Building fire-fighting facility leakage monitoring method and system

Abstract

The application discloses a kind of building fire-fighting facilities electric leakage monitoring method and system, it is related to fire safety monitoring technical field, including obtaining the power supply topology structure of target building fire-fighting facilities, and establishing initial electrical fingerprint atlas for each power supply loop;Electrical fingerprint atlas includes dynamic electric leakage baseline under different load conditions, transient response waveform envelope, load current harmonic distortion characteristics, and loop impedance-frequency characteristic curve and insulation dielectric loss factor frequency spectrum;Real-time acquisition of the electric leakage current time series signal of each power supply loop, load current time series signal, environmental temperature and humidity data, while synchronously collecting ultrasonic partial discharge signal and distributed optical fiber temperature measurement signal deployed at the key node of power supply loop.The application upgrades static threshold monitoring to dynamic, multidimensional, active intelligent early warning and control system, significantly reduces false alarm rate, realizes the early perception and active inhibition of trend risk, and guarantees the continuity and reliability of fire power supply.

Description

Technical Field

[0001] This invention relates to the field of fire safety monitoring technology, specifically to a method and system for monitoring leakage current in building fire protection facilities. Background Technology

[0002] Building fire protection facilities typically include fire pumps, smoke extraction fans, fire elevators, automatic fire alarm systems, emergency lighting equipment, and other fire-fighting linkage control equipment. These devices play a crucial role in fire monitoring, alarm systems, fire suppression, smoke extraction, and personnel evacuation. The continuity of their power supply and the reliability of their operation directly affect the overall emergency response capability of the building's fire protection system. To ensure the stable operation of building fire protection facilities, leakage current monitoring of their power supply circuits is usually required.

[0003] Currently, the mainstream solution for leakage current monitoring in fire protection power supply circuits is to set one or more fixed residual current action thresholds. When the detected leakage current exceeds the preset threshold, the system will issue an audible and visual alarm, or even directly cut off the faulty circuit. However, high-power equipment such as fire pumps and smoke exhaust fans inevitably generate large non-faulty transient leakage currents at the moment of startup, making the fixed threshold solution prone to false alarms. Furthermore, existing monitoring systems can only react to "events" that have reached the threshold, and are completely unable to detect "trend" risks caused by slow insulation aging, environmental moisture intrusion, etc., that have been below the threshold for a long time. To address the above problems, a method and system for monitoring leakage current in building fire protection facilities is proposed. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for monitoring leakage current in building fire protection facilities. This solves the problem that the mainstream approach for monitoring leakage current in fire protection power supply circuits still involves setting one or more fixed residual current action thresholds. When the detected leakage current exceeds the preset threshold, the system will issue an audible and visual alarm, or even directly cut off the faulty circuit. However, high-power equipment such as fire pumps and smoke exhaust fans inevitably generate significant non-faulty transient leakage currents at startup, making the fixed threshold approach prone to false alarms. Furthermore, existing monitoring systems can only react to "events" that have already reached the threshold, completely failing to detect "trend" risks caused by slow insulation aging, environmental moisture intrusion, etc., where the current remains below the threshold for an extended period.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a method for monitoring leakage current in building fire protection facilities, comprising the following steps: S100: Obtain the power supply topology of the target building's fire protection facilities and establish an initial electrical fingerprint spectrum for each power supply circuit; the electrical fingerprint spectrum includes the dynamic leakage current baseline, transient response waveform envelope, load current harmonic distortion characteristics, circuit impedance-frequency characteristic curve, and insulation dielectric loss factor spectrum under different load conditions. S200: Real-time acquisition of leakage current timing signals, load current timing signals, and ambient temperature and humidity data of each power supply circuit, while simultaneously acquiring ultrasonic partial discharge signals and distributed fiber optic temperature measurement signals deployed at key nodes of the power supply circuit; S300: Based on the leakage current timing signal and the ultrasonic partial discharge signal, extract high-frequency transient peak features, power frequency period modulation features, time-frequency domain energy entropy features, and partial discharge source location features based on multi-sensor fusion; S400: The extracted high-frequency transient spike features, power frequency periodic modulation features, energy entropy features, and partial discharge source location features are compared with the corresponding features in the electrical fingerprint map to perform multidimensional difference quantification, generating a prediction risk index for leakage current evolution trend. S500: Input the predicted risk index and the real-time status of each adjacent circuit in the power supply topology into a pre-constructed graph neural network to predict the propagation path and intensity of leakage risk along the power supply topology, and perform propagation correction on the predicted risk index. S600: When the predicted risk index exceeds the preset intervention threshold but is lower than the protection action threshold, suppressive intervention measures are executed according to the task priority and current operating redundancy of the load carried by the power supply circuit. The suppressive intervention measures include adjusting the reactive power compensation parameters of the power supply circuit, controlling the injection of dynamic compensation current into the active filter, temporarily switching to harmonic suppression mode, and adjusting the output harmonic characteristics of the distributed energy storage unit.

