A Fire Alarm Method and Storage Medium for Charging Pile Lines Based on Big Data

Abstract

The present invention provides a method and a storage medium for fire warning of charging pile lines based on big data. By collecting a line operation data set of a target charging pile cluster, the line operation data set including multi-source time-series monitoring data, performing cross-modal feature alignment on the multi-source time-series monitoring data to generate a spatio-temporal correlation feature matrix, inputting the spatio-temporal correlation feature matrix into a pre-trained fire risk prediction model to generate a line anomaly risk probability distribution map, generating a hierarchical warning signal set according to the risk levels corresponding to each spatial node in the line anomaly risk probability distribution map, and triggering a dynamic protection mechanism based on the hierarchical warning signal set. The dynamic protection mechanism includes performing a current cutoff operation on high-risk line segments and performing a power attenuation operation on adjacent line segments. The present invention can comprehensively improve the reliability and intelligent level of the fire warning system for charging pile lines.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to a method and storage medium for fire warning of charging pile lines based on big data. Background Art

[0002] With the rapid development of new energy infrastructure, the fire warning technology for charging pile lines has become the core link to ensure the charging safety of electric vehicles. The current mainstream warning systems mainly rely on single-parameter threshold monitoring or simple multi-source data superposition analysis, and perform passive alarms by setting current over-limit values or temperature warning lines. However, such methods are difficult to capture the dynamic coupling relationship between current fluctuations and temperature changes, and ignore the continuous impact of line aging on electrical and thermal parameters, resulting in a lag in the identification of hidden fire risks and a high false alarm rate. The existing processing methods based on fixed weights to fuse multi-source data cannot adapt to the non-linear characteristic evolution law of charging pile lines under complex working conditions, and the protection measures mostly adopt the extensive disposal of global power-off, which seriously restricts the operation efficiency of charging pile clusters. More critically, traditional methods lack the ability to predict the risk conduction path, and it is difficult to block the chain reaction caused by local faults in a timely manner, and there is an irreconcilable contradiction between ensuring system safety and maintaining service continuity. These technical defects make the existing warning systems difficult to meet the urgent needs for accurate warning and intelligent protection in high-density charging scenarios. Summary of the Invention

[0003] In view of this, the embodiments of this application provide a method and storage medium for fire warning of charging pile lines based on big data. The technical solution of the embodiments of this application is realized as follows:

[0004] On the one hand, the present invention provides a method for fire warning of charging pile lines based on big data. The method includes: collecting a line operation data set of a target charging pile cluster, where the line operation data set includes multi-source time-series monitoring data; performing cross-modal feature alignment on the multi-source time-series monitoring data to generate a spatio-temporal correlation feature matrix, where the cross-modal feature alignment includes dynamically adjusting the weight distribution of temperature change features based on the current fluctuation amplitude, and correcting the correlation relationship between current and temperature features based on the line aging index; inputting the spatio-temporal correlation feature matrix into a pre-trained fire risk prediction model to generate a line anomaly risk probability distribution map, and the fire risk prediction model is trained through the mapping relationship between the historical fire event data set and the multi-source monitoring data; generating a hierarchical warning signal set according to the risk levels corresponding to each spatial node in the line anomaly risk probability distribution map, and the hierarchical warning signal set includes differential control instructions for different charging pile line segments; triggering a dynamic protection mechanism based on the hierarchical warning signal set, and the dynamic protection mechanism includes performing a current truncation operation on high-risk line segments and performing a power attenuation operation on adjacent line segments.

[0005] On the other hand, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the above method are implemented.

[0006] The fire warning method for charging pile lines based on big data provided by the present invention performs spatio-temporal correlation modeling on multi-source time-series monitoring data such as current fluctuations, temperature changes, and line aging through a dynamic cross-modal feature alignment mechanism, generates a spatio-temporal correlation feature matrix integrating physical degradation characteristics, and accurately identifies the spatial distribution characteristics of line anomaly risks based on a pre-trained fire risk prediction model. Combining with a hierarchical warning signal set to trigger a dynamic protection mechanism significantly improves the warning accuracy and response efficiency of fire hazards in charging pile lines. This method actively blocks the risk diffusion path through the coordinated operation of current truncation and power attenuation, effectively overcoming the false alarm and missed alarm problems caused by static feature correlation in traditional methods. At the same time, a feedback mechanism is used to dynamically optimize the protection strategy, maximizing the operation stability of the charging pile system on the basis of ensuring the timeliness of risk disposal. In addition, training the risk prediction model based on the mapping relationship between historical fire events and multi-source monitoring data can adaptively learn the evolution law of hidden risk factors, breaking through the limitations of artificial experience rules, so that the warning decision-making process not only conforms to the physical laws of electrical and thermal coupling but also has the forward-looking judgment ability driven by data, thus comprehensively improving the reliability and intelligent level of the fire warning system for charging pile lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of the implementation process of a fire warning method for charging pile lines based on big data provided by an embodiment of the present application.

[0008] Figure 2 It is a schematic diagram of the composition structure of a fire warning device for charging pile lines based on big data provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0009] In order to make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be further elaborated in detail below in conjunction with the drawings and embodiments. The described embodiments should not be regarded as limitations of the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0010] The fire warning method for charging pile lines based on big data provided by an embodiment of the present application can be executed by a processor in a computer system on the edge side of a charging station or by a cloud server remotely connected to the charging station. Figure 1 It is a schematic diagram of the implementation process of a fire warning method for charging pile lines based on big data provided by an embodiment of the present application. As Figure 1 shown, the method includes the following steps:

[0011] Step S100: Collect the line operation data set of the target charging pile cluster. The line operation data set includes multi-source time-series monitoring data, and the multi-source time-series monitoring data includes a current fluctuation feature sequence, a temperature change feature sequence, and a line aging index sequence.

[0012] In the embodiment of the present invention, the target charging pile cluster is a set of charging piles that need to be monitored for fire warning. These charging piles are related in terms of physical location or function, such as being located in the same charging station or the same area. The line operation data set is a set formed after collecting data on the line operation status of the target charging pile cluster. The multi-source time-series monitoring data refers to data collected from multiple different sources in chronological order, which includes a current fluctuation feature sequence, a temperature change feature sequence, and a line aging index sequence.

[0013] The current fluctuation feature sequence is a data sequence obtained by continuously monitoring the current fluctuation in the charging pile line within a set time range. For example, a high-precision current sensor can be used to sample the current at a frequency of once per minute (only for example, high-frequency sampling can also be used, such as at the kHz level). The current values obtained from each sampling are arranged in chronological order to form the current fluctuation feature sequence. This sequence can reflect the magnitude change of the current at different times, such as sudden increases or decreases in the current. The temperature change feature sequence is a data sequence obtained by monitoring the temperature change of the charging pile line over time. A temperature sensor can be used to measure the line temperature at a frequency of once per minute. The measured temperature values are arranged in chronological order to form the temperature change feature sequence. The line aging index sequence is a data sequence used to measure the aging degree of the charging pile line. Line aging may be caused by various factors, such as long-term use and environmental factors. The line aging index can be obtained by detecting aspects such as the electrical conductivity and insulation performance of the line. For example, the resistance value and insulation resistance value of the line are regularly detected, and these detection results are arranged in chronological order to form the line aging index sequence. This sequence can reflect the degree and trend of line aging.

[0014] In the actual collection process, corresponding sensors, such as current sensors and temperature sensors, can be installed at each key line node of the target charging pile cluster. These sensors transmit the real-time collected data to the data collection device, and the data collection device performs preliminary processing and storage on these data to form the line operation data set.

[0015] Step S200: Perform cross-modal feature alignment on multi-source time-series monitoring data to generate a spatio-temporal correlation feature matrix, where the cross-modal feature alignment includes dynamically adjusting the weight allocation of temperature change features based on the current fluctuation amplitude and correcting the correlation relationship between current and temperature features based on the line aging index.

[0016] Cross-modal feature alignment refers to processing different types of feature data (such as current fluctuation features, temperature change features, and line aging index in the embodiments of the present invention) so that they have consistency and correlation in the spatio-temporal dimension. In the embodiments of the present invention, the specific method of cross-modal feature alignment includes dynamically adjusting the weight allocation of temperature change features based on the current fluctuation amplitude and correcting the correlation relationship between current and temperature features based on the line aging index. Dynamically adjusting the weight allocation of temperature change features based on the current fluctuation amplitude is to dynamically change the importance of temperature change features in subsequent processing according to the magnitude of current fluctuations.

[0017] Correcting the correlation relationship between current and temperature features based on the line aging index is because line aging will affect the normal relationship between current and temperature. For example, as the line ages, the resistance of the line may increase, resulting in a greater increase in temperature under the same current. Therefore, it is necessary to correct the correlation relationship between current and temperature features according to the line aging index to more accurately reflect the actual operating conditions of the line.

[0018] Generating a spatio-temporal correlation feature matrix is to integrate the multi-source time-series monitoring data after cross-modal feature alignment processing to form a matrix containing time and space information. This matrix can more comprehensively reflect the operating state of the lines of the target charging pile cluster. In actual operation, corresponding algorithms and models can be used to achieve cross-modal feature alignment and the generation of spatio-temporal correlation feature matrices. For example, the attention mechanism in deep learning can be used to dynamically adjust the weight allocation of temperature change features based on the current fluctuation amplitude, establish a regression model to correct the correlation relationship between current and temperature features based on the line aging index, and finally arrange the processed data according to time and space dimensions to form a spatio-temporal correlation feature matrix.

[0019] As an implementation method, step S200, performing cross-modal feature alignment on multi-source time-series monitoring data to generate a spatio-temporal correlation feature matrix, can specifically include the following steps S210~S250:

[0020] Step S210: Construct a time-dimensional sliding window according to the amplitude change rate of the current fluctuation feature sequence, and extract the current fluctuation peak-valley difference and duration interval within the sliding window as the first dynamic feature vector.

[0021] The time - dimension sliding window is a method for local analysis on time - series data. It can slide on the data sequence according to a preset step size, and each time intercept a data segment of a fixed length for analysis. In the embodiments of the present invention, a time - dimension sliding window is constructed according to the amplitude change rate of the current fluctuation feature sequence. The amplitude change rate reflects the degree of change of the current fluctuation amplitude with time. For example, by calculating the ratio of the difference between the current fluctuation amplitudes at two adjacent time points to the time interval, the amplitude change rate is obtained. According to the magnitude of the amplitude change rate, the length and step size of the sliding window can be determined.

[0022] Extract the peak - valley difference and duration interval of the current fluctuation within the sliding window as the first dynamic feature vector. The peak - valley difference of the current fluctuation is the difference between the maximum value and the minimum value of the current fluctuation within the sliding window, which reflects the amplitude of the current fluctuation. The duration interval refers to the time elapsed from the peak to the valley or from the valley to the peak of the current fluctuation. Combining these two features forms the first dynamic feature vector.

[0023] Step S220: Extract the temperature gradient change amount and extreme - value distribution density of the temperature change feature sequence within the sliding window as the second dynamic feature vector.

[0024] The temperature gradient change amount refers to the rate of change of temperature with time within the sliding window, which reflects the speed of temperature change. The temperature gradient change amount can be obtained by calculating the ratio of the difference between the temperature values at two adjacent time points to the time interval.

[0025] The extreme - value distribution density refers to the frequency of occurrence of temperature extreme values (maximum and minimum values) within the sliding window. By counting the number of times the temperature maximum and minimum values appear within the sliding window and dividing by the total time of the sliding window, the extreme - value distribution density is obtained. Combining the temperature gradient change amount and the extreme - value distribution density forms the second dynamic feature vector.

[0026] Step S230: Decompose the line aging index sequence into two independent components, namely the conductive material degradation coefficient and the insulation layer damage rate, and calculate the cross - influence factors between the independent components and the first dynamic feature vector and the second dynamic feature vector respectively.

