Battery state of charge estimation method and system based on embedded domain knowledge neural network
By embedding a domain knowledge neural network, SOC estimation is performed using battery operating data, which solves the problem of insufficient accuracy of traditional methods in complex environments. This achieves high-precision and high-reliability SOC estimation, which is suitable for lithium battery management systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- QUALTECH
- Filing Date
- 2024-11-06
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional lithium battery SOC estimation methods are not accurate enough in complex and variable working environments, especially under extreme temperature conditions, and deep learning models lack interpretability.
An embedded domain knowledge neural network is used to acquire battery operating data such as voltage, current, temperature and capacity. The SocNet neural network model is trained and combined with a feature extraction module and a SOC lookback module to estimate SOC, including processing by convolutional neural networks and multilayer perceptrons.
It improves the accuracy and reliability of SOC estimation, enhances the model's adaptability and stability in changing environments, and provides interpretable prediction results.
Smart Images
Figure CN119511113B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of lithium battery health management technology, and more specifically to a battery state of charge estimation method and system based on an embedded domain knowledge neural network. Background Technology
[0002] With the widespread application of electric vehicles, renewable energy storage systems, and portable electronic devices, lithium batteries have become the energy storage device for numerous systems. The State of Charge (SOC) of a lithium battery is a key indicator measuring its remaining capacity and health status, crucial for optimizing lithium battery prediction and PHM (Prognostic and Health Management). Accurate SOC estimation not only helps extend battery life but also ensures device safety and performance. However, traditional SOC estimation methods exhibit significant limitations when facing complex and variable operating environments.
[0003] Traditional SOC estimation methods mainly include coulomb counting, open-circuit voltage methods, and model-based methods. These methods typically rely on fixed parameters and ideal operating conditions. However, in practical applications, the operating environment of lithium batteries is often affected by external factors such as temperature and load variations, leading to inaccurate estimation results from traditional methods. In lithium battery applications, geographical location, day-night cycles, and seasonal changes significantly alter ambient temperature, affecting the electrochemical reaction rate and internal resistance of lithium batteries, thus impacting the accurate estimation of SOC. Especially under extreme temperature conditions, the errors of traditional methods are more significant. Therefore, data-driven methods, particularly those based on deep learning, are widely used.
[0004] In the field of temperature-robust SOC estimation methods, although deep learning models have demonstrated strong predictive capabilities in many applications, their "black box" nature and lack of interpretability often lead users to doubt their effectiveness.
[0005] Therefore, it is necessary to design a new method that can more accurately estimate SOC in complex and variable working environments, effectively reduce errors, improve the accuracy of SOC estimation, and has high reliability. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for estimating the state of charge of a battery based on an embedded domain knowledge neural network.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: a battery state-of-charge estimation method based on an embedded domain knowledge neural network, comprising:
[0008] Acquire battery operating data, wherein the battery operating data includes battery discharge terminal voltage, current, temperature data and corresponding battery capacity data;
[0009] The battery operating data is input into an embedded domain knowledge neural network to estimate the battery state of charge (SOC) and obtain the estimation result. The embedded domain knowledge neural network is trained by inputting historical battery operating data, corresponding time values, and historical SOC as a sample set into the SocNet neural network model. The SocNet neural network model includes a feature extraction module with a special path, a SOC lookback module, and a summation module.
[0010] Output the estimation results.
[0011] Its further technical solution is as follows: the embedded domain knowledge neural network is obtained by inputting historical battery operation data, corresponding time values, and historical SOC as a sample set into the SocNet neural network model for training, including:
[0012] Define the current SOC of the battery;
[0013] Acquire historical battery operating data, corresponding time values, and historical SOC;
[0014] The historical battery operating data, corresponding time values, and historical SOC are normalized and combined with the current SOC of the battery to form a sample set;
[0015] The sample set is divided into a training set, a validation set, and a test set;
[0016] Construct the SocNet neural network model;
[0017] 10-fold cross-validation is performed using the training and validation sets. The average network loss is calculated. When the average network loss is less than a set threshold, the validation ends, and the optimal hyperparameters of the SocNet neural network model are recorded.
[0018] The training set and validation set are combined into a new training set. The SocNet neural network model is trained using the optimal network hyperparameter settings, and the model is saved.
[0019] The saved model is tested using the test set, and the estimated battery capacity value is recorded to obtain the embedded domain knowledge neural network.
[0020] The further technical solution is as follows: the feature extraction module includes a convolutional neural network layer, a pooling and flattening layer, a special path, and a first multilayer perceptron module connected in sequence; wherein, the first multilayer perceptron module includes an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0021] A further technical solution is as follows: The step of inputting the battery operating data into an embedded domain knowledge neural network to estimate the battery state of charge, in order to obtain the estimation result, includes:
[0022] The battery operating data is input into an embedded domain knowledge neural network;
[0023] The first hidden state is generated using the feature extraction module;
[0024] The second hidden state is generated using the SOC lookback module;
[0025] The first hidden state and the second hidden state are summed to obtain the estimation result.
[0026] The further technical solution is as follows: the generation of the first hidden state using the feature extraction module includes:
[0027] The spatial and temporal relationships between voltage, current, temperature, and historical SOC in the battery operating data are extracted using convolutional neural network layers.
