CN116660757A - Battery voltage estimation method, device and storage medium - Google Patents
Battery voltage estimation method, device and storage medium Download PDFInfo
- Publication number
- CN116660757A CN116660757A CN202310799998.5A CN202310799998A CN116660757A CN 116660757 A CN116660757 A CN 116660757A CN 202310799998 A CN202310799998 A CN 202310799998A CN 116660757 A CN116660757 A CN 116660757A
- Authority
- CN
- China
- Prior art keywords
- data
- voltage
- battery
- vehicle
- characteristic parameters
- 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
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000012360 testing method Methods 0.000 claims abstract description 95
- 238000012549 training Methods 0.000 claims abstract description 78
- 230000002159 abnormal effect Effects 0.000 claims abstract description 53
- 238000004140 cleaning Methods 0.000 claims abstract description 12
- 238000006073 displacement reaction Methods 0.000 claims description 39
- 238000005070 sampling Methods 0.000 claims description 28
- 238000003062 neural network model Methods 0.000 claims description 25
- 239000000523 sample Substances 0.000 claims description 21
- 210000004027 cell Anatomy 0.000 claims description 20
- 125000004122 cyclic group Chemical group 0.000 claims description 16
- 239000000178 monomer Substances 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 8
- 210000002569 neuron Anatomy 0.000 claims description 6
- 230000000306 recurrent effect Effects 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 230000000873 masking effect Effects 0.000 claims description 4
- 238000001556 precipitation Methods 0.000 claims description 4
- 230000006399 behavior Effects 0.000 description 6
- 230000007613 environmental effect Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000005856 abnormality Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 2
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000013524 data verification Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 208000032953 Device battery issue Diseases 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/382—Arrangements for monitoring battery or accumulator variables, e.g. SoC
- G01R31/3835—Arrangements for monitoring battery or accumulator variables, e.g. SoC involving only voltage measurements
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/396—Acquisition or processing of data for testing or for monitoring individual cells or groups of cells within a battery
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/60—Other road transportation technologies with climate change mitigation effect
- Y02T10/70—Energy storage systems for electromobility, e.g. batteries
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Secondary Cells (AREA)
- Tests Of Electric Status Of Batteries (AREA)
Abstract
The invention discloses a battery voltage estimation method, a device and a storage medium, wherein the battery voltage estimation method comprises the following steps: constructing a training data set and a testing data set according to historical data of a plurality of vehicles in preset time, wherein the historical data comprise vehicle driving data, battery data and weather data; performing data cleaning on the training data set and the test data set and determining a plurality of characteristic parameters, wherein the characteristic parameters comprise training characteristic parameters; inputting the training characteristic parameters into a pre-constructed voltage prediction mixed model and outputting a voltage predicted value; and outputting early warning information when the voltage pre-estimated value is within a preset abnormal voltage range. The technical scheme provided by the invention can solve the technical problems that the battery voltage cannot be estimated accurately in advance according to various types of historical data of the vehicle and the abnormal output voltage alarm is realized in the prior art.
Description
Technical Field
The present invention relates to the field of battery data analysis technologies, and in particular, to a method and apparatus for estimating battery voltage, and a storage medium.
Background
The battery system is an indispensable component of the electric vehicle, and largely determines drivability, safety, durability, and cost effectiveness of the electric vehicle. In recent years, automobile power battery technology has been rapidly developed, and in order to satisfy a required voltage and capacity, a plurality of battery cells are generally connected in series/parallel. Due to aging or improper operation of the driver, various failures may occur on each of the same battery cells. Failure to be inspected, if not timely disposed of, can adversely affect battery safety and even lead to thermal runaway of the battery system. Studies have shown that voltage anomalies can lead to various battery failures, typical voltage anomalies include over-voltage, under-voltage, and poor battery voltage uniformity. Therefore, the method and the device can timely and accurately predict the abnormal voltage in the running process of the vehicle, and have important significance for preventing the occurrence of thermal runaway caused by voltage faults and ensuring the safe running of the electric vehicle.
Many voltage fault diagnosis methods are proposed in the prior art, and can be classified into a threshold-based method, a model-based method, and a data-driven method. Wherein the threshold-based method is used for diagnosing overvoltage and undervoltage battery faults, but cannot predict occurrence of battery faults in advance, and it is difficult to determine an appropriate threshold in practical application; the model-based method can describe the change of the abnormal battery state more accurately, but relies on conventional simulation of battery state estimation and can only be used for diagnosing specific fault types; based on the data driving method, although the fault diagnosis is carried out on the battery voltage by using an artificial intelligence technology, only the influence of the parameters of the battery on the battery state is considered, and the great influence of the driver behavior in actual driving and the change of the running environment of the electric automobile on the power battery is not considered, so that the voltage prediction of the model under the actual working condition of the electric automobile is not accurate enough.
In summary, the prior art has the technical problem that the battery voltage cannot be estimated accurately in advance according to various types of historical data of the vehicle and the abnormal voltage alarm is output.
Disclosure of Invention
The invention provides a battery voltage estimation method, a battery voltage estimation device and a storage medium, and aims to effectively solve the technical problem that in the prior art, battery voltage cannot be estimated accurately in advance according to various types of historical data of a vehicle and abnormal voltage alarm is output.
According to an aspect of the present invention, there is provided a battery voltage estimation method including:
constructing a training data set and a testing data set according to historical data of a plurality of vehicles in preset time, wherein the historical data comprise vehicle driving data, battery data and weather data;
performing data cleaning on the training data set and the test data set and determining a plurality of characteristic parameters, wherein the characteristic parameters comprise training characteristic parameters;
inputting the training characteristic parameters into a pre-constructed voltage prediction mixed model and outputting a voltage predicted value;
and outputting early warning information when the voltage pre-estimated value is within a preset abnormal voltage range.
Further, the plurality of vehicles at least includes a first normal vehicle, a second normal vehicle, a first abnormal vehicle, a second abnormal vehicle, and a first faulty vehicle, a second faulty vehicle, and constructing the training data set and the testing data set according to the historical data of the plurality of vehicles in a preset time includes:
constructing the training data set from historical data of the first normal vehicle, the first abnormal vehicle, and the first failed vehicle;
constructing the test data set from historical data of the second normal vehicle, the second abnormal vehicle, the second failed vehicle;
and normalizing the training data set and the test data set based on a maximum and minimum normalization method.