[0006] Preferably, the process of establishing the dynamic leakage baseline in the electrical fingerprint spectrum includes: Obtain normal leakage current data of the power supply circuit within a historical continuous time window. The normal leakage current data has been filtered out for non-fault disturbances caused by normal start-up and shutdown of fire-fighting equipment, motor commutation, or frequency converter adjustment. The distributed fiber optic temperature measurement signal is converted into the conductor temperature of the power supply circuit, and together with the load rate, ambient temperature, and equipment commissioning time, it is used as an auxiliary feature. This feature is then input into a long short-term memory network based on an attention mechanism to perform time series modeling, generating a dynamic baseline interval that adaptively adjusts with the load rate, conductor temperature, and equipment commissioning time. The initial parameters of the long short-term memory network are obtained through transfer learning using a baseline model trained on a similar type of fire protection power supply circuit.

[0007] Preferably, when a new fire-fighting device is connected to the power supply circuit or when the existing device is overhauled, a first-time learning mode is automatically triggered. In the first-time learning mode, the power supply circuit is controlled to execute a set of preset standard load sequences and a simulated partial discharge pulse with known parameters is injected. The electrical response characteristics and ultrasonic response characteristics under the sequence are collected. An incremental learning algorithm based on elastic weight consolidation is used to prevent catastrophic forgetting of the characteristic knowledge of the old circuit. The collected features are used to incrementally update the electrical fingerprint spectrum and the dynamic baseline interval.

[0008] Preferably, the extraction of high-frequency transient peak features and partial discharge source localization features includes: The leakage current timing signal and the synchronously acquired ultrasonic partial discharge signal are respectively subjected to noise-assisted multivariate empirical mode decomposition to obtain their respective intrinsic mode function components. The first k intrinsic mode function components representing high-frequency noise components are selected from the leakage current signal, their instantaneous energy accumulation curves are calculated, and the amplitude density and repetition frequency of the spike pulses generated by partial discharge of the insulating medium are identified. The high-frequency component of the leakage current and the corresponding narrowband component of the ultrasonic signal are subjected to time-frequency coherence analysis to extract the phase-locked value. Based on the time difference and attenuation characteristics of the ultrasonic signal arriving at different sensors, the spatial position of the partial discharge power source in the power supply circuit is calculated to form the positioning feature of the partial discharge power source.

[0009] Preferably, the implementation of the inhibitory intervention measures includes: Determine whether a controllable reactive power compensation device or active filter exists in the power supply circuit. If it does, and the digital twin simulation results show that the power supply voltage quality of the key fire-fighting equipment meets the requirements after intervention, then issue a command to the active power filter to inject a compensation current into the circuit with a phase opposite to the original leakage current signal and an amplitude dynamically optimized by model prediction control, so as to offset part of the capacitive leakage current component and specific harmonic leakage current. If a distributed energy storage unit is connected in the circuit, control the energy storage converter to output a preset low-order harmonic current to actively suppress the additional leakage current spikes caused by load harmonic distortion.

[0010] Preferably, it also includes protective switching measures. When the corrected predicted risk index exceeds the protection action threshold, if there is a physically redundant backup power supply channel in the power supply circuit, the load will be switched to the backup power supply channel without interruption. If there is no physically redundant channel, but the power supply circuit is part of a multi-input power supply system, the solid-state power switch will be controlled by a dynamic reconfiguration strategy based on deep reinforcement learning to dynamically redistribute the critical load on the circuit to an adjacent low-risk circuit. The deep reinforcement learning takes minimizing the probability of failure of critical fire protection functions and the transient impact of switching as the joint reward objective.

[0011] Preferably, the dynamic load redistribution includes: Obtain the current load rate of adjacent power supply circuits, real-time predicted risk index, and risk propagation sensitivity derived from graph neural network; A load redistribution optimization model is constructed with the dual objectives of minimizing the overall system residual risk and maximizing the availability of fire-fighting linkage. The load redistribution optimization model is solved using a multi-objective particle swarm optimization algorithm to generate dynamic reconfiguration instructions that include switching sequence, switching time, and solid-state power switch soft switching duty cycle curves.

[0012] Preferably, the current operational redundancy of the fire protection facilities is a weighted coefficient determined comprehensively based on the fire alarm status of the area where the facilities are located, the execution stage of the fire linkage plan, the online status of backup equipment, and the remaining integrity life predicted by the insulation aging trend. Specifically, when the area is in the unattended inspection stage and the remaining integrity life of the equipment is greater than a preset threshold, the redundancy weight coefficient is set to the first low value to prioritize triggering inhibitory intervention; when the area is in the fire confirmation and linkage stage, the redundancy weight coefficient is set to the second high value to prioritize maintaining power supply continuity.