[0028] The output of the convolutional neural network layer is pooled and flattened to form a feature vector.
[0029] The current and time values at the current moment are concatenated with the feature vector to form a new input vector;
[0030] The concatenated feature vectors are processed to generate the first hidden state.
[0031] The further technical solution is as follows: the generation of the second hidden state using the SOC lookback module includes:
[0032] Save the SOC value over a past period of time;
[0033] The SOC value is processed by a second multilayer perceptron module to obtain a second hidden state. The second multilayer perceptron module includes an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer connected in sequence.
[0034] This invention also provides a battery state-of-charge estimation system based on an embedded domain knowledge neural network, comprising:
[0035] A data acquisition unit is used to acquire battery operating data, wherein the battery operating data includes battery discharge terminal voltage, current, temperature data and corresponding battery capacity data;
[0036] An estimation unit is used to input the battery operating data into an embedded domain knowledge neural network to estimate the battery state of charge (SOC) and obtain an estimation result. The embedded domain knowledge neural network is trained by inputting historical battery operating data, corresponding time values, and historical SOC as a sample set into a SocNet neural network model. The SocNet neural network model includes a feature extraction module with a special path, a SOC lookback module, and a summation module.
[0037] The output unit is used to output the estimation result.
[0038] The further technical solution is as follows: the estimation unit includes:
[0039] An input subunit is used to input the battery operation data into an embedded domain knowledge neural network;
[0040] The feature extraction subunit is used to generate the first hidden state using the feature extraction module.
[0041] The lookback subunit is used to generate the second hidden state using the SOC lookback module;
[0042] The summation subunit is used to sum the first hidden state and the second hidden state to obtain the estimation result.
[0043] The further technical solution is as follows: the feature extraction subunit includes:
[0044] The relationship extraction module is used to extract the spatial and temporal relationships between voltage, current, temperature and historical SOC in the battery operating data using a convolutional neural network layer;
[0045] The pooling and flattening module is used to pool and flatten the output of the convolutional neural network layer to form a feature vector.
[0046] The connection module is used to connect the current and time values at the current moment with the feature vector to form a new input vector;
[0047] The vector processing module is used to process the concatenated feature vectors to generate the first hidden state.
[0048] Its further technical solution is as follows: the retrospective subunit includes:
[0049] The storage module is used to save the SOC value over a past period of time;
[0050] The SOC value processing module is used to process the SOC value using a second multilayer perceptron module to obtain a second hidden state. The second multilayer perceptron module includes a sequentially connected input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0051] The advantages of this invention compared to existing technologies are as follows: This invention acquires battery discharge terminal voltage, current, temperature, and capacity data to form a complete battery operation dataset. This data is then input into a SocNet neural network with embedded domain knowledge for state of charge (SOC) estimation. SocNet, trained using historical data and a SOC sample set, features feature extraction, SOC retrospective, and summation modules, effectively identifying complex patterns. The model's dynamic adaptability enhances its accuracy and stability in changing environments. Finally, the output estimation results achieve high-precision and high-reliability SOC estimation.
[0052] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description
[0053] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0054] Figure 1 This is a schematic diagram illustrating an application scenario of the battery state-of-charge estimation method based on an embedded domain knowledge neural network provided in an embodiment of the present invention.
[0055] Figure 2 A flowchart illustrating the battery state-of-charge estimation method based on an embedded domain knowledge neural network provided in an embodiment of the present invention;
[0056] Figure 3 A schematic block diagram of a battery state-of-charge estimation system based on an embedded domain knowledge neural network provided in an embodiment of the present invention;
[0057] Figure 4 A schematic block diagram of a computer device provided in an embodiment of the present invention;
[0058] Figure 5 A schematic diagram illustrating the workflow of the embedded domain knowledge neural network provided in an embodiment of the present invention;
[0059] Figure 6 This is a schematic diagram of the structure of the first multilayer sensor module provided in an embodiment of the present invention;
[0060] Figure 7 This is a schematic diagram of the structure of the second multilayer sensor module provided in an embodiment of the present invention;
[0061] Figure 8 This is a schematic diagram of the training process of an embedded domain knowledge neural network provided in an embodiment of the present invention. Detailed Implementation
[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0063] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.
[0064] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0065] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0066] Please see Figure 1 and Figure 2 , Figure 1 This is a schematic diagram illustrating an application scenario of the battery state-of-charge estimation method based on an embedded domain knowledge neural network provided in an embodiment of the present invention. Figure 2This is a schematic flowchart illustrating a battery state-of-charge (SOC) estimation method based on an embedded domain knowledge neural network, provided in an embodiment of the present invention. This SOC estimation method is applied in a server. The server interacts with the terminal. The current and time values are combined with the output of the convolutional neural network to reduce potential information loss during network layer processing. Simultaneously, a SOC lookback mechanism uses historical SOC data as input to a multilayer perceptron to adjust the current SOC based on past SOCs. Finally, the output of the feature extraction module with a specific path is added to the SOC lookback output to obtain the current SOC value. This structural design incorporates domain knowledge from the Coulomb counting method. Specifically, the SOC estimation model calculates battery capacity by considering the relationship between voltage, current, temperature, time during battery discharge, and historical SOCs. By monitoring battery discharge parameters and combining time with recorded historical SOCs, the battery's SOC can be estimated more accurately. This method integrates domain knowledge into the neural network, significantly improving the model's interpretability and robustness, enabling it to effectively learn complex nonlinear relationships and maintain stable performance under different temperature conditions. By combining the advantages of data-driven and knowledge-driven approaches, the model can not only provide accurate SOC estimates but also offer reasonable interpretations of the results.
