Battery state estimation methods, devices, terminal equipment and storage media
By constructing a SOC estimation model based on CNN and Informer networks, the model outputs SOC estimates and ranges, solving the problem of low SOC estimation accuracy in existing technologies and achieving high-precision and fast battery state estimation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-13
- Publication Date
- 2026-03-06
AI Technical Summary
Existing deep learning methods have low accuracy in estimating battery state of charge (SOC) and cannot accurately reflect the state of electric vehicle batteries.
A SOC estimation model based on CNN and Informer networks is adopted, which combines convolutional encoding and decoding modules to output SOC estimates and intervals. The model is trained using a Laplace distribution loss function to achieve high-precision and fast SOC estimation.
It improves the accuracy and speed of SOC estimation, can output deterministic and uncertain measures, achieves interval estimation, and adapts to the impact of sensor detection errors.
Smart Images
Figure CN115186587B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of battery state estimation technology, and in particular to a battery state estimation method, apparatus, terminal device and storage medium. Background Technology
[0002] In recent years, with the development of energy conservation and emission reduction strategies, new energy electric vehicles have gradually replaced fuel vehicles powered by fossil fuels. To ensure the safe and stable operation of the battery pack in electric vehicles, a Battery Management System (BMS) is often required to monitor and control it, effectively extending the battery's cycle life and ensuring safety during charging and discharging. State of Charge (SOC), as a key parameter in the BMS, is defined as the ratio of current remaining capacity to maximum usable capacity, quantifying the remaining energy within the battery. Accurate SOC estimates can predict the remaining driving range of an electric vehicle, providing a basis for reasonable driving; additionally, it can be used to assess the performance differences of individual cells in the battery pack, providing a basis for balanced control of individual cells, thereby maximizing the performance of the battery pack. However, SOC cannot be obtained through direct measurement and needs to be estimated indirectly using measurable variables (such as voltage, current, and temperature). Furthermore, the strong nonlinear and time-varying characteristics of electric vehicle batteries under extreme dynamic operating conditions pose a significant challenge to accurate SOC estimation.
[0003] Current research has identified deep learning methods for SOC estimation, including recurrent neural networks (RNNs), long short-term memory neural networks (LSTMs), and gated recurrent unit neural networks (GRUs). However, these methods often result in low SOC estimation accuracy. Summary of the Invention
[0004] This application provides a battery state estimation method, apparatus, terminal device, and storage medium, the purpose of which is to improve the accuracy of SOC estimation.
[0005] In a first aspect, embodiments of this application provide a battery state estimation method, including:
[0006] Obtain battery operating data information to estimate the current state of the battery;
[0007] Input the battery operation data into the pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range;
[0008] The SOC estimation model includes a convolutional coding module and a convolutional decoding module. The convolutional coding module includes a first CNN layer, a first positional coding layer, a coding layer and a distillation layer connected in sequence. The convolutional decoding module includes a second CNN layer, a second positional coding layer, a decoding layer and an output layer connected in sequence, and the output of the distillation layer is connected to the input of the decoding layer.
[0009] Optionally, battery operating data can be input into a pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range, including:
[0010] The battery operating data is input into a pre-trained SOC estimation model to obtain the battery's SOC estimate and the scale parameter used to calculate the battery's SOC estimation interval.
[0011] Based on the battery's SOC estimate and scale parameters, the battery's SOC estimation range is obtained.
[0012] Optionally, before inputting battery operating data into a pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range, the following steps are included:
[0013] Construct a SOC estimation model;
[0014] Collect battery operation data samples for training the SOC estimation model;
[0015] The battery operation data samples were divided into training and testing sets.
[0016] The SOC estimation model is iteratively trained using the training set to obtain the trained SOC estimation model;
[0017] The trained SOC estimation model was tested using a test set, and the test results were obtained.
[0018] If the test results indicate that the performance of the trained SOC estimation model meets the expected target, then the trained SOC estimation model will be used as the trained SOC estimation model.
[0019] If the test results indicate that the performance of the trained SOC estimation model does not meet the expected target, then the network parameters of the trained SOC estimation model are adjusted, and the SOC model with adjusted network parameters is used as the SOC estimation model to be trained. The process is then repeated to perform iterative training of the SOC estimation model using the training set to obtain the trained SOC estimation model.
