Battery health state estimation method and system, electronic equipment and storage medium
By building a deep learning model in the cloud, using deep convolutional neural network and gated network to extract battery aging characteristics, the stability and accuracy of lithium-ion battery health status estimation under complex working conditions of the vehicle are solved, and high-precision Battery health status estimate.
Patent Information
- Application Number
- CN202411075886.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-26
AI Technical Summary
The prior art is difficult to accurately estimate the health status of lithium-ion batteries in complex vehicle-side working conditions, resulting in poor estimation stability and accuracy.
Using deep learning algorithms to build a battery health state estimation model in the cloud, extract key features related to battery aging through improved deep convolutional neural networks and gated networks, and combine the residual network to alleviate the degradation and gradient disappearance problems to establish end-to-end batteries Health status estimate model.
Under actual operating conditions, high-precision and reliable estimation of the battery's health status are achieved, and the generalization ability and scope of application are improved.
Smart Images

Figure CN119026468A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of battery status evaluation, and in particular to a battery health status estimation method, system, electronic device and storage medium. Background Art
[0002] With the rapid development of new energy vehicles, lithium-ion batteries are widely used in vehicle power supply due to their high energy density, low self-discharge rate, long cycle life, energy saving and environmental protection. It is extremely important to accurately evaluate the battery health (SOH) of lithium-ion batteries and realize accurate health management of lithium-ion batteries. However, when lithium-ion batteries are installed in vehicles, the maximum available capacity of the battery cannot be directly measured. At the same time, there are many complex chemical reactions and physical changes inside lithium-ion batteries, such as lithium ion migration, structural changes of electrode materials, and loss of active materials. In addition, the aging mechanism of lithium-ion batteries is very complex. Its performance gradually decays nonlinearly with the use of cycles and calendars. Its health status is affected by ambient temperature, charge and discharge rate, and charge and discharge depth. These factors make the estimation of SOH extremely complicated.
[0003] At present, two solutions are usually used to estimate SOH: battery model-based and data-driven. The battery model-based solution uses a resistor-capacitor circuit model to estimate the terminal voltage through parameter identification, thereby identifying the SOH of the lithium battery. However, the complex chemical and physical processes of the battery cannot be accurately expressed by the equivalent circuit model, and as the battery ages, the identification parameters will change significantly, resulting in a significant decrease in the stability and accuracy of the SOH estimation. The data-driven solution mainly analyzes the aging data of the battery cells in the laboratory, establishes a mapping relationship between the cell and time, and extracts the battery portrait to estimate the SOH. Or extract important factors related to the degree of aging during the charging and discharging process, establish the relationship between the important factors and SOH, and thus estimate the battery's SOH. Or use a neural network model based on laboratory data to establish a mapping relationship between SOH and features through deep learning methods to predict SOH. However, under laboratory operating conditions, it is impossible to generalize to complex operating scenarios on the vehicle side, and it is impossible to predict the SOH of the battery over its entire life cycle, and stability and accuracy are still difficult to guarantee. Summary of the invention
[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a battery health status estimation method, system, electronic device and storage medium, which effectively solve the problems of poor stability and accuracy of battery health status estimation under complex actual working conditions on the vehicle side.
[0005] In a first aspect, the present invention provides a method for estimating a battery health state, the method comprising:
[0006] Acquire actual charging condition data of the battery from the cloud, and obtain an actual charging curve according to the actual charging condition data, wherein the actual charging curve includes at least an actual voltage curve and an actual temperature curve;
[0007] Extracting a voltage segment curve from the actual voltage curve according to a set threshold, converting the voltage segment curve to obtain an incremental capacity segment curve, and extracting a temperature segment curve from the actual temperature curve;
[0008] normalizing the voltage segment curve, the incremental capacity segment curve, and the temperature segment curve;
[0009] The processed voltage segment curve, the processed incremental capacity segment curve and the processed temperature segment curve are input into the battery state estimation model in the cloud, and the battery health state value of the battery is obtained through the battery state estimation model.
[0010] Furthermore, before obtaining the actual charging condition data of the battery from the cloud, the method further includes:
[0011] Construct an initial battery state estimation model;
[0012] Acquire experimental charging condition data of the battery, train the initial battery state estimation model according to the experimental charging condition data, and obtain a battery state estimation model;
[0013] The battery state estimation model is packaged into a function file, and the function file is deployed to the cloud.
[0014] Furthermore, the initial battery state estimation model includes an input layer, a deep gated convolutional neural network layer, a residual layer, a fully connected layer and an output layer, wherein:
[0015] The input layer acquires the voltage segment curve, the incremental capacity segment curve and the temperature segment curve;
[0016] The deep gated convolutional neural network layer extracts local key features related to battery aging in the segment curve;
[0017] The residual layer deepens the network in the deep gated convolutional neural network layer;
[0018] The fully connected layer reintegrates the local key features to obtain global key features;
[0019] The output layer outputs a battery health curve and uses labels for supervised learning.
