Battery SOC estimation method and device, electronic equipment, medium and program product

Abstract

The invention relates to the field of artificial intelligence, in particular to a battery SOC estimation method and device, electronic equipment, a medium and a program product. The method comprises the following steps: acquiring a first parameter of a battery; the first parameter is input into a pre-trained battery SOC model, SOC information is obtained, and the SOC information at least comprises an SOC estimation value; wherein the battery SOC model has the capability of compensating ampere-hour integral accumulative errors. Therefore, by acquiring the first parameter of the battery and inputting the first parameter into the pre-trained battery SOC model with the capacity of compensating the ampere-hour integral accumulative error, the problem of error accumulation of a traditional ampere-hour integral method can be effectively solved, and the accuracy of SOC estimation is improved, so that the battery management system can perform charging and discharging control more reasonably, the service life of the battery is prolonged, and the battery management efficiency is improved. The use efficiency of the battery is improved and safe and stable operation of the battery is ensured.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence, and more particularly to a battery SOC estimation method, apparatus, electronic device, medium, and program product. Background Technology

[0002] In fields such as electric vehicles and energy storage systems, accurately estimating the state of charge (SOC) of a battery is crucial. Currently, the ampere-hour integration method is a commonly used SOC estimation method, which calculates the change in SOC by integrating the current. However, this method has a significant drawback: current measurement errors accumulate over time, causing the SOC estimation error to increase continuously. This affects the battery management system's judgment of the battery's state, and consequently, the battery's efficiency and safety. Summary of the Invention

[0003] To overcome the problems existing in related technologies, this disclosure provides a battery SOC estimation method, apparatus, electronic device, medium, and program product.

[0004] According to a first aspect of the present disclosure, a battery SOC estimation method is provided, comprising: Obtain the first parameters of the battery; The first parameter is input into a pre-trained battery SOC model to obtain SOC information, which includes at least an estimated SOC value. The battery SOC model has the ability to compensate for accumulated errors in ampere-hour integration.

[0005] In the above technical solution, by obtaining the first parameter of the battery and inputting it into a pre-trained battery SOC model with the ability to compensate for the cumulative error of ampere-hour integration, the problem of error accumulation in the traditional ampere-hour integration method can be effectively overcome, the accuracy of SOC estimation can be improved, so that the battery management system can perform charging and discharging control more reasonably, extend battery life, improve battery efficiency, and ensure the safe and stable operation of the battery.

[0006] In some possible implementations, the SOC information may also include the uncertainty corresponding to the SOC estimate, which is obtained based on the current deviation coefficient and the SOC prediction deviation coefficient; The current deviation coefficient is used to characterize the reliability of the original input features; The SOC prediction deviation coefficient is used to characterize the degree of fluctuation in the deviation between the SOC estimate and the external reference SOC result.

[0007] In the above technical solution, the uncertainty obtained based on the current deviation coefficient and the SOC prediction deviation coefficient can be introduced to more comprehensively evaluate the reliability of the SOC estimate, provide users with risk reference, and improve the safety and scientific nature of battery management.

[0008] In some possible implementations, the battery SOC model includes a SOC determination model, which includes a biased temporal neural network and an ampere-hour integral neural network; inputting the first parameter into the pre-trained battery SOC model to obtain an estimated SOC value includes: The original ampere-hour integral value in the first parameter is input into the ampere-hour integral neural network to obtain the basic SOC; The other parameters in the first parameter are input into the biased temporal neural network to obtain the SOC correction value; The estimated SOC value is determined based on the base SOC and the corrected SOC value. Wherein, the original ampere-hour integral value is a value obtained by directly integrating the current measurement signal without conversion or correction; the other parameters include at least one of the following: current current, current voltage, time-series average current, time-series average voltage, current power, original ampere-hour integral change difference, current cell temperature, and cell temperature at a preset SOC point.

[0009] In the above technical solution, the basic SOC is determined by using an ampere-hour integral neural network, and the complex factors such as current error and temperature influence are dynamically compensated by a bias timing neural network. This can achieve accurate correction of the accumulated error of the ampere-hour integral and improve the accuracy of the SOC estimate and the robustness of the model.

[0010] In some possible implementations, the time-averaged current is the average current from the preset SOC start point to the current time; the time-averaged voltage is the average voltage from the preset SOC start point to the current time.

[0011] In the above technical solution, by determining the time-series average current and time-series average voltage, the characteristics of the battery at this stage can be more comprehensively reflected, which helps to accurately compensate for the ampere-hour integration error and improve the accuracy of SOC estimation.

[0012] In some possible implementations, the battery SOC model includes a SOC determination model and an uncertainty estimation model. The step of inputting the first parameter into the pre-trained battery SOC model to obtain SOC information includes: The first parameter is input into the SOC determination model to obtain the SOC estimate; The second parameter is input into the uncertainty estimation model to obtain the current deviation weight and the SOC deviation weight, wherein the second parameter includes the estimated SOC value; The uncertainty is determined based on the current deviation weight, the SOC deviation weight, the current deviation coefficient, and the SOC prediction deviation coefficient. The second parameter further includes at least one of the following: current current, current voltage, time-series average current, time-series average voltage, original ampere-hour integral value, original ampere-hour integral change difference, and current cell temperature.

[0013] In the above technical solution, the uncertainty estimation model enables real-time and accurate quantification of the uncertainty of the SOC estimate, providing users with risk references and improving the safety and scientific nature of battery management. Dynamically generated weights can adaptively adjust the influence ratio of current and SOC deviation, making the uncertainty assessment more closely reflect actual battery operating conditions and improving the reliability of the uncertainty.

[0014] In some possible implementations, determining the uncertainty based on the current deviation weight, the SOC deviation weight, the current deviation coefficient, and the SOC prediction deviation coefficient includes: The uncertainty is determined by the sum of the product of the current deviation weight and the current deviation coefficient, and the product of the SOC deviation weight and the SOC estimated deviation coefficient.

[0015] In the above technical solution, the uncertainty can be obtained simply and accurately.

[0016] In some possible implementations, the method further includes: The current deviation value is determined based on the difference between the current and the average current over a preset historical period. Based on the relationship between the estimated SOC value and the SOC accuracy division point, a corresponding strategy is adopted to determine the current deviation coefficient according to the current deviation value.

[0017] In the above technical solution, by comparing the magnitude of the SOC estimate and the SOC precision division point, a corresponding strategy is adopted to determine the current deviation coefficient, which can improve the reliability and accuracy of the current deviation coefficient.

[0018] In some possible implementations, the step of determining the current deviation coefficient based on the current deviation value using a corresponding strategy includes: In response to the SOC estimate being greater than the SOC precision division point, the current deviation coefficient is determined based on the current deviation value and the mean of the sample current disturbance. In response to the SOC estimate being less than or equal to the SOC accuracy division point, the current deviation coefficient is determined based on the current deviation value and the maximum value of the sample current disturbance.

[0019] In the above technical solution, by adaptively selecting the calculation benchmark in the SOC range, using the average characteristic to ensure stability in the high SOC range and the worst operating condition to ensure conservatism in the low SOC range, the current deviation coefficient assessment can be accurately adaptively evaluated, thereby improving the reliability of the determined uncertainty.

[0020] In some possible implementations, when the current deviation values ​​are the same, the current deviation coefficient obtained by using the first current deviation coefficient calculation strategy is greater than the current deviation coefficient obtained by using the second current deviation coefficient calculation strategy. The conditions for using the first current deviation coefficient calculation strategy are: the SOC estimate is greater than the SOC accuracy division point, and the SOC prediction deviation is less than or equal to the maximum allowable error of the external reference SOC result. The conditions for using the second current deviation coefficient calculation strategy are: the estimated SOC value is greater than the SOC accuracy division point, and the predicted SOC deviation value is greater than the maximum allowable error of the external reference SOC result.