[0020] Optionally, during the iterative training of the SOC estimation model, the loss value of the SOC estimation model can be calculated using the following formula:
[0021]
[0022] Where Loss is the loss value of the SOC estimation model, N is the total number of battery operation data samples in the training set, and x t For the battery operation data samples at time t in the training set, t = 1, ..., N, To train a set of battery operation data samples x t The resulting loss value, The two-dimensional output of the SOC estimation model. v(x t Let λ(x) be the estimated SOC value at time t. t ) is the scale parameter, where λ(x) is the scale parameter. t This is used to calculate the SOC estimation interval at time t, where SOC... t This represents the battery's true SOC value at time t. C max C is the maximum usable capacity of the battery. t This represents the amount of capacity consumed by the battery from its initial fully charged state to time t.
[0023] Optionally, the battery operation data samples can be divided into a training set and a test set, including:
[0024] Normalize each data sample in the battery operation data information sample;
[0025] The normalized battery operation data samples were divided into training and testing sets.
[0026] Optionally, the battery operation data information includes multiple sets of battery operation data, which are battery operation data at multiple different times, including the current time. Each set of battery operation data includes the battery's terminal voltage, current, and ambient temperature.
[0027] Secondly, embodiments of this application provide a battery state estimation device, comprising:
[0028] The acquisition module is used to acquire battery operating data information for estimating the current state of the battery;
[0029] The output module is used to input battery operating data into a pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range.
[0030] The SOC estimation model includes a convolutional coding module and a convolutional decoding module. The convolutional coding module includes a first CNN layer, a first positional coding layer, a coding layer and a distillation layer connected in sequence. The convolutional decoding module includes a second CNN layer, a second positional coding layer, a decoding layer and an output layer connected in sequence, and the output of the distillation layer is connected to the input of the decoding layer.
[0031] Optional, the output module includes:
[0032] The first processing unit is used to input battery operation data information into a pre-trained SOC estimation model to obtain the battery's SOC estimate and the scale parameter used to calculate the battery's SOC estimation interval.
[0033] The second processing unit is used to obtain the SOC estimation range of the battery based on the battery's SOC estimate and scale parameters.
[0034] Optionally, the battery state estimation device also includes:
[0035] The model building module is used to build SOC estimation models;
[0036] The acquisition module is used to collect battery operation data samples for training the SOC estimation model;
[0037] The preprocessing module is used to divide the battery operation data samples into training and testing sets;
[0038] The training module is used to iteratively train the SOC estimation model using the training set to obtain the trained SOC estimation model.
[0039] The testing module is used to test the trained SOC estimation model using a test set and obtain the test results.
[0040] The determination module is used to determine if the test results indicate that the performance of the trained SOC estimation model meets the expected target, and then use the trained SOC estimation model as the trained SOC estimation model.
[0041] The adjustment module is used to adjust the network parameters of the trained SOC estimation model if the test results indicate that the performance of the trained SOC estimation model does not meet the expected target. The adjusted SOC model is then used as the SOC estimation model to be trained, triggering the training module to iteratively train the SOC estimation model using the training set to obtain the trained SOC estimation model.
[0042] Optionally, during the iterative training of the SOC estimation model, the training module calculates the loss value of the SOC estimation model using the following formula:
[0043]
[0044] Where Loss is the loss value of the SOC estimation model, N is the total number of battery operation data samples in the training set, and x t For the battery operation data samples at time t in the training set, t = 1, ..., N, To train a set of battery operation data samples x t The resulting loss value, The two-dimensional output of the SOC estimation model. v(x t Let λ(x) be the estimated SOC value at time t. t ) is the scale parameter, where λ(x) is the scale parameter. t This is used to calculate the SOC estimation interval at time t, where SOC... t This represents the battery's true SOC value at time t. C max C is the maximum usable capacity of the battery. t This represents the capacity consumed by the battery from its initial fully charged state until time t.
[0045] Optional, the preprocessing module includes:
[0046] The normalization unit is used to normalize each data sample in the battery operation data information sample.
[0047] The partitioning unit is used to divide the normalized battery operation data samples into training and testing sets.
[0048] Optionally, the battery operation data information includes multiple sets of battery operation data, which are battery operation data at multiple different times, including the current time. Each set of battery operation data includes the battery's terminal voltage, current, and ambient temperature.
[0049] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned battery state estimation method.
[0050] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned battery state estimation method.
[0051] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes the terminal device to execute the battery state estimation method described in the first aspect.
