CN114355199A - Battery thermal runaway risk prediction method and system based on recurrent neural network - Google Patents
Battery thermal runaway risk prediction method and system based on recurrent neural network Download PDFInfo
- Publication number
- CN114355199A CN114355199A CN202111661180.4A CN202111661180A CN114355199A CN 114355199 A CN114355199 A CN 114355199A CN 202111661180 A CN202111661180 A CN 202111661180A CN 114355199 A CN114355199 A CN 114355199A
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- recurrent neural
- network model
- sample data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Combined Controls Of Internal Combustion Engines (AREA)
Abstract
The invention relates to a battery thermal runaway risk prediction method and system based on a recurrent neural network, wherein the method comprises the following steps: selecting characteristic data from pre-collected historical data used for training a recurrent neural network model, and performing data cleaning on the characteristic data; generating sample data with a fixed time length based on the characteristic data after cleaning, wherein the sample data comprises sample data of a normal vehicle and sample data of a problem vehicle; constructing a recurrent neural network model; defining an output of a recurrent neural network model; acquiring sample data of a normal vehicle by the recurrent neural network model and training; and verifying the trained recurrent neural network model by using the sample data of the vehicle in question. The method comprises the steps of acquiring acquired real-time data through a model, calculating according to a certain frequency, then obtaining a prediction result of the model, comparing the prediction result with an actual result, and calculating the deviation degree which can be used as the probability that the thermal runaway fault of the battery can occur.
Description
Technical Field
The invention belongs to the technical field of thermal runaway risk monitoring of power batteries of electric vehicles, and particularly relates to a battery thermal runaway risk prediction technology based on a recurrent neural network.
Background
The performance of the power battery, which is taken as a core component of the electric automobile, directly determines most key performances of the electric automobile. With the increasing popularity of electric vehicles, various problems caused by power batteries are highlighted and receive much attention. In order to reduce the fault level of the power battery, in addition to better raw materials are selected in the production and manufacturing process of the battery, more advanced production process and stricter quality control are adopted, the thermal runaway risk of the battery also needs to be found in advance during the use process of the battery, and the battery is prevented from getting ill.
At present, the fault early warning of the battery is mainly divided into two modes of local BMS early warning and big data online early warning. With the wide application of big data technology, the online early warning method based on big data is more and more emphasized. The chinese patent publication CN109978229A discloses a technique entitled "a method for predicting thermal runaway of a full-cell multi-point temperature and a connection point temperature of a power battery pack", which mainly includes acquiring real-time data of a vehicle, building a model and a training verification algorithm through big data machine learning, building an XGBoost model, a weibull distribution and a bayesian network model, and performing thermal runaway prediction of an electric vehicle battery according to temperature characteristics obtained by the model. The technology is mainly used for predicting the thermal runaway of the battery according to the temperature, and the prediction parameter is single.
Chinese patent publication No. CN110161414B discloses a technique entitled "power battery thermal runaway online prediction method and system", which mainly calculates a voltage deviation matrix at each time according to a voltage value of each battery cell in a power battery, and then inputs voltage deviation growth rates of each cell corresponding to a current driving mileage of an automobile, a current temperature average value of a temperature probe, and a current voltage deviation growth rate matrix at a current time T into a thermal runaway cell prediction model to obtain a power battery thermal runaway prediction result. The technology mainly focuses on how to construct the voltage offset growth rate matrix, and a specific construction method of a model is not given.
Disclosure of Invention
The invention aims to provide a battery thermal runaway risk prediction method and system based on a recurrent neural network, which solve the technical problems that: in the related art, a method for better predicting the thermal runaway risk of the battery does not exist, so that the thermal runaway risk of the battery cannot be accurately and timely predicted.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows: a battery thermal runaway risk prediction method based on a recurrent neural network comprises the following steps:
s01, selecting characteristic data from pre-collected historical data used for training the recurrent neural network model, and cleaning the characteristic data;
s02: generating sample data with a fixed time length based on the characteristic data after cleaning, wherein the sample data comprises sample data of a normal vehicle and sample data of a problem vehicle;
s03: constructing the recurrent neural network model;
s04: defining an output of the recurrent neural network model;
s05: the cyclic neural network model obtains sample data of the normal vehicle and trains the sample data;
s06: verifying the trained recurrent neural network model by using sample data of the problem vehicle;
s07: repeating the S05 to S06 until an optimal recurrent neural network model is obtained;
s08: the optimal recurrent neural network model obtains vehicle real-time data, a prediction result is obtained through calculation according to a set frequency, the prediction result is compared with an actual result to obtain a deviation degree, and when the deviation degree is larger than a threshold value, the battery can be judged to have a thermal runaway risk.
