Lithium ion battery state prediction method
By clustering temperature difference data and building neural network models for different categories, combining the second-order RC equivalent circuit model and LSTM model, the problems of environmental temperature impact and overfitting in lithium-ion battery status monitoring are solved, achieving more accurate The temperature and voltage prediction enhances the applicability and generalization capabilities of the model.
Patent Information
- Application Number
- CN202411511888.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-28
- Publication Date
- 2024-12-31
AI Technical Summary
The existing lithium-ion battery status monitoring technology ignores the impact of ambient temperature, and neural network models are prone to overfitting, resulting in inaccurate prediction results and difficult to generalize.
Using a neural network-based temperature prediction method, we use clustering temperature difference data and building models for different categories, combining the second-order RC equivalent circuit model and the LSTM model to predict voltage, and using the penalty term to optimize the loss function to reduce overfitting.
It improves the accuracy of temperature prediction and the accuracy of voltage prediction, enhances the applicability and generalization capabilities of the model, and can promptly identify temperature anomalies and reduce feature confusion.
Smart Images
Figure CN119224595A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of lithium ion battery state monitoring, and in particular to a lithium ion battery state prediction method. Background Art
[0002] Lithium-ion batteries are a type of battery widely used in the world today. They play a key role in many energy-related fields such as electric vehicles, energy storage systems, and portable electronic devices. In my country, in particular, the sales volume of electric vehicles has jumped to the first place in the world, making electric vehicles and their core components, the lithium-ion battery industry, one of the country's important pillar industries. Therefore, ensuring the safe operation of lithium-ion batteries has become particularly critical, and effective monitoring of their status is an important means to alleviate users' "safety anxiety".
[0003] Among the parameters of lithium-ion batteries, voltage and temperature are the two most important indicators. Accurately monitoring these parameters can not only detect potential safety hazards in a timely manner, but also help extend battery life and optimize its performance.
[0004] In response to this, the prior art has developed a lithium-ion battery temperature prediction technology based on a neural network. However, it still has the following shortcomings: on the one hand, it ignores the impact of ambient temperature on battery temperature; on the other hand, some studies have failed to fully solve the problem of overfitting of the neural network model, resulting in inaccurate prediction results or poor generalization ability.
[0005] In addition, in terms of lithium-ion battery voltage monitoring, existing technologies mainly rely on different types of battery models to achieve this. Depending on the construction method, battery models can be divided into two categories: physical models and machine learning models. Among them, physical models are based on known physical or chemical principles, with lower development difficulty and good stability. However, such models often find it difficult to fully consider the various complex interference factors that may be encountered during actual use. In contrast, machine learning models built using data-driven methods can better adapt to a variety of application scenarios and can generally provide higher prediction accuracy. However, machine learning models also face a significant challenge: how to avoid feature extraction bias caused by overfitting training data. Summary of the invention
[0006] The purpose of the present invention is to provide a lithium-ion battery state prediction method to solve the above technical problems.
[0007] To achieve the above object, the present invention provides a lithium ion battery state prediction method, comprising the following steps:
[0008] S1, predicting the current temperature of the lithium-ion battery based on the neural network model considering the ambient temperature;
[0009] S2, predicting the current voltage of the lithium-ion battery based on the second-order RC equivalent circuit model of the lithium-ion battery and the LSTM model;
[0010] S3, integrating the current temperature of the lithium-ion battery predicted in step S1 and the current voltage of the lithium-ion battery predicted in step S2, and outputting the prediction result of the current lithium-ion battery state.