[0052] The above-mentioned solution in this application has the following beneficial effects:
[0053] In the embodiments of this application, battery operating data information for estimating the current state of the battery is obtained, and this data is input into a pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation interval. Since the SOC estimation model is based on CNN and Informer networks, it has the advantages of high estimation accuracy and fast operation, thus greatly improving the accuracy of SOC estimation when estimating battery state. Furthermore, in practical applications, due to factors such as sensor detection errors, describing a time series with random fluctuations using a fixed number is unreliable. The SOC estimation model of this application, in addition to outputting the SOC estimate, can also output the SOC estimation interval. Therefore, this application can measure the uncertainty of SOC estimation and achieve interval estimation.
[0054] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description
[0055] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0056] Figure 1 A flowchart of a battery state estimation method provided in an embodiment of this application;
[0057] Figure 2 This is a schematic diagram of the structure of a SOC estimation model provided in an embodiment of this application;
[0058] Figure 3 A flowchart illustrating the training process of a SOC estimation model provided in an embodiment of this application;
[0059] Figure 4 This is the estimated curve for HWFET driving conditions in an example of this application;
[0060] Figure 5 This is the estimated curve for the US06 driving condition in one example of this application;
[0061] Figure 6 This is the estimated curve of the UUDS driving condition in an example of this application;
[0062] Figure 7This is the estimated curve of the LA92 under driving conditions in an example of this application;
[0063] Figure 8 This is a schematic diagram of the structure of a battery state estimation device provided in an embodiment of this application;
[0064] Figure 9 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation
[0065] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.
[0066] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.
[0067] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0068] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."
[0069] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0070] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0071] Currently, deep learning methods used in battery state estimation (SOC) research include recurrent neural networks (RNNs), long short-term memory neural networks (LSTMs), and gated recurrent unit neural networks (GRUs). However, these methods result in low SOC estimation accuracy.
[0072] To address the aforementioned issues, this application's embodiments acquire battery operating data for estimating the battery's current state and input this data into a pre-trained SOC estimation model to obtain the battery's estimated SOC value and SOC estimation interval. Since the SOC estimation model is based on CNN and Informer networks, it boasts advantages such as high estimation accuracy and fast operation, significantly improving the accuracy of SOC estimation when estimating battery state. Furthermore, in practical applications, describing a time series with random fluctuations using a fixed number is unreliable due to factors such as sensor detection errors. This application's SOC estimation model, in addition to outputting the estimated SOC value, also outputs the SOC estimation interval, thus enabling the measurement of SOC estimation uncertainty and achieving interval estimation.
[0073] The battery state estimation method provided in this application will be described exemplarily below with reference to specific embodiments.
[0074] This application provides a battery state estimation method, which can be executed by a terminal device or by a device (such as a chip) applied in the terminal device. The following embodiments use the execution of the method by a terminal device as an example. As an example, the terminal device can be a tablet, server, or laptop, etc., and this application does not limit it.
[0075] like Figure 1 As shown, the battery state estimation method provided in this application includes the following steps:
[0076] Step 101: Obtain battery operating data information for estimating the current state of the battery.
[0077] In some embodiments of this application, the battery is a vehicle battery, such as a lithium battery; the battery operation data information includes information such as the terminal voltage, current, and ambient temperature of the battery at different times (including the current time).
[0078] In some embodiments of this application, the battery operation data information specifically includes multiple sets of battery operation data, which are battery operation data at multiple different times, including the current time and multiple historical times. Each set of battery operation data includes the battery's terminal voltage, current and ambient temperature.
[0079] Specifically, information such as terminal voltage, current, and ambient temperature can be obtained through onboard sensors. More specifically, terminal voltage can be acquired using a Hall voltage sensor, current using a Hall current sensor, and ambient temperature using an infrared temperature sensor.
[0080] Step 102: Input the battery operation data into the pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range.
[0081] In some embodiments of this application, such as Figure 2 As shown, the SOC estimation model includes a convolutional encoding module 200 and a convolutional decoding module 201. The convolutional encoding module 200 includes a first CNN layer 202, a first positional encoding layer 203, an encoding layer 204, and a distillation layer 205 connected in sequence. The convolutional decoding module 201 includes a second CNN layer 206, a second positional encoding layer 207, a decoding layer 208, and an output layer 209 connected in sequence. The output of the distillation layer 205 is connected to the input of the decoding layer 208. The convolutional encoding module 200 generates a cascaded feature vector, which is output to the convolutional decoding module 201 through the distillation layer 205. The convolutional decoding module 201 outputs the SOC estimate and the scale parameter used to calculate the SOC estimation interval.