[0067] Figure 2 This is a flowchart illustrating the battery state-of-charge estimation method based on an embedded domain knowledge neural network provided in an embodiment of the present invention. Figure 2 As shown, the method includes the following steps S110 to S150.
[0068] S110. Obtain battery operating data, wherein the battery operating data includes battery discharge terminal voltage, current, temperature data and corresponding battery capacity data.
[0069] In addition, the battery operation data also includes the corresponding time.
[0070] The process of acquiring battery operating data is crucial, and this data includes the following aspects:
[0071] Battery discharge terminal voltage: Monitoring changes in the battery's terminal voltage during discharge reflects the battery's health and discharge efficiency. Voltage variations help analyze battery performance under different loads and identify potential faults.
[0072] Current: Recording current data helps understand the battery's discharge intensity and rate. The relationship between current and battery capacity is fundamental to SOC (State of Charge) estimation; by analyzing current changes, the remaining energy of the battery can be better predicted.
[0073] Temperature data: Temperature is a crucial factor affecting battery performance. Batteries may operate differently at different temperatures. Monitoring temperature changes can help optimize battery usage conditions and provide appropriate protection and management in high or low temperature environments.
[0074] Battery capacity data: Obtaining battery capacity data is crucial for understanding battery status, as capacity reflects the total amount of energy a battery can store. Capacity data allows for the assessment of remaining battery power and lifespan, thereby optimizing battery usage strategies.
[0075] Corresponding time: Recording the timestamps for each set of data is crucial, as it establishes a correlation between battery operating data and time. This is significant for analyzing battery performance trends, making time-series predictions, and optimizing charging / discharging strategies.
[0076] By integrating various operational data, the battery's State of Charge (SOC) can be estimated more accurately, reducing the risk of information loss and improving the reliability of the estimation results. Visualizing the relationship between various battery operating parameters and SOC makes the model output more transparent, facilitating user understanding and analysis. Monitoring operational data under different environmental conditions allows the battery management system to adjust its operating strategies in real time to adapt to temperature changes and load demands, ensuring the battery operates at its optimal state. Real-time analysis of battery operating data enables the rapid identification of potential faults or anomalies, allowing for preventative measures and extending battery life.
[0077] The accumulated operational data can be used for subsequent algorithm optimization and model training, improving the intelligence level of the battery management system and making it more adaptable to the needs of different application scenarios.
[0078] In summary, obtaining comprehensive battery operation data not only improves the intelligence level of battery management, but also significantly enhances battery performance and safety, providing users with a better user experience.
[0079] S120. The battery operating data is input into an embedded domain knowledge neural network to estimate the battery state of charge (SOC) and obtain the estimation result. The embedded domain knowledge neural network is trained by inputting historical battery operating data, corresponding time values, and historical SOC as a sample set into the SocNet neural network model. The SocNet neural network model includes a feature extraction module with a special path, a SOC lookback module, and a summation module.
[0080] In one embodiment, such as Figure 8 As shown, the aforementioned embedded domain knowledge neural network is trained by inputting historical battery operation data, corresponding time values, and historical SOC as a sample set into the SocNet neural network model, including:
[0081] Define the current SOC of the battery; specifically, Among them, C t C0 represents the battery capacity at time t, and C0 represents the initial capacity of the battery. If capacity loss occurs, the capacity value of the previous time step is assigned to the time of loss.
[0082] Acquire historical battery operating data, corresponding time values, and historical SOC;
[0083] The historical battery operating data, corresponding time values, and historical SOC are normalized and combined with the current SOC of the battery to form a sample set;
[0084] The sample set is divided into a training set, a validation set, and a test set; specifically, the sample set is divided into training, validation, and test sets according to temperature. The battery dataset is collected within the range of the lowest operating temperature T. min Up to the highest operating temperature T max Choose an ambient temperature of T min T max And in T min To T max The temperature selected at the middle interval is used as the training set, and the other temperatures are used as the validation set and test set, so that the ratio of the training set, validation set and test set is close to 6:1:3.
[0085] Construct the SocNet neural network model;
[0086] 10-fold cross-validation is performed using the training and validation sets. The average network loss is calculated. When the average network loss is less than a set threshold, the validation ends, and the optimal hyperparameters of the SocNet neural network model are recorded. These include the learning rate, batch size, network backtracking length, convolutional kernel size, and number of hidden layer neurons.
[0087] The training set and validation set are combined into a new training set. The SocNet neural network model is trained using the optimal network hyperparameter settings, and the model is saved.
[0088] The saved model is tested using the test set, and the estimated battery capacity value is recorded to obtain the embedded domain knowledge neural network.
