Intelligent early warning fire extinguishing device and system based on electrical equipment

Abstract

This invention relates to the field of intelligent early warning and fire extinguishing equipment technology, specifically disclosing an intelligent early warning and fire extinguishing equipment and system based on electrical equipment. The system includes a multi-dimensional perception and early warning module, an edge intelligent data processing and analysis module, an intelligent hierarchical decision-making module, a multi-mode early warning execution module, a precise fire extinguishing control and execution module, a system linkage and feedback module, and a system self-checking and predictive maintenance module that is communicatively connected to all of the above modules. In this invention, a multi-dimensional perception architecture and multi-sensor fusion design replace a single perception mode, achieving accurate identification of potential hazards and early warning, significantly reducing false alarm rates and preventing managers from ignoring early warning information. An edge computing architecture enables local real-time data processing, eliminating reliance on the cloud, reducing transmission latency, ensuring rapid response in high-risk scenarios, and preventing system failures caused by network interruptions.

Description

Technical Field

[0001] This invention relates to the field of intelligent early warning fire extinguishing equipment technology, and in particular to intelligent early warning fire extinguishing equipment and systems based on electrical equipment. Background Technology

[0002] Electrical equipment is widely used in industrial production, building construction, power transmission and other fields. Its operational safety is directly related to the safety of people's lives and property and the stability of production and operation.

[0003] At the same time, existing intelligent early warning and fire extinguishing technologies still have many limitations: First, the early warning methods are limited, with most systems using only a single temperature or smoke sensor without combining core operating parameters of electrical equipment (such as residual current, fault arc, and three-phase voltage and current), resulting in a high false alarm rate; Second, the data processing capabilities are weak, relying heavily on centralized cloud processing, which has bandwidth bottlenecks, transmission delays (generally exceeding 10 seconds), and single-point failure risks. Summary of the Invention

[0004] In view of the technical problems of existing intelligent early warning and fire extinguishing equipment mentioned in the background art, the present invention provides an intelligent early warning and fire extinguishing equipment and system based on electrical equipment.

[0005] The technical solution adopted in this invention is: an intelligent early warning and fire extinguishing system based on electrical equipment, comprising a multi-dimensional perception and early warning module, an edge intelligent data processing and analysis module, an intelligent hierarchical decision-making module, a multi-mode early warning execution module, a precise fire extinguishing control and execution module, a system linkage and feedback module, and a system self-checking and predictive maintenance module that is communicatively connected to all of the above modules; the multi-dimensional perception and early warning module is used to collect electrical parameters, environmental parameters, and fire characteristic data of electrical equipment and transmit them to the edge intelligent data processing and analysis module; the edge intelligent data processing and analysis module is used to preprocess the collected data, identify anomalies, and perform hierarchical analysis of potential hazards, and output the analysis results; the intelligent hierarchical decision-making module is used to conduct a comprehensive risk assessment based on the analysis results and multi-dimensional indicators, and output differentiated early warning and fire extinguishing decision instructions; the multi-mode early warning execution module is used to execute hierarchical multi-channel early warnings according to the decision instructions; the precise fire extinguishing control and execution module is used to execute precise fire extinguishing operations according to the decision instructions; the system linkage and feedback module is used to realize the coordinated linkage of each module and external equipment, and to provide feedback on the operating status data; the system self-checking and predictive maintenance module is used to perform periodic self-checks, fault prediction, and maintenance reminders for each module of the system.

[0006] The present invention is further configured such that the multi-dimensional perception and early warning module includes a MEMS integrated detection module, an electrical parameter sensor, an infrared thermal imaging sensor, a signal conditioning circuit, and a data acquisition module; the MEMS integrated detection module integrates temperature, smoke, humidity, and fault arc sensors; the electrical parameter sensor includes a residual current sensor, a three-phase current / voltage sensor, and an insulation resistance sensor; the multi-dimensional perception and early warning module adopts a multi-sensor data fusion algorithm and sets the fault arc detection threshold in combination with the 3σ principle.

[0007] The invention is further configured such that the edge intelligent data processing and analysis module includes an edge computing node, a data preprocessing unit, a GA-BP neural network identification unit, an anomaly detection unit, and an edge-cloud collaborative communication unit; the edge computing node has a built-in lightweight inference engine to realize local real-time data processing; the data preprocessing unit uses Min-Max normalization and sliding window denoising algorithms to process the collected data; the GA-BP neural network identification unit combines the global search capability of genetic algorithms with the local optimization capability of BP neural networks for accurate identification of hazard types and levels; the GA-BP neural network is trained and optimized using the Sigmoid activation function and the mean square error function.

[0008] The present invention is further configured such that the intelligent hierarchical decision-making module includes a risk level assessment unit, a decision logic unit, a strategy storage unit, and an emergency linkage triggering unit; the risk level assessment unit integrates four dimensions of indicators—hazard level, equipment importance, environmental factors, and fault arc characteristics—and uses the analytic hierarchy process (AHP) to determine the weight of each indicator and quantify the comprehensive risk level; the decision logic unit outputs four-level handling strategies based on the comprehensive risk level, corresponding to normal state, first-level warning, second-level warning, and third-level warning, respectively.

[0009] The present invention is further configured such that the precision fire extinguishing control and execution module includes a Venturi valve, a directional nozzle array, a delivery pipeline, a flow detection unit, a pressure detection unit, and a Venturi valve drive unit; the Venturi valve has a pressure adaptive adjustment function, adjusting the opening of the Venturi valve according to the pressure difference of the delivery pipeline and the target flow rate; the directional nozzle array consists of multiple independently controllable nozzles, each nozzle corresponding to a part of the internal components of the electrical equipment.

[0010] The present invention is further configured such that the system linkage and feedback module includes a power linkage unit, a monitoring linkage unit, an emergency platform linkage unit, an operating status feedback unit, a data storage unit, and a dual power redundancy unit; the power linkage unit dynamically adjusts the power cut-off delay time according to the decision command and the operating current of the electrical equipment to avoid equipment damage and electric shock risk; the dual power redundancy unit adopts a main power supply and a backup power supply, and automatically switches to the backup power supply when the main power supply voltage is lower than a set threshold for a duration of ≥0.5s; the operating status feedback unit adopts an error correction algorithm.

[0011] The present invention is further configured such that the multi-mode early warning execution module includes an audible and visual early warning unit, a remote push unit, an early warning intensity adjustment unit, an NB-IoT+LoRa dual-mode communication unit, and an early warning log recording unit; the early warning intensity adjustment unit can automatically adjust the audible and visual early warning intensity according to the on-site noise and light intensity; the dual-mode communication unit can monitor the packet loss rate of early warning information transmission in real time, and automatically switch the transmission mode when the packet loss rate exceeds 0.5%; the remote push unit supports three push channels: APP, SMS, and emergency platform, and matches the corresponding push method according to the early warning level.

[0012] The present invention is further configured such that the system self-inspection and predictive maintenance module includes a periodic self-inspection triggering unit, a hardware fault detection unit, an algorithm validity verification unit, a communication link detection unit, a fault level determination unit, a maintenance strategy generation unit, and a maintenance push unit; the system self-inspection and predictive maintenance module automatically calculates the optimal self-inspection cycle and performs a full-coverage self-inspection of all hardware, software algorithms, and communication links of the system; it adopts a multi-dimensional fault determination and component remaining service life prediction algorithm to generate and push differentiated maintenance strategies based on the fault level.

[0013] A further embodiment of the present invention is an intelligent early warning fire extinguishing device based on electrical equipment. The intelligent early warning fire extinguishing device includes a fire extinguishing medium storage tank and a control module. The output end of the fire extinguishing medium storage tank is fixedly connected to a delivery pipeline. A Venturi valve is installed in the delivery pipeline. A directional nozzle array is fixedly connected to the output end of the delivery pipeline, and the nozzles are arranged corresponding to the internal component positions of the electrical equipment. A pressure detection unit and a flow detection unit are respectively installed on the delivery pipeline for real-time acquisition of pressure and fire extinguishing medium flow data in the delivery pipeline. The control module is electrically connected to the Venturi valve, the directional nozzle array, the pressure detection unit, the flow detection unit, and the Venturi valve drive unit. It can receive decision commands from external systems, control the opening degree of the Venturi valve and the start and stop of the directional nozzle array, adjust the fire extinguishing medium spray volume according to pressure and flow data, and simultaneously provide feedback on the device's own operating status data.

[0014] The beneficial effects of this invention are as follows: By employing a multi-dimensional perception architecture and multi-sensor fusion design, replacing the single-sensor mode, this invention achieves accurate identification of potential hazards and early warning, significantly reducing false alarm rates and preventing managers from ignoring warning information. The edge computing architecture enables local real-time data processing, eliminating reliance on the cloud, reducing transmission latency, ensuring rapid response in high-risk scenarios, and preventing system failures caused by network interruptions. The precise fire suppression control design can accurately adjust the spray volume and range according to the fire situation, reducing waste of extinguishing agents and preventing damage to undamaged electrical equipment. Industrial-grade protection and redundancy design improve the system's environmental adaptability and reliability, extending trouble-free operating time. Fully automatic self-inspection and predictive maintenance functions reduce manual inspection costs, realizing a shift from passive emergency repairs to proactive prevention. Attached Figure Description

[0015] Figure 1 This is a top view of the structure of the present invention.

[0016] The diagram is marked as follows: 1. Fire extinguishing medium storage tank; 2. Delivery pipeline; 3. Venturi valve; 4. Directional nozzle array; 5. Electrical equipment. 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 is in conjunction with the appendix Figure 1 The present invention will be further described below.

[0019] To address the problems existing in the background technology, this application proposes the following technical solution: an intelligent early warning and fire extinguishing system based on electrical equipment, comprising a multi-dimensional perception and early warning module, an edge intelligent data processing and analysis module, an intelligent hierarchical decision-making module, a multi-mode early warning execution module, a precise fire extinguishing control and execution module, a system linkage and feedback module, and a system self-testing and predictive maintenance module that is communicatively connected to all of the above modules; the multi-dimensional perception and early warning module is used to collect electrical parameters, environmental parameters, and fire characteristic data of the electrical equipment 5 and transmit them to the edge intelligent data processing and analysis module; the edge intelligent data processing and analysis module is used to... The collected data undergoes preprocessing, anomaly identification, and hazard classification analysis, and outputs the analysis results. The intelligent classification decision-making module is used to conduct a comprehensive risk assessment based on the analysis results and multi-dimensional indicators, and output differentiated early warnings and fire extinguishing decision instructions. The multi-mode early warning execution module is used to execute graded multi-channel early warnings according to the decision instructions. The precise fire extinguishing control and execution module is used to execute precise fire extinguishing operations according to the decision instructions. The system linkage and feedback module is used to realize the coordinated linkage of various modules and external devices, and to provide feedback on the operating status data. The system self-check and predictive maintenance module is used to perform periodic self-checks, fault prediction, and maintenance reminders for each module of the system.

[0020] The modules in the above technical solution are explained as follows: The multi-dimensional perception and early warning module includes a MEMS integrated detection module, electrical parameter sensors, infrared thermal imaging sensors, signal conditioning circuits, and a data acquisition module. The MEMS integrated detection module integrates temperature, smoke, humidity, and fault arc sensors. The electrical parameter sensors include a residual current sensor, a three-phase current / voltage sensor, and an insulation resistance sensor. The multi-dimensional perception and early warning module uses a multi-sensor data fusion algorithm and combines the 3σ principle to set the fault arc detection threshold.

[0021] This module is the foundation for the system to realize hazard identification and fire early warning. Unlike the existing single sensor perception mode, it adopts a three-dimensional perception architecture built with electrical parameters, environmental parameters and fire characteristics. It integrates multiple high-precision sensors and MEMS integrated detection modules to simultaneously collect various key parameters and environmental and fire information during the operation of electrical equipment 5. This provides an accurate data source for subsequent data processing and decision-making. At the same time, through sensor error correction and redundancy design, the perception accuracy and reliability are improved.

