A Monitoring Method for the Preparation Process of a Carbon Anode Material for a Lithium Battery in a New Energy Vehicle

Abstract

The present invention relates to the technical field of data processing, and particularly relates to a method for monitoring the preparation process of a carbon negative electrode material for a lithium battery of a new energy vehicle. The method obtains a temperature data sequence in a reaction chamber; performs time series decomposition on the temperature data sequence to obtain a seasonal component change curve, and divides the seasonal component change curve into multiple sub-curves; obtains the similarity between two adjacent sub-curves to obtain a similarity sequence, performs adaptive smoothing processing on the similarity sequence to obtain a smoothed similarity sequence, performs subsequence division on the smoothed similarity sequence to obtain a target subsequence; analyzes the similarity change factor of the target subsequence, and obtains an adaptive gamma parameter according to the similarity change factor; based on the adaptive gamma parameter, uses a support vector regression model to obtain a temperature prediction value in the future time, and adjusts the temperature in the reaction chamber according to the temperature prediction value, improving the accuracy of temperature prediction and controlling the temperature in the reaction chamber more timely.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to a method for monitoring the preparation process of a carbon negative electrode material for a lithium battery of a new energy vehicle. Background Art

[0002] The carbon negative electrode material of a lithium battery for a new energy vehicle is a conductive material used for the negative electrode of a lithium-ion battery, mainly composed of graphite materials (such as natural graphite, artificial graphite, hard carbon, soft carbon, etc.). Its main function is to provide a place for lithium ions to be embedded or de-embedded during the charge and discharge process. The preparation process of the carbon negative electrode material of a lithium battery for a new energy vehicle generally includes steps such as raw material selection, pretreatment, carbonization, modification, surface treatment, and final product packaging. During the preparation process of the carbon negative electrode material, the temperature control of the reaction chamber is crucial, which can directly affect the crystallinity of the material, etc., thereby affecting the stability of various physical properties of the carbon negative electrode material. Therefore, during the preparation process of the lithium battery carbon negative electrode material, when it is monitored that the temperature in the reaction chamber is lower or higher than the set temperature threshold, the temperature in the reaction chamber can be warned and regulated to ensure that the reaction chamber is at the optimal temperature. However, this method ignores the dynamic changes in the temperature in the reaction chamber and the characteristics such as the delay in temperature monitoring, resulting in possible under-compensation or over-compensation of the temperature in the reaction chamber, making the temperature stability in the reaction chamber worse, and further affecting the efficiency and quality of the preparation of the carbon negative electrode material, etc.

[0003] Currently, for the temperature regulation during the preparation process of the lithium battery carbon negative electrode material, first, the support vector regression (SVR) algorithm is used to predict the temperature in the reaction chamber during the preparation process of the lithium battery carbon negative electrode material to obtain the predicted temperature, and then the temperature in the reaction chamber is regulated in real time according to the predicted temperature. However, during the temperature prediction process using the support vector regression (SVR) algorithm, a suitable gamma parameter needs to be selected to control the range of action of the kernel function in the support vector regression (SVR) algorithm to ensure the accuracy of the temperature prediction. Since there may be multiple reaction states and stages during the preparation process of the lithium battery carbon negative electrode material, and the change pattern of the temperature in its reaction chamber over time may also be different in each stage, it is necessary to adaptively adjust the gamma parameter according to the real-time collected temperature data to ensure the accuracy of the temperature prediction of the reaction chamber by the support vector regression (SVR) algorithm. Summary of the Invention

[0004] In view of this, an embodiment of the present invention provides a method for monitoring the preparation process of a carbon negative electrode material for a lithium battery of a new energy vehicle to solve the problem of how to improve the accuracy of predicting the temperature in the reaction chamber using the support vector regression (SVR) algorithm.