[0021] In the above technical solution, when the estimated SOC value is greater than the precision division point, different calculation strategies are used to obtain the current deviation coefficient based on the relationship between the estimated SOC deviation value and the maximum allowable error. This can more flexibly adapt to different error scenarios, optimize the calculation results, and meet the data accuracy requirements under different SOC ranges.

[0022] In some possible implementations, when the SOC estimate is greater than the SOC accuracy cutoff point, the current deviation coefficient is determined using the following formula. :

[0023] in, The current deviation value, The mean of the sample current disturbance. This represents the maximum value of the sample current disturbance. This represents the deviation from the SOC estimate. The maximum allowable error for the external reference SOC result. The SOC estimate is... These are the points used to define the precision of the SOC.

[0024] In the above technical solution, when the SOC estimate is greater than the SOC precision division point, different calculation methods are used according to the SOC prediction deviation to accurately determine the current deviation coefficient, thereby improving the accuracy and stability of SOC estimation.

[0025] In some possible implementations, when the SOC estimate is less than or equal to the SOC accuracy division point, the value of the current deviation coefficient changes exponentially with the change of the current deviation relative to the maximum value of the sample current disturbance.

[0026] In the above technical solution, the current deviation coefficient changes exponentially with the change of the current deviation value relative to the maximum value of the sample, which can amplify the current situation to meet the data accuracy requirements under the current SOC range.

[0027] In some possible implementations, when the SOC estimate is less than or equal to the SOC accuracy division point, the current deviation coefficient is determined by the following formula. :

[0028] The current deviation value, This represents the maximum value of the sample current disturbance. The SOC estimate is... These are the points used to define the precision of the SOC.

[0029] In the above technical solution, the coefficient is determined by exponentially correlating the current deviation value with the maximum disturbance value, which can more sensitively reflect the current change, so as to use the obtained current deviation coefficient to help to more accurately evaluate the SOC.

[0030] In some possible implementations, the method further includes: Based on the deviation between the SOC estimate and the external reference SOC result, the SOC prediction deviation value is determined; Based on the relationship between the estimated SOC value and the SOC precision division point, and the relationship between the predicted SOC deviation value and the allowable error of SOC, a corresponding calculation strategy is adopted to determine the predicted SOC deviation coefficient based on the predicted SOC deviation value.

[0031] In the above technical solution, the adaptive selection of calculation strategy through a dual judgment mechanism can accurately quantify the degree of SOC estimation deviation and improve the reliability and accuracy of the SOC prediction deviation coefficient.

[0032] In some possible implementations, when the SOC prediction deviation values ​​are the same, the SOC prediction deviation coefficient obtained using the first SOC prediction deviation coefficient calculation strategy is greater than the SOC prediction deviation coefficient obtained using other SOC prediction deviation coefficient calculation strategies. The conditions for using the first SOC prediction deviation coefficient calculation strategy are: the SOC estimate is greater than the SOC precision division point, and the SOC prediction deviation is greater than the allowable error of SOC.

[0033] In the above technical solution, when the SOC prediction deviation is the same, the calculation strategy for the first SOC prediction deviation coefficient is clearly defined to obtain a larger coefficient under specific conditions, thereby amplifying the SOC deviation fluctuation and meeting the data accuracy requirements under different SOC ranges.

[0034] In some possible implementations, the step of determining the SOC prediction deviation coefficient based on the SOC prediction deviation value using a corresponding calculation strategy includes: The SOC prediction deviation coefficient is determined using the following formula. :

[0035] in, The deviation value of the SOC prediction. The mean of the sample SOC error. Maximum SOC error of the sample The allowable error for the SOC, The SOC estimate is... These are the points used to define the precision of the SOC.

[0036] In the above technical solution, based on the above formula, the estimated deviation coefficient can be determined according to the relationship between the SOC estimate and the accuracy division point, the estimated deviation value and the allowable error, so as to accurately adapt to different scenarios, improve the accuracy and stability of the SOC estimated deviation coefficient, and thus ensure the reliability of uncertainty.

[0037] In some possible implementations, the uncertainty estimation model is trained using a negative log-likelihood loss function, the loss function... for:

[0038] in, Let the uncertainty be... For variance, For the true value, These are predicted values.

[0039] The above technical solutions can enable uncertainty estimation models to more accurately capture the difference between predictions and actual values, thereby improving the accuracy and effectiveness of uncertainty estimation.

[0040] In some possible implementations, the first parameter is input into a pre-trained battery SOC model to obtain SOC information, including: In response to the battery management system displaying that the SOC has reached a first preset value, the first parameter is input into the pre-trained battery SOC model to obtain SOC information.

[0041] In the above technical solution, if the SOC displayed by the battery management system does not reach the first preset value, the battery SOC model can be omitted to confirm the SOC information, thereby reducing the amount of model calculation and reducing the occupation of system resources.

[0042] In some possible implementations, the method further includes: The parameters input to the battery SOC model are obtained by using data collected when the SOC is greater than a second preset value.

[0043] The above technical solutions can focus on data from specific battery charge stages, improve the quality of input data, make model analysis more targeted, and enhance the accuracy and reliability of SOC estimation.

[0044] According to a second aspect of the present disclosure, a battery SOC estimation apparatus is provided, comprising: The acquisition module is used to acquire the first parameters of the battery; The first determining module is used to input the first parameter into a pre-trained battery SOC model to obtain SOC information, wherein the SOC information includes at least an estimated SOC value. The battery SOC model has the ability to compensate for accumulated errors in ampere-hour integration.

[0045] According to a third aspect of the present disclosure, an electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to execute the executable instructions in the memory to implement the steps of the battery SOC estimation method provided in the first aspect of this disclosure.

[0046] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the battery SOC estimation method provided in the first aspect of the present disclosure.

[0047] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the battery SOC estimation method provided in the first aspect of the present disclosure.

[0048] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0049] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0050] Figure 1 This is a flowchart illustrating a battery SOC estimation method according to an exemplary embodiment.

[0051] Figure 2 This is a schematic diagram illustrating a battery SOC estimation method with charging correction according to an exemplary embodiment.

[0052] Figure 3 This is a schematic diagram illustrating a SOC determination model according to an exemplary embodiment.

[0053] Figure 4 This is a schematic diagram illustrating an uncertainty estimation model according to an exemplary embodiment.

[0054] Figure 5 This is a flowchart illustrating a battery SOC estimation method according to an exemplary embodiment.

[0055] Figure 6 This is a schematic diagram illustrating a comparison between a BSNN model and the Mae of the ampere-hour integral SOC according to an exemplary embodiment.

[0056] Figure 7 This is a schematic diagram illustrating the maximum error comparison between a BSNN model and the ampere-hour integral SOC according to an exemplary embodiment.

[0057] Figure 8 This is a schematic diagram illustrating a comparison of the maximum error in estimating SOC between a U-BSNN model and a BSNN model, according to an exemplary embodiment.

[0058] Figure 9 This is a block diagram illustrating a battery SOC estimation device according to an exemplary embodiment.

[0059] Figure 10 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0060] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0061] It should be noted that all actions involving the acquisition of signals, information, or data in this disclosure are carried out in compliance with the relevant data protection laws and policies of the country where the location is situated, and with authorization from the owner of the relevant device.

[0062] The battery SOC estimation method disclosed herein can be applied to battery management systems, which can be automotive lithium iron phosphate (LiFePO4, LFP) batteries, or other types of batteries. Figure 1 This is a flowchart illustrating a battery SOC estimation method according to an exemplary embodiment, such as... Figure 1 As shown, the battery SOC estimation method includes steps S101 and S102.