[0022] Module Composition: The module consists of a MEMS integrated detection module (integrating temperature, smoke, humidity, and fault arc sensors), electrical parameter sensors (residual current sensor, three-phase current / voltage sensor, and insulation resistance sensor), an infrared thermal imaging sensor, sensor signal conditioning circuitry, and a data acquisition module. The MEMS detection module is manufactured using MEMS technology and can be directly embedded inside or into critical parts of the electrical equipment 5. The fault arc sensor is designed according to the IEC62606 standard, supports high-frequency sampling (10kHz), and can accurately capture series / parallel arcs caused by poor contact or insulation degradation. The infrared thermal imaging sensor uses non-contact detection to monitor the surface temperature distribution of the electrical equipment 5 in real time, capturing potential localized overheating hazards.

[0023] The MEMS integrated detection module adopts a modular integrated design, separating the sensor and supporting circuit. The infrared thermal imaging sensor has been upgraded to 320×240 pixels, 30fps frame rate, and temperature measurement accuracy of ±1°C, which can accurately detect local overheating of small components such as wiring terminals. The details are explained below: The electrical parameter sensor group is specifically designed to collect five core operating parameters of electrical equipment. It includes three types of targeted sensors: residual current sensors, three-phase current / voltage sensors, and insulation resistance sensors. The specific parameters and functions of each sensor are as follows: The residual current sensor uses a Rogowski coil structure, with a measurement range of 0~1000mA, an accuracy class of 0.5, and a response time ≤10ms. It can detect leakage current in the line in real time. When the leakage current exceeds the set threshold of 30mA by default, it can be adjusted via the local management terminal or cloud platform, and an abnormal signal will be output immediately. The three-phase current / voltage sensor uses a through-hole structure, with a current measurement range of 0~10... The 0A voltage measurement range is 0~500V with an accuracy class of 0.2. The frequency measurement range is 50Hz±10%. It can detect electrical parameters such as current, voltage, power, and power factor of three-phase lines in real time, and capture parameter changes corresponding to potential hazards such as overload current exceeding 1.2 times the rated value and short-circuit current exceeding 10 times the rated value. The insulation resistance sensor adopts a high-voltage pulse measurement method, with a measurement range of 1MΩ~1000MΩ, a measurement voltage of 500VDC, and an accuracy class of 1.0. It can detect the resistance value of the insulation layer of electrical equipment in real time. When the insulation resistance is lower than the set threshold of 1MΩ, it is judged as a potential insulation degradation hazard. All sensors adopt a waterproof and dustproof design with an IP65 protection rating. Installation can be completed without complicated operations. It can be fixed by clips or screws, which is convenient for on-site construction and later maintenance. It can accurately capture parameter changes corresponding to various potential hazards such as line overload, short circuit, leakage, and insulation degradation, providing core data support for hazard identification.

[0024] The infrared thermal imaging sensor employs a non-contact detection method, with a resolution of 160×120 pixels, a temperature measurement range of -20℃ to 200℃, a measurement accuracy of ±2℃, and a frame rate of 15fps. It can capture real-time images of the temperature distribution on the surface of electrical equipment. By analyzing temperature distribution differences, such as local temperatures exceeding the surrounding area by more than 50℃, it accurately identifies potential localized overheating hazards. It is particularly suitable for detecting hidden hazards such as loose wiring terminals and aging coils, overcoming the limitations of contact temperature measurement. The sensor signal conditioning circuit uses the INA128 instrumentation amplifier, effectively suppressing electromagnetic interference and power frequency interference. The amplification factor is adjustable from 1 to 1000 times, amplifying weak signals output by the sensor, such as mV-level voltage signals. Simultaneously, an RC low-pass filter circuit with a cutoff frequency of 1kHz filters out useless high-frequency interference signals, ensuring the stability and accuracy of the sensor signal. The high-speed data acquisition module uses an STM32L476 microcontroller as its core controller. The sampling frequency can be adaptively adjusted from 100Hz to 1MHz according to the sensor type. It has a built-in 2KB cache space to avoid data loss. At the same time, it connects to subsequent edge computing nodes through an SPI communication interface with a communication rate of 1Mbps to achieve real-time data transmission and ensure the timeliness of subsequent data processing.

[0025] Smoke detection is based on the principle of light scattering. The infrared emitting tube inside the sensor emits infrared light with a wavelength of 940nm. When the electrical equipment 5 produces smoke particles with a diameter of 0.1~10μm during the initial combustion, the smoke particles will scatter the light. The scattered light is received by the photoelectric receiving tube and converted into an electrical signal. The intensity of the scattered light is positively correlated with the smoke concentration. By detecting the intensity of the scattered light, the smoke concentration can be determined, thus realizing early fire identification. The smoke concentration detection threshold can be adjusted to a default of 0.1mg / m³.

[0026] Residual current and three-phase current / voltage detection are based on electromagnetic induction theory. When current flows through the coil inside the sensor, an induced electromotive force is generated. The magnitude of the induced electromotive force satisfies the formula: ,in For the induced electromotive force V, The number of coil turns. The magnetic flux change rate Wb / s is used to calculate the measured current and voltage values ​​by measuring the induced electromotive force, enabling real-time monitoring of electrical parameters. Fault arc detection is based on the current waveform characteristics of arc discharge. Fault arc current waveforms are irregular, have high peak values, and long durations. By acquiring the current waveform and extracting characteristic parameters, and comparing them with normal arc waveform parameters, normal arcs and fault arcs can be accurately distinguished. Simultaneously, the 3σ principle is introduced to set the fault arc detection threshold to avoid false alarms caused by sensor drift. Insulation resistance detection is based on Ohm's law. By applying a standard test voltage of 500VDC to the insulation layer of electrical equipment, the leakage current of the insulation layer is measured. And then through the formula Calculate the insulation resistance Infrared thermal imaging is used to assess the insulation performance of the insulation layer and prevent short-circuit fires caused by insulation deterioration. Based on the theory of thermal radiation, all objects radiate infrared radiation, and the intensity of this radiation is positively correlated with the object's temperature, following the Stefan-Boltzmann law. ,in The radiative exitance is W / m². The Stefan-Boltzmann constant is 5.67 × 10⁻ 8 W / (m²·K) 4 ), Given the absolute temperature of an object in K, by receiving the infrared radiation emitted by the object and converting it into an electrical signal, and then processing the signal to generate a visualized temperature distribution image, potential local overheating hazards in equipment can be identified.

[0027] This module, based on sensor perception theory, electromagnetic induction theory, light scattering theory, and fault arc identification theory, achieves comprehensive and thorough detection of potential hazards in electrical equipment through the collaborative work of different types of sensors. Temperature detection utilizes the PT100 platinum resistance thermometer principle, leveraging the characteristic of metal resistance changing with temperature to achieve accurate temperature measurement; smoke detection employs the light scattering principle, determining smoke concentration by detecting the intensity of scattering of incident light by smoke particles; residual current and three-phase current / voltage detection are based on the electromagnetic induction principle, acquiring electrical parameters through coil induction; fault arc detection is based on the current waveform characteristics of arc discharge, capturing fault arc signals through high-frequency sampling and pattern recognition; insulation resistance detection is based on Ohm's law, detecting leakage current by applying a test voltage to determine the insulation performance of electrical equipment; and infrared thermal imaging, based on the principle of thermal radiation, converts the surface temperature distribution of equipment into a visual image, identifying localized overheating areas.

[0028] To reduce the detection error of a single sensor, a multi-sensor redundancy design and error correction theory are adopted. Data fusion algorithms are used to integrate the detection data of each sensor to improve the sensing accuracy. At the same time, the 3σ principle is used to set the detection threshold to avoid false alarms caused by sensor drift and ensure the reliability of the sensing data.

[0029] Formula and explanation: PT100 temperature sensor resistance-temperature conversion formula: ; Formula function: Converts the resistance signal of the PT100 sensor into an actual temperature value, enabling accurate detection of the internal and surrounding temperature of electrical equipment 5, and is used to determine whether there is a risk of overheating in the equipment.

[0030] Letter meanings: For the PT100 sensor at temperature The unit of resistance value at that time: ; For PT100 sensor in The standard resistance value is fixed at . ; The first-order temperature coefficient of the PT100 sensor is fixed at 1. ; The second-order temperature coefficient of the PT100 sensor is fixed at 1. ; The actual unit of temperature measurement is: .

[0031] Formula for detecting smoke concentration by light scattering: ; Formula function: By detecting the scattering intensity of incident light by smoke particles, the smoke concentration is calculated to determine whether electrical equipment 5 has started to burn and produce smoke, thus enabling early fire identification.

[0032] Letter meanings: The unit for the intensity of light scattered by smoke particles: lx; The initial intensity of incident light is expressed in lx. The scattering coefficient of smoke particles is related to the size and type of smoke particles, and its value ranges from [value range missing]. ; Units for smoke concentration: ; The optical path length, i.e., the distance between the incident light and the received light, is fixed at 1. .

[0033] Residual current leakage detection formula: ; Formula function: Based on Kirchhoff's current law, it detects the vector sum of the three-phase currents to determine whether there is a potential leakage current in electrical equipment 5. Under normal circumstances, the vector sum of the three-phase currents is 0, and when there is a leakage current, the vector sum is equal to the residual current.

[0034] Letter meanings: The unit for residual current is mA, which is leakage current. , , The units for the current values ​​in a three-phase circuit are A, respectively. If the threshold is exceeded, it is considered a potential leakage current hazard.

[0035] Insulation resistance test formula: ; Formula function: To detect the resistance value of the insulation layer of electrical equipment, determine whether the insulation layer is aging or damaged, and avoid short circuit fires caused by insulation deterioration.

[0036] Letter meanings: The unit for insulation resistance of electrical equipment 5 is MΩ; The test voltage applied to the insulation layer is fixed at 500V, which complies with electrical safety testing standards; The unit for leakage current of the insulation layer is μA. The smaller the value, the worse the insulation performance and the higher the risk level.

[0037] The formula for the fault arc detection threshold is based on the 3σ principle; ; Formula function: Sets the detection threshold for fault arcs to avoid false alarms caused by normal arcs such as those caused by switch operation, and accurately identifies fault arc signals.

[0038] Letter meanings: The unit for fault arc current threshold is A; The average value of the normal arc current is expressed in A, and the range is 0.5~1A calibrated using historical data. The standard deviation of normal arc current is measured in A, with a range of 0.1 to 0.2 A. When the detected arc current consistently exceeds... If the duration is ≥5ms, it is judged as a potential fault arc.

[0039] Multi-sensor data fusion weighted average formula ; Formula function: It integrates detection data of five core parameters, namely temperature, smoke concentration, residual current, insulation resistance, and fault arc, to reduce the detection error of a single sensor and improve the accuracy of hazard identification.

[0040] Letter meanings: This is the combined detection value after fusion; The weighting coefficients for each sensor are set according to the sensor's accuracy. temperature, Smoke concentration Residual current, Insulation resistance Faulty arc; The standardized detection values ​​for each sensor range from 0 to 1 and are obtained through normalization.

[0041] Compared with existing technologies, this module combines fault arc detection (high-frequency sampling >10kHz) with local temperature measurement using infrared thermal imaging, breaking through the limitations of existing single temperature / smoke sensing. It can accurately identify five core hidden dangers in electrical equipment (such as insulation degradation and poor contact) and provide fire warnings 76 seconds in advance. Secondly, it adopts a MEMS integrated detection module, which is small in size and low in power consumption, and can be embedded inside small electrical equipment. This solves the problems of large size, inconvenient installation, and high power consumption of existing sensors. At the same time, through multi-sensor redundancy design and error correction, the false alarm rate is controlled to within 0.18 times / channel. Thirdly, it introduces the 3σ principle to set the fault arc detection threshold and combines it with a weighted average data fusion algorithm to effectively avoid false alarms caused by sensor drift and normal arcs.

[0042] In this embodiment, the edge intelligent data processing and analysis module includes an edge computing node, a data preprocessing unit, a GA-BP neural network identification unit, an anomaly detection unit, and an edge and cloud collaborative communication unit. The edge computing node has a built-in lightweight inference engine to realize local real-time data processing. The data preprocessing unit uses Min-Max normalization and sliding window denoising algorithms to process the collected data. The GA-BP neural network identification unit combines the global search capability of the genetic algorithm with the local optimization capability of the BP neural network for accurate identification of hazard types and levels. The GA-BP neural network is trained and optimized using the Sigmoid activation function and the mean square error function.