[0013] Preferred features also include digital twin-enhanced human-computer interaction and situational simulation: In the visualization interface of the fire control center, the predicted risk index and evolution trend of each power supply circuit are dynamically rendered using the power supply topology as the framework. In response to operator clicks on a specific circuit, the system simulates and displays the evolution path of leakage current risk and the prediction of equipment functional integrity over a future period under three different strategies: continuous operation, implementation of inhibitory intervention, or implementation of protective switching.

[0014] A building fire protection facility leakage current monitoring system is provided to implement the above-mentioned building fire protection facility leakage current monitoring method. The system includes: The fingerprint modeling module is used to obtain the power supply topology of building fire protection facilities and to establish an initial electrical fingerprint map for each power supply circuit, including circuit impedance-frequency characteristics, insulation dielectric loss factor spectrum, and dynamic leakage baseline. The signal acquisition and feature extraction module is used to acquire leakage current timing signals, load current timing signals, ultrasonic partial discharge signals, fiber optic temperature measurement signals and environmental data in real time, and extract high-frequency transient spike features, power frequency periodic modulation features, time-frequency domain energy entropy features and partial discharge source location features. The risk prediction engine has a built-in multidimensional differential measurement unit, graph neural network risk propagation prediction unit, and digital twin simulation unit, which are used to generate a corrected leakage event prediction risk index and deduce the effect of intervention strategies. The strategy decision-making module is used to decide and select the optimal inhibitory intervention or protective switching strategy based on the revised predicted risk index, task priority, and the current operational redundancy of the facility corrected by the remaining insulation life. The execution control module is used to issue commands to the reactive power compensation device, active filter, solid-state power switch and distributed energy storage converter to execute the inhibitory intervention measures or protective switching measures.

[0015] Compared with existing technologies, the advantages of this invention are as follows: This invention establishes an electrical fingerprint spectrum containing a dynamic leakage baseline for each power supply circuit. This baseline is adaptively generated by fusing conductor temperature, load rate, ambient temperature and humidity, and equipment commissioning time through an attention LSTM network. It can accurately distinguish between non-faulty transient leakage current caused by the start-up and shutdown of high-power equipment such as fire pumps and smoke exhaust fans and actual insulation degradation leakage current. This fundamentally solves the problem of frequent false alarms in fixed threshold schemes, ensuring the accuracy and reliability of alarms. By synchronously collecting multi-dimensional signals such as leakage current, ultrasonic partial discharge, and distributed fiber optic temperature measurement, it extracts high-frequency transient spikes, power frequency modulation, energy entropy, and partial discharge source location features, and performs multi-dimensional difference quantification with the electrical fingerprint spectrum to generate a continuous predictive risk index. This index can sensitively reflect the slowly evolving fault trends such as insulation aging and moisture intrusion. It can issue an early warning before the leakage current reaches the traditional threshold, realizing proactive early warning from "post-event alarm" to "pre-event prediction". In summary, this invention upgrades static threshold monitoring into a dynamic, multi-dimensional, and proactive intelligent early warning and control system, significantly reducing the false alarm rate, realizing early perception and proactive suppression of trend risks, and ensuring the continuity and reliability of fire protection power supply. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the method for monitoring leakage current in building fire protection facilities provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the building fire protection facility leakage current monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0017] In the description of this invention, it should be noted that the terms "front", "up", "down", "left", "right", "vertical", "horizontal", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0018] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0019] Reference Figure 1 As shown, a method for monitoring leakage current in building fire protection facilities includes the following steps: S100: Obtain the power supply topology of the target building's fire protection facilities and establish an initial electrical fingerprint map for each power supply circuit. Specifically, the power supply topology of the target building's fire protection facilities is obtained by importing building electrical design drawings or through on-site scanning and identification, clarifying the connection relationship, load type, and interrelationships of each power supply circuit; electrical parameters of the corresponding power supply circuits are collected under different load conditions, including no-load, half-load, full-load, and fire equipment start-stop transition conditions; based on the collected electrical parameters, dynamic leakage current baseline, transient response waveform envelope, load current harmonic distortion characteristics, circuit impedance-frequency characteristic curve, and insulation dielectric loss factor spectrum are generated under different load conditions; all the above features are integrated to form the initial electrical fingerprint map of the power supply circuit and stored in the system database.