[0011] Preferably, step S1 specifically includes the following steps:
[0012] S11, collecting lithium-ion battery temperature probe historical data and ambient temperature historical data;
[0013] S12, analyzing temperature difference data: calculating the difference between the lithium-ion battery temperature probe historical data and the ambient temperature historical data, and classifying them using a clustering method;
[0014] S13, constructing a neural network model: setting the neural network model hyperparameters, optimizer and loss function according to the length of the temperature difference data;
[0015] S14, training a neural network model: for the categories divided in step S12, the temperature difference data of the corresponding category of the mechanism is input into the neural network model constructed in step S13 for model training to obtain a neural network model of the corresponding category;
[0016] S15, collecting current ambient temperature data, and inputting the collected current ambient temperature data into the neural network model trained in step S14 to predict the current temperature of the lithium-ion battery;
[0017] S16, abnormal temperature warning: determine whether the difference between the current lithium-ion battery temperature predicted in step S15 and the temperature of the lithium-ion battery collected by the probe is greater than a threshold value, if so, output a temperature abnormality warning signal, if not, return to step S15.
[0018] Preferably, in step S12, the categories corresponding to each day are divided according to the category ratio of the temperature difference data of the day, and the temperature difference data category with the largest proportion is regarded as the category of the day.
[0019] Preferably, in step S14, a neural network model is separately established for each category of temperature difference data obtained in step S12, and each category of temperature difference data is used to train the corresponding neural network model.
[0020] Preferably, in step S16, the threshold of the temperature abnormality warning includes a first-level threshold range and a second-level threshold range, wherein the first-level threshold range is -2°C to 2°C, and the second-level threshold range is -4°C to -2°C and 2°C to 4°C, and when the difference between the current lithium-ion battery temperature and the temperature of the lithium-ion battery collected by the probe is within the first-level threshold range, a reminder warning signal is issued, and when the difference between the current lithium-ion battery temperature and the temperature of the lithium-ion battery collected by the probe is within the second-level threshold range, a warning warning signal is issued.
[0021] Preferably, step S2 specifically includes the following steps:
[0022] S21. Construct a second-order RC equivalent circuit model of a lithium-ion battery;
[0023] S22. Considering the voltage data and current data of the lithium-ion battery, an LSTM model for predicting the voltage of the lithium-ion battery is constructed;
[0024] S23, optimizing the LSTM model constructed in step S22 using a loss function with a penalty term;
[0025] S24, collecting and preprocessing the historical voltage and current data of the lithium-ion battery, and then dividing the preprocessed historical voltage and current data of the lithium-ion battery into a training set and a verification set according to a ratio;
[0026] S25, inputting the training set described in step S24 into the LSTM model optimized in step S23 for training, and then using the verification set to verify the trained LSTM model;
[0027] S26. Model generalization ability evaluation: Train LSTM models based on various voltage acquisition components respectively, calculate the error between the lithium-ion battery voltage data predicted by the LSTM model corresponding to the voltage acquisition component and the actual voltage data, and evaluate the generalization ability of the LSTM model based on the error;
[0028] S27, using the LSTM model after the generalization evaluation in step S26 to predict the current lithium-ion battery voltage data.
[0029] Preferably, the second-order RC equivalent circuit model of the lithium-ion battery described in step S21 includes two parallel RC circuits, and each RC circuit includes a resistor R and a capacitor C connected in series.
[0030] Preferably, the LSTM model described in step S22 includes an input layer for inputting voltage data and current data of the lithium-ion battery, a hidden layer for processing the input data, and an output layer for outputting the voltage data of the lithium-ion battery.
[0031] Preferably, the penalty term in step S23 is calculated as follows:
[0032] First, the logarithm of the absolute value of the difference between the output data of the lithium-ion battery second-order RC equivalent circuit model and the LSTM model is calculated. If the calculated result is within 5%, the penalty term is set to 0. Otherwise, the penalty term is calculated according to the following formula:
[0033]
[0034] In the formula, P represents the penalty term, V nn Represents the output voltage of the neural network model, V pm represents the output voltage of the second-order RC equivalent circuit model of the lithium-ion battery, and Δ is a threshold value, which is equal to 5% of the output data of the second-order RC equivalent circuit model of the lithium-ion battery.