[0043] This module takes over the collection of multi-dimensional perception data. Unlike the existing centralized data processing mode in the cloud, it adopts an architecture of edge nodes and GA-BP neural network to realize real-time data processing, anomaly identification and hidden danger classification. At the same time, through lightweight inference and edge-cloud collaboration, it balances processing speed and system scalability, and solves the problems of high data processing latency, low identification accuracy and easy getting trapped in local optima in the existing technology.

[0044] Module Composition: The module consists of an edge computing node (based on an ARM Cortex-A series industrial computer), a data preprocessing unit, a GA-BP neural network recognition unit, an anomaly detection unit, and an edge-cloud collaborative communication unit. The edge computing node has a built-in lightweight inference engine (TensorFlow LiteMicro) that enables real-time local data processing without cloud reliance. The data preprocessing unit is responsible for normalizing and denoising the sensed data to eliminate interference signals. The GA-BP neural network recognition unit analyzes the processed data to identify the type and level of potential hazards. The anomaly detection unit uses the 3σ principle to quickly determine data anomalies. The edge-cloud collaborative communication unit uses an encrypted transmission protocol, uploading only alarm events, device status summaries, and key feature vectors to the central platform, significantly reducing uplink traffic, while simultaneously receiving model updates and policy configurations from the cloud.

[0045] This module leverages edge computing, deep learning, data preprocessing, and anomaly detection theories to achieve efficient processing and accurate analysis of perceived data. Edge computing theory decentralizes data processing tasks to local edge nodes, avoiding cloud transmission delays and bandwidth bottlenecks, thus achieving a local closed loop of "perception-analysis response" with response latency controlled within 1.8 seconds. Data preprocessing theory eliminates the influence of data of different magnitudes through normalization and removes environmental interference through sliding window denoising, improving data quality. GA-BP neural network theory combines the global search capability of genetic algorithms (GA) with the local optimization capability of BP neural networks, addressing the problems of random initial parameters, susceptibility to local optima, and slow convergence speed in traditional BP neural networks, thereby improving the accuracy of hazard identification. Anomaly detection theory, based on the 3σ principle, quickly determines whether data is abnormal by statistically analyzing the distribution characteristics of perceived data, enabling rapid hazard identification.

[0046] The data preprocessing unit is crucial for ensuring data quality. It is responsible for standardizing the collected multi-dimensional sensor data, eliminating interference signals and data biases, and providing high-quality data for subsequent neural network recognition. The preprocessing process mainly includes three core steps: denoising, normalization, and outlier removal. Denoising employs a dedicated algorithm to effectively filter out errors caused by random interference. Normalization uses a dedicated algorithm to map detection data of different magnitudes to the same interval, eliminating the impact of data magnitude differences on subsequent neural network training and recognition. Outlier removal follows specific principles, performing statistical analysis on the processed data to remove abnormal and abrupt data, ensuring data validity. The preprocessing unit has a high processing speed, meeting the data processing requirements of various sensor sampling frequencies, ensuring real-time data transmission and processing, and guaranteeing the timeliness of subsequent processes.

[0047] The neural network identification unit is the core of this module, responsible for deep analysis of preprocessed data to identify the type and level of hazards. This unit employs a dedicated optimized neural network architecture, using a genetic algorithm to optimize the initial parameters of traditional neural networks, addressing the problems of random initial parameters, susceptibility to local optima, and slow convergence speed in traditional neural networks. The neural network adopts a hierarchical structure: the input layer corresponds to standardized data for various core parameters, the hidden layer extracts complex hazard features, and the output layer corresponds to different operating states of the equipment, determining the current operating state and hazard level through output probability values. The identification unit incorporates a pre-trained neural network model, trained on a large amount of diverse sample data to ensure identification accuracy, precisely distinguishing between the normal operating state of the equipment and different types and levels of hazards.

[0048] The neural network identification unit adopts a GA-BP neural network architecture. It optimizes the initial weights and thresholds of the traditional BP neural network using a genetic algorithm (GA), addressing the problems of random initial parameters, susceptibility to local optima, and slow convergence speed in traditional BP neural networks. The GA-BP neural network employs a hierarchical structure: 5 neurons in the input layer → 10 neurons in the hidden layer → 4 neurons in the output layer. The input layer corresponds to the standardized data of the five core parameters in step 1: temperature, smoke concentration, residual current, insulation resistance, and fault arc. The hidden layer uses the Sigmoid activation function to extract complex hazard features. The output layer uses the Softmax activation function, corresponding to the four operating states of the equipment: normal, level 1 hazard, level 2 hazard, and level 3 hazard. The identification result is determined by the output probability value, which corresponds to the state with the highest probability value for the current operating state of the equipment and the hazard level. The identification unit has a built-in trained neural network model, which is trained on 10,000 sets of sample data, including 2,000 sets of normal state data, 2,500 sets of first-level hidden danger data, 2,500 sets of second-level hidden danger data, and 3,000 sets of third-level hidden danger data. The training iterations are 1,000 times, and the convergence error is ≤0.001, ensuring the identification accuracy. It can accurately distinguish the normal operating state of the equipment from different types and levels of hidden dangers such as overload, short circuit, leakage, insulation deterioration, local overheating, and fault arc, with a hidden danger identification accuracy of ≥99.21%.

[0049] To ensure the optimization effect of the neural network model, the training process of the GA-BP neural network is divided into two steps: First, a genetic algorithm is used to optimize the initial weights and thresholds. The parameters of the genetic algorithm are set as follows: population size of 50, crossover probability of 0.8, mutation probability of 0.05, and the fitness function is the mean squared error of the neural network. Optimal initial weights and thresholds are selected through selection, crossover, and mutation operations. Second, the optimized initial weights and thresholds are substituted into the BP neural network, and the backpropagation algorithm is used to update the weights. The weight update formula is as follows: , ,in This represents the weight update amount between the i-th neuron in the input layer and the j-th neuron in the hidden layer. For learning rate, Let j be the error term of the j-th neuron in the hidden layer. Let be the input value of the i-th neuron in the input layer. The derivative of the activation function. The number of neurons in the output layer. This is the error term for the k-th neuron in the output layer. The weights are the weights between the j-th neuron in the hidden layer and the k-th neuron in the output layer. Through multiple rounds of iterative training, the prediction error of the neural network is brought to a preset level.

[0050] The anomaly detection unit, based on a dedicated principle, enables rapid identification of data anomalies. As a supplement to neural network recognition, it avoids missed detections due to neural network inference delays, achieving "double verification." This unit receives the fused comprehensive detection data in real time, sets anomaly judgment threshold by statistically analyzing the characteristic parameters of historical normal data, and immediately triggers a rapid anomaly warning when the real-time comprehensive detection data exceeds the normal range. Simultaneously, the anomaly data is sent to the neural network recognition unit for further analysis to determine the type and level of the potential hazard. The anomaly detection unit has an extremely fast response speed, enabling rapid capture of anomaly data and ensuring no potential hazard is overlooked.

[0051] The anomaly detection unit uses the 3σ principle to quickly determine data anomalies, supplementing neural network recognition and avoiding missed detections due to neural network inference delays, thus achieving "double verification." This unit receives the integrated detection value fused in step 1 in real time. The average value of the comprehensive test was calculated by statistically analyzing ≥1000 historical normal data samples from the past 30 days. with standard deviation The calculation formulas are as follows: , This represents the number of historical normal data samples. Let be the comprehensive detection value of the i-th historical normal sample, and set the anomaly detection threshold as . When the real-time comprehensive detection value If the anomaly exceeds this range, a rapid anomaly warning is immediately triggered with a response time of ≤100ms. Simultaneously, the anomaly data is sent to the neural network recognition unit for further analysis to determine the type and level of the hazard. The anomaly detection unit has an extremely fast response speed, enabling rapid capture of anomaly data and ensuring that no hazard is overlooked.

[0052] The edge and cloud collaborative communication unit adopts a dual-mode communication design, supporting automatic switching between two communication modes. The communication protocol uses a lightweight, low-power dedicated protocol, and data encryption employs a dedicated encryption algorithm to ensure the security and confidentiality of data transmission. The core function of this unit is to enable collaboration between edge nodes and the cloud central platform. Edge nodes only upload key alarm information, device status summaries, and key characteristic data to the cloud, significantly reducing uplink traffic and avoiding bandwidth bottlenecks. Simultaneously, the cloud platform can issue various commands to edge nodes, enabling remote optimization of system algorithms and parameter adjustments, improving system scalability and adaptability. When the signal in one communication mode is weak, it automatically switches to the other, ensuring the stability of edge-cloud collaboration, meeting the communication needs of different scenarios, and guaranteeing the stability and timeliness of data transmission.

[0053] The edge-cloud collaborative communication unit adopts a dual-mode communication design of NB-IoT and LoRa, supporting automatic switching between the two communication modes. The communication protocol uses the lightweight, low-power dedicated MQTT protocol, and the data encryption method employs the national standard SM4 encryption algorithm to ensure the security and confidentiality of data transmission. The encryption key length is 128 bits, making it unbreakable. The core function of this unit is to enable collaboration between edge nodes and the cloud central platform. Edge nodes only upload key alarm information, device status summaries such as the operating status of each module, the current risk level, and key characteristic data such as abnormal parameter values ​​and fault arc waveforms to the cloud. The data upload frequency is adjustable, defaulting to once per minute, with real-time uploads upon alarm, significantly reducing uplink traffic. The amount of data uploaded per instance is ≤100 bytes, avoiding bandwidth bottlenecks. Simultaneously, the cloud platform can issue various commands to the edge nodes, including algorithm parameter updates such as neural network model upgrades, detection threshold adjustments, strategy configurations such as data fusion weight adjustments, and self-test commands, enabling remote optimization and parameter adjustment of the system algorithm, improving the system's scalability and adaptability. When the signal strength of one communication mode is weak (≤-80dBm), it automatically switches to another communication mode with a switching time of ≤1s, ensuring the stability of edge-cloud collaboration and meeting the communication needs of different scenarios. In urban areas, NB-IoT mode is preferred, while in remote areas, LoRa mode is preferred to ensure the stability and timeliness of data transmission.

[0054] Formula and explanation: Data normalization formula: Min-Max normalization: The formula's function is to standardize various collected sensor data, such as temperature, smoke concentration, and residual current, mapping them to the 0-1 range. This eliminates the influence of different data volumes, avoids neural network training bias due to data differences, and improves recognition accuracy. The letters represent: The data is standardized after normalization; This refers to the raw detection data from the sensor. This is the minimum detection range value of the sensor; The maximum detection range value of the sensor, such as a temperature sensor. , .

[0055] Sliding window denoising formula: The formula's purpose is to eliminate random interference in sensed data, such as electromagnetic interference and errors caused by environmental fluctuations, thereby improving data stability and reliability and providing high-quality data for subsequent analysis. The letters represent: The test data is after noise reduction; The length of the sliding window is fixed at 5, which means taking the average of 5 consecutive sampling points; For the first in the sliding window The original detection data of each sampling point.

[0056] The sigmoid function is the activation function for GA-BP neural networks. The formula's function is to introduce nonlinear characteristics, converting the input signal of the neural network into an output signal in the 0-1 range, enabling nonlinear identification of hazard types and levels, such as distinguishing between overheating, leakage, and short circuits. The letters represent: The output value of the activation function; The input signal of the neural network neuron is the product of the preprocessed data and the weights, plus a bias. The natural constant is approximately 2.718.

[0057] Mean squared error of GA-BP neural network error function: The formula's function is to calculate the deviation between the predicted output value of the neural network and the actual label value, guiding the genetic algorithm to optimize the initial weights and the backpropagation neural network to update the weights, thereby reducing recognition errors. The letters represent: The smaller the mean square error, the higher the recognition accuracy. This represents the number of training samples; For the first The actual label values ​​for each sample are as follows: "Normal" is 0, "Level 1 Hidden Danger" is 1, "Level 2 Hidden Danger" is 2, and "Level 3 Hidden Danger" is 3. For the first The neural network predicts the output value of each sample.