Preferably, the first and second electrodes are formed of a metal,
in the S01, the historical data includes laboratory data and user real vehicle data;
the characteristic data comprises total current, total voltage, SOC, vehicle speed, accumulated mileage, monomer voltage and temperature collected by a plurality of temperature sensors;
the method for performing data cleaning on the characteristic data comprises the following steps: clearing invalid data, supplementing missing data and normalizing.
Preferably, the first and second electrodes are formed of a metal,
the formula for carrying out normalization processing on the characteristic data is as follows:
Preferably, the first and second electrodes are formed of a metal,
in S02, a total of P1 sample data of a fixed time length are generated by a sliding window method, where the number of sample data of a normal vehicle is P2, the number of sample data of a problem vehicle is P3, and the size of each sample data is an m × n matrix, where m is a time length and n is a feature number.
Preferably, the first and second electrodes are formed of a metal,
in S03, the recurrent neural network model is composed of a plurality of basic recurrent neural units and a full connection layer, where the basic recurrent neural units include simple recurrent neural units, LSTM units, and GRU units;
the recurrent neural network model can also adopt a self-defined recurrent neural unit.
Preferably, the first and second electrodes are formed of a metal,
in S04, the output of the defined recurrent neural network model, i.e., the prediction mode, is two, one of which predicts all data at the time of m1 based on the acquired data at one time, and the other of which predicts the next time by using a recurrent prediction method, i.e., the output of each time is the prediction of the next time, and then predicts the result of the next time by using the data of the last m-P4 length and the predicted data at the time of P4 in the sample data until all data at the time of m1 are obtained.
Preferably, the first and second electrodes are formed of a metal,
in the step S05, the recurrent neural network model obtains sample data of P2 normal vehicles, and trains the samples, where an evaluation index of the recurrent neural network model is an average square error MSE; optimizing the training process of the recurrent neural network model by adopting a regularization and batch normalization mode, wherein the MSE is defined as a parameter MSEIs normal;
In the S06, verifying the trained recurrent neural network model by using the sample data of the P3 problematic vehicles, wherein the evaluation index of the verification result is MSE, and the MSE is defined as a parameter MSEProblem(s);
In the S07, repeating the S05 to S06 until MSE is foundIs normalSmaller, MSEProblem(s)A larger optimal recurrent neural network model;
in the step S08, the optimal recurrent neural network model obtains vehicle real-time data, and the MSE is obtained through calculationReal timeThe calculation formula of the deviation degree R is as follows:
the invention also provides a battery thermal runaway risk prediction system based on the recurrent neural network, which comprises the following steps:
the cleaning module is used for selecting characteristic data from pre-collected historical data used for training the recurrent neural network model and cleaning the characteristic data;
the generating module is used for generating sample data with a fixed time length based on the characteristic data after cleaning, wherein the sample data comprises sample data of a normal vehicle and sample data of a problem vehicle;
the building module is used for building the recurrent neural network model;
a definition module for defining an output of the recurrent neural network model;
the training module is used for acquiring the sample data of the normal vehicle by the recurrent neural network model and training the sample data;
the verification module is used for verifying the trained recurrent neural network model by using sample data of the problem vehicle;
the circulating module is used for repeating the training module to the verification module until the optimal circulating neural network model is obtained;
and the comparison module is used for acquiring vehicle real-time data by the optimal cyclic neural network model, calculating according to a set frequency to obtain a prediction result, comparing the prediction result with an actual result to obtain a deviation degree, and judging that the battery has a thermal runaway risk when the deviation degree is greater than a threshold value.
Preferably, the first and second electrodes are formed of a metal,
in the cleaning module, the historical data comprises laboratory data and user real vehicle data;
the characteristic data comprises total current, total voltage, SOC, vehicle speed, accumulated mileage, monomer voltage and temperature collected by a plurality of temperature sensors;
the method for performing data cleaning on the characteristic data comprises the following steps: clearing invalid data, supplementing missing data and normalizing.