[0035] Therefore, the present invention adopts the above, and has the beneficial effects of:
[0036] 1. In terms of temperature prediction, the data is divided into different categories according to temperature differences, and the corresponding model is trained for each category. This helps to distinguish the characteristics of different seasons and ambient temperatures, reduce the interference of temperature characteristics of other seasons, and avoid feature confusion, thereby reducing the difficulty of model training. In addition, the overfitting ability of the neural network is used to quickly identify that the battery has deviated from the normal operating state reflected by the training data when a temperature anomaly occurs, so that an inspection signal can be issued in time;
[0037] 2. In terms of voltage prediction, it not only improves the training speed of the neural network model, but also enables the machine learning model to learn the prior features in the physical model, effectively alleviating the overfitting problem and enhancing the applicability of the model on different data sets.
[0038] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 The present invention is a flowchart of a lithium-ion battery state prediction method. DETAILED DESCRIPTION
[0040] In the description of the present invention, it should be noted that the terms "upper", "lower", "inside", "outside", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, or the positions or positional relationships in which the invented product is usually placed when in use. 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 orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In the description of the present invention, it should also be noted that, unless otherwise clearly specified and limited, the terms "setting", "installation", and "connection" 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 an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0041] The embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.
[0042] like Figure 1 As shown, a lithium-ion battery state prediction method comprises the following steps:
[0043] S1, predicting the current temperature of the lithium-ion battery based on the neural network model considering the ambient temperature;
[0044] Step S1 specifically includes the following steps:
[0045] S11, collecting lithium-ion battery temperature probe historical data and ambient temperature historical data;
[0046] In this embodiment, the lithium-ion battery temperature probe data and the ambient temperature data are collected throughout the year.
[0047] S12, analyzing temperature difference data: calculating the difference between the lithium-ion battery temperature probe historical data and the ambient temperature historical data, and classifying them using a clustering method;
[0048] In step S12, the categories corresponding to each day are divided according to the category ratio of the temperature difference data of the day, and the temperature difference data category with the largest proportion is regarded as the category of the day.
[0049] S13, constructing a neural network model: setting the neural network model hyperparameters, optimizer and loss function according to the length of the temperature difference data;
[0050] S14, training the neural network model: for the categories divided in step S12, the temperature difference data of the corresponding categories of the mechanism are input into the neural network model constructed in step S13 for model training to obtain the neural network model of the corresponding category; and because the temperature anomaly of the lithium-ion battery is usually caused by thermal runaway and there is less fault data, the training model is mainly based on normal operating data.
[0051] In this embodiment, the temperature difference data is first classified, and then the model of the data classification corresponding to the day is used to predict the temperature of the lithium-ion battery probe.
[0052] In step S14, a neural network model is established separately for each category of temperature difference data obtained in step S12, and each category of temperature difference data is used to train the corresponding neural network model.
[0053] S15, collecting current ambient temperature data, and inputting the collected current ambient temperature data into the neural network model trained in step S14 to predict the current temperature of the lithium-ion battery;
[0054] S16, abnormal temperature warning: determine whether the difference between the current lithium-ion battery temperature predicted in step S15 and the temperature of the lithium-ion battery collected by the probe is greater than a threshold value, if so, output a temperature abnormality warning signal, if not, return to step S15.
[0055] In step S16, the threshold of the temperature abnormality warning includes a first-level threshold range and a second-level threshold range, wherein the first-level threshold range is -2°C to 2°C, and the second-level threshold range is -4°C to -2°C and 2°C to 4°C, and when the difference between the current lithium-ion battery temperature and the temperature of the lithium-ion battery collected by the probe is within the first-level threshold range, a reminder warning signal is issued, i.e., to remind the user to pay attention, and when the difference between the current lithium-ion battery temperature and the temperature of the lithium-ion battery collected by the probe is within the second-level threshold range, a warning warning signal is issued, i.e., to remind the user to stay away from the lithium-ion battery.