[0020] Furthermore, the deep gated convolutional neural network layer includes a convolution module, a gated convolution module, a pooling module, a regularization module and a feature splicing module, wherein:
[0021] The convolution module performs interval scanning and inner product operation on the fragment curve to extract the local characteristic factors of the fragment curve;
[0022] The gated convolution module is used to perform a gated inner product operation to control the temporal memory of the weight of the local characteristic factor to obtain battery aging related characteristics;
[0023] The pooling module reduces the characteristic dimensions of the battery aging-related characteristics and extracts characteristic factors;
[0024] The regularization module performs regularization processing on the characteristic parameters to reduce the dependency of neurons in the network;
[0025] The feature splicing module performs feature splicing on the battery aging related features to obtain new local key features.
[0026] Further, the acquiring of experimental charging condition data of the battery, training the initial battery state estimation model according to the experimental charging condition data, and obtaining the battery state estimation model includes:
[0027] Acquiring the experimental charging condition data of the battery, and generating an experimental charging curve according to the experimental charging condition data;
[0028] Extracting an experimental voltage segment curve, an experimental incremental capacity segment curve, and an experimental temperature segment curve according to the experimental charging curve to obtain a training set;
[0029] Training and validating the initial battery state estimation model using a cross-training validation method according to the training set;
[0030] The trained initial battery state estimation model is evaluated using a root mean square error method to obtain a battery state estimation model.
[0031] Furthermore, the actual charging condition data of the battery is obtained from the cloud, and the actual charging curve is obtained according to the actual charging condition data, including:
[0032] Acquire the actual charging condition data of the battery from the cloud, and generate an initial charging curve according to the actual charging condition data;
[0033] The initial charging curve is subjected to abnormal values, illegal values and repeated values elimination, and missing values completion to obtain the actual charging curve.
[0034] Further, the extracting a voltage segment curve from the actual voltage curve, converting the incremental capacity segment curve according to the voltage segment curve, and extracting a temperature segment curve from the actual temperature curve includes:
[0035] Setting the interval point, extraction point threshold and segment voltage range threshold of the actual voltage curve;
[0036] extracting the voltage segment curve from the actual voltage curve according to the extraction point threshold and the segment voltage range threshold;
[0037] Obtaining voltage values and capacity values of the interval points in the voltage segment curve, and converting the voltage curve into an incremental capacity segment curve;
[0038] The temperature segment curve is extracted from the actual temperature curve according to the voltage segment curve.
[0039] In a second aspect, the present invention provides a battery health status estimation system, the system comprising:
[0040] A data processing module, used to obtain actual charging condition data of the battery from the cloud, and obtain an actual charging curve according to the actual charging condition data, wherein the actual charging curve at least includes an actual voltage curve and an actual temperature curve;
[0041] A curve generating module, configured to extract a voltage segment curve from the actual voltage curve according to a set threshold, convert the voltage segment curve to obtain an incremental capacity segment curve, and extract a temperature segment curve from the actual temperature curve;
[0042] A curve processing module, used for normalizing the voltage segment curve, the incremental capacity segment curve and the temperature segment curve;
[0043] The state estimation module is used to input the processed voltage segment curve, the processed incremental capacity segment curve and the processed temperature segment curve into the battery state estimation model in the cloud, and obtain the battery health state value of the battery through the battery state estimation model.
[0044] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the battery health status estimation method as described in the first aspect of the present invention.
[0045] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the battery health status estimation method as described in the first aspect of the present invention.
[0046] The battery health state estimation method, system, electronic device and storage medium provided by the present invention estimate the health state value of the battery in the cloud based on a deep learning algorithm, and realize end-to-end battery health state estimation based on a fragment timing charging curve. The method uses an improved deep convolutional neural network and a gated network to control the temporal memory of the input fragment curve, and uses multi-layer gated convolution to automatically extract key features related to battery aging from the fragment curve. The residual network further improves the expression ability of the model and alleviates the model degradation and gradient vanishing problems, further promotes the training of the deep convolutional neural network, and finally establishes an end-to-end battery health state estimation model. The model is deployed to the cloud engineering application, and the accuracy and reliability of the battery health state estimation under actual operating conditions are excellent, the generalization ability is stronger, and the scope of application is wider. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.
[0048] Figure 1 This is a first schematic diagram of the battery health status estimation method process provided by an embodiment of the present invention;
[0049] Figure 2 is a schematic diagram of the structure of an initial battery state estimation model provided by an embodiment of the present invention;
[0050] Figure 3 is a second schematic diagram of the battery health status estimation method process provided by an embodiment of the present invention;
[0051] Figure 4 is a schematic diagram of a voltage segment curve provided by an embodiment of the present invention;
[0052] Figure 5 is a schematic diagram of an incremental capacity segment curve provided by an embodiment of the present invention;
[0053] Figure 6 is a curve distribution diagram of predicted values and true values in an embodiment of the present invention;
[0054] Figure 7 Schematic diagram of the error distribution between the predicted value and the true value in the embodiment of the present invention;
[0055] Figure 8 is a schematic diagram of the distribution of absolute error and statistical times in an embodiment of the present invention;
[0056] Fig. 9is a third schematic diagram of the battery health status estimation method process provided by an embodiment of the present invention;
[0057] Fig.10 is a fourth schematic diagram of a battery health status estimation method process provided by an embodiment of the present invention;
[0058] Fig.11 is a fifth schematic diagram of the battery health status estimation method process provided by an embodiment of the present invention;
[0059] Fig.12 is a schematic diagram of the error distribution between the real value and the predicted value under actual working conditions in an embodiment of the present invention;
[0060] Fig.13 is a schematic diagram of the distribution of absolute error and statistical times under actual working conditions in an embodiment of the present invention;
[0061] Fig.14 is a schematic diagram of the structure of a battery health status estimation system provided by an embodiment of the present invention;
[0062] Fig.15 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention.