Further, the vehicle travel data includes one or more of a brake pedal displacement rate, an accelerator pedal displacement rate, a motor speed, a travel speed, and an accumulated travel distance, the battery data includes one or more of a battery pack total voltage, a plurality of battery cell voltages, a plurality of temperature probe values, a battery charge, and a battery charge, the weather data includes one or more of an air temperature value, an air humidity value, an air pressure value, and a precipitation amount, and the data cleaning of the training data set and the test data set includes:
Deleting the data with null values in the historical data;
deleting data of the brake pedal displacement rate which is not in a preset first displacement interval;
deleting the data of the accelerator pedal displacement rate which is not in a preset second displacement interval;
deleting the data of the driving speed which is not in a preset vehicle speed interval;
deleting the data of the voltage of the battery cell which is not in a preset voltage interval;
deleting the data of which the temperature probe value is not in a preset temperature interval;
and deleting the data of the battery electric quantity which is not in the preset electric quantity interval.
Further, the determining the plurality of characteristic parameters includes:
calculating one or more of a single voltage average value, a single voltage median, a single voltage range and a single voltage standard deviation of the single voltages of the plurality of batteries;
calculating one or more of a temperature mean, a temperature median, a temperature range, and a temperature standard deviation of the plurality of temperature probe values;
determining the plurality of characteristic parameters to be one or more of the brake pedal displacement rate, the accelerator pedal displacement rate, the motor rotating speed, the running speed, the accumulated stroke, the total battery pack voltage, the battery electric quantity, the air temperature value, the air humidity value, the air pressure value, the monomer voltage average value, the monomer voltage median, the monomer voltage range, the monomer voltage standard, the temperature average value, the temperature median, the temperature range and the temperature standard deviation.
Further, the battery voltage estimation method further includes:
sampling the vehicle travel data and the battery data according to a first sampling frequency prior to the constructing the training data set and the test data set; sampling the weather data according to a second sampling frequency, wherein the first sampling frequency is greater than the second sampling frequency;
after the plurality of characteristic parameters are determined, interpolation operation is performed on the weather data based on a Lagrangian difference method so that the number of the weather data is consistent with the number of the vehicle running data.
Further, the method further comprises:
the masking multi-headed attention layer of the decoder module in the transducer model is replaced with a recurrent neural network model to construct the voltage prediction hybrid model.
Further, the inputting the training characteristic parameters into the pre-constructed voltage prediction hybrid model and outputting the voltage predicted value comprises:
constructing the training characteristic parameters into input time series data based on a sliding window algorithm;
setting training super parameters of the voltage prediction hybrid model, wherein the training super parameters comprise an input step length of the input time series data, an output step length of the output time series data, a maximum iteration number, the multi-head number of multi-head attention models, the data quantity of each batch, a random reject data probability value, an initial learning rate, an adjustment coefficient and the neuron quantity of the cyclic neural network model;
Inputting the input time series data into the voltage prediction mixed model, and outputting output time series data corresponding to the voltage predicted value;
and calculating a loss function according to the actual total voltage of the battery pack in the historical data and the voltage estimated value.
Further, the characteristic parameters further include test characteristic parameters, and the method further includes:
acquiring a first test characteristic parameter corresponding to the second normal vehicle in the test data set;
inputting the first test characteristic parameters into the voltage prediction hybrid model and outputting a first voltage test value;
calculating precision parameters of the voltage prediction hybrid model according to the actual total voltage of the battery pack in the historical data and the first voltage test value, wherein the precision parameters comprise one or more of average absolute error, root mean square error, average relative error and goodness of fit;
acquiring second test characteristic parameters corresponding to the second abnormal vehicle and the second fault vehicle in the test data set;
and inputting the second test characteristic parameters into the voltage prediction hybrid model, outputting a second voltage test value, judging whether the second voltage test value is in the abnormal voltage range, and detecting whether the early warning information is output.
According to another aspect of the present invention, there is also provided a battery voltage estimating apparatus including:
the data acquisition module is used for constructing a training data set and a test data set according to historical data of a plurality of vehicles in preset time, wherein the historical data comprise vehicle driving data, battery data and weather data;
the data processing module is used for carrying out data cleaning on the training data set and the test data set and determining a plurality of characteristic parameters, wherein the characteristic parameters comprise training characteristic parameters;
the voltage estimation module is used for inputting the training characteristic parameters into a pre-constructed voltage prediction mixed model and outputting a voltage estimated value;
and the voltage early warning module is used for outputting early warning information when the voltage pre-estimated value is within a preset abnormal voltage range.
According to another aspect of the present invention, there is also provided a storage medium having stored therein a plurality of instructions adapted to be loaded by a processor to perform any of the battery voltage estimation methods as described above.
Through one or more of the above embodiments of the present invention, at least the following technical effects can be achieved:
In the technical scheme disclosed by the invention, not only is the influence of battery parameters on the battery pack voltage considered, but also the driving behavior data of a driver and the environmental data of a vehicle are combined, so that the model application scene is more similar to the actual application scene of the electric automobile, and the accuracy of voltage prediction is improved. In the aspect of a neural network model, a transducer model is applied to the voltage prediction of an automobile power battery, and a cyclic neural network (RNN) network is introduced for time series prediction to improve the transducer model so as to form a voltage prediction hybrid model, so that the model is suitable for the time series prediction, and the voltage prediction accuracy of the power battery is improved. According to the scheme, the voltage change trend of the automobile power battery pack at the future time can be predicted, the abnormal grade and type of battery voltage are divided, the fault risk of the automobile power battery is identified in advance, and therefore enough emergency treatment time is obtained.
In summary, the method combines driving behavior data and environmental data to construct a voltage prediction hybrid model of a cyclic neural network model (RNN) -converter model, improves the accuracy of future voltage prediction of the battery pack, detects and early warns battery faults in advance according to the predicted voltage, and improves the daily use safety of the electric automobile.
Drawings
The technical solution and other advantageous effects of the present invention will be made apparent by the following detailed description of the specific embodiments of the present invention with reference to the accompanying drawings.
