Method and device for evaluating the risk of swelling of active sites of activated carbon

Abstract

This application belongs to the field of activated carbon technology, specifically disclosing a method and apparatus for assessing the bulging risk of activated carbon regeneration active sites. This application generates a side reaction feature sequence based on the leakage current time-series data and bulging response data of the activated carbon sample under test; performs one-dimensional spline mapping based on an improved novel neural network and fuses the results within the same backbone; determines the continuous satisfaction period based on the judgment time range; calculates the re-entry judgment value based on the time-series logic of the differentiable signal after writing the features; and performs a risk assessment based on the current bulging risk value. Through the above method, a side reaction feature sequence is generated around the side reaction evolution of activated carbon regeneration active sites during short-term float charging. An improved novel neural network is introduced for one-dimensional spline mapping and result fusion, enabling stable discrimination of the microscopic side reaction evolution. Furthermore, a risk assessment is performed based on the current bulging risk value, thereby effectively improving the reliability and accuracy of bulging risk assessment.

Description

Technical Field

[0001] This application belongs to the field of activated carbon technology, and more specifically, relates to a method and apparatus for assessing the bulging risk of activated carbon regeneration active sites. Background Technology

[0002] Activated carbon, due to its large specific surface area, abundant pore structure, and relatively moderate production cost, has long been a commonly used electrode active material for electrochemical double-layer capacitors. Studies have shown that the oxygen-containing functional groups and active sites on the surface of activated carbon not only affect capacitance performance but also leakage current, gas evolution, and aging durability. Under high voltage or continuous voltage holding conditions, side reactions gradually induce gas generation, leading to increased internal pressure and performance degradation, ultimately causing bulging and safety issues. Therefore, assessing the bulging risk of regenerated active sites within a limited testing period during the pre-assembly stage has become an indispensable core operation in the manufacturing and quality control of activated carbon-based supercapacitors.

[0003] Currently, common methods for assessing bulging risk rely on capacitance, internal resistance, single voltage-current response, or post-aging analysis results to measure macroscopic electrical characteristics and observe mechanisms. However, in the specific scenario of short-term float charging before assembly, these methods struggle to unify the changes in leakage current (first decaying and then returning), the hysteretic cumulative changes in bulging response, and their continuous coupling relationship into a single, continuous judgment object. Furthermore, they fail to translate experimental interpretations and aging attributions into a single, usable intermediate judgment quantity. Therefore, the reliability and accuracy of these methods in assessing bulging risk are relatively low. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this application is to provide a method and apparatus for assessing the swelling risk of active sites in activated carbon regeneration. This aims to solve the problems of existing technologies, which struggle to unify the changes in leakage current that first decays and then returns, the hysteretic cumulative changes in swelling response, and the continuous coupling relationship between the two into a single continuous judgment object, and the difficulty in directly converting experimental interpretation and aging attribution into a single usable intermediate judgment quantity.

[0005] To achieve the above objectives, in a first aspect, this application provides a method for assessing the bulging risk of active sites during activated carbon regeneration, comprising: A side reaction characteristic sequence is generated based on the leakage current time series data and bulging response data of the activated carbon sample to be tested; Based on an improved novel neural network, one-dimensional spline mapping is performed on each reaction feature in the side reaction feature sequence, and the leakage current related feature mapping result and the swelling current coupling feature mapping result are fused in the same main branch to obtain the side reaction intensity sequence. Based on the side reaction intensity sequence, a differentiable signal timing logic is constructed, and the starting point and ending point of the determination of the side reaction intensity sequence are constrained according to the test duration constraint. Based on the constrained starting point and ending point, the determination time range of the differentiable signal timing logic is determined. The continuous satisfaction period is determined according to the determination time range. The swelling current coupling feature corresponding to the continuous satisfaction period is written to the differentiable signal timing logic determination. The re-entry determination value is calculated according to the differentiable signal timing logic determination after writing the feature. The current bulging risk value of the activated carbon sample to be tested is calculated based on the reentry determination value, and a risk assessment is performed based on the current bulging risk value.

[0006] In one embodiment, the step of generating a side reaction characteristic sequence based on the leakage current time-series data and bulging response data of the activated carbon sample to be tested includes: The activated carbon sample to be tested is subjected to constant pressure floating charge according to the preset temperature and preset floating charge voltage to form the initial state of data collection. Leakage current is continuously collected from the starting state of acquisition according to a preset sampling interval, and leakage current timing data is generated based on the leakage current at each sampling time. Bulging response is also collected synchronously from the starting state of acquisition according to a preset sampling interval, and bulging response data is generated based on the bulging response at each sampling time. The leakage current change rate and leakage current change curvature are calculated based on the leakage current time series data, and the bulging change rate is calculated based on the bulging response data. The correlation degree of the rate of change at each time offset position is calculated based on the leakage current change rate and the bulging change rate, and the hysteresis correlation characteristics are determined based on the maximum rate of change correlation degree and its corresponding time offset position. The proportional relationship between the bulging response and the leakage current change at each sampling time is calculated based on the hysteresis correlation characteristics, the leakage current change rate, and the bulging change rate. The bulging current coupling characteristics corresponding to the leakage current change rate are then determined based on the proportional relationship. A side reaction feature sequence is generated based on the leakage current change rate, the leakage current change curvature, the bulging change rate, the hysteresis correlation feature, and the bulging current coupling feature.

[0007] In one embodiment, the step of performing one-dimensional spline mapping on each reaction feature in the side reaction feature sequence based on an improved novel neural network, and fusing the leakage current-related feature mapping results and the swelling current coupling feature mapping results in the same backbone to obtain the side reaction intensity sequence includes: The improved novel neural network performs one-dimensional spline mapping on the leakage current change rate and leakage current change curvature in the side reaction feature sequence, respectively; and the improved novel neural network performs one-dimensional spline mapping on the bulging change rate, hysteresis correlation feature and bulging current coupling feature in the side reaction feature sequence, respectively. The current contribution of the leakage current related feature mapping result at each sampling time is constrained based on the swelling current coupling feature mapping result, and the constrained leakage current related feature mapping result, swelling change rate mapping result, hysteresis related feature mapping result and swelling current coupling feature mapping result are fused in the same backbone; Within the same backbone, the fusion results are continuously transmitted according to the sampling time sequence, and the continuous strength changes of the side reaction state are calculated based on the continuous representation generated after transmission. The side reaction intensity values ​​are generated based on the continuous changes in the strength of the side reaction states, and the side reaction intensity values ​​are arranged in chronological order according to the side reaction characteristic sequence to obtain the side reaction intensity sequence.

[0008] In one embodiment, the steps of determining a continuous satisfaction period based on the determination time range, writing the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination, and calculating the re-entry determination value based on the differentiable signal timing logic determination after writing the feature, include: Multiple candidate apparent stability thresholds are set within the determination time range, and the side reaction intensity sequence is compared with the multiple candidate apparent stability thresholds respectively. The apparent stable time period is determined based on the first comparison result, and the recurrence time interval is determined based on the apparent stable time period and the termination time of the determination time range. Multiple candidate re-emergence thresholds are set within the re-emergence time interval, and the side reaction intensity sequence is compared with the multiple candidate re-emergence thresholds respectively. Based on the second comparison result, a re-rise time period that connects with the apparent stable time period is determined, and multiple candidate duration thresholds are set based on the re-rise time period and the continuous holding length within the re-occurrence time interval. The continuous satisfaction time period is determined based on the multiple candidate duration thresholds, and the bulging current coupling feature corresponding to the continuous satisfaction time period is written into the differentiable signal timing logic determination. The re-entry determination value is calculated based on the differentiable signal timing logic determination after the feature is written.

[0009] In one embodiment, the step of determining a continuous satisfaction time period based on the plurality of candidate duration thresholds, writing the bulging current coupling feature corresponding to the continuous satisfaction time period to the differentiable signal timing logic determination, and calculating the re-entry determination value based on the differentiable signal timing logic determination after writing the feature includes: The continuous duration within the recurrence time interval is compared with the multiple candidate duration thresholds, and the continuous satisfaction time period is determined based on the third comparison result; The bulging current coupling characteristic corresponding to the continuous satisfaction time period is written into the differentiable signal timing logic determination; Traverse each time segment of the time sequence logic determination of the differentiable signal after writing the features; The smoothness robustness of each time segment is calculated, and the results of each smoothness robustness calculation are aggregated to obtain the re-entry determination value.

[0010] In one embodiment, the step of calculating the current bulging risk value of the activated carbon sample to be tested based on the reentry determination value, and performing a risk assessment based on the current bulging risk value, includes: Calculate the difference between the reentry determination value and the preset risk threshold; The deviation direction code is determined based on the sign of the difference, and the absolute magnitude is calculated based on the re-entry determination value and the preset risk threshold. A risk comparison result is generated based on the absolute magnitude and the deviation direction encoding, and a risk offset is determined based on the risk comparison result; The reentry determination value is continuously mapped based on the risk comparison results and the risk offset to calculate the current bulging risk value of the activated carbon sample to be tested. The current bulging risk value is compared with the preset risk threshold, and a risk assessment is performed based on the fourth comparison result.