[0063] Description of main component symbols:
[0064] 200, battery health status estimation system; 210, data processing module; 220, curve generation module; 230, curve processing module; 240, status estimation module; 300, electronic device; 310, processor; 320, communication interface; 330, memory; 340, communication bus. DETAILED DESCRIPTION
[0065] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be further clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. It should be noted that the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0066] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0067] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.
[0068] At present, two solutions, battery model-based and data-driven, are usually used to estimate the battery state of health (SOH). The battery model-based solution uses a resistor-capacitor circuit model to estimate the terminal voltage through parameter identification, thereby identifying the battery's SOH. However, the complex chemical and physical processes of the battery cannot be accurately expressed by the equivalent circuit model, and as the battery ages, the identification parameters will change significantly, resulting in a significant decrease in the stability and accuracy of the SOH estimation. Based on the data-driven solution, the aging data of the battery cells is analyzed in the laboratory to establish a mapping relationship between the cell and time, and extract the battery portrait to estimate the SOH. Or extract the important factors related to the degree of aging during the charging and discharging process, establish the relationship between the important factors and the SOH, and estimate the battery's SOH. Or use a neural network model through laboratory data, and establish a mapping relationship between SOH and features through deep learning methods to predict SOH. However, under laboratory operating conditions, it is impossible to generalize to the complex operating conditions of the vehicle, and it is impossible to predict the SOH of the battery over its entire life cycle, and stability and accuracy are still difficult to guarantee.
[0069] Example 1
[0070] In view of the above-mentioned deficiencies in the prior art, an embodiment of the present invention provides a battery health status estimation method, which effectively solves the problems of poor stability and accuracy of battery health status estimation under complex actual working conditions on the vehicle side. Figure 1 FIG. 1 is a first schematic diagram of a method for estimating a battery health state provided by an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:
[0071] S100: construct an initial battery state estimation model.
[0072] Optionally, the initial battery state estimation model of an embodiment of the present invention is constructed based on a convolutional neural network, which is particularly suitable for processing data with a grid structure, can automatically and adaptively learn hierarchical structure features, and can capture local features in sequence data.
[0073] The initial battery state estimation model of the embodiment of the present invention adds a multi-layer gated convolutional network (Gated CNN) network on the basis of the convolutional neural network, and controls the input timing memory through different activation functions to obtain the key timing information related to battery aging hidden in the fragment curve. Figure 2 is a schematic diagram of the structure of the initial battery state estimation model provided by an embodiment of the present invention, such as Figure 2 As shown, the initial battery state estimation model includes an input layer, a deep gated convolutional neural network layer, a residual layer, a fully connected layer and an output layer, where:
[0074] The input layer obtains fragment curves such as voltage fragment curve, incremental capacity fragment curve and temperature fragment curve.
[0075] The deep gated convolutional neural network layer extracts local key features related to battery aging in the fragment curve. The deep gated convolutional neural network layer includes a convolution module, a gated convolution module, a pooling module, a regularization module and a feature splicing module, where:
[0076] The convolution module uses multiple convolution kernels to scan the fragment curve and perform inner product operations to extract the local feature factors of the fragment curve, while reducing the feature dimension and suppressing input noise. The maximum pooling layer is used to extract local key factors, effectively reducing complexity and overfitting risks, while strengthening important features and suppressing non-important features.
[0077] The gated convolution module is a multi-level structure, and the convolution layer is divided into two layers, upper and lower. Specifically, let the feature matrix E = D w1 , where D w1 It represents a global feature matrix composed of multiple feature matrices in a sequential relationship. After layering the feature matrix E, two feature matrices A and B are obtained, where:
[0078] A=E×W+b
[0079] B=E×V+c
[0080] In the above formula, W and b represent the weight parameters of feature matrix A, V and c represent the weight parameters of feature matrix B. Feature matrix A and feature matrix B are input into different activation functions respectively, and gated inner product operations are performed respectively to control the temporal memory of feature weights. The formula of gated inner product operation is as follows:
[0081]
[0082] In the above formula, H0 represents the result matrix of gated convolution, and σ represents the activation function of the feature matrix B.
[0083] The activation function of the feature matrix A is:
[0084] tanh(x)=sinh(x) / cosh(x)=(e x -e -x ) / (e x +e -x )
[0085] The activation function of the feature matrix B is:
[0086] S(x)=1 / (1+e -x )
[0087] In the above formulas, tanh(x) represents the activation function of the feature matrix A, S(x) represents the activation function of the feature matrix B, e represents the base of the natural logarithm function, and x represents the function independent variable.