FIG. 1 is a flowchart illustrating steps of a method for estimating a battery voltage according to an embodiment of the present invention;
FIG. 2 is a block diagram of a voltage prediction hybrid model according to an embodiment of the present invention;
fig. 3 is a schematic diagram of input and output of time series data according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a battery voltage estimation device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.
In the description of the present invention, it should be noted that, unless explicitly specified and defined otherwise, the term "and/or" herein is merely an association relationship describing associated objects, and means that three relationships may exist, for example, a and/or B may mean: a exists alone, A and B exist together, and B exists alone. The character "/" herein generally indicates that the associated object is an "or" relationship unless otherwise specified.
Fig. 1 is a flowchart illustrating steps of a battery voltage estimation method according to an embodiment of the present invention, where according to an aspect of the present invention, a battery voltage estimation method is provided, and the battery voltage estimation method includes:
step 101: constructing a training data set and a testing data set according to historical data of a plurality of vehicles in preset time, wherein the historical data comprise vehicle driving data, battery data and weather data;
step 102: performing data cleaning on the training data set and the test data set and determining a plurality of characteristic parameters, wherein the characteristic parameters comprise training characteristic parameters;
step 103: inputting the training characteristic parameters into a pre-constructed voltage prediction mixed model and outputting a voltage predicted value;
step 104: and outputting early warning information when the voltage pre-estimated value is within a preset abnormal voltage range.
The existing electric automobile battery cannot diagnose faults in advance, and the voltage prediction research is less combined with driving behaviors and driving environments, so that the voltage prediction precision is lower. In order to solve the problem, the invention transfers a transducer model into voltage prediction, and evaluates the battery voltage based on the correlation of the vehicle multiclass data and the battery by combining the actual historical driving data collected by natural driving and the environmental parameters. Meanwhile, in order to enable the model to be suitable for prediction of time series data, a network layer of a cyclic neural network (RNN) model is introduced into a transducer model to construct a voltage prediction hybrid model, accuracy of predicting future battery voltage is improved, and corresponding battery fault grades and types are classified according to voltage abnormality grades and types.
The above steps 101 to 104 are specifically described below.
In step 101, a training data set and a test data set are constructed according to historical data of a plurality of vehicles in a preset time, wherein the historical data comprise vehicle driving data, battery data and weather data;
illustratively, the training data set is data for training a neural network model and the test data set is data for testing a neural network model. Historical data of a plurality of vehicles is acquired, for example, historical data of 3 vehicles over at least the past 2 years, specifically, 3 vehicles are vehicles whose data is normal, vehicles whose data is abnormal but which have not failed, and vehicles whose failure has occurred once.
In order to improve the estimation accuracy, the scheme combines actual historical driving data acquired by natural driving, such as vehicle driving data including vehicle speed, accelerator pedal displacement, brake pedal displacement and the like, and explores the correlation between environmental weather data and a power battery.
In step 102, data cleansing is performed on the training data set and the test data set, and a plurality of characteristic parameters are determined, wherein the characteristic parameters comprise training characteristic parameters;
for example, in order to improve the estimation accuracy, it is necessary to perform data cleaning on the historical data, delete the abnormal data, and sort out the training feature parameters and the test feature parameters after the data cleaning. The characteristic parameters are data which are directly input into the neural network model after data preprocessing, and can be optimized and trained, wherein the two types of characteristic parameters are the same in type and comprise driving speed, accumulated journey, air temperature value, battery electric quantity and the like. In practical application, the characteristic parameters can be determined according to practical requirements, and the invention is not limited to the above.
In step 103, inputting the training characteristic parameters into a pre-constructed voltage prediction mixed model and outputting a voltage predicted value;
illustratively, the neural network model is built in advance, and the method does not use a single data processing model, but combines a transducer model and a recurrent neural network model. The transform model is a main neural network model, but the Multi-Head Attention mechanism (Multi-Head Attention) in the conventional transform model is not suitable for processing time series data, and the data in the application is based on time ordering, so that a cyclic neural network model (RNN) suitable for processing time series is selected to replace the Mask Multi-Head Attention mechanism (Mask Multi-Head Attention) in the transform model, so that the model obtains better accuracy in time series data prediction.
In step 104, early warning information is output when the voltage pre-estimated value is within a preset abnormal voltage range.
Illustratively, the present approach aims to estimate the total voltage of the battery pack in the future from historical data, for example, the total voltage of the battery pack in the future 1 hour from the data of the past 2.5 hours. And judging the vehicle fault in advance, and outputting early warning information when the voltage pre-estimated value is in a preset abnormal voltage range, wherein the abnormal voltage range can be divided into a plurality of sections, and correspondingly, the early warning information is divided into multiple levels for early warning.
Further, the plurality of vehicles at least includes a first normal vehicle, a second normal vehicle, a first abnormal vehicle, a second abnormal vehicle, and a first faulty vehicle, a second faulty vehicle, and constructing the training data set and the testing data set according to the historical data of the plurality of vehicles in a preset time includes:
constructing the training data set from historical data of the first normal vehicle, the first abnormal vehicle, and the first failed vehicle;
constructing the test data set from historical data of the second normal vehicle, the second abnormal vehicle, the second failed vehicle;
and normalizing the training data set and the test data set based on a maximum and minimum normalization method.
Illustratively, in order to improve the optimization effect of the model, at least 3 vehicles, namely, a data normal vehicle, a data abnormal but non-faulty vehicle, and a vehicle that has once failed, are included. At least 2 vehicles per vehicle.
For example, actual operation data of 6 electric vehicles with the same specification are collected in a cloud platform, the data collection time granularity is 30s, and each vehicle collects data of the last two years. According to the maintenance records of the cloud platform on the battery pack, the vehicles are divided into three types:
The first type of normal vehicles, for example, a first normal vehicle (vehicle No. 1) and a second normal vehicle (vehicle No. 2) are normal vehicles, namely, no abnormal early warning exists in the cloud platform in history;
the second type of abnormal vehicles, for example, a first abnormal vehicle (No. 3 vehicle) and a second abnormal vehicle (No. 4 vehicle) are potential fault vehicles, and the two vehicles have abnormal early warning of overlarge internal resistance in a cloud platform, but have no thermal runaway;
third, faulty vehicles, for example, a first faulty vehicle (No. 5 vehicle) and a second faulty vehicle (No. 6 vehicle) are vehicles that have failed, and No. 5 and No. 6 vehicles are vehicles that have thermally run away.