[0005] In an embodiment of the present invention, a method for monitoring the preparation process of a carbon negative electrode material for a new energy vehicle lithium battery is provided. The method includes the following steps:

[0006] During the preparation process of the carbon negative electrode material for the lithium battery, obtain the temperature data sequence in the reaction chamber;

[0007] Perform STL time series decomposition on the temperature data sequence to obtain the seasonal component change curve constructed by the seasonal components of each temperature data in the temperature data sequence. According to the peak points on the seasonal component change curve, divide the seasonal component change curve into multiple sub-curves;

[0008] Obtain the similarity between two adjacent sub-curves to obtain a similarity sequence. Perform adaptive smoothing processing on the similarity sequence to obtain a smoothed similarity sequence. Divide the smoothed similarity sequence into at least one subsequence, and use the last subsequence as the subsequence corresponding to the last temperature data in the temperature data sequence, denoted as the target subsequence;

[0009] Analyze the similarity change factor of the target subsequence. According to the similarity change factor and the element difference between the target subsequence and each subsequence, obtain the adaptive gamma parameter in the support vector regression model;

[0010] Based on the adaptive gamma parameter, use the support vector regression model to predict the temperature data sequence to obtain the temperature prediction value in the future time, and adjust the temperature in the reaction chamber in real time according to the temperature prediction value.

[0011] Preferably, the obtaining the similarity between two adjacent sub-curves includes:

[0012] Obtain the duration corresponding to each sub-curve respectively, and obtain the change degree of the corresponding sub-curve according to the difference between the maximum value and the minimum value on each sub-curve;

[0013] For any two adjacent sub-curves, obtain the absolute value of the duration difference and the absolute value of the change degree difference between the two adjacent sub-curves, and perform inverse proportional normalization on the product of the absolute value of the duration difference and the absolute value of the change degree difference to obtain the similarity between the two adjacent sub-curves.

[0014] Preferably, the performing adaptive smoothing processing on the similarity sequence to obtain a smoothed similarity sequence includes:

[0015] Perform a first-order difference on the similarity sequence to obtain a corresponding difference sequence, and remove the first similarity in the similarity sequence to obtain a target similarity sequence;

[0016] For any similarity in the target similarity sequence, according to the position number of the any similarity in the target similarity sequence, obtain the difference value corresponding to the same position number in the difference sequence, denoted as the target difference value. In the difference sequence, obtain at least one neighborhood difference value of the target difference value, calculate the absolute value of the difference between the target difference value and each of the neighborhood difference values respectively, obtain the average absolute difference value, perform normalization processing on the average absolute difference value to obtain the corresponding normalized value, and round up the product of the normalized value and the preset hyperparameter to obtain the smoothing window size of the any similarity;

[0017] Obtain the smoothing window size of each similarity in the target similarity sequence. According to the smoothing window size of each similarity in the target similarity sequence, perform median filtering on the similarity sequence to obtain the filtered similarity sequence, denoted as the smoothed similarity sequence.

[0018] Preferably, the dividing the smoothed similarity sequence into at least one subsequence includes:

[0019] Obtain the similarity corresponding to the first derivative being 0 and the second derivative not being 0 in the smoothed similarity sequence as the division point. If the number of division points is 0, then take the smoothed similarity sequence as one subsequence; if the number of division points is not 0, then divide the smoothed similarity sequence into at least two subsequences according to all the division points.

[0020] Preferably, the analyzing the similarity change factor of the target subsequence includes:

[0021] Calculate the ratio between two adjacent similarities in the target subsequence to obtain the average ratio value, and obtain the similarity change factor of the target subsequence according to the absolute value of the difference between the average ratio value and the constant 1.

[0022] Preferably, the obtaining the adaptive gamma parameter in the support vector regression model according to the similarity change factor and the element difference between the target subsequence and each of the subsequences includes:

[0023] According to the number of elements in each subsequence, calculate the average number of elements, calculate the ratio between the average number of elements and the number of elements in the target subsequence, perform normalization processing on the product of the reciprocal of the similarity change factor and the ratio to obtain the corresponding normalized value, and obtain the adaptive gamma parameter in the support vector regression model according to the product of the normalized value and the preset conversion hyperparameter of the gamma mapping.