[0058] The weight update formula for the GA-BP neural network is backpropagation (BP): ; The formula's function is to adjust the weights of each layer in the neural network using the backpropagation algorithm, reducing the error function value and improving the accuracy of hazard identification. The Genetic Algorithm (GA) is used to optimize the initial weights, preventing the backpropagation neural network from getting trapped in local optima. The letters represent: For the input layer The first neuron and the hidden layer Weight update amount between neurons; The learning rate is fixed at 0.01 to control the speed of weight updates; For the hidden layer Error term of each neuron; For the input layer The input values ​​of each neuron; The derivative of the activation function ( ); This represents the number of neurons in the output layer. For the output layer Error term of each neuron; For the hidden layer The nth neuron and the output layer The weights between neurons.

[0059] based on The principle of anomaly detection formula: ; ; Anomaly detection: or The formula's function is to quickly determine whether the integrated detection value after fusion in step 1 is abnormal, enabling rapid early warning of potential risks and avoiding missed detections due to neural network inference delays.

[0060] Letter meanings: This is the average of the comprehensive test values ​​(calibrated using historical normal data). The standard deviation of the overall test values; This represents the sample size of historical normal data. For the first The combined test value of a historical normal sample; The current comprehensive detection value, when Exceeding If the data falls within the specified range, it is considered an anomaly, and an early warning process is immediately triggered.

[0061] Compared with existing technologies, this module has three main advantages: First, it adopts an edge computing architecture combined with a lightweight inference engine to achieve real-time local data processing, compressing response latency to less than 1.8 seconds, which is 78% faster than existing cloud processing systems. Simultaneously, through edge-cloud collaboration, it balances local response with global management, solving the problems of high latency and cloud dependence in existing technologies. Second, it introduces a GA-BP neural network, optimizing the initial weights of the BP neural network through a genetic algorithm, addressing the issues of traditional BP neural networks easily getting trapped in local optima and having slow convergence speed, thus improving the accuracy of hazard identification compared to traditional BP neural networks. Third, it integrates motion-free window denoising, 3σ principle anomaly detection, and GA-BP neural network recognition to form a "dual verification" mechanism, ensuring both the speed of anomaly detection and the accuracy of hazard identification. Furthermore, it eliminates data interference through normalization processing, solving the problems of large identification errors, false negatives, and missed detections in existing technologies.

[0062] In this embodiment, the intelligent hierarchical decision-making module includes a risk level assessment unit, a decision logic unit, a strategy storage unit, and an emergency linkage triggering unit. The risk level assessment unit integrates four dimensions of indicators: hazard level, equipment importance, environmental factors, and fault arc characteristics, and uses the analytic hierarchy process (AHP) to determine the weight of each indicator and quantify the comprehensive risk level. The decision logic unit outputs four-level handling strategies based on the comprehensive risk level, corresponding to normal status, level one warning, level two warning, and level three warning, respectively.

[0063] This module takes into account the data processing and analysis results, and combines multiple dimensions such as the importance of electrical equipment, the level of hidden danger, and environmental factors to formulate differentiated early warning and fire extinguishing strategies. This enables "hidden danger classification and precise measures" to avoid equipment damage and media waste caused by accidental activation of fire extinguishing equipment, while ensuring that major hidden dangers are dealt with in a timely manner.

[0064] Module Composition: The module consists of a risk level assessment unit, a decision logic unit, a strategy storage unit, and an emergency response triggering unit. The risk level assessment unit calculates the comprehensive risk level based on the analysis results from the edge intelligent data processing and analysis module, combined with equipment importance weights and environmental factors. The decision logic unit calls upon preset handling strategies in the strategy storage unit according to the comprehensive risk level to formulate differentiated early warning and fire suppression plans. The emergency response triggering unit, based on the decision results, triggers corresponding early warning, fire suppression, and power cut-off commands, linking with other modules.

[0065] The decision logic unit is responsible for formulating differentiated warning, fire suppression, and power cut-off commands based on the comprehensive risk level calculated by the risk level assessment unit and the preset response strategies stored in the strategy storage unit. The decision logic is implemented using a dedicated programmable logic controller (PLC), possessing high reliability and rapid response capabilities. Decision commands are output using standardized digital signals, allowing direct interface with subsequent warning execution and fire suppression control modules to achieve real-time command issuance and ensure timely response actions. The decision logic unit incorporates a logic verification module to rigorously verify decision commands, preventing erroneous operations caused by command errors, ensuring the accuracy of decision commands, and guaranteeing the safety and reliability of system operation.

[0066] The decision logic unit is responsible for calculating the comprehensive risk level based on the risk level assessment unit. The system invokes preset response strategies from the strategy storage unit to formulate differentiated warning, fire extinguishing, and power cut-off instructions. The decision logic is implemented using an STM32F407 programmable logic controller (PLC), boasting high reliability (mean time between failures ≥ 50,000 hours) and rapid response capability (instruction generation time ≤ 100ms). Decision instructions are output using standardized TTL digital signals, allowing direct interface with subsequent warning execution and fire extinguishing control modules for real-time instruction issuance and timely response. The decision logic unit incorporates a logic verification module to rigorously verify decision instructions, including instruction format, execution module address, and action sequence, preventing erroneous operations such as accidental activation of fire extinguishing equipment or accidental power cut-off due to instruction errors. This ensures the accuracy of decision instructions and guarantees the safety and reliability of system operation.

[0067] The strategy storage unit uses non-volatile memory to store various core data such as preset handling strategies, indicator weight parameters, equipment importance classification standards, and environmental factor judgment standards. Data is not lost even after power failure, ensuring normal operation after system restart. Strategy updates and parameter adjustments can be performed via a cloud platform or local management terminal, facilitating future system optimization and upgrades to adapt to changing needs in different scenarios. The storage unit has a built-in data encryption module using a dedicated encryption algorithm to ensure that strategy data is not tampered with, guaranteeing the security of decision-making logic and preventing malicious tampering that could lead to system failure or misoperation. Preset handling strategies are divided into four levels based on comprehensive risk levels: normal state, Level 1 warning, Level 2 warning, and Level 3 warning. Each level of strategy clearly defines the warning method, power cut-off requirements, and fire extinguishing equipment activation requirements, ensuring the targeting and operability of the handling strategies and achieving "tiered policy implementation and precise handling."

[0068] The policy storage unit uses NOR Flash non-volatile memory with a storage capacity of 4GB. It can store various core data such as preset handling policies, indicator weight parameters, equipment importance classification standards, and environmental factor judgment standards. Data is not lost even when power is off, and the storage life is ≥10 years, ensuring normal operation after system restart. Policy updates and parameter adjustments can be performed via the cloud platform or the local management terminal RS232 interface. Updates support both online and offline upgrades, facilitating future system optimization and upgrades to adapt to changing needs in different scenarios. The storage unit has a built-in data encryption module using the national cryptographic SM2 encryption algorithm to encrypt and protect the stored policy data and parameters. This ensures that the policy data cannot be tampered with. Tampering will trigger data verification failure, and the system will start the default policy, ensuring the security of the decision-making logic and preventing malicious tampering that could lead to system failure or misoperation.

[0069] The emergency response triggering unit is responsible for triggering corresponding early warning, fire extinguishing, and power cut-off actions based on the instructions output by the decision logic unit, and for coordinating subsequent multi-mode early warning execution modules, precision fire extinguishing control and execution modules, and system linkage and feedback modules. This unit uses a dedicated relay output interface, supporting simultaneous triggering of multiple commands. It can simultaneously trigger multiple actions such as early warning, power cut-off, and fire extinguishing activation, with a fast trigger response speed, ensuring coordinated work among modules and improving emergency response efficiency. The emergency response triggering unit has a built-in status feedback module that can collect the action status of each execution module in real time and feed it back to the decision logic unit, realizing closed-loop control of decision-making, execution, and feedback. If an execution module fails to act according to the instructions, a fault alarm is immediately triggered, reminding personnel to handle the situation promptly, ensuring the effectiveness of the response and preventing the escalation of potential hazards due to execution module failure.

[0070] The emergency linkage triggering unit is responsible for triggering corresponding early warning, fire extinguishing, and power cut-off actions based on the instructions output by the decision logic unit, and for linking subsequent multi-mode early warning execution modules, precision fire extinguishing control and execution modules, and system linkage and feedback modules. This unit uses a solid-state relay output interface, supporting simultaneous triggering of 8 commands. Each interface has a maximum load current of 10A and can simultaneously trigger multiple actions such as early warning, power cut-off, and fire extinguishing activation, with a trigger response speed of ≤50ms, ensuring coordinated operation of all modules and improving emergency response efficiency. The emergency linkage triggering unit has a built-in status feedback module that collects the action status of each execution module, such as whether the early warning has been activated, whether Venturi valve 3 is open, and whether the power has been cut off, via an optocoupler, and feeds this information back to the decision logic unit, achieving closed-loop control of decision-making, execution, and feedback. If an execution module fails to act according to the instructions, such as Venturi valve 3 not opening or power not being cut off, a fault alarm is immediately triggered with audible and visual alarms and remote push notifications. Simultaneously, fault information, fault time, fault module, and fault type are recorded to remind personnel to handle the situation promptly, ensuring the effectiveness of the response and preventing the expansion of potential hazards due to execution module failure.

[0071] This module, based on risk assessment theory, the analytic hierarchy process (AHP), and decision theory, enables comprehensive evaluation and differentiated decision-making across multiple dimensions. Risk assessment theory quantifies the fire risk of electrical equipment (5) by integrating indicators such as hazard level, equipment importance, and environmental factors, avoiding the limitations of single-indicator decision-making. The AHP is used to determine the weights of each assessment indicator, ensuring the scientific rigor and rationality of the risk assessment. Decision theory, based on the risk assessment results, formulates a "tiered response, graded handling" strategy, balancing the timeliness of early warning, the effectiveness of fire suppression, and equipment protection needs. Specifically, minor hazards only require an early warning, without fire suppression; moderate hazards trigger an enhanced early warning and cut off non-critical power; severe hazards immediately trigger fire suppression and cut off all power, achieving "precise handling and avoiding waste."

[0072] Formula and explanation: Comprehensive risk level assessment formula: Formula function: To quantify the comprehensive fire risk level of electrical equipment 5, providing a core basis for decision-making logic and avoiding decision-making biases caused by a single hazard indicator. Letter meanings: The comprehensive risk level ranges from 0 to 10 points, with higher scores indicating greater risk. The hazard level is weighted at 0.4, which is the core weight. The hazard level score is obtained by the GA-BP neural network in step 2, with 0 points for normal, 3 points for level 1 hazard, 6 points for level 2 hazard, and 10 points for level 3 hazard. The equipment importance weight is 0.3; The importance score for equipment is as follows: critical equipment is worth 10 points, important equipment is worth 7 points, and ordinary equipment is worth 3 points. The environmental factor weight is 0.2; Environmental factors are scored as follows: high humidity / high dust / high temperature environment is 10 points, general environment is 5 points, and dry and clean environment is 2 points. The fault arc weight is 0.1; The score for fault arc is 10 points if a fault arc exists and 0 points if it does not exist.

[0073] Decision logic judgment formula hierarchical response; ; Formula function: Based on the comprehensive risk level Develop differentiated response strategies, clarify the action instructions corresponding to each level of risk, and achieve "tiered response and precise handling." Explanation: Under normal conditions, the system only monitors in real time and does not trigger any actions; at Level 1 warning, an audible and visual alarm is triggered, and warning information is simultaneously pushed to management personnel, but fire extinguishing equipment is not activated and power is not cut off; at Level 2 warning, a high-intensity audible and visual alarm is triggered, along with SMS notifications and platform pushes, cutting off the power supply to non-critical electrical equipment 5, such as auxiliary equipment, to prevent the hazard from escalating, but fire extinguishing equipment is not activated; at Level 3 warning, the highest intensity alarm is immediately triggered, fire extinguishing equipment is activated, all power to electrical equipment 5 is cut off, and the monitoring system and emergency platform are linked to achieve emergency response.