[0063] In step S101, the first parameter of the battery is obtained.

[0064] For example, high-precision voltage, current, and temperature sensors can be used to collect battery voltage, current, and temperature data, and the battery's initial parameters can be obtained based on this data. For instance, during electric vehicle operation, the sensors can collect data every second to ensure timeliness and accuracy. Accurately obtaining the battery's initial parameters provides a reliable basis for subsequent SOC estimation.

[0065] In step S102, the first parameter is input into the pre-trained battery SOC model to obtain SOC information, which includes at least the SOC estimate.

[0066] Among them, the battery SOC model has the ability to compensate for the cumulative error of ampere-hour integration.

[0067] For example, the battery SOC model can be a neural network model trained on a large amount of historical data. This battery SOC model takes a first parameter as input and obtains a predicted compensation amount for the accumulated error of the ampere-hour integral. This compensation is then combined with the calculated ampere-hour integral value to obtain the SOC estimate. This battery SOC model can be stored locally on the electronic device and invoked locally each time it is used, or it can be stored on a third-party platform and invoked from the third party each time it is used; no specific limitation is made here.

[0068] In the above technical solution, by obtaining the first parameter of the battery and inputting it into a pre-trained battery SOC model with the ability to compensate for the cumulative error of ampere-hour integration, the problem of error accumulation in the traditional ampere-hour integration method can be effectively overcome, the accuracy of SOC estimation can be improved, so that the battery management system can perform charging and discharging control more reasonably, extend battery life, improve battery efficiency, and ensure the safe and stable operation of the battery.

[0069] In some possible implementations, in step S102, the first parameter is input into the pre-trained battery SOC model to obtain SOC information, including: In response to the battery management system displaying that the SOC has reached a first preset value, the first parameter is input into the pre-trained battery SOC model to obtain SOC information.

[0070] For example, the first preset value can be preset based on actual needs, such as 80%. For battery protection, users typically stop charging when it reaches 80% to 90%. The traditional SOC correction range is 3% before full charge (i.e., the correction range is 97%-100%). Figure 2 The BMS calibration range in the data is used. This means that when the SOC reaches above 80%, the accumulated error may persist because correction cannot be triggered. By setting the first preset value, the SOC correction range is expanded from 3% before full charge to 20% before full charge (corresponding to...). Figure 2 (The AI ​​estimation algorithm calibration range in the text). Thus, by expanding the SOC correction range, the SOC estimation can be corrected in a timely manner over a wider range of battery capacity, effectively improving the accuracy of SOC estimation and ensuring the proper use and safety of the battery.

[0071] If the SOC displayed in the battery management system does not reach the first preset value, the SOC information can be confirmed without using the battery SOC model. This reduces the amount of model calculation and lowers the system resource consumption.

[0072] In some possible implementations, the battery SOC estimation method provided in this disclosure may further include: The parameters input to the battery SOC model are obtained by using data collected when the SOC is greater than the second preset value.

[0073] For example, the second preset value can be preset based on actual needs, such as 70%. By collecting data from points with SOC ≥ 70% as the data source, it is possible to focus on data at specific battery charge stages, improve the quality of input data, make model analysis more targeted, and enhance the accuracy and reliability of SOC estimation.

[0074] In some possible implementations, the first parameter may include the raw ampere-hour integral value. And other parameters. The original ampere-hour integral value. The value is the direct integration of the current measurement signal without conversion or correction. The other parameters are parameters that are highly correlated with the battery SOC and are used to compensate for the cumulative error of the ampere-hour integration.

[0075] Other parameters may include current current I, current voltage V, and time-averaged current. Time-series average voltage Current power P, difference in original ampere-hour integral change At least one of the current cell temperature T and the cell temperature at the preset SOC point.

[0076] For example, the current power P is the product of the current current I and the current voltage V. The current cell temperature T and the cell temperature at a preset SOC point are collected as temperature characteristics. For example, the preset SOC points can be 50%, 60%, and 70%, and the cell temperatures at these preset SOC points can be denoted as follows: , , The cell temperature can be the average temperature of multiple cells in the battery.

[0077] For example, time-series average current This is the average current from a preset SOC starting point (e.g., 70%) to the current time. In other words, the time-series average current can be defined as the average current from the preset SOC starting point to each time point. The corresponding expression is: .

[0078] For example, time-series average voltage This refers to the average voltage from a preset SOC (for example, 70%) to the current time. In other words, the time-averaged voltage can be defined as the average voltage from the preset SOC to each time point. The corresponding expression is: .

[0079] in, For the current moment, The time corresponding to the preset SOC start point, For the first Current at any moment For the first Voltage at a given moment.

[0080] Thus, by determining the time-series average current and time-series average voltage, the characteristics of the battery at this stage can be more comprehensively reflected, which can help to accurately compensate for the ampere-hour integration error and improve the accuracy of SOC estimation.

[0081] For example, the first parameter may include 11 input features, namely... It can construct feature groups with a time series length of L, for example, L is 300 (sampling length: 300s, sampling period: 1s).

[0082] In some possible implementations, the battery SOC model includes a SOC determination model. This SOC determination model is used to estimate the SOC based on a first input parameter, obtaining an estimated SOC value. For example... Figure 3As shown, the SOC determination model may include a biased temporal neural network and an ampere-hour integral neural network.

[0083] like Figure 3 As shown, the feature input module receives input features (i.e., the first parameter) and divides them into two types of input features. The first type of input is time-series data containing the original ampere-hour integral value, with a dimension of (batch_size, 1); the second type of input is time-series data of other parameters, including but not limited to current, voltage, temperature, etc., with a dimension of (batch_size, time_len, feature_dim-1).

[0084] In a biased temporal neural network, the time-series data of other parameters are first transformed by a reshape layer, then standardized by a batch normalization layer (BatchNormld), and then reshaped again. The preprocessed time-series data of other battery parameters are sequentially input into the first LSTM layer (LSTM1) and the second LSTM layer (LSTM2), extracting 256-dimensional and 64-dimensional temporal features, respectively. Then, the output features of the last 16 time steps are taken and reshaped. The reshaped features are then fed into a subsequent multilayer perceptron (MLP) for nonlinear transformation, outputting the SOC correction value. The MLP can be sequentially connected to a linear transformation layer (Linear), a batch normalization layer (BatchNormld), and a ReLU activation function; nonlinear feature transformation is achieved through the stacking of multiple such structures. Figure 3 As shown, an MLP may include at least two fully connected layers with the number of neurons decreasing sequentially to obtain a SOC correction value.

[0085] In the ampere-hour integral neural network, the original ampere-hour integral value can be nonlinearly mapped by a lightweight multilayer perceptron (MLP) and projected onto a new feature space to obtain the basic SOC.

[0086] The corrected SOC value can be summed with the basic SOC, and the result is the final SOC estimate. By combining the results of the biased temporal neural network and the ampere-hour integral neural network, on the one hand, the correspondence between the ampere-hour integral and the actual SOC is statistically determined; on the other hand, the SOC prediction is continuously corrected based on the temporal feature values, which can make the SOC estimate more accurate and more stable.

[0087] Correspondingly, in step S102, the first parameter is input into the pre-trained battery SOC model to obtain SOC information, including: The original ampere-hour integral value in the first parameter is input into the ampere-hour integral neural network to obtain the basic SOC; The other parameters in the first parameter are input into the biased temporal neural network to obtain the SOC correction value; The estimated SOC value is determined based on the base SOC and the SOC correction value.

[0088] Thus, by using an ampere-hour integral neural network to determine the basic SOC, and by using a bias timing neural network to dynamically compensate for complex factors such as current error and temperature influence, it is possible to accurately correct the accumulated error of the ampere-hour integral, thereby improving the accuracy of the SOC estimate and the robustness of the model.