[0024] Preferably, the adjusting the temperature in the reaction chamber in real time according to the temperature prediction value includes:

[0025] Obtain the maximum temperature threshold and the minimum temperature threshold of the reaction chamber during the preparation process of the carbon anode material for lithium batteries. When the temperature prediction value is greater than the maximum temperature threshold, heat dissipation treatment is performed on the reaction chamber; when the temperature prediction value is less than the minimum temperature threshold, temperature compensation treatment is performed on the reaction chamber.

[0026] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0027] In the preparation process of the carbon anode material for lithium batteries of the present invention, a temperature data sequence in the reaction chamber is obtained; the STL time series decomposition is performed on the temperature data sequence to obtain a seasonal component change curve constructed by the seasonal components of each temperature data in the temperature data sequence. According to the peak points on the seasonal component change curve, the seasonal component change curve is divided into multiple sub-curves; the similarity between two adjacent sub-curves is obtained to obtain a similarity sequence, the similarity sequence is adaptively smoothed to obtain a smoothed similarity sequence, the smoothed similarity sequence is divided into at least one subsequence, and the last subsequence is used as the subsequence corresponding to the last temperature data in the temperature data sequence, denoted as the target subsequence; the similarity change factor of the target subsequence is analyzed, and according to the similarity change factor and the element difference between the target subsequence and each subsequence, an adaptive gamma parameter in the support vector regression model is obtained; based on the adaptive gamma parameter, the support vector regression model is used to predict the temperature data sequence to obtain a temperature prediction value in the future time, and the temperature in the reaction chamber is adjusted in real time according to the temperature prediction value. Among them, when using the SVR support vector regression model to predict the temperature in the reaction chamber during the preparation process of the carbon anode material for lithium batteries, the STL decomposition is performed on the temperature data sequence in time series to obtain a seasonal component change curve, which is used to characterize the periodic change of temperature. Then, according to the peak points of the seasonal component change curve, the seasonal component change curve is divided into multiple sub-curves, and one sub-curve represents a temperature change mode. Furthermore, by calculating the similarity between two adjacent sub-curves, the difference in the change law and mode between the latest temperature data and the historical temperature data is analyzed. The greater the difference, the smaller the reference value for subsequent temperature prediction. Therefore, according to the local fluctuation of the similarity sequence corresponding to the temperature data sequence, an adaptive gamma parameter that is more suitable for this scenario is obtained. Furthermore, when using the SVR support vector regression model with the adaptive gamma parameter to predict the temperature of the reaction chamber, the prediction accuracy can be improved, and at the same time, the temperature control in the reaction chamber according to the prediction result is also more timely and accurate. Description of the Drawings

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for use in the embodiments or the description of the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0029] Figure 1 It is a flowchart of a method for monitoring the preparation process of a carbon negative electrode material for a new energy vehicle lithium battery provided in Embodiment 1 of the present invention. Specific embodiments

[0030] The following will describe in detail the embodiments of the present disclosure, and the examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0031] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above accompanying drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are only examples of devices and methods consistent with some aspects of the present disclosure.

[0032] To illustrate the technical solutions of the present invention, the following will be described through specific embodiments.

[0033] The specific scenario targeted by the present invention is as follows: When using a support vector regression model (SVR) to predict the temperature in the reaction chamber during the preparation process of the carbon negative electrode material for the lithium battery, according to the change pattern and law of the temperature in the reaction chamber in time series, a suitable gamma parameter is obtained to control the action range of the kernel function of the SVR model, thereby improving the accuracy of temperature prediction.

[0034] See Figure 1 , which is a flowchart of a method for monitoring the preparation process of a carbon negative electrode material for a new energy vehicle lithium battery provided in Embodiment 1 of the present invention. As Figure 1 shown, the method may include:

[0035] Step S101, during the preparation process of the carbon negative electrode material for the lithium battery, obtain the temperature data sequence in the reaction chamber.

[0036] In the process of preparing the carbon anode material for lithium batteries, temperature data in the reaction chamber is collected through temperature sensors arranged in the reaction chamber, forming a sequential temperature data series. The sampling frequency and the acquisition duration of the temperature data series are not restricted, and implementers can set them according to the implementation scenario.