Preferably, the first and second electrodes are formed of a metal,
the formula for carrying out normalization processing on the characteristic data is as follows:
Preferably, the first and second electrodes are formed of a metal,
in the generation module, a total of P1 sample data with a fixed time length are generated in a sliding window mode, the number of the sample data of a normal vehicle is P2, the number of the sample data of a problem vehicle is P3, and the size of each sample data is an m × n matrix, wherein m is the time length, and n is a characteristic number.
Preferably, the first and second electrodes are formed of a metal,
in the building module, the recurrent neural network model consists of a plurality of basic recurrent neural units and a full connection layer, wherein the basic recurrent neural units comprise simple recurrent neural units, LSTM units and GRU units;
the recurrent neural network model can also adopt a self-defined recurrent neural unit.
Preferably, the first and second electrodes are formed of a metal,
in the definition module, the output of the defined recurrent neural network model, namely the prediction mode, is two, one, all data at the time with the length of m1 are predicted once based on the acquired data, and the other, the recurrent prediction method is adopted, namely, the output of each time is the prediction of the next time, and then the result of the next time is predicted by using the data with the length of m-P4 and the predicted data with the length of P4 in the sample data until all data at the time with the predicted length of m1 are obtained.
Preferably, the first and second electrodes are formed of a metal,
in the training module, the cyclic neural network model acquires sample data of P2 normal vehicles for training, and the evaluation index of the cyclic network model is the Mean Square Error (MSE); optimizing the training process of the recurrent neural network model by adopting a regularization and batch normalization mode, wherein the MSE is defined as a parameter MSEIs normal;
In the verification module, verifying the trained recurrent neural network model by using sample data of P3 problematic vehicles, wherein the evaluation index of the verification result is MSE, and the MSE is defined as a parameter MSEProblem(s);
In the circulation module, the training module is repeated to the verification module until the training module is obtainedMSE generationIs normalSmaller, MSEProblem(s)A larger optimal recurrent neural network model;
in the comparison module, the optimal cyclic neural network model obtains vehicle real-time data, and MSE is obtained through calculationReal timeThe calculation formula of the deviation degree R is as follows:
by adopting the technical scheme, the invention has the following beneficial effects: according to the method, a cyclic neural network model is constructed by a deep learning method by utilizing the characteristic that vehicle data belong to time series data, the data of a normal vehicle is used as training data, the data of a problem vehicle is used as verification data, and the characteristics of the data comprise multiple dimensions such as SOC (system on chip), current, monomer voltage, temperature and the like; in practical application, a trained model is deployed to a big data platform, calculation is carried out according to certain frequency by using collected real-time data, then a prediction result of the model is obtained, the prediction result is compared with an actual result, and the deviation degree is calculated, wherein the deviation degree can be used as the probability that the thermal runaway fault of the battery is possible to occur.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The invention will be further explained with reference to the drawings.
As shown in fig. 1, the invention provides a battery thermal runaway risk prediction method based on a recurrent neural network, which includes the following steps:
and S01, selecting characteristic data from the pre-collected historical data for training the recurrent neural network model, and performing data cleaning on the characteristic data.
Specifically, the historical data includes laboratory data and user real vehicle data. The data of vehicles with the same type of power batteries are randomly extracted, the vehicles can be layered in advance according to mileage, the selected vehicles are normal vehicles without faults, the number of the selected vehicles accounts for not less than 10% of the total number of the vehicles, and the duration of the selected data is at least one year.
The characteristic data comprises total current, total voltage, SOC, vehicle speed, accumulated mileage, monomer voltage, temperature collected by a plurality of temperature sensors and the like. In addition, full life cycle data is collected for the thermally runaway vehicle in the same manner.
The method for performing data cleaning on the characteristic data comprises the following steps: clearing invalid data, complementing missing data and normalizing, and adopting
The specific formula is as follows by using a max-min normalization mode:
S02: and generating sample data of a fixed time length based on the characteristic data after cleaning, wherein the sample data comprises the sample data of the normal vehicle and the sample data of the problem vehicle.
Specifically, a total of P1 sample data with a fixed time length are generated in a sliding window manner, the number of the sample data of a normal vehicle is P2, the number of the sample data of a problem vehicle is P3, and the size of each sample data is an m × n matrix, where m is the time length and n is a feature number. The interval of the sliding window should be as small as possible so that as much data as possible can be generated, but also the requirements of computing power and computing duration should be met.
S03: and constructing a recurrent neural network model.
