A fire warning device and fire extinguishing system for electrical distribution cabinet

Abstract

A fire warning device and fire extinguishing system for an electrical distribution cabinet, belonging to the field of alarm technology. The fire warning device includes a fire recognition model, which is trained by a Bayesian network, and the training process includes the following steps: inferring the distribution of a hidden variable Z with the parameters of the Bayesian network at the current time and the i-th training set data vector, and calculating the expectation of the log likelihood about Z; obtaining the parameters of the Bayesian network at the next time t+1 according to the expectation; calculating the error of the Bayesian network trained by the i-th training set data vector; further updating the Bayesian network parameters according to the error; repeating the above steps, training the Bayesian network, and verifying it on a verification set until the stop condition is reached, completing the training of the Bayesian network to obtain a fire recognition model. The present invention trains the Bayesian network into a fire recognition model through observable variables and hidden variables, thereby improving the ability of fire recognition.

Description

Technical Field

[0001] The invention relates to a fire early warning device and a fire extinguishing system for an electrical distribution cabinet, belonging to the technical field of alarm devices. Background Art

[0002] For example, the Chinese invention patent application with publication number CN118397775A discloses an early warning and fire extinguishing system for electrical distribution cabinets, and its technical solution is: receiving video data of the distribution cabinet and temperature data of key nodes in the distribution cabinet; meshing the three-dimensional model of the distribution cabinet, obtaining the internal temperature distribution of the model through thermoelectric coupling simulation, and obtaining each spatial node and corresponding temperature data based on the internal temperature distribution result of the model; comparing the temperature corresponding to each spatial node and the temperature of the key node with the set threshold, if any temperature is greater than the set first threshold, responding to the early warning control instruction; if the temperature is greater than the set second threshold, sending a fire extinguishing instruction to the fire extinguishing execution subsystem, and executing the fire extinguishing. This invention patent application comprehensively judges the fire extinguishing triggering time, and timely extinguishes the fire on the basis of ensuring the safety of the lines inside the distribution cabinet.

[0003] However, the main causes of fires in electrical distribution cabinets include: after the insulator in the wire is damaged, the live wire contacts the adjacent wire at a certain point, causing a sudden increase in current and causing a short circuit, resulting in instantaneous heat, burning the insulation layer and causing a fire; a certain position of the line is due to moisture, high temperature, impact and pressure, etc., resulting in a decrease in the insulation capacity of the wire insulation or bracket material, resulting in leakage and causing a fire; when the current passing through the wire exceeds the safe current carrying capacity, the temperature of the wire is still rising, causing a load, accelerating the aging of the wire, igniting combustibles and causing a fire; with the increase in usage time, the components of the distribution box may age, be damaged, etc. and cause a fire.

[0004] However, the above prior art only extinguishes fire based on temperature, which often causes misoperation and wastes fire extinguishing agents. In addition, fires in electrical distribution cabinets are often caused by some unobserved hidden variables, and how to identify these hidden variables is not reported in the prior art. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the present invention provides a fire warning device and a fire extinguishing system for an electrical distribution cabinet, which trains a Bayesian network into a fire recognition model through observable variables and latent variables, thereby improving the ability of fire recognition.

[0006] To achieve the above-mentioned purpose, the present invention provides a fire warning device for an electrical distribution cabinet, which includes: a fire recognition model, which recognizes the probability of fire according to the spark spectrum parameters in the electrical distribution cabinet, the aging degree of the material in the electrical distribution cabinet and the temperature value in the electrical distribution cabinet; the fire recognition model is based on a Bayesian network Trained, are the parameters of the Bayesian network, and the training process includes the following steps:

[0007] S2-1: Bayesian Network Parameters of the current time t and the i-th training set data vector Infer the distribution of the latent variable Z , and calculate the log-likelihood Expectations for Z:

[0008] ,

[0009] In the formula, , is the i-th training set data vector The wavelength of the ignitable spectrum in the medium spark spectrum, is the i-th training set data vector The peak of the ignitable spectrum in the medium spark spectrum, is the i-th training set data vector The temperature inside the electrical distribution cabinet; is the i-th training set data vector The aging degree of materials in the electrical distribution cabinet; Indicates that the parameter is , the training set data vector is Time Z's expectations;

[0010] S2-2: The Bayesian network is obtained according to the following formula Parameters for the next time t+1 :

[0011] ;

[0012] S2-3: Calculate the i-th training set data vector Training a Bayesian Network The error generated ;

[0013] S2-4: Update according to the following formula , complete a training:

[0014] ,

[0015] In the formula, Express The gradient of; η is the learning rate of training; is the i-th test set data vector; i=1,…,N, N is the total number of training set data vectors or test set data vectors;

[0016] S2-5: Assignment , repeat steps S2-1 to S2-4 to perform Bayesian network training, and on the validation set Verify on until the stopping condition is reached and the sample Bayesian network is completed The fire recognition model is obtained by training.