[0056] S2, predicting the current voltage of the lithium-ion battery based on the second-order RC equivalent circuit model of the lithium-ion battery and the LSTM model (long short-term memory neural network model);
[0057] Step S2 specifically includes the following steps:
[0058] S21. Construct a second-order RC equivalent circuit model of a lithium-ion battery;
[0059] The second-order RC equivalent circuit model of the lithium-ion battery described in step S21 includes two parallel RC loops, and each RC loop includes a resistor R and a capacitor C connected in series.
[0060] S22. Considering the voltage data and current data of the lithium-ion battery, an LSTM model for predicting the voltage of the lithium-ion battery is constructed;
[0061] The LSTM model described in step S22 includes an input layer for inputting voltage data and current data of the lithium-ion battery, a hidden layer for processing the input data, and an output layer for outputting the voltage data of the lithium-ion battery.
[0062] In this embodiment, the hyperparameters (learning rate, batch size, and number of iterations) of the LSTM model are adjusted according to the length of the input data and the length of the predicted data. For example, if the input data length is 24 hours and the predicted data length is 1 hour, a smaller learning rate (such as 0.001) and a larger batch size (such as 64) can be selected to ensure the stability of the model.
[0063] S23, optimizing the LSTM model constructed in step S22 using a loss function with a penalty term;
[0064] The penalty term calculation method described in step S23 is as follows:
[0065] First, the logarithm of the absolute value of the difference between the output data of the lithium-ion battery second-order RC equivalent circuit model and the LSTM model is calculated. If the calculated result is within 5%, the penalty term is set to 0. Otherwise, the penalty term is calculated according to the following formula:
[0066]
[0067] In the formula, P represents the penalty term, V nn Represents the output voltage of the neural network model, V pm represents the output voltage of the second-order RC equivalent circuit model of the lithium-ion battery, and Δ is a threshold value, which is equal to 5% of the output data of the second-order RC equivalent circuit model of the lithium-ion battery.
[0068] S24, collecting historical voltage and current data of the lithium-ion battery and performing preprocessing (denoising, normalization, and missing value processing), and then dividing the preprocessed historical voltage and current data of the lithium-ion battery into a training set and a validation set according to a ratio;
[0069] S25. Input the training set described in step S24 into the LSTM model optimized in step S23 for training, and then use the verification set to verify the trained LSTM model, use the verification set to evaluate the performance of the LSTM model, and use the early stopping method to prevent overfitting. The stopping condition of the LSTM model training is that the performance on the verification set no longer improves or reaches a preset number of iterations; and in this embodiment, when the voltage and current acquisition equipment is replaced or the environmental conditions change, the LSTM model needs to be retrained.
[0070] S26. Model generalization ability evaluation: Train LSTM models based on various voltage acquisition components respectively, calculate the error between the lithium-ion battery voltage data predicted by the LSTM model corresponding to the voltage acquisition component and the actual voltage data, and evaluate the generalization ability of the LSTM model based on the error;
[0071] S27, using the LSTM model after the generalization evaluation in step S26 to predict the current lithium-ion battery voltage data.
[0072] S3, integrating the current temperature of the lithium-ion battery predicted in step S1 and the current voltage of the lithium-ion battery predicted in step S2, and outputting the prediction result of the current lithium-ion battery state.
[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.
Claims
1. A lithium-ion battery state prediction method, characterized in that: The following steps are involved: S1, predicting the current temperature of the lithium-ion battery based on the neural network model considering the ambient temperature; S2, predicting the current voltage of the lithium-ion battery based on the second-order RC equivalent circuit model of the lithium-ion battery and the LSTM model; S3, integrating the current temperature of the lithium-ion battery predicted in step S1 and the current voltage of the lithium-ion battery predicted in step S2, and outputting the prediction result of the current lithium-ion battery state.
