Lithium ion battery state-of-charge estimation method and device based on transfer learning

Abstract

The invention discloses a lithium ion battery state-of-charge estimation method and device based on transfer learning, and belongs to the technical field of battery management systems, and the method comprises the steps: obtaining current, voltage and temperature original data, and carrying out the normalization processing of the current, voltage and temperature original data, and constructing time sequence data; sequentially performing convolution processing and pooling processing on the time sequence data, and substituting the time sequence data into an LSTM network to extract time sequence features; carrying out transfer learning processing on the time sequence characteristics to obtain an SOC value of primary version prediction; and the SOC value after Kalman filtering correction is obtained from the SOC value predicted by the primary edition through a Kalman filter.

Description

Technical Field

[0001] The invention discloses a method and a device for estimating the state of charge of a lithium-ion battery based on transfer learning, belonging to the technical field of battery management systems. Background Art

[0002] Against the backdrop of energy shortages and tightening regulations on greenhouse gas emissions, electric vehicles have received extensive attention. Lithium-ion batteries are widely used in electric vehicles due to their high energy density, high charge and discharge efficiency, low self-discharge rate, long life, and high cycle stability. Lithium-ion batteries are affected by many factors during operation, such as temperature, charge and discharge rate, charge and discharge depth, and number of cycles. Therefore, in order to ensure the safe and stable operation of lithium-ion batteries, it is very important to establish a reliable battery management system (BMS) to monitor the operating status of the battery. The battery state of charge is used to reflect the remaining capacity of the battery. It is numerically defined as the ratio of the remaining capacity to the battery capacity and is one of the key states of lithium-ion batteries. Accurately estimating the battery SOC plays an important role in extending battery life, improving system performance, and improving safety. However, complex chemical reactions occur inside lithium ions during operation, which are affected by factors such as discharge current and temperature. Therefore, inferring SOC from measurable signals such as battery terminal voltage, load current, and ambient temperature is a challenging task.

[0003] At present, there are mainly the following methods for estimating battery SOC: open circuit voltage method, coulomb counting method, model-based method, and data-driven method.

[0004] The basic principle of the open circuit voltage (OCV) method is to estimate the SOC by using the relationship between the open circuit voltage of the battery and its state of charge; the coulomb counting method tracks the current during the charging and discharging process based on the internal chemical reactions of the battery; the model-based SOC estimation method includes several steps: model establishment, parameter identification and estimation, and state estimation; the data-driven method estimates the battery SOC by collecting real-time current and voltage data of the battery, preprocessing the data, extracting features, and establishing a regression model.

[0005] In summary, the open circuit voltage method directly uses the relationship between the voltage and SOC of the battery under no-load conditions for estimation. It is simple to operate and has a fast response speed, but its accuracy is limited by the battery type and temperature changes, and it cannot be monitored in real time; the coulomb counting method calculates the SOC by integrating the current during the charge and discharge process. It is easy to implement and suitable for real-time monitoring, but the cumulative error increases with time, and it requires high accuracy for current measurement; the model-based method predicts the SOC by establishing an electrochemical model of the battery. It has a solid theoretical foundation and strong adaptability, but the model construction and parameter adjustment are complex, the calculation is large, and it is sensitive to environmental changes; the data-driven method uses historical data to train the model to estimate the SOC. It does not require a deep understanding of the internal mechanism of the battery and has a strong ability to handle nonlinear problems, but it is highly dependent on data quality and feature selection, and there may be a risk of overfitting. In addition to the problems described above, SOC estimation also has the following difficulties. One is temperature sensitivity, and the other is that the training process is relatively slow. Summary of the invention

[0006] In view of the defects of the prior art, the present invention proposes a lithium-ion battery state of charge estimation method and device based on transfer learning to solve the above-mentioned problems existing in the battery SOC estimation method.

[0007] The technical solution of the present invention is as follows:

[0008] According to a first aspect of an embodiment of the present invention, a method for estimating a state of charge of a lithium-ion battery based on transfer learning is provided, comprising:

[0009] Obtaining the original data of current, voltage and temperature, normalizing the original data of current, voltage and temperature respectively to construct time series data;

[0010] The time series data is sequentially subjected to convolution processing and pooling processing, and then brought into the LSTM network to extract time series features;

[0011] Performing transfer learning processing on the time series features to obtain a preliminary predicted SOC value;

[0012] The SOC value predicted by the initial version is passed through a Kalman filter to obtain a Kalman filter-corrected SOC value.

[0013] Preferably, the normalizing the original data of current, voltage and temperature to construct the time series data comprises:

[0014] The original data of current, voltage and temperature are normalized by Z-zero method to obtain normalized data set;

[0015] The sequence length N is selected as 416 for the time series extracted from the normalized data set, and the dimension of each extracted time series is N×4;

[0016] According to the selected batches, the extracted time series are arranged into a three-dimensional tensor to obtain time series data with a dimension of M×N×4.

