby enabling equipment status identification;
[0047] It can determine whether personnel are wearing protective equipment, identify violations of operating procedures, and identify unauthorized entry into dangerous areas, thereby achieving personnel behavior identification;
[0048] Environmental anomalies are identified based on detected smoke, water accumulation, and debris buildup.
[0049] The multimodal fusion recognition module includes:
[0050] By fusing infrared thermal imaging data with visible light, the potential for equipment overheating can be detected.
[0051] By fusing sound and video, abnormal vibration faults in equipment can be identified;
[0052] By combining vibration and pressure, it can identify pipe blockages or internal leaks in valves.
[0053] Pipeline leak identification targets the characteristics of water stains and steam leaks on the surface of pipelines in heating stations. A dedicated dataset for pipeline leak scenarios in heating stations is constructed. A multi-scale feature fusion network is used to identify water stains and steam of different sizes and shapes. The input image is divided into S×S grids, and each grid predicts B bounding boxes and their corresponding confidence scores, as well as C class probabilities and leak confidence scores. The calculation formula is:
[0054] in, This represents the probability of a leak within the grid. To predict the intersection-union ratio (IU) between the bounding box and the true bounding box.
[0055] When the confidence level of a leak in a certain area exceeds the threshold, it is determined to be a pipeline leak, and the leak location and leak level are output.
[0056] The equipment status recognition uses a combination of optical character recognition and target detection to identify the readings of instruments such as pressure gauges, thermometers, and flow meters. Based on the readings, it determines the valve opening / closing status and equipment operating status, compares the readings with preset thresholds to determine whether the valve is fully open, fully closed, or in an intermediate state, and whether the equipment is operating normally.
[0057] Personnel behavior recognition addresses the safety requirements of confined space operations at heating stations by identifying the wearing of personal protective equipment, violations of operating procedures, and unauthorized entry into hazardous areas. It employs a combination of human key point detection and behavioral sequence analysis to extract the coordinates of key skeletal points and construct behavioral feature vectors. A long short-term memory network is then used to identify personnel behavior sequences. The formula for determining violations is as follows:
[0058] in, For the first The weight of each violation feature For the first The detection result for each violation characteristic is represented by 1 indicating its presence and 0 indicating its absence. For the sequence of key points of human skeleton, when When the value exceeds the threshold, it is determined to be a violation and an alarm of the corresponding level is triggered.
[0059] Environmental anomaly detection enables the detection of smoke and fire, water accumulation, and debris buildup.
[0060] Fireworks detection uses a combination of color and motion characteristics to distinguish real fireworks from interference such as lights and reflections.
[0061] Water accumulation detection is based on image segmentation algorithms to identify areas of water accumulation on the ground;
[0062] Debris accumulation detection identifies abnormal accumulations around fire lanes and equipment through background modeling.
[0063] The multi-source data analysis module is used for preprocessing and analyzing the data after multimodal fusion;
[0064] Preprocessing includes:
[0065] The collected data is synchronized in time and aligned spatially.
[0066] Features from different modalities are extracted and weighted fusion is performed using an attention mechanism;
[0067] Decision-making is based on the DS evidence theory to fuse multiple identification results;
[0068] The confidence level of abnormal events is calculated by combining multimodal data, and an alarm is triggered when the confidence level exceeds a threshold.
[0069] The multi-source data analysis module includes a knowledge graph module, which constructs a safety knowledge graph for heating stations by acquiring entity relationships between equipment, personnel, environment, safety hazards and accident cases. Based on the knowledge graph, it performs abnormal event correlation analysis and root cause tracing, and conducts risk assessment based on rule reasoning and machine learning.
[0070] The overheat detection method for devices that fuse infrared thermal imaging and visible light images employs a combination of pixel-level and feature-level fusion. First, the infrared and visible light images are synchronized temporally and aligned spatially. Then, pixel-level fusion is performed using a weighted average method to obtain the fused image. The pixel-level fusion formula is as follows:
[0071] in, For infrared images in Pixel value at that location, For visible light images in Pixel value at that location, For weight fusion.
[0072] Temperature features from infrared images and texture features from visible light images are extracted separately, and feature-level fusion is performed using a convolutional neural network to finally output the location, temperature, and overheating level of the overheated area of the device.
[0073] The sound and video fusion method for identifying abnormal vibration faults in rotating equipment such as water pumps and motors involves simultaneously acquiring sound and video signals from the equipment. The sound signals are used to extract acoustic features through Mel-frequency cepstral coefficients, while the video signals are used to extract vibration features through optical flow.
[0074] An attention mechanism fusion network is used to fuse the two features. The network structure includes an audio feature branch, a video feature branch, and an attention fusion layer. The attention fusion layer automatically learns the weights of different modal features, highlighting the features that are more important for fault identification.