[0037] Step S102: Perform STL time series decomposition on the temperature data series to obtain the seasonal component change curve constructed by the seasonal components of each temperature data in the temperature data series. According to the peak points on the seasonal component change curve, divide the seasonal component change curve into multiple sub-curves.

[0038] In the process of preparing the carbon anode material for lithium batteries of new energy vehicles, some chemical reactions and the influence of the preparation operation process are usually involved, which will cause certain dynamic changes in the temperature in the reaction chamber. Therefore, it is necessary to ensure that the temperature in the reaction chamber during the preparation process of the carbon anode material for lithium batteries is in a relatively stable and appropriate state to ensure the quality of processes such as crystallization of the carbon anode material for lithium batteries. Since the temperature in the reaction chamber is dynamically changing during the monitoring process, and the temperature monitoring and transmission have a certain time continuity, while the traditional monitoring method may ignore the temperature changes in the reaction chamber in a future time period, resulting in overcompensation or undercompensation when monitoring and controlling the temperature in the reaction chamber, thereby affecting the accuracy of the temperature control in the reaction chamber during the preparation process of the carbon anode material.

[0039] Therefore, the prior art uses a support vector regression (SVR) model to predict the temperature in the reaction chamber. However, when using the SVR model for temperature prediction, the magnitude of the gamma parameter will affect the control range of the kernel function in the SVR model, thereby affecting the prediction accuracy of the SVR model. And due to the dynamic change of the temperature in the reaction chamber, the temperature change patterns and rules in different stages may be different. Therefore, the embodiments of the present invention need to adjust the gamma parameter of the SVR model according to the dynamic change characteristics of the obtained temperature data series to improve the accuracy of the temperature prediction in the reaction chamber.

[0040] First, perform STL time series decomposition on the temperature data series to obtain the seasonal component change curve constructed by the seasonal components of each temperature data in the temperature data series. The horizontal axis of the seasonal component change curve is time, and the vertical axis is the seasonal component of each temperature data in the temperature data series. Among them, STL time series decomposition is a well-known technology and will not be elaborated here in detail.

[0041] Then, use the AMPD algorithm (Amplitude Modulation Peak Detection) to obtain the peak points in the seasonal component change curve. Take the peak points as the division points to divide the seasonal component change curve into multiple sub-curves, which are used to subsequently judge the difference between the new period and the historical period of the temperature by analyzing the similarity between the sub-curves. Among them, one sub-curve represents the temperature change within one period, and it is default that the last temperature data in the temperature data sequence belongs to the newly obtained temperature data, that is, the new temperature data, and the sub-curve corresponding to the new temperature data also represents a new period.

[0042] Step S103: Obtain the similarity between two adjacent sub-curves to get a similarity sequence. Perform adaptive smoothing processing on the similarity sequence to obtain a smoothed similarity sequence. Divide the smoothed similarity sequence into at least one sub-sequence, and take the last sub-sequence as the sub-sequence corresponding to the last temperature data in the temperature data sequence, denoted as the target sub-sequence.

[0043] After dividing the seasonal component change curve into multiple sub-curves, obtain the similarity between two adjacent sub-curves to get a similarity sequence. The greater the similarity, the more similar the temperature changes between the two sub-curves, and the closer the corresponding temperature change rules and patterns are. Among them, the method for obtaining the similarity between two adjacent sub-curves is as follows:

[0044] Obtain the duration corresponding to each of the sub-curves respectively, and according to the difference between the maximum value and the minimum value on each sub-curve, obtain the change degree of the corresponding sub-curve;

[0045] For any two adjacent sub-curves, obtain the absolute value of the duration difference and the absolute value of the change degree difference between the two adjacent sub-curves, and perform inverse proportional normalization on the product of the absolute value of the duration difference and the absolute value of the change degree difference to obtain the similarity between the two adjacent sub-curves.

[0046] In an embodiment, taking the i-th sub-curve and the (i - 1)-th sub-curve as an example, the calculation expression for the similarity between the i-th sub-curve and the (i - 1)-th sub-curve is:

[0047]

[0048] Among them, represents the similarity between the i-th sub-curve and the (i - 1)-th sub-curve, exp() represents the exponential function with the natural constant as the base, represents the duration corresponding to the i-th sub-curve, represents the difference between the maximum value and the minimum value on the i-th sub-curve, represents the duration corresponding to the (i - 1)-th sub-curve, represents the difference between the maximum and minimum values on the (i - 1)-th sub-curve, and | | represents the absolute value symbol.