Specifically, the recurrent neural network model is composed of a plurality of basic recurrent neural units and a full connection layer, the basic recurrent neural units comprise simple recurrent neural units, LSTM units and GRU units, and the basic recurrent neural units can be directly called in a common machine learning framework. The recurrent neural network model can also employ custom recurrent neural units.
S04: the output of the recurrent neural network model is defined.
Specifically, there are two ways of defining the output of the recurrent neural network model, namely, prediction, one of which is to predict all data at the time when the length is m1 at one time based on the acquired data, and the other is to adopt a recurrent prediction method, namely, the output at each time is the prediction at the next time, and then predict the result at the next time by using the data at the last m-P4(P4 is 1) length and the data at the predicted time when the length is P4(P4 is 1) (the same data with the m length), and so on until all data at the time when the predicted length is m1 are obtained. The two modes can give a prediction result of any time length, and the mode is determined based on the number of the neurons in the full connection layer.
S05: and the recurrent neural network model acquires sample data of a normal vehicle and trains the sample data.
Specifically, the cyclic neural network model acquires sample data of P1 normal vehicles and trains the samples, the evaluation index of the cyclic network model is Mean Square Error (MSE), the smaller the MSE is, the better the MSE is, in order to avoid overfitting, the training process of the cyclic neural network model can be optimized in modes of regularization, batch normalization and the like, and the MSE is defined as parameter MSEIs normal;
S06: and verifying the trained recurrent neural network model by using the sample data of the vehicle in question.
Specifically, the cyclic neural network model after training is verified by using sample data of P2 problem vehicles, the evaluation index of the verification result is MSE, and the MSE is defined as a parameter MSEProblem(s)MSE because it is data of the problem vehicleProblem(s)The larger should be the better.
S07: repeating S05-S06 until the optimal recurrent neural network model is obtained.
Specifically, S05 through S06 are repeated until MSE is obtainedIs normalSmaller, MSEProblem(s)A larger optimal recurrent neural network model, theThe optimal recurrent neural network model is a model of final online application.
S08: the optimal cyclic neural network model obtains vehicle real-time data, a prediction result is obtained through calculation according to set frequency, the prediction result is compared with an actual result to obtain a deviation degree, and when the deviation degree is larger than a threshold value, the battery can be judged to have a thermal runaway risk.
Specifically, the optimal recurrent neural network model obtains real-time sample data of the vehicle, and MSE is obtained through calculationReal timeThe calculation formula of the deviation degree, namely R is as follows:
the invention also provides a battery thermal runaway risk prediction system based on the recurrent neural network, which comprises the following steps:
the cleaning module is used for selecting characteristic data from pre-collected historical data used for training the recurrent neural network model and cleaning the characteristic data;
the generating module is used for generating sample data with fixed time length based on the cleaned characteristic data, wherein the sample data comprises sample data of a normal vehicle and sample data of a problem vehicle;
the building module is used for building a recurrent neural network model;
the definition module is used for defining the output of the recurrent neural network model;
the training module is used for acquiring sample data of a normal vehicle by the recurrent neural network model and training the sample data;
the verification module is used for verifying the trained recurrent neural network model by using sample data of the vehicle in question;
the circulating module is used for repeatedly training the module to the verification module until an optimal circulating neural network model is obtained;
and the comparison module is used for acquiring the real-time data of the vehicle by the optimal cyclic neural network model, calculating according to the set frequency to obtain a prediction result, comparing the prediction result with an actual result to obtain a deviation degree, and judging that the battery has a thermal runaway risk when the deviation degree is greater than a threshold value.
In particular, the amount of the solvent to be used,
in the cleaning module, the historical data comprises laboratory data and user real vehicle data;
the characteristic data comprises total current, total voltage, SOC, vehicle speed, accumulated mileage, monomer voltage and temperature collected by a plurality of temperature sensors;
the method for performing data cleaning on the characteristic data comprises the following steps: clearing invalid data, supplementing missing data and normalizing.
In particular, the amount of the solvent to be used,
the formula for normalizing the characteristic data is as follows:
In particular, the amount of the solvent to be used,
in the generation module, a total of P1 sample data with a fixed time length are generated in a sliding window mode, the number of the sample data of a normal vehicle is P2, the number of the sample data of a problem vehicle is P3, and the size of each sample data is an m × n matrix, wherein m is the time length and n is a characteristic number.