The data of one year in the vehicles 1, 3 and 5 can be used as training data of the model, and the data of the second year and the data of the vehicles 2, 4 and 6 can be used as test data of the model.
In order to make the data more resolution and normalization, the data is normalized. And carrying out normalization calculation on the formed training data and test data according to a maximum-minimum normalization formula, wherein the formula is as follows:
wherein x is im Representing the normalized sequence of the feature sequence,representing the original feature sequence, ++>Representing the minimum value of the characteristic sequence in the sample, < > >Represents the maximum value of the characteristic sequence in the sample, m represents the characteristic quantity, and +.>The file stored in a fixed format, such as a pkl file, is convenient for subsequent use in the inverse normalization process.
Further, the vehicle travel data includes one or more of a brake pedal displacement rate, an accelerator pedal displacement rate, a motor speed, a travel speed, and an accumulated travel distance, the battery data includes one or more of a battery pack total voltage, a plurality of battery cell voltages, a plurality of temperature probe values, a battery charge, and a battery charge, the weather data includes one or more of an air temperature value, an air humidity value, an air pressure value, and a precipitation amount, and the data cleaning of the training data set and the test data set includes:
deleting the data with null values in the historical data;
deleting data of the brake pedal displacement rate which is not in a preset first displacement interval;
deleting the data of the accelerator pedal displacement rate which is not in a preset second displacement interval;
deleting the data of the driving speed which is not in a preset vehicle speed interval;
deleting the data of the voltage of the battery cell which is not in a preset voltage interval;
deleting the data of which the temperature probe value is not in a preset temperature interval;
And deleting the data of the battery electric quantity which is not in the preset electric quantity interval.
Illustratively, the data collected in the historical data includes a variety of field information such as cell voltage, total battery voltage, temperature probe value, brake pedal displacement rate, accelerator pedal displacement rate, motor speed, vehicle speed, current, SOC, vehicle cumulative range, etc. The battery cell voltage and the temperature probe value are multiple at the same time, and because a plurality of battery cells are arranged in a battery pack, each battery cell corresponds to one battery cell voltage; the battery is internally provided with a plurality of temperature probes at a plurality of positions, and the temperatures of the batteries at different positions are different. Because meteorological factors can affect battery voltage, it is necessary to collect weather data to participate in training and testing of the model.
The data cleaning is an indispensable step in the data analysis process, and can identify erroneous, inaccurate or incomplete parts in the data so as to ensure the quality and reliability of the data.
Firstly, eliminating null values existing in data; and secondly, deleting the data of which each field does not accord with the normal range of the field. For example, the SOC is controlled within 0% -100%, the temperature probe value is controlled at-40-210 ℃, the cell voltage is less than 5000mv, the total voltage of the battery pack is less than 480v, the vehicle speed is 0-200km/h, and the acceleration/brake pedal displacement rate is 0% -100%.
Further, the determining the plurality of characteristic parameters includes:
calculating one or more of a single voltage average value, a single voltage median, a single voltage range and a single voltage standard deviation of the single voltages of the plurality of batteries;
calculating one or more of a temperature mean, a temperature median, a temperature range, and a temperature standard deviation of the plurality of temperature probe values;
determining the plurality of characteristic parameters to be one or more of the brake pedal displacement rate, the accelerator pedal displacement rate, the motor rotating speed, the running speed, the accumulated stroke, the total battery pack voltage, the battery electric quantity, the air temperature value, the air humidity value, the air pressure value, the monomer voltage average value, the monomer voltage median, the monomer voltage range, the monomer voltage standard, the temperature average value, the temperature median, the temperature range and the temperature standard deviation.
For example, since there are a plurality of battery cell voltage and temperature probe values at the same time, it is necessary to separately perform data processing on the two sets of data, and sort one set of data into a representative input value. Specifically, the average value, the median, the standard deviation and the range of the cell voltage and the probe temperature in each frame of data can be calculated to generate 8 characteristic parameters representing the cell voltage and the battery pack temperature.
Finally, a plurality of characteristic parameters are obtained by combining other data, wherein the 20 characteristic parameters are listed in the application, and can be determined according to actual requirements in specific application, and the application is not limited to the above.
Further, the battery voltage estimation method further includes:
sampling the vehicle travel data and the battery data according to a first sampling frequency prior to the constructing the training data set and the test data set; sampling the weather data according to a second sampling frequency, wherein the first sampling frequency is greater than the second sampling frequency;
after the plurality of characteristic parameters are determined, interpolation operation is performed on the weather data based on a Lagrangian difference method so that the number of the weather data is consistent with the number of the vehicle running data.
For example, the sampling interval may be 20s or 30s due to the rapid change in vehicle data. But the weather data generally varies less and so the sampling interval may be 0.5 hours or 1 hour. Thus, training of the neural network model requires maintaining consistency of the data, with each set of data having the same number of characteristic parameters, with a different number of samples of the data. According to the application, the Lagrange interpolation method is adopted to interpolate the weather parameters, so that the sampling time of the weather parameters and the sampling time of the weather parameters are kept consistent, and the weather parameters and the vehicle data are combined according to time calibration.
The lagrangian interpolation method is a polynomial interpolation method, and a set of polynomials which just take values at each observation point can be found according to a given function.
Further, the method further comprises:
the masking multi-headed attention layer of the decoder module in the transducer model is replaced with a recurrent neural network model to construct the voltage prediction hybrid model.
For example, in order to more reasonably process input data based on a time sequence, the scheme constructs the voltage prediction hybrid model based on a transducer model and a recurrent neural network model, and fig. 2 is a frame diagram of the voltage prediction hybrid model provided by the embodiment of the present invention, as shown in fig. 2, a conventional transducer model structure includes an input layer (input encoding), a position coding layer (Positional Encoding), an encoding module, and a Decoder module.
Wherein the Encoder has a total of 6 layers of network, each layer is composed of two sublayers: multi-Head Attention mechanisms (Multi-Head Attention) and Feed Forward layers, and each sub-layer adds an Add & Norm layer. The Decoder also has 6 layers, each layer containing three sublayers: the modulated Multi-Head Attention layer, multi-Head Attention, feed Forward layer, each sub-layer also adds an Add & Norm layer.