[0088] The tanh(x) activation function maps the input value to the range of [-1,1], improving the symmetry of the model output and reducing the deviation when updating the weights. Its nonlinearity provides the neural network with more powerful expression capabilities and reduces the risk of gradient explosion. The S(x) activation function is used to implement the gating mechanism, allowing the network to dynamically select important features to enhance expression and suppress unimportant features. It controls the temporal memory of weights through inner product operations to extract battery aging-related features. At the same time, it can better express the long-term dependencies in the segment curve and reduce the risk of overfitting.
[0089] The pooling module is used to reduce the complexity of the model, reduce the feature dimensions of battery aging-related features, further extract important feature factors, and suppress noise information.
[0090] The regularization module regularizes the feature parameters by randomly activating or eliminating neurons in the network during training, reducing the complex co-adaptation relationship between neurons, thereby improving the generalization ability of the model and the speed of model convergence.
[0091] The feature splicing module is used to splice the battery aging-related features output by different gated convolutional layers and recombine them into new local key features, effectively improving the expressiveness of the model and thus improving the accuracy of the model.
[0092] The residual layer encapsulates the convolutional module in the deep gated convolutional neural network layer and the deep gated convolutional neural network into a residual module through skip-layer connections. Multiple residual modules are established for stacking and connection, which can build a very deep network, maintain the stability and efficiency of network training, and improve the model's expressiveness. At the same time, the residual connection allows the gradient to be directly back-propagated through shortcuts, effectively alleviating the problems of gradient vanishing and gradient explosion. The residual block also makes it easier for the network to learn the identity mapping, which is used to reduce model degradation and greatly improves the model convergence efficiency and model learning ability.
[0093] The fully connected layer reintegrates the local key features after a series of convolution, pooling, and gating operations to obtain global key features and updates the parameters through back propagation.
[0094] The output layer outputs the battery health curve and uses labels for supervised learning. The feature parameters are continuously updated through back propagation, thereby encouraging the model to iterate step by step and find the optimal feature parameters.
[0095] The initial battery state estimation model of the embodiment of the present invention captures the local features of the curve by sliding the convolution kernel on the fragment curve, constructs a unique mapping relationship between the local common features and the battery health status, and can process multiple time inputs in parallel to process longer sequence curve data. At the same time, the weight sharing mechanism of convolution greatly reduces the parameters and significantly improves the training efficiency. Through convolution kernels of different sizes, multi-scale feature learning and capture can be achieved. When processing local fine features, due to position independence, similar features at different positions can be extracted, which greatly improves the generalization ability of the model and effectively prevents overfitting.
[0096] S200, obtaining experimental charging condition data of the battery, training an initial battery state estimation model according to the experimental charging condition data, and obtaining a battery state estimation model;
[0097] Figure 3 FIG. 2 is a second schematic diagram of a method for estimating a battery health state provided by an embodiment of the present invention. Figure 3 As shown, the training of the battery state estimation model includes the following steps:
[0098] S210: Acquire experimental charging condition data of the battery, and generate an experimental charging curve according to the experimental charging condition data.
[0099] Obtain cycle experimental data of cycle tests in the laboratory under different working conditions, where different working conditions include cycle conditions, calendar conditions, and cycle calendar conditions. In an embodiment of the present invention, cycle conditions refer to repeatedly charging the battery cell in a laboratory environment, and discharging the battery cell after standing for more than 2 hours to accelerate the aging of the battery cell; calendar conditions refer to charging the battery cell and standing it for more than 15 days at a high temperature of 45°C; cycle calendar conditions refer to repeatedly charging the battery cell and discharging it after standing for more than 15 days at a high temperature of 45°C. By setting different pretreatment conditions under different working conditions to perform cyclic charging and discharging, the battery health status value is aged from 100% to less than 80%, and the experimental charge and discharge condition data of the battery are obtained. In an embodiment of the present invention, different pretreatment conditions include 0.1C, 0.3C and 1C, where C represents the charge and discharge rate.
[0100] The charging, discharging and static events are divided according to the charging and discharging currents and experimental conditions, and the charging curve is generated with the charging event as the research object. The charging curve is preprocessed. The preprocessing mainly includes eliminating abnormal values, illegal values, and repeated values, and linearly interpolating the missing values of the time series to fill the missing values, so as to obtain the experimental charging curve for model input, which at least includes a voltage curve and a temperature curve.
[0101] S220 , extracting an experimental voltage segment curve, an experimental incremental capacity segment curve, and an experimental temperature segment curve according to the experimental charging curve to obtain a training set.
[0102] Figure 4 It is a schematic diagram of a voltage segment curve provided in an embodiment of the present invention. Assuming that the charging voltage range of the battery cell is 2.8V to 4.2V, corresponding to SOC charging from 0% to 100%, since most batteries cannot be charged from 0% to 100% in actual working conditions in the cloud, the embodiment of the present invention only needs one segment of the complete charging curve, such as charging from 3.6V to 3.8V across 200mV, without the entire charging curve, thereby increasing the generalization performance.