[0075] The identification of pipe blockage or valve leakage through vibration and pressure fusion is based on the fact that pipe blockage or valve leakage can cause changes in pipe vibration characteristics and pressure distribution. Vibration and pressure signals are simultaneously acquired at different locations in the pipe, and the time-domain and frequency-domain characteristics of the vibration signals, as well as the rate of change characteristics of the pressure signals, are extracted.
[0076] A multi-input neural network is constructed, taking vibration and pressure features as inputs, and outputting the location, degree of blockage, and leakage rate of the valve.
[0077] Multimodal feature fusion employs an attention-based multimodal feature fusion method to extract deep features from different modalities and automatically learn the weights of each feature. The feature fusion formula is as follows:
[0078] in, For the first Feature vectors of each modality For the first Attention weights for each modality feature, .
[0079] The formula for calculating attention weights is:
[0080] in, and These are learnable parameters.
[0081] Decision fusion based on DS evidence theory addresses the problem that multiple recognition modules may output different recognition results. By using DS evidence theory for decision fusion, the reliability of the recognition results can be improved.
[0082] Let the recognition framework be ,in For the first One possible abnormal event;
[0083] Each recognition module outputs a Basic Probability Assignment (BPA) function. ,satisfy:
[0084] Multiple BPA functions are merged using Dempster's combination rules:
[0085] in, This represents the conflict coefficient.
[0086] After fusion, the event with the highest basic probability allocation is selected as the final identification result, and the confidence level of that event is calculated. When the confidence level exceeds a preset threshold, an alarm of the corresponding level is triggered.
[0087] The security monitoring module establishes physical and network security protection models to perform physical security detection and network security monitoring.
[0088] Physical security inspections include:
[0089] Based on video recognition, it can be determined whether there is unauthorized intrusion or malfunction. If so, the damage status of vibration sensors and video analysis equipment can be further assessed.
[0090] The analysis results are matched with the association rules between security events and valuation parameters to locate equipment failure points and analyze the causes of failures.
[0091] Network security monitoring first involves in-depth analysis and anomaly detection of industrial control protocols, then real-time monitoring and auditing of network traffic based on the detection results. If abnormal communication behavior is detected, network intrusion detection is performed based on the abnormal characteristics and behaviors. If a network intrusion event occurs, the risk assessment results are correlated with the network security event to generate a security risk assessment report.
[0092] The safety monitoring module includes a simulation and contingency plan optimization module, which simulates the development process and impact range of different types of safety hazards, conducts simulation evaluation of different emergency response plans in a digital twin space, and optimizes the emergency response plans based on the simulation results.
[0093] Physical security inspections include:
[0094] Based on the fusion of video recognition and vibration sensors, the method detects whether unauthorized personnel have entered the heating station and whether the equipment is malfunctioning. When the video identifies unauthorized personnel entering, it triggers the vibration sensor to collect vibration signals in the relevant area to further confirm whether there is personnel activity or equipment damage.
[0095] When potential equipment damage is detected, the damage status of the equipment is determined by combining video analysis and vibration signal analysis.
[0096] The analysis results are matched with a pre-established association rule base to locate the equipment failure point and analyze the cause of the failure. The association rule base is built based on historical failure data and expert experience and contains thousands of failure rules.
[0097] Network security monitoring includes:
[0098] We conduct in-depth analysis of commonly used industrial control protocols in heating stations, extract protocol field information, establish a baseline for normal communication behavior, and detect abnormal protocol messages.
[0099] Real-time monitoring and auditing of traffic on the industrial control network of the heating station, including statistics on traffic volume, direction, and protocol type. Establishing a normal traffic baseline and detecting abnormal network traffic.
[0100] Based on anomaly features and behavior analysis, network intrusion events are detected. A deep learning intrusion detection model is used to improve the ability to detect unknown attacks.
[0101] When a network intrusion incident occurs, the risk assessment results are correlated with the network security incident to assess the scope and severity of the incident's impact, generate a detailed security risk assessment report, and propose corresponding handling recommendations.
[0102] The 3D visualization module constructs a 3D geometric model of the heating station and displays the monitoring results on the 3D geometric model of the heating station.
[0103] The three-dimensional geometric model of the heating station maps equipment attributes, operating parameters, and the location and status of security equipment. It provides a three-dimensional visualization of the security status of the heating station, and also performs abnormal event location, visualization annotation, and display of safety risk heat maps.
[0104] Example 2, from a hardware perspective, this application provides an embodiment of an electronic device comprising all or part of an intelligent security alarm monitoring system for a heating station. The electronic device includes a service processor and a distributed memory. The service processor is connected to the memory. The distributed memory stores a service self-management program configured to store machine-readable instructions. The service processor executes the service self-management program. When the instructions are executed by the processor, an intelligent security alarm monitoring system for a heating station as described above can be implemented.