[0027] The line aging index sequence reflects the aging degree of the charging pile line over time. In the embodiments of the present invention, the line aging index sequence is decomposed into two independent components, namely the conductive material degradation coefficient and the insulation layer damage rate. The conductive material degradation coefficient is used to measure the degree of decline in the performance of the conductive material in the line. For example, as the usage time increases, the resistance of the conductive material may increase, and the conductive performance will decline. The conductive material degradation coefficient can be determined by measuring the resistance change of the line, etc. The insulation layer damage rate is the degree of damage to the line insulation layer, which can be evaluated by detecting the insulation resistance, etc.

[0028] Calculating the cross - influence factors between the independent components and the first dynamic eigenvector and the second dynamic eigenvector is to analyze the influence of line aging on current fluctuations and temperature changes. The cross - influence factor reflects the degree of influence of one variable on another. For example, calculating the cross - influence factor between the conductive material degradation coefficient and the first dynamic eigenvector can analyze the influence of the decline in the performance of the conductive material on current fluctuations; calculating the cross - influence factor between the insulation layer breakage rate and the second dynamic eigenvector can analyze the influence of the breakage of the insulation layer on temperature changes. In actual calculations, methods such as correlation analysis can be used to calculate the cross - influence factor. For example, by calculating the Pearson correlation coefficient between the conductive material degradation coefficient and the peak - valley difference of current fluctuations in the first dynamic eigenvector, the cross - influence factor between the two can be obtained. If the correlation coefficient is positive and the value is large, it indicates that an increase in the conductive material degradation coefficient will lead to an increase in the peak - valley difference of current fluctuations; if the correlation coefficient is negative and the value is large, it indicates that an increase in the conductive material degradation coefficient will lead to a decrease in the peak - valley difference of current fluctuations.

[0029] As an implementation manner, in step S230, the line aging index sequence is decomposed into two independent components, namely the conductive material degradation coefficient and the insulation layer breakage rate, and the cross - influence factors between the independent components and the first dynamic eigenvector and the second dynamic eigenvector are calculated respectively. Specifically, it may include the following steps S231 - S235:

[0030] Step S231: Perform frequency - domain decomposition on the line aging index sequence, extract the low - frequency fluctuation component as the conductive material degradation coefficient, and separate the high - frequency mutation component as the insulation layer breakage rate.

[0031] Frequency - domain decomposition is a method of converting a time - domain signal to the frequency domain for analysis. Through frequency - domain decomposition, a signal can be decomposed into components of different frequencies. In the embodiment of the present invention, for the line aging index sequence, frequency - domain decomposition is performed. For example, the fast Fourier transform (FFT) is used to convert the line aging index sequence from the time domain to the frequency domain. The low - frequency fluctuation component usually reflects the slow - changing trend of the signal. In this scenario, the degradation of the conductive material is a relatively slow process, so the low - frequency fluctuation component is extracted as the conductive material degradation coefficient. The high - frequency mutation component reflects the rapid change of the signal. The breakage of the insulation layer may cause sudden changes in the line aging index, so the high - frequency mutation component is separated as the insulation layer breakage rate.

[0032] For example, assume that the line aging index sequence is a time series containing multiple data points. After performing FFT transformation on it, a frequency-domain signal is obtained. By setting an appropriate frequency threshold, the signal in the low-frequency part is extracted and converted back to the time domain through inverse Fourier transform to obtain the conductive material degradation coefficient sequence; the signal in the high-frequency part is extracted and also converted back to the time domain through inverse Fourier transform to obtain the insulation layer damage rate sequence.

[0033] In an alternative embodiment, in order to overcome the decomposition failure generated in low-frequency sampling and the noise contained in high-frequency components, empirical mode decomposition (EMD) can also be used to replace FFT to adapt to non-uniformly sampled data and extract more reasonable low-frequency (conductive degradation) and high-frequency (insulation damage) components.

[0034] Step S232: Align the conductive material degradation coefficient with the peak-valley difference of current fluctuations in the first dynamic feature vector in terms of time, calculate the phase correlation between the two within a sliding window, and generate a current-degradation interaction feature vector.

[0035] Time alignment is to match the data of different time series in the time dimension so that they have a corresponding relationship at the same time point. In the embodiments of the present invention, the conductive material degradation coefficient sequence is aligned with the peak-valley difference sequence of current fluctuations in the first dynamic feature vector in terms of time. For example, if the sampling time interval of the conductive material degradation coefficient sequence is 1 hour and the sampling time interval of the peak-valley difference sequence of current fluctuations is 30 minutes, then one of the sequences is interpolated or sampled so that the time points of the two sequences are consistent.

[0036] Phase correlation refers to the degree of association between two signals in terms of phase. Calculate the phase correlation between the conductive material degradation coefficient and the peak-valley difference of current fluctuations within a sliding window. For example, use the cross-correlation function to calculate the phase correlation between the two. Generate a current-degradation interaction feature vector based on the calculated phase correlation result. This vector reflects the interaction relationship between the conductive material degradation and current fluctuations. For example, if the phase correlation is strong, it indicates that the degradation of the conductive material has a greater impact on current fluctuations; if the phase correlation is weak, it indicates that the impact between the two is smaller.

[0037] Step S233: Spatially match the insulation layer damage rate with the temperature gradient change amount in the second dynamic feature vector, detect the synchronous offset amount in the extreme point distribution of the two, and generate a temperature-damage interaction feature vector.

[0038] Spatial matching is to correspond and compare different data in the spatial dimension. In the embodiments of the present invention, the insulation layer breakage rate is spatially matched with the temperature gradient change amount in the second dynamic feature vector. The space here can be understood as different positions of the charging pile line. For example, the insulation layer breakage rate and the temperature gradient change amount are measured respectively on different line segments of the charging pile, and the insulation layer breakage rate and the temperature gradient change amount on the same line segment are corresponded.

[0039] Detecting the synchronous offset amount in the extreme point distribution of the two is to analyze the offset situation of the extreme points (maximum or minimum value) of the insulation layer breakage rate and the extreme points of the temperature gradient change amount in time or space. For example, if the maximum value of the insulation layer breakage rate appears at a certain time point, and the maximum value of the temperature gradient change amount appears at a later time point, there is a synchronous offset amount in time between the two. By detecting the synchronous offset amount, a temperature-breakage interaction feature vector is generated. This vector reflects the interaction relationship between the insulation layer breakage and the temperature change.

[0040] Step S234: Perform multi-scale convolution processing on the current-degradation interaction feature vector, extract the lag response pattern of the conductive material degradation coefficient to the change in the current fluctuation amplitude, and generate a first cross-influence factor sequence.

[0041] Multi-scale convolution processing is a method of performing convolution operations on data at different scales, which can capture the features of data at different scales. In the embodiments of the present invention, multi-scale convolution processing is performed on the current-degradation interaction feature vector. For example, convolution operations are performed on the current-degradation interaction feature vector using convolution kernels of different sizes (such as 3×3, 5×5, etc.) to obtain feature representations at different scales.

[0042] Extracting the lag response pattern of the conductive material degradation coefficient to the change in the current fluctuation amplitude means analyzing how the change in the conductive material degradation coefficient affects the change in the current fluctuation amplitude and whether there is a time lag in this influence. Through multi-scale convolution processing, this lag response pattern can be extracted from different scales. For example, at a certain scale, it may be found that the change in the conductive material degradation coefficient will cause an obvious change in the current fluctuation amplitude after a period of time, which is a lag response pattern. The extracted lag response patterns are sorted out to generate a first cross-influence factor sequence. This sequence reflects the influence degree and pattern of the conductive material degradation on the current fluctuation.

[0043] Step S235: Perform bidirectional gated recurrent processing on the temperature-breakage interaction feature vector, capture the cumulative effect of the insulation layer breakage rate on the temperature extreme value distribution, and generate a second cross-influence factor sequence.

[0044] Bidirectional gated recurrent processing is a processing method based on recurrent neural networks, which can consider both past and future information of data. In the embodiments of the present invention, bidirectional gated recurrent processing is performed on the temperature-damage interaction feature vector. For example, a bidirectional gated recurrent unit (GRU) is used to process the temperature-damage interaction feature vector. The bidirectional GRU can process the input data from two directions (forward and backward), and can better capture the context information of the data.

[0045] Capturing the cumulative effect of the insulation layer damage rate on the temperature extreme value distribution means analyzing how the change of the insulation layer damage rate has a cumulative impact on the temperature extreme value distribution over time. Through bidirectional gated recurrent processing, the temperature-damage interaction feature vector can be modeled to capture this cumulative effect. For example, as the insulation layer damage rate gradually increases, it may lead to a gradual expansion of the range of the temperature extreme value distribution, and this change is a kind of cumulative effect. Quantify the captured cumulative effect to generate a second cross-influence factor sequence. This sequence reflects the degree and pattern of the impact of the insulation layer damage on the temperature change.

[0046] Step S240: Dynamically assign weights to the cross-influence factors through a multi-channel attention mechanism to generate an association mapping relationship among the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence.

[0047] The multi-channel attention mechanism can dynamically adjust the importance of each feature channel according to the specific situation of the data. In the embodiments of the present invention, the multi-channel attention mechanism is used to dynamically assign weights to the first cross-influence factor sequence and the second cross-influence factor sequence. For example, an attention module based on a convolutional neural network is adopted, and this module can automatically learn the importance weights of each factor according to the input cross-influence factor sequence.

[0048] Generating an association mapping relationship among the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence means establishing the mutual association among these three sequences by dynamically assigning weights to the cross-influence factors. For example, through the weights calculated by the attention mechanism, the first cross-influence factor sequence and the second cross-influence factor sequence are weighted and combined with the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence to obtain a mapping matrix reflecting their association relationship. This mapping matrix can be used for subsequent analysis and processing, such as predicting the abnormal risk of the line, etc.

[0049] As an implementation manner, in step S240, dynamically assign weights to the cross-influence factors through a multi-channel attention mechanism to generate an association mapping relationship among the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence, which may specifically include the following steps S241 to S245:

[0050] Step S241: Concatenate the first cross - influence factor sequence and the current fluctuation feature sequence along the channel dimension to form a current - mode correlation feature block, and concatenate the second cross - influence factor sequence and the temperature change feature sequence along the channel dimension to form a temperature - mode correlation feature block.

[0051] In the embodiment of the present invention, the first cross - influence factor sequence and the current fluctuation feature sequence are concatenated along the channel dimension. Similarly, the second cross - influence factor sequence and the temperature change feature sequence are concatenated along the channel dimension to form a temperature - mode correlation feature block. Channel concatenation can integrate relevant feature information together, facilitating subsequent processing and analysis. For example, by concatenating the first cross - influence factor sequence and the current fluctuation feature sequence, the influence information of conductive material degradation on current fluctuation and the information of the current fluctuation itself can be combined to more comprehensively reflect the characteristics of the current mode.

[0052] Perform the following processing on the current - mode correlation feature block and the temperature - mode correlation feature block through a cross - modal attention gating mechanism:

[0053] Step S242: In the current - dominant attention branch, calculate the weight influence value of each time step of the current - mode correlation feature block on the temperature - mode correlation feature block, and generate an attention distribution map from current to temperature.

[0054] The current - dominant attention branch is a branch in the cross - modal attention gating mechanism, which mainly focuses on the influence of the current - mode correlation feature block on the temperature - mode correlation feature block. In this branch, calculate the weight influence value of each time step of the current - mode correlation feature block on the temperature - mode correlation feature block. For example, through an attention calculation module that adopts the dot - product attention mechanism, taking the current - mode correlation feature block and the temperature - mode correlation feature block as inputs, calculate the similarity between the feature vectors of each time step in the current - mode correlation feature block and the feature vectors of the temperature - mode correlation feature block, and use the similarity as the weight influence value.

[0055] Arrange the calculated weight influence values according to time steps and spatial positions to generate an attention distribution map from current to temperature. This distribution map can intuitively show the distribution of the influence degree of the current - mode correlation feature block on the temperature - mode correlation feature block in terms of time and space. For example, if the weight influence value is large at a certain time step and spatial position, it indicates that at this moment and position, the current - mode correlation feature has a greater influence on the temperature - mode correlation feature.

[0056] Step S243: In the temperature - dominant attention branch, calculate the weight influence value of each spatial node of the temperature - mode correlation feature block on the current - mode correlation feature block, and generate an attention distribution map from temperature to current.