[0011] Secondly, this application provides a device for assessing the bulging risk of active sites during activated carbon regeneration, comprising: The generation module is used to generate a sequence of side reaction characteristics based on the leakage current time-series data and bulging response data of the activated carbon sample to be tested; The mapping module is used to perform one-dimensional spline mapping on each reaction feature in the side reaction feature sequence based on an improved novel neural network, and to fuse the leakage current related feature mapping results and the swelling current coupling feature mapping results in the same main branch to obtain the side reaction intensity sequence. The determination module is used to construct a differentiable signal timing logic determination based on the side reaction intensity sequence, and to constrain the determination start point and determination end point of the side reaction intensity sequence according to the test duration constraint, and to determine the determination time range of the differentiable signal timing logic determination based on the constrained determination start point and determination end point; The determining module is further configured to determine a continuous satisfaction period based on the determination time range, write the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination, and calculate the re-entry determination value based on the differentiable signal timing logic determination after writing the feature. The evaluation module is used to calculate the current bulging risk value of the activated carbon sample to be tested based on the reentry determination value, and to perform a risk assessment based on the current bulging risk value.

[0012] Thirdly, this application provides an electronic device, comprising: at least one memory for storing a program; and at least one processor for executing the program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to execute the method described in the first aspect or any possible implementation thereof.

[0013] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0014] Fifthly, this application provides a computer program product that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0015] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0016] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art: (1) This application focuses on the evolution of side reactions of activated carbon regeneration active sites in short-term floating charge. Based on the leakage current time series data and bulging response data of the activated carbon sample to be tested, a side reaction feature sequence is generated. Since the side reaction feature sequence is still a scattered change information, it cannot be directly used as the basis for the bulging risk value. Therefore, an improved new neural network is introduced. The reaction features in the side reaction feature sequence are mapped in one dimension spline, so that the above features are transformed from the original numerical changes into nonlinear responses reflecting the dynamics of the side reaction. Then, the leakage current related feature mapping results and the bulging current coupled feature mapping results are fused in the same backbone, so that the change information in the leakage current time series data and the cumulative information in the bulging response data form a continuous expression on the same time axis. The "attenuation", "return" and "continuation" are retained in the same intermediate object through the side reaction intensity sequence. From the side reaction feature sequence to the side reaction intensity sequence, and then to the re-entry judgment value, a continuous processing chain around the same failure mechanism is formed, which can effectively improve the reliability and accuracy of assessing the bulging risk.

[0017] (2) After obtaining the side reaction intensity sequence, this application uses the swelling current coupling feature in the side reaction characteristic sequence to construct a differentiable signal time-series logic judgment based on the side reaction intensity sequence. It requires that the side reaction intensity sequence meets the thresholds for re-rise and duration, while the swelling current coupling feature also reflects the continuous correlation between the swelling response and the leakage current change. Local fluctuations unrelated to swelling are excluded from the judgment to avoid the defects of not being able to distinguish between the two completely different risk forms of "always high" and "first decrease and then rise and continue", and misjudgment due to amplitude scaling, time drift and duration changes between different batches. Furthermore, the degree of common satisfaction of the re-occurrence time interval, apparent stability threshold, re-rise threshold, duration threshold and swelling current coupling feature is quantified into a re-entry judgment value through smoothing robustness calculation. At this time, the re-entry judgment value is not a simple comparison result of a single threshold, but a quantification result of the entire time structure, which constitutes the key bridging object between the side reaction intensity sequence and the activated carbon batch screening result, thereby effectively improving the accuracy of calculating the re-entry judgment value.

[0018] In summary, this application generates a side reaction feature sequence based on the leakage current time-series data and bulging response data of the activated carbon sample under test; it performs one-dimensional spline mapping on each reaction feature in the side reaction feature sequence based on an improved novel neural network, and fuses the leakage current related feature mapping results and the bulging current coupling feature mapping results in the same backbone to obtain a side reaction intensity sequence; it constructs a differentiable signal time-series logic determination based on the side reaction intensity sequence, and constrains the determination start point and determination end point of the side reaction intensity sequence according to the test duration constraint, and determines the determination time range of the differentiable signal time-series logic determination based on the constrained determination start point and determination end point; it determines a continuous satisfaction period based on the determination time range, writes the bulging current coupling feature corresponding to the continuous satisfaction period into the differentiable signal time-series logic determination, and calculates the re-entry determination value based on the differentiable signal time-series logic determination after writing the feature; it calculates the current bulging risk value of the activated carbon sample under test based on the re-entry determination value, and performs a risk assessment based on the current bulging risk value. By using the above method, a side reaction characteristic sequence is generated around the side reaction evolution of the active sites of activated carbon regeneration during short-term floating charging. This allows for the joint expression of leakage current return, bulging hysteresis, and continuous coupling relationship. An improved novel neural network is introduced to perform one-dimensional spline mapping and fusion of the mapping results, enabling the stable identification of microscopic side reaction evolution. Then, risk assessment is performed based on the current bulging risk value, thereby effectively improving the reliability and accuracy of bulging risk assessment. Attached Figure Description

[0019] Figure 1 This is one of the flowcharts illustrating the method for assessing the bulging risk of activated carbon regeneration active sites provided in the embodiments of this application; Figure 2 This is a contour line and color-coded diagram of bulging risk provided in the embodiments of this application; Figure 3 This is a dual heatmap of the average batch swelling risk and uncertainty distribution provided in the embodiments of this application; Figure 4 This is the second flowchart illustrating the method for assessing the bulging risk of activated carbon regeneration active sites provided in the embodiments of this application; Figure 5 This is a thermodynamic diagram of the side reaction intensity provided in the embodiments of this application; Figure 6 This is a schematic diagram of the module structure of the activated carbon regeneration active site swelling risk assessment device provided in the embodiments of this application; Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.

[0022] The terms "first" and "second," etc., used in the specification and claims herein are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.

[0023] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0024] Based on this, embodiments of this application provide a method for assessing the bulging risk of active sites during activated carbon regeneration, referring to... Figure 1 , Figure 1 This is one of the flowcharts illustrating the method for assessing the bulging risk of activated carbon regeneration active sites provided in this application embodiment. In this embodiment, the method for assessing the bulging risk of activated carbon regeneration active sites includes steps S10 to S50: Step S10: Generate a side reaction characteristic sequence based on the leakage current time series data and bulging response data of the activated carbon sample to be tested.

[0025] It should be noted that the application scenario in this embodiment is short-term float charging before assembly. The leakage current time series data refers to the time series data formed by arranging the continuously collected leakage currents according to the order of sampling times. Similarly, the bulging response data refers to the data formed by arranging the synchronously collected bulging responses according to the same time order as the leakage current time series data. When collecting leakage current and bulging response data, it is necessary to perform constant-pressure float charging on the activated carbon sample under test according to the preset temperature and preset float charging voltage, and maintain the preset temperature and preset float charging voltage during the collection process to form the initial state of collection. The side reaction characteristic sequence refers to the time series formed by arranging the leakage current related characteristics, bulging change rate, hysteresis related characteristics, and bulging current coupling characteristics according to the corresponding relationship of the same sampling time.

[0026] Further, step S10 includes: performing constant-pressure float charging on the activated carbon sample to be tested according to a preset temperature and a preset float charging voltage to form a data acquisition start state; continuously acquiring leakage current from the data acquisition start state at preset sampling intervals, generating leakage current time-series data based on the leakage current at each sampling time; synchronously acquiring bulging response from the data acquisition start state at preset sampling intervals, generating bulging response data based on the bulging response at each sampling time; calculating the leakage current change rate and leakage current change curvature based on the leakage current time-series data, and calculating the bulging change rate based on the bulging response data; and calculating the leakage current... The correlation between the rate of change of flow and the rate of change of bulging at each time offset position is calculated, and the hysteresis correlation feature is determined based on the correlation of the largest rate of change and its corresponding time offset position. The proportional relationship between the bulging response and the change of leakage current at each sampling time is calculated based on the hysteresis correlation feature, the rate of change of leakage current, and the rate of change of bulging, and the bulging current coupling feature corresponding to the rate of change of leakage current is determined based on the proportional relationship. The side reaction feature sequence is generated based on the rate of change of leakage current, the curvature of the change of leakage current, the rate of change of bulging, the hysteresis correlation feature, and the bulging current coupling feature.

[0027] It should be understood that before data acquisition, the activated carbon sample to be tested needs to be subjected to constant-pressure float charging according to a preset temperature and preset float charging voltage. This preset temperature and preset float charging voltage should be maintained throughout the acquisition process to establish the initial acquisition state. The preset temperature and preset float charging voltage refer to the temperature and float charging voltage required for constant-pressure float charging of the activated carbon sample to ensure that the leakage current timing data and bulging response data are acquired under consistent conditions. Specifically, the activated carbon sample to be tested is placed in the constant-pressure float charging acquisition environment before assembly, and the preset temperature and preset float charging voltage are used as the consistent acquisition conditions maintained throughout the entire acquisition period. When the constant-pressure float charging has reached the preset float charging voltage and the acquisition environment has reached the preset temperature, the current moment is recorded as... At this time This can represent the start time of data collection, and the state of the activated carbon sample to be tested corresponding to the start time of data collection is determined as the start state of data collection. The preset sampling interval is denoted as... At this time This represents the fixed time interval between two adjacent samples; the total number of sampling times is denoted as... At this time This represents the number of discrete samples from the start of data acquisition to the end of the acquisition process. In short-time float charging scenarios, We set 120 to construct a unified timeline. ,in, Indicates the first Each sampling time, Indicates the sampling index.