[0089] By using historical battery operating data, corresponding time values, and historical SOC as input, the model can learn battery behavior more comprehensively, thereby improving the accuracy of SOC estimation. Handling capacity loss allows the model to still provide reasonable SOC estimates when facing battery performance degradation, enhancing its robustness. Normalizing the input data eliminates the influence between different units, making model training more stable and convergence faster. Dividing the training, validation, and test sets according to temperature ensures the model's generalization ability under different environmental conditions, making it more reliable in practical applications. Using 10-fold cross-validation effectively evaluates model performance, reduces the risk of overfitting, and improves the model's predictive ability on new data. Recording the optimal network hyperparameters improves the model's learning efficiency and accuracy, making the training process more efficient. Saving the trained model facilitates rapid deployment and use in subsequent practical applications, enhancing the model's practicality.
[0090] This method can adapt to different operating temperature ranges, enhancing the model's applicability under various environmental conditions and ensuring good performance in different application scenarios. Evaluating the capabilities of the saved model through a test set provides reliable battery capacity predictions for practical applications, helping users better manage and utilize batteries.
[0091] In summary, this method not only improves the accuracy and reliability of SOC estimation, but also enhances the model's adaptability under various environmental conditions, providing strong support for the optimization of battery management systems.
[0092] In addition, the feature extraction module includes a convolutional neural network layer, a pooling and flattening layer, a special path, and a first multilayer perceptron module connected in sequence; wherein, the first multilayer perceptron module includes an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0093] In one embodiment, such as Figure 5 As shown, step S120 above may include steps S121 to S124.
[0094] S121. Input the battery operation data into the embedded domain knowledge neural network;
[0095] S122. Generate the first hidden state using the feature extraction module.
[0096] In this embodiment, the first hidden state refers to the output of the multilayer perceptron (MLP) module, representing an intermediate representation obtained after feature extraction and processing. This representation contains important features and information of battery operating data (such as voltage, current, temperature, and historical SOC).
[0097] Specifically, the first hidden state is generated through the following steps:
[0098] Feature extraction: Convolutional neural network (CNN) layers are used to extract the spatial and temporal relationships of the input data.
[0099] Pooling and flattening: The CNN output is pooled and then flattened into a feature vector, forming a vector that can be input into subsequent models.
[0100] Feature concatenation: Concatenates the current and time information at the current moment with the feature vector to form a new input vector (length is...). ).
[0101] Multilayer perceptron processing: The concatenated vector is input into the MLP module and processed through a fully connected layer and activation function to generate the first hidden state (h1).
[0102] The role of the first hidden state is to serve as input to the subsequent model, helping the model capture the latent features of the battery state, thereby improving the ability to estimate and predict battery performance.
[0103] In one embodiment, step S122 described above may include steps S1221 to S1224.
[0104] S1221. Use a convolutional neural network layer to extract the spatial and temporal relationships between voltage, current, temperature and historical SOC in the battery operating data.
[0105] The spatial and temporal relationships between voltage, current, temperature, and historical SOC refer to the interactions and dependencies of these variables at different points in time.
[0106] Spatial relationships: This involves the interactions between voltage, current, temperature, and state of charge (SOC) at any given moment. For example, a change in voltage at a given point in time may affect the state of current and SOC. Convolutional neural networks (CNNs) can effectively capture the correlations and patterns between these variables through their convolutional layers.
[0107] Temporal relationships: This refers to the dynamic changes of these variables over time. For example, changes in current may affect changes in SOC after a certain delay, or changes in temperature may show a certain pattern with changes in voltage. By processing historical data, convolutional neural networks can identify these temporal dependencies and trends.
[0108] In summary, convolutional neural networks can learn the complex spatial and temporal relationships between these variables through feature extraction, thereby more accurately estimating the state and performance of the battery.
[0109] Specifically, the Convolutional Neural Network (CNN) layer structure is used as the starting point for feature extraction to extract the spatial relationships between different inputs and the temporal relationships between different historical quantities. The input sequences of the neural network, namely voltage, current, temperature, and SOC sequences of historical length L, and their corresponding time sequences, are stacked into a 5×L matrix H. This matrix passes through two convolutional layers, named Conv1 and Conv2, with a kernel size of (10, 3).
[0110] S1222. Pool and flatten the output of the convolutional neural network layer to form a feature vector.
[0111] In this embodiment, the output of the second convolutional layer Conv2 is flattened into a feature vector after passing through a pooling layer. Its length is
[0112] S1223. Connect the current and time values at the current moment with the feature vector to form a new input vector.
[0113] In this embodiment, a special path is established to connect the current at the current moment and the corresponding time value with... Connect them to form a length of The vector.
[0114] S1224. Process the concatenated feature vectors to generate the first hidden state.
[0115] In this embodiment, the first multi-layer perceptron (MLP) module is used at the end of feature extraction and is named MLP1. The input of MLP1 is the length described in S213. The feature vectors of a specific path are connected, and the output is the hidden state h1.
[0116] Specifically, such as Figure 6 As shown, the structure of the MLP1 module consists of an input layer, a fully connected hidden layer with ReLU activation, and a linear output layer. The input to the MLP1 module is denoted as x1, and the hidden layer weight matrix and bias vector are W1 and W2, respectively. 1,1 and B 1,1 .