The conventional transform model structure includes an input layer (input encoding), a position encoding layer (Positional Encoding), an Encoder module, and a Decoder module.
Wherein the Encoder has a total of 6 layers of network, each layer is composed of two sublayers: multi-Head Attention mechanisms (Multi-Head Attention) and Feed Forward layers, and each sub-layer adds an Add & Norm layer. The Decoder also has 6 layers, each layer containing three sublayers: the modulated Multi-Head Attention layer, multi-Head Attention, feed Forward layer, each sub-layer also adds an Add & Norm layer. For Self-Attention, when matrix X is taken as input, X is transformed linearly to get the matrix.
The scheme is mainly modified for a Decoder in a traditional converter model, and because mapping based on sliding window data is different from a machine translation task, a modulated Multi-Head attribute cannot be used, meanwhile, considering that a cyclic neural network (RNN) network can capture characteristics of a time sequence, the output nature of the Encoder is also a sequence, the modulated Multi-Head attribute is modified into the cyclic neural network (RNN) layer in the Decoder layer, a new hybrid prediction model is constructed, the model structure is shown in figure 2, an Nx module is provided in a complete model structure, and only one of the Nx modules is listed in figure 2.
Because the cyclic neural network model (RNN) structure can well utilize the relation between sequences, and has achieved great success in processing long-time sequence data, embedding the cyclic neural network model (RNN) network layer into the transform model structure enables the model to obtain better accuracy in time sequence data prediction.
Further, the inputting the training characteristic parameters into the pre-constructed voltage prediction hybrid model and outputting the voltage predicted value comprises:
constructing the training characteristic parameters into input time series data based on a sliding window algorithm;
setting training super parameters of the voltage prediction hybrid model, wherein the training super parameters comprise an input step length of the input time series data, an output step length of the output time series data, a maximum iteration number, the multi-head number of multi-head attention models, the data quantity of each batch, a random reject data probability value, an initial learning rate, an adjustment coefficient and the neuron quantity of the cyclic neural network model;
inputting the input time series data into the voltage prediction mixed model, and outputting output time series data corresponding to the voltage predicted value;
And calculating a loss function according to the actual total voltage of the battery pack in the historical data and the voltage estimated value.
Illustratively, this step obtains the voltage estimate through a voltage prediction hybrid model.
In one aspect, the training feature parameters are structured as input time series data based on a sliding window algorithm. I.e. the original data is structured into time series data in a sliding window manner. The invention can provide a fixed sliding window, can also self-define the size of the sliding window according to the requirement to determine the time span of the historical data for predicting the future voltage, and can self-define the time step of the predicted data according to the requirement, namely the time span of model output.
Fig. 3 is a schematic diagram of input and output of time series data according to an embodiment of the present invention, assuming that the default sliding window size is IW and the prediction step length is OW, the input data set of the model constructed according to the original data is x= { X t-IW ，X t-IW+1 ，...，X t Model output dataset y= { Y } t+1 ，y t+2 ，...，y t+OW }. Wherein X is t-Iw ＝(x t-IW，1 ，x t-IW，2 ，...，x t-IW，m ) Where IW represents the time step of the historical data contained by each data sample and m represents the data signature format. Specifically, assuming that the input step iw=240 and the output step ow=120 are selected, if the employed time is 30s, the past 2h history is used to predict the future 1 hour battery pack voltage. Assuming that the start time is 9 points on a day, X t-IW Representing characteristic parameter set acquired at 9 points, X t-IW+1 Then the characteristic parameter set at 9 points for 30s is represented, X t-IW+2 Then the characteristic parameter set of 9-point 1-time sharing is represented, X t A feature parameter set representing 11-point integer.
X t-IW ＝(x t-IW，1 ，x t-IW，2 ，...，x t-IW，m ) Assuming 20 feature parameters, m is 20, e.g., x t-IW，1 Indicating the brake pedal displacement rate, x t-IW，2 Representing the accelerator pedal displacement rate, x t-IW，m Representing the standard deviation of temperature, etc., for a total of 20 characteristic parameters.
Model output dataset y= { Y t+1 ，y t+2 ，...，y t+OW -wherein y t+1 The battery voltage at time t+1 is shown. Assuming that the input step iw=240 and the output step ow=120 are selected, if the time is 30s, the past 2h history is used to predict the future 1 hour pack voltage. Suppose the start time is 9 points on a day, where y t+1 The battery voltage at time t+1 is indicated as the voltage estimated value at 11 points.
On the other hand, setting training super parameters of the voltage prediction hybrid model, wherein the training super parameters comprise an input step length of the input time series data, an output step length of the output time series data, a maximum iteration number, the multi-head number of multi-head attention models, each batch of data amount, random reject data probability values, an initial learning rate, adjustment coefficients and the neuron number of the cyclic neural network model.
The model training hyper-parameters are set, for example, the input step iw=300 of the input time series data is selected, and the input step ow=120 of the output time series data, that is, the battery pack voltage of 1 hour in the future is predicted using the history data of the past 2.5 hours.
In the voltage prediction hybrid model structure, for example, the multi-head number nhead=10 of the multi-head attention model is set, that is, 10-head attention layers are used in the Encoder module; setting a maximum iteration number epochs=300, and setting batch_size=64 for each batch of data; setting the random reject data probability value dropout to 0.2 can prevent overfitting; initial learning rate lr=0.001, and adjustment coefficient gamma=0.95; a cyclic neural network model (RNN) in the model sets the number of neurons to 128;
inputting the input time series data into the voltage prediction mixed model, and outputting output time series data corresponding to the voltage predicted value;
and calculating a loss function according to the actual total voltage of the battery pack in the historical data and the voltage estimated value.
Finally, the full connection layer activation function is set asThe function, adam, is set as the model optimizer, and the loss function is set as MSE, with the following formula:
Wherein n is the number of samples,and y is the actual total voltage of the battery pack in the historical data and is a voltage predicted value.