[0028] Understandably, starting from the initial state of data collection, in each... Simultaneous acquisition of leakage current and bulging response, and the first The leakage current collected at each sampling time is denoted as . ,Should Indicates the location at the sampling time A single leakage current sample value; the first The bulging response collected at each sampling time is denoted as ,Should Indicates the location at the sampling time Individual bulging response sample values. Arranged in ascending order of sampling index. To generate leakage current timing data Arranged according to the sampling index order that is completely consistent with the leakage current timing data. To generate bulging response data ,because and Sharing the same data acquisition start time The same preset sampling interval and the same sampling endpoint Each sampling index Each uniquely corresponds to a set of simultaneously collected results. and Therefore, the leakage current timing data and the bulging response data already meet the corresponding requirements of the same sampling start point, the same sampling interval, and the same sampling end point when they are generated.

[0029] It should be noted that the sampling index corresponding and The sampling index is retained as the original baseline value at the start of the acquisition process. to corresponding and The leakage current timing data is retained as time-series variation data. According to the sampling settings of this embodiment, the leakage current timing data can be written as a discrete sequence of length 120. The bulging response data can be written as a discrete sequence of length 120. Both sets of data can be represented as The data is presented in column vector form. The correspondence between leakage current time-series data and bulging response data on the same sampling index axis is directly used for time-by-time calculation of leakage current change rate, current change curvature, and bulging change rate, and is also directly used to determine hysteresis correlation characteristics and bulging current coupling characteristics within a preset time delay range. Therefore, under the condition that the preset temperature and preset float charge voltage remain unchanged, the acquisition process does not form isolated sampled values, but rather two sets of synchronous discrete sequences sharing a unified time base, so that each sampling index position simultaneously corresponds to a leakage current value and a bulging response value.

[0030] Understandably, within the sampling start to sampling end range of the leakage current time-series data, starting from the second sampling time, based on the leakage current value at the current sampling time, the leakage current value at the previous sampling time, and the corresponding sampling time interval, the leakage current change rate at each current sampling time is calculated to generate leakage current change rates arranged continuously according to sampling time. For example, in the sampling index... to Within the range, the leakage current value of the current sampling index is subtracted from the leakage current value of the previous sampling index, and then... Converted to change per unit time, the leakage current change rate is obtained. ,in, Indicates sampling index The corresponding leakage current change rate; Write sampling index The location forms a leakage current change rate of equal length to the leakage current time-series data. Based on the rate of change value of the leakage current at the current sampling moment, the rate of change value of the previous sampling moment, and the corresponding sampling time interval, the current change curvature at each current sampling moment is calculated to generate the leakage current change curvature corresponding to the leakage current rate of change at each moment. For example, in the sampling index... to Within the range, the difference between the leakage current change rate of the current sampling index and the leakage current change rate of the previous sampling index is calculated, and then... Converted to change per unit time, the curvature of the current change is obtained. ,in, Indicates sampling index The corresponding leakage current change curvature; Write sampling index and The position forms At this time and It is determined to be a one-dimensional component of the leakage current-related characteristics at each sampling index.

[0031] It should be understood that, within the sampling start and end range of the bulging response data, starting from the second sampling time, based on the bulging response value at the current sampling time, the bulging response value at the previous sampling time, and the corresponding sampling time interval, the bulging change rate at each current sampling time is calculated to generate the bulging change rate corresponding to the leakage current change rate at each moment. For example, in the sampling index... to Within the range, the difference between the bulging response value of the current sampling index and the bulging response value of the previous sampling index is calculated, and then... Converted to change per unit time, the rate of change of bulge is obtained. , Indicates sampling index The corresponding rate of change of bulge; Write sampling index The location is chosen to generate a length equal to the rate of change of the leakage current. After calculating the leakage current change rate, leakage current change curvature, and bulging change rate respectively, the preset time delay range is set according to... Converted to the number of sampling intervals, the maximum sampling interval offset is obtained. ,in, This represents the integer boundary of the preset time delay range on the sampling index axis; the set of time offset positions is written as... , Represents the set of all candidate time offset positions; the minimum comparison length for multiple consecutive sampling times is preset to... , Indicates the number of consecutive samples used for consistency filtering in the direction of change; the starting sampling index of the consecutive window is written as... , This indicates the starting point of the window that satisfies the continuous comparison condition. For each time offset position... Shift the rate of change of bulging along the sampling index axis Each sampling interval is used, and the shifted bulging change rate and leakage current change rate are compared point by point on a common valid sampling index; when If the rate of change of bulge after translation is positive or negative at the same sampling index, it is determined that the direction of change is consistent; if either value is zero, it is determined that the direction of change is inconsistent. Only retain at least one continuous window. The time offset positions that consistently maintain the same direction of change are selected, while those that cannot maintain the same direction of change across multiple consecutive sampling times are removed.

[0032] Understandably, for each retained time offset, the correlation degree is calculated using a correlation metric that removes the mean and normalizes to the fluctuation amplitude. The correlation degree is expressed as... , Indicates the time offset position The corresponding degree of correlation is calculated as follows: .

[0033] in, Indicates the time offset position The degree of correlation; Indicates the time offset position; Indicates the sampling index; Indicates the time offset position The set of valid sampling indices below By simultaneously satisfying the sampling index and sampling index All fall into to Within range and sampling index The sampling indexes that fall within at least one continuous window that satisfies the same direction of change constitute the sample indexes. Indicates sampling index The rate of change of leakage current at the location; Indicates the time offset position Downsampling index Alignment bulge rate of change; express The arithmetic mean of the rate of change of internal leakage current; express The arithmetic mean of the rates of change of internal bulging; superscript Indicates the time offset position The following statistics; Indicates to All valid sampled indexes within the index are accumulated; This indicates that the square root normalization is applied to the accumulated result.

[0034] It should be noted that when any term in the denominator is zero, the current time offset is not retained; instead, the time offset corresponding to the given time offset is retained. Compare and take the maximum value as , This indicates the degree of association when the association is at its maximum; it will produce The time offset position is denoted as , This indicates the time offset position corresponding to the highest degree of correlation; and The characteristics were jointly identified as hysteresis correlation features. These hysteresis correlation features characterize the time-shift relationship between the leakage current change rate and the bulging change rate, which is jointly represented by the correlation degree at its maximum and the corresponding time offset position. The corresponding time offset position at its maximum correlation degree was obtained. After that, it can be based on Time-align the leakage current rate of change and the bulging rate of change. The leakage current rate of change consistent with hysteresis characteristics is written as... , Represents the rate of change of leakage current after time alignment; the rate of change of bulging consistent with the hysteresis correlation characteristic is written as , This represents the bulging change rate after time alignment. During time alignment, the sampling index of the leakage current change rate remains unchanged, and the bulging change rate is adjusted accordingly. The sampling index is shifted along the sampling index axis to a position matching the leakage current change rate. For sampling indices that exceed the sampling start or end point due to the shift, the bulging change rate corresponding to the nearest valid sampling index is used for boundary preservation. At each sampling index... First, put and Perform proportional calculations; then compare the leakage current values ​​at the same sampling index. Write the validity check only. Non-zero and Non-zero sampling indexes retain the calculated proportion, forming a bulging current coupling characteristic. , Indicates sampling index The corresponding bulging current coupling characteristics; in zero or For a sampling index that is zero, the bulging current coupling characteristic of the previous valid sampling index is written to the current sampling index. If there is no previous valid sampling index at the starting position, the bulging current coupling characteristic of the first valid sampling index is written to the starting position.

[0035] In the sampling index Construct a six-dimensional eigenvector , Indicates sampling index The corresponding row of side reaction features. The first dimension of the six-dimensional feature vector is the one-dimensional leakage current change rate, the second dimension is the one-dimensional current change curvature, the third dimension is the one-dimensional bulging change rate, the fourth dimension is the maximum correlation value in the hysteresis correlation feature, the fifth dimension is the corresponding time offset position in the hysteresis correlation feature, and the sixth dimension is the one-dimensional bulging current coupling feature. (The rest of the text appears to be a list of parameters and is left untranslated.) Arranged in ascending order of sampling index, forming a side reaction characteristic sequence. ,in, This represents a feature matrix organized by time. In a short-time float charging embodiment, the side reaction feature sequence... for The characteristic matrix of .

[0036] Step S20: Based on the improved novel neural network, perform one-dimensional spline mapping on each reaction feature in the side reaction feature sequence, and fuse the leakage current related feature mapping result and the swelling current coupling feature mapping result in the same main branch to obtain the side reaction intensity sequence.

[0037] Understandably, when generating the side reaction intensity sequence, an improved novel neural network is introduced. This improved novel neural network can be a Kolmogorov-Arnold network, receiving one row at each sampling time. Feature vectors, and according to the sampling index from arrive The input of the Kolmogorov-Arnold network is to receive the six-dimensional feature vectors in the side reaction feature sequence in sequence. At each sampling time, only the six input components corresponding to the current sampling index are used, without changing the temporal order of the side reaction feature sequence. For example, the input of the Kolmogorov-Arnold network can be a 120×6 side reaction feature sequence, and the output can be a 120×1 side reaction intensity sequence. The one-dimensional spline mapping represents the mapping method for establishing nonlinear correspondences for the leakage current change rate, current change curvature, bulging change rate, hysteresis correlation feature and bulging current coupling feature, respectively. The same backbone represents the same computation chain in the Kolmogorov-Arnold network that continuously calculates, constrains and fuses the leakage current correlation feature and the bulging current coupling feature.