[0057] The temperature-dominated attention branch is another branch in the cross-modal attention gating mechanism, mainly focusing on the influence of the temperature-modal correlation feature block on the current-modal correlation feature block. In this branch, the weight influence values of each spatial node of the temperature-modal correlation feature block on the current-modal correlation feature block are calculated. The dot product attention mechanism can also be used. Taking the temperature-modal correlation feature block and the current-modal correlation feature block as inputs, the similarity between the feature vectors of each spatial node in the temperature-modal correlation feature block and the feature vectors of the current-modal correlation feature block is calculated, and the similarity is used as the weight influence value.

[0058] Arrange the calculated weight influence values according to spatial nodes and time steps to generate an attention distribution map from temperature to current. This distribution map can intuitively show the distribution of the influence degree of the temperature-modal correlation feature block on the current-modal correlation feature block in space and time. For example, if the weight influence value is large at a certain spatial node and time step, it means that at this position and moment, the temperature-modal feature has a greater influence on the current-modal feature.

[0059] Step S244: Perform a dot product operation on the attention distribution map from current to temperature and the temperature-modal correlation feature block to obtain a temperature feature enhancement vector, and perform a dot product operation on the attention distribution map from temperature to current and the current-modal correlation feature block to obtain a current feature enhancement vector.

[0060] In the embodiment of the present invention, a dot product operation is performed on the attention distribution map from current to temperature and the temperature-modal correlation feature block. Through the dot product operation, the influence information of the current-modal correlation feature block on the temperature-modal correlation feature block is incorporated into the temperature-modal correlation feature block to obtain a temperature feature enhancement vector. Similarly, a dot product operation is performed on the attention distribution map from temperature to current and the current-modal correlation feature block to obtain a current feature enhancement vector. In this way, the correlation information between different modal features can be enhanced, enabling subsequent analysis and processing to more accurately reflect the actual operating conditions of the circuit.

[0061] Step S245: Perform cross-projection fusion on the temperature feature enhancement vector and the current feature enhancement vector to generate an association mapping relationship among the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence.

[0062] In the embodiment of the present invention, cross-projection fusion is performed on the temperature feature enhancement vector and the current feature enhancement vector. For example, a linear projection method can be used to project the temperature feature enhancement vector and the current feature enhancement vector into a common feature space, and then operations such as adding or splicing the projected vectors are performed to obtain a fused feature vector.

[0063] Through cross - projection fusion, the characteristic information of the temperature mode and the current mode is integrated to generate the correlation mapping relationship among the current fluctuation characteristic sequence, the temperature change characteristic sequence, and the line aging index sequence. This correlation mapping relationship can more comprehensively reflect the mutual relationship among these three sequences, providing a more accurate basis for subsequent fire risk prediction.

[0064] Step S250: Based on the correlation mapping relationship, perform spatial interpolation fusion on the first dynamic feature vector and the second dynamic feature vector to generate a spatio - temporal correlation feature matrix containing timestamp marks.

[0065] In the embodiment of the present invention, based on the previously generated correlation mapping relationship, spatial interpolation fusion is performed on the first dynamic feature vector and the second dynamic feature vector. The correlation mapping relationship reflects the mutual relationship among the current fluctuation characteristics, the temperature change characteristics, and the line aging index. By using this relationship, the first dynamic feature vector and the second dynamic feature vector can be more reasonably fused. Generating a spatio - temporal correlation feature matrix containing timestamp marks is to arrange the feature vectors after spatial interpolation fusion in time and space dimensions and add timestamp marks to each element. The timestamp mark can record the acquisition time of each feature data, enabling the spatio - temporal correlation feature matrix to more accurately reflect the operating state of the target charging pile cluster line at different times and in different spaces.

[0066] As an implementation manner, in step S250, based on the correlation mapping relationship, perform spatial interpolation fusion on the first dynamic feature vector and the second dynamic feature vector to generate a spatio - temporal correlation feature matrix containing timestamp marks, which may specifically include the following steps:

[0067] Step S251: Perform timestamp alignment processing on the current fluctuation peak - valley difference sequence in the first dynamic feature vector and the temperature gradient change amount sequence in the second dynamic feature vector to generate a time - synchronized current - temperature feature pair sequence.

[0068] Timestamp alignment processing is to match the data of different time series in the time dimension so that they have a corresponding relationship at the same time point. In the embodiment of the present invention, timestamp alignment processing is performed on the current fluctuation peak - valley difference sequence in the first dynamic feature vector and the temperature gradient change amount sequence in the second dynamic feature vector. For example, if the sampling time interval of the current fluctuation peak - valley difference sequence is 10 minutes and the sampling time interval of the temperature gradient change amount sequence is 15 minutes, then one of the sequences needs to be interpolated or sampled so that the timestamps of the two sequences are consistent.

[0069] The sequence of the difference between the peaks and valleys of current fluctuations and the sequence of the change in temperature gradient after timestamp alignment are combined into pairs of feature pairs to generate a time-synchronized current-temperature feature pair sequence. This sequence can more accurately reflect the corresponding relationship between current fluctuations and temperature changes at the same time point.

[0070] Step S252: According to the attention distribution map from current to temperature in the association mapping relationship, extract the interpolation weight distribution of the sequence of the difference between the peaks and valleys of current fluctuations in the spatial dimension to generate a current-dominated spatial interpolation mask matrix.

[0071] According to the attention distribution map from current to temperature in the association mapping relationship, this distribution map reflects the distribution of the influence degree of the current modal association feature block on the temperature modal association feature block in time and space. In the embodiment of the present invention, the interpolation weight distribution of the sequence of the difference between the peaks and valleys of current fluctuations in the spatial dimension is extracted from this attention distribution map. For example, by performing statistical analysis on the attention distribution map in the spatial dimension, the interpolation weight of the sequence of the difference between the peaks and valleys of current fluctuations at each spatial position can be obtained. The extracted interpolation weights are arranged according to the spatial positions to generate a current-dominated spatial interpolation mask matrix. This matrix can be used for subsequent spatial interpolation operations to achieve the smoothing and fusion of the sequence of the difference between the peaks and valleys of current fluctuations in the spatial dimension.

[0072] Step S253: According to the attention distribution map from temperature to current in the association mapping relationship, extract the interpolation weight distribution of the sequence of the change in temperature gradient in the time dimension to generate a temperature-dominated time interpolation mask matrix.

[0073] Similarly, according to the attention distribution map from temperature to current in the association mapping relationship, this distribution map reflects the distribution of the influence degree of the temperature modal association feature block on the current modal association feature block in space and time. In the embodiment of the present invention, the interpolation weight distribution of the sequence of the change in temperature gradient in the time dimension is extracted from this attention distribution map. For example, by performing statistical analysis on the attention distribution map in the time dimension, the interpolation weight of the sequence of the change in temperature gradient at each time point can be obtained.

[0074] The extracted interpolation weights are arranged according to the time points to generate a temperature-dominated time interpolation mask matrix. This matrix can be used for subsequent time interpolation operations to achieve the smoothing and fusion of the sequence of the change in temperature gradient in the time dimension.

[0075] Step S254: Perform a spatial dimension convolution operation on the current-dominated spatial interpolation mask matrix and the time-synchronized current-temperature feature pair sequence to generate an intermediate fusion matrix with enhanced current features.

[0076] Spatial dimension convolution operation is a method for performing convolution operations on data in the spatial dimension, which can achieve spatial smoothing of data and feature extraction. In the embodiments of the present invention, a current-dominated spatial interpolation mask matrix is subjected to a spatial dimension convolution operation with a time-synchronized current-temperature feature pair sequence. For example, a two-dimensional convolution kernel is used to perform a convolution operation on the current-temperature feature pair sequence in the spatial dimension, and the weights of the convolution kernel are determined by the current-dominated spatial interpolation mask matrix.

[0077] Through the spatial dimension convolution operation, the spatial interpolation weight information of the current fluctuation peak-valley difference sequence is incorporated into the time-synchronized current-temperature feature pair sequence to generate an intermediate fusion matrix with enhanced current features. This matrix enhances the current features in the spatial dimension and can more accurately reflect the spatial distribution of current fluctuations.

[0078] Step S255: Perform a time dimension sliding window superposition on the temperature-dominated time interpolation mask matrix and the intermediate fusion matrix with enhanced current features to generate a spatio-temporal joint feature tensor.

[0079] Time dimension sliding window superposition is a method for processing data in the time dimension, which can fuse data at different time points. In the embodiments of the present invention, a temperature-dominated time interpolation mask matrix is subjected to a time dimension sliding window superposition with the intermediate fusion matrix with enhanced current features. For example, a sliding window with a fixed length is used to slide on the intermediate fusion matrix with enhanced current features in the time dimension, and the data within the window is weighted and summed with the temperature-dominated time interpolation mask matrix each time it slides.

[0080] Through the time dimension sliding window superposition, the time interpolation weight information of the temperature gradient change amount sequence is incorporated into the intermediate fusion matrix with enhanced current features to generate a spatio-temporal joint feature tensor. This tensor integrates the feature information in the time and spatial dimensions and can more comprehensively reflect the operating state of the target charging pile cluster line.

[0081] Perform the following interpolation operations on the spatio-temporal joint feature tensor:

[0082] Step S256: In the spatial dimension, according to the cross-projection fusion result in the association mapping relationship, perform bidirectional weighted interpolation on the feature vectors of adjacent line segments to generate a spatial continuity feature layer.

[0083] In the spatial dimension, according to the cross-projection fusion result in the association mapping relationship, this result reflects the comprehensive association relationship among the current fluctuation characteristics, temperature change characteristics, and line aging indicators. In the embodiments of the present invention, this result is used to perform bidirectional weighted interpolation on the feature vectors of adjacent line segments.

[0084] Generate a spatially continuous feature layer by performing bidirectional weighted interpolation on the feature vectors of all adjacent line segments. This feature layer smooths the feature information in the spatial dimension, making the feature changes between adjacent line segments more continuous and enabling a more accurate reflection of the overall operating condition of the line in space.

[0085] Step S257: In the time dimension, based on the time stamp intervals of the original multi-source time-series monitoring data, perform time-domain interpolation and completion on the spatially continuous feature layer to generate a feature distribution map with aligned timestamps.

[0086] In the time dimension, based on the time stamp intervals of the original multi-source time-series monitoring data, which record the data collection time intervals. In the embodiments of the present invention, perform time-domain interpolation and completion on the spatially continuous feature layer.

[0087] Generate a feature distribution map with aligned timestamps through time-domain interpolation and completion. This distribution map completes the feature information in the time dimension, making the feature data more continuous and accurate in time and enabling a more comprehensive reflection of the operating state of the line at different times.

[0088] Step S258: Perform channel superposition on the spatially continuous feature layer and the feature distribution map with aligned timestamps to generate a spatio-temporal correlation feature matrix containing timestamp marks, where each timestamp corresponds to the original acquisition moment of the multi-source time-series monitoring data.

[0089] Channel superposition refers to merging different feature layers in the channel dimension. In the embodiments of the present invention, perform channel superposition on the spatially continuous feature layer and the feature distribution map with aligned timestamps. For example, if the spatially continuous feature layer is a two-dimensional matrix and the feature distribution map with aligned timestamps is also a two-dimensional matrix, then after superposing them in the channel dimension, a three-dimensional matrix is obtained, and this matrix is the spatio-temporal correlation feature matrix containing timestamp marks.

[0090] Each timestamp corresponds to the original acquisition moment of the multi-source time-series monitoring data, which enables the spatio-temporal correlation feature matrix to accurately reflect the operating state of the target charging pile cluster line at different times and in different spaces.

[0091] Step S259: Perform edge feature compensation processing on the spatio-temporal correlation feature matrix, detect the feature attenuation regions at the matrix boundaries, and perform eigenvalue repair according to the dynamic weight allocation ratio in the correlation mapping relationship to generate a spatio-temporal correlation feature matrix that completely covers the topology of the target charging pile line.