[0103] The interval points, extraction point threshold and segment voltage range threshold of the voltage curve are set, and then a charging segment curve is extracted from the voltage curve at intervals of a certain number of points according to the extraction point threshold. If the segment voltage range threshold is X mV, each voltage segment curve is a charging curve spanning X mV. In the embodiment of the present invention, the extraction point threshold is set to 50, that is, a voltage segment curve is extracted every 50 interval points, and the segment voltage range threshold is set to 200mV, that is, the extracted voltage segment curve is a voltage curve spanning 200mV.
[0104] Figure 5 : is a schematic diagram of an incremental capacity segment curve provided by an embodiment of the present invention. The voltage value and capacity value of each interval point in the voltage segment curve are obtained, and the voltage segment curve is converted into an incremental capacity segment curve. The conversion formula is as follows:
[0105] dqdv=dq / dv=(Q end -Q strat ) / (V end -V strat )
[0106] In the above formula, dqdv represents the incremental capacity, dq represents the capacitance increment, dv represents the voltage increment, Q end and Q start Respectively represent the starting capacity and ending capacity of each interval point, V end and V start Respectively represent the starting voltage and ending voltage of each interval point.
[0107] The temperature values under the voltage values represented by all interval points of the voltage segment curve are obtained, and the temperature segment curves of the same interval points are extracted from the temperature curve.
[0108] Since the scales of the three segment curves are quite different, normalization to different degrees is required. In the embodiment of the present invention, the three segment curves are respectively normalized after the values are extracted. The normalization methods include but are not limited to one or more methods of [0, 1] normalization, [-1, 1] normalization and standard normalization, among which:
[0109] The calculation formula for [0, 1] normalization is as follows:
[0110] X new1 =(X i -X min ) / (X max -X min )
[0111] The calculation formula for [-1, 1] normalization is as follows:
[0112] X new2 =(X i -X mean ) / (X max -X min )
[0113] The calculation formula for standard normalization is as follows:
[0114] X new3 =(X i -X mean ) / X std
[0115] In the above formula, X new1 represents the updated value after normalization to [0, 1], X new2 represents the updated value after normalization to [-1, 1], X new3 represents the updated value after standard normalization, X i is the eigenvalue of each column in the feature matrix. In the embodiment of the present invention, three columns of features are included, namely, the experimental voltage segment curve, the experimental incremental capacity segment curve and the experimental temperature segment curve. min represents the minimum value of the matrix, X max Represents the maximum value of the matrix, X mean represents the matrix mean, X std Represents the standard deviation of the matrix.
[0116] Optionally, the parameters are adaptively adjusted and an optimal normalization method is selected to perform normalization processing to obtain a training set.
[0117] S230 , training and verifying the initial battery state estimation model using a cross-training verification method according to the training set.
[0118] The initial battery state estimation model is trained using K-fold cross-validation based on the training set. The entire training set is randomly divided into K subsets, while ensuring that the distribution of each subset is similar to the original training set, and the training and validation process is performed K times. In each iteration, one of the subsets is selected as the validation set, and the remaining K-1 subsets are used as training sets for model training. This process is repeated K times, and a different subset is selected as the validation set each time to ensure that each sample is used as a validation set once. Through multiple training and validation, the performance of the model on unknown data can be more accurately evaluated, avoiding the deviation caused by using only one division. Since each sample is used as a validation set, it is more difficult for the model to overfit the training data during the training process, thereby improving the generalization ability of the model and reducing overfitting. At the same time, when the neural network is back-propagating, the initialized gradient uses multiple random assignments to reduce model contingency.
[0119] S240, using a root mean square error method to evaluate the trained initial battery state estimation model to obtain a battery state estimation model.
[0120] In the embodiment of the present invention, the root mean square error is used to measure the difference between the model prediction value and the true value. The calculation formula of the root mean square error is as follows:
[0121]
[0122] In the above formula, m represents the sample size of battery cells, that is, the predicted number of battery cells, y i represents the true battery health status value of the i-th battery cell sample, y pred_i Represents the predicted battery health status value of the i-th battery cell sample.
[0123] If the RMSE of the initial battery state estimation model obtained after K trainings is ≤ 1% and the error distribution is similar, the model with the smallest RMSE is selected as the battery state estimation model. The battery state estimation model is used to output the predicted value of the test battery, and the actual value of the corresponding test battery is tested in a laboratory environment. Figure 6 is a curve distribution diagram of the predicted value and the true value in the embodiment of the present invention, such as Figure 6 As shown, the shaded part is the space of plus or minus 1%. From the figure, it can be seen that all the predicted value points fall within the range of ±1% of the true value; Figure 7 Schematic diagram of the error distribution between the predicted value and the true value in the embodiment of the present invention. Figure 7 As shown, from 80% to 100%, all points are concentrated on the straight line y=x, that is, the predicted value is extremely close to the true value; Figure 8: is a schematic diagram of the distribution of absolute error and statistical times in an embodiment of the present invention, such as Figure 8 As shown in the figure, the maximum error is ≤1%, more than 95% of the errors are less than 0.5%, and the RMSE of the validation set is ≤0.5%, indicating that the prediction effect on the validation set is extremely good and the model has excellent accuracy.