[0038] It should be noted that in this implementation, the Kolmogorov-Arnold network receives a 1x6 feature vector at each sampling time. The first mapping layer has 24 neurons, of which 16 receive leakage current-related features and 8 receive swelling current coupling features. Each neuron is fully connected to the received input component, and each connection contains a set of one-dimensional spline weights and a bias, used to map a single feature value to a nonlinear response corresponding to the side reaction morphology. After the first mapping layer outputs a 1x24 intermediate representation, it enters the second mapping layer. The second mapping layer has 16 fully connected neurons, which simultaneously receive the 24 intermediate components and continue to perform one-dimensional spline mapping, outputting a 1x16 fused representation. The 1x16 fused representation is processed by a linear output neuron to form a one-dimensional side reaction intensity value at the current sampling time. After 120 sampling times, this is calculated sequentially to form a 120x1 side reaction intensity sequence. The scenario-based modification occurs within the existing mapping link from the first mapping layer to the second mapping layer. The first mapping layer does not perform indiscriminate mapping on the six input features. Instead, it first divides the six input features into four-dimensional leakage current-related features and two-dimensional swelling current-coupled features based on their physical meaning. Then, 24 neurons receive the two types of input features respectively. The four-dimensional leakage current-related features correspond to whether the side reaction changes from decay to recurrence, while the two-dimensional swelling current-coupled features correspond to whether the swelling response continuously follows the changes in leakage current. The second mapping layer merges the 24 intermediate columns into a 16-column fused representation within the same backbone, so that the recurrence information in the leakage current time series data and the cumulative information in the swelling response data jointly determine the side reaction intensity sequence in the same time expression. This structure makes the side reaction intensity sequence no longer correspond to the strength of a single segment, but to the continuous state of "whether it rises again after decay and is accompanied by continuous changes in swelling", providing a unified input object for the timing logic determination of differentiable signals.

[0039] Step S30: Construct a differentiable signal timing logic determination based on the side reaction intensity sequence, and constrain the determination start point and determination end point of the side reaction intensity sequence according to the test duration constraint, and determine the determination time range of the differentiable signal timing logic determination based on the constrained determination start point and determination end point.

[0040] It should be understood that the time-series logic decision for differentiable signals refers to a decision-making method that continuously calculates the satisfying relationships of the apparent stability threshold, recurrence time interval, re-rise threshold, duration threshold, and swelling current coupling characteristics within the decision time range. The test duration constraint represents the time range for the execution of the time-series logic decision for differentiable signals, which is jointly defined by the decision start point and the decision end point. For the side reaction intensity sequence, its decision start point and decision end point can be constrained according to the test duration, and the time range between the decision start point and the decision end point can be determined as the decision time range of the time-series logic decision for differentiable signals, so that the sampling times within the decision time range maintain a time order consistent with the side reaction intensity sequence.

[0041] Step S40: Determine the continuous satisfaction period according to the determination time range, write the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination, and calculate the re-entry determination value according to the differentiable signal timing logic determination after writing the feature.

[0042] It is understandable that the continuous satisfaction period refers to the period during which the continuous holding length of the re-rise period reaches the duration threshold. The re-entry judgment value represents the result of quantifying the degree of satisfaction of the apparent stability threshold, the re-occurrence time interval, the re-rise threshold, the duration threshold, and the swelling current coupling feature. After writing the swelling current coupling feature corresponding to the continuous satisfaction period into the differentiable signal timing logic judgment, the re-entry judgment value is further calculated.

[0043] Further, step S40 includes: setting multiple candidate apparent stability thresholds within the determination time range, and comparing the side reaction intensity sequence with each of the multiple candidate apparent stability thresholds; determining an apparent stability time period based on a first comparison result, and determining a re-occurrence time interval based on the apparent stability time period and the termination time of the determination time range; setting multiple candidate re-rise thresholds within the re-occurrence time interval, and comparing the side reaction intensity sequence with each of the multiple candidate re-rise thresholds; determining a re-rise time period that connects to the apparent stability time period based on a second comparison result, and setting multiple candidate duration thresholds based on the re-rise time period and the continuous holding length within the re-occurrence time interval; determining a continuous satisfaction time period based on the multiple candidate duration thresholds, and writing the swelling current coupling feature corresponding to the continuous satisfaction time period into the differentiable signal timing logic determination, and calculating a re-entry determination value based on the differentiable signal timing logic determination after writing the feature.

[0044] It should be noted that the sequence convergence stage is based on the side reaction intensity sequence. Coupling characteristics of bulging current For input, where, Indicates sampling index The corresponding side reaction intensity value, Indicates sampling index The corresponding bulging current coupling characteristics, This represents the total number of sampling moments, in the short-time float charging embodiment. The sequence convergence stage comprises one parameter layer, one smoothness robustness calculation layer, and one convergence neuron. The parameter layer consists of five scalar parameter neurons, each providing a decision starting point. Determine the endpoint Apparent stability threshold center value The threshold center value rises again and duration threshold center value Each scalar parameter neuron contains one learnable weight and one learnable bias. Determine the starting point. and determining the endpoint After the parameter layer output, it is discretized into sampling indices according to the nearest neighbor integer rule, and restricted to... Within the range; when the discrete decision starting point is greater than the discrete decision ending point, the decision ending point is rewritten with the same sampling index as the decision starting point. Duration threshold center value. After the parameter layer output, it is discretized into positive integer sample lengths according to the nearest neighbor integer rule, and limited to... to Within the range. The threshold parameter can be written as , This refers to the bulging current coupling characteristic comparison parameters that are pre-set and fixed before the execution of the differentiable signal timing logic determination. The apparent stability threshold step size is written as... The threshold step size is then written as... The duration threshold step size is written as , , and All of these are positive scalars that are pre-set before the determination begins, among which Use positive integer sampling lengths. Write the decision time range as... The smoothness robustness calculation layer is expanded according to the sampling index as follows: There are 1 time node, and each time node receives 1 side reaction intensity value. One bulging current coupling characteristic value and the parameter layer given , , , , The converging neuron receives the local satisfaction level formed by the smoothness robustness calculation layer over the effective time segment and outputs a 1D reentry decision value. .

[0045] It should be noted that the judgment time range refers to the time range used for performing the timing logic judgment of differentiable signals from the judgment start point to the judgment end point. The apparent stability threshold refers to the threshold used to determine whether the side reaction intensity sequence has entered the apparent stable time period. The falling time segment refers to the time segment in which the side reaction intensity sequence continuously falls before the apparent stable time period. The apparent stable time period refers to the time period in which the side reaction intensity sequence continuously falls below the apparent stability threshold and is continuously connected with the falling time segment. The reappearance time interval refers to the time interval after the apparent stable time period and is constrained by the test duration. The re-rise threshold refers to the threshold used to determine whether the side reaction intensity sequence crosses and enters a continuously rising state within the reappearance time interval. The threshold parameter refers to the parameter in which the swelling current coupling feature is written into the timing logic judgment of differentiable signals and compared time by time. The constraint-satisfied continuous time period refers to the time period in which the constraint-satisfied continuous time period is retained after comparing the swelling current coupling feature with the threshold parameter time by time.

[0046] It should be understood that after determining the decision time range for the time-series logic of the differentiable signal, multiple candidate apparent stability thresholds are set within the decision time range. The side reaction intensity sequence is compared with each candidate apparent stability threshold time-by-time, and time segments that continuously fall below the candidate apparent stability threshold and are continuously connected with the falling time segments preceding the apparent stability time period are matched. The apparent stability threshold is determined based on the matching results, forming the apparent stability time period. For example, with Centered on, take , and Three candidate apparent stability thresholds are formed. Then, the side reaction intensity sequence is scanned from smallest to largest sampling index. The largest continuous sampling segment satisfying "the side reaction intensity value at the current sampling index is not greater than the side reaction intensity value at the previous sampling index" is extracted as a falling time segment. The side reaction intensity sequence is compared with each candidate apparent stability threshold index by index. Sampling segments that continuously fall below the candidate apparent stability threshold and whose starting sampling index is continuously connected to the ending sampling index of a falling time segment are retained from the first comparison results. The continuous length of all retained sampling segments is compared, and the candidate apparent stability threshold corresponding to the retained sampling segment with the largest continuous length is selected as the apparent stability threshold. The determined apparent stability threshold is denoted as […]. , will with The corresponding retained sampling segment is determined as the apparent stable time period, and the starting sampling index of the apparent stable time period is denoted as... The termination sampling index of the apparent stable time period is denoted as When multiple candidate apparent stability thresholds have the same continuous length, the candidate apparent stability threshold with the smaller value is retained.