[0049] It should be noted that The larger the value of, the greater the duration difference between the two sub-curves, the greater the corresponding period duration difference between the two sub-curves, and the smaller the similarity between the two sub-curves; The smaller the value of, it indicates that the change ranges between the two sub-curves are closer, and the corresponding similarity between the two sub-curves is greater.

[0050] Similarly, the similarity between every two adjacent sub-curves can be obtained to get a similarity sequence in time series, which is used to characterize the similarity between adjacent periods in the temperature data sequence. When the similarity sequence is relatively stable, that is, there is no obvious change amplitude and trend, it indicates that the change rules and patterns of the latest temperature data and the historical temperature data are relatively similar. The historical temperature data has a greater reference value for future temperature prediction. Correspondingly, the larger the gamma parameter should be selected, so as to control a larger range of action when the kernel function of the SVR model is used for prediction, and then use more historical data as a reference to improve the accuracy of future temperature prediction; on the contrary, a smaller gamma parameter needs to be selected to avoid a larger range of action of the kernel function, resulting in more historical data participating in the prediction of future temperature, thus ignoring the pattern of the latest temperature change in the reaction chamber, and then causing the predicted temperature to be too smooth and reducing the accuracy of future temperature prediction.

[0051] Therefore, in the embodiments of the present invention, an adaptive gamma parameter is obtained based on the fluctuation change of the similarity sequence. However, before obtaining the adaptive gamma parameter, in order to avoid the influence of local fluctuations of the similarity sequence, the similarity sequence is first subjected to adaptive smoothing processing to obtain a smoothed similarity sequence. The method of adaptive smoothing processing is as follows:

[0052] Perform a first-order difference on the similarity sequence to obtain a corresponding difference sequence, and remove the first similarity in the similarity sequence to obtain a target similarity sequence;

[0053] For any similarity in the target similarity sequence, according to the position number of the any similarity in the target similarity sequence, obtain the difference value corresponding to the same position number in the difference sequence, denoted as the target difference value. Obtain at least one neighborhood difference value of the target difference value in the difference sequence, calculate the absolute value of the difference between the target difference value and each neighborhood difference value, obtain the average absolute difference value, perform normalization processing on the average absolute difference value to obtain a corresponding normalized value, and round up the product of the normalized value and a preset hyperparameter to obtain the smoothing window size of the any similarity;

[0054] Obtain the smoothing window size for each similarity in the target similarity sequence. According to the smoothing window size of each similarity in the target similarity sequence, perform median filtering on the similarity sequence to obtain a filtered similarity sequence, denoted as the smoothed similarity sequence.

[0055] In one embodiment, perform a first-order difference on the similarity sequence to obtain a difference sequence. Considering that the first similarity in the similarity sequence belongs to the first data and has no impact on the overall change trend of this similarity sequence, and in order to monitor the change trend and pattern of the similarity sequence, therefore, when performing median filtering on the similarity sequence, the first similarity does not need to be smoothed and remains unchanged. Only adaptively obtain a smoothing window for each similarity in the similarity sequence except the first similarity for median filtering.

[0056] Take the sequence after removing the first similarity from the similarity sequence as the target similarity sequence. It should be noted that the length of the difference sequence is equal to the length of the target similarity sequence, and the j-th difference value in the difference sequence corresponds to the adjacent change difference of the j-th similarity in the target similarity sequence. Therefore, taking the j-th similarity as an example, obtain the j-th difference value in the difference sequence as the difference value corresponding to the j-th similarity, and then, centered on the j-th difference value in the difference sequence, select the 5 difference values before and after it as neighborhood difference values. According to the difference between the j-th difference value and each neighborhood difference value, the calculation expression for the smoothing window size of the j-th similarity is:

[0057]

[0058] Where, represents the smoothing window size of the j-th similarity, norm() represents the normalization function, represents the number of neighborhood difference values of the difference value corresponding to the j-th similarity in the difference sequence, represents the difference value corresponding to the j-th similarity in the difference sequence, represents the y-th neighborhood difference value of the difference value corresponding to the j-th similarity in the difference sequence, || represents the absolute value symbol, and K represents a hyperparameter.