[0092] Edge feature compensation processing is to make up for the possible missing or attenuated feature information at the boundary of the spatio-temporal correlation feature matrix. In the embodiments of the present invention, edge feature compensation processing is performed on the spatio-temporal correlation feature matrix. First, the feature attenuation region at the matrix boundary is detected. For example, by calculating the difference between the eigenvalue at the matrix boundary and the internal eigenvalue, the position and degree of feature attenuation are determined.

[0093] Then, eigenvalue repair is performed according to the dynamic weight distribution ratio in the correlation mapping relationship. The dynamic weight distribution ratio in the correlation mapping relationship reflects the importance and correlation degree between different features, and this ratio is used to adjust and repair the eigenvalues in the feature attenuation region. For example, for a certain feature attenuation region, according to the weight distribution ratio of the relevant features in this region in the correlation mapping relationship, preset information is extracted from the eigenvalues in the adjacent region to supplement the eigenvalues in this region.

[0094] Through edge feature compensation processing, a spatio-temporal correlation feature matrix that completely covers the target charging pile line topology is generated. This matrix can more accurately reflect the overall operating state of the target charging pile cluster line and provides more reliable data support for subsequent fire risk prediction.

[0095] Step S300: Input the spatio-temporal correlation feature matrix into a pre-trained fire risk prediction model to generate a line anomaly risk probability distribution map. The fire risk prediction model is trained through the mapping relationship between the historical fire event dataset and multi-source monitoring data.

[0096] The pre-trained fire risk prediction model is a trained model that can predict the line anomaly risk probability according to the input spatio-temporal correlation feature matrix. The model is trained through the mapping relationship between the historical fire event dataset and multi-source monitoring data. The historical fire event dataset contains relevant data when fires occurred in the past, such as the time, location of the fire, and the operating state of the line; the multi-source monitoring data refers to various monitoring data collected from the target charging pile cluster, such as current fluctuation characteristics, temperature change characteristics, etc.

[0097] By establishing the mapping relationship between the historical fire event dataset and multi-source monitoring data, the model is trained using machine learning or deep learning algorithms. For example, model structures such as convolutional neural network (CNN) or recurrent neural network (RNN) can be used. Taking the historical fire event dataset and multi-source monitoring data as inputs, after multiple iterative trainings, the parameters of the model are adjusted so that the model can learn the relationship between the occurrence of fires and the operating state of the line.

[0098] Input the spatio-temporal correlation feature matrix into the pre-trained fire risk prediction model. The model will predict the abnormal risk probability of the line according to its internal parameters and the learned relationships. The prediction result is presented in the form of a distribution map of the abnormal risk probability of the line, which can intuitively show the abnormal risk probability of the target charging pile cluster line at different positions and times. For example, in the distribution map, the darker the color area, the higher the abnormal risk probability, and the lighter the color area, the lower the abnormal risk probability.

[0099] As an implementation, the training process of the fire risk prediction model can specifically include the following steps:

[0100] Step S10: Obtain a historical fire event dataset, which contains multi-source monitoring data slices within a set time range before the fire and the corresponding fire location annotation information.

[0101] The historical fire event dataset is an important data source for training the fire risk prediction model. In the embodiment of the present invention, a historical fire event dataset is obtained, which contains multi-source monitoring data slices within a set time range before the fire and the corresponding fire location annotation information. The set time range can be determined according to the actual situation, such as 1 hour, 2 hours, etc. before the fire.

[0102] The multi-source monitoring data slices refer to various monitoring data collected within the set time range, such as current fluctuation characteristics, temperature change characteristics, line aging indicators, etc. These data slices record the operating state of the line before the fire. The corresponding fire location annotation information specifies the specific location where the fire occurred, such as the specific number of a charging pile or the specific location of a line segment. For example, in a historical fire event dataset, it contains multi-source monitoring data slices within 2 hours before the fire, and the annotation information that the fire occurred in the line segment of the 5th charging pile in a certain charging station. By collecting a large number of such historical fire event data, a complete historical fire event dataset is formed.

[0103] Step S20: Perform cross-modal feature alignment processing on the multi-source monitoring data slices to generate a set of historical spatio-temporal correlation feature matrices.

[0104] Performing cross-modal feature alignment processing on the multi-source monitoring data slices is similar to the cross-modal feature alignment processing performed on the multi-source time-series monitoring data in the previous step S200. Through this processing, different types of monitoring data (such as current fluctuation characteristics, temperature change characteristics, line aging indicators, etc.) are aligned and correlated in the spatio-temporal dimension.

[0105] The specific processing process includes dynamically adjusting the weight distribution of temperature change characteristics based on the amplitude of current fluctuations, and correcting the correlation between current and temperature characteristics based on the line aging index, etc. Through these processes, a set of historical spatio-temporal correlation feature matrices is generated. This matrix set contains the line operation status information at different times and spaces before historical fire events, providing more accurate and comprehensive data for subsequent model training.

[0106] Step S30: Construct a deep spatio-temporal convolutional network, which includes a parallel local feature extraction branch and a global correlation branch. The local feature extraction branch uses a three-dimensional dilated convolutional kernel to capture local abnormal patterns of the line, and the global correlation branch uses a graph attention mechanism to model the topological relationship across line segments.

[0107] The deep spatio-temporal convolutional network is a deep learning network structure for processing spatio-temporal data. In the embodiment of the present invention, a deep spatio-temporal convolutional network is constructed, which includes a parallel local feature extraction branch and a global correlation branch.

[0108] The local feature extraction branch uses a three-dimensional dilated convolutional kernel to capture local abnormal patterns of the line. The three-dimensional dilated convolutional kernel can perform convolutional operations on data in three-dimensional space (two dimensions of time and space), and by setting the dilation rate, the receptive field can be expanded without increasing the parameters of the convolutional kernel, so as to better capture local abnormal patterns of the line. For example, when there is a sudden increase in current or an abnormal rise in temperature in a local area of the line, the local feature extraction branch can extract these abnormal patterns through the three-dimensional dilated convolutional kernel.

[0109] The global correlation branch uses a graph attention mechanism to model the topological relationship across line segments. The graph attention mechanism can dynamically assign attention weights according to the relationship between nodes, so as to better model the topological relationship across line segments. In this scenario, the charging pile line segments are regarded as the nodes of the graph, and the electrical connection relationship between adjacent line segments is regarded as the edges of the graph. Through the graph attention mechanism, the mutual influence and correlation between different line segments can be learned. For example, when an abnormality occurs in a certain line segment, the degree of influence of this abnormality on adjacent line segments can be analyzed through the graph attention mechanism.

[0110] Step S40: Input the set of historical spatio-temporal correlation feature matrices into the deep spatio-temporal convolutional network for forward propagation, output the predicted fire probability distribution map, and calculate the cross-entropy loss with the real fire location annotation information.

[0111] Input the historical spatio-temporal correlation feature matrix set into the deep spatio-temporal convolutional network for forward propagation. Forward propagation refers to the process where data passes through each hidden layer from the input layer of the network and finally reaches the output layer. During this process, the local feature extraction branch and the global correlation branch of the deep spatio-temporal convolutional network will process the input data respectively to extract local features and model the global topological relationship.

[0112] Output the predicted fire probability distribution map, which reflects the prediction results of the model on the occurrence probability of historical fire events at different times and spaces. Then, compare the predicted fire probability distribution map with the true fire location annotation information and calculate the cross-entropy loss. Cross-entropy loss is a loss function used to measure the difference between the prediction result and the true label. By calculating the cross-entropy loss, the prediction accuracy of the model can be evaluated. For example, if the predicted probability at a certain position in the predicted fire probability distribution map is inconsistent with the true fire location annotation information, the cross-entropy loss will increase.

[0113] Step S50: Iteratively optimize the parameters of the deep spatio-temporal convolutional network through the backpropagation algorithm until the spatial coincidence degree between the predicted fire probability distribution map and the true annotation reaches a preset threshold.

[0114] In the embodiment of the present invention, the parameters of the deep spatio-temporal convolutional network are iteratively optimized through the backpropagation algorithm. In each iteration, according to the calculated cross-entropy loss, calculate the gradient of the loss function with respect to the network parameters, and then use an optimization algorithm (such as the stochastic gradient descent method) to update the network parameters. The iteration process will be repeated continuously until the spatial coincidence degree between the predicted fire probability distribution map and the true annotation reaches a preset threshold. The spatial coincidence degree can be measured by calculating the overlapping degree between the predicted fire probability distribution map and the true fire location annotation information in space. The preset threshold can be determined according to the actual situation. For example, when the spatial coincidence degree reaches 80%, it is considered that the prediction result of the model is accurate enough and the iterative optimization is stopped.

[0115] As an implementation method, the global correlation branch of the deep spatio-temporal convolutional network performs the following processing:

[0116] Step S31: Construct a line topology graph according to the physical connection relationship of the charging pile lines. The nodes of the line topology graph represent individual charging pile line segments, and the edges represent the electrical connection relationships between adjacent line segments.

[0117] The line topology diagram is a graph structure used to represent the connection relationship between the charging pile lines. In the embodiments of the present invention, the line topology diagram is constructed according to the physical connection relationship of the charging pile lines. Each single charging pile line segment is regarded as a node of the graph, and the electrical connection relationship between adjacent line segments is regarded as an edge of the graph. For example, in a charging station containing multiple charging piles, the line segments at the incoming end and the outgoing end of each charging pile can be regarded as a node. If there is an electrical connection between two line segments, then an edge is added between their corresponding nodes.

[0118] By constructing the line topology diagram, the topological structure of the charging pile lines can be represented more intuitively, providing a basis for subsequent processing by the graph attention mechanism.

[0119] Step S32: Assign an initial feature vector to each node. The initial feature vector includes the statistical quantities of the current fluctuation feature, the temperature change feature, and the aging index corresponding to the charging pile line segment.

[0120] An initial feature vector is assigned to each node of the line topology diagram. The initial feature vector includes the statistical quantities of the current fluctuation feature, the temperature change feature, and the aging index corresponding to the charging pile line segment. For example, the statistical quantities of the current fluctuation feature can include the average value, the standard deviation, etc. of the current; the statistical quantities of the temperature change feature can include the maximum value, the minimum value, etc. of the temperature; the statistical quantities of the aging index can include the average value of the conductive material degradation coefficient, the average value of the insulation layer breakage rate, etc.

[0121] By assigning an initial feature vector to each node, the operating state information of the line can be incorporated into the graph structure, enabling the graph attention mechanism to better process this information.

[0122] Step S33: Iteratively update the node features through a multi-layer graph attention network, where each layer of the graph attention network performs the following operations:

[0123] Step S331: Calculate the attention coefficient between the current node and its neighbor nodes. The attention coefficient is dynamically adjusted based on the current phase difference between nodes and the temperature conduction rate.

[0124] In the multi-layer graph attention network, each layer of the graph attention network first calculates the attention coefficient between the current node and its neighbor nodes. The attention coefficient is used to measure the degree of attention of the current node to the features of the neighbor nodes. In the embodiments of the present invention, the attention coefficient is dynamically adjusted based on the current phase difference between nodes and the temperature conduction rate.

[0125] The current phase difference between nodes reflects the phase difference of the current between adjacent line segments, and the temperature conduction rate reflects the heat transfer speed between adjacent line segments. For example, if the current phase difference between two adjacent nodes is large, it indicates that there may be significant differences in their current operating states, and at this time, the attention coefficient may be relatively small; if the temperature conduction rate between two adjacent nodes is fast, it indicates that the heat transfer relationship between them is relatively close, and at this time, the attention coefficient may be relatively large.

[0126] By dynamically adjusting the attention coefficient based on the current phase difference between nodes and the temperature conduction rate, the mutual influence relationship between adjacent line segments can be more accurately reflected.

[0127] Step S332: Aggregate the neighbor node features weighted according to the attention coefficient, and perform a residual connection with the current node features.

[0128] According to the calculated attention coefficient, aggregate the features of the neighbor nodes weighted. Then, perform a residual connection between the weighted aggregated feature vector and the feature vector of the current node. Residual connection is a commonly used technique in deep learning, which can alleviate the problem of gradient disappearance and make the network easier to train. Specifically, add the weighted aggregated feature vector to the feature vector of the current node to obtain a new node feature vector.