[0082] In some embodiments of this application, the aforementioned battery operation data can be input into a pre-trained SOC estimation model in the form of a sequence. Specifically, the first CNN layer 202 of the convolutional coding module 200 is mainly used to extract spatial features from the input terminal voltage, current, and ambient temperature data; the first positional coding layer 203 is mainly used to perform positional coding on the spatial features extracted by the first CNN layer 202; the coding layer 204 is mainly used to extract temporal features; and the distillation layer 205 is mainly used to stack the outputs of the coding layer 204 to output a concatenated feature sequence. The second CNN layer 206 of the convolutional decoding module 201 is mainly used to extract spatial features from historical battery operation data; the second positional coding layer 207 is mainly used to perform positional coding on the spatial features extracted by the second CNN layer 206; and the decoding layer 208, after receiving the positionally encoded spatial features, can further decode the concatenated feature vector generated in the convolutional coding module 200 and output the decoded result to the output layer 209. The output layer 209 is mainly used to convert the decoded result into a two-dimensional output, namely, the SOC estimate and the scale parameter used to calculate the SOC estimation interval.
[0083] As described above, after inputting the battery operating data information obtained in step 101 into the pre-trained SOC estimation model, the SOC estimation model can output the current SOC estimate of the battery, as well as the scale parameter used to calculate the current SOC estimation interval of the battery. It should be noted that after obtaining the SOC estimate and scale parameter, the probability distribution function can be used to calculate the SOC estimate and scale parameter to obtain the current SOC estimation interval of the battery.
[0084] It is worth mentioning that, in some embodiments of this application, since the SOC estimation model is based on a CNN and Informer network, it has the advantages of high estimation accuracy and fast running speed, thus greatly improving the accuracy of SOC estimation when estimating battery state. Furthermore, in practical applications, due to factors such as sensor detection errors, describing a time series with random fluctuations using a fixed number is unreliable. The SOC estimation model of this application, in addition to outputting the estimated SOC value, can also output the SOC estimation interval. Therefore, this application can measure the uncertainty of SOC estimation and achieve interval estimation.
[0085] In some embodiments of this application, the specific implementation process of step 102 is as follows: input the battery operation data information obtained in step 101 into the pre-trained SOC estimation model to obtain the current SOC estimate of the battery and the scale parameter used to calculate the current SOC estimation interval of the battery; then, based on the battery's SOC estimate and the scale parameter, the battery's SOC estimation interval is obtained.
[0086] The process of determining the SOC estimation interval mainly includes determining the upper and lower bounds of the SOC estimation interval. For example, taking a 95% confidence interval, the upper bound of the SOC estimation interval can be calculated using the formula laplace.pdf(0.975,loc,scale)=1 / (2*scale)*exp(-abs(0.975-loc) / scale), and the lower bound of the SOC estimation interval can be calculated using the formula laplace.pdf(0.025,loc,scale)=1 / (2*scale)*exp(-abs(0.025-loc) / scale). Here, loc is the estimated SOC value, scale is the scaling parameter, exp represents the exponential function, and abs represents taking the absolute value.
[0087] The training process of the SOC estimation model is illustrated below with specific examples.
[0088] In some embodiments of this application, before performing step 102, a process of training the SOC estimation model is also included, such as... Figure 3 As shown, the training process of the SOC estimation model includes the following steps:
[0089] Step 301: Construct the SOC estimation model.
[0090] In some embodiments of this application, the convergence of the training model can be accelerated by utilizing a deep learning framework network to build the SOC estimation model. As a preferred example, the aforementioned SOC estimation model can be a convolutional neural network (CNN)-Informer network based on PyTorch (PyTorch is a deep learning framework with a Python interface). Here, Informer is an improved Transformer (Transformer is a deep network based on a self-attention mechanism).
[0091] Step 302: Collect battery operation data samples for training the SOC estimation model.
[0092] In some embodiments of this application, the aforementioned battery operating data information samples include battery operating data under different temperatures and operating conditions. Specifically, the battery operating data information samples include multiple data samples, each including terminal voltage, current, and ambient temperature.
