Charging pile fire warning system and method based on AI

Abstract

The present invention belongs to the technical field of charging warning, and provides a charging pile fire warning system and method based on AI. The method includes: obtaining the vehicle identity information of the charging vehicle entering the charging pile area, and determining the fire warning sensitivity coefficient according to the vehicle identity information; obtaining the real-time temperature of the charging vehicle during the charging process and the real-time power supply parameters of the corresponding charging pile, using the fire prediction model to process the real-time temperature and the real-time power supply parameters, and predicting the first fire probability; determining the second fire probability according to the first fire probability and the fire warning sensitivity coefficient, and outputting a fire warning signal when the second fire probability is higher than the fire probability threshold. The scheme of the present invention accurately analyzes the probability of fire in the charging vehicle based on the monitoring parameters of both the vehicle and the charging pile, and issues an early warning, which is conducive to the management personnel to handle it in advance, thereby reducing the loss of the fire.

Description

Technical Field

[0001] The present invention relates to the field of charging warning technology, and in particular to an AI-based charging pile fire warning system and method. Background Art

[0002] The charging pile fire warning system is an important technical measure to ensure the safety of electric vehicle charging facilities. Existing charging pile fire warning systems are based on temperature sensors, smoke sensors, video surveillance, etc. to alarm existing fires, but cannot achieve early prediction and warning of fires, making it difficult to effectively reduce fire losses. The present invention aims to solve or improve this technical problem. Summary of the invention

[0003] In this regard, the present invention provides an AI-based charging pile fire warning method, system, electronic device, computer storage medium and computer program product to solve the above technical problems.

[0004] The present invention discloses an AI-based charging pile fire warning method, which comprises the following steps:

[0005] Obtaining vehicle identity information of a charging vehicle entering a charging pile area, and determining a fire warning sensitivity coefficient according to the vehicle identity information; wherein the fire warning sensitivity coefficient is associated with a health status of the charging vehicle;

[0006] Acquire the real-time temperature of the charging vehicle during the charging process and the real-time power supply parameters of the corresponding charging pile, use the fire prediction model to process the real-time temperature and the real-time power supply parameters, and predict the probability of occurrence of a first fire;

[0007] A second fire occurrence probability is determined according to the first fire occurrence probability and the fire warning sensitivity coefficient, and a fire warning signal is output when the second fire occurrence probability is higher than a fire probability threshold.

[0008] In some embodiments, determining the fire warning sensitivity coefficient according to the vehicle identity information includes:

[0009] Retrieving battery health parameters of the charging vehicle according to the vehicle identity information, the battery health parameters including the number of charge and discharge cycles, the battery usage and storage time, and the battery charge and discharge depth;

[0010] Using a generative adversarial network to uniformly convert the structured parameters in the battery health parameters into natural language text, and using a BERT model to perform semantic understanding on each parameter in the converted battery health parameters to extract a battery health feature matrix;

[0011] The battery health feature matrix is ​​input into a health status classification model to obtain a health score of a battery of a charging vehicle, and the fire warning sensitivity coefficient is calculated based on the health score and a standard health score.

[0012] In some embodiments, the fire warning sensitivity coefficient is calculated based on the health score and the standard health score, including:

[0013] ;

[0014] In the formula, is the fire warning sensitivity coefficient of the charging vehicle, is the basic value of the fire warning sensitivity coefficient; is the health score of the charging vehicle, is the standard health score corresponding to the brand of the charging vehicle. Provides standard health scores for all brands of charging-type vehicles; is the weight coefficient.

[0015] In some embodiments, is calculated based on the number of vehicles corresponding to the standard health score and the corresponding direct proportional function, and The proportional coefficient of the direct proportional function is greater than The proportional coefficient of the direct proportional function.

[0016] In some embodiments, the using a fire prediction model to process the real-time temperature and the real-time power supply parameter to predict a first fire occurrence probability includes:

[0017] Collect charging data of multiple groups of charging vehicles, and label each group of charging data according to the fire occurrence record, wherein the label includes a positive label and a negative label, so as to divide the charging data into positive data and negative data; wherein the charging data includes temperature data of the vehicle and power supply parameters of the charging pile, wherein the power supply parameters include a voltage curve, a current curve, and a power fluctuation curve;

[0018] Mix the positive data and the negative data in a certain proportion to construct a training data set, and use the training data set to train the fire prediction model until a convergence condition is met;

[0019] The real-time temperature and the real-time power supply parameter are processed using the fire prediction model to predict the first fire occurrence probability.