[0089] In some possible implementations, historical first parameters and corresponding true SOC values ​​can be used as training data to train the SOC determination model until training termination conditions are met. This allows the SOC determination model to autonomously learn the variation pattern of ampere-hour integral error from the data, thereby gaining generalization ability to unknown data and improving the accuracy and adaptability of the SOC determination model in practical applications. The training termination conditions may include at least one of the following: the model's loss function output value is less than or equal to a preset threshold, or the number of iterations reaches a preset threshold.

[0090] In some possible implementations, the SOC information also includes the uncertainty corresponding to the SOC estimate, which is obtained based on the current deviation coefficient and the SOC prediction deviation coefficient.

[0091] Among them, the current deviation coefficient is used to characterize the reliability of the original input characteristics; the SOC prediction deviation coefficient is used to characterize the degree of fluctuation of the deviation between the SOC estimate and the external reference SOC result.

[0092] Therefore, by introducing the uncertainty based on the current deviation coefficient and the SOC prediction deviation coefficient, the reliability of the SOC estimate can be more comprehensively evaluated, providing users with risk reference and improving the safety and scientific nature of battery management.

[0093] In some possible implementations, the current deviation coefficient is determined in the following manner: The current deviation value is determined based on the difference between the current and the average current over a preset historical period. Based on the relationship between the estimated SOC value and the SOC accuracy division point, a corresponding strategy is adopted to determine the current deviation coefficient according to the current deviation value.

[0094] In one embodiment, the preset history duration can be pre-set based on actual needs, for example, it can be set to 30 seconds. Correspondingly, the current deviation value can be determined using the following formula. : Where t is the current time, For the first Current at any given moment.

[0095] Considering that the accuracy of the mapping between current, voltage, and SOC improves with increasing SOC value, but the model's tolerance for SOC error decreases as the SOC value increases, the SOC value can be divided into multiple intervals for evaluation. For example, based on the relationship between the estimated SOC value and the SOC accuracy division point, a corresponding strategy can be adopted to determine the current deviation coefficient based on the current deviation value, ensuring that the most suitable benchmark is used for current deviation coefficient evaluation in different SOC intervals. The SOC accuracy division point can be set based on actual needs, for example, it can be set to 85%, without limitation.

[0096] In one embodiment, based on the relationship between the estimated SOC value and the SOC accuracy division point, a corresponding strategy is adopted to determine the current deviation coefficient according to the current deviation value, which may include: In response to the SOC estimate being greater than the SOC precision cutoff point, the current deviation coefficient is determined based on the current deviation value and the mean of the sample current disturbance. In response to the SOC estimate being less than or equal to the SOC accuracy cutoff point, the current deviation coefficient is determined based on the current deviation value and the maximum value of the sample current disturbance.

[0097] Among them, the mean of the sample current disturbance can be the average of the maximum values ​​of the current disturbance under each operating condition in the sample data, and the maximum value of the sample current disturbance can be the maximum value among the maximum values ​​of the current disturbance under each operating condition in the sample data.

[0098] Thus, by adaptively selecting the calculation benchmark within the SOC range, using average characteristics to ensure stability in the high SOC range and the worst-case scenario to ensure conservatism in the low SOC range, accurate adaptive evaluation of the current deviation coefficient can be achieved, thereby improving the reliability of the determined uncertainty.

[0099] When the estimated SOC value is greater than the SOC precision division point, to further ensure the stability of SOC, the estimated SOC deviation value can be used for further segmentation. When the current deviation values ​​are the same, the current deviation coefficient obtained using the first current deviation coefficient calculation strategy is greater than the current deviation coefficient obtained using the second current deviation coefficient calculation strategy. The conditions for using the first current deviation coefficient calculation strategy are: the estimated SOC value is greater than the SOC accuracy division point, and the predicted SOC deviation value is less than or equal to the maximum allowable error of the external reference SOC result, i.e. .

[0100] The conditions for using the second current deviation coefficient calculation strategy are: the estimated SOC value is greater than the SOC accuracy division point, and the predicted SOC deviation value is greater than the maximum allowable error of the external reference SOC result, i.e. .

[0101] In the above technical solution, when the estimated SOC value is greater than the precision division point, different calculation strategies are used to obtain the current deviation coefficient based on the relationship between the estimated SOC deviation value and the maximum allowable error. This can more flexibly adapt to different error scenarios, optimize the calculation results, and meet the data accuracy requirements under different SOC ranges.

[0102] For example, the current deviation coefficient can be determined using the following formula. :

[0103] in, This is the current deviation value. The mean of the sample current disturbance. This represents the maximum value of the sample current disturbance. This represents the deviation from the SOC estimate. The maximum allowable error for the external reference SOC result. This is the estimated SOC value. The SOC precision division points.

[0104] Among them, SOC prediction deviation The deviation can be determined based on the difference between the SOC estimate and the external reference SOC result. For example, the SOC prediction deviation can be determined using the following formula. : ,in, As an external reference SOC result, This indicates taking the absolute value.

[0105] For example, The value can be 5%. In the case of, if If this is the case, it can be determined that the current estimation accuracy of the model is good and within the confidence interval. At this point, the first current deviation coefficient calculation strategy can be adopted (i.e., To calculate the current deviation coefficient; In the case of, if If the model estimation accuracy is poor and the results are unreliable, then a second current deviation coefficient calculation strategy can be adopted (i.e., To calculate the current deviation coefficient, a more conservative risk assessment can be performed.

[0106] Thus, when the SOC estimate is greater than the SOC precision division point, different calculation methods can be used based on the deviation of the SOC prediction to accurately determine the current deviation coefficient, thereby improving the accuracy and stability of the SOC estimate.

[0107] When the estimated SOC is less than or equal to the SOC precision cutoff point, the current deviation coefficient can change exponentially with the change in current deviation relative to the maximum value of the sample current disturbance. This allows for amplification of the current data to meet the precision requirements within the current SOC range.

[0108] For example, when the SOC estimate is less than or equal to the SOC accuracy cutoff point, the current deviation coefficient is determined using the following formula. :

[0109] This is the current deviation value. This represents the maximum value of the sample current disturbance. This is the estimated SOC value. The SOC precision division points.

[0110] Therefore, by determining the coefficients by exponentially correlating the current deviation with the maximum disturbance value, current changes can be reflected more sensitively, and the obtained current deviation coefficients can be used to assist in a more accurate assessment of SOC.

[0111] In some possible implementations, the SOC prediction deviation factor is determined in the following manner: The deviation of the SOC estimate is determined based on the discrepancy between the SOC estimate and the external reference SOC result; Based on the relationship between the SOC estimate and the SOC accuracy division point, and the relationship between the SOC prediction deviation and the allowable error of SOC, a corresponding calculation strategy is adopted to determine the SOC prediction deviation coefficient based on the SOC prediction deviation.

[0112] Thus, by using a dual-judgment mechanism to adaptively select the calculation strategy, the degree of SOC estimation deviation can be accurately quantified, thereby improving the reliability and accuracy of the SOC prediction deviation coefficient.

[0113] To meet the data accuracy requirements under different SOC ranges, it can be set that, when the SOC prediction deviation value is the same, the SOC prediction deviation coefficient obtained by using the first SOC prediction deviation coefficient calculation strategy is greater than the SOC prediction deviation coefficient obtained by using other SOC prediction deviation coefficient calculation strategies. The conditions for using the first SOC prediction deviation coefficient calculation strategy are: the SOC estimate is greater than the SOC precision division point, and the SOC prediction deviation is greater than the allowable error of SOC. .