In particular, the amount of the solvent to be used,
in the building module, the recurrent neural network model consists of a plurality of basic recurrent neural units and a full connection layer, wherein the basic recurrent neural units comprise simple recurrent neural units, LSTM units and GRU units;
the recurrent neural network model can also employ custom recurrent neural units.
In particular, the amount of the solvent to be used,
in the definition module, the output of the recurrent neural network model is defined, namely, the prediction mode is two, one is that all data at the time with the length of m1 are predicted once based on the acquired data, and the other is that the recurrent prediction method is adopted, namely, the output at each time is the prediction of the next time, and then the result of the next time is predicted by using the data with the length of m-P4 and the predicted data at the time with the length of P4 in the sample data until all the data at the time with the predicted length of m1 are obtained.
In particular, the amount of the solvent to be used,
in a training module, a cyclic neural network model acquires sample data of P2 normal vehicles and trains the samples, and an evaluation index of the cyclic network model is an average square error (MSE); optimizing the training process of the recurrent neural network model by adopting a regularization and batch normalization mode, wherein MSE is defined as a parameter MSEIs normal;
In a verification module, verifying a trained recurrent neural network model by using sample data of P3 problem vehicles, wherein the evaluation index of a verification result is MSE, and the MSE is defined as a parameter MSEProblem(s);
In the circulation module, the training module is repeated to the verification module until MSE is obtainedIs normalSmaller, MSEProblem(s)A larger optimal recurrent neural network model;
in the comparison module, the optimal cyclic neural network model obtains vehicle real-time data, and MSE is obtained through calculationReal timeThe calculation formula of the deviation degree, namely R is as follows:
Claims (14)
1. a battery thermal runaway risk prediction method based on a recurrent neural network is characterized by comprising the following steps:
s01, selecting characteristic data from pre-collected historical data used for training the recurrent neural network model, and cleaning the characteristic data;
s02: generating sample data with a fixed time length based on the characteristic data after cleaning, wherein the sample data comprises sample data of a normal vehicle and sample data of a problem vehicle;
s03: constructing the recurrent neural network model;
s04: defining an output of the recurrent neural network model;
s05: the cyclic neural network model obtains sample data of the normal vehicle and trains the sample data;
s06: verifying the trained recurrent neural network model by using sample data of the problem vehicle;
s07: repeating the S05 to S06 until an optimal recurrent neural network model is obtained;
s08: the optimal recurrent neural network model obtains vehicle real-time data, a prediction result is obtained through calculation according to a set frequency, the prediction result is compared with an actual result to obtain a deviation degree, and when the deviation degree is larger than a threshold value, the battery can be judged to have a thermal runaway risk.
2. The method of claim 1, wherein the method comprises the steps of,
in the S01, the historical data includes laboratory data and user real vehicle data;
the characteristic data comprises total current, total voltage, SOC, vehicle speed, accumulated mileage, monomer voltage and temperature collected by a plurality of temperature sensors;
the method for performing data cleaning on the characteristic data comprises the following steps: clearing invalid data, supplementing missing data and normalizing.
4. The method of claim 1, wherein the method comprises the steps of,
in S02, a total of P1 sample data of a fixed time length are generated by a sliding window method, where the number of sample data of a normal vehicle is P2, the number of sample data of a problem vehicle is P3, and the size of each sample data is an m × n matrix, where m is a time length and n is a feature number.
5. The method of claim 1, wherein the method comprises the steps of,
in S03, the recurrent neural network model is composed of a plurality of basic recurrent neural units and a full connection layer, where the basic recurrent neural units include simple recurrent neural units, LSTM units, and GRU units;
the recurrent neural network model can also adopt a self-defined recurrent neural unit.
6. The method of claim 4, wherein the method comprises the steps of,
in S04, the output of the defined recurrent neural network model, i.e., the prediction mode, is two, one of which predicts all data at the time of m1 based on the acquired data at one time, and the other of which predicts the next time by using a recurrent prediction method, i.e., the output of each time is the prediction of the next time, and then predicts the result of the next time by using the data of the last m-P4 length and the predicted data at the time of P4 in the sample data until all data at the time of m1 are obtained.