[0129] Step S334: Input the updated node features into a gated linear unit for non-linear transformation.

[0130] In the embodiment of the present invention, the updated node features are input into a gated linear unit for non-linear transformation. GLU can control the transmission of input information through a gating mechanism, thereby realizing non-linear processing of node features. For example, GLU can dynamically adjust the output feature values according to the input node features, making the feature information richer and more complex. Through the non-linear transformation of the gated linear unit, the expression ability of node features can be enhanced, and the performance of the graph attention network can be improved.

[0131] Step S34: Concatenate the node features output by the last layer of the graph attention network with the output features of the local feature extraction branch to generate a fused global-local joint feature.

[0132] Concatenate the node features output by the last layer of the graph attention network with the output features of the local feature extraction branch. Channel concatenation is to merge different feature vectors in the channel dimension. Through channel concatenation, the global topological relationship information extracted by the global association branch and the local abnormal pattern information extracted by the local feature extraction branch are integrated to generate a fused global-local joint feature. This feature can more comprehensively reflect the operating state of the charging pile line and provide a more accurate basis for subsequent fire risk prediction.

[0133] Step S400: Generate a hierarchical warning signal set according to the risk levels corresponding to each spatial node in the line anomaly risk probability distribution map. The hierarchical warning signal set includes differential control instructions for different segments of the charging pile lines.

[0134] The line anomaly risk probability distribution map shows the anomaly risk probability conditions of the lines of the target charging pile cluster at different locations and times. In the embodiments of the present invention, a hierarchical warning signal set is generated according to the risk levels corresponding to each spatial node in the line anomaly risk probability distribution map. The risk levels can be divided according to the magnitude of the anomaly risk probability. For example, the risk levels are divided into three levels: high, medium, and low. The hierarchical warning signal set includes differential control instructions for different segments of the charging pile lines. Different risk levels correspond to different control instructions. For example, for a line segment with a high risk level, more stringent control measures may be required, such as cutting off the power supply; for a line segment with a medium risk level, measures such as adjusting the load may be required; for a line segment with a low risk level, only monitoring and warning prompts may be needed.

[0135] By generating a hierarchical warning signal set, different control measures can be taken according to the actual risk situation of the line, improving the pertinence and effectiveness of fire warning.

[0136] As an implementation manner, in step S400, to generate a hierarchical warning signal set according to the risk levels corresponding to each spatial node in the line anomaly risk probability distribution map, the following steps may specifically include:

[0137] Step S410: Divide the line anomaly risk probability distribution map into multiple risk level regions, and each risk level region corresponds to a preset warning response strategy.

[0138] The line anomaly risk probability distribution map is divided into multiple risk level regions. For example, the distribution map can be divided into a high-risk region, a medium-risk region, and a low-risk region according to the magnitude of the anomaly risk probability. Each risk level region corresponds to a preset warning response strategy, and the warning response strategy stipulates the specific measures to be taken at this risk level. For example, for the high-risk region, the preset warning response strategy may be to immediately cut off the power supply of this region and notify relevant personnel for inspection and repair; for the medium-risk region, the warning response strategy may be to adjust the load of this region to reduce the operating pressure of the line; for the low-risk region, the warning response strategy may be to strengthen the monitoring of this region and issue a warning prompt.

[0139] Step S420: For the first risk level region, extract the real-time current spectrum characteristics of the line segments therein, and detect whether there is a harmonic resonance phenomenon; if a harmonic resonance is detected, generate a first-level warning signal including a harmonic suppression instruction.

[0140] For the first risk level area (such as a high-risk area), extract the real-time current spectrum characteristics of the line segments within it. The real-time current spectrum characteristics reflect the frequency component distribution of the current in the line. The real-time current spectrum characteristics can be obtained by methods such as Fourier transform of the current signal in the line.

[0141] Detect whether there is a harmonic resonance phenomenon. Harmonic resonance occurs when the harmonic frequency in the line matches the natural resonance frequency of the line, resulting in a sharp increase in harmonic energy, which may damage the line. Whether there is a harmonic resonance phenomenon can be detected by calculating the matching degree between the harmonic frequencies in the real-time current spectrum characteristics and the natural resonance frequency spectrum of the line.

[0142] If harmonic resonance is detected, generate a first-level warning signal containing a harmonic suppression instruction. The harmonic suppression instruction is used to suppress the harmonic components in the line. For example, it can be achieved by starting a filter, etc. The first-level warning signal will promptly notify relevant personnel to take measures to avoid damage to the line caused by harmonic resonance.

[0143] Step S430: For the second risk level area, calculate the deviation degree between the temperature change rate and the current load of the line segments in this area. When the deviation degree exceeds the dynamic threshold, generate a second-level warning signal containing a load balancing instruction.

[0144] For the second risk level area, calculate the deviation degree between the temperature change rate and the current load of the line segments in this area. The temperature change rate reflects the speed of change of the line temperature over time, and the current load reflects the magnitude of the current in the line. The deviation degree is used to measure the difference between the temperature change rate and the current load. For example, the deviation degree can be obtained by calculating the ratio of the temperature change rate to the current load and comparing it with a standard ratio.

[0145] The dynamic threshold is a threshold dynamically adjusted according to the actual operating conditions of the line. When the deviation degree exceeds the dynamic threshold, it indicates that the operating state of the line may be abnormal. At this time, generate a second-level warning signal containing a load balancing instruction. The load balancing instruction is used to adjust the load of the line to make the load of each line segment more balanced, thereby reducing the operating risk of the line. For example, the load balancing can be achieved by adjusting the charging power of the charging pile, etc.

[0146] Step S440: For the third risk level area, monitor the cumulative change amount of the line aging index. When the cumulative change amount reaches the preset critical value, generate a third-level warning signal containing a preventive maintenance reminder.

[0147] For the third risk level area (e.g., low-risk area), monitor the cumulative change amount of the line aging index. The cumulative change amount of the line aging index reflects the gradual increase in the degree of line aging. For example, by continuously accumulating the change amounts of the conductive material degradation coefficient and the insulation layer breakage rate, the cumulative change amount of the line aging index is obtained. The preset critical value is a threshold determined based on factors such as the design life and safety standards of the line. When the cumulative change amount reaches the preset critical value, it indicates that the line may have approached the limit of aging and preventive maintenance is required. At this time, a third-level warning signal containing preventive maintenance prompts is generated to prompt relevant personnel to inspect and maintain the line to ensure the safe operation of the line.

[0148] Step S450: Topologically sort the first-level warning signal, the second-level warning signal, and the third-level warning signal according to the spatial position to generate a hierarchical warning signal set that matches the physical layout of the charging pile line.

[0149] Topologically sort the first-level warning signal, the second-level warning signal, and the third-level warning signal according to the spatial position. Topological sorting refers to sorting the warning signals according to the physical connection relationship and spatial position of the line. For example, sort according to the line segment number or geographical location so that the order of the warning signals matches the physical layout of the charging pile line.

[0150] Through topological sorting, generate a hierarchical warning signal set that matches the physical layout of the charging pile line. This signal set can more conveniently correspond to the actual charging pile line, enabling relevant personnel to quickly and accurately take corresponding control measures according to the prompts of the signal set.

[0151] As an implementation manner, in step S420, detecting whether there is a harmonic resonance phenomenon may specifically include:

[0152] Step S421: Obtain the current waveform sampling data of the target line segment, and extract the frequency-amplitude pair set of the fundamental wave component and each harmonic component through time-frequency conversion processing.

[0153] Obtain the current waveform sampling data of the target line segment. A current sensor can be used to sample the current waveform of the target line segment to obtain a series of discrete current values. Time-frequency conversion processing is a method of converting the current waveform data in the time domain to the frequency domain for analysis. Commonly used time-frequency conversion methods include the fast Fourier transform (FFT).

[0154] Through time-frequency conversion processing, convert the current waveform sampling data of the target line segment to the frequency domain, and extract the frequency-amplitude pair set of the fundamental wave component and each harmonic component. The fundamental wave component refers to the component with the lowest frequency in the current waveform, and each harmonic component refers to the component with a frequency that is an integer multiple of the fundamental wave frequency. The frequency-amplitude pair set records the frequency and corresponding amplitude information of each harmonic component.

[0155] Step S422: Calculate the matching degree between the harmonic frequencies in the frequency-amplitude pair set and the line's inherent resonance frequency spectrum, and generate the frequency offset and resonance tendency index corresponding to each harmonic component.

[0156] Calculate the matching degree between the harmonic frequencies in the frequency-amplitude pair set and the line's inherent resonance frequency spectrum. The line's inherent resonance frequency spectrum refers to the set of resonance frequencies that the line itself has, and these frequencies are determined by the physical parameters of the line (such as inductance, capacitance, etc.). The matching degree calculation can be carried out by calculating the difference between the harmonic frequency and the line's inherent resonance frequency.

[0157] Based on the matching degree calculation result, generate the frequency offset and resonance tendency index corresponding to each harmonic component. The frequency offset refers to the difference between the harmonic frequency and the line's inherent resonance frequency, which reflects the proximity of the harmonic frequency to the resonance frequency. The resonance tendency index is an index comprehensively calculated based on factors such as the frequency offset and the harmonic amplitude, and is used to measure the likelihood of a harmonic component resonating. For example, if the frequency offset is small and the harmonic amplitude is large, then the resonance tendency index will be high.

[0158] Step S423: Screen out potential dangerous harmonic components based on the resonance tendency index, and input the amplitudes of the potential dangerous harmonic components into the energy accumulation predictor according to the time series to generate the harmonic energy accumulation trajectory.

[0159] Screen out potential dangerous harmonic components based on the resonance tendency index. A resonance tendency index threshold can be set, and when the resonance tendency index of a certain harmonic component exceeds this threshold, it is regarded as a potential dangerous harmonic component. Input the amplitudes of the potential dangerous harmonic components into the energy accumulation predictor according to the time series. The energy accumulation predictor is a model used to predict the harmonic energy accumulation situation, and can predict the cumulative change of harmonic energy over time based on the input harmonic amplitude time series.

[0160] Through the energy accumulation predictor, generate the harmonic energy accumulation trajectory. This trajectory reflects the accumulation of the energy of potential dangerous harmonic components over a period of time. For example, if the harmonic energy accumulation trajectory shows a rapidly rising trend, it indicates that the harmonic energy is continuously accumulating and may trigger harmonic resonance phenomena.

[0161] Step S424: Conduct a trend analysis on the harmonic energy accumulation trajectory. When it is detected that the continuous rising slope exceeds the safety threshold, trigger the multi-band interference detection process.

[0162] Perform trend analysis on the harmonic energy accumulation trajectory. Trend analysis can be carried out by calculating methods such as the slope of the trajectory to determine whether the trend of harmonic energy accumulation is rising, falling, or remaining stable. The safety threshold is a threshold determined based on factors such as the safety standards and design requirements of the line, and is used to judge whether the rising speed of harmonic energy accumulation is too fast.

[0163] When it is detected that the continuous rising slope exceeds the safety threshold, it indicates that the speed of harmonic energy accumulation is too fast, which may trigger harmonic resonance. At this time, the multi-band interference detection process is triggered. The multi-band interference detection process is used to further analyze the interaction between harmonics in different frequency bands to determine whether there is a risk of harmonic resonance.

[0164] Step S425: Extract the harmonic phase difference sequence of adjacent frequency bands, and calculate the phase synchronization index and the energy superposition effect coefficient.

[0165] In the multi-band interference detection process, extract the harmonic phase difference sequence of adjacent frequency bands. The harmonic phase difference sequence records the phase difference between harmonics in adjacent frequency bands. By performing phase analysis on the harmonic signals of adjacent frequency bands, the harmonic phase difference sequence is obtained.

[0166] Calculate the phase synchronization index and the energy superposition effect coefficient. The phase synchronization index is used to measure the degree of phase synchronization between harmonics in adjacent frequency bands. If the phase synchronization is high, it indicates that the harmonics in adjacent frequency bands may be superimposed on each other, increasing the harmonic energy. The energy superposition effect coefficient is an index comprehensively calculated based on factors such as the phase synchronization index and harmonic amplitude, and is used to measure the degree of harmonic energy superposition between adjacent frequency bands. For example, if the phase synchronization is high and the harmonic amplitude is large, then the energy superposition effect coefficient will be high.