[0020] In some embodiments, the mixing ratio of the positive data and the negative data is determined by:

[0021] Mixing the positive data and the negative data according to a preset ratio;

[0022] During the training process, after the positive data or the negative data is input into the fire prediction model, the corresponding prediction deviation is recorded, and a first deviation cumulative value corresponding to the positive data and a second deviation cumulative value corresponding to the negative data are calculated;

[0023] If the difference between the first deviation cumulative value and the second deviation cumulative value is greater than the difference threshold, the preset ratio is increased, that is, the amount of negative data in the training data that is not input into the fire prediction model for training is reduced.

[0024] The present invention also discloses an AI-based charging pile fire warning system, which includes a sensitivity analysis module, a fire prediction module, and a fire warning processing module;

[0025] The sensitivity analysis module is used to obtain the vehicle identity information of the charging vehicle entering the charging pile area, and determine the fire warning sensitivity coefficient according to the vehicle identity information; wherein the fire warning sensitivity coefficient is associated with the health status of the charging vehicle;

[0026] The fire prediction module is used to obtain the real-time temperature of the charging vehicle during the charging process and the real-time power supply parameters of the corresponding charging pile, and use the fire prediction model to process the real-time temperature and the real-time power supply parameters to predict the first fire occurrence probability;

[0027] The fire warning processing module is used to determine a second fire occurrence probability based on the first fire occurrence probability and the fire warning sensitivity coefficient, and output a fire warning signal when the second fire occurrence probability is higher than a fire probability threshold.

[0028] The present invention also discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement any of the above methods.

[0029] The present invention also discloses a computer storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement any of the above methods.

[0030] The present invention also discloses a computer program product, which implements any of the above methods when the computer program product is run on a terminal.

[0031] The solution of the present invention accurately analyzes the probability of fire in a charging vehicle based on monitoring parameters of both the vehicle and the charging pile, and provides early warning, which is beneficial for management personnel to handle in advance, thereby reducing the loss caused by the fire. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 It is a flow chart of a charging pile fire warning method based on AI disclosed in an embodiment of the present invention;

[0034] Figure 2 It is a schematic diagram of a flow chart of determining a fire warning sensitivity coefficient disclosed in an embodiment of the present invention;

[0035] Figure 3 is a flow chart of a method for determining a mixing ratio of positive data and negative data disclosed in an embodiment of the present invention;

[0036] Figure 4 It is a structural schematic diagram of an AI-based charging pile fire warning system disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following is a description of the implementation of the present application by specific specific embodiments. People familiar with the technology can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0038] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0039] like Figure 1 As shown, an embodiment of the present invention discloses an AI-based charging pile fire warning method, the method comprising the following steps:

[0040] Obtaining vehicle identity information of a charging vehicle entering a charging pile area, and determining a fire warning sensitivity coefficient according to the vehicle identity information; wherein the fire warning sensitivity coefficient is associated with a health status of the charging vehicle;

[0041] Acquire the real-time temperature of the charging vehicle during the charging process and the real-time power supply parameters of the corresponding charging pile, use the fire prediction model to process the real-time temperature and the real-time power supply parameters, and predict the probability of occurrence of a first fire;

[0042] A second fire occurrence probability is determined according to the first fire occurrence probability and the fire warning sensitivity coefficient, and a fire warning signal is output when the second fire occurrence probability is higher than a fire probability threshold.

[0043] After the charging vehicle enters the charging pile area, the monitoring equipment deployed in the charging pile area can obtain the identity information of the charging vehicle, based on which the fire warning sensitivity coefficient of the charging vehicle can be analyzed. The fire warning sensitivity coefficient of the charging vehicle is a coefficient determined based on the probability of fire in a vehicle in a standard health state. In the process of the charging vehicle being connected to the charging pile for charging, the monitoring equipment deployed in the charging pile area can obtain the real-time temperature of the charging vehicle and the power supply parameters of the corresponding charging pile, and the fire prediction module can predict the preliminary probability of fire in the charging vehicle based on this. Considering the fire warning sensitivity coefficient associated with the health status of the vehicle itself, the fire warning sensitivity coefficient is multiplied by the preliminary probability to obtain the final probability of fire. When the probability of fire is higher than the fire probability threshold, it is determined that the charging vehicle has a high probability of fire. At this time, the warning information is output in an appropriate manner to remind the management personnel to check and handle it in advance.