[0020] The process of establishing a dynamic leakage current baseline in the electrical fingerprint spectrum is as follows: Normal leakage current data of the power supply circuit within a historical continuous time window is acquired, and non-fault disturbance data caused by normal start-up and shutdown of fire-fighting equipment, motor commutation, or inverter regulation are pre-selected; the distributed fiber optic temperature measurement signal is converted into the conductor temperature of the power supply circuit, and conductor temperature, load rate, ambient temperature, and equipment commissioning time are used as auxiliary features, input into a long short-term memory network based on an attention mechanism for time series modeling, generating a dynamic baseline interval that adaptively adjusts with load rate, conductor temperature, and equipment commissioning time; the initial parameters of the long short-term memory network are obtained through transfer learning from a baseline model trained on a similar type of fire-fighting power supply circuit, in order to shorten the model training cycle and improve baseline prediction accuracy.

[0021] When new fire-fighting equipment is connected to the power supply circuit or when existing equipment undergoes major repairs, the system automatically triggers a first-time learning mode. In this mode, the system controls the power supply circuit to execute a set of preset standard load sequences and injects simulated partial discharge pulses with known parameters. Simultaneously, it collects the electrical and ultrasonic response characteristics under this sequence. An incremental learning algorithm based on elastic weight consolidation is used to process the newly collected features, preventing catastrophic forgetting of old circuit feature knowledge. The collected new features are then used to incrementally update the electrical fingerprint spectrum and dynamic baseline interval.

[0022] S200: Real-time acquisition of leakage current timing signals, load current timing signals, and ambient temperature and humidity data for each power supply circuit. Simultaneously, it acquires ultrasonic partial discharge signals and distributed fiber optic temperature measurement signals deployed at key nodes of the power supply circuit. Specifically, residual current transformers are installed at the incoming end of each power supply circuit to acquire leakage current timing signals in real time; current transformers are installed on each fire equipment load branch to acquire current timing signals of each load in real time; temperature and humidity sensors are installed in the power distribution room, cable trays, and key power distribution nodes to acquire ambient temperature and humidity data in real time; ultrasonic partial discharge sensors are deployed at key nodes prone to partial discharge, such as cable joints, switch contacts, transformer windings, and busbar connections in the power supply circuit; distributed fiber optic temperature measurement sensors are laid along the entire length of the power supply cable to achieve continuous monitoring of the cable temperature; all acquired signals are synchronously transmitted to the signal acquisition and feature extraction module of the monitoring system via wired industrial Ethernet or wireless IoT communication.

[0023] S300: Based on leakage current time-series signals and ultrasonic partial discharge signals, it extracts high-frequency transient peak features, power frequency cycle modulation features, time-frequency domain energy entropy features, and partial discharge source localization features based on multi-sensor fusion. The specific extraction process is as follows: noise-assisted multivariate empirical mode decomposition is performed on the leakage current time-series signal and the synchronously acquired ultrasonic partial discharge signal, decomposing the original signal into multiple intrinsic mode function components at different frequency scales; the first k components representing high-frequency noise components are selected from the intrinsic mode function components of the leakage current signal, and their instantaneous energy accumulation curves are calculated. By setting an adaptive energy threshold, peak pulses generated by partial discharge of the insulating medium are identified, and the amplitude density and repetition frequency of the peak pulses are statistically analyzed to form high-frequency transient peak features; the amplitude variation law and phase shift characteristics of the leakage current signal within the power frequency cycle are extracted to analyze the partial discharge signal. The modulation effect of the signal on the power frequency current forms a power frequency periodic modulation feature; short-time Fourier transform or wavelet transform is performed on the leakage current signal and the ultrasonic partial discharge signal to obtain the time-frequency distribution of the signal, and the energy distribution entropy value in different time-frequency intervals is calculated to form a time-frequency domain energy entropy feature; time-frequency coherence analysis is performed on the high-frequency component of the leakage current and the corresponding narrowband component of the ultrasonic signal to extract the phase lock value of the two; at the same time, based on the time difference and signal attenuation characteristics of the ultrasonic partial discharge sensor at different locations, the spatial position of the partial discharge source in the power supply circuit is calculated by the triangulation algorithm to form a partial discharge source positioning feature based on multi-sensor fusion.

[0024] S400: The extracted high-frequency transient spike features, power frequency periodic modulation features, energy entropy features, and partial discharge source location features are compared with the corresponding features in the electrical fingerprint map using multidimensional difference quantification to generate a predictive risk index for leakage current evolution trend. Specifically, the features extracted in real time are compared one by one with the corresponding normal operating condition features stored in the initial electrical fingerprint map; the difference degree of each feature dimension is calculated using methods such as Euclidean distance, cosine similarity, or Mahalanobis distance; weight coefficients are pre-assigned according to the influence of each feature dimension on the occurrence of leakage current faults, and the difference degrees of each dimension are weighted and summed to generate a predictive risk index that can comprehensively reflect the leakage current evolution trend; the predictive risk index ranges from 0 to 1, with higher values ​​indicating greater leakage current risk and more severe insulation degradation.