[0167] Step S426: Input the phase synchronization index and the energy superposition effect coefficient into the resonance risk classifier to generate a comprehensive resonance risk level.

[0168] Input the phase synchronization index and the energy superposition effect coefficient into the resonance risk classifier. The resonance risk classifier is a trained classification model that can classify the risk of harmonic resonance according to the input phase synchronization index and energy superposition effect coefficient. For example, the resonance risk classifier can divide the risk level into three levels: high, medium, and low.

[0169] Through the resonance risk classifier, generate a comprehensive resonance risk level. This level comprehensively considers the phase synchronization and energy superposition effects between harmonics in adjacent frequency bands, and can more accurately evaluate the degree of risk of harmonic resonance.

[0170] Step S427: Adjust the response intensity parameters of the harmonic suppression instruction according to the comprehensive resonance risk level. The response intensity parameters include the filter startup priority and the attenuation depth gradient.

[0171] Adjust the response intensity parameters of the harmonic suppression instruction according to the comprehensive resonance risk level. The response intensity parameters include the filter startup priority and the attenuation depth gradient. The filter startup priority is used to determine the startup order of the filters. When the comprehensive resonance risk level is high, the startup priority of the filters can be increased to suppress harmonics more quickly. The attenuation depth gradient is used to control the attenuation degree of the filters for harmonics. When the comprehensive resonance risk level is high, the attenuation depth gradient can be increased to suppress harmonics more effectively.

[0172] For example, if the comprehensive resonance risk level is high, the filter startup priority can be set to the highest level and the attenuation depth gradient can be set to a relatively large value; if the comprehensive resonance risk level is medium, the filter startup priority can be set to medium and the attenuation depth gradient can be adjusted appropriately; when the comprehensive resonance risk level is low, the filter startup priority is reduced and the attenuation depth gradient is also decreased. Exemplarily, for a high risk level, the filter startup priority can be set to immediately start all available filters, and the attenuation depth gradient can be set to attenuate 3 dB (voltage) per 100 Hz; for a medium risk level, the priority can be set to preferentially start some key filters, and the attenuation depth gradient is 2 dB (voltage) per 100 Hz; at a low risk level, the filter is started only when necessary, and the attenuation depth gradient is 1 dB (voltage) per 100 Hz.

[0173] Step S428: Bind the adjusted response intensity parameters with the harmonic suppression instruction in the first-level warning signal to generate an optimized suppression instruction set that dynamically adapts to the electrical characteristics of the target line segment.

[0174] Bind the adjusted response intensity parameters with the harmonic suppression instruction in the first-level warning signal. In actual operation, the response intensity parameters can be associated with the harmonic suppression instruction in the form of a data structure, such as using key-value pairs, taking the filter startup priority and the attenuation depth gradient as keys, and the corresponding specific parameter values as values, and integrating them with the harmonic suppression instruction.

[0175] Generate an optimized suppression instruction set that dynamically adapts to the electrical characteristics of the target line segment. The electrical characteristics of the target line segment include parameters such as the resistance, inductance, and capacitance of the line, which can affect the propagation and characteristics of harmonics. By combining the electrical characteristics of the target line segment, the bound instructions are further optimized. For example, for a line segment with a relatively large resistance, it may be necessary to emphasize more the suppression of high-frequency harmonics, and the start priority and attenuation depth gradient of the filter for high-frequency harmonics can be appropriately increased; for a line segment with a relatively large inductance, low-frequency harmonics are more likely to resonate, and the suppression instructions for low-frequency harmonics can be optimized. The optimized suppression instruction set generated in this way can more accurately target the actual situation of the target line segment for harmonic suppression and improve the suppression effect.

[0176] Step S500: Trigger a dynamic protection mechanism based on the hierarchical warning signal set. The dynamic protection mechanism includes performing a current truncation operation on the high-risk line segment and a power attenuation operation on the adjacent line segment.

[0177] The hierarchical warning signal set contains differential control instructions for line segments with different risk levels, and based on this, a dynamic protection mechanism is triggered. The core goal of the dynamic protection mechanism is to take timely measures to avoid accidents such as fires when abnormal risks are found in the line. By performing a current truncation operation on the high-risk line segment and a power attenuation operation on the adjacent line segment, the risk of the entire charging pile cluster line is reduced.

[0178] When triggering the dynamic protection mechanism, the system will first analyze the hierarchical warning signal set to determine the line segments corresponding to each warning signal and the specific control instructions. For the high-risk line segment, immediately perform a current truncation operation, which can be achieved by controlling devices such as circuit breakers to quickly cut off the current supply of this line segment and prevent the further deterioration of possible overheating, short-circuit, etc. At the same time, in order to avoid excessive impact on the adjacent line segment caused by the current truncation, a power attenuation operation is performed on the adjacent line segment. By adjusting the load of the adjacent line segment, its power consumption is reduced to maintain the stability of the entire line system.

[0179] As an implementation method, in step S500, triggering the dynamic protection mechanism based on the hierarchical warning signal set may specifically include the following steps:

[0180] Step S510: Analyze the line segment identifiers and warning response strategy parameters corresponding to each warning signal in the hierarchical warning signal set, and extract the set of line segments covered by the first-level warning signals containing harmonic suppression instructions, as well as the electrical connection topology corresponding to the set of line segments.

[0181] Analyze the line segment identifiers and early warning response strategy parameters corresponding to each early warning signal in the hierarchical early warning signal set. Each early warning signal in the hierarchical early warning signal set is associated with a line segment and contains corresponding early warning response strategy parameters, such as the specific parameters of the harmonic suppression instruction, the adjustment range of the load balancing instruction, etc. The analysis process can be achieved by performing encoding and decoding operations on the signal set. For example, the signal is encoded using a preset protocol and decoded according to the protocol rules during analysis to extract the line segment identifier and the early warning response strategy parameters.

[0182] Extract the set of line segments covered by the first-level early warning signal containing the harmonic suppression instruction, and the electrical connection topology corresponding to the set of line segments. The set of line segments covered by the first-level early warning signal containing the harmonic suppression instruction can be found by screening the early warning signals, and the line segment identifiers covered by it are extracted to form the set of line segments. The electrical connection topology describes the connection method and mutual relationship between these line segments, which can be obtained through a pre-constructed line topology diagram. This topology diagram records the connection information of each line segment, such as the identifiers of adjacent line segments, the connection method, etc.

[0183] Step S520: Send a current truncation instruction to the target charging pile in the set of line segments. The current truncation instruction includes the truncation phase angle range and duration parameters generated based on the harmonic suppression instruction, and the current waveform distortion rate and temperature gradient change amount are collected in real time after the truncation operation is executed.

[0184] Send a current truncation instruction to the target charging pile in the set of line segments. When sending the instruction, it can be transmitted to the controller of the target charging pile through a communication network to ensure accurate transmission of the instruction. The current truncation instruction includes the truncation phase angle range and duration parameters generated based on the harmonic suppression instruction. The truncation phase angle range refers to the phase interval of the current waveform where the truncation operation is performed, which can be determined according to the requirements of harmonic suppression. For example, the truncation can be selected in the phase interval with a large harmonic amplitude. The duration parameter specifies the duration of the current truncation to avoid unnecessary impacts on the charging pile and the line due to long-term truncation.

[0185] Collect the current waveform distortion rate and temperature gradient change amount in real time after the truncation operation is executed. Current sensors and temperature sensors can be used to monitor the current waveform and temperature in real time respectively. The current waveform distortion rate reflects the deviation degree of the current waveform from the ideal sine wave, and the distortion rate is calculated by analyzing the collected current waveform. The temperature gradient change amount refers to the change rate of the temperature before and after the truncation operation, and the temperature gradient change amount is calculated by comparing the temperature values at different time points. These data can be used to evaluate the effect of the current truncation operation and the real-time state of the line.

[0186] Step S530: When the current waveform distortion rate exceeds the preset distortion threshold or the temperature gradient change amount fails to reach the expected decrease, activate the power attenuation compensation process for adjacent line segments.

[0187] When the current waveform distortion rate exceeds the preset distortion threshold or the temperature gradient change amount fails to reach the expected decrease, it indicates that the current truncation operation may not have achieved the expected effect, and there are still significant risks in the line. The preset distortion threshold is a threshold determined based on the safety standards and design requirements of the line. When the current waveform distortion rate exceeds this threshold, it indicates that the distortion degree of the current waveform is too large, which may cause damage to the line and equipment. The expected decrease refers to the reduction amplitude that the temperature gradient change amount should reach after the current truncation operation. If the temperature gradient change amount fails to reach this decrease, it means that the heating condition of the line has not been effectively controlled.

[0188] In this case, activate the power attenuation compensation process for adjacent line segments. The purpose of this process is to reduce the power consumption of adjacent line segments, relieve the load pressure on the line, and further reduce the risks of the line.

[0189] Step S540: Determine the propagation path sequence of power attenuation according to the electrical connection topology relationship, and dynamically allocate the power attenuation ratio of each path node based on the load balancing instruction in the early warning response strategy parameters.

[0190] Determine the propagation path sequence of power attenuation according to the electrical connection topology relationship. The electrical connection topology relationship describes the connection methods and mutual relationships between line segments. By analyzing the topology relationship, the propagation path of power attenuation between adjacent line segments can be determined. For example, if line segment A is adjacent to line segments B and C, and B and C are respectively connected to other line segments, then the propagation paths of power attenuation can be determined as A - B -... and A - C -... according to these connection relationships.

[0191] Dynamically allocate the power attenuation ratio of each path node based on the load balancing instruction in the early warning response strategy parameters. The load balancing instruction stipulates how to distribute the load between different line segments to achieve the purpose of load balancing. According to this instruction, the power attenuation ratio of each path node can be calculated. For example, if the load balancing instruction requires distributing the total power attenuation amount according to a preset ratio to each adjacent line segment, then the power attenuation ratio of each path node can be dynamically adjusted according to factors such as the load capacity and importance of the line segment.

[0192] Step S550: Convert the power attenuation ratio into a pulse width modulation waveform, inject it into the power regulation loop of the corresponding charging pile, and generate a stepped attenuation signal synchronized with the truncation phase angle range.

[0193] Convert the power attenuation ratio into a Pulse Width Modulation (PWM) waveform. The PWM waveform is a waveform that controls the average power by adjusting the width of the pulse. The duty cycle of the PWM waveform, which is the ratio of the pulse width to the period, can be calculated based on the power attenuation ratio. For example, if the power attenuation ratio is 50%, the duty cycle of the PWM waveform can be set to 50%. Inject the generated PWM waveform into the power regulation circuit of the corresponding charging pile. The power regulation circuit is the circuit in the charging pile used to control the power output. By injecting the PWM waveform, the power output of the charging pile can be adjusted. Generate a stepped attenuation signal synchronized with the truncated phase angle range, that is, the start and end phases of the stepped attenuation signal are consistent with the truncated phase angle range in the current truncation instruction. This can ensure that the power attenuation operation is synchronized with the current truncation operation, improving the effect of the dynamic protection mechanism.

[0194] Step S560: Collect the line status feedback data under the action of the stepped attenuation signal, which may specifically include the following steps: the distribution of current harmonic components after attenuation and the temperature field uniformity index, and input them into the fire risk prediction model for real-time risk re-evaluation.

[0195] Collect the line status feedback data under the action of the stepped attenuation signal. Through various sensors installed on the line, such as current sensors, temperature sensors, etc., the distribution of current harmonic components after attenuation and the temperature field uniformity index are collected in real time. The distribution of current harmonic components reflects the content and distribution of different frequency harmonics in the line after the action of the stepped attenuation signal, which can be obtained by performing spectral analysis on the collected current signal. The temperature field uniformity index describes the degree of spatial uniformity of the line temperature. By installing temperature sensors at different positions, measuring the temperature values at each point, and then calculating statistical quantities such as the standard deviation of the temperature, it can be evaluated. Input the collected line status feedback data into the fire risk prediction model for real-time risk re-evaluation. The fire risk prediction model is a previously trained model that can predict the abnormal risk probability of the line based on the input line status data. By inputting the line status feedback data in real time, the model can update the evaluation result of the line risk in a timely manner, providing a basis for subsequent decisions.