[0047] Understandably, after determining the apparent stable time period based on the first comparison result, the end time of the apparent stable time period is used as the starting position of the recurrence time interval. Combined with the end time of the judgment time range, the recurrence time interval is determined, ensuring it is located after the apparent stable time period and constrained by the test duration. Multiple candidate recurrence thresholds are set within the recurrence time interval. The side reaction intensity sequence is compared with each of these candidate recurrence thresholds time-by-time. Time segments crossing candidate recurrence thresholds and those in a continuously rising state are matched. Based on the matching results, the recurrence threshold is determined to generate a recurrence time period that connects to the apparent stable time period. For example, after determining the apparent stable time period, the recurrence time interval is written as... The time interval reappeared The starting sampling index is directly taken from the ending sampling index of the apparent stable period. The time interval reappeared The termination sampling index is directly taken from the determination endpoint. During the time interval of recurrence Inside, with Centered on, take , and Three candidate re-rise thresholds are generated, and a comparison is performed on a sampling index-by-sampling basis for each candidate re-rise threshold, while simultaneously checking the continuous rise state and the swelling current coupling characteristic comparison condition. The sampling position that crosses a candidate re-rise threshold is defined as: the side reaction intensity value at the previous sampling index is lower than the candidate re-rise threshold, and the side reaction intensity value at the current sampling index reaches or exceeds the candidate re-rise threshold. A continuous rise state is defined as: within a continuous sampling segment after crossing a candidate re-rise threshold, the side reaction intensity value at the current sampling index is not less than the side reaction intensity value at the previous sampling index. The swelling current coupling characteristic comparison condition is defined as: the swelling current coupling characteristic at the same sampling index satisfies... Only candidate re-rise thresholds that simultaneously meet the conditions of crossing the candidate re-rise threshold, being in a continuous rising state, and satisfying the comparison condition of bulging current coupling characteristics are retained. The continuous lengths of all retained sampling segments are compared, and the candidate re-rise threshold corresponding to the retained sampling segment with the largest continuous length is selected as the re-rise threshold. The determined re-rise threshold is denoted as... and will be with The corresponding retained sampling segment is determined as the second rising time period, and the starting sampling index of the second rising time period is denoted as... The terminating sampling index of the next rising period is denoted as When multiple candidates raise the threshold again and obtain the same continuous length, the candidate with the larger value is retained when raising the threshold again.

[0048] It should be noted that after determining the re-rise time period that connects to the apparent stable time period based on the second comparison result, multiple candidate duration thresholds are set based on the continuous holding length of the re-rise time period within the re-occurrence time interval. The continuous holding length of the re-rise time period is compared with each candidate duration threshold, and the duration threshold is determined based on the candidate duration threshold that meets the comparison conditions, forming a continuous satisfaction time period. The swelling current coupling characteristic is compared with the threshold parameter time-by-time. If the comparison conditions are met, the continuous satisfaction time period is retained; if the comparison conditions are not met, the corresponding time segment is removed from the continuous satisfaction time period, forming a constrained continuous satisfaction time period. For example, the continuous holding length of the re-rise time period is denoted as... , From the sampling index range to The number of sampling points is determined. , and As the initial candidate duration threshold source, and will be less than The numerical value is rewritten as , will be greater than The numerical value is rewritten as The rewritten positive integer sampling length is used as the threshold for three candidate durations. The continuous hold length of the re-rising time interval is then determined. Comparing each of the three candidate duration thresholds, only those samples with a continuous holding length reaching the candidate duration threshold and all bulging current coupling characteristics within the corresponding sampling segment are retained. The candidate duration threshold is determined by selecting the maximum candidate duration threshold that meets the conditions, and the determined duration threshold is denoted as . , will reach The continuous sampling segments are determined as the time period that is continuously satisfied.

[0049] Further, the steps of determining the continuous satisfaction period based on the multiple candidate duration thresholds, writing the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination, and calculating the re-entry determination value based on the differentiable signal timing logic determination after writing the feature, include: comparing the continuous holding length within the re-occurrence time interval with the multiple candidate duration thresholds, and determining the continuous satisfaction period based on the third comparison result; writing the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination; traversing each time segment of the differentiable signal timing logic determination after writing the feature; calculating the smoothing robustness of the satisfaction degree of each time segment, and aggregating the smoothing robustness calculation results to obtain the re-entry determination value.

[0050] Understandably, after determining the continuous satisfaction period based on the third comparison result, the swelling current coupling characteristics corresponding to each sampling index within the continuous satisfaction period are then written into the differentiable signal timing logic for judgment, and the process is executed again. and Comparison, eliminate those that do not meet the requirements. The sampling points are defined; when multiple consecutive sampling segments appear after removal, the sampling segment with the longest consecutive length is retained as the time period for continuous satisfaction of the constraint. The starting sampling index of the time period for continuous satisfaction of the constraint is denoted as... Let the termination sampling index of the time period during which the constraint is continuously satisfied be denoted as .

[0051] It should also be emphasized that if any of the following time periods—the apparent stability period, the re-rise period, or the period of continued satisfaction after constraint—is absent, the re-entry determination value is directly set to 0. When the apparent stability period, the re-emergence period, and the period of sustained satisfaction after constraint all exist, the smoothing robustness calculation layer performs smoothing robustness calculations on the satisfaction degree of the apparent stability threshold, the re-emergence time interval, the re-emergence threshold, the duration threshold, and the bulging current coupling feature according to the sampling index. The converging neuron aggregates the satisfaction degree of all time segments into a re-entry judgment value. Specifically: .

[0052] in, Indicates the reentry determination value; This represents the apparent stability threshold after it has been determined. This indicates the threshold for further increases after a certain point has been reached; This represents the duration threshold after it has been determined. Indicates the threshold parameter; Indicates sampling index The corresponding side reaction intensity value; Indicates sampling index The corresponding bulging current coupling characteristics; and These represent the starting and ending sampling indices for the apparent stable time period, respectively. and These represent the starting and ending sampling indices for the next upward period, respectively. and These represent the starting and ending sampling indices of the time period during which the constraint is continuously satisfied. This indicates the sampling index corresponding to the endpoint. Indicates the sampling index used for summation; This indicates that the summation is performed on all sample points within the corresponding sample index range; Indicates taking The larger of the values ​​within the parentheses; Indicates taking The smaller of the values ​​in parentheses; The constant in the formula represents the absolute value. The expression represents the smoothing normalization constant; in the formula... This indicates that the fifth root of the product of the five satisfaction levels is converged.

[0053] It should be noted that in the above formula, the first term corresponds to the degree of satisfaction of the apparent stability threshold during the apparent stability period; the second term corresponds to the proportion of the re-rise period in the re-occurrence time interval; the third term corresponds to the degree of satisfaction of the re-rise threshold during the re-rise period; the fourth term corresponds to the degree of satisfaction of the duration threshold during the period of continuous satisfaction after constraint; and the fifth term corresponds to the degree of satisfaction of the bulging current coupling feature and the threshold parameter during the period of continuous satisfaction after constraint. The time nodes in the smoothness robustness calculation layer are compared and normalized point by point according to their respective sampling indices. The converging neurons perform unified convergence on the five satisfaction levels to obtain a scalar form re-entry judgment value. .

[0054] Step S50: Calculate the current bulging risk value of the activated carbon sample to be tested based on the reentry determination value, and conduct a risk assessment based on the current bulging risk value.

[0055] It should be understood that the current swelling risk value refers to the value of the risk of swelling at the regenerated active sites of the activated carbon sample to be tested. After calculating the current swelling risk value of the activated carbon sample to be tested based on the re-entry judgment value, a risk assessment is carried out based on the current swelling risk value. At this time, the batch-level swelling risk control of activated carbon can be achieved based on the risk assessment results.

[0056] Further, step S50 includes: calculating the difference between the reentry determination value and a preset risk threshold; determining a deviation direction code based on the sign of the difference, and calculating an absolute magnitude based on the reentry determination value and the preset risk threshold; generating a risk comparison result based on the absolute magnitude and the deviation direction code, and determining a risk offset based on the risk comparison result; continuously mapping the reentry determination value based on the risk comparison result and the risk offset to calculate the current bulging risk value of the activated carbon sample to be tested; comparing the current bulging risk value with the preset risk threshold, and performing a risk assessment based on the fourth comparison result.

[0057] It is understandable that the re-entry determination value is calculated. With preset risk threshold The difference between ,in, This represents the signed difference between the reentry determination value and the preset risk threshold. The sign of the difference can be used to determine the deviation direction code, and the absolute magnitude can be calculated based on the reentry determination value and the preset risk threshold. Specifically: at the reentry determination value... When the deviation direction is coded, it is recorded as And the deviation range is recorded as ; In the re-entry determination value When the deviation direction is coded, it is recorded as And the deviation range is recorded as ; In the re-entry determination value When the deviation direction is coded, it is recorded as And the deviation range is recorded as .in, A benchmark value used to distinguish between high and low risk levels. Fixed write and satisfy ; A directional marker indicating whether the reentry determination value is above, below, or coincides with a preset risk threshold. This indicates the absolute extent to which the re-entry determination value deviates from the preset risk threshold.

[0058] It should be noted that after determining the absolute magnitude and deviation direction codes, a risk comparison result is generated, which can be expressed as: ,in, This indicates that it consists of the deviation direction code and the deviation magnitude. The risk comparison result refers to the comparison result of the reentry determination value relative to the preset risk threshold in terms of both the direction and magnitude of deviation. Further, based on the risk comparison result, a risk offset is determined. This risk offset is represented by a signed scalar that simultaneously characterizes the direction and magnitude of deviation. At that time, the risk offset is recorded as a positive deviation. ; in direction markers When this happens, the risk offset is recorded as a negative deviation. When the direction marker At that time, the risk offset is recorded as With this construction method, the absolute value of the risk offset directly corresponds to the degree to which the re-entry determination value deviates from the preset risk threshold, and the sign of the risk offset directly corresponds to whether the re-entry determination value is above or below the preset risk threshold.