[0074] The analytic hierarchy process (AHP) is used to calculate the importance weights of equipment. The formula's purpose is to determine the weight of equipment importance in the comprehensive risk assessment, ensuring the scientific nature of the weight allocation and avoiding biases caused by subjective judgment. The letters represent: Assign weights to equipment importance; For the first Each assessment expert scored the equipment importance index from 1 to 10 points. The number of evaluation experts is fixed at 5; There are four categories of evaluation indicators: hazard level, equipment importance, environmental factors, and fault arc; For the first The expert on the first Scoring of similar indicators.

[0075] Compared with existing technologies, this module has three main advantages: First, it constructs a multi-dimensional comprehensive risk assessment system, combining equipment importance, environmental factors, fault arc characteristics, and hazard levels for the first time. Weights are determined using the analytic hierarchy process (AHP) to quantify risk levels, addressing the problems of single-threshold decision-making and lack of scientific rigor in existing technologies. Second, it adopts a "tiered response, gradient handling" decision-making logic, developing differentiated strategies based on risk levels. This avoids equipment damage and media waste caused by the existing technology's approach of "activating fire suppression whenever the risk exceeds the limit," while ensuring timely handling of major hazards and balancing safety and economy. Third, it introduces a multi-expert weight calibration method, improving the scientific rigor and versatility of decision-making. This allows it to adapt to the electrical equipment protection needs of different scenarios (industrial plants, buildings, substations), solving the problems of single-strategy and poor adaptability in existing technologies.

[0076] In this embodiment, the precision fire extinguishing control and execution module includes a Venturi valve 3, a directional nozzle array 4, a delivery pipeline 2, a flow detection unit, a pressure detection unit, and a Venturi valve 3 drive unit; the Venturi valve 3 has a pressure adaptive adjustment function, adjusting the opening of the Venturi valve 3 according to the pressure difference of the delivery pipeline 2 and the target flow rate; the directional nozzle array 4 consists of multiple independently controllable nozzles, each nozzle corresponding to a part of the internal components of the electrical equipment 5.

[0077] Module Composition: The module consists of a Venturi valve 3, a directional sprinkler array 4, a delivery pipeline 2, a flow detection unit, a pressure detection unit, and a Venturi valve 3 drive unit. The Venturi valve 3, designed based on the Venturi principle, features adaptive pressure adjustment, automatically adjusting its opening according to the pressure in the delivery pipeline 2. The directional sprinkler array 4 comprises multiple independently controllable sprinklers, each corresponding to a key component of the electrical equipment 5, allowing for precise activation based on the potential hazard area. The flow detection unit and pressure detection unit monitor the flow rate of the extinguishing medium and the pressure in the delivery pipeline 2 in real time, providing a basis for flow adjustment. The Venturi valve 3 drive unit utilizes existing electromagnetic drive technology, offering fast response and enabling precise control of the Venturi valve 3's opening.

[0078] Examples of independently controllable nozzles are as follows: 1. The models are based on ZSTK open sprinkler heads, such as ZSTK-15 / 80 and ZSTK-20 / 115. They are individually controlled by a solenoid valve to accurately spray the extinguishing medium and are suitable for the rapid extinguishing of local fires in electrical equipment.

[0079] II. Closed-type printheads controlled by solenoid valves: Based on traditional closed-type printheads such as the ZSTZ and ZSTX series, a special miniature solenoid valve is added, such as the ZSTZ-15 / 68+DC24V solenoid valve and the ZSTX-15 / 79+DC24V solenoid valve, to achieve non-thermal active control opening.

[0080] This module, based on fluid mechanics theory, pressure adaptive control theory, and precision spraying theory, achieves efficient and precise delivery of the extinguishing medium. It calculates the relationship between the opening of the Venturi valve 3 and the flow rate and pressure using formulas to achieve precise flow control. The pressure adaptive control theory automatically adjusts the opening of the Venturi valve 3 by monitoring the pressure in the delivery pipeline 2 in real time, ensuring a stable flow rate of the extinguishing medium even during pressure fluctuations. The precision spraying theory, combined with sensor location information, identifies potential hazard areas and activates directional nozzles in the corresponding areas, achieving precise "point-to-point" fire suppression and avoiding waste of the extinguishing medium and secondary damage to undamaged equipment. Simultaneously, considering the characteristics of electrical equipment 5, it selects extinguishing media suitable for electrical fires (such as carbon dioxide and heptafluoropropane) to prevent electric shock accidents during fire suppression.

[0081] Core formulas and explanations: Venturi valve flow regulation formula; ; Formula function: Based on the pressure difference in delivery pipeline 2 and the opening degree of Venturi valve 3, calculate the spray flow rate of the extinguishing medium, and adjust the opening coefficient of Venturi valve 3 accordingly. This achieves precise flow control while utilizing the pressure adaptive characteristics of the Venturi valve 3 to ensure stable flow. Letter meanings: The unit for the flow rate of the extinguishing medium is L / s; The flow coefficient is determined by the material and structure of the Venturi valve 3 and is fixed at 0.85; The area unit for a Venturi valve with a 3-degree opening is m². The pressure difference before and after Venturi valve 3 is expressed in Pa. The density unit for extinguishing agents is kg / m³, for example, carbon dioxide. heptafluoropropane ; The unit for the maximum opening area of ​​the Venturi valve 3 is m². The opening coefficient of the Venturi valve 3 ranges from 0 to 1, where 0 represents fully closed and 1 represents fully open. It is controlled by the decision result and flow detection data.

[0082] Venturi valve 3 opening coefficient adjustment formula: Formula function: Based on the pressure and flow deviations in the delivery pipeline 2, automatically adjust the opening coefficient of the Venturi valve 3 to achieve pressure self-adaptation and precise flow control, ensuring that the flow rate of the extinguishing medium remains stable at the target value; Letter meanings: This represents the current opening coefficient of the Venturi valve. The baseline opening coefficient is set according to the target flow rate, and its value ranges from 0.3 to 0.8. The pressure proportionality coefficient is fixed at 0.05; This represents the actual pressure difference; The target pressure difference is fixed at 1. Pa; The flow rate ratio coefficient is fixed at 0.08; The target traffic volume is determined by the risk level in step 3, and a level 3 warning is issued. It will not be activated during a Level II warning. This represents the actual traffic volume.

[0083] Directional nozzle start-up determination formula: Startup conditions: and Formula function: Based on the comprehensive detection values ​​from step 1, determine the sprinkler heads corresponding to the potential hazard areas, enabling precise activation of directional sprinkler heads and avoiding media waste caused by activating all sprinkler heads. Letter meanings: For the first The comprehensive test value of the area corresponding to each nozzle; This represents the average of the comprehensive test values; The standard deviation of the overall test values; The overall risk level is determined by the risk level of a certain area. Exceeding Scope, and overall risk level When a Level 3 warning is issued, the corresponding directional sprinklers in that area will be activated.

[0084] Formula for calculating the spraying time of extinguishing medium: The formula's purpose is to calculate the spraying time of the extinguishing agent, ensuring that the agent fully covers the affected area for effective fire suppression, while avoiding excessive spraying time that could lead to agent waste. Letter meanings: Spraying time (unit: s); Volume of extinguishing medium storage tank 1 (unit: L); The effective injection coefficient is fixed at 0.8, taking into account media residue and loss. The actual injection flow rate (unit: L / s) is given when the injection time reaches [a certain value]. Or, if the hazard was detected and eliminated in step 1, then... Stop spraying when the time comes.

[0085] Compared with existing technologies, this module has four main advantages: First, it intelligently upgrades the existing control Venturi valve 3 by combining it with a pressure adaptive adjustment formula to achieve precise control of the valve's opening and pressure self-adaptation, solving the problems of the existing Venturi valve 3 lacking pressure self-adaptation capability and experiencing large flow fluctuations. Second, it adopts a directional nozzle array 4 design, combined with a nozzle activation judgment formula, to achieve precise "point-to-point" fire extinguishing in hazardous areas, avoiding media waste and secondary equipment damage caused by activating all nozzles, saving more than 30% of fire extinguishing media compared to existing technologies. Third, it constructs a flow closed-loop adjustment mechanism, adjusting the spray flow in real time through flow detection and opening coefficient adjustment formulas to ensure fire extinguishing efficiency, while setting target flow rates according to risk levels to achieve differentiated fire extinguishing. Fourth, it optimizes the layout of the delivery pipeline 2, shortening the delivery path of the fire extinguishing media and improving response speed, controlling the fire extinguishing response time to within 3 seconds, a 50% improvement over existing technologies.

[0086] In this embodiment, the system linkage and feedback module includes a power linkage unit, a monitoring linkage unit, an emergency platform linkage unit, an operating status feedback unit, a data storage unit, and a dual power redundancy unit. The power linkage unit dynamically adjusts the power cut-off delay time based on the decision command and the operating current of the electrical equipment 5 to avoid equipment damage and electric shock risk. The dual power redundancy unit adopts a main power supply and a backup power supply. When the main power supply voltage is lower than a set threshold and the duration is ≥0.5s, it automatically switches to the backup power supply. The operating status feedback unit adopts an error correction algorithm.

[0087] Module Composition: The module consists of a power supply linkage unit, a monitoring linkage unit, an emergency platform linkage unit, an operational status feedback unit, a data storage unit, and a dual-power redundancy unit. Specifically, the power supply linkage unit, based on decision-making results, enables tiered power cutoff for the five electrical equipment components; the monitoring linkage unit, linked to the video surveillance system, captures real-time images of potential hazard areas and uploads them to management personnel and the emergency platform; the emergency platform linkage unit interfaces with the city's emergency platform and the enterprise's fire protection platform, reporting early warning and firefighting information; the operational status feedback unit collects operational parameters from each module in real-time for optimization and adjustment; the data storage unit stores perceived data, decision results, early warning information, firefighting records, etc., facilitating subsequent traceability and analysis; the dual-power redundancy unit employs a supercapacitor design for both the main and backup power supplies, enabling continuous operation for over 96 hours after a power outage, ensuring reliability in high-risk environments.

[0088] This module, based on linkage control theory, closed-loop feedback control theory, and redundancy design theory, achieves full-process coordination and stable operation of the system. Linkage control theory, through the establishment of a unified linkage protocol, enables the coordinated work of various modules, electrical equipment, power supplies, monitoring systems, and emergency platforms, ensuring the synchronization of actions such as early warning, fire suppression, and power cut-off. Closed-loop feedback control theory, through real-time feedback of system operating status and handling results, adjusts decision-making and execution strategies, improving the system's adaptability and reliability. Redundancy design theory, through dual power supply redundancy, avoids system failure due to power outages, ensuring that the system can normally complete early warning and fire suppression operations in the event of a sudden power outage. Simultaneously, a national-level encryption module is used to ensure the security and confidentiality of linkage data.

[0089] Formula and explanation: Power cut-off delay formula: The formula's function is to set the power cut-off delay time based on the operating current of electrical equipment 5, to avoid damage to electrical equipment 5, such as motors and frequency converters, due to sudden power outages, while ensuring that there is no risk of electric shock during firefighting.

[0090] Letter meanings: The unit for power cut-off delay time is seconds (s). The baseline delay time is fixed at 1 second; The current correction factor is fixed at 0.005s / A; The unit for the actual operating current of electrical equipment 5 is A. The larger the current, the longer the delay time, in order to avoid damage to the equipment due to sudden current changes. The delay time shall not exceed 3 seconds.

[0091] Operating status feedback error correction formula: The formula's function is to correct errors in the operational status feedback data of each module, such as flow rate, pressure, and Venturi valve opening, thereby improving the accuracy of the feedback data and providing a reliable basis for data processing and decision optimization. The letters represent: The feedback data has been corrected; This is the original feedback data; The error correction factor is fixed at 0.02; This feedback data is the historical average value. Through correction, the error of the feedback data can be controlled within [a certain range]. Within.

[0092] Dual power supply switching determination formula and switching conditions: and Formula function: Real-time monitoring of main power supply voltage. When the main power supply voltage drops below a set threshold for more than 0.5 seconds, it automatically switches to the backup power supply to ensure normal system operation and prevent system failure due to power outages; Letter meanings: The actual voltage unit of the main power supply is V; The minimum allowable voltage of the main power supply is fixed at Single-phase power supply; Main power supply voltage is lower than The duration is measured in seconds. When the switching conditions are met, the system immediately switches to the backup power supercapacitor, which can provide power for more than 96 hours.