[0114] For example, in When the SOC prediction deviation is less than or equal to the SOC allowable error, the corresponding calculation strategy can be determined based on the relationship between the SOC prediction deviation and the SOC allowable error. And a relatively large denominator, to determine a relatively stable SOC prediction deviation coefficient; when the SOC prediction deviation value is greater than the allowable error of SOC, it can be based on And a relatively small denominator, which amplifies the deviation of the SOC prediction.

[0115] For example, the SOC prediction deviation factor can be determined using the following formula. :

[0116] in, This represents the deviation from the SOC estimate. The mean of the sample SOC error. Maximum SOC error of the sample For the allowable error of SOC, This is the estimated SOC value. The SOC precision division points.

[0117] Among them, the mean of sample SOC error This can be the average of the maximum errors of the external reference SOC results for each operating condition in the sample data; the maximum value of the sample SOC error. It can be the maximum value among the maximum errors of the external reference SOC results for each working condition in the sample data.

[0118] For example, The value can be 3%. At that time, if If this is the case, it can be determined that the current estimation accuracy of the model is good and within the confidence interval. At this point, it can be proceeded according to... (i.e., the second SOC prediction deviation coefficient calculation strategy) is used to calculate to obtain a relatively stable result; if If this is the case, it can be determined that the model estimation accuracy is poor and the results have low reliability. In this case, it can be followed... (That is, the first SOC prediction deviation coefficient calculation strategy) is used to calculate the deviation coefficient, making it more sensitive to errors for amplified risk assessment. And... Similarly, the second SOC prediction deviation coefficient calculation strategy can be adopted to make the results relatively stable and conservative, thereby improving the reliability of the uncertainty. The second SOC prediction deviation coefficient calculation strategy corresponds to the other SOC prediction deviation coefficient calculation strategies mentioned above.

[0119] Thus, based on the above formula, the estimated deviation coefficient can be determined according to the relationship between the SOC estimate and the precision division point, the estimated deviation value and the allowable error, and different scenarios can be accurately adapted to improve the accuracy and stability of the SOC estimated deviation coefficient, thereby ensuring the reliability of uncertainty.

[0120] In some possible implementations, the battery SOC model includes an SOC determination model and an uncertainty estimation model. The SOC determination model has been described in detail in the embodiments above and will not be elaborated upon here. The uncertainty estimation model can be constructed based on a neural network model, which is used to determine the current deviation weight and SOC deviation weight according to the second parameter, so as to combine the current deviation coefficient and the SOC prediction deviation coefficient to obtain the uncertainty corresponding to the SOC estimate.

[0121] The second parameter may include the SOC estimate obtained from the SOC determination model. It may also include the current current I, the current voltage V, and the time-series average current. Time-series average voltage Original ampere-hour integral value Difference in the original ampere-hour integral and at least one of the current cell temperature T. For example, the second parameter input to the uncertainty estimation model may include nine input features, namely... .

[0122] The neural network model in the uncertainty estimation model may include a first fully connected layer, a batch normalization layer, a first activation function layer, a second fully connected layer, a second activation function layer, a third fully connected layer, a third activation function layer, and a fourth fully connected layer. For example... Figure 4 As shown, the activation function layers can all use the ReLU function. The first fully connected layer, Linear1, maps the 9-dimensional input features to a 256-dimensional high-dimensional space; the second fully connected layer, Linear2, compresses the 256-dimensional features to 64 dimensions; the third fully connected layer, Linear3, compresses the 64-dimensional features to 16 dimensions; and the fourth fully connected layer, Linear4, maps the 16-dimensional features to a 2-dimensional output, obtaining the current deviation weights. and SOC deviation weight Deviate the current from the weight and SOC deviation weight By fusing the current deviation coefficient and the SOC prediction deviation coefficient, a value equal to the SOC estimate can be obtained. The corresponding uncertainty.

[0123] based on Figure 4Given a structure, the deviation coefficient is used, and the output is learned through a neural network to obtain the uncertainty, which in turn yields the variance of the model error. On one hand, this can filter out points with large errors; on the other hand, the variance of the error can be used for model fusion.

[0124] Thus, by using the uncertainty estimation model, the uncertainty of the SOC estimate can be quantified in real time and accurately, providing users with risk references and improving the safety and scientific nature of battery management.

[0125] Correspondingly, in step S102, the first parameter is input into the pre-trained battery SOC model to obtain SOC information, including: Input the first parameter into the SOC determination model to obtain the SOC estimate; The second parameter is input into the uncertainty estimation model to obtain the current deviation weight and SOC deviation weight. The second parameter includes the SOC estimate. The uncertainty is determined based on the current deviation weight, the SOC deviation weight, the current deviation coefficient, and the SOC prediction deviation coefficient.

[0126] In this way, the influence of the deviation between current and SOC can be adaptively adjusted by using dynamically generated weights, making the uncertainty assessment more in line with the actual operating conditions of the battery and improving the reliability of the uncertainty.

[0127] In one embodiment, the uncertainty can be determined as the sum of the product of the current deviation weight and the current deviation coefficient, and the product of the SOC deviation weight and the estimated SOC deviation coefficient. Alternatively, the uncertainty can be determined using the following formula. : In this way, the uncertainty can be obtained simply and accurately.

[0128] The uncertainty estimation model can be trained using historical second parameters and corresponding true SOC values ​​as training data until the training termination condition is met. This allows the uncertainty estimation model to accurately learn the error distribution pattern, thereby achieving precise quantification of the uncertainty in SOC estimation. The training termination condition can include at least one of the following: the model's loss function output value is less than or equal to a preset threshold, or the number of iterations reaches a preset threshold.

[0129] In one embodiment, an uncertainty estimation model can be trained using uncertainty loss to obtain the uncertainty value corresponding to SOC, and then converted into error.

[0130] First, we can assume that the output value of the SOC estimation model is... Follow the mean variance is The Gaussian distribution of can be written as:

[0131] Taking the negative logarithm of the probability density function, the expression is:

[0132] Simplified, we get:

[0133] Using negative log-likelihood loss, the loss function can be written as:

[0134] in, This is the residual term, used to measure the error between the predicted and actual values, and the variance at this point. The smaller the size, the greater the punishment; This is a regularization term that prevents the model from predicting infinitesimally small variances in order to minimize residuals, thus serving as a constraint.

[0135] To prevent variance When the value approaches zero, it becomes unstable. Therefore, the uncertainty can be defined as: Substituting this into the loss function, we get: ,in, For the true value, These are predicted values.

[0136] In this way, the uncertainty estimation model can more accurately capture the difference between the prediction and the actual value, thereby improving the accuracy and effectiveness of uncertainty estimation.

[0137] Figure 5 This is a flowchart illustrating a battery SOC estimation method according to an exemplary embodiment. Through this... Figure 5 This provides a clearer understanding of the implementation process of the battery SOC estimation method provided in this disclosure. During data preprocessing, data filtering is performed, and parameters are generated for input into the model and for determining the current offset coefficient and the SOC prediction deviation coefficient. In the model building phase, the current offset coefficient and the SOC prediction deviation coefficient are determined, and the SOC estimation value is obtained using the SOC determination model. This SOC estimation value is then input into the uncertainty estimation model for calculation. In the model training phase, the uncertainty estimation model obtains the uncertainty loss corresponding to the SOC estimation value based on the obtained information, thus obtaining the numerical value of the uncertainty corresponding to the SOC estimation value.

[0138] The correction range provided in this disclosure (SOC>80%, where 80% is the first preset value) is 6 times larger than the traditional full-charge correction range (SOC>97%). Compared to the SOC estimate obtained by ampere-hour integration, the maximum error of the SOC estimate obtained based on the battery SOC estimation method provided in this disclosure is smaller. This disclosure uses uncertainty to characterize the error range of the SOC estimate, which is used for SOC estimate screening and subsequent fusion, and can further improve the accuracy of the SOC estimate.