[0093] In some embodiments of this application, the process of collecting battery operation data samples can be as follows: First, the vehicle battery is fully charged; then, under laboratory conditions, the battery operating temperature is set to 0 degrees Celsius, 10 degrees Celsius, and 25 degrees Celsius respectively. Under constant ambient temperature, the battery is charged and discharged according to different power configuration curves. The power configurations used include HWFET, US06, UUDS, LA92 driving conditions, and Cycle1-4 and NN driving conditions, which are a mixture of the above four driving conditions. These driving conditions (i.e., HWFET, US06, UUDS, LA92, Cycle1-4, and NN) cover most of the operating conditions of electric vehicles and are highly representative. Specifically, the HWFET condition simulates a car traveling at 60 km / h on a highway, the US06 condition simulates a car undergoing rapid acceleration and deceleration, the UUDS condition simulates a car traveling on general urban roads, and the LA92 condition simulates a car traveling under energy-saving conditions. The sensor devices in the car detect the terminal voltage, current, and surface temperature (i.e., ambient temperature) of each battery in the battery pack during operation at a sampling frequency of 100 Hz.
[0094] It should be noted that during the training process, each data sample also includes the battery's actual SOC value. Specifically, the real-time power and capacity changes of the battery can be measured using an Arbin BT2000 battery tester, and then the battery's actual SOC value is obtained based on the capacity changes.
[0095] Step 303: Divide the battery operation data information samples into training set and test set.
[0096] In some embodiments of this application, each data sample in the battery operation data information sample can be normalized first, and then the normalized battery operation data information sample can be divided into a training set and a test set. Specifically, the data under the four driving conditions of HWFET, US06, UUDS and HWFET can be used as the test set, while the data under the other driving conditions can be used as the training set.
[0097] The formula for normalizing the battery operation data samples is as follows:
[0098]
[0099] Where, x min and x max x represents the minimum and maximum values in each data set (i.e., a data set consisting of all terminal voltages, all currents, or all ambient temperatures). inFor the normalized raw data value (i.e., the normalized terminal voltage, the normalized current, or the normalized ambient temperature), x norm The normalized data value (i.e., the normalized terminal voltage, the normalized current, or the normalized ambient temperature).
[0100] It should be noted that if the terminal voltage is normalized, then x min and x max x represents the minimum and maximum values in the data set consisting of all terminal voltages. in x is the normalized terminal voltage. norm This is the normalized terminal voltage; if the current is normalized, then x... min and x max x represents the minimum and maximum values in the data set consisting of all currents. in For the normalized current, x norm This is the normalized current; if the ambient temperature is normalized, then x... min and x max x represents the minimum and maximum values in the dataset consisting of all ambient temperatures. in For the normalized ambient temperature, x norm This is the normalized ambient temperature.
[0101] Step 304: Iteratively train the SOC estimation model using the training set to obtain the trained SOC estimation model.
[0102] In some embodiments of this application, before performing the SOC estimation model, the model parameters of the SOC estimation model are set. These model parameters include model structure parameters (such as the number of encoder layers in the encoding layer, the number of decoder layers in the decoding layer, the number of CNN layers, the number of heads, etc.) and other hyperparameters (such as the length of the input information sequence, model dimension, batch size, learning rate, and number of iterations, etc.).
[0103] For example, in some embodiments of this application, the encoder layer has 2 layers, the decoder layer has 1 layer, the CNN layer has 4 layers, the head has 5 heads, the length of the input information sequence is 40, the model dimension is 64, the batch size is 50, the learning rate is 0.0001, and the number of iterations is 100.
[0104] During iterative training, the SOC estimation model is trained using backpropagation and a loss function based on the Laplace distribution. In the process of training the network model using the backpropagation algorithm, the Adam (Adam is a stochastic optimization method with adaptive momentum) stochastic gradient descent algorithm is introduced to iteratively adjust the model weights according to the set learning rate to minimize the error.
[0105] During the iterative training of the SOC estimation model, the loss value of the SOC estimation model is calculated using the following formula:
[0106]
[0107] Where Loss is the loss value of the SOC estimation model, N is the total number of battery operation data samples in the training set (which can be understood as the collection time of the last data sample collected in the training set), and x t For the battery operation data samples at time t in the training set, t = 1, ..., N, To train a set of battery operation data samples x t The resulting loss value, The two-dimensional output of the SOC estimation model. v(x t Let λ(x) be the estimated SOC value at time t. t ) is the scale parameter, where λ(x) is the scale parameter. t This is used to calculate the SOC estimation interval at time t, where SOC... t This represents the battery's true SOC value at time t. C max C is the maximum usable capacity of the battery. t This represents the capacity consumed by the battery from its initial fully charged state until time t.