7. The method of claim 4, wherein the method comprises the steps of,
in the step S05, the recurrent neural network model obtains sample data of P2 normal vehicles, and trains the samples, where an evaluation index of the recurrent neural network model is an average square error MSE; optimizing the training process of the recurrent neural network model by adopting a regularization and batch normalization mode, wherein the MSE is defined as a parameter MSEIs normal;
In the S06, verifying the trained recurrent neural network model by using the sample data of the P3 problematic vehicles, wherein the evaluation index of the verification result is MSE, and the MSE is defined as a parameter MSEProblem(s);
In the S07, repeating the S05 to S06 until MSE is foundIs normalSmaller, MSEProblem(s)A larger optimal recurrent neural network model;
in the step S08, the optimal recurrent neural network model obtains vehicle real-time data, and the MSE is obtained through calculationReal timeThe calculation formula of the deviation degree R is as follows:
8. a battery thermal runaway risk prediction system based on a Recurrent Neural Network (RNN), comprising:
the cleaning module is used for selecting characteristic data from pre-collected historical data used for training the recurrent neural network model and cleaning the characteristic data;
the generating module is used for generating sample data with a fixed time length based on the characteristic data after cleaning, wherein the sample data comprises sample data of a normal vehicle and sample data of a problem vehicle;
the building module is used for building the recurrent neural network model;
a definition module for defining an output of the recurrent neural network model;
the training module is used for acquiring the sample data of the normal vehicle by the recurrent neural network model and training the sample data;
the verification module is used for verifying the trained recurrent neural network model by using sample data of the problem vehicle;
the circulating module is used for repeating the training module to the verification module until the optimal circulating neural network model is obtained;
and the comparison module is used for acquiring vehicle real-time data by the optimal cyclic neural network model, calculating according to a set frequency to obtain a prediction result, comparing the prediction result with an actual result to obtain a deviation degree, and judging that the battery has a thermal runaway risk when the deviation degree is greater than a threshold value.
9. The recurrent neural network-based battery thermal runaway risk prediction system of claim 8,
in the cleaning module, the historical data comprises laboratory data and user real vehicle data;
the characteristic data comprises total current, total voltage, SOC, vehicle speed, accumulated mileage, monomer voltage and temperature collected by a plurality of temperature sensors;
the method for performing data cleaning on the characteristic data comprises the following steps: clearing invalid data, supplementing missing data and normalizing.
11. The recurrent neural network-based battery thermal runaway risk prediction system of claim 8,
in the generation module, a total of P1 sample data with a fixed time length are generated in a sliding window mode, the number of the sample data of a normal vehicle is P2, the number of the sample data of a problem vehicle is P3, and the size of each sample data is an m × n matrix, wherein m is the time length, and n is a characteristic number.
12. The recurrent neural network-based battery thermal runaway risk prediction system of claim 8,
in the building module, the recurrent neural network model consists of a plurality of basic recurrent neural units and a full connection layer, wherein the basic recurrent neural units comprise simple recurrent neural units, LSTM units and GRU units;
the recurrent neural network model can also adopt a self-defined recurrent neural unit.
13. The recurrent neural network-based battery thermal runaway risk prediction system of claim 11,
in the definition module, the output of the defined recurrent neural network model, namely the prediction mode, is two, one, all data at the time with the length of m1 are predicted once based on the acquired data, and the other, the recurrent prediction method is adopted, namely, the output of each time is the prediction of the next time, and then the result of the next time is predicted by using the data with the length of m-P4 and the predicted data with the length of P4 in the sample data until all data at the time with the predicted length of m1 are obtained.