[0025] Let the real-time extraction of the first The features are The corresponding normal baseline features in the electrical fingerprint spectrum are The differences between each feature dimension are calculated using Mahalanobis distance: ,in Let be the historical covariance matrix of this feature. Then, the predicted risk index... The calculation formula is: ; In the formula, The preset weight coefficients for each feature dimension satisfy... , The function maps the weighted result to Interval.

[0026] S500: The predicted risk index and the real-time status of each adjacent circuit in the power supply topology are input into a pre-constructed graph neural network to predict the propagation path and intensity of leakage risk along the power supply topology, and the predicted risk index is corrected for propagation. Specifically, a graph structure of the power supply system is pre-constructed with power supply circuits as nodes and electrical connections between circuits as edges, and the graph neural network is trained using historical leakage fault data, enabling the model to learn the propagation law of leakage risk in the power supply topology; the currently generated predicted risk index, the real-time operating status of each adjacent circuit in the power supply topology, and the real-time risk index are input into the trained graph neural network; the graph neural network outputs the propagation path of the current leakage risk to adjacent circuits and the propagation intensity of each path; based on the predicted propagation intensity, the original predicted risk index is corrected to obtain the corrected predicted risk index, which can more accurately reflect the overall risk status of the entire power supply system and the possibility of fault propagation.

[0027] Let the risk propagation strength of the neighboring loop j to the current loop k in the graph neural network output be: The original predicted risk index for the current loop is The revised risk index The calculation is as follows: ; in, Let k represent the set of loops adjacent to loop k. Propagation attenuation factor (range of values) ), used to control the strength of risk coupling.

[0028] S600: When the predicted risk index exceeds the preset intervention threshold but is lower than the protection action threshold, suppressive intervention measures are implemented based on the task priority of the load on the power supply circuit and the current operating redundancy. Specifically, intervention thresholds and protection action thresholds are pre-set in the system, with the intervention threshold being lower than the protection action threshold. When the corrected predicted risk index exceeds the preset intervention threshold but is lower than the protection action threshold, the system first obtains the task priority of the load connected to the power supply circuit. The task priority is pre-divided according to the importance of the fire-fighting equipment. At the same time, the system obtains the current operational redundancy of the power supply circuit and its connected fire-fighting facilities. The current operational redundancy of the fire-fighting facilities is a weighted coefficient determined comprehensively based on the fire alarm status of the area where the facilities are located, the execution stage of the fire-fighting linkage plan, the online status of the backup equipment, and the remaining integrity life predicted by the insulation aging trend. Among them, when the area is in the unattended inspection stage and the remaining integrity life of the equipment is greater than the preset threshold, the redundancy weight coefficient is set to the first low value to prioritize triggering the inhibitory intervention. When the area is in the fire confirmation and linkage stage, the redundancy weight coefficient is set to the second high value to prioritize maintaining power supply continuity. The system selects the optimal inhibitory intervention measure from the preset intervention strategy library and executes it according to the load task priority and operational redundancy.

[0029] The specific implementation of suppressive intervention measures includes: determining whether there is a controllable reactive power compensation device or active filter in the power supply circuit; if so, and the digital twin simulation results show that the power supply voltage quality of key fire-fighting equipment meets the requirements after intervention, then issuing a command to the active power filter to inject a compensation current into the circuit with a phase opposite to the original leakage current signal and an amplitude dynamically optimized by model prediction control, so as to offset part of the capacitive leakage current component and specific harmonic leakage current; if there is a distributed energy storage unit in the circuit, then controlling the energy storage converter to output a preset low-order harmonic current to actively suppress the additional leakage current spikes caused by load harmonic distortion; in addition, the reactive power compensation parameters of the power supply circuit can be adjusted according to the actual situation, or the active filter can be temporarily switched to harmonic suppression mode.

[0030] The building fire protection facility leakage monitoring method in this embodiment also includes protective switching measures. When the corrected predicted risk index exceeds the protection action threshold, if there is a physically redundant backup power supply channel in the power supply circuit, the load will be switched to the backup power supply channel without interruption. If there is no physically redundant channel, but the power supply circuit is part of a multi-input power supply system, the solid-state power switch will be controlled by a dynamic reconstruction strategy based on deep reinforcement learning to dynamically redistribute the critical load on the circuit to an adjacent low-risk circuit. The deep reinforcement learning takes minimizing the failure probability of critical fire protection functions and the switching transient impact as the joint reward objective.