[0017] Preferably, the step of sequentially performing convolution processing and pooling processing on the time series data and then bringing the data into the LSTM network to extract time series features comprises:

[0018] The time series data is subjected to a dot product operation on the input data through a convolution kernel to extract local features and patterns;

[0019] The maximum pooling method is used for the local features and patterns. Assuming that the dimension of the output Y of the convolution layer is m×n, m is the number of features, and n is the dimension of the feature, the value of the pooled feature at position (i, j) is obtained by formula (1):

[0020]

[0021] Where: P(i, j) is the value of the feature at position (i, j) after pooling, i and j represent the height and width of the output respectively, m i Represents all features in the i-th window of the input features;

[0022] The value of the pooled feature at position (i, j) is fed into the LSTM network to obtain the time series feature.

[0023] Preferably, the performing transfer learning processing on the time series features to obtain the initial predicted SOC value includes:

[0024] The time series features are transferred and learned to obtain the initial predicted SOC value through formula (2):

[0025] SOC t =σ(W h ·h+b) (2)

[0026] Where: W h represents the weight matrix of the fully connected layer, h is the time series feature, b is the bias vector, σ is the sigmoid activation function, SOC t This is the SOC value predicted by the initial version.

[0027] Preferably, the SOC value predicted by the initial version is obtained by using a Kalman filter to obtain a Kalman filter-corrected SOC value, including:

[0028] The SOC value predicted by the initial version is obtained by using formulas (3) and (4) to obtain the SOC value after Kalman filtering:

[0029]

[0030] Where: Pk is the optimal estimate of the error covariance matrix at time point k, A is the transfer matrix, B is the system input matrix, all of which are calculated by the first-order equivalent circuit model of the battery, I k is the current control variable, P k-1 is the error covariance, Q is the variance of the process noise, is the SOC value after Kalman filtering;

[0031] The SOC value after Kalman filtering is obtained by using formulas (5) to (7) to obtain the SOC value corrected by Kalman filtering:

[0032]

[0033] Where: K k is the Kalman gain, R is the variance of the measurement noise, C is the system output matrix calculated by the first-order equivalent circuit model of the battery, is the SOC value corrected by Kalman filter.

[0034] According to a second aspect of an embodiment of the present invention, a lithium-ion battery state of charge estimation device based on transfer learning is provided, comprising:

[0035] Construct a time series module to obtain the original data of current, voltage and temperature, and normalize the original data of current, voltage and temperature to construct the time series data;

[0036] A time series feature extraction module is used to perform convolution processing and pooling processing on the time series data in sequence and then bring them into the LSTM network to extract time series features;

[0037] A transfer learning processing module, used for performing transfer learning processing on the time series features to obtain a preliminary predicted SOC value;

[0038] The filtering correction module is used to obtain the SOC value corrected by Kalman filtering through the Kalman filter based on the SOC value predicted in the initial version.

[0039] Preferably, the timing sequence building module is used to:

[0040] The original data of current, voltage and temperature are normalized by Z-zero method to obtain normalized data set;

[0041] The sequence length N is selected as 416 for the time series extracted from the normalized data set, and the dimension of each extracted time series is N×4;

[0042] According to the selected batches, the extracted time series are arranged into a three-dimensional tensor to obtain time series data with a dimension of M×N×4.

[0043] Preferably, the module for extracting time series features is used to:

[0044] The time series data is subjected to a dot product operation on the input data through a convolution kernel to extract local features and patterns;

[0045] The maximum pooling method is used for the local features and patterns. Assuming that the dimension of the output Y of the convolution layer is m×n, m is the number of features, and n is the dimension of the feature, the value of the pooled feature at position (i, j) is obtained by formula (1):

[0046]

[0047] Where: P(i, j) is the value of the feature at position (i, j) after pooling, i and j represent the height and width of the output respectively, m i Represents all features in the i-th window of the input features;

[0048] The value of the pooled feature at position (i, j) is fed into the LSTM network to obtain the time series feature.

[0049] According to a third aspect of an embodiment of the present invention, a terminal is provided, including:

[0050] one or more processors;

[0051] a memory for storing the one or more processor-executable instructions;

[0052] Wherein, the one or more processors are configured to:

[0053] Execute the method described in the first aspect of the embodiment of the present invention.

[0054] According to a fourth aspect of an embodiment of the present invention, a non-temporary computer-readable storage medium is provided. When instructions in the storage medium are executed by a processor of a terminal, the terminal is enabled to execute the method described in the first aspect of the embodiment of the present invention.