Further, the characteristic parameters further include test characteristic parameters, and the method further includes:
acquiring a first test characteristic parameter corresponding to the second normal vehicle in the test data set;
inputting the first test characteristic parameters into the voltage prediction hybrid model and outputting a first voltage test value;
calculating precision parameters of the voltage prediction hybrid model according to the actual total voltage of the battery pack in the historical data and the first voltage test value, wherein the precision parameters comprise one or more of average absolute error, root mean square error, average relative error and goodness of fit;
acquiring second test characteristic parameters corresponding to the second abnormal vehicle and the second fault vehicle in the test data set;
and inputting the second test characteristic parameters into the voltage prediction hybrid model, outputting a second voltage test value, judging whether the second voltage test value is in the abnormal voltage range, and detecting whether the early warning information is output.
The evaluation accuracy of the voltage prediction hybrid model is tested based on the test characteristic parameters, specifically, two verification modes are adopted, wherein one mode is based on the accuracy of a data verification model of a normal vehicle, and the other mode is based on whether the data verification model of an abnormal vehicle and a fault vehicle can detect faults in advance and send alarm information.
And feeding the constructed training data into a voltage prediction mixed model for training, and storing the trained model as a model file. And using the accuracy of the verification model prediction of the first test characteristic parameters corresponding to the second normal vehicle (No. 2 vehicle).
The evaluation criteria employs a plurality of precision parameters to evaluate the accuracy of the model, including Mean Absolute Error (MAE), root Mean Square Error (RMSE), mean relative error (MAPE), and goodness of fit (R 2 ) Four indexes.
Let y be i The true total voltage of the battery pack at the moment i,for the voltage test value at time i, +.>And m is the average value of the time sequence corresponding to the actual total voltage of the battery pack, and represents the length of the time sequence.
The calculation formulas of the four indexes are respectively as follows:
the Mean Absolute Error (MAE) is calculated according to the following formula:
root Mean Square Error (RMSE) is calculated according to the following formula:
average relative error (MAPE) was calculated according to the following formula:
the goodness of fit (R) is calculated according to the following formula 2 )：
Then, voltage abnormality detection is carried out, and according to historical operation and maintenance experience of the cloud platform, a voltage abnormality early warning level and an early warning threshold value are given in a table 1, wherein U is as follows pack And the voltage test value is represented, and the voltage abnormality is set to three-level early warning according to the threshold value.
TABLE 1 Voltage anomaly early warning threshold and early warning level (n in the Table indicates the number of battery cells)
Acquiring second test characteristic parameters corresponding to the second abnormal vehicle and the second fault vehicle in the test data set; and inputting the second test characteristic parameters into the voltage prediction hybrid model, outputting a second voltage test value, judging whether the second voltage test value is in the abnormal voltage range, and detecting whether the early warning information is output.
The trained model and the threshold value determined in table 1 are used for predicting the data of the second abnormal vehicle (No. 4 vehicle) and the second fault vehicle (No. 5 vehicle) in the test set through the normalized input model, the predicted data is reversely normalized to form voltage prediction data (a second voltage test value), the prediction data is calculated and evaluated according to the accuracy parameter calculation formula, and meanwhile, the early warning detection process is completed by comparing the threshold value in table 1, so that the early warning of the fault of the vehicle-mounted battery pack is realized.
Through one or more of the above embodiments of the present invention, at least the following technical effects can be achieved:
in the technical scheme disclosed by the invention, not only is the influence of battery parameters on the battery pack voltage considered, but also the driving behavior data of a driver and the environmental data of a vehicle are combined, so that the model application scene is more similar to the actual application scene of the electric automobile, and the accuracy of voltage prediction is improved. In the aspect of a neural network model, a transducer model is applied to the voltage prediction of an automobile power battery, and a cyclic neural network (RNN) network is introduced for time series prediction to improve the transducer model so as to form a voltage prediction hybrid model, so that the model is suitable for the time series prediction, and the voltage prediction accuracy of the power battery is improved. According to the scheme, the voltage change trend of the automobile power battery pack at the future time can be predicted, the abnormal grade and type of battery voltage are divided, the fault risk of the automobile power battery is identified in advance, and therefore enough emergency treatment time is obtained.
In summary, the method combines driving behavior data and environmental data to construct a voltage prediction hybrid model of a cyclic neural network model (RNN) -converter model, improves the accuracy of future voltage prediction of the battery pack, detects and early warns battery faults in advance according to the predicted voltage, and improves the daily use safety of the electric automobile.
Based on the same inventive concept as the battery voltage estimation method of the embodiment of the present invention, the embodiment of the present invention provides a battery voltage estimation device, please refer to fig. 4, which includes:
a data acquisition module 201, configured to construct a training data set and a test data set according to historical data of a plurality of vehicles within a preset time, wherein the historical data includes vehicle driving data, battery data and weather data;
a data processing module 202, configured to perform data cleaning on the training data set and the test data set and determine a plurality of feature parameters, where the feature parameters include training feature parameters;
the voltage estimation module 203 is configured to input the training feature parameter to a pre-constructed voltage prediction hybrid model and output a voltage estimated value;
the voltage early warning module 204 is configured to output early warning information when the voltage pre-estimated value is within a preset abnormal voltage range.
Further, the plurality of vehicles includes at least a first normal vehicle, a second normal vehicle, a first abnormal vehicle, a second abnormal vehicle, and a first faulty vehicle, a second faulty vehicle, and the data acquisition module 201 is further configured to:
constructing the training data set from historical data of the first normal vehicle, the first abnormal vehicle, and the first failed vehicle;
constructing the test data set from historical data of the second normal vehicle, the second abnormal vehicle, the second failed vehicle;
and normalizing the training data set and the test data set based on a maximum and minimum normalization method.
Further, the vehicle travel data includes one or more of a brake pedal displacement rate, an accelerator pedal displacement rate, a motor speed, a travel speed, and an accumulated travel distance, the battery data includes one or more of a battery pack total voltage, a plurality of battery cell voltages, a plurality of temperature probe values, a battery charge, and a battery charge, the weather data includes one or more of an air temperature value, an air humidity value, an air pressure value, and a precipitation amount, and the data processing module 202 is further configured to:
deleting the data with null values in the historical data;
Deleting data of the brake pedal displacement rate which is not in a preset first displacement interval;
deleting the data of the accelerator pedal displacement rate which is not in a preset second displacement interval;
deleting the data of the driving speed which is not in a preset vehicle speed interval;
deleting the data of the voltage of the battery cell which is not in a preset voltage interval;
deleting the data of which the temperature probe value is not in a preset temperature interval;
and deleting the data of the battery electric quantity which is not in the preset electric quantity interval.