[0196] Step S570: Update the line abnormal risk probability distribution map according to the real-time risk re-evaluation result. If the diffusion trend of the harmonic energy accumulation trajectory is detected, trigger the secondary protection strategy.

[0197] Update the line anomaly risk probability distribution map according to the real-time risk re-evaluation results. The line anomaly risk probability distribution map shows the anomaly risk probability of the target charging pile cluster lines at different locations and times. Integrate the real-time risk re-evaluation results with the original distribution map, update the risk probability values of each spatial node, so that the distribution map can more accurately reflect the current risk state of the line. If the diffusion trend of the harmonic energy accumulation trajectory is detected, trigger the secondary protection strategy. The diffusion trend of the harmonic energy accumulation trajectory indicates that the harmonic energy is increasing and spreading to surrounding line segments, which may cause more serious problems. The secondary protection strategy can include further strengthening the protection measures for high-risk line segments, such as increasing the current cutoff range, performing power attenuation on more adjacent line segments, etc., to prevent the further spread of harmonic energy and the occurrence of accidents.

[0198] Step S580: Perform voltage amplitude limit operations on the key nodes in the propagation path sequence to suppress the conduction path of abnormal harmonic components.

[0199] Perform voltage amplitude limit operations on the key nodes in the propagation path sequence. Key nodes refer to the line segments or nodes that play an important role in the harmonic conduction in the power attenuation propagation path. By limiting the voltage amplitude of these key nodes, the conduction path of abnormal harmonic components can be effectively suppressed. Voltage regulators and other devices can be used to achieve the voltage amplitude limit, and control the voltage amplitude of the key nodes within a safe range. For example, set a voltage upper limit value. When the voltage of the key node exceeds this upper limit value, the voltage regulator automatically adjusts the voltage to keep it within the safe range.

[0200] Step S590: After the voltage amplitude limit operation takes effect, recalculate the power attenuation ratio and superimpose the high-frequency oscillation suppression component to eliminate the residual harmonic interference.

[0201] After the voltage amplitude limit operation takes effect, recalculate the power attenuation ratio. Since the voltage amplitude limit will affect the power transmission and load conditions of the line, it is necessary to recalculate the power attenuation ratio according to the new line state to ensure the effectiveness of the power attenuation operation. The power attenuation ratio of each path node can be recalculated based on the real-time collected line state data, such as current, voltage, etc., combined with the load balancing instruction.

[0202] Superimpose the high-frequency oscillation suppression component to eliminate the residual harmonic interference. High-frequency oscillation refers to the high-frequency harmonic components that may still exist after the voltage amplitude limit operation, and these components may cause interference to the line and equipment. The high-frequency oscillation suppression component can be superimposed in the power regulation loop, such as using filters and other devices, to further suppress the high-frequency harmonics, eliminate the residual harmonic interference, and improve the stability of the line.

[0203] Step S5100: Align the updated power attenuation parameter and the current cut-off instruction in time sequence for calibration, and generate a dynamic protection instruction sequence including the connection relationship of multi-stage protection logic.

[0204] Align the updated power attenuation parameter and the current cut-off instruction in time sequence for calibration. Time sequence alignment calibration means ensuring the coordination of the power attenuation operation and the current cut-off operation in time, avoiding operation conflicts or excessive time differences. It can be achieved by comparing and adjusting the time parameters of the updated power attenuation parameter and the current cut-off instruction to make their execution times match each other.

[0205] Generate a dynamic protection instruction sequence including the connection relationship of multi-stage protection logic. The dynamic protection mechanism includes multiple stages of protection measures, such as current cut-off, power attenuation, voltage amplitude limitation, etc. There needs to be a reasonable connection relationship between these measures to ensure the effectiveness and stability of the entire protection process. Integrate the updated power attenuation parameter and the current cut-off instruction, etc., to generate a dynamic protection instruction sequence including the connection relationship of multi-stage protection logic, clarify the operation sequence, time interval, and specific parameters of each stage, and provide clear guidance for subsequent control operations.

[0206] Step S5110: Control the target charging pile cluster to perform collaborative protection operations through the dynamic protection instruction sequence, and continuously monitor the dynamic response curves of the current waveform distortion rate and the temperature gradient change amount.

[0207] Control the target charging pile cluster to perform collaborative protection operations through the dynamic protection instruction sequence. Send the generated dynamic protection instruction sequence to each controller of the target charging pile cluster. The controller executes operations such as current cut-off, power attenuation, and voltage amplitude limitation in sequence according to the requirements of the instruction sequence to achieve the collaborative protection of the target charging pile cluster.

[0208] Continuously monitor the dynamic response curves of the current waveform distortion rate and the temperature gradient change amount. By collecting current waveform and temperature data in real time, calculate the current waveform distortion rate and the temperature gradient change amount, and plot their changes over time as dynamic response curves. Through the analysis of the dynamic response curves, the effect of the collaborative protection operation can be evaluated, and it can be judged whether the risk of the line is effectively controlled. If the dynamic response curves show that the current waveform distortion rate and the temperature gradient change amount gradually decrease and tend to be stable, it indicates that the collaborative protection operation has achieved good results; if the curves show abnormal fluctuations or continuous increases, the protection strategy needs to be adjusted in time to further reduce the risk of the line.

[0209] It should be noted that those skilled in the art can implement the unrefined technical details without obstacles based on their own technical knowledge in the field during the process of reading the above-mentioned embodiments of the invention. For example, when it comes to the calculation of variables with different dimensions, those skilled in the art can use common normalization or standardization means to uniformly eliminate the dimension differences and then perform subsequent operations. Another example is that for scenarios not involved, the technical means already disclosed in the present invention can be used for continuous adaptive extension. For example, step S520 requires "truncating the phase angle range" control. If in some scenarios, the charging pile lacks fast phase control hardware, resulting in the inability to execute the instruction. The "truncated phase angle" can be changed to time window truncation (such as cutting off the current within 10 ms) to adapt to the response ability of conventional circuit breakers. For calculations with inconsistent dimensions, such as different numbers of topological nodes (such as 10) and sensors (such as 20), the sensor data can be aggregated to the topological node dimension through principal component analysis (PCA) to ensure spatial consistency, or a graph convolutional network (GCN) can be used to directly process the topological graph structure to avoid explicit spatial interpolation. For the specific numerical values of the parameter adjustment examples, more reasonable numerical selections can also be made according to the actual situation. The present invention is only an example and is not limited thereto.

[0210] Based on the foregoing embodiments, an embodiment of the present application provides a big data-based charging pile line fire warning device. Each unit included in the device and each module included in each unit can be implemented by a processor in a computer device; of course, it can also be implemented by specific logic circuits; during implementation, the processor can be a central processing unit (CPU), a microprocessor unit (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0211] Figure 2 It is a schematic diagram of the composition structure of a big data-based charging pile line fire warning device provided by an embodiment of the present application, as Figure 2 shown. The big data-based charging pile line fire warning device 200 includes:

[0212] A data acquisition module 210, configured to acquire a line operation data set of a target charging pile cluster, where the line operation data set includes multi-source time series monitoring data;

[0213] A feature alignment module 220 for performing cross-modal feature alignment on multi-source time-series monitoring data to generate a spatio-temporal correlation feature matrix, where the cross-modal feature alignment includes dynamically adjusting the weight assignment of temperature change features based on the current fluctuation amplitude and correcting the correlation relationship between current and temperature features based on the line aging index;

[0214] A risk prediction module 230 for inputting the spatio-temporal correlation feature matrix into a pre-trained fire risk prediction model to generate a line anomaly risk probability distribution map, and the fire risk prediction model is trained through the mapping relationship between the historical fire event dataset and the multi-source monitoring data;

[0215] An early warning signal generation module 240 for generating a hierarchical early warning signal set according to the risk levels corresponding to each spatial node in the line anomaly risk probability distribution map, and the hierarchical early warning signal set includes differential control instructions for different charging pile line segments;

[0216] A protection trigger module 250 for triggering a dynamic protection mechanism based on the hierarchical early warning signal set, and the dynamic protection mechanism includes performing current truncation operations on high-risk line segments and performing power attenuation operations on adjacent line segments.

[0217] The description of the above device embodiments is similar to the description of the above method embodiments and has similar beneficial effects to the method embodiments. In some embodiments, the functions or modules included in the device provided in the embodiments of the present application can be used to execute the methods described in the above method embodiments. For the technical details not disclosed in the device embodiments of the present application, please refer to the description of the method embodiments of the present application for understanding. It should be noted that in the embodiments of the present application, if the above-mentioned big data-based charging pile line fire warning method is implemented in the form of software function modules and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present application, in essence, or the part that contributes to the related technology, can be embodied in the form of a software product. The software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the various embodiments of the present application. And the aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), magnetic disks, or optical discs, etc., which can store program codes. In this way, the embodiments of the present application are not limited to any specific hardware, software, or firmware, or any combination of hardware, software, and firmware.

[0218] An embodiment of the present application provides a computer system, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, some or all of the steps in the above method are implemented.

[0219] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, some or all of the steps in the above method are implemented. The computer-readable storage medium can be transient or non-transient.

[0220] An embodiment of the present application provides a computer program, including computer-readable code. When the computer-readable code runs in a computer device, the processor in the computer device executes to implement some or all of the steps in the above method.

[0221] An embodiment of the present application provides a computer program product. The computer program product includes a non-transient computer-readable storage medium storing a computer program. When the computer program is read and executed by a computer, some or all of the steps in the above method are implemented. The computer program product can be specifically implemented in a manner of hardware, software, or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium. In other embodiments, the computer program product is specifically embodied as a software product, such as a Software Development Kit (SDK), etc.

[0222] It should be noted here that: the descriptions of the above embodiments tend to emphasize the differences between the embodiments, and their similarities or similarities can be referred to each other. The descriptions of the above device, storage medium, computer program, and computer program product embodiments are similar to the descriptions of the above method embodiments and have beneficial effects similar to those of the method embodiments. For the technical details not disclosed in the embodiments of the device, storage medium, computer program, and computer program product of the present application, please refer to the descriptions of the method embodiments of the present application for understanding.

[0223] If the above integrated unit of the present application is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the related technology, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present application. The foregoing storage medium includes: various media such as a removable storage device, a ROM, a magnetic disk, or an optical disc that can store program codes.

[0224] As described above, it is only the implementation mode of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application.

Claims

1. A method for fire warning of charging pile lines based on big data, characterized in that, The method includes: Collecting a line operation data set of a target charging pile cluster, where the line operation data set includes multi-source time-series monitoring data; Performing cross-modal feature alignment on the multi-source time-series monitoring data to generate a spatio-temporal correlation feature matrix, where the cross-modal feature alignment includes dynamically adjusting the weight allocation of temperature change features based on the current fluctuation amplitude and correcting the correlation relationship between current and temperature features based on the line aging index; Inputting the spatio-temporal correlation feature matrix into a pre-trained fire risk prediction model to generate a line anomaly risk probability distribution map, where the fire risk prediction model is trained through the mapping relationship between the historical fire event data set and the multi-source monitoring data; Generating a hierarchical warning signal set according to the risk levels corresponding to each spatial node in the line anomaly risk probability distribution map, where the hierarchical warning signal set contains differential control instructions for different charging pile line segments; Triggering a dynamic protection mechanism based on the hierarchical warning signal set, where the dynamic protection mechanism includes performing a current truncation operation on high-risk line segments and performing a power attenuation operation on adjacent line segments; Wherein, the multi-source time-series monitoring data includes a current fluctuation feature sequence, a temperature change feature sequence, and a line aging index sequence, and performing cross-modal feature alignment on the multi-source time-series monitoring data to generate a spatio-temporal correlation feature matrix includes: Constructing a time-dimensional sliding window according to the amplitude change rate of the current fluctuation feature sequence, and extracting the peak-valley difference and duration interval of the current fluctuation within the sliding window as the first dynamic feature vector; Extracting the temperature gradient change amount and extreme value distribution density of the temperature change feature sequence within the sliding window as the second dynamic feature vector; Decomposing the line aging index sequence into two independent components, namely a conductive material degradation coefficient and an insulation layer breakage rate, and calculating the cross-influence factors between the independent components and the first dynamic feature vector and the second dynamic feature vector respectively; Performing dynamic weight allocation on the cross-influence factors through a multi-channel attention mechanism to generate an association mapping relationship between the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence; Performing spatial interpolation fusion on the first dynamic feature vector and the second dynamic feature vector based on the association mapping relationship to generate the spatio-temporal correlation feature matrix including timestamp marks.