[0059] It should be noted that the value of K is 11, and there is no restriction here; is used to characterize the overall difference between the j-th difference value and the neighborhood difference values. The larger the overall difference, the worse the smoothness of the change between the j-th similarity in the target similarity sequence and the similarities in its front and back local ranges, and the larger the window for filtering and smoothing the j-th similarity.

[0060] Similarly, the smoothing window size of each similarity in the target similarity sequence can be obtained, and then, based on the smoothing window size of each similarity in the target similarity sequence, median filtering is performed on the similarity sequence to obtain a filtered similarity sequence, denoted as the smoothed similarity sequence. Median filtering belongs to the prior art and will not be elaborated here.

[0061] To more intuitively present the change trend of the new temperature data (i.e., the last temperature data) in the temperature data sequence, the smoothed similarity sequence is divided into subsequences, and the last subsequence obtained by the division is used as the subsequence corresponding to the new temperature data, denoted as the target subsequence. Among them, the method for dividing the smoothed similarity sequence into subsequences is as follows:

[0062] In the smoothed similarity sequence, the similarity corresponding to the first derivative being 0 and the second derivative not being 0 is obtained as the division point. If the number of division points is 0, the smoothed similarity sequence is used as a subsequence; if the number of division points is not 0, the smoothed similarity sequence is divided into at least two subsequences according to all the division points. Among them, the judgment of the first derivative being 0 and the second derivative being 0 belongs to the prior art and will not be elaborated here.

[0063] It should be noted that when there is no division point in the smoothed similarity sequence, it indicates that the fluctuation change of the smoothed similarity sequence is relatively stable, and the smoothed similarity sequence is directly used as the target subsequence.

[0064] So far, the similarity subsequence corresponding to the new temperature data, that is, the target subsequence, has been obtained.

[0065] Step S104: Analyze the similarity change factor of the target subsequence, and obtain the adaptive gamma parameter in the support vector regression model according to the similarity change factor and the element difference between the target subsequence and each subsequence.

[0066] If the target subsequence shows an upward or downward trend, it indicates that the change law and pattern of the new temperature data are more different from the change pattern of the historical temperature data, and thus the dependence relationship between the future temperature data prediction and the historical temperature data may be weaker. Therefore, the similarity change factor of the target subsequence is analyzed to detect the change of the new temperature data. The method for obtaining the similarity change factor of the target subsequence is as follows:

[0067] Calculate the ratio between two adjacent similarities in the target subsequence to obtain the average ratio, and obtain the similarity change factor of the target subsequence according to the absolute value of the difference between the average ratio and the constant 1.

[0068] In an embodiment, the calculation expression of the similarity change factor of the target subsequence is:

[0069]

[0070] Among them, represents the similarity change factor of the target subsequence, represents the number of similarities in the target subsequence, represents the r-th similarity in the target subsequence, represents the (r + 1)-th similarity in the target subsequence, 1 represents a constant, and | | represents the absolute value symbol.

[0071] It should be noted that is used to characterize the average ratio between adjacent similarities in the similarity subsequence corresponding to the new temperature data. The less similar this average ratio is to the constant 1, the more severe the upward or downward change trend of the target subsequence may be, and the larger the similarity change factor of the target subsequence.

[0072] Furthermore, by combining the similarity change factor of the target subsequence and the proportional relationship of the number of similarities in each subsequence corresponding to the temperature data sequence, an adaptive gamma parameter of the SVR model is obtained for predicting the temperature data in the future. Among them, the method for obtaining the adaptive gamma parameter is as follows:

[0073] According to the number of elements in each of the said subsequences, the average value of the number of elements is calculated, the ratio between the average value of the number of elements and the number of elements in the target subsequence is calculated, and the product of the reciprocal of the similarity change factor and the ratio is normalized to obtain a corresponding normalized value. According to the product between the normalized value and the preset conversion hyperparameter of the gamma mapping, the adaptive gamma parameter in the support vector regression model is obtained.