Further, the data processing module 202 is further configured to:
calculating one or more of a single voltage average value, a single voltage median, a single voltage range and a single voltage standard deviation of the single voltages of the plurality of batteries;
calculating one or more of a temperature mean, a temperature median, a temperature range, and a temperature standard deviation of the plurality of temperature probe values;
determining the plurality of characteristic parameters to be one or more of the brake pedal displacement rate, the accelerator pedal displacement rate, the motor rotating speed, the running speed, the accumulated stroke, the total battery pack voltage, the battery electric quantity, the air temperature value, the air humidity value, the air pressure value, the monomer voltage average value, the monomer voltage median, the monomer voltage range, the monomer voltage standard, the temperature average value, the temperature median, the temperature range and the temperature standard deviation.
Further, the device is further configured to:
sampling the vehicle travel data and the battery data according to a first sampling frequency prior to the constructing the training data set and the test data set; sampling the weather data according to a second sampling frequency, wherein the first sampling frequency is greater than the second sampling frequency;
after the plurality of characteristic parameters are determined, interpolation operation is performed on the weather data based on a Lagrangian difference method so that the number of the weather data is consistent with the number of the vehicle running data.
Further, the voltage estimation module 203 is further configured to:
the masking multi-headed attention layer of the decoder module in the transducer model is replaced with a recurrent neural network model to construct the voltage prediction hybrid model.
Further, the voltage estimation module 203 is further configured to:
constructing the training characteristic parameters into input time series data based on a sliding window algorithm;
setting training super parameters of the voltage prediction hybrid model, wherein the training super parameters comprise an input step length of the input time series data, an output step length of the output time series data, a maximum iteration number, the multi-head number of multi-head attention models, the data quantity of each batch, a random reject data probability value, an initial learning rate, an adjustment coefficient and the neuron quantity of the cyclic neural network model;
Inputting the input time series data into the voltage prediction mixed model, and outputting output time series data corresponding to the voltage predicted value;
and calculating a loss function according to the actual total voltage of the battery pack in the historical data and the voltage estimated value.
Further, the characteristic parameters further include test characteristic parameters, and the voltage estimation module 203 is further configured to:
acquiring a first test characteristic parameter corresponding to the second normal vehicle in the test data set;
inputting the first test characteristic parameters into the voltage prediction hybrid model and outputting a first voltage test value;
calculating precision parameters of the voltage prediction hybrid model according to the actual total voltage of the battery pack in the historical data and the first voltage test value, wherein the precision parameters comprise one or more of average absolute error, root mean square error, average relative error and goodness of fit;
acquiring second test characteristic parameters corresponding to the second abnormal vehicle and the second fault vehicle in the test data set;
and inputting the second test characteristic parameters into the voltage prediction hybrid model, outputting a second voltage test value, judging whether the second voltage test value is in the abnormal voltage range, and detecting whether the early warning information is output.
Other aspects and implementation details of the battery voltage estimating apparatus are the same as or similar to those of the battery voltage estimating method described above, and are not described herein.
According to another aspect of the present invention, there is also provided a storage medium having stored therein a plurality of instructions adapted to be loaded by a processor to perform any of the battery voltage estimation methods as described above.
In summary, although the present invention has been described in terms of the preferred embodiments, the preferred embodiments are not limited to the above embodiments, and various modifications and changes can be made by one skilled in the art without departing from the spirit and scope of the invention, and the scope of the invention is defined by the appended claims.
Claims (10)
1. The battery voltage estimation method is characterized by comprising the following steps of:
constructing a training data set and a testing data set according to historical data of a plurality of vehicles in preset time, wherein the historical data comprise vehicle driving data, battery data and weather data;
performing data cleaning on the training data set and the test data set and determining a plurality of characteristic parameters, wherein the characteristic parameters comprise training characteristic parameters;
Inputting the training characteristic parameters into a pre-constructed voltage prediction mixed model and outputting a voltage predicted value;
and outputting early warning information when the voltage pre-estimated value is within a preset abnormal voltage range.
2. The battery voltage estimation method according to claim 1, wherein the plurality of vehicles includes at least a first normal vehicle, a second normal vehicle, a first abnormal vehicle, a second abnormal vehicle, and a first faulty vehicle, a second faulty vehicle, and the constructing the training data set and the test data set based on the history data of the plurality of vehicles within a preset time includes:
constructing the training data set from historical data of the first normal vehicle, the first abnormal vehicle, and the first failed vehicle;
constructing the test data set from historical data of the second normal vehicle, the second abnormal vehicle, the second failed vehicle;
and normalizing the training data set and the test data set based on a maximum and minimum normalization method.
3. The battery voltage estimation method of claim 1, wherein the vehicle travel data includes one or more of a brake pedal displacement rate, an accelerator pedal displacement rate, a motor speed, a travel speed, and an accumulated trip, the battery data includes one or more of a battery pack total voltage, a plurality of battery cell voltages, a plurality of temperature probe values, a battery charge, and a battery charge, the weather data includes one or more of an air temperature value, an air humidity value, an air pressure value, and a precipitation amount, and the data cleansing the training data set and the test data set includes:
Deleting the data with null values in the historical data;
deleting data of the brake pedal displacement rate which is not in a preset first displacement interval;
deleting the data of the accelerator pedal displacement rate which is not in a preset second displacement interval;
deleting the data of the driving speed which is not in a preset vehicle speed interval;
deleting the data of the voltage of the battery cell which is not in a preset voltage interval;
deleting the data of which the temperature probe value is not in a preset temperature interval;
and deleting the data of the battery electric quantity which is not in the preset electric quantity interval.
4. The battery voltage estimation method of claim 3, wherein the determining a plurality of characteristic parameters includes:
calculating one or more of a single voltage average value, a single voltage median, a single voltage range and a single voltage standard deviation of the single voltages of the plurality of batteries;
calculating one or more of a temperature mean, a temperature median, a temperature range, and a temperature standard deviation of the plurality of temperature probe values;
determining the plurality of characteristic parameters to be one or more of the brake pedal displacement rate, the accelerator pedal displacement rate, the motor rotating speed, the running speed, the accumulated stroke, the total battery pack voltage, the battery electric quantity, the air temperature value, the air humidity value, the air pressure value, the monomer voltage average value, the monomer voltage median, the monomer voltage range, the monomer voltage standard, the temperature average value, the temperature median, the temperature range and the temperature standard deviation.