[0017] In order to achieve the above-mentioned purpose of the invention, the present invention also provides a fire extinguishing system, including the above-mentioned fire warning device for electrical distribution cabinets and automatic fire extinguishers connected through a network. When the automatic fire extinguisher receives the fire warning information from the fire warning device for the electrical distribution cabinet, it makes preparations for fire disasters. When receiving the fire extinguishing information, the ignition unit ignites the fire extinguishing agent in the fire extinguishing agent box.

[0018] Compared with the prior art, the fire warning device and fire extinguishing system for an electrical distribution cabinet provided by the present invention trains a Bayesian network into a fire recognition model through observable variables and latent variables, thereby improving the ability of fire recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The present invention is a block diagram of the fire extinguishing system for an electrical distribution cabinet.

[0020] Figure 2 The present invention is a block diagram of the fire warning device for an electrical distribution cabinet.

[0021] Figure 3 It is a block diagram of the composition of the material aging detection module provided by the present invention.

[0022] Figure 4 The present invention is a block diagram of the automatic extinguisher for an electrical distribution cabinet. DETAILED DESCRIPTION

[0023] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0024] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. The meaning of "several" is one or more, unless otherwise clearly and specifically defined.

[0025] Figure 1 FIG. 1 is a block diagram of a fire extinguishing system for an electrical distribution cabinet provided by an example of the present invention, such as Figure 1 As shown, the fire extinguishing system for an electrical distribution cabinet provided by the present invention includes a fire warning device, an automatic fire extinguisher and a host computer for the electrical distribution cabinet connected through a network, and the fire warning device and the automatic fire extinguisher for the electrical distribution cabinet are paired. Optionally, the fire warning device and the automatic fire extinguisher for the electrical distribution cabinet are made into one. The network includes a local area network, a 4G network, a 5G network, and also includes a wireless local area network of the IEEE802.11 standard.

[0026] Figure 2 The block diagram of the fire warning device for an electrical distribution cabinet provided by the present invention is as follows: Figure 2 As shown, the fire warning device for an electrical distribution cabinet provided by the present invention includes: a detection module, a first data processor and a first communication unit, wherein the detection module includes a light detector, an odor detector and a temperature detector, wherein the light detector detects the spectrum of the spark discharge from the incident light beam and converts the spectrum into a first electrical information; the odor detector is used to detect the odor in the electrical distribution cabinet and convert the odor into a second electrical information; the first data processor includes an ignitable spectrum recognition module, a material aging detection module, a fire recognition model and a comparison module, wherein the ignitable spectrum recognition module determines the parameters of the ignitable spark spectrum according to the first electrical information, wherein the parameters include the wavelength and peak value of the ignitable spectrum; the material aging detection module determines the aging degree value of the material in the electrical distribution cabinet according to the second electrical information; the fire recognition model generates the probability of fire according to the parameters of the ignitable spark spectrum, the aging degree value of the material in the electrical distribution cabinet and the temperature value provided by the temperature detector, and the comparison module is used to compare the probability of fire with the first threshold value, and if the probability of fire is greater than or equal to the first threshold value, the first data processor sends a fire warning message to the automatic fire extinguisher and the host computer through the first communication unit. In the present invention, the first threshold is 0.8-0.9.

[0027] In the present invention, the comparison module is used to compare the probability of fire with the second threshold. If the probability of fire is greater than or equal to the second threshold, the first data processor sends fire extinguishing information to the fire extinguisher and the host computer through the first communication unit. The second threshold is greater than the first threshold, and the second threshold is usually 0.9 to 1.

[0028] If an electric spark is generated in an electrical distribution cabinet, and a peak is observed in at least one of the wavelengths of the spark spectrum within the first band of 552nm to 555nm, the wavelength of the spark spectrum within the second band of 588nm to 591nm, and the third band of 619nm to 623nm, the following relationship is confirmed: regardless of the material that generates the spark, the spark is flammable. Therefore, the wavelength and peak of the spark spectrum are input as characteristic values ​​into the fire identification model. The fire identification model calculates the probability of fire in the electrical distribution cabinet based on the parameters of the spark spectrum in the electrical distribution cabinet, the temperature value in the electrical distribution cabinet, and the aging degree of various components, wires, etc. in the electrical distribution cabinet.

[0029] In the present invention, the aging degree values ​​of various components, wires and other materials in the electrical distribution cabinet are identified by a neural network, such as an LLM model.

[0030] The fire warning device for an electrical distribution cabinet provided by the present invention further includes a first storage unit, which is used to store control programs, module programs and data, such as the ignitable spectrum recognition module, the material aging detection module, the fire recognition model and the comparison module program. The controlled control program, for example, controls the odor detector, the light detector, the temperature sensor and the first communication unit program. The data includes intermediate data and result data of the module and control program operation process, and also includes the first threshold value and the second threshold value data provided to the comparison module.