[0055] According to a fifth aspect of an embodiment of the present invention, an application product is provided. When the application product is running on a terminal, the terminal executes the method described in the first aspect of the embodiment of the present invention.

[0056] The beneficial effects of the present invention are:

[0057] The present invention provides a lithium-ion battery state of charge estimation method and device based on transfer learning, by acquiring original data of current, voltage and temperature, respectively normalizing the original data of current, voltage and temperature to construct time series data, performing convolution processing and pooling processing on the time series data in turn, and then bringing them into an LSTM network to extract time series features, performing transfer learning processing on the time series features to obtain a preliminary predicted SOC value, and finally obtaining a Kalman filter-corrected SOC value from the preliminary predicted SOC value, so that the transfer learning theory can be used to adapt a pre-trained model under room temperature conditions to other temperature scenes, so that the models in other temperature scenes can use the prior knowledge learned under room temperature conditions to quickly adapt to the new environment, reducing the time and resources required for training the model from scratch at the new temperature.

[0058] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 The present invention is a flowchart of a method for estimating the state of charge of a lithium-ion battery based on transfer learning according to an exemplary embodiment.

[0060] Figure 2 It is a schematic diagram of constructing a timing sequence in a method for estimating the state of charge of a lithium-ion battery based on transfer learning according to an exemplary embodiment.

[0061] Figure 3 It is a schematic diagram of constructing a CNN-LSTM network in a lithium-ion battery state of charge estimation method based on transfer learning according to an exemplary embodiment.

[0062] Figure 4 It is a schematic diagram of LSTM construction in a lithium-ion battery state of charge estimation method based on transfer learning according to an exemplary embodiment.

[0063] Figure 5 It is a schematic diagram of a transfer learning method in a lithium-ion battery state of charge estimation method based on transfer learning according to an exemplary embodiment.

[0064] Figure 6 It is a schematic diagram of applying Kalman filtering in a lithium-ion battery state of charge estimation method based on transfer learning according to an exemplary embodiment.

[0065] Figure 7 The present invention is a schematic block diagram of a lithium-ion battery state of charge estimation device based on transfer learning according to an exemplary embodiment.

[0066] Figure 8The present invention is a schematic block diagram of a terminal structure according to an exemplary embodiment. DETAILED DESCRIPTION

[0067] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, 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 creative work are within the scope of protection of the present invention.

[0068] In the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they cannot be understood as limitations on the present invention.

[0069] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0070] An embodiment of the present invention provides a lithium-ion battery state of charge estimation method based on transfer learning. The method is implemented by a terminal, which may be a desktop computer or a laptop computer, etc. The terminal at least includes a CPU, etc.

[0071] Embodiment 1

[0072] Figure 1 The present invention is a flow chart of a method for estimating the state of charge of a lithium-ion battery based on transfer learning according to an exemplary embodiment. The method is used in a terminal, and the method includes the following steps:

[0073] Step 101, obtaining the original data of current, voltage and temperature, normalizing the original data of current, voltage and temperature respectively to construct time series data, the specific contents are as follows:

[0074] like Figure 2 As shown in the figure, when building a neural network model for estimating the battery status, three key physical quantities are first selected as feature inputs to obtain the original data of current, voltage and temperature, which can effectively reflect the working status and health of the battery and is an important basis for estimating the battery SOC.

[0075] In order to improve the training efficiency and prediction accuracy of the neural network, the original data was first Z-zero normalized. Z-zero normalization is a commonly used data preprocessing method that can scale the data to a fixed range, thereby eliminating dimensional differences in the data and improving the convergence speed of the network. During the Z-zero normalization process, the mean and standard deviation of the data were calculated, and each data point was subtracted from the mean and divided by the standard deviation to make it conform to the normal distribution. Next, the normalized data was constructed into a time series format that can be input to the neural network using the keras.preprocessing.timeseries_dataset_from_array module. This module can convert a one-dimensional array into a time series dataset, which contains input data and output targets. For the SOC estimation problem, the input data is the normalized current, voltage, and temperature series, and the output target is the estimated value of SOC. The formula (1) for data normalization is as follows:

[0076]

[0077] Among them, X * =[V * , I * , T * ] is the normalized data set, X = [V, I, T] is the original value, μ is the overall mean of the data, and σ is the overall standard deviation of the data.

[0078] In this way, the raw data is successfully converted into a format suitable for neural network training, laying the foundation for subsequent model construction and training. This data preprocessing method not only improves the training efficiency of the model, but also helps the network learn more effective features, thereby improving the accuracy of SOC estimation.