[0105] Example 3: This application also provides a computer-readable storage medium capable of implementing a smart security alarm monitoring system for a heating station with a server or client as the execution subject in the above embodiments. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements all the contents of the smart security alarm monitoring system for a heating station with a server or client as the execution subject in the above embodiments.
[0106] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications, improvements, and substitutions without departing from the concept of this application, and these all fall within the protection scope of this application.
Claims
1. A smart security alarm monitoring system for a heating station, characterized in that, include: The multimodal recognition module performs multimodal data recognition and acquisition, and performs multimodal fusion on the acquired data; The multi-source data analysis module is used for preprocessing and analyzing the data after multimodal fusion; The security monitoring module establishes physical and network security protection models to perform physical security detection and network security monitoring. The 3D visualization module constructs a 3D geometric model of the heating station and displays the monitoring results on the 3D geometric model of the heating station.
2. The intelligent security alarm monitoring system for a heating station according to claim 1, characterized in that, The multimodal recognition module includes a visual recognition module and a multimodal fusion recognition module; The visual recognition module includes: Pipeline leak identification based on water stains on the pipe surface using deep learning algorithms; The valve status is determined based on the identified instrument readings, thereby enabling equipment status identification; It can determine whether personnel are wearing protective equipment, identify violations of operating procedures, and identify unauthorized entry into dangerous areas, thereby achieving personnel behavior identification; Environmental anomalies are identified based on detected smoke, water accumulation, and debris buildup. The multimodal fusion recognition module includes: By fusing infrared thermal imaging data with visible light, the potential for equipment overheating can be detected. By fusing sound and video, abnormal vibration faults in equipment can be identified; By combining vibration and pressure, it can identify pipe blockages or internal leaks in valves.
3. The intelligent security alarm monitoring system for a heating station according to claim 2, characterized in that, The personnel behavior recognition system addresses the safety requirements of confined space operations at heating stations by identifying the wearing of personal protective equipment, violations of operating procedures, and unauthorized entry into hazardous areas. It employs a combination of human key point detection and behavior sequence analysis to extract the coordinates of key skeletal points and construct a behavior feature vector. A long short-term memory network is then used to identify personnel behavior sequences. The formula for determining violations is as follows: in, For the first The weight of each violation feature For the first The detection result for each violation characteristic is represented by 1 indicating its presence and 0 indicating its absence. For the sequence of key points of human skeleton, when When the value exceeds the threshold, it is determined to be a violation and an alarm of the corresponding level is triggered.
4. The intelligent security alarm monitoring system for a heating station according to claim 1, characterized in that, The preprocessing includes: The collected data is synchronized in time and aligned spatially. Features from different modalities are extracted and weighted fusion is performed using an attention mechanism; Decision-making is based on the DS evidence theory to fuse multiple identification results; The confidence level of abnormal events is calculated by combining multimodal data, and an alarm is triggered when the confidence level exceeds a threshold.
5. The intelligent security alarm monitoring system for a heating station according to claim 4, characterized in that, The multi-source data analysis module includes a knowledge graph module, which constructs a safety knowledge graph for heating stations by acquiring entity relationships between equipment, personnel, environment, safety hazards and accident cases. Based on the knowledge graph, it performs abnormal event association analysis and root cause tracing, and conducts risk assessment based on rule reasoning and machine learning.
6. The intelligent security alarm monitoring system for a heating station according to claim 1, characterized in that, The physical security detection includes: Based on video recognition, it can be determined whether there is unauthorized intrusion or malfunction. If so, the damage status of vibration sensors and video analysis equipment can be further assessed. The analysis results are matched with the association rules between security events and valuation parameters to locate equipment failure points and analyze the causes of failures. Network security monitoring first involves in-depth analysis and anomaly detection of industrial control protocols, then real-time monitoring and auditing of network traffic based on the detection results. If abnormal communication behavior is detected, network intrusion detection is performed based on the abnormal characteristics and behaviors. If a network intrusion event occurs, the risk assessment results are correlated with the network security event to generate a security risk assessment report.
7. The intelligent security alarm monitoring system for a heating station according to claim 6, characterized in that, The safety monitoring module includes a simulation and contingency plan optimization module, which simulates the development process and impact range of different types of safety hazards, conducts simulation evaluation of different emergency response plans in a digital twin space, and optimizes the emergency response plans based on the simulation results.
8. The intelligent security alarm monitoring system for a heating station according to claim 1, characterized in that, The three-dimensional geometric model of the heating station maps equipment attributes, operating parameters, and the location and status of security equipment, providing a three-dimensional visualization of the heating station's security status, as well as abnormal event location, visual annotation, and a safety risk heat map display.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the content of the intelligent security alarm monitoring system for a heating station as described in claim 1.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the content of the intelligent security alarm monitoring system for a heating station as described in claim 1.