[0074] In an embodiment, the calculation expression of the adaptive gamma parameter is:

[0075]

[0076] Among them, represents the adaptive gamma parameter, norm() represents the normalization function, represents the similarity change factor of the target subsequence, represents the number of subsequences, represents the number of elements in the target subsequence, that is, the number of similarities in the target subsequence, represents the average value of the number of elements in all subsequences, represents the preset conversion hyperparameter of the gamma mapping.

[0077] It should be noted that the value of G is 5, and there is no limitation here; The smaller the value, the weaker the dependence of the local change of the new temperature data on the historical temperature data, and the smaller the corresponding gamma parameter value should be; The smaller the value, the faster the local change rate of the new temperature data relative to the local change rate of the historical temperature data, that is, the shorter the time for the seasonal cycle of the new temperature data to exist in the time series. The value of the gamma parameter should be smaller, and more attention should be given to the new temperature data during prediction, so that the dependence on the historical temperature data for future temperature data prediction is smaller.

[0078] Thus, the adaptive gamma parameter for future temperature prediction of the temperature data sequence using the SVR model is obtained.

[0079] Step S105: Based on the adaptive gamma parameter, use the support vector regression model to predict the temperature data sequence to obtain the temperature prediction value within the future time, and adjust the temperature in the reaction chamber in real time according to the temperature prediction value.

[0080] After determining the gamma parameter of the SVR model, the SVR model can be used to predict the temperature data sequence to obtain the temperature prediction value within the future time. The SVR model belongs to the prior art and will not be elaborated here in detail. Usually, the heat treatment temperature of the lithium battery carbon anode material is between 500 degrees and 1000 degrees. Therefore, the maximum temperature threshold of the reaction chamber in the preparation process of the lithium battery carbon anode material is set to 1000 degrees and the minimum temperature threshold is 500 degrees. Considering that the temperature change in the reaction chamber is different in the time series and each preparation stage, in the embodiments of the present invention, the maximum temperature threshold and the minimum temperature threshold are not limited, and the implementer can set them according to the implementation scenario.

[0081] Furthermore, according to the maximum temperature threshold, the minimum temperature threshold and the temperature prediction value, the temperature in the reaction chamber is adjusted in real time: obtain the maximum temperature threshold and the minimum temperature threshold of the reaction chamber in the preparation process of the lithium battery carbon anode material. When the temperature prediction value is greater than the maximum temperature threshold, heat dissipation treatment is performed on the reaction chamber; when the temperature prediction value is less than the minimum temperature threshold, temperature compensation treatment is performed on the reaction chamber.

[0082] In one embodiment, when the temperature prediction value is greater than 1000 degrees, heat dissipation is performed on the temperature in the reaction chamber through a corresponding temperature control device (such as a ventilation and heat dissipation device) to ensure that the temperature in the reaction chamber is below 1000 degrees; when the temperature prediction value is less than 500 degrees, heating treatment is performed on the temperature in the reaction chamber through a temperature control device such as a heating furnace, so that the temperature in the reaction chamber is higher than 500 degrees. Thus, the temperature in the reaction chamber can be dissipated or compensated to maintain the temperature in the reaction chamber in balance and stability.