[0139] The following section compares the SOC estimate determined based on the battery SOC model provided in this disclosure with the maximum SOC error of the related ampere-hour integral SOC calculation method, and verifies the effectiveness and coverage of the uncertainty estimation model. An ablation experiment and coverage statistics of the uncertainty estimation model are conducted, and the experimental results are as follows.

[0140] For the sake of simplicity, the SOC determination model provided in this disclosure is referred to as the BSNN model below, and the battery SOC model provided in this disclosure, which includes the SOC determination model and the uncertainty estimation model, is referred to as the U-BSNN model below.

[0141] A comparison of the Mae (mean absolute error) and maximum error of SOC estimation using two methods—the BSNN model and the ampere-hour integral method—is as follows: Figure 6 , 7 As shown.

[0142] Comparing the U-BSNN model and the BSNN model with and without adding uncertainty estimation, based on the definition of uncertainty and the Gaussian distribution... The rule for uncertainty The true value in over 97% of cases Compared with model predictions The relationship is as follows:

[0143] When the SOC is 80%-90%, the portion with uncertainty greater than 1 is filtered out (equivalent to approximately 5 times the standard deviation). When the SOC is 90%-100%, the portion with uncertainty greater than 0 is filtered out (equivalent to 3 times the standard deviation). The maximum error comparison is as follows. Figure 8 As shown.

[0144] from Figure 6 , 7 It can be seen that, compared with the ampere-hour integration method, the BSNN model constructed by this scheme has a smaller and more concentrated SOC error, and a smaller maximum error under each operating condition; from Figure 8It can be seen that adding the uncertainty estimation model reduces the maximum error of the model by filtering out points with high uncertainty in the operating conditions. Table 1 shows a comparison of the maximum error and Mae index of the ampere-hour integral method, BSNN, and U-BSNN models.

[0145] Table 1

[0146] The coverage and accuracy of the uncertainty estimation model for the SOC intervals [80,85), [85,90), [90,95), and [95,100) were statistically analyzed. The uncertainty of the intervals [80,85) and [85,90) is less than 1, and the uncertainty of the intervals [90,95) and [95,100) is less than 0, as shown in Table 2.

[0147] Table 2

[0148] As shown in Table 2, the uncertainty coverage exceeds 90% and the accuracy exceeds 99% in each SOC interval. The overall uncertainty coverage exceeds 95% and the overall accuracy exceeds 99.6%.

[0149] Therefore, the battery SOC estimation method provided in this disclosure can obtain an accurate SOC estimate, enabling the battery management system to more rationally control charging and discharging, extend battery life, improve battery efficiency, and ensure the safe and stable operation of the battery.

[0150] Figure 9 This is a block diagram illustrating a battery SOC estimation device 300 according to an exemplary embodiment. (Refer to...) Figure 9 The device includes an acquisition module 301 and a first determination module 302.

[0151] Module 301 is used to acquire the first parameters of the battery; The first determining module 302 is used to input the first parameter into a pre-trained battery SOC model to obtain SOC information, wherein the SOC information includes at least an estimated SOC value. The battery SOC model has the ability to compensate for accumulated errors in ampere-hour integration.

[0152] In the above technical solution, by obtaining the first parameter of the battery and inputting it into a pre-trained battery SOC model with the ability to compensate for the cumulative error of ampere-hour integration, the problem of error accumulation in the traditional ampere-hour integration method can be effectively overcome, the accuracy of SOC estimation can be improved, so that the battery management system can perform charging and discharging control more reasonably, extend battery life, improve battery efficiency, and ensure the safe and stable operation of the battery.

[0153] In some possible implementations, the SOC information may also include the uncertainty corresponding to the SOC estimate, which is obtained based on the current deviation coefficient and the SOC prediction deviation coefficient; The current deviation coefficient is used to characterize the reliability of the original input features; The SOC prediction deviation coefficient is used to characterize the degree of fluctuation in the deviation between the SOC estimate and the external reference SOC result.

[0154] In some possible implementations, the battery SOC model includes a SOC determination model, which includes a biased temporal neural network and an ampere-hour integral neural network; the first determination module 302 includes: The first determining submodule is used to input the original ampere-hour integral value in the first parameter into the ampere-hour integral neural network to obtain the basic SOC; The second determining submodule is used to input other parameters in the first parameter into the biased temporal neural network to obtain the SOC correction value; The third determining submodule is used to determine the estimated SOC value based on the basic SOC and the SOC correction value; Wherein, the original ampere-hour integral value is a value obtained by directly integrating the current measurement signal without conversion or correction; the other parameters include at least one of the following: current current, current voltage, time-series average current, time-series average voltage, current power, original ampere-hour integral change difference, current cell temperature, and cell temperature at a preset SOC point.

[0155] In some possible implementations, the time-averaged current is the average current from the preset SOC start point to the current time; the time-averaged voltage is the average voltage from the preset SOC start point to the current time.

[0156] In some possible implementations, the battery SOC model includes an SOC determination model and an uncertainty estimation model; the first determination module 302 includes: The fourth determination submodule is used to input the first parameter into the SOC determination model to obtain the SOC estimate; The fifth determining submodule is used to input the second parameter into the uncertainty estimation model to obtain the current deviation weight and the SOC deviation weight, wherein the second parameter includes the estimated SOC value; The sixth determining submodule is used to determine the uncertainty based on the current deviation weight, the SOC deviation weight, the current deviation coefficient, and the SOC prediction deviation coefficient; The second parameter further includes at least one of the following: current current, current voltage, time-series average current, time-series average voltage, original ampere-hour integral value, original ampere-hour integral change difference, and current cell temperature.

[0157] In some possible implementations, the sixth determining submodule is used to determine the uncertainty in the following ways: The uncertainty is determined by the sum of the product of the current deviation weight and the current deviation coefficient, and the product of the SOC deviation weight and the SOC estimated deviation coefficient.

[0158] In some possible implementations, the device 300 further includes: The second determining module is used to determine the current deviation value based on the difference between the current and the average current value within a preset historical time period; and to determine the current deviation coefficient based on the current deviation value using a corresponding strategy according to the relationship between the SOC estimate and the SOC precision division point.

[0159] In some possible implementations, the second determining module is used to determine the current deviation coefficient based on the current deviation value using a corresponding strategy in the following manner: In response to the SOC estimate being greater than the SOC precision division point, the current deviation coefficient is determined based on the current deviation value and the mean of the sample current disturbance. In response to the SOC estimate being less than or equal to the SOC accuracy division point, the current deviation coefficient is determined based on the current deviation value and the maximum value of the sample current disturbance.

[0160] In some possible implementations, when the current deviation values ​​are the same, the current deviation coefficient obtained by using the first current deviation coefficient calculation strategy is greater than the current deviation coefficient obtained by using the second current deviation coefficient calculation strategy. The conditions for using the first current deviation coefficient calculation strategy are: the SOC estimate is greater than the SOC accuracy division point, and the SOC prediction deviation is less than or equal to the maximum allowable error of the external reference SOC result. The conditions for using the second current deviation coefficient calculation strategy are: the estimated SOC value is greater than the SOC accuracy division point, and the predicted SOC deviation value is greater than the maximum allowable error of the external reference SOC result.

[0161] In some possible implementations, when the SOC estimate is greater than the SOC accuracy cutoff point, the current deviation coefficient is determined using the following formula. :

[0162] in, The current deviation value, The mean of the sample current disturbance. This represents the maximum value of the sample current disturbance. This represents the deviation from the SOC estimate. The maximum allowable error for the external reference SOC result. The SOC estimate is... These are the points used to define the precision of the SOC.