[0031] The dynamic load reallocation specifically includes: obtaining the current load rate of adjacent power supply circuits, real-time predicted risk index, and risk propagation sensitivity derived from graph neural networks; constructing a load reallocation optimization model with the dual objectives of minimizing the overall system residual risk and maximizing fire linkage availability; solving the load reallocation optimization model using a multi-objective particle swarm optimization algorithm to generate dynamic reconfiguration instructions that include switching sequence, switching time, and solid-state power switch soft switching duty cycle curves; and executing the control module to control the corresponding solid-state power switch actions according to the dynamic reconfiguration instructions to complete the dynamic load reallocation.

[0032] The building fire protection facility leakage monitoring method in this embodiment also includes digital twin-enhanced human-computer interaction and situation simulation. Specifically, in the visualization interface of the fire control center, the predicted risk index and evolution trend of each power supply circuit are dynamically rendered using lines of different colors and thicknesses, with the power supply topology as the framework. In response to the operator's click operation on a specific circuit, the system calls the digital twin simulation unit to simulate and display the leakage risk evolution path and equipment functional integrity prediction for a period of time under three different strategies: continuous operation, implementation of inhibitory intervention, or implementation of protective switching, providing an intuitive basis for the operator's decision-making.

[0033] This embodiment also provides a building fire protection facility leakage current monitoring system to implement the above-described building fire protection facility leakage current monitoring method, referring to... Figure 2 As shown, the system includes a fingerprint modeling module, a signal acquisition and feature extraction module, a risk prediction engine, a strategy decision-making module, and an execution control module.

[0034] The fingerprint modeling module is used to acquire the power supply topology of building fire protection facilities and establish an initial electrical fingerprint map for each power supply circuit, including circuit impedance-frequency characteristics, insulation dielectric loss factor spectrum, and dynamic leakage baseline. The fingerprint modeling module supports importing building electrical design drawings or automatically generating power supply topology maps through on-site scanning and recognition; it can control power supply circuits to execute preset standard load sequences and collect electrical parameters and ultrasonic response parameters under different load conditions; it has a built-in incremental learning unit that can update the electrical fingerprint map and dynamic baseline range using an incremental learning algorithm based on elastic weight consolidation after the connection of new equipment or after equipment overhaul.

[0035] The signal acquisition and feature extraction module is used to acquire leakage current time-series signals, load current time-series signals, ultrasonic partial discharge signals, fiber optic temperature measurement signals, and environmental data in real time, and extract high-frequency transient peak features, power frequency periodic modulation features, time-frequency domain energy entropy features, and partial discharge source location features. The module communicates with various sensors and includes a built-in signal preprocessing unit and feature extraction unit. The signal preprocessing unit filters, denoises, and synchronizes the acquired raw signals. The feature extraction unit extracts the aforementioned features from the preprocessed signals and transmits the extracted features to the risk prediction engine.

[0036] The risk prediction engine incorporates a multidimensional difference quantification unit, a graph neural network risk propagation prediction unit, and a digital twin simulation unit. These units generate a revised leakage event prediction risk index and deduce the effectiveness of intervention strategies. The multidimensional difference quantification unit receives feature data transmitted from the signal acquisition and feature extraction module, compares it with the electrical fingerprint spectrum stored in the fingerprint modeling module, calculates the multidimensional difference degree, and generates an initial predicted risk index. The graph neural network risk propagation prediction unit predicts the propagation path and intensity of leakage risk based on the power supply topology and the real-time status of each circuit, thus correcting the initial predicted risk index. The digital twin simulation unit constructs a digital twin model of the target building's fire protection power supply system, simulating the system's operating status under different intervention measures, deduce the effectiveness of intervention strategies, and provide a basis for strategy decision-making.

[0037] The strategy decision-making module is used to determine and select the optimal inhibitory intervention or protective switching strategy based on the revised predicted risk index, task priority, and the current operational redundancy of the facility corrected by the remaining insulation life. The strategy decision-making module has a built-in strategy library containing various inhibitory intervention and protective switching strategies; it can select the optimal execution strategy from the strategy library according to different risk levels, load priorities, and operational redundancy, and generate corresponding control commands to transmit to the execution control module.

[0038] The execution control module is used to issue commands to reactive power compensation devices, active power filters, solid-state power switches, and distributed energy storage converters to execute inhibitory intervention measures or protective switching measures. The execution control module communicates with various execution devices, converting control commands from the strategy decision module into standard signals that the execution devices can recognize, thereby controlling the execution devices to complete the corresponding actions.