[0124] S300: Encapsulate the battery state estimation model into a function file and deploy the function file to the cloud. In an embodiment of the present invention, the battery state estimation model is encapsulated into an hdf5 format file and the hdf5 format file is deployed to a cloud server for estimating the battery health state value.
[0125] Fig. 9 FIG. 3 is a third schematic diagram of a method for estimating a battery health state provided by an embodiment of the present invention. Fig. 9 As shown, the method also includes the following steps:
[0126] S400: Acquire actual charging condition data of the battery from the cloud, and obtain an actual charging curve according to the actual charging condition data, where the actual charging curve at least includes an actual voltage curve and an actual temperature curve.
[0127] Fig.10 4 is a schematic diagram of a battery health status estimation method according to an embodiment of the present invention. Fig.10 As shown, the acquisition of the actual charging curve is similar to the acquisition of the experimental charging curve, which specifically includes the following steps:
[0128] S410: Acquire actual charging condition data of the battery from the cloud, and generate an initial charging curve according to the actual charging condition data.
[0129] S420: Remove abnormal values, illegal values and duplicate values from the initial charging curve, and complete missing values to obtain an actual charging curve.
[0130] The actual charging curve at least includes an actual voltage curve and an actual temperature curve.
[0131] S500 , extracting a voltage segment curve from the actual voltage curve according to a set threshold, converting the voltage segment curve to obtain an incremental capacity segment curve, and extracting a temperature segment curve from the actual temperature curve.
[0132] Fig.11 5 is a fifth schematic diagram of a method for estimating a battery health state provided by an embodiment of the present invention. Fig.11 As shown, the acquisition of the three actual fragment curves is similar to the acquisition of the experimental fragment curve, which specifically includes the following steps:
[0133] S510 , setting interval points, extraction point thresholds, and segment voltage range thresholds of the actual voltage curve.
[0134] In the embodiment of the present invention, the interval point of the actual voltage curve, the extraction point threshold and the segment voltage range threshold can be set according to the actual situation of the battery.
[0135] S520 , extracting a voltage segment curve from the actual voltage curve according to the extraction point threshold and the segment voltage range threshold.
[0136] S530, obtaining voltage values and capacity values of interval points in the voltage segment curve, and converting the voltage segment curve into an incremental capacity segment curve;
[0137] S540 , extracting a temperature segment curve from the actual temperature curve according to the voltage segment curve.
[0138] In actual working conditions on the cloud, each battery pack is usually composed of multiple cells connected in series. For example, 96 cells, the same operation is performed 96 times to extract segment curve data of all cells in a battery pack.
[0139] S600 , normalizing the voltage segment curve, the incremental capacity segment curve, and the temperature segment curve.
[0140] In the embodiment of the present invention, one or more of the three normalization methods of [0, 1] normalization, [-1, 1] normalization and standard normalization in step S220 are used to normalize the three segment curves.
[0141] S700: Input the processed voltage segment curve, the processed incremental capacity segment curve, and the processed temperature segment curve into a battery state estimation model in the cloud, and obtain a battery health state value of the battery through the battery state estimation model.
[0142] In the embodiment of the present invention, the segment curve data of all cells in a battery pack is input into the battery state estimation model packaged in the cloud, the battery state estimation model is called and the battery health state value of each cell is output to complete the battery health state value estimation at the cell level. The segment curve of the battery pack is extracted according to the actual working conditions and input into the cloud model for prediction to obtain the predicted value. The data returned by the battery pack to the laboratory for calibration is taken as the true value, and the data of more than 100 battery packs are statistically counted. Fig.12 is a schematic diagram of the error distribution between the true value and the predicted value under actual working conditions in an embodiment of the present invention, Fig.13 : is a schematic diagram of the distribution of absolute error and statistical times under actual working conditions in an embodiment of the present invention, such as Fig.12 and Fig.13 As shown, after verification of a large number of battery samples under actual operating conditions, the RMSE is ≤1.5%, and the maximum does not exceed 2%, which is an improvement of more than 50% compared with the existing methods, with excellent accuracy and reliability.
[0143] The battery health status estimation method provided in the embodiment of the present invention estimates the battery health status value in the cloud based on a deep learning algorithm, and realizes end-to-end battery health status estimation based on a fragment timing charging curve. The method uses an improved deep convolutional neural network and a gated network to control the temporal memory of the input fragment curve, and uses multi-layer gated convolution to automatically extract key features related to battery aging from the fragment curve. The residual network further improves the expression ability of the model and alleviates the model degradation and gradient vanishing problems, further promotes the training of the deep convolutional neural network, and finally establishes an end-to-end battery health status estimation model.