[0163] In some possible implementations, when the SOC estimate is less than or equal to the SOC accuracy division point, the value of the current deviation coefficient changes exponentially with the change of the current deviation relative to the maximum value of the sample current disturbance.

[0164] In some possible implementations, when the SOC estimate is less than or equal to the SOC accuracy division point, the current deviation coefficient is determined by the following formula. :

[0165] The current deviation value, This represents the maximum value of the sample current disturbance. The SOC estimate is... These are the points used to define the precision of the SOC.

[0166] In some possible implementations, the device 300 further includes: The third determining module is used to determine the SOC prediction deviation value based on the deviation between the SOC estimate and the external reference SOC result; and to determine the SOC prediction deviation coefficient based on the relationship between the SOC estimate and the SOC precision division point, and the relationship between the SOC prediction deviation value and the allowable error of SOC, using a corresponding calculation strategy.

[0167] In some possible implementations, when the SOC prediction deviation values ​​are the same, the SOC prediction deviation coefficient obtained using the first SOC prediction deviation coefficient calculation strategy is greater than the SOC prediction deviation coefficient obtained using other SOC prediction deviation coefficient calculation strategies. The conditions for using the first SOC prediction deviation coefficient calculation strategy are: the SOC estimate is greater than the SOC precision division point, and the SOC prediction deviation is greater than the allowable error of SOC.

[0168] In some possible implementations, the third determining module is used to determine the SOC prediction deviation coefficient based on the SOC prediction deviation value using a corresponding calculation strategy in the following manner: The SOC prediction deviation coefficient is determined using the following formula. :

[0169] in, The deviation value of the SOC prediction. The mean of the sample SOC error. Maximum SOC error of the sample The allowable error for the SOC, The SOC estimate is... These are the points used to define the precision of the SOC.

[0170] In some possible implementations, the uncertainty estimation model is trained using a negative log-likelihood loss function, the loss function... for:

[0171] in, Let the uncertainty be... For variance, For the true value, These are predicted values.

[0172] In some possible implementations, the first determining module 301 is used to input the first parameter into a pre-trained battery SOC model to obtain SOC information in the following manner: In response to the battery management system displaying that the SOC has reached a first preset value, the first parameter is input into the pre-trained battery SOC model to obtain SOC information.

[0173] In some possible implementations, device 300 further includes: The fourth determining module is used to obtain the parameters input to the battery SOC model by using the data collected when the SOC is greater than the second preset value.

[0174] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0175] Figure 10 This is a block diagram illustrating an electronic device 1900 according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. (Refer to...) Figure 10The electronic device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by memory 1932 for storing instructions, such as application programs, that can be executed by the processing component 1922. The application programs stored in memory 1932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 1922 is configured to execute instructions to perform the aforementioned battery SOC estimation method.

[0176] Electronic device 1900 may also include a power supply component 1926 configured to perform power management of electronic device 1900, a wired or wireless network interface 1950 configured to connect electronic device 1900 to a network, and an input / output interface 1958. Electronic device 1900 can operate on an operating system, such as Windows Server, stored in memory 1932. TM Mac OS X TM Unix TM Linux TM FreeBSD TM Or similar.

[0177] In another exemplary embodiment, this disclosure also provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the steps of the battery SOC estimation method provided in this disclosure.

[0178] In another exemplary embodiment, this disclosure also provides a computer program product comprising a computer program executable by a programmable device, the computer program having a code portion for performing the battery SOC estimation method described above when executed by the programmable device.

[0179] Those skilled in the art will also understand that the various illustrative logical blocks and steps listed in the embodiments of this application can be implemented by electronic hardware, computer software, or a combination of both. Whether such functionality is implemented through hardware or software depends on the specific application and the overall system design requirements. Those skilled in the art can implement the described functionality using various methods for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of this application.

[0180] It should be understood that, unless otherwise specifically indicated, features of various embodiments of this disclosure described herein can be combined with each other. As used herein, the term “and / or” includes any one of the relevant listed items and any combination of any two or more; similarly, “at least one of…” includes any one of the relevant listed items and any combination of any two or more.

[0181] Although terms such as “first,” “second,” and “third” may be used herein to describe various components, parts, regions, layers, or sections, these components, parts, regions, layers, or sections are not limited to these terms. Rather, these terms are used only to distinguish one component, part, region, layer, or section from another. Therefore, without departing from the teachings of the examples described herein, the first component, part, region, layer, or section mentioned in the examples may also be referred to as the second component, part, region, layer, or section. Furthermore, the terms “first” and “second” are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as “first” or “second” may explicitly or implicitly include at least one of that feature. In the description herein, “a plurality” means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0182] Similarly, although this disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding this specification and the accompanying drawings. This disclosure includes all such modifications and variations and is limited only by the scope of the claims. In particular, with respect to the various functions performed by the components described above (e.g., elements, resources, etc.), unless otherwise indicated, the terminology used to describe such components is intended to correspond to any component (functionally equivalent) that performs the specific function of the described component, even if structurally not equivalent to the disclosed structure. Furthermore, although specific features of this disclosure may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations, as may be desired and advantageous to any given or particular application. Moreover, with regard to the terms “comprising,” “owning,” “having,” “having,” or variations thereof as used in the detailed description or claims, such terms are intended to be inclusive in a manner similar to the term “including.”

[0183] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

[0184] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for estimating battery SOC, characterized in that, include: Obtain the first parameters of the battery; The first parameter is input into a pre-trained battery SOC model to obtain SOC information, which includes at least an estimated SOC value. The battery SOC model has the ability to compensate for accumulated errors in ampere-hour integration.

2. The method according to claim 1, characterized in that, The SOC information also includes the uncertainty corresponding to the SOC estimate, which is obtained based on the current deviation coefficient and the SOC prediction deviation coefficient. The current deviation coefficient is used to characterize the reliability of the original input features; The SOC prediction deviation coefficient is used to characterize the degree of fluctuation in the deviation between the SOC estimate and the external reference SOC result.

3. The method according to claim 1 or 2, characterized in that, The battery SOC model includes a SOC determination model, which includes a biased temporal neural network and an ampere-hour integral neural network. The first parameter is input into a pre-trained battery SOC model to obtain an estimated SOC value, including: The original ampere-hour integral value in the first parameter is input into the ampere-hour integral neural network to obtain the basic SOC; The other parameters in the first parameter are input into the biased temporal neural network to obtain the SOC correction value; The estimated SOC value is determined based on the base SOC and the corrected SOC value. Wherein, the original ampere-hour integral value is a value obtained by directly integrating the current measurement signal without conversion or correction; the other parameters include at least one of the following: current current, current voltage, time-series average current, time-series average voltage, current power, original ampere-hour integral change difference, current cell temperature, and cell temperature at a preset SOC point.

4. The method according to claim 3, characterized in that, The time-series average current is the average current from the preset SOC start point to the current time; the time-series average voltage is the average voltage from the preset SOC start point to the current time.

5. The method according to claim 2, characterized in that, The battery SOC model includes a SOC determination model and an uncertainty estimation model. The step of inputting the first parameter into the pre-trained battery SOC model to obtain SOC information includes: The first parameter is input into the SOC determination model to obtain the SOC estimate; The second parameter is input into the uncertainty estimation model to obtain the current deviation weight and the SOC deviation weight, wherein the second parameter includes the estimated SOC value; The uncertainty is determined based on the current deviation weight, the SOC deviation weight, the current deviation coefficient, and the SOC prediction deviation coefficient. The second parameter further includes at least one of the following: current current, current voltage, time-series average current, time-series average voltage, original ampere-hour integral value, original ampere-hour integral change difference, and current cell temperature.