2. A lithium-ion battery state prediction method according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11, collecting lithium-ion battery temperature probe historical data and ambient temperature historical data; S12, analyzing temperature difference data: calculating the difference between the lithium-ion battery temperature probe historical data and the ambient temperature historical data, and classifying them using a clustering method; S13, constructing a neural network model: setting the neural network model hyperparameters, optimizer and loss function according to the length of the temperature difference data; S14, training a neural network model: for the categories divided in step S12, the temperature difference data of the corresponding category of the mechanism is input into the neural network model constructed in step S13 for model training to obtain a neural network model of the corresponding category; S15, collecting current ambient temperature data, and inputting the collected current ambient temperature data into the neural network model trained in step S14 to predict the current temperature of the lithium-ion battery; S16, abnormal temperature warning: determine whether the difference between the current lithium-ion battery temperature predicted in step S15 and the temperature of the lithium-ion battery collected by the probe is greater than a threshold value, if so, output a temperature abnormality warning signal, if not, return to step S15.
3. A lithium-ion battery state prediction method according to claim 2, characterized in that: In step S12, the categories corresponding to each day are divided according to the category ratio of the temperature difference data of the day, and the temperature difference data category with the largest proportion is regarded as the category of the day.
4. A lithium-ion battery state prediction method according to claim 3, characterized in that: In step S14, a neural network model is established separately for each category of temperature difference data obtained in step S12, and each category of temperature difference data is used to train the corresponding neural network model.
5. A lithium-ion battery state prediction method according to claim 4, characterized in that: In step S16, the threshold of the temperature abnormality warning includes a first-level threshold range and a second-level threshold range, wherein the first-level threshold range is -2°C to 2°C, and the second-level threshold range is -4°C to -2°C and 2°C to 4°C, and when the difference between the current lithium-ion battery temperature and the temperature of the lithium-ion battery collected by the probe is within the first-level threshold range, a reminder warning signal is issued, and when the difference between the current lithium-ion battery temperature and the temperature of the lithium-ion battery collected by the probe is within the second-level threshold range, a warning warning signal is issued.
6. A lithium-ion battery state prediction method according to claim 5, characterized in that: Step S2 specifically includes the following steps: S21. Construct a second-order RC equivalent circuit model of a lithium-ion battery; S22. Considering the voltage data and current data of the lithium-ion battery, an LSTM model for predicting the voltage of the lithium-ion battery is constructed; S23, optimizing the LSTM model constructed in step S22 using a loss function with a penalty term; S24, collecting and preprocessing the historical voltage and current data of the lithium-ion battery, and then dividing the preprocessed historical voltage and current data of the lithium-ion battery into a training set and a verification set according to a ratio; S25, inputting the training set described in step S24 into the LSTM model optimized in step S23 for training, and then using the verification set to verify the trained LSTM model; S26. Model generalization ability evaluation: Train LSTM models based on various voltage acquisition components respectively, calculate the error between the lithium-ion battery voltage data predicted by the LSTM model corresponding to the voltage acquisition component and the actual voltage data, and evaluate the generalization ability of the LSTM model based on the error; S27, using the LSTM model after the generalization evaluation in step S26 to predict the current lithium-ion battery voltage data.
7. A lithium-ion battery state prediction method according to claim 6, characterized in that: The second-order RC equivalent circuit model of the lithium-ion battery described in step S21 includes two parallel RC circuits, and each RC circuit includes a resistor R and a capacitor C connected in series.
8. A lithium-ion battery state prediction method according to claim 7, characterized in that: The LSTM model described in step S22 includes an input layer for inputting voltage data and current data of the lithium-ion battery, a hidden layer for processing the input data, and an output layer for outputting the voltage data of the lithium-ion battery.
9. A lithium-ion battery state prediction method according to claim 8, characterized in that: The penalty term in step S23 is calculated as follows: First, the logarithm of the absolute value of the difference between the output data of the lithium-ion battery second-order RC equivalent circuit model and the LSTM model is calculated. If the calculated result is within 5%, the penalty term is set to 0. Otherwise, the penalty term is calculated according to the following formula: In the formula, P represents the penalty term, V nn Represents the output voltage of the neural network model, V pm represents the output voltage of the second-order RC equivalent circuit model of the lithium-ion battery, and Δ is a threshold value, which is equal to 5% of the output data of the second-order RC equivalent circuit model of the lithium-ion battery.