[0044] The monitoring equipment used to obtain identity information in the charging pile area may be a monitoring camera, a radio frequency reader, etc. The monitoring camera can determine the identity information of the charging vehicle based on the extracted vehicle license plate, and the radio frequency reader can obtain the vehicle's identity information by reading the electronic tag arranged on the charging vehicle or the radio frequency card manually swiped by the passengers in the vehicle. The charging pile area is also equipped with a temperature sensor (preferably arranged in the monitoring camera, i.e., an infrared camera) and a communication device. The infrared camera directly collects the real-time temperature of the charging vehicle during the charging process, and the communication device communicates with the corresponding charging pile to read the power supply parameters of the charging pile, which include voltage curve, current curve, power fluctuation curve, etc.

[0045] In some embodiments, Figure 2 As shown, the determining of the fire warning sensitivity coefficient according to the vehicle identity information includes:

[0046] Retrieving battery health parameters of the charging vehicle according to the vehicle identity information, the battery health parameters including the number of charge and discharge cycles, the battery usage and storage time, and the battery charge and discharge depth;

[0047] Using a generative adversarial network to uniformly convert the structured parameters in the battery health parameters into natural language text, and using a BERT model to perform semantic understanding on each parameter in the converted battery health parameters to extract a battery health feature matrix;

[0048] The battery health feature matrix is ​​input into a health status classification model to obtain a health score of a battery of a charging vehicle, and the fire warning sensitivity coefficient is calculated based on the health score and a standard health score.

[0049] In this embodiment, the fire warning sensitivity coefficient is calculated based on the standard health status of the battery of the vehicle group and the health status of the battery of the charging vehicle itself. The parameters that affect the health status of the battery of the charging vehicle mainly include:

[0050] 1) Number of battery charge and discharge cycles: The life of a battery is closely related to the number of charge and discharge cycles it has experienced. As the number of charge and discharge cycles increases, the active substances inside the battery gradually decrease, resulting in a decrease in battery capacity and an increase in battery internal resistance, which increases the risk of battery overheating and thermal runaway. The more cycles a battery has, the worse its health status is, and the higher the probability of a fire.

[0051] 2) Battery usage and storage time: As the battery is used and stored for a longer period of time, its health will gradually deteriorate. Long-term use and storage will lead to instability in the chemical reaction inside the battery, increasing the risk of battery failure. When the battery is not in use, if the storage conditions are improper, such as too high or too low temperature, it will also accelerate the aging process of the battery, thereby increasing the possibility of fire.

[0052] 3) Battery charge and discharge depth: The charge and discharge depth has a significant effect on the battery health. Different charge and discharge depths have different effects on battery life. Deep charge and discharge will activate more active substances inside the battery and release more power, but it will also cause the battery health to decay faster. If the battery is frequently deeply discharged, especially if it is not charged for a long time when the battery is at low power, the risk of fire may increase due to the drastic changes in the chemical reaction inside the battery.

[0053] Since the collected battery health parameters may contain both structured data and unstructured data, in order to facilitate the subsequent extraction of battery health features, the present invention first uses a generative adversarial network to uniformly convert the structured parameters into natural language text, and then uses the BERT model to perform semantic analysis on all the converted unstructured battery health parameters, thereby extracting each group of battery health features and forming a battery health feature matrix.

[0054] After obtaining the battery health feature matrix, the health status classification model is used to classify the battery health status of the charging vehicle to obtain the corresponding score value. The battery health score of the vehicle group based on the charging type is pre-derived by statistical analysis to obtain the standard health score. The fire warning sensitivity coefficient of the charging vehicle can be calculated based on the health score of the charging vehicle and the standard health score. Therefore, the fire warning sensitivity coefficient represents the deviation of the general battery health status of the charging vehicle compared to the charging vehicle group. Generally, it is a negative deviation, that is, the battery health status is worse, and the deviation is indirectly related to the probability of battery fire.

[0055] Among them, the health status classification model preferably uses models such as multi-layer perceptron (MLP), random forest, and recurrent neural network (RNN) that can handle complex classification tasks involving multiple factors.