[0108] It is worth mentioning that using a loss function based on the Laplace distribution as the objective function for optimizing model parameters can, on the one hand, take error characteristics into account during the modeling process, reduce the impact of outliers on the estimation results, and thus improve the accuracy and reliability of the model; on the other hand, the model simultaneously outputs location parameters (i.e., SOC estimates) and scale parameters (i.e., SOC estimation intervals), which can measure the uncertainty of SOC estimation and achieve interval estimation.
[0109] Step 305: Test the trained SOC estimation model using the test set to obtain the test results.
[0110] In some embodiments of this application, the SOC estimation model can be tested using general evaluation criteria formulas. Specifically, the evaluation criteria formulas use the mean absolute error (MAE) and maximum absolute error (MAX) to evaluate the SOC estimate, and the interval coverage probability (CP) and interval average width percentage (MWP) to evaluate the SOC estimation interval.
[0111] The formula for calculating MAE is as follows:
[0112]
[0113] The formula for calculating MAX is:
[0114]
[0115] Where N′ represents the total number of battery operation data samples in the test set (which can be understood as the acquisition time of the last data sample collected in the test set). MAX represents the SOC estimate output by the SOC estimation model at time t. The data selected by MAX is all the data from the start time to the end of the acquisition.
[0116] The formula for calculating CP is:
[0117]
[0118] The formula for calculating MWP is:
[0119]
[0120] in, It is the lower bound of the SOC estimation interval for the i-th data sample in the test set. It is the upper bound of the SOC estimation interval for the i-th data sample in the test set, where SOC i It is the SOC estimate of the i-th data sample in the test set.
[0121] Step 306: Determine whether the test results indicate that the performance of the trained SOC estimation model has reached the expected target. If the test results indicate that the performance of the trained SOC estimation model has reached the expected target, proceed to step 307; otherwise, proceed to step 308.
[0122] In some embodiments of this application, when the value of MAE is less than 1%, the value of MAX is less than 5%, the value of CP with a given 95% confidence level is greater than 95%, and the value of MWP is less than 20, it is determined that the performance of the SOC estimation model has reached the expected target; otherwise, it is determined that the performance of the SOC estimation model has not reached the expected target.
[0123] Step 307: Use the trained SOC estimation model as the trained SOC estimation model.
[0124] Step 308: Adjust the network parameters of the trained SOC estimation model, and use the SOC model with adjusted network parameters as the SOC estimation model to be trained, then return to step 304. It should be noted that the SOC estimation model in step 304 at this point is the SOC model with adjusted network parameters.
[0125] The accuracy of the battery state estimation method of this application is illustrated below with specific experimental data.
[0126] like Figures 4 to 7 As shown, where, Figure 4 The curves represent the SOC estimation results of the model under HWFET driving conditions, with ambient temperatures from left to right of 0 degrees Celsius, 10 degrees Celsius, and 25 degrees Celsius. Figure 5 The curves represent the SOC estimation results of the model under the US06 driving conditions, with ambient temperatures from left to right of 0 degrees Celsius, 10 degrees Celsius, and 25 degrees Celsius. Figure 6 The curves represent the SOC estimation results of the model under UUDS driving conditions, with ambient temperatures from left to right of 0 degrees Celsius, 10 degrees Celsius, and 25 degrees Celsius. Figure 7 The curves represent the SOC estimation results of the model under the LA92 driving conditions, with ambient temperatures from left to right being 0 degrees Celsius, 10 degrees Celsius, and 25 degrees Celsius.
[0127] The experimental data above shows that the model's output SOC estimate almost completely falls within the generated SOC estimation range, and the SOC estimate is very close to the true SOC value (since the SOC estimate and the true SOC value are extremely close, they basically overlap in the figure, so only the SOC estimate is marked in the figure), indicating that the model has high accuracy in SOC estimation. For four driving conditions under three different temperatures, the SOC estimation model achieved an average MAE of 0.77%, an average MAX of 3.79%, an average CP of 95.19%, and an average MWP of 17.10.
[0128] In summary, the battery state estimation method of this application can fully learn the mapping relationship between measurable variables (such as terminal voltage, current and ambient temperature) and SOC values, which not only achieves high-precision SOC estimation, but also quantifies the uncertainty of the estimation results.