[0031] Figure 3 is a block diagram of the material aging detection module provided by the present invention, such as Figure 3 As shown, the material aging detection module includes a normalization layer, an input mapping layer and a neural network, wherein the normalization layer is used to normalize the second electrical information provided by the odor sensor to obtain a normalized vector; the input mapping layer is used to perform a convolution operation on the normalized vector to align the input layer dimension of the neural network; the neural network predicts the material aging degree value according to the vector output by the input mapping layer. In the present invention, the neural network is preferably an LLM model.

[0032] In the present invention, the LLM model is trained through the following steps to generate a material aging detection module:

[0033] S1-1: multiple data sets are taken from the database, each data set includes odor measurement data in the electrical distribution cabinet used as training data M and test data A, The test data is the measurement data of material aging degree determined based on expert experience or past occurrence of a certain odor, that is, the training data corresponds to the data provided by the odor sensor, and the test data is the estimated material aging degree value. The pre-trained data vector M is input into the input embedding layer of the LLM model, and the vector is output from the second residual connection layer. , ,in

[0034] ,

[0035] in, , is the current parameter vector of the LLM model; the parameters of the LLM model include: the key value K, query value Q, value V of the multi-head attention mechanism layer, and the matrix coefficient W and bias B of the feedforward neural network of the LLM model, etc. K is a positive integer greater than or equal to 2;

[0036] S1-2: Assume the loss function L is:

[0037] ,

[0038] The gradients of the mean and variance are calculated as follows:

[0039] ,

[0040] S1-3: Update the current parameter vector of the LLM model according to the following formula:

[0041] ,

[0042] In the formula, is the learning rate, Express The gradient of

[0043] S1-4: Determine whether the loss function L is the minimum. If so, end the training. If not, replace the current parameter vector with a new parameter vector and return to S1-1.

[0044] In the present invention, the fire identification model is based on the Bayesian network Trained, are the parameters of the Bayesian network, and the training process includes the following steps:

[0045] S2-1: Bayesian Network Parameters of the current time t and the i-th training set data vector Infer the distribution of the latent variable Z , and calculate the log-likelihood Expectations for Z:

[0046] ,

[0047] In the formula, , is the i-th training set data vector The wavelength of the ignitable spectrum in the medium spark spectrum, is the i-th training set data vector The peak of the ignitable spectrum in the medium spark spectrum, is the i-th training set data vector The temperature inside the electrical distribution cabinet; is the i-th training set data vector The aging degree of materials in the electrical distribution cabinet; Indicates that the parameter is , the training set data vector is Time Z's expectations;

[0048] S2-2: The Bayesian network is obtained according to the following formula Parameters for the next time t+1 :

[0049] ;

[0050] S2-3: Calculate the i-th training set data vector Training a Bayesian Network The error generated ;

[0051] S2-4: Update according to the following formula , complete a training:

[0052] ,

[0053] In the formula, Express The gradient of; η is the learning rate of training; is the i-th test set data vector; i=1,…,N, N is the total number of training set data vectors or test set data vectors;

[0054] S2-5: Assignment , repeat steps S2-1 to S2-4 to perform Bayesian network training, and on the validation set Verify on until the stopping condition is reached and the sample Bayesian network is completed The fire recognition model is obtained by training.

[0055] In the present invention, the i-th training set data vector is calculated by the following formula: Training a Bayesian Network The error generated :

[0056] ,

[0057] In the formula, , T represents transposition, h is the Bayesian network The feature layer hidden code vector; C is the fire probability subdivision level, for example, C = 100; Table norm.

[0058] The fire warning device for an electrical distribution cabinet provided by the present invention trains a Bayesian network into a fire recognition model through observable variables and latent variables, thereby improving the ability of fire recognition.

[0059] Figure 4 1 is a block diagram of an automatic extinguisher for an electrical distribution cabinet provided by the present invention, such as Figure 4 As shown, the automatic fire extinguisher provided by the present invention includes a second data processor, an ignition unit, a fire extinguishing agent box and a second communication unit. The second communication unit is used to communicate with the fire warning device for the electrical distribution cabinet. When the automatic fire extinguisher receives the fire warning information from the fire warning device for the electrical distribution cabinet, it is prepared for fire disasters. When receiving the fire extinguishing information, the fire extinguishing agent in the fire extinguishing agent box is ignited through the ignition unit. The fire extinguishing agent in the fire extinguishing agent box is potassium oxide fire extinguishing agent, which is non-toxic potassium oxide fire extinguishing agent.

[0060] The automatic fire extinguisher provided by the present invention also includes a second storage unit, which is used to store control programs and data, such as a program for controlling the ignition unit, a program for controlling the second communication unit, etc. The data includes intermediate data and result data of the control program running process.