[0083] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for monitoring the preparation process of carbon negative electrode materials for lithium batteries of new energy vehicles, characterized in that: The method comprises: During the preparation of carbon negative electrode materials for lithium batteries, a temperature data sequence is obtained in the reaction chamber; Performing STL time series decomposition on the temperature data sequence to obtain a seasonal component variation curve constructed by the seasonal component of each temperature data in the temperature data sequence, and dividing the seasonal component variation curve into a plurality of sub-curves according to peak points on the seasonal component variation curve; Acquire the similarity between two adjacent sub-curves to obtain a similarity sequence, perform adaptive smoothing on the similarity sequence to obtain a smoothed similarity sequence, divide the smoothed similarity sequence into at least one sub-sequence, and use the last sub-sequence as the sub-sequence corresponding to the last temperature data in the temperature data sequence, recorded as a target sub-sequence; Calculate the ratio between two adjacent similarities in the target subsequence to obtain a ratio mean, obtain a similarity change factor of the target subsequence according to the absolute value of the difference between the ratio mean and a constant 1, and obtain an adaptive gamma parameter in a support vector regression model according to the similarity change factor and the element difference between the target subsequence and each of the subsequences; Based on the adaptive gamma parameter, the temperature data sequence is predicted using the support vector regression model to obtain a temperature prediction value in the future, and the temperature in the reaction chamber is adjusted in real time according to the temperature prediction value; The step of obtaining an adaptive gamma parameter in a support vector regression model according to the similarity change factor and the element difference between the target subsequence and each of the subsequences includes: According to the number of elements in each of the subsequences, the mean number of elements is calculated, the ratio between the mean number of elements and the number of elements in the target subsequence is calculated, the product between the inverse of the similarity change factor and the ratio is normalized to obtain a corresponding normalized value, and the adaptive gamma parameters in the support vector regression model are obtained according to the product between the normalized value and the preset conversion hyperparameters of the gamma map.

2. The method for monitoring the preparation process of carbon negative electrode materials for lithium batteries of new energy vehicles according to claim 1, characterized in that: The obtaining the similarity between two adjacent sub-curves includes: Obtaining the duration corresponding to each of the sub-curves respectively, and obtaining the degree of change of the corresponding sub-curve according to the difference between the maximum value and the minimum value on each of the sub-curves; For any two adjacent sub-curves, the absolute value of the duration difference and the absolute value of the change degree difference between the two adjacent sub-curves are obtained, and the product of the absolute value of the duration difference and the absolute value of the change degree difference is inversely normalized to obtain the similarity between the two adjacent sub-curves.

3. The method for monitoring the preparation process of carbon negative electrode materials for lithium batteries of new energy vehicles according to claim 1, characterized in that: The step of performing adaptive smoothing on the similarity sequence to obtain a smoothed similarity sequence includes: Performing a first-order difference on the similarity sequence to obtain a corresponding difference sequence, removing the first similarity in the similarity sequence to obtain a target similarity sequence; For any similarity of the target similarity sequence, according to the position number of any similarity in the target similarity sequence, obtain the differential value corresponding to the same position number in the differential sequence, record it as the target differential value, obtain at least one neighborhood differential value of the target differential value in the differential sequence, calculate the absolute value of the difference between the target differential value and each of the neighborhood differential values, obtain the average absolute value of the difference, normalize the average absolute value of the difference, obtain the corresponding normalized value, round up the product of the normalized value and the preset hyperparameter, and obtain the smoothing window size of any similarity; The smoothing window size of each similarity in the target similarity sequence is obtained, and according to the smoothing window size of each similarity in the target similarity sequence, the similarity sequence is subjected to median filtering to obtain a filtered similarity sequence, which is recorded as a smoothed similarity sequence.

4. The method for monitoring the preparation process of carbon negative electrode materials for lithium batteries of new energy vehicles according to claim 1, characterized in that: The dividing the smoothed similarity sequence into at least one subsequence comprises: In the smooth similarity sequence, the similarity corresponding to the first-order derivative being 0 and the second-order derivative being not 0 is obtained as a division point. If the number of the division points is 0, the smooth similarity sequence is taken as a subsequence; if the number of the division points is not 0, the smooth similarity sequence is divided into at least two subsequences according to all the division points.

5. The method for monitoring the preparation process of carbon negative electrode materials for lithium batteries of new energy vehicles according to claim 1, characterized in that: The real-time adjustment of the temperature in the reaction chamber according to the temperature prediction value comprises: The maximum temperature threshold and the minimum temperature threshold of the reaction chamber in the preparation process of the carbon negative electrode material of the lithium battery are obtained. When the temperature prediction value is greater than the maximum temperature threshold, the reaction chamber is subjected to heat dissipation treatment; when the temperature prediction value is less than the minimum temperature threshold, the reaction chamber is subjected to temperature compensation treatment.

Images