[0129] In summary, the battery state estimation method provided in this application has the following effects:
[0130] First, since the SOC estimation model is based on CNN and Informer network, it has the advantages of high estimation accuracy and fast running speed, which greatly improves the accuracy of SOC estimation when estimating battery state.
[0131] Secondly, in practical applications, due to factors such as sensor detection errors, it is unreliable to describe a time series with random fluctuation characteristics using a definite number. In addition to outputting the SOC estimate value, the SOC estimation model of this application can also output the SOC estimation interval. Therefore, this application can measure the uncertainty of SOC estimation and realize interval estimation.
[0132] The battery state estimation device provided in this application will be described exemplarily below with reference to specific embodiments.
[0133] like Figure 8 As shown, an embodiment of this application provides a battery state device, the battery state estimation device 800 including:
[0134] The acquisition module 801 is used to acquire battery operating data information for estimating the current state of the battery;
[0135] The output module 802 is used to input battery operation data information into a pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range.
[0136] The SOC estimation model includes a convolutional coding module and a convolutional decoding module. The convolutional coding module includes a first CNN layer, a first positional coding layer, a coding layer and a distillation layer connected in sequence. The convolutional decoding module includes a second CNN layer, a second positional coding layer, a decoding layer and an output layer connected in sequence, and the output of the distillation layer is connected to the input of the decoding layer.
[0137] Optionally, output module 802 includes:
[0138] The first processing unit is used to input battery operation data information into a pre-trained SOC estimation model to obtain the battery's SOC estimate and the scale parameter used to calculate the battery's SOC estimation interval.
[0139] The second processing unit is used to obtain the SOC estimation range of the battery based on the battery's SOC estimate and scale parameters.
[0140] Optionally, the battery state estimation device 800 further includes:
[0141] The model building module is used to build SOC estimation models;
[0142] The acquisition module is used to collect battery operation data samples for training the SOC estimation model;
[0143] The preprocessing module is used to divide the battery operation data samples into training and testing sets;
[0144] The training module is used to iteratively train the SOC estimation model using the training set to obtain the trained SOC estimation model.
[0145] The testing module is used to test the trained SOC estimation model using a test set and obtain the test results.
[0146] The determination module is used to determine if the test results indicate that the performance of the trained SOC estimation model meets the expected target, and then use the trained SOC estimation model as the trained SOC estimation model.
[0147] The adjustment module is used to adjust the network parameters of the trained SOC estimation model if the test results indicate that the performance of the trained SOC estimation model does not meet the expected target. The adjusted SOC model is then used as the SOC estimation model to be trained, triggering the training module to iteratively train the SOC estimation model using the training set to obtain the trained SOC estimation model.
[0148] Optionally, during the iterative training of the SOC estimation model, the training module calculates the loss value of the SOC estimation model using the following formula:
[0149]
[0150] Where Loss is the loss value of the SOC estimation model, N is the total number of battery operation data samples, and x t For the battery operation data samples at time t in the training set, t = 1, ..., N, To train a set of battery operation data samples x t The resulting loss value, The two-dimensional output of the SOC estimation model. v(x t Let λ(x) be the estimated SOC value at time t. t ) is the scale parameter, where λ(x) is the scale parameter. t This is used to calculate the SOC estimation interval at time t, where SOC... t This represents the battery's true SOC value at time t. C max C is the maximum usable capacity of the battery. t This represents the capacity consumed by the battery from its initial fully charged state until time t.
[0151] Optional, the preprocessing module includes:
[0152] The normalization unit is used to normalize each data sample in the battery operation data information sample.
[0153] The partitioning unit is used to divide the normalized battery operation data samples into training and testing sets.
[0154] Optionally, the battery operation data information includes multiple sets of battery operation data, which are battery operation data at multiple different times, including the current time. Each set of battery operation data includes the battery's terminal voltage, current, and ambient temperature.
[0155] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.
[0156] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0157] like Figure 9 As shown, embodiments of this application provide a terminal device, such as... Figure 9 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 9 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 executes the computer program D102 to implement the steps in any of the above method embodiments.
[0158] Specifically, when the processor D100 executes the computer program D102, it acquires battery operating data information used to estimate the current state of the battery, and inputs the battery operating data into a pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range. Because the SOC estimation model integrates CNN and Informer networks and adopts a loss function based on Laplace distribution, it reduces the impact of outliers on the estimation results, thereby significantly improving the SOC estimation accuracy when estimating the battery state.
[0159] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.
[0160] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.
[0161] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.
[0162] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0163] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.