[0039] The working principle of this invention is as follows: First, the system obtains the power supply topology of the target building's fire protection facilities through the fingerprint modeling module and establishes an initial electrical fingerprint map for each power supply circuit. Then, the signal acquisition and feature extraction module collects signals such as leakage current, load current, ambient temperature and humidity, ultrasonic partial discharge, and fiber optic temperature measurement of each power supply circuit in real time, and extracts various features that can reflect the insulation status from the collected signals. The risk prediction engine compares the extracted features with the electrical fingerprint map to generate an initial predicted risk index, and uses a graph neural network to predict the propagation path and intensity of leakage risk, and corrects the risk index. The strategy decision module decides the optimal intervention or switching strategy based on the corrected risk index, load task priority, and system operation redundancy. The execution control module controls the corresponding execution equipment to perform actions according to the instructions of the strategy decision module, realizing active suppression or protective switching of leakage risk. At the same time, the system uses a digital twin-enhanced human-machine interface to intuitively display the system's risk status and the deduction results of different strategies to the operator, assisting the operator in making decisions.

[0040] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for monitoring leakage current in building fire protection facilities, characterized in that, Includes the following steps: S100: Obtain the power supply topology of the target building's fire protection facilities and establish an initial electrical fingerprint spectrum for each power supply circuit; the electrical fingerprint spectrum includes the dynamic leakage current baseline, transient response waveform envelope, load current harmonic distortion characteristics, circuit impedance-frequency characteristic curve, and insulation dielectric loss factor spectrum under different load conditions. S200: Real-time acquisition of leakage current timing signals, load current timing signals, and ambient temperature and humidity data of each power supply circuit, while simultaneously acquiring ultrasonic partial discharge signals and distributed fiber optic temperature measurement signals deployed at key nodes of the power supply circuit; S300: Based on the leakage current timing signal and the ultrasonic partial discharge signal, extract high-frequency transient peak features, power frequency period modulation features, time-frequency domain energy entropy features, and partial discharge source location features based on multi-sensor fusion; S400: The extracted high-frequency transient spike features, power frequency periodic modulation features, energy entropy features, and partial discharge source location features are compared with the corresponding features in the electrical fingerprint map to perform multidimensional difference quantification, generating a prediction risk index for leakage current evolution trend. S500: Input the predicted risk index and the real-time status of each adjacent circuit in the power supply topology into a pre-constructed graph neural network to predict the propagation path and intensity of leakage risk along the power supply topology, and perform propagation correction on the predicted risk index. S600: When the predicted risk index exceeds the preset intervention threshold but is lower than the protection action threshold, suppressive intervention measures are executed according to the task priority and current operating redundancy of the load carried by the power supply circuit. The suppressive intervention measures include adjusting the reactive power compensation parameters of the power supply circuit, controlling the injection of dynamic compensation current into the active filter, temporarily switching to harmonic suppression mode, and adjusting the output harmonic characteristics of the distributed energy storage unit.

2. The method for monitoring leakage current in building fire protection facilities according to claim 1, characterized in that, The process of establishing the dynamic leakage current baseline in the electrical fingerprint spectrum includes: Obtain normal leakage current data of the power supply circuit within a historical continuous time window. The normal leakage current data has been filtered out for non-fault disturbances caused by normal start-up and shutdown of fire-fighting equipment, motor commutation, or frequency converter adjustment. The distributed fiber optic temperature measurement signal is converted into the conductor temperature of the power supply circuit, and together with the load rate, ambient temperature, and equipment commissioning time, it is used as an auxiliary feature. This feature is then input into a long short-term memory network based on an attention mechanism to perform time series modeling, generating a dynamic baseline interval that adaptively adjusts with the load rate, conductor temperature, and equipment commissioning time. The initial parameters of the long short-term memory network are obtained through transfer learning using a baseline model trained on a similar type of fire protection power supply circuit.

3. The method for monitoring leakage current in building fire protection facilities according to claim 2, characterized in that, When new fire-fighting equipment is connected to the power supply circuit or when existing equipment is overhauled, a learning mode is automatically triggered. In the learning mode, the power supply circuit is controlled to execute a set of preset standard load sequences and simulated partial discharge pulses with known parameters are injected. The electrical response characteristics and ultrasonic response characteristics under the sequence are collected. An incremental learning algorithm based on elastic weight consolidation is used to prevent catastrophic forgetting of the characteristic knowledge of the old circuit. The collected features are used to incrementally update the electrical fingerprint spectrum and the dynamic baseline interval.

4. The method for monitoring leakage current in building fire protection facilities according to claim 1, characterized in that, The extraction of high-frequency transient peak features and partial discharge source localization features includes: The leakage current timing signal and the synchronously acquired ultrasonic partial discharge signal are respectively subjected to noise-assisted multivariate empirical mode decomposition to obtain their respective intrinsic mode function components. The first k intrinsic mode function components representing high-frequency noise components are selected from the leakage current signal, their instantaneous energy accumulation curves are calculated, and the amplitude density and repetition frequency of the spike pulses generated by partial discharge of the insulating medium are identified. The high-frequency component of the leakage current and the corresponding narrowband component of the ultrasonic signal are subjected to time-frequency coherence analysis to extract the phase-locked value. Based on the time difference and attenuation characteristics of the ultrasonic signal arriving at different sensors, the spatial position of the partial discharge power source in the power supply circuit is calculated to form the positioning feature of the partial discharge power source.