[0059] It should be understood that after determining the risk offset based on the risk comparison results, a branch calculation rule with a preset risk threshold as the mapping center can be used to continuously map the re-entry determination value based on the risk comparison results and the risk offset, with the direction marker... At that time, the risk mapping coefficient and deviation range Multiply to obtain the mapping increment, then add the mapping increment to the preset risk threshold. Above, when the sum is greater than At that time, the current inflation risk value will be... Cut off as In the direction marker At that time, the risk mapping coefficient and deviation range Multiply to obtain the mapping increment, then reduce the mapping increment from the preset risk threshold. Deduct from the middle, when the deduction result is less than At that time, the current inflation risk value will be... Cut off as When the direction marker At that time, directly set the current bulging risk value. Recorded as the preset risk threshold Through the above continuous mapping, as the risk offset increases, the current bulging risk value... Compared to the preset risk threshold The distance increases synchronously; in the risk offset When the sign is positive, the current inflation risk value is... Maintain at the preset risk threshold The above; risk offset When the sign is negative, the current inflation risk value Maintain at the preset risk threshold the following.

[0060] Understandably, after calculating the current bulging risk value of the activated carbon sample to be tested, the current bulging risk value is... With preset risk threshold Compare again, and adjust the current bulging risk value based on the comparison results. Assign to the corresponding risk interval. To ensure that the risk interval assignment is consistent with the preset risk threshold position, first, according to the boundary index, place it in the preset risk interval boundary sequence. Search for values ​​containing the current bulge risk value. The only risk range. At the current inflation risk value. At that time, only at the boundary index To Boundary Index Search within the corresponding risk range to find the satisfying Location; in Preset risk threshold At that time, only at the boundary index To Boundary Index Search within the corresponding risk range to find the satisfying Location; in At that time, directly set the current bulging risk value. Assign to the last risk zone and find the current bulging risk value. After the risk range is defined, its position index in the preset order is determined. Recorded as bulging risk level .in, This represents a 1-dimensional discrete-level scalar. The preset risk interval boundary sequence is denoted as... ,in, Indicates the first Each risk zone boundary, This represents the total number of risk intervals. All risk interval boundaries are written in ascending order and satisfy the following conditions: , In the preset risk interval boundary sequence, a boundary that coincides with the preset risk threshold is set, denoted as... ,in, This indicates the boundary index of the preset risk threshold within the preset risk interval boundary sequence.

[0061] It should be noted that the batch number of the activated carbon sample to be tested is denoted as follows: , This refers to the code value written during the testing and registration phase and used to uniquely identify the current activated carbon batch. The preset screening conditions are recorded as follows: , This indicates a discrete level threshold code pre-set before batch screening is performed. (Bloating risk level) With preset screening conditions Using the same risk level order and coding scale, the higher the code value, the higher the risk level, for batch identification of activated carbon. Risk level of bulging Bind them to the same decision unit, and record the filtering decision status as , Indicates batch labeling of activated carbon When comparing the corresponding screening and judgment status, the size of the discrete level code is directly determined, and the swelling risk level is no longer considered. Perform remapping. When Greater than the preset screening condition At that time, the filtering and judgment status will be determined. Recorded as ;when Equal to preset screening conditions At that time, the filtering and judgment status will be determined. Also recorded as ;when Less than the preset screening condition At that time, the filtering and judgment status will be determined. Recorded as After applying this comparison rule, the state is filtered and determined. It is a one-dimensional binary scalar, where, This indicates that the risk level of bloating has reached the preset screening criteria. This indicates that the risk level of bloating has not met the preset screening criteria. (In screening judgment status) After formation, a batch category code is generated. , Indicates batch labeling of activated carbon Batch category. Batch category code. Value selection rules and filtering judgment status Maintain consistency: when At that time, the batch category code will be used. Recorded as ;when At that time, the batch category code will be used. Recorded as .in, Indicates the batches to be screened out. This indicates the batch to be released. This process determines the risk level of bulging. The comparison results are directly converted into discrete category codes that can be written into the batch processing results.

[0062] It should be understood that, in the screening and judgment state At that time, the batch number of activated carbon will be marked. The corresponding activated carbon batch is identified as the screening batch, and screening results before assembly are generated. Screening results before assembly Written as , express Result vector. Results filtered out before assembly. The first dimension is the batch identifier for activated carbon. The second dimension is the risk level of bloating. The third dimension represents the preset screening conditions. The fourth dimension is the filtering and judgment status. The fifth dimension is the batch category code. Results were screened out before assembly. In the middle, the fifth dimension has a fixed value of This indicates that the current activated carbon batch has been identified as a rejection batch; no assembly release result is generated within this decision branch. (Screening decision status) At that time, the batch number of activated carbon will be marked. The corresponding activated carbon batch was identified as the release batch, and an assembly release result was generated. Assembly release results Written as , express Result vector. Assembly release result. The first dimension is the batch identifier for activated carbon. The second dimension is the risk level of bloating. The third dimension represents the preset screening conditions. The fourth dimension is the filtering and judgment status. The fifth dimension is the batch category code. Assembly release results In the middle, the fifth dimension has a fixed value of This indicates that the current activated carbon batch has been determined as a release batch; within this decision branch, no pre-assembly screening results are generated.

[0063] It should be noted that the results were screened out before assembly. Or assembly release results After formation, the batch screening results of activated carbon are recorded as follows: , Indicates the final batch disposal Filter result vector. In the filtering decision state. At that time, the batch screening results of activated carbon will be... Write it directly as the screening result before assembly. In the screening and judgment state At that time, the batch screening results of activated carbon will be... Write it directly as assembly release result Activated carbon batch screening results The five components remain consistent with the pre-assembly screening results. and assembly release results Record activated carbon batch identifiers in the same field order. , bloating risk level Preset screening conditions Filter and determine status and batch category code .

[0064] It should be understood that, reference Figure 2 , Figure 2 This diagram illustrates the contour lines and color-coded representation of bulging risk. Specifically, it predicts the bulging risk level within a two-dimensional operating space defined by temperature T and float charge voltage U. The horizontal axis represents float charge voltages U1 to U5, and the vertical axis represents temperatures T1 to T5. The background is a 10×10 grid. Multiple risk contour lines are drawn on this plane, categorized by grayscale into safe, low-risk, medium-risk, and high-risk zones. Darker colors indicate higher predicted bulging risk. This diagram visually demonstrates the combined trend of risk with temperature and voltage, helps determine recommended safe operating ranges and critical boundaries, and explains why float charge voltage or float charge time needs to be reduced or shortened in high-temperature, high-pressure areas. (Reference) Figure 3 , Figure 3A dual heatmap is used to represent the mean and uncertainty distribution of batch bulging risk. Specifically, in the left 10×10 grayscale image, the horizontal axis represents batch numbers 1-10, and the vertical axis represents batch groups (group 1-10). The grayscale value of each grid represents the average bulging risk of that batch, with high grayscale bands corresponding to "high mean areas." The right grayscale image uses the same coordinate axis, but the grayscale value represents prediction uncertainty (such as standard deviation), with darker areas representing "high uncertainty areas." By comparing the two images, batches with "high risk and low uncertainty" (which can be directly rejected) and batches with "medium risk but high uncertainty" can be distinguished, thus providing a basis for setting sorting thresholds on the production line.

[0065] This embodiment generates a side reaction feature sequence based on the leakage current time-series data and bulging response data of the activated carbon sample under test. A novel, improved neural network is used to perform one-dimensional spline mapping on each reaction feature in the side reaction feature sequence. The leakage current-related feature mapping results and the bulging current coupling feature mapping results are fused in the same backbone to obtain a side reaction intensity sequence. A differentiable signal timing logic determination is constructed based on the side reaction intensity sequence. The determination start and end points of the side reaction intensity sequence are constrained according to the test duration. The determination time range of the differentiable signal timing logic determination is determined based on the constrained determination start and end points. A continuous satisfaction period is determined based on the determination time range. The bulging current coupling feature corresponding to the continuous satisfaction period is written into the differentiable signal timing logic determination. A re-entry determination value is calculated based on the written feature. The current bulging risk value of the activated carbon sample under test is calculated based on the re-entry determination value, and a risk assessment is performed based on the current bulging risk value. By using the above method, a side reaction characteristic sequence is generated around the side reaction evolution of the active sites of activated carbon regeneration during short-term floating charging. This allows for the joint expression of leakage current return, bulging hysteresis, and continuous coupling relationship. An improved novel neural network is introduced to perform one-dimensional spline mapping and fusion of the mapping results, enabling the stable identification of microscopic side reaction evolution. Then, risk assessment is performed based on the current bulging risk value, thereby effectively improving the reliability and accuracy of bulging risk assessment.