[0056] In some embodiments, the fire warning sensitivity coefficient is calculated based on the health score and the standard health score, including:

[0057]

[0058] In the formula, is the fire warning sensitivity coefficient of the charging vehicle, is the basic value of the fire warning sensitivity coefficient; is the health score of the charging vehicle, is the standard health score corresponding to the brand of the charging vehicle. Provides standard health scores for all brands of charging-type vehicles; is the weight coefficient.

[0059] In this embodiment, the fire warning sensitivity coefficient of the charging vehicle can be calculated based on the above formula. The above formula divides the standard health score into the standard health score corresponding to the brand of the charging vehicle and the standard health score corresponding to all brands of charging type vehicles. The numerator of the above formula represents the deviation of the health status of the charging vehicle from other vehicle groups of the same brand, and the denominator represents the deviation of the health status of the charging vehicle of this brand from all brands of charging vehicles. The fire warning sensitivity coefficient calculated in this way can more accurately reflect the deviation in health status between the charging vehicle and the charging type vehicle group, which is conducive to determining a more accurate probability of fire.

[0060] It should be noted that the method for determining the standard health scores corresponding to charging type vehicles of all brands is the same as the method for determining the standard health scores corresponding to charging type vehicles of specific brands mentioned above, and the details will not be repeated here.

[0061] In some embodiments, is calculated based on the number of vehicles corresponding to the standard health score and the corresponding direct proportional function, and The proportional coefficient of the direct proportional function is greater than The proportional coefficient of the direct proportional function.

[0062] In this embodiment, It is calculated based on the number of vehicles corresponding to the standard health score. Specifically, the more vehicles used in the score, the higher the confidence of the corresponding standard health score, and the larger the corresponding weight coefficient is set, and vice versa. At the same time, since the number of vehicles corresponding to the standard health score of all brands is significantly higher than the number of vehicles corresponding to the standard health score of a specific brand, in order to avoid the denominator of the above calculation formula being too large, resulting in a fire warning sensitivity coefficient that is too small, the present invention sets a weight coefficient determined according to the corresponding number of vehicles. The function of the size should be different. Specifically, setting The proportionality coefficient of the function is larger, while The proportionality coefficient of the function is smaller, for example , ,in, is the number of vehicles used for scoring,

[0063] is the proportionality coefficient of the function, and .

[0064] In some embodiments, the using a fire prediction model to process the real-time temperature and the real-time power supply parameter to predict a first fire occurrence probability includes:

[0065] Collect charging data of multiple groups of charging vehicles, and label each group of charging data according to the fire occurrence record, wherein the label includes a positive label and a negative label, so as to divide the charging data into positive data and negative data; wherein the charging data includes temperature data of the vehicle and power supply parameters of the charging pile, wherein the power supply parameters include a voltage curve, a current curve, and a power fluctuation curve;

[0066] Mix the positive data and the negative data in a certain proportion to construct a training data set, and use the training data set to train the fire prediction model until a convergence condition is met;

[0067] The real-time temperature and the real-time power supply parameter are processed using the fire prediction model to predict the first fire occurrence probability.

[0068] In this embodiment, it is preferred to use a Transformer model with an attention mechanism to construct the fire prediction model in the present invention, and the constructed fire prediction model needs to be trained. Collect charging data of multiple groups of charging vehicles, and each group of charging data is configured with corresponding fire occurrence records, such as "fire occurred" and "no fire occurred". These records can be used to add labels to each group of charging data. "Fire occurred" corresponds to a positive label, and "no fire occurred" corresponds to a negative label. Based on these labels, the charging data is divided into positive data and negative data.

[0069] Positive data and negative data are mixed in a certain ratio, such as 2:1, 3:2, etc. During the training process, the mutual coordination of positive data and negative data can guide the fire prediction model to deeply analyze the functional relationship between vehicle temperature and charging pile power supply parameters and fire probability in the specified direction, thereby having the ability to accurately analyze the probability of fire.