[0061] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A fire warning device for an electrical distribution cabinet, characterized in that: include: A fire identification model that identifies the probability of a fire based on the spark spectrum parameters in the electrical distribution cabinet, the aging degree of the material in the electrical distribution cabinet, and the temperature value in the electrical distribution cabinet; Fire identification model via Bayesian network Trained, are the parameters of the Bayesian network, and the training process includes the following steps: S2-1: Bayesian Network Parameters of the current time t and the i-th training set data vector Infer the distribution of the latent variable Z , and calculate the log-likelihood Expectations for Z: ， In the formula, , is the i-th training set data vector The wavelength of the ignitable spectrum in the medium spark spectrum, is the i-th training set data vector The peak of the ignitable spectrum in the medium spark spectrum, is the i-th training set data vector The temperature inside the electrical distribution cabinet; is the i-th training set data vector The aging degree of materials in the electrical distribution cabinet; Indicates that the parameter is , the training set data vector is Time Z's expectations; S2-2: The Bayesian network is obtained according to the following formula Parameters for the next time t+1 : ； S2-3: Calculate the i-th training set data vector Training a Bayesian Network The error generated ; S2-4: Update according to the following formula , complete a training: ， In the formula, Express The gradient of; η is the learning rate of training; is the i-th test set data vector; i=1,…,N, N is the total number of training set data vectors or test set data vectors; S2-5: Assignment , repeat steps S2-1 to S2-4 to perform Bayesian network training, and on the validation set Verify on until the stopping condition is reached and the sample Bayesian network is completed A fire recognition model is obtained through training; a spark is generated in the electrical distribution cabinet, and when a peak is observed in at least one of the first band of the spark spectrum within the wavelength of 552nm and below 555nm, the second band of the spark spectrum within the wavelength of 588nm and below 591nm, and the third band within the wavelength of 619nm and below 623nm, it is determined that the spark is flammable.

2. The fire warning device for an electrical distribution cabinet according to claim 1, characterized in that: The i-th training set data vector is calculated by the following formula Training a Bayesian Network The error generated : ， In the formula, , T represents transposition, h is the Bayesian network The feature layer hidden code vector of; C is the fire probability subdivision level; Table norm.

3. The fire warning device for an electrical distribution cabinet according to claim 2, characterized in that: It also includes a detection module, which includes a light detector, an odor detector and a temperature detector. The light detector detects a spectrum in the spark discharge from the incident light beam and converts the spectrum into first electrical information. The odor detector is used to detect odor in the electrical distribution cabinet and convert the odor into second electrical information. The temperature detector is used to detect the temperature value inside the electrical distribution cabinet.

4. The fire warning device for an electrical distribution cabinet according to claim 3, characterized in that: The device also includes an ignitable spectrum identification module, which determines parameters of the ignitable spectrum of the spark according to the first electrical information, wherein the parameters include a wavelength and a peak value of the ignitable spectrum.

5. The fire warning device for an electrical distribution cabinet according to claim 4, characterized in that: It also includes a material aging detection module, which determines the aging degree of the material in the electrical distribution cabinet according to the second electrical information.

6. The fire warning device for an electrical distribution cabinet according to claim 5, characterized in that: It also includes a comparison module, which is used to compare the probability of fire with a first threshold value. If the probability of fire is greater than or equal to the first threshold value, the first data processor sends fire warning information to the automatic fire extinguisher and the host computer through the first communication unit; The comparison module is also used to compare the probability of fire with a second threshold. If the probability of fire is greater than or equal to the second threshold, the first data processor sends fire extinguishing information to the fire extinguisher and the host computer through the first communication unit, and the second threshold is greater than the first threshold.

7. The fire warning device for an electrical distribution cabinet according to claim 6, characterized in that: The material aging detection module includes a normalization layer, an input mapping layer and a neural network, wherein the normalization layer is used to normalize the second electrical information provided by the odor sensor to obtain a normalized vector; the input mapping layer is used to perform a convolution operation on the normalized vector to align the input layer dimension of the neural network; the neural network predicts the normalized vector of the material aging degree based on the vector output by the input mapping layer.

8. A fire extinguishing system, characterized in that: It comprises a fire warning device and an automatic fire extinguisher for an electrical distribution cabinet as described in any one of claims 1 to 7 connected via a network. When the automatic fire extinguisher receives fire warning information from the fire warning device for the electrical distribution cabinet, it makes preparations for fire disasters, and when it receives fire extinguishing information, it ignites the fire extinguishing agent in the fire extinguishing agent box through the ignition unit.

9. The fire extinguishing system according to claim 8, characterized in that: The fire extinguishing agent in the fire extinguishing agent box is potassium oxide fire extinguishing agent.

Images