[0144] Example 2
[0145] Based on the same technical concept as the method embodiment of the above embodiment 1, the embodiment of the present invention provides a battery health status estimation system. Fig.14 FIG. 1 is a schematic diagram of the structure of a battery health status estimation system provided by an embodiment of the present invention. Fig.14 As shown, the battery health status estimation system 200 includes:
[0146] The data processing module 210 is used to obtain the actual charging condition data of the battery from the cloud, and obtain the actual charging curve according to the actual charging condition data. The actual charging curve at least includes an actual voltage curve and an actual temperature curve.
[0147] The curve generating module 220 is used to extract a voltage segment curve from the actual voltage curve according to a set threshold, obtain an incremental capacity segment curve according to the voltage segment curve, and extract a temperature segment curve from the actual temperature curve.
[0148] The curve processing module 230 is used to perform normalization processing on the voltage segment curve, the incremental capacity segment curve and the temperature segment curve.
[0149] The state estimation module 240 is used to input the processed voltage segment curve, the processed incremental capacity segment curve and the processed temperature segment curve into the battery state estimation model in the cloud, and obtain the battery health state value of the battery through the battery state estimation model.
[0150] The battery health state estimation system provided by the embodiment of the present invention has excellent accuracy and reliability in battery health state estimation under actual operating conditions, stronger generalization ability, and wider scope of application.
[0151] It can be understood that the implementation method of the battery health status estimation method described in the above embodiment 1 is also applicable to this embodiment and can achieve the same technical effect, so it will not be repeated here.
[0152] Example 3
[0153] Based on the same concept, an embodiment of the present invention further provides an electronic device, Fig.15 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Fig.15 As shown, the electronic device 300 may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute the steps of the battery health status estimation method described in the above embodiments. For example, it includes:
[0154] S100, constructing an initial battery state estimation model;
[0155] S200, obtaining experimental charging condition data of the battery, training an initial battery state estimation model according to the experimental charging condition data, and obtaining a battery state estimation model;
[0156] S300, encapsulating the battery state estimation model into a function file, and deploying the function file to the cloud;
[0157] S400, obtaining actual charging condition data of the battery from the cloud, and obtaining an actual charging curve according to the actual charging condition data, where the actual charging curve at least includes an actual voltage curve and an actual temperature curve;
[0158] S500, extracting a voltage segment curve from the actual voltage curve according to a set threshold, converting the voltage segment curve to obtain an incremental capacity segment curve, and extracting a temperature segment curve from the actual temperature curve;
[0159] S600, normalizing the voltage segment curve, the incremental capacity segment curve, and the temperature segment curve;
[0160] S700: Input the processed voltage segment curve, the processed incremental capacity segment curve, and the processed temperature segment curve into a battery state estimation model in the cloud, and obtain a battery health state value of the battery through the battery state estimation model.
[0161] The processor 310 may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.
[0162] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0163] The memory 330 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0164] Example 4
[0165] Based on the same concept, an embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program, and the computer program includes at least one code segment, which can be executed by a main control device to control the main control device to implement the steps of the battery health status estimation method described in the above embodiments. For example, it includes:
[0166] S100, constructing an initial battery state estimation model;
[0167] S200, obtaining experimental charging condition data of the battery, training an initial battery state estimation model according to the experimental charging condition data, and obtaining a battery state estimation model;
[0168] S300, encapsulating the battery state estimation model into a function file, and deploying the function file to the cloud;
[0169] S400, obtaining actual charging condition data of the battery from the cloud, and obtaining an actual charging curve according to the actual charging condition data, where the actual charging curve at least includes an actual voltage curve and an actual temperature curve;
[0170] S500, extracting a voltage segment curve from the actual voltage curve according to a set threshold, converting the voltage segment curve to obtain an incremental capacity segment curve, and extracting a temperature segment curve from the actual temperature curve;
[0171] S600, normalizing the voltage segment curve, the incremental capacity segment curve, and the temperature segment curve;
[0172] S700: Input the processed voltage segment curve, the processed incremental capacity segment curve, and the processed temperature segment curve into a battery state estimation model in the cloud, and obtain a battery health state value of the battery through the battery state estimation model.
[0173] Based on the same technical concept, an embodiment of the present invention further provides a computer program, which is used to implement the above method embodiment when executed by a main control device.
[0174] The computer program may be stored in whole or in part on a computer-readable storage medium packaged with the processor, or may be stored in whole or in part on a memory not packaged with the processor.
[0175] Based on the same technical concept, an embodiment of the present invention further provides a processor, which is used to implement the above method embodiment. The above processor can be a chip.
[0176] In summary, the battery health status estimation method provided by the embodiment of the present invention estimates the battery health status value in the cloud based on a deep learning algorithm, and realizes end-to-end battery health status estimation based on a fragment timing charging curve. This method uses an improved deep convolutional neural network and a gated network to control the temporal memory of the input fragment curve, and uses multi-layer gated convolution to automatically extract key features related to battery aging from the fragment curve. The residual network further improves the expression ability of the model and alleviates the model degradation and gradient vanishing problems, further promotes the training of deep convolutional neural networks, and finally establishes an end-to-end battery health status estimation model.
[0177] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0178] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.