[0070] In some embodiments, Figure 3 As shown, the mixing ratio of the positive data and the negative data is determined by:

[0071] Mixing the positive data and the negative data according to a preset ratio;

[0072] During the training process, after the positive data or the negative data is input into the fire prediction model, the corresponding prediction deviation is recorded, and a first deviation cumulative value corresponding to the positive data and a second deviation cumulative value corresponding to the negative data are calculated;

[0073] If the difference between the first deviation cumulative value and the second deviation cumulative value is greater than the difference threshold, the preset ratio is increased, that is, the amount of negative data in the training data that is not input into the fire prediction model for training is reduced.

[0074] In this embodiment, positive data and negative data are mixed in advance according to a preset ratio. Since positive data plays a major role in the training of the fire prediction model, the proportion of positive data in the preset ratio is set to be slightly higher than that of negative data, such as 2:1, 3:2, etc. The preset ratio can be determined according to empirical values.

[0075] At the same time, during the training process, it is also necessary to monitor the training effect. Specifically, after each positive data or negative data is input into the fire prediction model, the corresponding prediction deviation (that is, the deviation between the predicted value output by the fire prediction model and the corresponding label) is recorded, and the first deviation cumulative value corresponding to the positive data and the second deviation cumulative value corresponding to the negative data are calculated. If the difference between the two deviation cumulative values ​​is too large (ideally, the fire prediction model should have close prediction accuracy for positive and negative data), it means that the proportion of negative data is too high. At this time, the proportion of negative data can be appropriately reduced. The way to reduce it is to reduce the number of negative data that are not input into the fire prediction model for training until the adjusted preset ratio is reached (based on all training data before training).

[0076] Among them, the degree of increase of the preset ratio can be a preset value, for example, only 2 units are increased at a time. For example, the preset ratio is 6:4, and this time the preset ratio is increased by 2 units to 8:2. The number of adjustments can be multiple times, which can be determined according to the monitoring effect of the difference, but an upper limit value should be set, for example, 5 times. In addition, if the amount of negative data in the remaining training data is not enough to adjust the preset ratio to the specified ratio, it can be achieved by inputting positive data and negative data at the same time, which will not be described in detail.

[0077] like Figure 4 As shown, an embodiment of the present invention further discloses an AI-based charging pile fire warning system, the system comprising a sensitive analysis module, a fire prediction module, and a fire warning processing module;

[0078] The sensitivity analysis module is used to obtain the vehicle identity information of the charging vehicle entering the charging pile area, and determine the fire warning sensitivity coefficient according to the vehicle identity information; wherein the fire warning sensitivity coefficient is associated with the health status of the charging vehicle;

[0079] The fire prediction module is used to obtain the real-time temperature of the charging vehicle during the charging process and the real-time power supply parameters of the corresponding charging pile, and use the fire prediction model to process the real-time temperature and the real-time power supply parameters to predict the first fire occurrence probability;

[0080] The fire warning processing module is used to determine a second fire occurrence probability based on the first fire occurrence probability and the fire warning sensitivity coefficient, and output a fire warning signal when the second fire occurrence probability is higher than a fire probability threshold.

[0081] An embodiment of the present invention further discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method described in the above embodiment.

[0082] An embodiment of the present invention further discloses a computer storage medium, wherein the computer storage medium stores a computer program, and the computer program is executed by a processor to implement the method described in the above embodiment.

[0083] The embodiment of the present invention further discloses a computer program product. When the computer program product is run on a terminal, the method described in the above embodiment is implemented.

[0084] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all functions of some or all components in the device according to an embodiment of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., computer program and computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0085] References herein to "one embodiment," "embodiment," or "one or more embodiments" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present invention. In addition, please note that examples of the term "in one embodiment" do not necessarily all refer to the same embodiment.

[0086] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0087] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc., does not indicate any order. These words may be interpreted as names.

[0088] In addition, it should be noted that the language used in this specification is selected primarily for readability and teaching purposes, rather than for explaining or limiting the subject matter of the present invention. Therefore, many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the appended claims. The disclosure of the present invention is illustrative rather than restrictive with respect to the scope of the present invention, and the scope of the present invention is defined by the appended claims.