[0117] The output of the hidden layer is represented as: h 1,0 =ReLU(W 1,1 x1+b 1,1 );
[0118] The ReLU activation function is defined as: ReLU(z) = max(0,z);
[0119] The weight matrix and bias vector of the output layer are W, respectively. 1,2 and b 1,2 The output h1 of MLP1 is expressed as: h1 = W 1,2 h 1,0 +b 1,2 ;
[0120] Substituting the output of the hidden layer into the expression, the complete MLP1 model is represented as: h1 = W 1,2 ReLU(W 1,1 x1+b 1,1 )+b 1,2 .
[0121] S123. Use the SOC lookback module to generate the second hidden state.
[0122] In this embodiment, the second hidden state refers to the hidden layer output generated by the second multilayer perceptron module. This process is to incorporate information from historical SOC values to improve the estimation of the current state.
[0123] Specifically, the second hidden state is the processed output of MLP2, reflecting the understanding and modeling of past SOC values. Combined with the first hidden state, this provides more comprehensive information for estimating the current SOC.
[0124] By adding the first and second hidden states, the model can comprehensively utilize current information and historical SOC data, thereby improving the accuracy of SOC estimation. This mechanism ensures that the model has stronger robustness in the face of dynamic and complex environments.
[0125] In one embodiment, step S123 described above may include steps S1231 to S1232.
[0126] S1231, Save the SOC value for a past period of time.
[0127] In this embodiment, the SOC value of a past time period of length L is saved as SOC. t-L:t-1 During the training of the SocNet neural network, the true SOC values from the training data are used; during the testing and application of the SocNet neural network, past estimation results of the model are saved as the SOC. t-L:t-1 .
[0128] S1232. The SOC value is processed by a second multilayer perceptron module to obtain a second hidden state, wherein the second multilayer perceptron module includes an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer connected in sequence.
[0129] like Figure 7 As shown, the second multilayer perceptron module MLP2 is used at the end of the SOC lookback mechanism. The input of NLP2 is the SOC described in S221. t-L:t-1 The output is the hidden state h2.
[0130] The input to the MLP2 module is denoted as x2, and the hidden layer weight matrix and bias vector are W, respectively. 2,1 and b 2,1 .
[0131] The output of the hidden layer is represented as: h 2,0 =ReLU(W 2,1 x2+b 2,1 );
[0132] The ReLU activation function is defined as: ReLU(z) = max(0,z);
[0133] The weight matrix and bias vector of the output layer are W, respectively. 2,2 and b 2,2 The output h2 of MLP2 is represented as:
[0134] h2 = W 2,2 h 2,0 +b 2,2 ;
[0135] Substituting the output of the hidden layer into the expression, the complete MLP1 model is represented as: h2 = W 2,2 ReLU(W 2,1 x1+b 2,1 )+b 2, 2.
[0136] S124. Sum the first hidden state and the second hidden state to obtain the estimation result.
[0137] Specifically, Among them, I t Let be the current at time t. The output SOC of the neural network model SocNet is... t , is the sum of the outputs of MLP1 and MLP2.
[0138] By extracting key features from battery operating data (such as voltage, current, temperature, and SOC), the model can more accurately reflect the true state of the battery, thereby improving the ability to predict battery performance.
[0139] Convolutional neural networks effectively learn the spatial and temporal relationships in battery data, which helps to identify dependencies and dynamic changes between variables, thus providing a more comprehensive understanding of battery performance.
[0140] By storing SOC values from a past period, the model can leverage historical information to enhance its judgment of the current state and improve the reliability of its predictions.
[0141] The adoption of a step-by-step modular approach (feature extraction, SOC retrospective, etc.) results in a clear system structure that is easy to maintain and improve. Different modules can be optimized independently, flexibly adapting to different application scenarios.
[0142] Using a multilayer perceptron to process features can effectively reduce computational complexity, while the ReLU activation function can accelerate the training process.
[0143] In summary, this method, through the application of deep learning technology, can effectively improve the accuracy of battery state estimation and the predictive ability of the model, which is of great significance to the development of battery management systems.
[0144] S130. Output the estimation result.
[0145] The estimation results are output to the terminal for display.
[0146] The method in this embodiment, by embedding domain knowledge into a combination of coulomb counting and deep learning, can more accurately estimate the state of charge (SOC) of a battery in complex and variable working environments. Particularly under extreme temperature conditions, this method can significantly reduce errors and improve the accuracy of SOC estimation. The model's robustness to temperature variations is enhanced; the feature extraction module for specific pathways and the SOC lookback mechanism enable the model to maintain stable performance at different temperatures.
[0147] The SOC retrospective mechanism utilizes historical SOC data to dynamically correct and optimize the current SOC, ensuring the continuity and accuracy of the estimation results. This is particularly true under long-term operation and significant environmental changes, providing more reliable estimates. Geographical location differences, day-night cycles, and seasonal variations all affect ambient temperature, and this method effectively addresses these changes, ensuring the stability and reliability of the SOC estimation.