[0079] Step 102, the time series data is sequentially subjected to convolution processing and pooling processing, and then brought into the LSTM network to extract time series features. The specific contents are as follows:

[0080] like Figure 3 and 4 As shown in the figure, the normalized data set contains key physical quantities such as voltage (V), current (I) and temperature (T) and is input into the convolutional neural network CNN. CNN performs convolution operations on the input data through a series of convolution kernels, which perform dot product operations on the data to capture local features and patterns in the data. This process can not only extract key information from the data, but also enhance the nonlinear expression ability of the data. The operation process of the convolution operation is shown in the following formula (2):

[0081] Y=ReLU(W*X*+b) (2)

[0082] Among them, Y is the output data, * represents the convolution operation, W represents the convolution kernel weight, b represents the bias of each convolution kernel, and the values ​​of W and b are calculated and updated through the back propagation process of the neural network.

[0083] After the convolution operation, the data will enter the pooling layer. The role of the pooling layer is to reduce the data dimension while retaining the important information of the data. Through the pooling layer, CNN can further reduce the size of the data while maintaining the main features of the data. This step helps the network process data more efficiently in subsequent operations and helps the network learn more abstract and useful features. The operation process of the pooling operation is shown in the following formula (3):

[0084]

[0085] Among them, P(i, j) is the value of the feature at position (i, j) after pooling, i and j represent the height and width of the output respectively, and m i Represents all features within the i-th window of the input features.

[0086] As the convolution kernel slides over the time length, the network finally outputs the operation results of the convolutional neural network. These results represent the key features and patterns extracted by the network from the input data. Subsequently, these features will be fed into the LSTM network to further extract time series features. The LSTM network has a strong ability to extract time series features and can effectively process and memorize long-distance dependencies in time series data. By combining the local features extracted by CNN with the time series features extracted by LSTM, the network can understand the data more comprehensively, thereby improving the accuracy of SOC estimation. Assume that the length of the sequence after the above convolution and pooling is T, and the input vector of each time step is X t state, the hidden state is H t , the output of the forget gate is F t , the input gate input is I t , the output gate output is O t , the hidden state of the previous time step is H t -1, the value of the pooled feature at position (i, j) is fed into the LSTM network to obtain the time series feature. The operation process of the LSTM network is shown in the following formula:

[0087] Forget Gate:

[0088] F t =σ(W f ·[H t -1,X t ]+b f )(4)

[0089] Input Gate:

[0090] I t =σ(W i ·[H t-1 , X t ]+b i ) (5)

[0091] C t =f t ·C t-1 +i t tanh(W c ·[H t-1 , X t ]+b c ) (6)

[0092] Output Gate:

[0093] O t =σ(W o ·[H t-1 , X t ]+b o ) (7)

[0094] H t =O t tanh(C t ) (8)

[0095] Among them, σ is the Sigmoid activation function, W f , W i , W c , W o is the weight matrix corresponding to the forget gate and the input gate, bx, b i , b c , b o is the bias vector corresponding to the forget gate and the input gate. The above weight matrix and bias vector are obtained by neural network training. t is the output of the forget gate, i t is the output of the input gate, C t is the cell state at the current time step, C t-1 is the cell state at the previous time step, o t is the output of the output gate.

[0096] Step 103, transfer learning is performed on the time series features to obtain a preliminary predicted SOC value, the specific contents of which are as follows:

[0097] like Figure 5As shown in Figure 1, in order to obtain better training results when data is scarce and computing resources are limited, the idea of ​​transfer learning is considered. Transfer learning can apply the knowledge learned in one task to another different but related task, and use the existing knowledge to accelerate model training and improve performance. Specifically, two important concepts in transfer learning are source domain and target domain. The source domain refers to the domain corresponding to the trained model, which has abundant data and excellent model performance. The target domain is the domain where knowledge needs to be transferred, data is scarce, and model performance needs to be improved. By establishing a connection between the source domain and the target domain, transfer learning can enable the model in the target domain to achieve better performance with fewer samples. In the task of SOC estimation, the rate of decrease of battery SOC may vary at different temperatures. However, due to the inherent stability of the electrochemical reaction of lithium-ion batteries during operation, there is a close interdependence between SOC estimation tasks at different temperatures. Inspired by this, the basic idea of ​​the proposed transfer learning is to inherit the general prediction ability of the source task, while retraining part of the model for the target task to meet the specific prediction ability. Specifically, the first three layers of parameters in the source domain model are frozen and migrated to the target domain. Only the last fully connected layer is trained, that is, only the parameters of the last layer of the network are updated. The last layer of the network is a fully connected layer. Let the output of the LSTM layer be h. The SOC value of the initial version prediction is obtained by formula (9):

[0098] SOC t =σ(W h ·h+b) (9)

[0099] Where: W h represents the weight matrix of the fully connected layer, h is the time series feature, b is the bias vector, σ is the sigmoid activation function, SOC t For the initial predicted SOC value, the values ​​of the weight matrix and the bias vector are calculated and updated through the back propagation process of the neural network.