[0164] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.
[0165] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a battery state estimation device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.
[0166] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0167] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0168] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.
[0169] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0170] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. A battery state estimation method characterized by, The method comprises: acquiring battery operation data information for estimating a current state of the battery; inputting the battery operation data information into a pre-trained SOC estimation model to obtain an SOC estimation value and an SOC estimation interval of the battery; wherein the SOC estimation model comprises a convolutional encoding module and a convolutional decoding module, the convolutional encoding module comprises a first CNN layer, a first position encoding layer, an encoding layer and a distillation layer connected in sequence, the convolutional decoding module comprises a second CNN layer, a second position encoding layer, a decoding layer and an output layer connected in sequence, and an output end of the distillation layer is connected to an input end of the decoding layer; the first CNN layer is used to extract spatial features in input terminal voltage, current and environmental temperature data information, and the second CNN layer is used to extract spatial features from historical terminal voltage, current and environmental temperature data information.
2. The method of claim 1, wherein, The inputting of the battery operation data information into the pre-trained SOC estimation model to obtain the SOC estimation value and the SOC estimation interval of the battery comprises: inputting the battery operation data information into the pre-trained SOC estimation model to obtain the SOC estimation value of the battery and a scale parameter used to calculate the SOC estimation interval of the battery; obtaining the SOC estimation interval of the battery according to the SOC estimation value and the scale parameter.
3. The method of claim 1, wherein, Before the inputting of the battery operation data information into the pre-trained SOC estimation model to obtain the SOC estimation value and the SOC estimation interval of the battery, the method further comprises: constructing the SOC estimation model; collecting battery operation data information samples for training the SOC estimation model; dividing the battery operation data information samples into a training set and a test set; iteratively training the SOC estimation model using the training set to obtain a trained SOC estimation model; testing the trained SOC estimation model using the test set to obtain a test result; if the test result represents that the performance of the trained SOC estimation model reaches an expected target, taking the trained SOC estimation model as the pre-trained SOC estimation model; if the test result represents that the performance of the trained SOC estimation model does not reach the expected target, adjusting network parameters of the trained SOC estimation model, taking the SOC model with the adjusted network parameters as a to-be-trained SOC estimation model, and returning to execute the step of iteratively training the SOC estimation model using the training set to obtain the trained SOC estimation model.
4. The method of claim 3, wherein, In the process of iteratively training the SOC estimation model, the loss value of the SOC estimation model is calculated by the following formula: ； in, Estimate the loss value of the SOC model. The total number of battery operation data samples in the training set. For the training set Sample of battery operation data at any given time. , Sample battery operation data information in the training set The resulting loss value, The two-dimensional output of the SOC estimation model. , for SOC estimate at time t. As a scale parameter, this Used for calculation The SOC estimation interval at time 10:
00. For the battery in The actual SOC value at any given moment. , This represents the maximum usable capacity of the battery. The capacity consumed by the battery from its initial fully charged state to time t.
5. The method of claim 3, wherein, The dividing of the battery operation data information samples into the training set and the test set comprises: performing normalization processing on each data sample in the battery operation data information samples; dividing the normalized battery operation data information samples into the training set and the test set.
6. The method of claim 1, wherein, The battery operation data information includes a plurality of sets of battery operation data, each set of battery operation data including battery operation data of the battery at a plurality of different time points, the plurality of different time points including a current time point, and each set of battery operation data including an end voltage, a current, and an ambient temperature of the battery.
7. A battery state estimation device characterized by comprising: The method comprises: an acquisition module configured to acquire battery operation data information used to estimate a current state of the battery; an output module configured to input the battery operation data information into a pre-trained SOC estimation model to obtain an SOC estimation value and an SOC estimation interval of the battery; wherein the SOC estimation model comprises a convolutional encoding module and a convolutional decoding module, the convolutional encoding module comprises a first CNN layer, a first position encoding layer, an encoding layer, and a distillation layer connected in sequence, the convolutional decoding module comprises a second CNN layer, a second position encoding layer, a decoding layer, and an output layer connected in sequence, and an output end of the distillation layer is connected to an input end of the decoding layer; the first CNN layer is configured to extract spatial features in input end voltage, current, and ambient temperature data information, and the second CNN layer is configured to extract spatial features from historical end voltage, current, and ambient temperature data information.
8. A terminal 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 state estimation method of any one of claims 1 to 6.
9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the battery state estimation method of any one of claims 1 to 6.