[0093] Linked data transmission delay formula Formula function: Calculates the time delay from when a linkage command is sent from the decision-making module to when it is received by the execution module, monitors the system's linkage efficiency in real time, and automatically switches the communication channel when the delay exceeds a threshold of 2 seconds to ensure that emergency commands are issued in real time. Letter meanings: The unit for the delay time of the linked data transmission is seconds (s). The timestamp of the instruction received by the execution module; The system uses edge computing and local closed-loop design to stably control the transmission latency to within 0.5s, which is far lower than the industry's conventional latency level, by providing the timestamp for sending instructions to the decision-making module.

[0094] Closed-loop feedback regulation formula Formula Function: Based on the difference between the actual operating data and the set target data, the decision output value is dynamically optimized and adjusted to achieve closed-loop control of the entire system process, improving the accuracy of early warning and fire suppression execution. Letter Meaning: The optimized decision output value; This is the decision output value from the previous cycle; The feedback adjustment coefficient is fixed at 0.1; The feedback data has been corrected; The target operating data set for the system.

[0095] Compared with existing technologies, this module has three main advantages: First, it constructs a closed-loop linkage architecture that breaks down the barriers of independent operation of modules in traditional fire extinguishing systems, achieving synchronous coordination of early warning, fire extinguishing, power cut-off, monitoring upload, and emergency reporting, with a linkage response delay of ≤0.5s, which is an improvement over traditional systems. Second, it designs a tiered power linkage delay strategy, dynamically adjusting the power outage time based on current parameters, eliminating the risk of electric shock during fire extinguishing and avoiding damage to precision electrical equipment caused by instantaneous power outages, thus solving the problem of secondary equipment damage caused by direct power outages in traditional systems. Third, it adopts a dual-power redundancy and real-time transmission delay monitoring dual-protection mechanism, seamlessly switching to the backup power supply within 0.5s of the main power failure, and automatically switching the communication channel when the linkage command delay exceeds the standard, completely solving the problem of early warning and fire extinguishing failures caused by power outages and communication lags in traditional systems.

[0096] In this embodiment, the multi-mode early warning execution module includes an audible and visual early warning unit, a remote push unit, an early warning intensity adjustment unit, an NB-IoT+LoRa dual-mode communication unit, and an early warning log recording unit. The early warning intensity adjustment unit can automatically adjust the audible and visual early warning intensity according to the on-site noise and light intensity. The dual-mode communication unit can monitor the packet loss rate of early warning information transmission in real time, and automatically switch the transmission mode when the packet loss rate exceeds 0.5%. The remote push unit supports three push channels: APP, SMS, and emergency platform, and matches the corresponding push method according to the early warning level.

[0097] This module receives the results of intelligent decision-making and is the core execution unit for transmitting early warning information. Unlike the existing technology's "single sound and light warning mode," it adopts a multi-mode early warning architecture of "local sound and light, remote push, and hierarchical prompts." Combined with environmental adaptive adjustment and dual-mode transmission technology, it ensures that early warning information can be transmitted to management personnel in a timely and accurate manner. At the same time, it adjusts the warning intensity according to the warning level to avoid the warning information being ignored, thus solving the problems of poor early warning effect, untimely information transmission, and narrow coverage of existing technologies.

[0098] Module Composition: The module consists of an audible and visual warning unit, a remote push unit (SMS, APP, emergency platform), a warning intensity adjustment unit, a dual-mode communication unit (NB-IoT and LoRa), and a warning log recording unit. The audible and visual warning unit uses high-brightness LEDs and a high-decibel speaker, supporting intensity adjustment. The remote push unit uses NB-IoT and LoRa dual-mode transmission to achieve multi-channel delivery of warning information. The alarm intensity adjustment unit automatically adjusts the warning intensity according to the on-site environment (noise, light). The warning log recording unit records the warning time, level, and cause in real time, facilitating subsequent traceability and analysis.

[0099] The high-decibel audible alarm and high-brightness LED warning light work together to output sound and light signals of different intensities and frequencies depending on the warning level. Specific parameters and corresponding warning modes are as follows: 1. Level 1 Warning (Low Risk): The audible alarm outputs a gentle 60dB volume, equivalent to normal conversation, with an intermittent 1Hz frequency, sounding for 1 second followed by a 1-second pause. The LED is green and flashes at 1Hz. 2. Level 2 Warning (Medium Risk): The audible alarm outputs a moderate 80dB volume, equivalent to a noisy office environment, with a 2Hz frequency, sounding for 0.5 seconds followed by a 0.5-second pause. The LED is yellow and flashes at 2Hz. 3. Level 3 Warning (High Risk): The audible alarm outputs a high-intensity 100dB volume, equivalent to a motorcycle starting sound, with a continuous 4Hz frequency and no pause. The LED is red and flashes at 4Hz. The audible and visual warning unit adopts an industrial-grade protective design. The audible alarm uses a waterproof and dustproof speaker, and the LED uses high-brightness surface-mount LEDs with a luminous intensity ≥1000mcd. It has excellent anti-interference, waterproof, and dustproof capabilities, and can adapt to harsh industrial environments such as high dust, high humidity, and high temperature, ensuring stable output of the warning signal without distortion or interruption.

[0100] The local display and early warning unit uses a 2.4-inch TFT high-definition display terminal with a resolution of 320×240 pixels and a viewing angle of ≥170°. It can display detailed early warning information in real time, including the early warning level marked by color: green = Level 1, yellow = Level 2, red = Level 3; the equipment where the hazard is located, such as "Workshop No. 1 power distribution cabinet"; the type of hazard, such as "poor contact leading to local overheating"; the early warning trigger time accurate to the second, such as "2026-02-11 10:30:45"; and the current abnormal parameter values, such as "temperature 78℃, residual current 45mA". This allows on-site management personnel to quickly grasp the details of the hazard and take targeted measures. The display terminal supports touch operation with a touch response time of ≤200ms. Management personnel can use the terminal to query early warning history records and confirm early warning information. After confirmation, the early warning sound and light signals automatically stop and display a "confirmed" mark. At the same time, the early warning signal can be manually controlled to be turned off only for early warnings that have been dealt with. The operation is convenient. The display terminal uses a high-definition display screen with a wide viewing angle, making it easy for people in different positions on site to view the screen at a distance of ≥5m. It also has a screen protection function that automatically enters sleep mode after 30 seconds of inactivity, with a power consumption of ≤0.5W, extending the service life of the equipment and reducing power consumption.

[0101] The remote early warning push unit is responsible for pushing early warning information to the mobile terminals of management personnel and the cloud management platform, realizing remote early warning prompts and ensuring that management personnel can receive early warning information in a timely manner and quickly arrange disposal work even when they are not on site. Remote push notifications include three channels: SMS, dedicated app push, and cloud platform alarm notifications. This ensures that warning information can be transmitted through multiple channels, avoiding the loss of warning information due to the failure of a single push method, such as no mobile signal or the app not being launched. The specific implementation of the three push methods is as follows: 1. SMS push: Connecting to the operator's SMS gateway through the NB-IoT module, warning SMS messages can be sent to up to 10 preset management personnel mobile phone numbers. The SMS content includes the warning level, the potentially hazardous equipment, the type of hazard, and the trigger time, with a sending delay of ≤3s; 2. Dedicated app push: Management personnel install the system's dedicated app, which supports Android / iOS systems. The app runs continuously in the background. After a warning is triggered, a pop-up alarm is immediately pushed with an audio prompt. Clicking the pop-up window allows viewing detailed warning information. The app also supports confirming warnings and querying historical warnings; 3. Cloud platform alarm notifications: The cloud management platform receives warning information in real time and displays alarm pop-ups on the platform's homepage. The warning level is marked with different colors. The platform background automatically records warning information, allowing management personnel to view details and issue handling instructions on the platform.

[0102] The remote early warning push unit works in conjunction with the edge-cloud collaborative communication unit in step 2. After receiving early warning information from the decision module, it quickly processes the early warning details (processing time ≤ 500ms) and pushes them to relevant terminals via a dedicated communication link. This ensures fast push speed and high accuracy. It also supports push record queries, allowing users to check the push time, push channel, and reception status (received / not received) for each early warning message, facilitating later traceability. Furthermore, the remote early warning push unit supports push permission settings. Different push permissions can be set according to the responsibilities of management personnel; for example, on-site management personnel receive all levels of early warnings, while remote management personnel only receive level 2 and above warnings, preventing irrelevant personnel from receiving too many early warning messages and improving the targeting of push notifications.

[0103] The early warning recording and query unit is responsible for recording all early warning-related information in real time, including early warning trigger time, early warning level, hazard details, hazard equipment, hazard type, abnormal parameters, early warning method (audio / visual / SMS / APP / cloud), push recipient, confirmation time, and handling results such as "investigated, no abnormality" or "handled, hazard eliminated." Recorded data is stored on a dedicated SD card with a storage capacity of 16GB, expandable to 32GB, and can be stored for at least one year. It supports queries based on various conditions, such as time (e.g., "February 2026"), early warning level (e.g., "Level 3 Early Warning"), equipment name (e.g., "Distribution Cabinet No. 1"), and handling results. The query response time is ≤1 second, facilitating later hazard tracing, system optimization, and accident review. Simultaneously, this unit can perform simple statistical analysis on early warning data, generating early warning statistical reports that can be generated daily / weekly / monthly. These reports include statistical information such as the number of early warnings at each level, the distribution of hazard types, and the timeliness of handling, providing data support for managers to understand the occurrence patterns of electrical equipment hazards and optimize protection strategies. For example, if a certain piece of equipment frequently triggers a Level 1 early warning, targeted inspections of that equipment can be strengthened.

[0104] This module's working logic aligns with the requirements of tiered early warning systems, strictly adhering to the early warning instructions issued in step 3 to execute corresponding early warning actions, ensuring the targeted and timely nature of early warning alerts. Upon receiving an early warning instruction from the decision-making module, the early warning execution module first parses the core information in the instruction, such as the early warning level and hazard details (parse time ≤ 100ms), and then synchronously triggers the corresponding level of multi-mode early warning actions. The specific workflow is as follows: 1. Level 1 Early Warning (Low Risk): Only a local soft audible and visual warning (green light), a 60dB intermittent audible alarm, and a local display showing detailed early warning information are activated. Early warning information is simultaneously pushed to the mobile terminals of on-site management personnel via SMS and APP, without needing to push to remote management personnel to avoid excessive early warnings interfering with normal work; 2. Level 2 Early Warning (Medium Risk): A local enhanced audible and visual warning (yellow light), an 80dB intermittent audible alarm, and a local display showing... 1. **Level 3 High-Risk Warning:** Simultaneously push warning information to the mobile terminals, SMS, APP, and cloud management platform of on-site and relevant remote management personnel, reminding them to arrive promptly for handling. 2. **Level 3 High-Risk Warning:** Activate the highest intensity local audible and visual warning (red light, 100dB continuous siren, local warning display), and simultaneously push warning information to the mobile terminals, SMS, APP, and cloud management platform of all relevant management personnel. Simultaneously, link the on-site broadcast system to broadcast warning messages such as "Warning! High-risk hazard has occurred in distribution cabinet No. 1. Immediately evacuate the danger zone. Relevant personnel should quickly take action!", ensuring all on-site personnel are quickly informed, promptly evacuate the danger zone, and take emergency response measures.