[0148] Existing deep learning models, due to their "black box" nature, often lack interpretability, leading users to doubt their effectiveness. The method in this embodiment, however, improves model transparency by incorporating domain knowledge, enabling users to better understand and trust the prediction results. This interpretability not only helps users understand the model's operating mechanism but also enhances their confidence in the results.
[0149] By combining data-driven and knowledge-driven approaches, and fully leveraging existing experience and data advantages, the overall performance of the model is improved, resulting in more accurate, stable, and easily understandable SOC estimation. This method is applicable to various lithium battery applications, including electric vehicles, renewable energy storage systems, and portable electronic devices, requiring only battery voltage, current, temperature, and corresponding time values.
[0150] The aforementioned battery state-of-charge (SOC) estimation method based on an embedded domain knowledge neural network acquires battery discharge terminal voltage, current, temperature, and capacity data to form a complete battery operation dataset. This data is then input into the SocNet neural network, which incorporates embedded domain knowledge, for SOC estimation. SocNet, trained on historical data and an SOC sample set, features feature extraction, SOC retrospective, and summation modules, effectively identifying complex patterns. The model's dynamic adaptability enhances its accuracy and stability in changing environments. Finally, the output estimation results achieve high-precision and high-reliability SOC estimation.
[0151] Figure 3 This is a schematic block diagram of a battery state-of-charge estimation system 300 based on an embedded domain knowledge neural network, provided in an embodiment of the present invention. Figure 3 As shown, corresponding to the above-described battery state-of-charge estimation method based on embedded domain knowledge neural networks, the present invention also provides a battery state-of-charge estimation system 300 based on embedded domain knowledge neural networks. This battery state-of-charge estimation system 300 includes a unit for executing the above-described battery state-of-charge estimation method based on embedded domain knowledge neural networks, and the system can be configured in a server. Specifically, please refer to... Figure 3 The battery state of charge estimation system 300 based on an embedded domain knowledge neural network includes a data acquisition unit 301, an estimation unit 302, and an output unit 303.
[0152] The data acquisition unit 301 is used to acquire battery operating data, including battery discharge terminal voltage, current, temperature data, and corresponding battery capacity data; the estimation unit 302 is used to input the battery operating data into an embedded domain knowledge neural network to estimate the battery state of charge (SOC) to obtain an estimation result; wherein the embedded domain knowledge neural network is trained by inputting historical battery operating data, corresponding time values, and historical SOC as a sample set into a SocNet neural network model, and the SocNet neural network model includes a feature extraction module with a special path, a SOC lookback module, and a summation module; the output unit 303 is used to output the estimation result.
[0153] In one embodiment, the estimation unit 302 includes:
[0154] The input subunit is used to input the battery operation data into the embedded domain knowledge neural network; the feature extraction subunit is used to generate a first hidden state using the feature extraction module; the lookback subunit is used to generate a second hidden state using the SOC lookback module; and the summation subunit is used to sum the first hidden state and the second hidden state to obtain an estimation result.
[0155] In one embodiment, the feature extraction subunit includes:
[0156] The relationship extraction module is used to extract the spatial and temporal relationships between voltage, current, temperature and historical SOC in the battery operating data using a convolutional neural network layer;
[0157] The pooling and flattening module is used to pool and flatten the output of the convolutional neural network layer to form a feature vector.
[0158] The connection module is used to connect the current and time values at the current moment with the feature vector to form a new input vector;
[0159] The vector processing module is used to process the concatenated feature vectors to generate the first hidden state.
[0160] In one embodiment, the retrospective subunit includes:
[0161] The storage module is used to save the SOC value over a past period of time;
[0162] The SOC value processing module is used to process the SOC value using a second multilayer perceptron module to obtain a second hidden state. The second multilayer perceptron module includes a sequentially connected input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0163] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the battery state of charge estimation system 300 based on embedded domain knowledge neural network and each unit can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.
[0164] The aforementioned battery state-of-charge estimation system 300 based on embedded domain knowledge neural networks can be implemented as a computer program, which can be used in various ways, such as... Figure 4 It runs on the computer device shown.
[0165] Please see Figure 4 , Figure 4This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.
[0166] See Figure 4 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.
[0167] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a battery state-of-charge estimation method based on an embedded domain knowledge neural network.
[0168] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.
[0169] The internal memory 504 provides an environment for the execution of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a battery state of charge estimation method based on an embedded domain knowledge neural network.
[0170] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0171] The processor 502 is used to run a computer program 5032 stored in the memory to perform the following steps:
[0172] The process involves acquiring battery operating data, including battery discharge terminal voltage, current, temperature data, and corresponding battery capacity data; inputting the battery operating data into an embedded domain knowledge neural network to estimate the battery state of charge (SOC) to obtain an estimation result; wherein the embedded domain knowledge neural network is trained by inputting historical battery operating data, corresponding time values, and historical SOC as a sample set into a SocNet neural network model, which includes a feature extraction module with a special path, a SOC lookback module, and a summation module; and outputting the estimation result.