[0066] In one specific embodiment, this application provides steps for determining the intensity sequence of side reactions. Please refer to... Figure 4 , Figure 4 This is the second schematic flowchart of the activated carbon regeneration active site swell risk assessment method provided in the embodiments of this application. Step S20 includes steps S201 to S204: Step S201: Based on the improved novel neural network, one-dimensional spline mapping is performed on the leakage current change rate and leakage current change curvature in the side reaction feature sequence, respectively; and based on the improved novel neural network, one-dimensional spline mapping is performed on the bulging change rate, hysteresis correlation feature and bulging current coupling feature in the side reaction feature sequence, respectively.

[0067] It should be noted that for the side reaction feature sequence, including features such as leakage current change rate, leakage current change curvature, bulging change rate, hysteresis correlation features, and bulging current coupling features, these features can be sequentially input into the improved novel neural network according to the sampling time sequence. Operations such as one-dimensional spline mapping and mapping result fusion are performed, ensuring that the input at each sampling time maintains a consistent time correspondence with the leakage current time series data and bulging response data. However, the side reaction feature sequence still represents scattered change information and cannot be directly used as the basis for bulging risk values. Furthermore, the key phenomenon of activated carbon regeneration active sites in short-term float charging is not a strong response at a particular moment, but rather the return of leakage current time series data during decay, while the bulging response data continues to accumulate changes. Therefore, risk evidence relies on the time sequence and the combined changes between multiple features. Common processing methods often involve directly feeding the original time series or feature sequence into a time series network and then representing the risk with the final classification or statistical results. This approach tends to treat the amplitude change of the initial response segment as the primary basis, making it difficult to retain the subsequent rise after decay and the continuous accumulation of bulging in the intermediate representation.

[0068] It should be understood that after each feature in the side reaction feature sequence is sequentially input into the improved novel neural network, the improved novel neural network performs one-dimensional spline mapping on the leakage current change rate and leakage current change curvature in the side reaction feature sequence, and performs one-dimensional spline mapping on the bulging change rate, hysteresis correlation feature, and bulging current coupling feature in the side reaction feature sequence, so that the above features are transformed from original numerical changes into nonlinear responses reflecting the dynamics of the side reaction. The physical relationship corresponding to the above one-dimensional spline mapping in this scenario is as follows: the leakage current change rate and leakage current change curvature reflect whether the side reaction changes from decay to reversion; the hysteresis correlation feature reflects the time delay between the leakage current and the bulging response; and the bulging current coupling feature reflects whether the bulging response continuously follows the leakage current change.

[0069] Step S202: Constrain the current contribution of the leakage current related feature mapping result at each sampling time according to the swelling current coupling feature mapping result, and fuse the constrained leakage current related feature mapping result, swelling change rate mapping result, hysteresis related feature mapping result and swelling current coupling feature mapping result in the same main trunk.

[0070] Understandably, after constraining the current contribution of the leakage current-related feature mapping results at each sampling time based on the bulging current coupling feature mapping results, the improved novel neural network fuses the constrained leakage current-related feature mapping results, bulging rate of change mapping results, hysteresis correlation feature mapping results, and bulging current coupling feature mapping results in the same backbone. This allows the change information in the leakage current time series data and the cumulative information in the bulging response data to form a continuous expression on a unified time axis, rather than forming independent judgments. Here, the current contribution refers to the degree of influence of the leakage current-related feature mapping results when forming the fused result at the current sampling time.

[0071] Step S203: In the same backbone, the fusion results are continuously transmitted according to the sampling time sequence, and the continuous strength change of the side reaction state is calculated based on the continuous representation generated after transmission.

[0072] It should be understood that after obtaining the fusion result, the fusion result is continuously transmitted in the same backbone of the improved new neural network according to the sampling time sequence, so that the fusion results of earlier sampling time and later sampling time are continuously connected on the same time axis to generate a continuous representation corresponding to each sampling time. Based on this continuous representation, the continuous strength change of the side reaction state is calculated time by time, so that the side reaction state at each sampling time simultaneously corresponds to the decay change, the re-rise change and the cumulative change in the bulging response data in the leakage current time series data.

[0073] Step S204: Generate a side reaction intensity value based on the continuous changes in the strength of the side reaction state, and arrange each side reaction intensity value according to the time sequence of the side reaction characteristic sequence to obtain a side reaction intensity sequence.

[0074] Understandably, after generating the side reaction intensity values, the side reaction intensity values ​​corresponding to each sampling time are arranged in chronological order according to the side reaction characteristic sequence to generate the side reaction intensity sequence. This side reaction intensity sequence is not a single label, nor is it a repetitive representation of the original response amplitude; rather, it provides a continuous change in the strength of the side reaction state at each time step. The side reaction intensity sequence can retain "attenuation," "rebound," and "persistence" within the same intermediate object, becoming a direct basis for determining the apparent stability threshold, the re-rise threshold, and the duration threshold. Without the side reaction intensity sequence, the various features in the side reaction characteristic sequence remain dispersed, and therefore cannot simultaneously represent the determination of the recurrence time interval.

[0075] It should be noted that the reference Figure 5 , Figure 5This is a thermodynamic diagram illustrating the intensity of side reactions. Specifically, under the same float charge condition, the intensity of side reactions / bulging of multiple samples over time is arranged into a two-dimensional thermogram. The horizontal axis represents the sampling time t1 to t6, and the vertical axis represents the sample or batch number (sample 1 to sample 10). Each grid corresponds to the intensity of the side reaction of a specific sample at a certain time, with gray levels ranging from light to dark indicating increasing intensity. By observing the color changes along the row direction, the start-up time, growth rate, and stabilization of the side reaction in a single sample can be determined. By observing the distribution along the column direction, the differences between different samples at the same time point can be compared, identifying abnormal batches or outlier samples.

[0076] This embodiment uses the improved novel neural network to perform one-dimensional spline mapping on the leakage current change rate and leakage current change curvature in the side reaction feature sequence, and also uses the improved novel neural network to perform one-dimensional spline mapping on the bulging change rate, hysteresis correlation feature, and bulging current coupling feature in the side reaction feature sequence. The current contribution of the leakage current correlation feature mapping result at each sampling time is constrained based on the bulging current coupling feature mapping result, and the constrained leakage current correlation feature mapping result, bulging change rate mapping result, hysteresis correlation feature mapping result, and bulging current coupling feature mapping result are fused in the same backbone. In the same backbone, the fused result is continuously transmitted according to the sampling time order, and the continuous strength change of the side reaction state is calculated based on the continuous representation generated after transmission. The side reaction intensity value is generated based on the continuous strength change of the side reaction state, and the various side reaction intensity values ​​are arranged according to the time order of the side reaction feature sequence to obtain the side reaction intensity sequence. By introducing an improved novel neural network, one-dimensional spline mapping is performed on each feature in the side reaction feature sequence. After the feature mapping results are fused in the same backbone, they need to be continuously transmitted in the same backbone to keep the fusion results of earlier and later sampling times connected on the same time axis. The side reaction intensity sequence is generated by arranging them, which can effectively improve the accuracy of generating the side reaction intensity sequence.

[0077] The following describes the apparatus for assessing the bulging risk of activated carbon regenerated active sites provided in this application. The apparatus described below corresponds to the method described above for assessing the bulging risk of activated carbon regenerated active sites. Please refer to... Figure 6 , Figure 6 This is a schematic diagram of the module structure of the activated carbon regeneration active site swelling risk assessment device provided in this application embodiment, including: The generation module T10 is used to generate a side reaction characteristic sequence based on the leakage current time series data and bulging response data of the activated carbon sample to be tested.

[0078] The mapping module T20 is used to perform one-dimensional spline mapping on each reaction feature in the side reaction feature sequence based on the improved novel neural network, and to fuse the leakage current related feature mapping results and the swelling current coupling feature mapping results in the same trunk to obtain the side reaction intensity sequence.

[0079] The determination module T30 is used to construct a differentiable signal timing logic determination based on the side reaction intensity sequence, and to constrain the determination start point and determination end point of the side reaction intensity sequence according to the test duration constraint, and to determine the determination time range of the differentiable signal timing logic determination based on the constrained determination start point and determination end point.

[0080] The determining module T30 is further configured to determine a continuous satisfaction period based on the determination time range, write the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination, and calculate the re-entry determination value based on the differentiable signal timing logic determination after writing the feature.

[0081] The evaluation module T40 is used to calculate the current bulging risk value of the activated carbon sample to be tested based on the reentry determination value, and to perform a risk assessment based on the current bulging risk value.

[0082] This embodiment generates a side reaction feature sequence based on the leakage current time-series data and bulging response data of the activated carbon sample under test. A novel, improved neural network is used to perform one-dimensional spline mapping on each reaction feature in the side reaction feature sequence. The leakage current-related feature mapping results and the bulging current coupling feature mapping results are fused in the same backbone to obtain a side reaction intensity sequence. A differentiable signal timing logic determination is constructed based on the side reaction intensity sequence. The determination start and end points of the side reaction intensity sequence are constrained according to the test duration. The determination time range of the differentiable signal timing logic determination is determined based on the constrained determination start and end points. A continuous satisfaction period is determined based on the determination time range. The bulging current coupling feature corresponding to the continuous satisfaction period is written into the differentiable signal timing logic determination. A re-entry determination value is calculated based on the written feature. The current bulging risk value of the activated carbon sample under test is calculated based on the re-entry determination value, and a risk assessment is performed based on the current bulging risk value. By using the above method, a side reaction characteristic sequence is generated around the side reaction evolution of the active sites of activated carbon regeneration during short-term floating charging. This allows for the joint expression of leakage current return, bulging hysteresis, and continuous coupling relationship. An improved novel neural network is introduced to perform one-dimensional spline mapping and fusion of the mapping results, enabling the stable identification of microscopic side reaction evolution. Then, risk assessment is performed based on the current bulging risk value, thereby effectively improving the reliability and accuracy of bulging risk assessment.