[0100] Step 104, the SOC value predicted in the initial version is passed through a Kalman filter to obtain a Kalman filter-corrected SOC value, the specific contents of which are as follows:

[0101] like Figure 6As shown, since the SOC output predicted by the neural network fluctuates greatly, it is possible to consider introducing a filtering idea to reduce output noise and improve prediction accuracy. Here, the Kalman filter is selected to modify the output of the neural network. It includes measuring SOC and estimating SOC, where the measured SOC is the output value of the neural network, and the estimated SOC is the output value after Kalman filtering. In the estimation process, the equivalent circuit of the battery is first modeled, and then the SOC value is calculated using the ampere-hour integration method. Finally, the SOC value calculated by the ampere-hour integration method is used to correct the output of the neural network to obtain the filtering result. The SOC value predicted by the initial version is obtained by formulas (10) and (11) to obtain the SOC value after Kalman filtering:

[0102]

[0103] Where: P k is the optimal estimate of the error covariance matrix at time point k, A is the transfer matrix, B is the system input matrix, all of which are calculated by the first-order equivalent circuit model of the battery, I k is the current control variable, P k-1 is the error covariance, Q is the variance of the process noise, is the SOC value after Kalman filtering;

[0104] The SOC value after Kalman filtering is obtained by using formulas (12) to (14) to obtain the SOC value corrected by Kalman filtering:

[0105]

[0106] Where: K k is the Kalman gain, C is the system output matrix calculated by the first-order equivalent circuit model of the battery, and R is the variance of the measurement noise. The larger the measurement noise, the smaller the correction amplitude of the algorithm. is the SOC value corrected by Kalman filter.

[0107] Embodiment 2

[0108] Figure 7 1 is a schematic block diagram of a lithium-ion battery state of charge estimation device based on transfer learning according to an exemplary embodiment, the device comprising:

[0109] A time series construction module 210 is used to obtain the original data of current, voltage and temperature, and respectively normalize the original data of current, voltage and temperature to construct the time series data;

[0110] A time series feature extraction module 220 is used to perform convolution processing and pooling processing on the time series data in sequence and then bring them into the LSTM network to extract time series features;

[0111] A transfer learning processing module 230 is used to perform transfer learning processing on the time series features to obtain a preliminary predicted SOC value;

[0112] The filtering correction module 240 is used to obtain a Kalman filter-corrected SOC value from the initial predicted SOC value through a Kalman filter.

[0113] Preferably, the timing sequence building module 210 is used to:

[0114] The original data of current, voltage and temperature are normalized by Z-zero method to obtain normalized data set;

[0115] The sequence length N is selected as 416 for the time series extracted from the normalized data set, and the dimension of each extracted time series is N×4;

[0116] According to the selected batches, the extracted time series are arranged into a three-dimensional tensor to obtain time series data with a dimension of M×N×4.

[0117] Preferably, the module for extracting time series features 220 is used to:

[0118] The time series data is subjected to a dot product operation on the input data through a convolution kernel to extract local features and patterns;

[0119] The maximum pooling method is used for the local features and patterns. Assuming that the dimension of the output Y of the convolution layer is m×n, m is the number of features, and n is the dimension of the feature, the value of the pooled feature at position (i, j) is obtained by formula (1):

[0120]

[0121] Where: P(i, j) is the value of the feature at position (i, j) after pooling, i and j represent the height and width of the output respectively, m i Represents all features in the i-th window of the input features;

[0122] The value of the pooled feature at position (i, j) is fed into the LSTM network to obtain the time series feature.

[0123] The present application obtains the original data of current, voltage and temperature, normalizes the original data of current, voltage and temperature respectively to construct time series data, performs convolution and pooling processing on the time series data in turn, and then brings them into the LSTM network to extract time series features, performs transfer learning processing on the time series features to obtain the SOC value predicted by the initial version, and finally obtains the SOC value corrected by the Kalman filter from the SOC value predicted by the initial version through the Kalman filter. Therefore, the transfer learning theory can be used to adapt the pre-trained model under room temperature conditions to other temperature scenarios, so that the models in other temperature scenarios can use the prior knowledge learned under room temperature conditions to quickly adapt to the new environment, reducing the time and resources required to train the model from scratch at the new temperature.

[0124] Embodiment 3

[0125] Figure 8 300 is a block diagram of a terminal provided in an embodiment of the present application, and the terminal may be a terminal in the above embodiment. The terminal 300 may be a portable mobile terminal, such as a smart phone or a tablet computer. The terminal 300 may also be referred to as a user equipment, a portable terminal, or other names.

[0126] Typically, the terminal 300 includes a processor 301 and a memory 302 .