Claims

1. An AI-based charging pile fire warning method, characterized by: The method comprises the following steps: Obtaining vehicle identity information of a charging vehicle entering a charging pile area, and determining a fire warning sensitivity coefficient according to the vehicle identity information; wherein the fire warning sensitivity coefficient is associated with a health status of the charging vehicle; Acquire the real-time temperature of the charging vehicle during the charging process and the real-time power supply parameters of the corresponding charging pile, use the fire prediction model to process the real-time temperature and the real-time power supply parameters, and predict the probability of occurrence of a first fire; Determining a second fire occurrence probability according to the first fire occurrence probability and the fire warning sensitivity coefficient, and outputting a fire warning signal when the second fire occurrence probability is higher than a fire probability threshold; The determining of the fire warning sensitivity coefficient according to the vehicle identity information includes: Retrieving battery health parameters of the charging vehicle according to the vehicle identity information, the battery health parameters including the number of charge and discharge cycles, the battery usage and storage time, and the battery charge and discharge depth; Using a generative adversarial network to uniformly convert the structured parameters in the battery health parameters into natural language text, and using a BERT model to perform semantic understanding on each parameter in the converted battery health parameters to extract a battery health feature matrix; Inputting the battery health feature matrix into a health status classification model to obtain a health score of the battery of the charging vehicle, and calculating the fire warning sensitivity coefficient according to the health score and a standard health score; The fire warning sensitivity coefficient is calculated based on the health score and the standard health score, including: In the formula, ε is the fire warning sensitivity coefficient of the charging vehicle, ε1 is the basic value of the fire warning sensitivity coefficient; I is the health score of the charging vehicle, I1 is the standard health score corresponding to the brand of the charging vehicle, and I2 is the standard health score corresponding to all brands of charging type vehicles; α1 and α2 are weight coefficients; α1 and α2 are calculated based on the corresponding number of vehicles for determining the standard health score and the corresponding direct proportional function, and the proportional coefficient of the direct proportional function of α1 is greater than the proportional coefficient of the direct proportional function of α2.

2. The Al-based charging pile fire warning method according to claim 1, characterized in that: The using a fire prediction model to process the real-time temperature and the real-time power supply parameter to predict a first fire occurrence probability includes: Collect charging data of multiple groups of charging vehicles, and label each group of charging data according to the fire occurrence record, wherein the label includes a positive label and a negative label, so as to divide the charging data into positive data and negative data; wherein the charging data includes temperature data of the vehicle and power supply parameters of the charging pile, wherein the power supply parameters include a voltage curve, a current curve, and a power fluctuation curve; Mix the positive data and the negative data in a certain proportion to construct a training data set, and use the training data set to train the fire prediction model until a convergence condition is met; The real-time temperature and the real-time power supply parameter are processed using the fire prediction model to predict the first fire occurrence probability.

3. The Al-based charging pile fire warning method according to claim 2, characterized in that: The mixing ratio of the positive data and the negative data is determined by: Mixing the positive data and the negative data according to a preset ratio; During the training process, after the positive data or the negative data is input into the fire prediction model, the corresponding prediction deviation is recorded, and a first deviation cumulative value corresponding to the positive data and a second deviation cumulative value corresponding to the negative data are calculated; If the difference between the first deviation cumulative value and the second deviation cumulative value is greater than the difference threshold, the preset ratio is increased, that is, the amount of negative data in the training data that is not input into the fire prediction model for training is reduced.

4. An AI-based charging pile fire warning system, the system is based on the method according to any one of claims 1 to 3, characterized in that: The system includes a sensitivity analysis module, a fire prediction module, and a fire warning processing module; The sensitivity analysis module is used to obtain the vehicle identity information of the charging vehicle entering the charging pile area, and determine the fire warning sensitivity coefficient according to the vehicle identity information; wherein the fire warning sensitivity coefficient is associated with the health status of the charging vehicle; The fire prediction module is used to obtain the real-time temperature of the charging vehicle during the charging process and the real-time power supply parameters of the corresponding charging pile, and use the fire prediction model to process the real-time temperature and the real-time power supply parameters to predict the first fire occurrence probability; The fire warning processing module is used to determine a second fire occurrence probability based on the first fire occurrence probability and the fire warning sensitivity coefficient, and output a fire warning signal when the second fire occurrence probability is higher than a fire probability threshold.

5. An electronic device comprising: At least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method according to any one of claims 1 to 3.

6. A computer storage medium storing a computer program, characterized in that: The computer program is executed by a processor to implement the method according to any one of claims 1 to 3.

7. A computer program product, characterized in that: When the computer program product runs on a terminal, the method according to any one of claims 1 to 3 is implemented.

Images