[0105] Once management personnel confirm the warning and complete the hazard handling, they can send a "Hazard Handling Completed" command via a local display terminal, dedicated APP, or cloud platform. Upon receiving the command, the warning execution module immediately stops the audible and visual warning signal and updates the handling result and confirmation time in the warning record, completing one warning handling process. If management personnel fail to complete the warning confirmation and hazard handling within the specified time after the warning is triggered, the system will activate a tiered timeout reminder mechanism to ensure that the hazard is taken seriously and handled promptly. The specific timeout thresholds are set differently according to the warning level: Level 1 warning (low risk) has a timeout threshold of 30 minutes, Level 2 warning (medium risk) has a timeout threshold of 15 minutes, and Level 3 warning (high risk) has a timeout threshold of 5 minutes. After the timeout, the system will automatically strengthen the warning prompt. The intensity of the local audible and visual warning will be increased by one level. For example, after a Level 1 warning expires, the volume will be increased to 80dB and the frequency to 2Hz, and the LED lights will switch to flashing yellow. After a Level 2 warning expires, the volume will be increased to 100dB and the frequency to 4Hz, and the LED lights will switch to flashing red. After a Level 3 warning expires, the highest intensity audible and visual warning will be maintained, and the on-site broadcast system will be linked to play the warning prompt in a loop every 10 seconds. The remote push unit will initiate a secondary push, adding the next level of management personnel to the push recipients. If the on-site management personnel do not handle the issue, the push will be sent to the department head. If the department head does not handle the issue, the push will be sent to the enterprise safety management person in charge. The secondary push will be marked with "overdue and unhandled". At the same time, an overdue alarm record will be generated on the cloud platform, clearly indicating the overdue duration, details of the unhandled warning, and the current responsible personnel.

[0106] For Level 3 high-risk warnings, if no response or action is received within 5 minutes, the system will automatically activate the system linkage and feedback module in step 6, sending a "warning timeout without action" linkage signal. Simultaneously, it will trigger temporary protective measures in non-critical areas of electrical equipment, such as shutting off power to surrounding auxiliary equipment and activating regional ventilation equipment, to reduce the risk of escalation. At the same time, the timeout information will be pushed to the city's emergency management platform in real time, requesting external emergency support to ensure that high-risk hazards do not escalate due to human negligence. The warning record and query unit will record the entire timeout process in detail, including the warning trigger time, timeout threshold, timeout start time, second push time, emergency platform linkage time, final action time (if any), and any reasons for non-action (if supplemented later). All timeout-related data will be linked to the warning record and stored long-term for future safety reviews, accountability, and management optimization, providing data support for adjusting warning response procedures and optimizing management personnel allocation.

[0107] In addition, this module also features a warning and mute function. For scenarios where management personnel have arrived at the scene but the issue has not yet been resolved, and the audible and visual alarms need to be temporarily disabled, management personnel can activate the mute mode via the local display terminal or a dedicated app. The mute duration can be manually set from 1 to 30 minutes, with a default of 10 minutes. During the mute period, the local audible and visual alarms stop, but the local display terminal continues to show the warning information, and the remote push unit will not push notifications repeatedly. After the mute duration ends, if the hazard has not been resolved, the system will automatically restore the audible and visual alarms and activate an overtime reminder mechanism to prevent the hazard from being forgotten due to the mute mode. The mute operation will be recorded in real time in the warning log, noting the mute activation time, mute duration, initiator, and recovery time, ensuring traceability and preventing safety risks caused by unauthorized mute actions.

[0108] This module, based on the early warning closed-loop control theory, realizes a complete closed-loop process of "early warning triggering – multi-channel notification – handling confirmation – timeout linkage – record traceability," ensuring that every early warning receives an effective response and handling, preventing early warnings from becoming mere formalities. Simultaneously, combining hierarchical management theory, it sets differentiated timeout thresholds and response measures for different early warning levels, avoiding excessive consumption of management resources by low-risk early warnings while ensuring rapid response to high-risk hazards, thus aligning with the management needs of actual industrial scenarios.

[0109] This module features clear technical improvements and non-obvious benefits. First, it overcomes the limitations of existing early warning systems that only "trigger without tracking," constructing a closed-loop process for early warning response. New functions such as timeout reminders, secondary push notifications, and emergency linkage are added, addressing the pain points of no response and delayed response after early warning information is issued in existing technologies. This ensures that potential hazards are promptly addressed and resolved, improving the effectiveness of early warnings. Second, the design of differentiated timeout thresholds and early warning reinforcement mechanisms matches different response requirements according to the early warning level. Low-risk early warnings provide ample time for investigation, while high-risk early warnings strictly control response time. The silent mode design also aligns with actual on-site operational needs, avoiding interference from audible and visual warnings while preventing the hazard from being forgotten, balancing practicality and safety. Third, the refined design of early warning records not only records core information about the early warning and response but also details operational aspects such as timeouts and silence, providing comprehensive data support for later safety reviews, accountability, and management optimization. This solves the problems of coarse and untraceable early warning records in existing systems. Furthermore, the combination of multi-channel push notifications and permission settings improves the targeting of early warning pushes, avoids interference from irrelevant personnel, and enhances management efficiency.

[0110] This module, based on acoustic-optical propagation theory, wireless communication theory, and environmental adaptive adjustment theory, achieves efficient transmission and accurate alerts. Acoustic-optical propagation theory guides the power adjustment of the acoustic-optical warning unit, ensuring effective perception of warning information under different environments (high noise, low light). Wireless communication theory combines the advantages of NB-IoT and LoRa dual-mode. NB-IoT, with its cellular network guarantee, achieves a 0.23% packet loss rate and 1.42-second latency, suitable for high-real-time urban scenarios. LoRa boasts lower power consumption and a theoretical battery life of up to 7.2 years, suitable for deployment in remote areas, overcoming the limitations of single transmission methods. Environmental adaptive adjustment theory automatically adjusts the warning intensity by detecting on-site noise and light intensity, improving warning effectiveness and avoiding resource waste.

[0111] Formula and explanation: Formula for adjusting the intensity of audible and visual warnings: ; Formula Function: Based on the ambient noise and light intensity, the power of the warning sound is automatically adjusted to ensure clear hearing in high-noise environments and clear visibility of lights in low-light environments, while avoiding unnecessary energy waste. Letter Meaning: The actual power unit for the warning sound is dB; The sound reference power is fixed at 80dB; The noise correction factor is fixed at 0.02 dB / dB; The actual noise level at the scene is measured in decibels (dB). The baseline noise level is fixed at 50 dB. The illumination correction factor is fixed at 0.01 dB / lux; The unit for actual on-site light intensity is lux. The baseline illuminance is fixed at 100 lux. When, increase sound power; when When increasing light brightness, the formula for adjusting light brightness is similar to that for sound power, only the coefficients are different.

[0112] Formula for calculating packet loss rate in early warning information transmission: ; Formula Function: Real-time monitoring of early warning information transmission quality. When the packet loss rate exceeds a set threshold of 0.5%, it automatically switches between NB-IoT and LoRa transmission modes to ensure accurate transmission of early warning information and prevent information loss due to packet loss. Letter Meaning: The number of warning information packets sent; The number of warning information packets received; the lower the packet loss rate, the better the transmission quality. In NB-IoT mode, the packet loss rate can be controlled within 0.23%, and in LoRa mode, the packet loss rate can be controlled within 0.3%.

[0113] Formula for the correspondence between warning level and warning method: ; Formula Function: Based on the warning level, the corresponding warning method is determined to achieve "tiered alerts," ensuring that managers can quickly assess the urgency of potential hazards and take appropriate measures. Explanation: Level 1 warnings use low-intensity sound and light (60dB), 50% light brightness, and APP push notifications to avoid disrupting normal production; Level 2 warnings use medium-intensity sound and light (80dB), 75% light brightness, APP push notifications, and SMS notifications to remind managers to investigate promptly; Level 3 warnings use high-intensity sound and light (100dB), 100% light brightness, APP push notifications, SMS notifications, and reporting to the emergency platform to ensure rapid response from emergency personnel.

[0114] Compared with existing technologies, this module has three main advantages: First, it adopts an environment-adaptive early warning intensity adjustment mechanism, which adjusts the intensity of audible and visual warnings in real time through a formula, solving the problem that existing technologies have fixed warning intensity and are easily ignored in complex environments. Second, it introduces NB-IoT+LoRa dual-mode transmission technology, combined with real-time packet loss rate monitoring and automatic switching, to achieve stable transmission of warning information, with the packet loss rate controlled within 0.3%, solving the problems of existing technologies having a single transmission method, high packet loss rate, and insufficient coverage in remote areas. Third, it constructs a multi-mode early warning system with "tiered prompts," matching different warning methods according to the warning level, which avoids excessive interference from low-level warnings while ensuring the emergency transmission of high-level warnings. At the same time, it adds warning log recording to facilitate subsequent traceability, solving the problems of existing technologies having a single warning method and lacking traceability.

[0115] The system self-inspection and predictive maintenance module includes a periodic self-inspection triggering unit, a hardware fault detection unit, an algorithm validity verification unit, a communication link detection unit, a fault level determination unit, a maintenance strategy generation unit, and a maintenance push unit. The system self-inspection and predictive maintenance module automatically calculates the optimal self-inspection cycle and performs a full-coverage self-inspection of all hardware, software algorithms, and communication links in the system. It adopts a multi-dimensional fault determination and component remaining service life prediction algorithm to generate and push differentiated maintenance strategies based on the fault level.

[0116] This module is the core support unit for ensuring the long-term stable operation of the system and reducing maintenance costs. It performs full-coverage testing on all hardware modules, software algorithms, and communication links, predicts the risk of component aging, failure, and inoperability in advance, and generates a visualized maintenance plan. This enables a shift from "passive emergency repair" to "proactive prevention," solving the problems of high maintenance costs, delayed fault detection, and poor long-term reliability of existing systems.

[0117] Module Composition: The module consists of a periodic self-test trigger unit, a hardware fault detection unit, an algorithm validity verification unit, a communication link detection unit, a fault level determination unit, a maintenance strategy generation unit, and a local / remote maintenance push unit. Specifically, the periodic self-test trigger unit can be set to daily, weekly, or monthly automatic self-test cycles; the hardware fault detection unit covers all hardware including sensors, Venturi valves 3, nozzles, power supplies, and communication modules; the algorithm validity verification unit verifies the operational accuracy of the GA-BP neural network and data fusion algorithm in real time; the communication link detection unit monitors the communication status of NB-IoT, LoRa, and the local bus; the fault level determination unit classifies faults into three levels: minor, moderate, and severe; the maintenance strategy generation unit automatically matches the corresponding maintenance plan; and the maintenance push unit simultaneously pushes maintenance instructions to the local display screen and the remote management terminal.

[0118] This module, based on reliability engineering theory, fault prediction theory, periodic inspection theory, and maintenance decision theory, enables intelligent maintenance throughout the system's entire lifecycle. Reliability engineering theory establishes a system reliability model by statistically analyzing the operating parameters and failure probabilities of each component. Fault prediction theory predicts component aging trends and remaining service life based on historical operating data and real-time monitoring data. Periodic inspection theory enables routine monitoring of the system's status by setting fixed inspection cycles, preventing the hidden development of faults. Maintenance decision theory generates differentiated maintenance strategies based on fault severity and component importance, prioritizing the handling of severe faults, rationally allocating maintenance resources, and minimizing maintenance costs and downtime while ensuring system safety.

[0119] Formula and explanation: Formula for periodic self-test time interval: ; Formula Function: Based on the system's designed safe lifespan and safety factor, automatically calculates the optimal self-test cycle, balancing detection frequency and system power consumption to ensure early fault detection without increasing additional energy consumption. Letter Meaning: Self-test time interval unit: h; The safe life unit for designing a single component of the system is h, with a standard value of 8760h. To ensure a safety factor of 12, a comprehensive self-inspection is to be completed within one year.

[0120] Hardware fault diagnosis formula: ; Formula Function: By comparing the difference between the test data and the standard normal data, it accurately determines whether there are faults, short circuits, open circuits, blockages, or leaks in hardware such as sensors, Venturi valves 3, nozzles, and power supplies. Letter Meaning: The hardware fault flag is set to 0 for normal operation and 1 for fault. This is actual hardware self-test data; This is normal data according to hardware standards; The allowable error threshold is calibrated by the hardware model and is generally ±3%.

[0121] Formula for predicting the remaining useful life of components: ; Formula purpose: Based on cumulative component power consumption, runtime, and load changes, predict remaining usable lifespan, providing a time basis for predictive maintenance. Letter meanings: The remaining service life of the component is measured in hours (h). The rated total life of a component is measured in hours (h). Real-time power consumption of components; This is the runtime; This refers to the rated total power consumption capacity of the component.