[0127] The processor 301 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 301 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 301 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 301 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 301 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0128] The memory 302 may include one or more computer-readable storage media, which may be tangible and non-transitory. The memory 302 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 302 is used to store at least one instruction, which is used to be executed by the processor 301 to implement a lithium-ion battery state of charge estimation method based on transfer learning provided in the present application.

[0129] In some embodiments, the terminal 300 may further include: a peripheral device interface 303 and at least one peripheral device. Specifically, the peripheral device includes: at least one of a radio frequency circuit 304 , a touch screen 305 , a camera 306 , an audio circuit 307 , a positioning component 308 and a power supply 309 .

[0130] The peripheral device interface 303 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 301 and the memory 302. In some embodiments, the processor 301, the memory 302, and the peripheral device interface 303 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 301, the memory 302, and the peripheral device interface 303 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0131] The radio frequency circuit 304 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 304 communicates with the communication network and other communication devices through electromagnetic signals. The radio frequency circuit 304 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 304 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, and the like. The radio frequency circuit 304 can communicate with other terminals through at least one wireless communication protocol. The wireless communication protocol includes but is not limited to: the World Wide Web, a metropolitan area network, an intranet, various generations of mobile communication networks (2G, 3G, 4G and 5G), a wireless local area network and / or a WiFi (Wireless Fidelity) network. In some embodiments, the radio frequency circuit 304 may also include circuits related to NFC (Near Field Communication), which is not limited in this application.

[0132] The touch display screen 305 is used to display the UI (User Interface). The UI may include graphics, text, icons, videos and any combination thereof. The touch display screen 305 also has the ability to collect touch signals on the surface or above the surface of the touch display screen 305. The touch signal can be input to the processor 301 as a control signal for processing. The touch display screen 305 is used to provide virtual buttons and / or virtual keyboards, also known as soft buttons and / or soft keyboards. In some embodiments, the touch display screen 305 can be one, and the front panel of the terminal 300 is set; in other embodiments, the touch display screen 305 can be at least two, which are respectively set on different surfaces of the terminal 300 or are folded; in some other embodiments, the touch display screen 305 can be a flexible display screen, which is set on the curved surface or folded surface of the terminal 300. Even, the touch display screen 305 can also be set to a non-rectangular irregular shape, that is, a special-shaped screen. The touch display screen 305 can be made of materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0133] The camera assembly 306 is used to capture images or videos. Optionally, the camera assembly 306 includes a front camera and a rear camera. Typically, the front camera is used to realize video calls or selfies, and the rear camera is used to realize the shooting of photos or videos. In some embodiments, there are at least two rear cameras, which are any one of a main camera, a depth of field camera, and a wide-angle camera, so as to realize the fusion of the main camera and the depth of field camera to realize the background blur function, and the fusion of the main camera and the wide-angle camera to realize panoramic shooting and VR (Virtual Reality) shooting functions. In some embodiments, the camera assembly 306 may also include a flash. The flash can be a monochrome temperature flash or a dual-color temperature flash. A dual-color temperature flash refers to a combination of a warm light flash and a cold light flash, which can be used for light compensation at different color temperatures.

[0134] The audio circuit 307 is used to provide an audio interface between the user and the terminal 300. The audio circuit 307 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, and convert the sound waves into electrical signals and input them into the processor 301 for processing, or input them into the radio frequency circuit 304 to achieve voice communication. For the purpose of stereo acquisition or noise reduction, there may be multiple microphones, which are respectively arranged at different parts of the terminal 300. The microphone may also be an array microphone or an omnidirectional acquisition microphone. The speaker is used to convert the electrical signal from the processor 301 or the radio frequency circuit 304 into sound waves. The speaker may be a traditional film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can not only convert the electrical signal into sound waves audible to humans, but also convert the electrical signal into sound waves inaudible to humans for purposes such as ranging. In some embodiments, the audio circuit 307 may also include a headphone jack.

[0135] The positioning component 308 is used to locate the current geographical location of the terminal 300 to implement navigation or LBS (Location Based Service). The positioning component 308 can be based on the GPS (Global Positioning System) of the United States or the Beidou system of China.

[0136] The power supply 309 is used to power various components in the terminal 300. The power supply 309 can be an alternating current, a direct current, a disposable battery, or a rechargeable battery. When the power supply 309 includes a rechargeable battery, the rechargeable battery can be a wired rechargeable battery or a wireless rechargeable battery. A wired rechargeable battery is a battery charged through a wired line, and a wireless rechargeable battery is a battery charged through a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0137] In some embodiments, the terminal 300 further includes one or more sensors 310 , including but not limited to: an acceleration sensor 311 , a gyroscope sensor 312 , a pressure sensor 313 , a fingerprint sensor 314 , an optical sensor 315 , and a proximity sensor 316 .