5. The battery voltage estimation method according to claim 1, further comprising:
sampling the vehicle travel data and the battery data according to a first sampling frequency prior to the constructing the training data set and the test data set; sampling the weather data according to a second sampling frequency, wherein the first sampling frequency is greater than the second sampling frequency;
after the plurality of characteristic parameters are determined, interpolation operation is performed on the weather data based on a Lagrangian difference method so that the number of the weather data is consistent with the number of the vehicle running data.
6. The battery voltage estimation method according to claim 1, further comprising:
the masking multi-headed attention layer of the decoder module in the transducer model is replaced with a recurrent neural network model to construct the voltage prediction hybrid model.
7. The battery voltage estimation method according to claim 6, wherein inputting the training characteristic parameter into a pre-constructed voltage prediction hybrid model and outputting a voltage estimated value includes:
constructing the training characteristic parameters into input time series data based on a sliding window algorithm;
Setting training super parameters of the voltage prediction hybrid model, wherein the training super parameters comprise an input step length of the input time series data, an output step length of the output time series data, a maximum iteration number, the multi-head number of multi-head attention models, the data quantity of each batch, a random reject data probability value, an initial learning rate, an adjustment coefficient and the neuron quantity of the cyclic neural network model;
inputting the input time series data into the voltage prediction mixed model, and outputting output time series data corresponding to the voltage predicted value;
and calculating a loss function according to the actual total voltage of the battery pack in the historical data and the voltage estimated value.
8. The battery voltage estimation method of claim 2, wherein the characteristic parameters further comprise test characteristic parameters, the method further comprising:
acquiring a first test characteristic parameter corresponding to the second normal vehicle in the test data set;
inputting the first test characteristic parameters into the voltage prediction hybrid model and outputting a first voltage test value;
calculating precision parameters of the voltage prediction hybrid model according to the actual total voltage of the battery pack in the historical data and the first voltage test value, wherein the precision parameters comprise one or more of average absolute error, root mean square error, average relative error and goodness of fit;
Acquiring second test characteristic parameters corresponding to the second abnormal vehicle and the second fault vehicle in the test data set;
and inputting the second test characteristic parameters into the voltage prediction hybrid model, outputting a second voltage test value, judging whether the second voltage test value is in the abnormal voltage range, and detecting whether the early warning information is output.
9. A battery voltage estimation apparatus, the apparatus comprising:
the data acquisition module is used for constructing a training data set and a test data set according to historical data of a plurality of vehicles in preset time, wherein the historical data comprise vehicle driving data, battery data and weather data;
the data processing module is used for carrying out data cleaning on the training data set and the test data set and determining a plurality of characteristic parameters, wherein the characteristic parameters comprise training characteristic parameters;
the voltage estimation module is used for inputting the training characteristic parameters into a pre-constructed voltage prediction mixed model and outputting a voltage estimated value;
and the voltage early warning module is used for outputting early warning information when the voltage pre-estimated value is within a preset abnormal voltage range.
10. A storage medium having stored therein a plurality of instructions adapted to be loaded by a processor to perform the battery voltage estimation method of any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310799998.5A CN116660757A (en) | 2023-06-30 | 2023-06-30 | Battery voltage estimation method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310799998.5A CN116660757A (en) | 2023-06-30 | 2023-06-30 | Battery voltage estimation method, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116660757A true CN116660757A (en) | 2023-08-29 |
Family
ID=87711914
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310799998.5A Pending CN116660757A (en) | 2023-06-30 | 2023-06-30 | Battery voltage estimation method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116660757A (en) |
-
2023
- 2023-06-30 CN CN202310799998.5A patent/CN116660757A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113343633B (en) | Thermal runaway fault classification and risk prediction method and system for power lithium battery | |
| CN111999657A (en) | Method for evaluating driving mileage of lithium ion battery of electric vehicle in residual life | |
| US11965935B2 (en) | Method and apparatus for operating a system for providing predicted states of health of electrical energy stores for a device using machine learning methods | |
| CN114559819A (en) | Electric vehicle battery safety early warning method based on signal processing | |
| CN116541790B (en) | New energy vehicle health assessment method and device based on multi-feature fusion | |
| CN114239377A (en) | Method and system for evaluating health state of urban rail electromechanical equipment and storage medium | |
| CN114791993B (en) | Power battery pack SOH prediction method and system | |
| Zhao et al. | Multi-step ahead voltage prediction and voltage fault diagnosis based on gated recurrent unit neural network and incremental training | |
| He et al. | Voltage abnormality-based fault diagnosis for batteries in electric buses with a self-adapting update model | |
| CN116184210A (en) | Battery abnormality detection method, device, system and electronic device | |
| CN114994543A (en) | Energy storage power station battery fault diagnosis method and device and storage medium | |
| Wei et al. | Multiscale dynamic construction for abnormality detection and localization of Li-ion batteries | |
| Wang et al. | Prognostics and health management system for electric vehicles with a hierarchy fusion framework: concepts, architectures, and methods | |
| DE102022120962A1 (en) | BATTERY ERROR DETECTION | |
| Yao et al. | Fault identification of lithium-ion battery pack for electric vehicle based on ga optimized ELM neural network | |
| CN109948194B (en) | High-voltage circuit breaker mechanical defect integrated learning diagnosis method | |
| JP7115346B2 (en) | Abnormality detection device | |
| CN114036647A (en) | Power battery safety risk assessment method based on real vehicle data | |
| CN116901707A (en) | Power battery pack fault early warning method, system and vehicle | |
| CN117665620A (en) | New energy automobile data-based battery health evaluation method | |
| CN117630672A (en) | Method and device for predicting a device battery of a diagnostic device using a converter model | |
| CN116660757A (en) | Battery voltage estimation method, device and storage medium | |
| CN116381493A (en) | Battery pack fault detection method, system, electronic equipment and storage medium | |
| CN115656831A (en) | Multi-step advanced prediction and fault diagnosis method for single battery voltage | |
| CN114994546A (en) | Battery pack SOH prediction system and method based on block chain and Informer neural network |
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 |