[0122] Fault severity quantification formula: ; Formula function: By considering the importance of components, the duration of the failure, and the scope of its impact, the formula quantifies the failure level and enables tiered maintenance. Letter meanings: The fault level is scored from 0 to 10 points. Assign importance weights to components; Assign an importance score to the faulty component; Weighted by fault duration; The score is based on the duration of the fault. Weighting for the impact of faults; The score is based on the scope of the fault's impact; 0-3 points indicate a minor fault, 4-7 points indicate a general fault, and 8-10 points indicate a serious fault.

[0123] Overall system reliability formula ; Formula Function: Calculates the overall real-time reliability of the system, intuitively reflecting the system's operational safety status. When the reliability falls below a threshold, it forcibly activates the highest priority maintenance alert. Letter Meaning: For the overall reliability of the system; For the first The reliability of each component; Where is the total number of system components; is the link loss coefficient, which is fixed at 0.02; when If the system is deemed unreliable, a maintenance alert will be immediately sent.

[0124] Compared with existing technologies, this module has three key advantages: First, it pioneers a predictive maintenance mechanism for the entire lifecycle of an intelligent electrical early warning and fire extinguishing system, replacing traditional manual inspections and achieving 24 / 7 fully automated self-inspection with 100% coverage, reducing manual maintenance costs by over 70%. Second, it establishes a multi-dimensional fault diagnosis and lifespan prediction model, using quantitative formulas to accurately locate faults, determine fault levels, and predict remaining lifespan, enabling the prediction of component failure risks 15-30 days in advance and preventing sudden system failures. Third, it enables automatic generation and tiered push of maintenance strategies, providing local alerts for minor faults and remote alarms for serious faults, completely solving the industry pain points of traditional fire extinguishing systems such as long-term lack of maintenance, difficulty in fault detection, and paralysis upon failure, increasing the overall system availability to over 99.5%.

[0125] Another possible embodiment is an intelligent early warning fire extinguishing device based on electrical equipment 5: the intelligent early warning fire extinguishing device includes a fire extinguishing medium storage tank 1 and a control module; the output end of the fire extinguishing medium storage tank 1 is fixedly connected to a delivery pipeline 2, the Venturi valve 3 is installed in the delivery pipeline 2, the directional nozzle array 4 is fixedly connected to the output end of the delivery pipeline 2, and the nozzles are arranged corresponding to the internal component positions of the electrical equipment 5; the pressure detection unit (pressure gauge) and the flow detection unit (flow meter) are respectively installed on the delivery pipeline 2 to collect pressure and fire extinguishing medium flow data in the delivery pipeline 2 in real time; the control module is electrically connected to the Venturi valve 3, the directional nozzle array 4, the pressure detection unit, the flow detection unit, and the Venturi valve 3 drive unit, and can receive decision commands from external systems to control the opening degree of the Venturi valve 3 and the start and stop of the directional nozzle array 4, adjust the fire extinguishing medium spray volume according to pressure and flow data, and simultaneously provide feedback on the device's own operating status data. The control module is a PLC controller.

[0126] In summary, after system installation, first check that all module connections are secure. Once confirmed, power on the main system. The system will automatically enter initialization mode, completing self-tests for each module. After successful self-tests, it will automatically switch to daily monitoring mode without manual intervention. During daily use, the system will continuously capture the operating parameters, environmental conditions, and fire characteristics of electrical equipment 5 through various sensors, performing real-time local data processing and analysis. No manual operation is required throughout the process; only management personnel need to periodically check the local display terminal or cloud management platform to understand the system's operating status and equipment monitoring. When the system detects a hazard and triggers an alarm, it will automatically activate the corresponding mode of audible and visual alarms, local display alarms, and remote push alarms based on the alarm level. After receiving the alarm information, management personnel must promptly check the alarm details, identify the equipment containing the hazard and the type of hazard, and take appropriate measures according to the alarm level. For low-risk alarms, only hazard investigation and confirmation of hazard removal are needed. For medium-risk alarms, non-critical power supplies to the equipment must be cut off while investigating the hazard. For high-risk alarms, personnel must be immediately evacuated from the relevant area. Simultaneously, it must be confirmed that the system's automatically activated precision fire extinguishing equipment and full power cut-off function are operating normally, and emergency support should be coordinated if necessary. After management personnel complete the investigation and handling, they need to send a completion command through the local display terminal, dedicated APP, or cloud platform. The system will then stop the warning signal and update the warning handling record. During routine maintenance, management personnel need to regularly check and maintain each module according to the maintenance reminders pushed by the system, replace aging components in a timely manner, and ensure the normal operation of the system. At the same time, they can use the system's record query function to view historical warning information, handling records, and system operation logs, which facilitates the tracing of potential problems and system optimization. If the system issues a fault alarm, management personnel need to promptly investigate the faulty module, confirm the fault type, and repair it. After the repair is completed, the corresponding module will be restarted to ensure that the system returns to normal monitoring status. Remote management personnel can issue commands such as parameter adjustment and policy update through the cloud platform to achieve remote management and optimization of the system without on-site operation.

[0127] Although embodiments of the invention have been shown and described, the scope of the invention will be defined by the appended claims and their equivalents by those skilled in the art.

Claims

1. An intelligent early warning and fire extinguishing system based on electrical equipment, characterized in that, It includes a multi-dimensional perception and early warning module, an edge intelligent data processing and analysis module, an intelligent hierarchical decision-making module, a multi-mode early warning execution module, a precise fire extinguishing control and execution module, a system linkage and feedback module, and a system self-test and predictive maintenance module that is connected to all of the above modules in sequence. The multi-dimensional perception and early warning module is used to collect electrical parameters, environmental parameters and fire characteristics data of electrical equipment (5) and transmit them to the edge intelligent data processing and analysis module; The edge intelligent data processing and analysis module is used to preprocess the collected data, identify anomalies, perform hazard classification analysis, and output the analysis results. The intelligent hierarchical decision-making module is used to conduct a comprehensive risk assessment based on the analysis results and multi-dimensional indicators, and output differentiated early warning and fire extinguishing decision instructions. The multi-mode early warning execution module is used to execute hierarchical multi-channel early warnings according to decision instructions; The precision fire suppression control and execution module is used to execute precision fire suppression operations according to decision instructions; The system linkage and feedback module is used to realize the coordinated linkage of various modules and external devices, and to provide feedback on the operating status data; The system self-check and predictive maintenance module is used to perform periodic self-checks, fault predictions, and maintenance reminders for each module of the system.

2. The intelligent early warning and fire extinguishing system based on electrical equipment according to claim 1, characterized in that, The multi-dimensional perception and early warning module includes a MEMS integrated detection module, electrical parameter sensors, infrared thermal imaging sensors, signal conditioning circuits, and a data acquisition module. The MEMS integrated detection module integrates temperature arc sensors, smoke arc sensors, humidity arc sensors, and fault arc sensors. The electrical parameter sensors include residual current sensors, three-phase current / voltage sensors, and insulation resistance sensors. The multi-dimensional perception and early warning module uses a multi-sensor data fusion algorithm and combines the 3σ principle to set the fault arc detection threshold.

3. The intelligent early warning and fire extinguishing system based on electrical equipment according to claim 2, characterized in that, The edge intelligent data processing and analysis module includes an edge computing node, a data preprocessing unit, a GA-BP neural network identification unit, an anomaly detection unit, and an edge-cloud collaborative communication unit. The edge computing node has a built-in lightweight inference engine to achieve local real-time data processing. The data preprocessing unit uses Min-Max normalization and sliding window denoising algorithms to process the collected data. The GA-BP neural network identification unit combines the global search capability of genetic algorithms with the local optimization capability of BP neural networks for accurate identification of hazard types and levels. The GA-BP neural network is trained and optimized using the Sigmoid activation function and the mean square error function.

4. The intelligent early warning and fire extinguishing system based on electrical equipment according to claim 3, characterized in that, The intelligent hierarchical decision-making module includes a risk level assessment unit, a decision logic unit, a strategy storage unit, and an emergency linkage triggering unit. The risk level assessment unit integrates four dimensions of indicators: hazard level, equipment importance, environmental factors, and fault arc characteristics. It uses the analytic hierarchy process to determine the weight of each indicator and quantifies the comprehensive risk level. The decision-making logic unit outputs a four-level response strategy based on the comprehensive risk level, corresponding to normal status, level one warning, level two warning, and level three warning, respectively.

5. The intelligent early warning and fire extinguishing system based on electrical equipment according to claim 4, characterized in that, The precision fire extinguishing control and execution module includes a Venturi valve (3), a directional nozzle array (4), a delivery pipeline (2), a flow detection unit, a pressure detection unit, and a Venturi valve (3) drive unit; the Venturi valve (3) has a pressure adaptive adjustment function, which adjusts the opening of the Venturi valve (3) according to the pressure difference of the delivery pipeline (2) and the target flow rate; the directional nozzle array (4) consists of multiple independently controllable nozzles, each nozzle corresponding to a part of the internal components of the electrical equipment (5).

6. The intelligent early warning and fire extinguishing system based on electrical equipment according to claim 5, characterized in that, The system linkage and feedback module includes a power linkage unit, a monitoring linkage unit, an emergency platform linkage unit, an operating status feedback unit, a data storage unit, and a dual power redundancy unit. The power linkage unit dynamically adjusts the power cut-off delay time according to the decision command and the operating current of the electrical equipment (5) to avoid equipment damage and electric shock risk. The dual power redundancy unit adopts a main power supply and a backup power supply. When the main power supply voltage is lower than the set threshold and the duration is ≥0.5s, it automatically switches to the backup power supply. The operating status feedback unit adopts an error correction algorithm.

7. The intelligent early warning and fire extinguishing system based on electrical equipment according to claim 6, characterized in that, The multi-mode early warning execution module includes an audible and visual early warning unit, a remote push unit, an early warning intensity adjustment unit, an NB-IoT and LoRa dual-mode communication unit, and an early warning log recording unit. The early warning intensity adjustment unit can automatically adjust the audible and visual early warning intensity according to the on-site noise and light intensity. The dual-mode communication unit can monitor the packet loss rate of early warning information transmission in real time, and automatically switch the transmission mode when the packet loss rate exceeds 0.5%. The remote push unit supports three push channels: APP, SMS, and emergency platform, and matches the corresponding push method according to the early warning level.

8. The intelligent early warning and fire extinguishing system based on electrical equipment according to claim 7, characterized in that, The system self-inspection and predictive maintenance module includes a periodic self-inspection triggering unit, a hardware fault detection unit, an algorithm validity verification unit, a communication link detection unit, a fault level determination unit, a maintenance strategy generation unit, and a maintenance push unit. The system self-inspection and predictive maintenance module automatically calculates the optimal self-inspection cycle and performs a full-coverage self-inspection of all hardware, software algorithms, and communication links in the system. It adopts a multi-dimensional fault determination and component remaining service life prediction algorithm to generate and push differentiated maintenance strategies based on the fault level.

9. An intelligent early warning and fire extinguishing device based on electrical equipment, characterized in that, Including the intelligent early warning and fire extinguishing system as described in claim 7; The intelligent early warning fire extinguishing equipment includes a fire extinguishing medium storage tank (1) and a control module; The output end of the fire extinguishing medium storage tank (1) is fixedly connected to the delivery pipeline (2), the Venturi valve (3) is installed in the delivery pipeline (2), the directional nozzle array (4) is fixedly connected to the output end of the delivery pipeline (2), and the nozzles are arranged in accordance with the internal component positions of the electrical equipment (5). The pressure detection unit and flow detection unit are respectively installed on the conveying pipeline (2) to collect pressure and fire extinguishing medium flow data in the conveying pipeline (2) in real time. The control module is electrically connected to the Venturi valve (3), the directional nozzle array (4), the pressure detection unit, the flow detection unit and the Venturi valve (3) drive unit respectively. It can receive decision commands from external systems, control the opening degree of the Venturi valve (3) and the start and stop of the directional nozzle array (4), adjust the amount of fire extinguishing medium sprayed according to pressure and flow data, and at the same time, it can feed back the device's own operating status data.

Images