[0138] The acceleration sensor 311 can detect the magnitude of acceleration on the three coordinate axes of the coordinate system established by the terminal 300. For example, the acceleration sensor 311 can be used to detect the components of gravity acceleration on the three coordinate axes. The processor 301 can control the touch display screen 305 to display the user interface in a horizontal view or a vertical view according to the gravity acceleration signal collected by the acceleration sensor 311. The acceleration sensor 311 can also be used for collecting game or user motion data.

[0139] The gyro sensor 312 can detect the body direction and rotation angle of the terminal 300. The gyro sensor 312 can cooperate with the acceleration sensor 311 to collect the user's 3D (3 Dimensions) action on the terminal 300. The processor 301 can implement the following functions based on the data collected by the gyro sensor 312: motion sensing (such as changing the UI according to the user's tilt operation), image stabilization during shooting, game control, and inertial navigation.

[0140] The pressure sensor 313 can be set on the side frame of the terminal 300 and / or the lower layer of the touch display screen 305. When the pressure sensor 313 is set on the side frame of the terminal 300, it can detect the user's holding signal of the terminal 300, and perform left and right hand recognition or shortcut operation according to the holding signal. When the pressure sensor 313 is set on the lower layer of the touch display screen 305, it can control the operability control on the UI interface according to the user's pressure operation on the touch display screen 305. The operability control includes at least one of a button control, a scroll bar control, an icon control, and a menu control.

[0141] The fingerprint sensor 314 is used to collect the user's fingerprint to identify the user's identity based on the collected fingerprint. When the user's identity is identified as a trusted identity, the processor 301 authorizes the user to perform relevant sensitive operations, including unlocking the screen, viewing encrypted information, downloading software, paying, and changing settings. The fingerprint sensor 314 can be set on the front, back, or side of the terminal 300. When a physical button or a manufacturer logo is set on the terminal 300, the fingerprint sensor 314 can be integrated with the physical button or the manufacturer logo.

[0142] The optical sensor 315 is used to collect the ambient light intensity. In one embodiment, the processor 301 can control the display brightness of the touch display screen 305 according to the ambient light intensity collected by the optical sensor 315. Specifically, when the ambient light intensity is high, the display brightness of the touch display screen 305 is increased; when the ambient light intensity is low, the display brightness of the touch display screen 305 is reduced. In another embodiment, the processor 301 can also dynamically adjust the shooting parameters of the camera assembly 306 according to the ambient light intensity collected by the optical sensor 315.

[0143] The proximity sensor 316, also called a distance sensor, is usually arranged on the front of the terminal 300. The proximity sensor 316 is used to collect the distance between the user and the front of the terminal 300. In one embodiment, when the proximity sensor 316 detects that the distance between the user and the front of the terminal 300 is gradually decreasing, the processor 301 controls the touch display screen 305 to switch from the screen-on state to the screen-off state; when the proximity sensor 316 detects that the distance between the user and the front of the terminal 300 is gradually increasing, the processor 301 controls the touch display screen 305 to switch from the screen-off state to the screen-on state.

[0144] Those skilled in the art will understand that Figure 8 The structure shown in the figure does not constitute a limitation on the terminal 300, and the terminal 300 may include more or less components than those shown in the figure, or combine some components, or adopt a different component arrangement.

[0145] Embodiment 4

[0146] In an exemplary embodiment, a computer-readable storage medium is also provided, on which a computer program is stored. When the program is executed by a processor, a lithium-ion battery state of charge estimation method based on transfer learning is implemented as provided in all the inventive embodiments of the present application.

[0147] Any combination of one or more computer-readable media may be used. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or device.

[0148] Computer-readable signal media may include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0149] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0150] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0151] Embodiment 5

[0152] In an exemplary embodiment, an application product is also provided, including one or more instructions, which can be executed by the processor 301 of the above-mentioned device to complete the above-mentioned lithium-ion battery state of charge estimation method based on transfer learning.

[0153] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and the embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily realized. Therefore, without departing from the general concept defined by the claims and equivalent scope, the present invention is not limited to the specific details and the illustrations shown and described here.

Claims

1. A lithium-ion battery state of charge estimation method based on transfer learning, characterized in that: include: Obtaining the original data of current, voltage and temperature, normalizing the original data of current, voltage and temperature respectively to construct time series data; The time series data is sequentially subjected to convolution processing and pooling processing, and then brought into the LSTM network to extract time series features; Performing transfer learning processing on the time series features to obtain a preliminary predicted SOC value; The SOC value predicted by the initial version is passed through a Kalman filter to obtain a Kalman filter-corrected SOC value.