[0173] In one embodiment, when the processor 502 implements the step of training the embedded domain knowledge neural network by inputting historical battery operation data, corresponding time values, and historical SOC as a sample set into the SocNet neural network model, the specific steps are as follows:
[0174] Define the current SOC of the battery; acquire historical battery operating data, corresponding time values, and historical SOC; normalize the historical battery operating data, corresponding time values, and historical SOC, and combine them with the current SOC to form a sample set; divide the sample set into a training set, a validation set, and a test set; construct a SocNet neural network model; perform 10-fold cross-validation using the training and validation sets, calculate the average network loss, and terminate the validation when the average network loss is less than a set threshold, recording the optimal hyperparameters of the SocNet neural network model; mix the training and validation sets into a new training set, set the SocNet neural network model using the optimal hyperparameters, train it, and save the model; test the saved model using the test set and record the estimated battery capacity value to obtain the embedded domain knowledge neural network.
[0175] The feature extraction module includes a convolutional neural network layer, a pooling and flattening layer, a special pathway, and a first multilayer perceptron module connected in sequence; wherein the first multilayer perceptron module includes an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0176] In one embodiment, when the processor 502 implements the step of inputting the battery operating data into an embedded domain knowledge neural network to estimate the battery state of charge and obtain the estimation result, the following steps are specifically implemented:
[0177] The battery operation data is input into the embedded domain knowledge neural network; a first hidden state is generated using the feature extraction module; a second hidden state is generated using the SOC lookback module; the first hidden state and the second hidden state are summed to obtain the estimation result.
[0178] In one embodiment, when implementing the step of generating the first hidden state using the feature extraction module, the processor 502 specifically implements the following steps:
[0179] The spatial and temporal relationships between voltage, current, temperature, and historical SOC in the battery operating data are extracted using a convolutional neural network layer; the output of the convolutional neural network layer is pooled and flattened to form a feature vector; the current current and time values are concatenated with the feature vector to form a new input vector; the concatenated feature vector is processed to generate the first hidden state.
[0180] In one embodiment, when implementing the step of generating the second hidden state using the SOC lookback module, the processor 502 specifically implements the following steps:
[0181] The SOC value over a past period is stored; the SOC value is processed by a second multilayer perceptron module to obtain a second hidden state, wherein the second multilayer perceptron module includes a sequentially connected input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0182] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may 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. The general-purpose processor may be a microprocessor or any conventional processor.
[0183] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.
[0184] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform the following steps:
[0185] The process involves acquiring battery operating data, including battery discharge terminal voltage, current, temperature data, and corresponding battery capacity data; inputting the battery operating data into an embedded domain knowledge neural network to estimate the battery state of charge (SOC) to obtain an estimation result; wherein the embedded domain knowledge neural network is trained by inputting historical battery operating data, corresponding time values, and historical SOC as a sample set into a SocNet neural network model, which includes a feature extraction module with a special path, a SOC lookback module, and a summation module; and outputting the estimation result.
[0186] In one embodiment, when the processor executes the computer program to implement the step of training the embedded domain knowledge neural network by inputting historical battery operation data, corresponding time values, and historical SOC as a sample set into the SocNet neural network model, the specific implementation steps are as follows:
[0187] Define the current SOC of the battery; acquire historical battery operating data, corresponding time values, and historical SOC; normalize the historical battery operating data, corresponding time values, and historical SOC, and combine them with the current SOC to form a sample set; divide the sample set into a training set, a validation set, and a test set; construct a SocNet neural network model; perform 10-fold cross-validation using the training and validation sets, calculate the average network loss, and terminate the validation when the average network loss is less than a set threshold, recording the optimal hyperparameters of the SocNet neural network model; mix the training and validation sets into a new training set, set the SocNet neural network model using the optimal hyperparameters, train it, and save the model; test the saved model using the test set and record the estimated battery capacity value to obtain the embedded domain knowledge neural network.
[0188] The feature extraction module includes a convolutional neural network layer, a pooling and flattening layer, a special pathway, and a first multilayer perceptron module connected in sequence; wherein the first multilayer perceptron module includes an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0189] In one embodiment, when the processor executes the computer program to implement the step of inputting the battery operating data into an embedded domain knowledge neural network to estimate the battery state of charge and obtain the estimation result, the specific implementation includes the following steps:
[0190] The battery operation data is input into the embedded domain knowledge neural network; a first hidden state is generated using the feature extraction module; a second hidden state is generated using the SOC lookback module; the first hidden state and the second hidden state are summed to obtain the estimation result.
[0191] In one embodiment, when the processor executes the computer program to implement the step of generating the first hidden state using the feature extraction module, it specifically implements the following steps:
[0192] The spatial and temporal relationships between voltage, current, temperature, and historical SOC in the battery operating data are extracted using a convolutional neural network layer; the output of the convolutional neural network layer is pooled and flattened to form a feature vector; the current current and time values are concatenated with the feature vector to form a new input vector; the concatenated feature vector is processed to generate the first hidden state.
[0193] In one embodiment, when the processor executes the computer program to implement the step of generating the second hidden state using the SOC lookback module, it specifically implements the following steps:
[0194] The SOC value over a past period is stored; the SOC value is processed by a second multilayer perceptron module to obtain a second hidden state, wherein the second multilayer perceptron module includes a sequentially connected input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer.
[0195] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.
[0196] 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, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. 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 implementations should not be considered beyond the scope of this invention.
[0197] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.
[0198] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the system of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention 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.