2. The method according to claim 1, characterized in that The decomposing the line aging index sequence into two independent components, namely a conductive material degradation coefficient and an insulation layer breakage rate, and calculating the cross-influence factors between the independent components and the first dynamic feature vector and the second dynamic feature vector respectively, includes: Performing frequency domain decomposition on the line aging index sequence, extracting the low-frequency fluctuation component as the conductive material degradation coefficient, and separating the high-frequency mutation component as the insulation layer breakage rate; Performing time alignment between the conductive material degradation coefficient and the peak-valley difference of the current fluctuation in the first dynamic feature vector, and calculating the phase correlation between the two within the sliding window to generate a current-degradation interaction feature vector; Spatially match the insulation layer breakage rate with the temperature gradient change amount in the second dynamic feature vector, detect the synchronous offset amount in the extreme point distribution of the two, and generate a temperature-breakage interaction feature vector; Perform multi-scale convolution processing on the current-degradation interaction feature vector, extract the lag response mode of the conductive material degradation coefficient to the change in the current fluctuation amplitude, and generate a first cross-influence factor sequence; Perform bidirectional gated recurrent processing on the temperature-breakage interaction feature vector, capture the cumulative effect of the insulation layer breakage rate on the temperature extreme distribution, and generate a second cross-influence factor sequence; The dynamic weight assignment to the cross-influence factors through a multi-channel attention mechanism to generate the association mapping relationship between the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence includes: Perform channel splicing on the first cross-influence factor sequence and the current fluctuation feature sequence to form a current modality association feature block, and perform channel splicing on the second cross-influence factor sequence and the temperature change feature sequence to form a temperature modality association feature block; Perform the following processing on the current modality association feature block and the temperature modality association feature block through a cross-modal attention gating mechanism: In the current-dominated attention branch, calculate the weight influence value of each time step of the current modality association feature block on the temperature modality association feature block to generate an attention distribution map from current to temperature; In the temperature-dominated attention branch, calculate the weight influence value of each spatial node of the temperature modality association feature block on the current modality association feature block to generate an attention distribution map from temperature to current; Perform a dot product operation on the attention distribution map from current to temperature and the temperature modality association feature block to obtain a temperature feature enhancement vector, and perform a dot product operation on the attention distribution map from temperature to current and the current modality association feature block to obtain a current feature enhancement vector; Perform cross-projection fusion on the temperature feature enhancement vector and the current feature enhancement vector to generate the association mapping relationship between the current fluctuation feature sequence, the temperature change feature sequence, and the line aging index sequence.

3. The method according to claim 1, wherein The spatial interpolation fusion of the first dynamic feature vector and the second dynamic feature vector based on the association mapping relationship to generate the spatio-temporal association feature matrix including time stamp marks includes: Perform time stamp alignment processing on the current fluctuation peak-valley difference sequence in the first dynamic feature vector and the temperature gradient change amount sequence in the second dynamic feature vector to generate a time-synchronized current-temperature feature pair sequence; According to the attention distribution map from current to temperature in the association mapping relationship, extract the interpolation weight distribution of the current fluctuation peak-valley difference sequence in the spatial dimension to generate a current-dominated spatial interpolation mask matrix; According to the attention distribution map from temperature to current in the association mapping relationship, extract the interpolation weight distribution of the temperature gradient change amount sequence in the time dimension to generate a temperature-dominated time interpolation mask matrix; Perform a spatial - dimensional convolution operation on the current - dominant spatial interpolation mask matrix and the time - synchronized current - temperature feature - pair sequence to generate an intermediate fusion matrix with enhanced current features; Perform a time - dimensional sliding - window superposition on the temperature - dominant time interpolation mask matrix and the intermediate fusion matrix with enhanced current features to generate a spatio - temporal joint feature tensor; Perform the following interpolation operations on the spatio - temporal joint feature tensor: In the spatial dimension, according to the cross - projection fusion result in the correlation mapping relationship, perform bidirectional weighted interpolation on the feature vectors of adjacent line segments to generate a spatially continuous feature layer; In the time dimension, based on the time - stamp interval of the original multi - source time - series monitoring data, perform time - domain interpolation and completion on the spatially continuous feature layer to generate a feature distribution map with aligned timestamps; Perform channel superposition on the spatially continuous feature layer and the feature distribution map with aligned timestamps to generate the spatio - temporal correlation feature matrix containing timestamp marks, where each timestamp corresponds to the original acquisition moment of the multi - source time - series monitoring data; Perform edge - feature compensation processing on the spatio - temporal correlation feature matrix, detect the feature - attenuation regions at the matrix boundaries, and perform eigenvalue repair according to the dynamic weight - allocation ratio in the correlation mapping relationship to generate a spatio - temporal correlation feature matrix that completely covers the target charging - pile line topology; 4. The method according to claim 1, wherein The training process of the fire - risk prediction model includes: Obtain a historical fire - event dataset, which contains multi - source monitoring data slices within a set time range before a fire occurs and the corresponding fire - location annotation information; Perform cross - modal feature alignment processing on the multi - source monitoring data slices to generate a set of historical spatio - temporal correlation feature matrices; Construct a deep spatio - temporal convolutional network, which includes a parallel local - feature extraction branch and a global - correlation branch. The local - feature extraction branch uses a three - dimensional dilated convolutional kernel to capture local line - anomaly patterns, and the global - correlation branch uses a graph - attention mechanism to model the topological relationship across line segments; Input the set of historical spatio - temporal correlation feature matrices into the deep spatio - temporal convolutional network for forward propagation, output a predicted fire - probability distribution map, and calculate the cross - entropy loss with the true fire - location annotation information; Iteratively optimize the parameters of the deep spatio - temporal convolutional network through the back - propagation algorithm until the spatial coincidence degree between the predicted fire - probability distribution map and the true annotation reaches a preset threshold; 5. The method according to claim 4, wherein The global - correlation branch of the deep spatio - temporal convolutional network performs the following processing: Construct a line - topology graph according to the physical connection relationship of the charging - pile lines. The nodes of the line - topology graph represent individual charging - pile line segments, and the edges represent the electrical connection relationships between adjacent line segments; Assign an initial feature vector to each node, and the initial feature vector includes the statistical quantities of the current - fluctuation feature, temperature - change feature, and aging index corresponding to the charging - pile line segment; Iteratively update the node features through a multi - layer graph - attention network, and each layer of the graph - attention network performs the following operations: Calculate the attention coefficient between the current node and its neighbor nodes, and the attention coefficient is dynamically adjusted based on the current - phase difference and temperature - conduction rate between nodes; Weightedly aggregate the neighbor node features according to the attention coefficient, and perform a residual connection with the current node features; Input the updated node features into a gated linear unit for non-linear transformation; Concatenate the node features output by the last layer of the graph attention network and the output features of the local feature extraction branch in the channel dimension to generate the fused global-local joint features.

6. The method according to claim 1, characterized in that The generating of the hierarchical warning signal set according to the risk levels corresponding to the respective spatial nodes in the line anomaly risk probability distribution map includes: Divide the line anomaly risk probability distribution map into multiple risk level regions, and each risk level region corresponds to a preset warning response strategy; For the first risk level region, extract the real-time current spectrum features of the line segments therein, and detect whether there is a harmonic resonance phenomenon; if a harmonic resonance is detected, generate a first-level warning signal including a harmonic suppression instruction; For the second risk level region, calculate the deviation degree of the temperature change rate of the line segments in this region from the current load, and generate a second-level warning signal including a load balancing instruction when the deviation degree exceeds the dynamic threshold; For the third risk level region, monitor the cumulative change amount of the line aging index, and generate a third-level warning signal including a preventive maintenance reminder when the cumulative change amount reaches the preset critical value; Topologically sort the first-level warning signal, the second-level warning signal, and the third-level warning signal according to the spatial position to generate the hierarchical warning signal set that matches the physical layout of the charging pile line.

7. The method according to claim 6, wherein The detecting whether there is a harmonic resonance phenomenon includes: Obtain the current waveform sampling data of the target line segment, and extract the frequency-amplitude pair set of the fundamental wave component and each harmonic component through time-frequency conversion processing; Calculate the matching degree between the harmonic frequencies in the frequency-amplitude pair set and the line inherent resonance frequency spectrum, and generate the frequency offset amount and resonance tendency index corresponding to each harmonic component; Screen the potentially dangerous harmonic components based on the resonance tendency index, and input the amplitudes of the potentially dangerous harmonic components into an energy accumulation predictor in time series to generate a harmonic energy accumulation trajectory; Perform a trend analysis on the harmonic energy accumulation trajectory, and when it is detected that the continuous rising slope exceeds the safety threshold, trigger a multi-band interference detection process: Extract the harmonic phase difference sequence of adjacent frequency bands, and calculate the phase synchronization degree index and the energy superposition effect coefficient; Input the phase synchronization degree index and the energy superposition effect coefficient into a resonance risk classifier to generate a comprehensive resonance risk level; Adjust the response intensity parameters of the harmonic suppression instruction according to the comprehensive resonance risk level, and the response intensity parameters include the filter start priority and the attenuation depth gradient; Bind the adjusted response intensity parameters to the harmonic suppression instruction in the first-level warning signal to generate an optimized suppression instruction set that dynamically adapts to the electrical characteristics of the target line segment.

8. The method according to claim 1, characterized in that, The triggering of the dynamic protection mechanism based on the hierarchical warning signal set includes: Parse the line segment identifiers and warning response strategy parameters corresponding to each warning signal in the hierarchical warning signal set, extract the set of line segments covered by the first-level warning signal including the harmonic suppression instruction, and the electrical connection topology corresponding to the set of line segments; Send a current cutoff instruction to the target charging pile in the set of line segments. The current cutoff instruction includes a cutoff phase angle range and a duration parameter generated based on the harmonic suppression instruction, and the current waveform distortion rate and the temperature gradient change amount after the cutoff operation are collected in real time; When the current waveform distortion rate exceeds the preset distortion threshold or the temperature gradient change amount fails to reach the expected decrease, activate the power attenuation compensation process for adjacent line segments: Determine the propagation path sequence of power attenuation according to the electrical connection topology relationship, and dynamically allocate the power attenuation ratio of each path node based on the load balancing instruction in the early warning response strategy parameters; Convert the power attenuation ratio into a pulse width modulation waveform, inject it into the power regulation loop of the corresponding charging pile, and generate a stepped attenuation signal synchronized with the cutoff phase angle range; Collect the line state feedback data under the action of the stepped attenuation signal, including the distribution of the attenuated current harmonic components and the temperature field uniformity index, and input them into the fire risk prediction model for real-time risk re-evaluation; Update the line abnormal risk probability distribution map according to the real-time risk re-evaluation result. If the diffusion trend of the harmonic energy accumulation trajectory is detected, trigger the secondary protection strategy: Perform a voltage amplitude limit operation on the key nodes in the propagation path sequence to suppress the conduction path of abnormal harmonic components; After the voltage amplitude limit operation takes effect, recalculate the power attenuation ratio and superimpose a high-frequency oscillation suppression component to eliminate residual harmonic interference; Align the updated power attenuation parameters with the current cutoff instruction in time sequence to generate a dynamic protection instruction sequence including the connection relationship of multi-stage protection logic; Control the target charging pile cluster to perform cooperative protection operations through the dynamic protection instruction sequence, and continuously monitor the dynamic response curves of the current waveform distortion rate and the temperature gradient change amount.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps in the method according to any one of claims 1 to 8.

Images