2. A lithium-ion battery state of charge estimation method based on transfer learning according to claim 1, characterized in that: The normalizing of the original current, voltage and temperature data to construct time series data includes: The original data of current, voltage and temperature are normalized by Z-zero method to obtain normalized data set; The sequence length N is selected as 416 for the time series extracted from the normalized data set, and the dimension of each extracted time series is N×4; According to the selected batches, the extracted time series are arranged into a three-dimensional tensor to obtain time series data with a dimension of M×N×4.

3. The method for estimating the state of charge of a lithium-ion battery based on transfer learning according to claim 1, characterized in that: The time series data is sequentially subjected to convolution and pooling processing and then brought into the LSTM network to extract time series features, including: The time series data is subjected to a dot product operation on the input data through a convolution kernel to extract local features and patterns; The maximum pooling method is used for the local features and patterns. Assuming that the dimension of the output Y of the convolution layer is m×n, m is the number of features, and n is the dimension of the feature, the value of the pooled feature at position (i, j) is obtained by formula (1): Where: P(i, j) is the value of the feature at position (i, j) after pooling, i and j represent the height and width of the output respectively, m i Represents all features in the i-th window of the input features; The value of the pooled feature at position (i, j) is fed into the LSTM network to obtain the time series feature.

4. The method for estimating the state of charge of a lithium-ion battery based on transfer learning according to claim 1, characterized in that: The performing transfer learning processing on the time series features to obtain the initial predicted SOC value includes: The time series features are transferred and learned to obtain the initial predicted SOC value through formula (2): SOC t =σ(W h ·h+b) (2) Where: W h represents the weight matrix of the fully connected layer, h is the time series feature, b is the bias vector, σ is the sigmoid activation function, SOC t This is the SOC value predicted by the initial version.

5. The method for estimating the state of charge of a lithium-ion battery based on transfer learning according to claim 1, characterized in that: The SOC value predicted in the initial version is passed through a Kalman filter to obtain a Kalman filter-corrected SOC value, including: The SOC value predicted by the initial version is obtained by using formulas (3) and (4) to obtain the SOC value after Kalman filtering: Where: P k is the optimal estimate of the error covariance matrix at time point k, A is the transfer matrix, B is the system input matrix, all of which are calculated by the first-order equivalent circuit model of the battery, I k is the current control variable, P k-1 is the error covariance, Q is the variance of the process noise, is the SOC value after Kalman filtering; The SOC value after Kalman filtering is obtained by using formulas (5)-(7) to obtain the SOC value corrected by Kalman filtering: Where: K k is the Kalman gain, R is the variance of the measurement noise, C is the system output matrix calculated by the first-order equivalent circuit model of the battery, is the SOC value corrected by Kalman filter.

6. A lithium-ion battery state of charge estimation device based on transfer learning, characterized in that: include: Construct a time series module to obtain the original data of current, voltage and temperature, and normalize the original data of current, voltage and temperature to construct the time series data; A time series feature extraction module is used to perform convolution processing and pooling processing on the time series data in sequence and then bring them into the LSTM network to extract time series features; A transfer learning processing module, used for performing transfer learning processing on the time series features to obtain a preliminary predicted SOC value; The filtering correction module is used to obtain the SOC value corrected by Kalman filtering through the Kalman filter based on the SOC value predicted in the initial version.

7. A lithium-ion battery state of charge estimation device based on transfer learning according to claim 6, characterized in that: The timing sequence module is constructed to: The original data of current, voltage and temperature are normalized by Z-zero method to obtain normalized data set; The sequence length N is selected as 416 for the time series extracted from the normalized data set, and the dimension of each extracted time series is N×4; According to the selected batches, the extracted time series are arranged into a three-dimensional tensor to obtain time series data with a dimension of M×N×4.

8. The device for estimating the state of charge of a lithium-ion battery based on transfer learning according to claim 6, characterized in that: The module for extracting time series features is used to: The time series data is subjected to a dot product operation on the input data through a convolution kernel to extract local features and patterns; The maximum pooling method is used for the local features and patterns. Assuming that the dimension of the output Y of the convolution layer is m×n, m is the number of features, and n is the dimension of the feature, the value of the pooled feature at position (i, j) is obtained by formula (1): Where: P(i, j) is the value of the feature at position (i, j) after pooling, i and j represent the height and width of the output respectively, m i Represents all features in the i-th window of the input features; The value of the pooled feature at position (i, j) is fed into the LSTM network to obtain the time series feature.

9. A terminal, characterized in that: include: one or more processors; a memory for storing the one or more processor-executable instructions; Wherein, the one or more processors are configured to: Execute a lithium-ion battery state of charge estimation method based on transfer learning as described in any one of claims 1 to 5.

10. A non-transitory computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by the processor of the terminal, the terminal is enabled to execute the lithium-ion battery state of charge estimation method based on transfer learning as described in any one of claims 1 to 5.

Images