[0199] 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 storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.
[0200] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for battery state of charge estimation based on an embedded domain knowledge neural network, characterized in that, The method comprises: acquiring battery operation data, wherein the battery operation data comprises battery discharge terminal voltage, current, temperature data and corresponding battery capacity data; inputting the battery operation data into an embedded domain knowledge neural network for battery state of charge estimation to obtain an estimation result; wherein the embedded domain knowledge neural network is obtained by inputting historical battery operation data, corresponding time values and historical SOC as a sample set into the embedded domain knowledge neural network for training, and the embedded domain knowledge neural network comprises a feature extraction module with a special path, an SOC lookback module and a summation module; outputting the estimation result; wherein the feature extraction module comprises a convolutional neural network layer, a pooling and flattening layer, a special path and a first multilayer perceptron module connected in sequence; wherein the first multilayer perceptron module comprises an input layer, a fully connected hidden layer with a ReLU activation function and a linear output layer; the inputting the battery operation data into the embedded domain knowledge neural network for battery state of charge estimation to obtain an estimation result comprises: inputting the battery operation data into the embedded domain knowledge neural network; generating a first hidden state by using the feature extraction module; generating a second hidden state by using the SOC lookback module; summing the first hidden state and the second hidden state to obtain an estimation result; the generating a first hidden state by using the feature extraction module comprises: extracting the spatial and temporal relationships between voltage, current, temperature and historical SOC in the battery operation data by using a convolutional neural network layer; performing pooling and flattening processing on the output of the convolutional neural network layer to form a feature vector; connecting the current and time values at the current time with the feature vector to form a new input vector; processing the connected feature vector to generate a first hidden state.
2. The battery state of charge estimation method based on an embedded domain knowledge neural network according to claim 1, characterized in that, The embedded domain knowledge neural network is obtained by inputting historical battery operation data, corresponding time values and historical SOC as a sample set into the embedded domain knowledge neural network for training, and comprises: defining a current SOC of the battery; acquiring historical battery operation data, corresponding time values and historical SOC; performing normalization processing on the historical battery operation data, corresponding time values and historical SOC, and combining the current SOC of the battery to form a sample set; dividing the sample set into a training set, a validation set and a test set; constructing an embedded domain knowledge neural network; performing 10-fold cross-validation using the training set and the validation set to calculate the average network loss value, ending the validation when the average network loss value is less than a set threshold, and recording the optimal network hyperparameters of the embedded domain knowledge neural network; mixing the training set and the validation set into a new training set, setting the embedded domain knowledge neural network using the optimal network hyperparameters, training and saving the model; testing the saved model using the test set and recording the estimated battery capacity value to obtain the embedded domain knowledge neural network.
3. The battery state of charge estimation method based on an embedded domain knowledge neural network according to claim 1, characterized in that, The generating a second hidden state by using the SOC lookback module comprises: saving SOC values in the past period of time; The SOC value is processed by a second multi-layer perception module to obtain a second hidden state, wherein the second multi-layer perception module comprises an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer connected in sequence.
4. A battery state of charge estimation system based on an embedded domain knowledge neural network, characterized in that, The system uses the battery state of charge estimation method based on the embedded field knowledge neural network according to any one of claims 1 to 3, comprising: a data acquisition unit configured to acquire battery operation data, wherein the battery operation data comprises battery discharge terminal voltage, current, temperature data, and corresponding battery capacity data; an estimation unit configured to input the battery operation data into the embedded field knowledge neural network to estimate the battery state of charge to obtain an estimation result; wherein the embedded field knowledge neural network is obtained by inputting historical battery operation data, corresponding time values, and historical SOC as a sample set into the embedded field knowledge neural network for training, and the embedded field knowledge neural network comprises a feature extraction module with a special path, an SOC lookback module, and a summation module; an output unit configured to output the estimation result.
5. The embedded domain knowledge based neural network battery state of charge estimation system of claim 4, wherein, The estimation unit comprises: an input subunit configured to input the battery operation data into the embedded field knowledge neural network; a feature extraction subunit configured to generate a first hidden state by using the feature extraction module; a lookback subunit configured to generate a second hidden state by using the SOC lookback module; a summation subunit configured to sum the first hidden state and the second hidden state to obtain an estimation result.
6. The embedded domain knowledge based neural network battery state of charge estimation system of claim 5, wherein, The feature extraction subunit comprises: a relationship extraction module configured to extract spatial and temporal relationships between voltage, current, temperature, and historical SOC in the battery operation data by using a convolutional neural network layer; a pooling and flattening module configured to perform pooling and flattening processing on the output of the convolutional neural network layer to form a feature vector; a connection module configured to connect the current and time values at the current time with the feature vector to form a new input vector; a vector processing module configured to process the connected feature vector to generate a first hidden state.
7. The embedded domain knowledge based neural network battery state of charge estimation system of claim 6, wherein, The lookback subunit comprises: a saving module configured to save SOC values in the past period of time; an SOC value processing module configured to process the SOC values by using a second multi-layer perception module to obtain a second hidden state, wherein the second multi-layer perception module comprises an input layer, a fully connected hidden layer with a ReLU activation function, and a linear output layer connected in sequence.