5. The method for monitoring leakage current in building fire protection facilities according to claim 1, characterized in that, The implementation of inhibitory intervention measures includes: Determine whether a controllable reactive power compensation device or active filter exists in the power supply circuit. If it does, and the digital twin simulation results show that the power supply voltage quality of the key fire-fighting equipment meets the requirements after intervention, then issue a command to the active power filter to inject a compensation current into the circuit with a phase opposite to the original leakage current signal and an amplitude dynamically optimized by model prediction control, so as to offset part of the capacitive leakage current component and specific harmonic leakage current. If a distributed energy storage unit is connected in the circuit, control the energy storage converter to output a preset low-order harmonic current to actively suppress the additional leakage current spikes caused by load harmonic distortion.

6. The method for monitoring leakage current in building fire protection facilities according to claim 1, characterized in that, It also includes protective switching measures. When the corrected predicted risk index exceeds the protection action threshold, if there is a physically redundant backup power supply channel in the power supply circuit, the load will be switched to the backup power supply channel without interruption. If there is no physically redundant channel, but the power supply circuit is part of a multi-input power supply system, the solid-state power switch will be controlled by a dynamic reconfiguration strategy based on deep reinforcement learning to dynamically redistribute the critical load on the circuit to an adjacent low-risk circuit. The deep reinforcement learning takes minimizing the probability of failure of critical fire protection functions and the transient impact of switching as the joint reward objective.

7. The method for monitoring leakage current in building fire protection facilities according to claim 6, characterized in that, The dynamic load reallocation includes: Obtain the current load rate of adjacent power supply circuits, real-time predicted risk index, and risk propagation sensitivity derived from graph neural network; A load redistribution optimization model is constructed with the dual objectives of minimizing the overall system residual risk and maximizing the availability of fire-fighting linkage. The load redistribution optimization model is solved using a multi-objective particle swarm optimization algorithm to generate dynamic reconfiguration instructions that include switching sequence, switching time, and solid-state power switch soft switching duty cycle curves.

8. The method for monitoring leakage current in building fire protection facilities according to claim 1, characterized in that, The current operational redundancy of the fire protection facilities is determined by a weighted coefficient based on the fire alarm status of the area where the facilities are located, the execution stage of the fire linkage plan, the online status of backup equipment, and the remaining integrity life predicted by the insulation aging trend. Specifically, when the area is in the unattended inspection stage and the remaining integrity life of the equipment is greater than a preset threshold, the redundancy weight coefficient is set to the first low value to prioritize triggering inhibitory intervention. When the area is in the fire confirmation and linkage stage, the redundancy weight coefficient is set to the second high value to prioritize maintaining power supply continuity.

9. A method for monitoring leakage current in building fire protection facilities according to claim 1, characterized in that, It also includes digital twin-enhanced human-computer interaction and situational simulation: In the visualization interface of the fire control center, the predicted risk index and evolution trend of each power supply circuit are dynamically rendered using the power supply topology as the framework. In response to operator clicks on a specific circuit, the system simulates and displays the evolution path of leakage current risk and the prediction of equipment functional integrity over a future period under three different strategies: continuous operation, implementation of inhibitory intervention, or implementation of protective switching.

10. A building fire protection facility leakage current monitoring system, used to implement the building fire protection facility leakage current monitoring method as described in any one of claims 1-9, the system comprising: The fingerprint modeling module is used to obtain the power supply topology of building fire protection facilities and to establish an initial electrical fingerprint map for each power supply circuit, including circuit impedance-frequency characteristics, insulation dielectric loss factor spectrum, and dynamic leakage baseline. The signal acquisition and feature extraction module is used to acquire leakage current timing signals, load current timing signals, ultrasonic partial discharge signals, fiber optic temperature measurement signals and environmental data in real time, and extract high-frequency transient spike features, power frequency periodic modulation features, time-frequency domain energy entropy features and partial discharge source location features. The risk prediction engine has a built-in multidimensional differential measurement unit, graph neural network risk propagation prediction unit, and digital twin simulation unit, which are used to generate a corrected leakage event prediction risk index and deduce the effect of intervention strategies. The strategy decision-making module is used to decide and select the optimal inhibitory intervention or protective switching strategy based on the revised predicted risk index, task priority, and the current operational redundancy of the facility corrected by the remaining insulation life. The execution control module is used to issue commands to the reactive power compensation device, active filter, solid-state power switch and distributed energy storage converter to execute the inhibitory intervention measures or protective switching measures.

Images