[0083] It is understood that the detailed functional implementation of each of the above modules can be found in the description of the aforementioned method embodiments, and will not be repeated here.

[0084] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.

[0085] Based on the methods in the above embodiments, this application provides an electronic device, please refer to... Figure 7 , Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0086] It should be noted that the system may include: a processor 10, a communications interface 20, a memory 30, and a communication bus 40. The processor 10, communications interface 20, and memory 30 communicate with each other via the communication bus 40. The processor 10 can invoke logical instructions stored in the memory 30 to execute the methods described in the above embodiments.

[0087] Furthermore, the logical instructions in the aforementioned memory 30 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0088] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0089] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0090] It is understood that the processor in the embodiments of this application can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, field-programmable gate arrays, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0091] The method steps in this application embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disks, portable hard disks, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor.

[0092] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application. Those skilled in the art will readily understand that the above descriptions are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for evaluating the risk of swelling of active sites of activated carbon, characterized by, include: A side reaction characteristic sequence is generated based on the leakage current time series data and bulging response data of the activated carbon sample to be tested; Based on an improved novel neural network, one-dimensional spline mapping is performed on each reaction feature in the side reaction feature sequence, and the leakage current related feature mapping result and the swelling current coupling feature mapping result are fused in the same main branch to obtain the side reaction intensity sequence. Based on the side reaction intensity sequence, a differentiable signal timing logic is constructed, and the starting point and ending point of the determination of the side reaction intensity sequence are constrained according to the test duration constraint. Based on the constrained starting point and ending point, the determination time range of the differentiable signal timing logic is determined. The continuous satisfaction period is determined according to the determination time range. The swelling current coupling feature corresponding to the continuous satisfaction period is written to the differentiable signal timing logic determination. The re-entry determination value is calculated according to the differentiable signal timing logic determination after writing the feature. The current bulging risk value of the activated carbon sample to be tested is calculated based on the reentry determination value, and a risk assessment is performed based on the current bulging risk value.

2. The bulge risk assessment method of claim 1, wherein, The step of generating a side reaction characteristic sequence based on the leakage current time-series data and bulging response data of the activated carbon sample to be tested includes: The activated carbon sample to be tested is subjected to constant pressure floating charge according to the preset temperature and preset floating charge voltage to form the initial state of data collection. Leakage current is continuously collected from the starting state of acquisition according to a preset sampling interval, and leakage current timing data is generated based on the leakage current at each sampling time. Bulging response is also collected synchronously from the starting state of acquisition according to a preset sampling interval, and bulging response data is generated based on the bulging response at each sampling time. The leakage current change rate and leakage current change curvature are calculated based on the leakage current time series data, and the bulging change rate is calculated based on the bulging response data. The correlation degree of the rate of change at each time offset position is calculated based on the leakage current change rate and the bulging change rate, and the hysteresis correlation characteristics are determined based on the maximum rate of change correlation degree and its corresponding time offset position. The proportional relationship between the bulging response and the leakage current change at each sampling time is calculated based on the hysteresis correlation characteristics, the leakage current change rate, and the bulging change rate. The bulging current coupling characteristics corresponding to the leakage current change rate are then determined based on the proportional relationship. A side reaction feature sequence is generated based on the leakage current change rate, the leakage current change curvature, the bulging change rate, the hysteresis correlation feature, and the bulging current coupling feature.

3. The bulge risk assessment method of claim 1, wherein, The step of performing one-dimensional spline mapping on each reaction feature in the side reaction feature sequence based on the improved novel neural network, and fusing the leakage current-related feature mapping results and the swelling current coupling feature mapping results in the same backbone to obtain the side reaction intensity sequence includes: The improved novel neural network performs one-dimensional spline mapping on the leakage current change rate and leakage current change curvature in the side reaction feature sequence, respectively; and the improved novel neural network performs one-dimensional spline mapping on the bulging change rate, hysteresis correlation feature and bulging current coupling feature in the side reaction feature sequence, respectively. The current contribution of the leakage current related feature mapping result at each sampling time is constrained based on the swelling current coupling feature mapping result, and the constrained leakage current related feature mapping result, swelling change rate mapping result, hysteresis related feature mapping result and swelling current coupling feature mapping result are fused in the same backbone; Within the same backbone, the fusion results are continuously transmitted according to the sampling time sequence, and the continuous strength changes of the side reaction state are calculated based on the continuous representation generated after transmission. The side reaction intensity values ​​are generated based on the continuous changes in the strength of the side reaction states, and the side reaction intensity values ​​are arranged in chronological order according to the side reaction characteristic sequence to obtain the side reaction intensity sequence.

4. The bulge risk assessment method of claim 1, wherein, The steps of determining a continuous satisfaction period based on the determination time range, writing the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination, and calculating the re-entry determination value based on the differentiable signal timing logic determination after writing the feature include: Multiple candidate apparent stability thresholds are set within the determination time range, and the side reaction intensity sequence is compared with the multiple candidate apparent stability thresholds respectively. The apparent stable time period is determined based on the first comparison result, and the recurrence time interval is determined based on the apparent stable time period and the termination time of the determination time range. Multiple candidate re-emergence thresholds are set within the re-emergence time interval, and the side reaction intensity sequence is compared with the multiple candidate re-emergence thresholds respectively. Based on the second comparison result, a re-rise time period that connects with the apparent stable time period is determined, and multiple candidate duration thresholds are set based on the re-rise time period and the continuous holding length within the re-occurrence time interval. The continuous satisfaction time period is determined based on the multiple candidate duration thresholds, and the bulging current coupling feature corresponding to the continuous satisfaction time period is written into the differentiable signal timing logic determination. The re-entry determination value is calculated based on the differentiable signal timing logic determination after the feature is written.

5. The bulge risk assessment method of claim 4, wherein, The steps of determining a continuous satisfaction time period based on the multiple candidate duration thresholds, writing the bulging current coupling feature corresponding to the continuous satisfaction time period to the differentiable signal timing logic determination, and calculating the re-entry determination value based on the differentiable signal timing logic determination after writing the feature include: The continuous duration within the recurrence time interval is compared with the multiple candidate duration thresholds, and the continuous satisfaction time period is determined based on the third comparison result; The bulging current coupling characteristic corresponding to the continuous satisfaction time period is written into the differentiable signal timing logic determination; Traverse each time segment of the time sequence logic determination of the differentiable signal after writing the features; The smoothness robustness of each time segment is calculated, and the results of each smoothness robustness calculation are aggregated to obtain the re-entry determination value.

6. The risk of ballooning evaluation method according to any one of claims 1 to 5, characterized in that, The steps of calculating the current bulging risk value of the activated carbon sample to be tested based on the reentry determination value, and performing a risk assessment based on the current bulging risk value, include: Calculate the difference between the reentry determination value and the preset risk threshold; The deviation direction code is determined based on the sign of the difference, and the absolute magnitude is calculated based on the re-entry determination value and the preset risk threshold. A risk comparison result is generated based on the absolute magnitude and the deviation direction encoding, and a risk offset is determined based on the risk comparison result; The reentry determination value is continuously mapped based on the risk comparison results and the risk offset to calculate the current bulging risk value of the activated carbon sample to be tested. The current bulging risk value is compared with the preset risk threshold, and a risk assessment is performed based on the fourth comparison result.

7. A device for assessing the bulging risk of active sites during activated carbon regeneration, characterized in that, include: The generation module is used to generate a sequence of side reaction characteristics based on the leakage current time-series data and bulging response data of the activated carbon sample to be tested; The mapping module is used to perform one-dimensional spline mapping on each reaction feature in the side reaction feature sequence based on an improved novel neural network, and to fuse the leakage current related feature mapping results and the swelling current coupling feature mapping results in the same main branch to obtain the side reaction intensity sequence. The determination module is used to construct a differentiable signal timing logic determination based on the side reaction intensity sequence, and to constrain the determination start point and determination end point of the side reaction intensity sequence according to the test duration constraint, and to determine the determination time range of the differentiable signal timing logic determination based on the constrained determination start point and determination end point; The determining module is further configured to determine a continuous satisfaction period based on the determination time range, write the bulging current coupling feature corresponding to the continuous satisfaction period to the differentiable signal timing logic determination, and calculate the re-entry determination value based on the differentiable signal timing logic determination after writing the feature. The evaluation module is used to calculate the current bulging risk value of the activated carbon sample to be tested based on the reentry determination value, and to perform a risk assessment based on the current bulging risk value.

8. An electronic device, comprising: include: At least one memory for storing computer programs; At least one processor is configured to execute a program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. When the computer program is run on the processor, it causes the processor to perform the method as described in any one of claims 1-6.

10. A computer program product, characterised in that, When the computer program product is run on a processor, the processor causes the processor to perform the method as described in any one of claims 1-6.

Images