6. The method according to claim 5, characterized in that, The step of determining the uncertainty based on the current deviation weight, the SOC deviation weight, the current deviation coefficient, and the SOC prediction deviation coefficient includes: The uncertainty is determined by the sum of the product of the current deviation weight and the current deviation coefficient, and the product of the SOC deviation weight and the SOC estimated deviation coefficient.

7. The method according to claim 2, characterized in that, The method further includes: The current deviation value is determined based on the difference between the current and the average current over a preset historical period. Based on the relationship between the estimated SOC value and the SOC accuracy division point, a corresponding strategy is adopted to determine the current deviation coefficient according to the current deviation value.

8. The method according to claim 7, characterized in that, The step of determining the current deviation coefficient based on the current deviation value using a corresponding strategy includes: In response to the SOC estimate being greater than the SOC precision division point, the current deviation coefficient is determined based on the current deviation value and the mean of the sample current disturbance. In response to the SOC estimate being less than or equal to the SOC accuracy division point, the current deviation coefficient is determined based on the current deviation value and the maximum value of the sample current disturbance.

9. The method according to claim 8, characterized in that, When the current deviation values ​​are the same, the current deviation coefficient obtained by using the first current deviation coefficient calculation strategy is greater than the current deviation coefficient obtained by using the second current deviation coefficient calculation strategy. The conditions for using the first current deviation coefficient calculation strategy are: the SOC estimate is greater than the SOC accuracy division point, and the SOC prediction deviation is less than or equal to the maximum allowable error of the external reference SOC result. The conditions for using the second current deviation coefficient calculation strategy are: the estimated SOC value is greater than the SOC accuracy division point, and the predicted SOC deviation value is greater than the maximum allowable error of the external reference SOC result.

10. The method according to claim 8, characterized in that, When the estimated SOC value is greater than the SOC accuracy division point, the current deviation coefficient is determined by the following formula. : in, The current deviation value, The mean of the sample current disturbance. This represents the maximum value of the sample current disturbance. This represents the deviation from the SOC estimate. The maximum allowable error for the external reference SOC result. The SOC estimate is... These are the points used to define the precision of the SOC.

11. The method according to claim 8, characterized in that, When the estimated SOC value is less than or equal to the SOC accuracy division point, the value of the current deviation coefficient changes exponentially with the change of the current deviation value relative to the maximum value of the sample current disturbance.

12. The method according to claim 8, characterized in that, When the estimated SOC value is less than or equal to the SOC accuracy division point, the current deviation coefficient is determined by the following formula. : The current deviation value, This represents the maximum value of the sample current disturbance. The SOC estimate is... These are the points used to define the precision of the SOC.

13. The method according to claim 2, characterized in that, The method further includes: Based on the deviation between the SOC estimate and the external reference SOC result, the SOC prediction deviation value is determined; Based on the relationship between the estimated SOC value and the SOC precision division point, and the relationship between the predicted SOC deviation value and the allowable error of SOC, a corresponding calculation strategy is adopted to determine the predicted SOC deviation coefficient based on the predicted SOC deviation value.

14. The method according to claim 13, characterized in that, When the SOC prediction deviation values ​​are the same, the SOC prediction deviation coefficient obtained by using the first SOC prediction deviation coefficient calculation strategy is greater than the SOC prediction deviation coefficient obtained by using other SOC prediction deviation coefficient calculation strategies. The conditions for using the first SOC prediction deviation coefficient calculation strategy are: the SOC estimate is greater than the SOC precision division point, and the SOC prediction deviation is greater than the allowable error of SOC.

15. The method according to claim 13, characterized in that, The step of determining the SOC prediction deviation coefficient based on the SOC prediction deviation value using the corresponding calculation strategy includes: The SOC prediction deviation coefficient is determined using the following formula. : in, The deviation value of the SOC prediction. The mean of the sample SOC error. Maximum SOC error of the sample The allowable error for the SOC, The SOC estimate is... These are the points used to define the precision of the SOC.

16. The method according to claim 5, characterized in that, The uncertainty estimation model is trained using a negative log-likelihood loss function. for: in, Let the uncertainty be... For variance, For the true value, These are predicted values.

17. The method according to claim 1, characterized in that, The first parameter is input into the pre-trained battery SOC model to obtain SOC information, including: In response to the battery management system displaying that the SOC has reached a first preset value, the first parameter is input into the pre-trained battery SOC model to obtain SOC information.

18. The method according to claim 1, characterized in that, The method further includes: The parameters input to the battery SOC model are obtained by using data collected when the SOC is greater than a second preset value.

19. A battery SOC estimation device, characterized in that, include: The acquisition module is used to acquire the first parameters of the battery; The first determining module is used to input the first parameter into a pre-trained battery SOC model to obtain SOC information, wherein the SOC information includes at least an estimated SOC value. The battery SOC model has the ability to compensate for accumulated errors in ampere-hour integration.

20. The apparatus according to claim 19, characterized in that, The SOC information also includes the uncertainty corresponding to the SOC estimate, which is obtained based on the current deviation coefficient and the SOC prediction deviation coefficient. The current deviation coefficient is used to characterize the reliability of the original input features; The SOC prediction deviation coefficient is used to characterize the degree of fluctuation in the deviation between the SOC estimate and the external reference SOC result.

21. The apparatus according to claim 19 or 20, characterized in that, The battery SOC model includes a SOC determination model, which includes a biased temporal neural network and an ampere-hour integral neural network. The first determining module includes: The first determining submodule is used to input the original ampere-hour integral value in the first parameter into the ampere-hour integral neural network to obtain the basic SOC; The second determining submodule is used to input other parameters in the first parameter into the biased temporal neural network to obtain the SOC correction value; The third determining submodule is used to determine the estimated SOC value based on the basic SOC and the SOC correction value; Wherein, the original ampere-hour integral value is a value obtained by directly integrating the current measurement signal without conversion or correction; the other parameters include at least one of the following: current current, current voltage, time-series average current, time-series average voltage, current power, original ampere-hour integral change difference, current cell temperature, and cell temperature at a preset SOC point.

22. The apparatus according to claim 20, characterized in that, The battery SOC model includes an SOC determination model and an uncertainty estimation model; the first determination module includes: The fourth determination submodule is used to input the first parameter into the SOC determination model to obtain the SOC estimate; The fifth determining submodule is used to input the second parameter into the uncertainty estimation model to obtain the current deviation weight and the SOC deviation weight, wherein the second parameter includes the estimated SOC value; The sixth determining submodule is used to determine the uncertainty based on the current deviation weight, the SOC deviation weight, the current deviation coefficient, and the SOC prediction deviation coefficient; The second parameter further includes at least one of the following: current current, current voltage, time-series average current, time-series average voltage, original ampere-hour integral value, original ampere-hour integral change difference, and current cell temperature.

23. The apparatus according to claim 20, characterized in that, The device further includes: The second determining module is used to determine the current deviation value based on the difference between the current and the average current value within a preset historical time period; and to determine the current deviation coefficient based on the current deviation value using a corresponding strategy according to the relationship between the SOC estimate and the SOC precision division point.

24. The apparatus according to claim 20, characterized in that, The device further includes: The third determining module is used to determine the SOC prediction deviation value based on the deviation between the SOC estimate and the external reference SOC result; and to determine the SOC prediction deviation coefficient based on the relationship between the SOC estimate and the SOC precision division point, and the relationship between the SOC prediction deviation value and the allowable error of SOC, using a corresponding calculation strategy.

25. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to execute the executable instructions in the memory to implement the steps of the battery SOC estimation method according to any one of claims 1-18.

26. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the battery SOC estimation method according to any one of claims 1-18.

27. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the steps of the battery SOC estimation method according to any one of claims 1-18.

Images