[0179] Finally, it should be noted that 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for estimating a battery health state, characterized in that: The method comprises: Acquire actual charging condition data of the battery from the cloud, and obtain an actual charging curve according to the actual charging condition data, wherein the actual charging curve includes at least an actual voltage curve and an actual temperature curve; Extracting a voltage segment curve from the actual voltage curve according to a set threshold, converting the voltage segment curve to obtain an incremental capacity segment curve, and extracting a temperature segment curve from the actual temperature curve; normalizing the voltage segment curve, the incremental capacity segment curve, and the temperature segment curve; The processed voltage segment curve, the processed incremental capacity segment curve and the processed temperature segment curve are input into the battery state estimation model in the cloud, and the battery health state value of the battery is obtained through the battery state estimation model.
2. The battery health status estimation method according to claim 1, characterized in that: Before obtaining the actual charging condition data of the battery from the cloud, the method further includes: Construct an initial battery state estimation model; Acquire experimental charging condition data of the battery, train the initial battery state estimation model according to the experimental charging condition data, and obtain a battery state estimation model; The battery state estimation model is packaged into a function file, and the function file is deployed to the cloud.
3. The battery health status estimation method according to claim 2, characterized in that: The initial battery state estimation model includes an input layer, a deep gated convolutional neural network layer, a residual layer, a fully connected layer and an output layer, wherein: The input layer acquires the voltage segment curve, the incremental capacity segment curve and the temperature segment curve; The deep gated convolutional neural network layer extracts local key features related to battery aging in the segment curve; The residual layer deepens the network in the deep gated convolutional neural network layer; The fully connected layer reintegrates the local key features to obtain global key features; The output layer outputs a battery health curve and uses labels for supervised learning.
4. The battery health status estimation method according to claim 3, characterized in that: The deep gated convolutional neural network layer includes a convolution module, a gated convolution module, a pooling module, a regularization module and a feature splicing module, wherein: The convolution module performs interval scanning and inner product operation on the segment curve to extract local characteristic factors of the segment curve; The gated convolution module is used to perform a gated inner product operation to control the temporal memory of the weight of the local characteristic factor to obtain battery aging related characteristics; The pooling module reduces the characteristic dimensions of the battery aging-related characteristics and extracts characteristic factors; The regularization module performs regularization processing on the characteristic parameters to reduce the dependency of neurons in the network; The feature splicing module performs feature splicing on the battery aging related features to obtain new local key features.
5. The battery health status estimation method according to claim 2, characterized in that: The step of acquiring experimental charging condition data of the battery, training the initial battery state estimation model according to the experimental charging condition data, and obtaining the battery state estimation model comprises: Acquiring the experimental charging condition data of the battery, and generating an experimental charging curve according to the experimental charging condition data; Extracting an experimental voltage segment curve, an experimental incremental capacity segment curve, and an experimental temperature segment curve according to the experimental charging curve to obtain a training set; Training and validating the initial battery state estimation model using a cross-training validation method according to the training set; The trained initial battery state estimation model is evaluated using a root mean square error method to obtain a battery state estimation model.
6. The battery health status estimation method according to claim 1, characterized in that: The obtaining of actual charging condition data of the battery from the cloud, and obtaining an actual charging curve according to the actual charging condition data, includes: Acquire the actual charging condition data of the battery from the cloud, and generate an initial charging curve according to the actual charging condition data; The initial charging curve is subjected to abnormal values, illegal values and repeated values elimination, and missing values completion to obtain the actual charging curve.
7. The battery health status estimation method according to claim 6, characterized in that: The extracting a voltage segment curve from the actual voltage curve, converting the voltage segment curve to obtain an incremental capacity segment curve, and extracting a temperature segment curve from the actual temperature curve includes: Setting the interval point, extraction point threshold and segment voltage range threshold of the actual voltage curve; extracting the voltage segment curve from the actual voltage curve according to the extraction point threshold and the segment voltage range threshold; Acquire the voltage value and capacity value of the interval point in the voltage segment curve, and convert the voltage segment curve into the incremental capacity segment curve; The temperature segment curve is extracted from the actual temperature curve according to the voltage segment curve.
8. A battery health status estimation system, characterized in that: The system comprises: A data processing module, used to obtain actual charging condition data of the battery from the cloud, and obtain an actual charging curve according to the actual charging condition data, wherein the actual charging curve at least includes an actual voltage curve and an actual temperature curve; A curve generating module, configured to extract a voltage segment curve from the actual voltage curve according to a set threshold, convert the voltage segment curve to obtain an incremental capacity segment curve, and extract a temperature segment curve from the actual temperature curve; A curve processing module, used for normalizing the voltage segment curve, the incremental capacity segment curve and the temperature segment curve; The state estimation module is used to input the processed voltage segment curve, the processed incremental capacity segment curve and the processed temperature segment curve into the battery state estimation model in the cloud, and obtain the battery health state value of the battery through the battery state estimation model.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the computer program to implement the battery health status estimation method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the battery health state estimation method according to any one of claims 1 to 7 is implemented.
Citation Information
Cited By
Retired battery secondary sorting method, system and equipment based on random charging fragments and medium
CN119500617A