14. The recurrent neural network-based battery thermal runaway risk prediction system of claim 11,
in the training module, the cyclic neural network model acquires sample data of P2 normal vehicles for training, and the evaluation index of the cyclic network model is the Mean Square Error (MSE); optimizing the training process of the recurrent neural network model by adopting a regularization and batch normalization mode, wherein the MSE is defined as a parameter MSEIs normal;
In the verification module, verifying the trained recurrent neural network model by using sample data of P3 problematic vehicles, wherein the evaluation index of the verification result is MSE, and the MSE is defined as a parameter MSEProblem(s);
In the circulation module, repeating the training module to the verification module until MSE is obtainedIs normalSmaller, MSEProblem(s)A larger optimal recurrent neural network model;
in the comparison module, the optimal cyclic neural network model obtains vehicle real-time data, and MSE is obtained through calculationReal timeThe calculation formula of the deviation degree R is as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111661180.4A CN114355199A (en) | 2021-12-30 | 2021-12-30 | Battery thermal runaway risk prediction method and system based on recurrent neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111661180.4A CN114355199A (en) | 2021-12-30 | 2021-12-30 | Battery thermal runaway risk prediction method and system based on recurrent neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114355199A true CN114355199A (en) | 2022-04-15 |
Family
ID=81106018
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111661180.4A Pending CN114355199A (en) | 2021-12-30 | 2021-12-30 | Battery thermal runaway risk prediction method and system based on recurrent neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114355199A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115782584A (en) * | 2022-11-22 | 2023-03-14 | 重庆长安新能源汽车科技有限公司 | New energy vehicle safety state determination method, system, equipment and medium |
CN116087782A (en) * | 2022-11-09 | 2023-05-09 | 苏州首帆电子科技有限公司 | Automobile battery fault early warning method, system, device and storage medium |
CN116430245A (en) * | 2023-06-14 | 2023-07-14 | 威海谱跃光电科技有限公司 | Battery thermal runaway prediction method based on gradient optimization multi-physical information neural network |
CN117872166A (en) * | 2024-03-11 | 2024-04-12 | 广东采日能源科技有限公司 | Method and device for detecting thermal runaway of energy storage battery and electronic equipment |
-
2021
- 2021-12-30 CN CN202111661180.4A patent/CN114355199A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116087782A (en) * | 2022-11-09 | 2023-05-09 | 苏州首帆电子科技有限公司 | Automobile battery fault early warning method, system, device and storage medium |
CN116087782B (en) * | 2022-11-09 | 2024-02-02 | 苏州首帆电子科技有限公司 | Automobile battery fault early warning method, system, device and storage medium |
CN115782584A (en) * | 2022-11-22 | 2023-03-14 | 重庆长安新能源汽车科技有限公司 | New energy vehicle safety state determination method, system, equipment and medium |
CN116430245A (en) * | 2023-06-14 | 2023-07-14 | 威海谱跃光电科技有限公司 | Battery thermal runaway prediction method based on gradient optimization multi-physical information neural network |
CN117872166A (en) * | 2024-03-11 | 2024-04-12 | 广东采日能源科技有限公司 | Method and device for detecting thermal runaway of energy storage battery and electronic equipment |
CN117872166B (en) * | 2024-03-11 | 2024-05-14 | 广东采日能源科技有限公司 | Method and device for detecting thermal runaway of energy storage battery and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114355199A (en) | Battery thermal runaway risk prediction method and system based on recurrent neural network | |
Vanem et al. | Data-driven state of health modelling—A review of state of the art and reflections on applications for maritime battery systems | |
CN112904219B (en) | Big data-based power battery health state prediction method | |
CN108197648A (en) | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models | |
CN109978229A (en) | The method that the full battery core multi-point temperature of a kind of pair of power battery pack and tie point temperature carry out thermal runaway prediction | |
CN111680848A (en) | Battery life prediction method based on prediction model fusion and storage medium | |
CN108556682B (en) | Driving range prediction method, device and equipment | |
CN111310387B (en) | Fuel cell life prediction method | |
CN113821875B (en) | Intelligent vehicle fault real-time prediction method and system based on end cloud cooperation | |
CN113378990B (en) | Flow data anomaly detection method based on deep learning | |
CN115859058B (en) | UPS fault prediction method and system based on width learning network | |
CN116882981B (en) | Intelligent battery management system based on data analysis | |
CN112766618B (en) | Abnormality prediction method and device | |
CN110516813A (en) | A method of batteries of electric automobile RDR prediction is carried out based on big data machine learning | |
CN116298936A (en) | Intelligent lithium ion battery health state prediction method in incomplete voltage range | |
CN115166534A (en) | Method and system for predicting remaining service life of lithium ion battery | |
CN117261599B (en) | Fault detection method and device of electric automobile, electronic equipment and electric automobile | |
CN114036647A (en) | Power battery safety risk assessment method based on real vehicle data | |
US20230305073A1 (en) | Method and apparatus for providing a predicted aging state of a device battery based on a predicted usage pattern | |
CN116736133A (en) | Early prediction method for capacity degradation track of lithium ion battery in full life cycle | |
Wang et al. | Ensemble machine learning based driving range estimation for real‐world electric city buses by considering battery degradation levels | |
CN113212244B (en) | New energy vehicle power battery life prediction method and system | |
CN114139783A (en) | Wind power short-term power prediction method and device based on nonlinear weighted combination | |
CN113989066A (en) | Galvanic pile energy consumption analysis based on big data | |
CN110837932A (en) | Thermal power prediction method of solar heat collection system based on DBN-GA model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |