CN117131687A - Electric automobile battery management method based on data analysis - Google Patents

Electric automobile battery management method based on data analysis Download PDF

Info

Publication number
CN117131687A
CN117131687A CN202311106776.7A CN202311106776A CN117131687A CN 117131687 A CN117131687 A CN 117131687A CN 202311106776 A CN202311106776 A CN 202311106776A CN 117131687 A CN117131687 A CN 117131687A
Authority
CN
China
Prior art keywords
battery
training
model
data
whale
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
Application number
CN202311106776.7A
Other languages
Chinese (zh)
Inventor
李卓
何荣
王伟民
冉洁
张根海
柏松
樊冬
许新桢
冯亮
代晓虹
王井
黄宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing Qingyi Microsystem Technology Co ltd
Chongqing Qingyi Microsystem Research Institute
Original Assignee
Chongqing Qingyi Microsystem Technology Co ltd
Chongqing Qingyi Microsystem Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chongqing Qingyi Microsystem Technology Co ltd, Chongqing Qingyi Microsystem Research Institute filed Critical Chongqing Qingyi Microsystem Technology Co ltd
Priority to CN202311106776.7A priority Critical patent/CN117131687A/en
Publication of CN117131687A publication Critical patent/CN117131687A/en
Pending legal-status Critical Current

Links

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
    • B60L58/00—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles
    • B60L58/10—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries
    • B60L58/12—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries responding to state of charge [SoC]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation
    • 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

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Mathematical Optimization (AREA)
  • Software Systems (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • Health & Medical Sciences (AREA)
  • Mechanical Engineering (AREA)
  • Transportation (AREA)
  • Power Engineering (AREA)
  • Sustainable Energy (AREA)
  • Sustainable Development (AREA)
  • Automation & Control Theory (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Secondary Cells (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

The invention relates to an electric automobile battery management method based on data analysis, which comprises the following steps: acquiring charging voltage data of a battery pack; extracting charging characteristic information of each single battery from charging voltage data of the battery pack through a continuous variable mode decomposition algorithm; inputting the charging characteristic information of the single battery into a risk prediction model constructed based on a deep learning model to obtain a corresponding predicted thermal runaway risk value; taking a single battery with a predicted thermal runaway risk value exceeding a risk threshold as a battery to be monitored; and acquiring data of the charge and discharge process of the battery to be monitored and feeding the data back to a user of the electric automobile so that the user knows the charge and discharge state of the battery to be monitored. According to the invention, the characteristic signals are effectively extracted from the battery charging voltage data through the continuous variable mode decomposition algorithm, the thermal runaway risk of the battery is comprehensively, accurately and efficiently predicted through the deep learning model, and meanwhile, the prediction of the thermal runaway risk of the battery can be assisted by a user.

Description

Electric automobile battery management method based on data analysis
Technical Field
The invention relates to the technical field of electric vehicle battery management, in particular to an electric vehicle battery management method based on data analysis.
Background
In recent years, the technology of electric vehicles is rapidly developed, and a battery system is used as the only energy source of the pure electric vehicles and provides energy required by running for the pure electric vehicles. Although the battery manufacturing and packaging technology is greatly developed at present, the power battery is difficult to avoid faults under the multi-factor coupling effect due to the fact that the vehicle operation working conditions are complex and changeable, and the conditions of heat abuse and electric abuse are frequent. If the vehicle cannot be processed in time, the occurrence of thermal runaway accidents can be finally caused, and the life and property safety of passengers in the vehicle can be seriously threatened.
At present, for the battery system of the electric automobile, certain faults have strong concealment early-stage performance almost no different from that of normal batteries, and after obvious characteristics of the faults appear, a large amount of heat energy can be generated in a short time to rapidly induce thermal runaway. For this reason, chinese patent publication No. CN114355199a discloses a method and system for predicting thermal runaway risk of battery based on cyclic neural network, which comprises: selecting characteristic data from the pre-collected historical data for training the cyclic neural network model and cleaning the characteristic data; generating sample data of a fixed time length based on the feature data after the cleaning; constructing a cyclic neural network model; defining the output of a cyclic neural network model; the cyclic neural network model acquires sample data of a normal vehicle and trains the sample data; the model of the recurrent neural network after training is validated using the sample data of the problem vehicle.
In the above-described conventional scheme, prediction of the risk of thermal runaway of the battery is achieved by constructing a deep learning (recurrent neural network) model. The thermal runaway risk prediction mode based on the deep learning model needs to construct a large amount of training data to realize model training, such as the fact that the charging voltage data of the battery in the charge-discharge cycle can effectively reflect the thermal runaway risk. However, there are many noise signals in the battery charging voltage data collected in the prior art, and the data is also easily disturbed and attenuated by the environment in the transmission process, which makes it difficult to accurately and effectively predict the thermal runaway risk of the battery in a manner of directly inputting the battery charging voltage data as a model.
Therefore, how to improve the accuracy and effectiveness of battery thermal runaway risk prediction is a technical problem that needs to be solved.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to solve the technical problems that: how to provide an electric automobile battery management method based on data analysis, through SVMD (continuous variable mode decomposition) algorithm effectively extract characteristic signals from battery charging voltage data to through the comprehensive, accurate and efficient prediction battery thermal runaway risk of deep learning model, can assist the user to make the prejudgement to the thermal runaway risk of battery simultaneously, thereby improve electric automobile battery management's effect.
In order to solve the technical problems, the invention adopts the following technical scheme:
an electric vehicle battery management method based on data analysis comprises the following steps:
s1: acquiring charging voltage data of a battery pack in the process of charging the electric automobile by a user;
s2: extracting charging characteristic information of each single battery from charging voltage data of the battery pack through a continuous variable mode decomposition algorithm;
s3: inputting the charging characteristic information of the single battery into a risk prediction model constructed based on a deep learning model to obtain a corresponding predicted thermal runaway risk value;
s4: taking a single battery with a predicted thermal runaway risk value exceeding a risk threshold as a battery to be monitored;
s5: and acquiring data of the charge and discharge process of the battery to be monitored and feeding the data back to a user of the electric automobile so that the user knows the charge and discharge state of the battery to be monitored.
Preferably, the discharge data of the unit cells are extracted by:
s201: dividing charging voltage data of each single battery from charging voltage data of the battery pack;
s202: decomposing the charging voltage data of the single battery by a continuous variable mode decomposition algorithm to obtain a plurality of IMF components;
s203: calculating kurtosis values of the IMF components and selecting the IMF component with the largest kurtosis value as a target IMF;
s204: and carrying out envelope demodulation on the target IMF, and further extracting and obtaining corresponding charging characteristic information.
Preferably, the optimization parameters of the continuous variable modal decomposition algorithm are adaptively searched through a whale optimization algorithm;
the method specifically comprises the following steps:
s2021: determining an fitness function of a whale optimization algorithm and initializing whale populations, wherein each whale represents an optimal balance parameter of a continuous variable modal decomposition algorithm;
s2022: calculating the fitness value of each whale through a fitness function, and selecting the optimal whale;
s2023: performing surrounding prey, bubble net attack and searching for prey based on the optimal whale location, updating whale location;
s2024: updating the fitness value based on the position of the whale, and recording the optimal balance parameter corresponding to the current optimal whale;
s2025: judging whether a termination condition is satisfied: if yes, outputting the optimal balance parameters; otherwise, the process returns to step S2023.
Preferably, the formula surrounding the prey is described as follows:
wherein: t represents the current iteration;and->Representing the coefficient vector; />A position vector representing the best solution currently obtained;a position vector representing a current solution; the absolute value is represented by; />Representing the distance between whale and prey;
the vector is calculated by the following formulaAnd->
Wherein:the value of (2) decreases linearly from 2 to 0; />Is [0,1]]Is a random vector in (a).
Preferably, the formula for a bubble attack is described as follows:
wherein:representing the distance from the ith whale to the prey; b represents a constant; l represents [ -1,1]Random numbers in (a);
the circle is contracted while the whale is swimming along the spiral path, assuming a 0.5 probability of choosing between a contracted envelope mechanism or spiral model to optimize the position of the updating whale;
the formula is described as follows:
wherein: p represents a random number in [0,1 ].
Preferably, the formula for searching for a prey is described as follows:
wherein:representing a random position vector selected from the current population.
Preferably, the kurtosis value of the IMF component is calculated by the following formula:
wherein: k represents the kurtosis value of the IMF component; x is x i Representing an ith value in the IMF component;representing IMF scoreThe average of the values in the quantity, n, represents the number of samples in the IMF component.
Preferably, the risk prediction model needs to be trained twice;
training for the first time: training a risk prediction model through charging characteristic information of a battery under a certain working condition to obtain a pre-training model;
training for the second time: freezing a core neural network layer of the pre-training model and calling model parameters thereof to obtain an optimized model; and then training an optimization model through the charging characteristic information of the battery under other different working conditions to reconstruct a full-connection output layer of the battery, so as to obtain a trained risk prediction model.
Preferably, the first training is performed by:
s301: performing battery charge-discharge cycle experiments under different working conditions to obtain corresponding charge voltage data; then extracting charging characteristic information from the charging voltage data to construct a charging characteristic data set under various working conditions;
s302: acquiring charging characteristic information from a charging characteristic data set under a certain working condition as primary training data;
s303: training a risk prediction model through primary training data;
s304: calculating corresponding training loss based on the predicted thermal runaway risk value and the real thermal runaway risk value output by the risk prediction model, and optimizing model parameters through the training loss;
s305: steps S302 to S304 are repeated until the risk prediction model converges.
Preferably, the second training is performed by:
s311: freezing nuclear network layer parameters of the pre-training model through a transfer learning fine tuning technology, calling model parameters of the pre-training model, connecting output of the pre-training model to a full-connection layer to form a full-connection output layer, and performing linear change on hidden layer neuron activation to obtain an optimized model;
s312: acquiring charging characteristic information from the charging characteristic data sets under different working conditions as secondary training data;
s313: training an optimization model through the secondary training data to adjust parameters of a full-connection output layer of the optimization model;
s314: calculating corresponding training loss based on the predicted thermal runaway risk value and the real thermal runaway risk value output by the optimization model, and optimizing model parameters through the training loss;
s315: steps S312 to S314 are repeated until the risk prediction model converges.
Compared with the prior art, the method and the system for managing the battery of the electric automobile based on data analysis have the following beneficial effects:
the method aims at the problems of noise signals, environmental interference and attenuation in battery charging voltage data. According to the invention, the charging characteristic information of each single battery is extracted from the charging voltage data of the battery pack through an SVMD (continuous variable mode decomposition) algorithm, wherein the optimization of the SVMD algorithm can be approximately regarded as K one-dimensional optimization problems, compared with the existing VMD, the SVMD algorithm has lower calculation complexity, the trouble of determining the number of IMFs related to the VMD can be well avoided, and meanwhile, the problem that the initial center frequency in the VME is difficult to determine is avoided, namely, the SVMD algorithm has better robustness to noise and interference compared with the prior art, and the charging characteristic information of the battery can be effectively extracted, so that the accuracy and the effectiveness of the subsequent thermal runaway risk prediction can be improved in an auxiliary manner.
According to the invention, on the basis of effectively extracting the battery charging characteristic information, the thermal runaway risk of the battery is automatically predicted through the deep learning model. Wherein the deep learning model can provide more accurate thermal runaway risk prediction by learning complex nonlinear relationships and patterns; meanwhile, the most relevant characteristic representation can be automatically learned, and the manual selection and extraction of the characteristics are not needed, so that the model is more suitable for the characteristics and the change conditions of different batteries, and the dependence on the knowledge in the related field is reduced; in addition, the deep learning model has strong robustness and generalization capability, can be used for predicting unseen thermal runaway risks, can be used for predicting under different batteries and different environmental conditions and has good adaptability, namely, the thermal runaway risks can be comprehensively, accurately and efficiently predicted through the deep learning model, and the effectiveness and the robustness of the battery management of the electric automobile are further improved, so that the effect of the battery management of the electric automobile is improved.
According to the invention, quantitative assessment of the thermal runaway risk of the battery is realized through the deep learning model, and the data of the charging and discharging process of the battery to be monitored is fed back to the user of the electric automobile, so that the user of the electric automobile is enabled to know the charging and discharging state of the battery to be monitored, and further, the user of the electric automobile can be assisted to intuitively and effectively know the thermal runaway risk of the battery, so that the user can predict the thermal runaway risk prediction of the battery, and the management effect of the battery of the electric automobile is further improved.
Drawings
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings, in which:
fig. 1 is a logic block diagram of an electric vehicle battery management method based on data analysis.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of 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 some, but not all, embodiments of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected 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 be within the scope of the invention.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. In the description of the present invention, it should be noted that, directions or positional relationships indicated by terms such as "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., are directions or positional relationships based on those shown in the drawings, or are directions or positional relationships conventionally put in use of the inventive product, are merely for convenience of describing the present invention and simplifying the description, and are not indicative or implying that the apparatus or element to be referred to must have a specific direction, be constructed and operated in a specific direction, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," "third," and the like are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance. Furthermore, the terms "horizontal," "vertical," and the like do not denote a requirement that the component be absolutely horizontal or overhang, but rather may be slightly inclined. For example, "horizontal" merely means that its direction is more horizontal than "vertical" and does not mean that the structure must be perfectly horizontal, but may be slightly tilted. In the description of the present invention, it should also be noted that, unless explicitly specified and limited otherwise, the terms "disposed," "mounted," "connected," and "connected" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
The following is a further detailed description of the embodiments:
examples:
the embodiment discloses an electric vehicle battery management method based on data analysis.
As shown in fig. 1, an electric vehicle battery management method based on data analysis includes:
s1: acquiring charging voltage data of a battery pack in the process of charging the electric automobile by a user;
s2: extracting charging characteristic information of each single battery from charging voltage data of the battery pack through a continuous variable mode decomposition algorithm;
s3: inputting the charging characteristic information of the single battery into a risk prediction model constructed based on a deep learning model to obtain a corresponding predicted thermal runaway risk value;
s4: taking a single battery with a predicted thermal runaway risk value exceeding a risk threshold as a battery to be monitored;
s5: and acquiring data of the charge and discharge process of the battery to be monitored and feeding the data back to a user of the electric automobile so that the user knows the charge and discharge state of the battery to be monitored.
The method aims at the problems of noise signals, environmental interference and attenuation in battery charging voltage data. According to the invention, the charging characteristic information of each single battery is extracted from the charging voltage data of the battery pack through an SVMD (continuous variable mode decomposition) algorithm, wherein the optimization of the SVMD algorithm can be approximately regarded as K one-dimensional optimization problems, compared with the existing VMD, the SVMD algorithm has lower calculation complexity, the trouble of determining the number of IMFs related to the VMD can be well avoided, and meanwhile, the problem that the initial center frequency in the VME is difficult to determine is avoided, namely, the SVMD algorithm has better robustness to noise and interference compared with the prior art, and the charging characteristic information of the battery can be effectively extracted, so that the accuracy and the effectiveness of the subsequent thermal runaway risk prediction can be improved in an auxiliary manner.
According to the invention, on the basis of effectively extracting the battery charging characteristic information, the thermal runaway risk of the battery is automatically predicted through the deep learning model. Wherein the deep learning model can provide more accurate thermal runaway risk prediction by learning complex nonlinear relationships and patterns; meanwhile, the most relevant characteristic representation can be automatically learned, and the manual selection and extraction of the characteristics are not needed, so that the model is more suitable for the characteristics and the change conditions of different batteries, and the dependence on the knowledge in the related field is reduced; in addition, the deep learning model has strong robustness and generalization capability, can be used for predicting unseen thermal runaway risks, can be used for predicting under different batteries and different environmental conditions and has good adaptability, namely, the thermal runaway risks can be comprehensively, accurately and efficiently predicted through the deep learning model, and the effectiveness and the robustness of the battery management of the electric automobile are further improved, so that the effect of the battery management of the electric automobile is improved.
According to the invention, quantitative assessment of the thermal runaway risk of the battery is realized through the deep learning model, and the data of the charging and discharging process of the battery to be monitored is fed back to the user of the electric automobile, so that the user of the electric automobile is enabled to know the charging and discharging state of the battery to be monitored, and further, the user of the electric automobile can be assisted to intuitively and effectively know the thermal runaway risk of the battery, so that the user can predict the thermal runaway risk prediction of the battery, and the management effect of the battery of the electric automobile is further improved.
In the specific implementation process, the discharge data of the single battery is extracted through the following steps:
s201: dividing charging voltage data of each single battery from charging voltage data of the battery pack;
in this embodiment, the charging voltage data of each unit cell is divided from the charging voltage data of the battery pack, and may be implemented by referring to the following steps:
1) Knowing the battery structure: the number and the connection mode of the battery cells of the battery pack are known, and the battery pack can be formed by connecting a plurality of battery cells in series or in parallel.
2) Measuring the battery pack charge voltage: and a proper testing instrument is used for measuring the charging voltage of the battery pack, so that the measured voltage value is accurate and reliable.
3) Calculating the voltage of the single battery: and calculating the voltage of each single battery according to the structure and the connection mode of the battery pack. If connected in series, the voltage of each cell should be equal. If connected in parallel, the voltage of each cell should be the average of the stack voltages.
3) Dividing the charging voltage data of the single battery: the division of the charging voltage data of the battery pack into the charging voltage data of each cell according to the calculated cell voltages may be implemented using a battery management system or self-programming.
When dividing the charging voltage data of the battery pack, the battery pack is ensured to be connected stably, and the conditions of poor contact, wiring error and the like are avoided so as to ensure the accuracy and reliability of the divided single battery voltage data. In addition, the method also accords with the advice and safety operation regulations of battery pack manufacturers, and ensures the operation safety.
S202: decomposing the charging voltage data of the single battery by a continuous variable mode decomposition algorithm to obtain a plurality of IMF components;
in this embodiment, the continuous variable modal decomposition algorithm (Successive variational mode decomposition, SVMD) adaptively decomposes a signal into a series of IMF components by continuously applying VMEs to the signal to decompose the signal. Compared with the existing VMD, the SVMD does not need to know the number of IMFs in advance, has lower computational complexity, and simultaneously avoids the problem that the initial center frequency in the VME is difficult to determine.
Specifically, the decomposition of the charging voltage data by the SVMD algorithm is implemented by an existing means, such as processing logic that can refer to the SVMD algorithm to decompose the vibration information, and the present invention does not make any improvement on the decomposition logic itself. The SVMD algorithm is completed by continuously executing VME on the data and adding certain constraints to the extraction process to prevent convergence to the previously extracted IMF until all IMF components are extracted or the reconstruction error (the error between the sum of all IMF components and the original data) is less than a given threshold, and the extraction process is ended.
S203: calculating kurtosis values of the IMF components and selecting the IMF component with the largest kurtosis value as a target IMF;
in this embodiment, the kurtosis value of the IMF component is calculated by the following formula:
wherein: k represents the kurtosis value of the IMF component; x is x i Representing an ith value in the IMF component;representing the average of the values in the IMF component and n representing the number of samples in the IMF component.
S204: and carrying out envelope demodulation on the target IMF, and further extracting and obtaining corresponding charging characteristic information.
In this example, a larger kurtosis value represents steeper data. The modal components may be specifically envelope demodulated (i.e., the signal is enveloped) by Hilbert transform. Firstly, taking a model of a signal subjected to Hilbert transformation to obtain an envelope curve; then performing fast Fourier transform fft on the envelope curve to obtain an envelope spectrum; and finally, extracting charge and discharge characteristic frequency and harmonic waves thereof based on the envelope spectrum as charge characteristic information.
According to the invention, through the steps, the decomposition of the charging voltage data of the single battery by utilizing the SVMD algorithm is realized, a plurality of IMF components can be obtained by high-precision signal decomposition, and meanwhile, through selecting the IMF component with the maximum kurtosis value for envelope demodulation analysis, a fault component with more obvious signal is observed, so that the efficiency and effect of the charging voltage data decomposition can be improved, and the accuracy of the charging and discharging feature extraction is improved. The optimization in the SVMD can be approximately regarded as K one-dimensional optimization problems, compared with the existing VMD, the method has lower computational complexity, the trouble of determining the number of IMFs related to the VMD can be well avoided, and meanwhile, the problem that the initial center frequency in the VME is difficult to determine is avoided.
In a specific implementation process, the SVMD has a problem that balance parameters are difficult to determine, different balance parameter values may cause different numbers of IMFs (eigen mode functions), and when a target IMF is selected for analysis according to a kurtosis maximum criterion, fault feature information carried by different target IMFs due to different balance parameter presets may be different, so that accuracy and effectiveness of fault feature extraction are low.
Therefore, the invention adaptively searches the optimal balance parameters of the continuous variable modal decomposition algorithm through the whale optimization algorithm. The WOA (Whale Optimization Algorithm ) algorithm is a bionics-based optimization algorithm with inspiration derived from whale population behavior. It is a new meta-heuristic optimization algorithm that uses random or optimal search agents to simulate hunting behavior to chase the prey, and spirals to simulate the bubble network attack mechanism of the whale of the seated user. The whale optimization algorithm is used as a group intelligent algorithm with relatively hot fire in recent years, and has excellent effect on the optimal balance parameters of the optimal SVMD algorithm.
The whale optimization algorithm specifically comprises the following steps:
s2021: determining an fitness function of a whale optimization algorithm and initializing whale populations, wherein each whale represents an optimal balance parameter of a continuous variable modal decomposition algorithm;
in this embodiment, parameters of the whale optimization algorithm further include: population number n=30, maximum iteration number max_iter=10, upper and lower bounds ub= [10,2] and lb= [100,7] of input parameter combination, target dimension dim=2.
S2022: calculating the fitness value of each whale through a fitness function, and selecting the optimal whale;
in this embodiment, an existing fitness function is selected, and the whale with the largest fitness value is selected as the optimal whale.
S2023: performing surrounding prey, bubble net attack and searching for prey based on the optimal whale location, updating whale location;
s2024: updating the fitness value based on the position of the whale, and recording the optimal balance parameter corresponding to the current optimal whale;
s2025: judging whether a termination condition is satisfied: if yes, outputting the optimal balance parameters; otherwise, the process returns to step S2023.
1) Surrounding the prey means that the whale at the base recognizes the prey locations and surrounds them, the whale optimization algorithm assuming the current optimal whale as the target prey, the other whales trying to update in the iteration;
the formula is described as follows:
wherein: t represents the current iteration;and->Representing the coefficient vector; />A position vector representing the best solution currently obtained;a position vector representing a current solution; the absolute value is represented by; />Representing the distance between whale and prey;
the vector is calculated by the following formulaAnd->
Wherein:the value of (2) decreases linearly from 2 to 0; />Is [0,1]]Is a random vector in (a).
2) The bubble net attack is to simulate the bubble net behavior of the whale of the seat head by a spiral updating position method;
the formula is described as follows:
wherein:representing the distance from the ith whale to the prey; b represents a constant; l represents [ -1,1]Random numbers in (a);
the circle is contracted while the whale is swimming along the spiral path, assuming a 0.5 probability of choosing between a contracted envelope mechanism or spiral model to optimize the position of the updating whale;
the formula is described as follows:
wherein: p represents a random number in [0,1 ].
3) Searching for prey means that whales are searched randomly according to each other's position, and can be usedForced searching is carried out when the random value is more than 1 or less than-1;
the formula is described as follows:
wherein:representing a random position vector selected from the current population.
According to the invention, on the basis of carrying out battery charging voltage data decomposition through a continuous variable mode decomposition algorithm, the problem that the SVMD algorithm needs to preset balance parameters and different balance parameter values determine SVMD decomposition precision is solved, and the parameter selection of SVMD is optimized through WOA (whale optimization algorithm), so that the optimal balance parameters can be generated for the SVMD algorithm by utilizing the global searching capability of WOA to fully exert the signal decomposition performance of the SVMD algorithm, and the accuracy of the SVMD algorithm on the extraction of charging and discharging characteristics is further improved.
In the specific implementation process, the working environment of the battery is complex and changeable, so that the risk prediction model is required to be applicable to battery thermal runaway risk prediction under various working conditions. However, in the prior art, when a risk prediction model suitable for various working conditions is trained, historical data under large-scale different working conditions is required to be acquired to train the risk prediction model, so that the thermal runaway risk prediction precision of the model can be ensured, and the model is difficult to train and high in training cost.
For this purpose, the risk prediction model in the present invention is constructed based on a long short-term memory neural network (LSTM), a gate cycle neural network (GRU), or a Convolutional Neural Network (CNN), and requires two training.
Training for the first time: training a risk prediction model through charging characteristic information of a battery under a certain working condition to obtain a pre-training model;
training for the second time: freezing a core neural network layer of the pre-training model and calling model parameters thereof to obtain an optimized model; and then training an optimization model through the charging characteristic information of the battery under other different working conditions to reconstruct a full-connection output layer of the battery, so as to obtain a trained risk prediction model.
The first training is performed by the following steps:
s301: performing battery charge-discharge cycle experiments under different working conditions to obtain corresponding charge voltage data; then extracting charging characteristic information from the charging voltage data (the charging characteristic information can be extracted by referring to the mode) to construct a charging characteristic data set under various working conditions;
s302: acquiring charging characteristic information from a charging characteristic data set under a certain working condition as primary training data;
in this embodiment, the charging characteristic information of 20% capacity before the battery is used as the primary training data, so as to improve the training effect of the model.
S303: training a risk prediction model through primary training data;
s304: calculating corresponding training loss based on the predicted thermal runaway risk value and the real thermal runaway risk value (which can be predefined through experiments) output by the risk prediction model, and optimizing model parameters through the training loss;
s305: steps S302 to S304 are repeated until the risk prediction model converges.
The second training is performed by the following steps:
s311: freezing the parameters of a nuclear network layer of the pre-training model and calling the model parameters of the pre-training model through a Fine tuning technology (Fine-Tune) of transfer learning, connecting the output of the pre-training model to a full-connection layer to form a full-connection output layer and performing linear change on hidden layer neuron activation to obtain an optimized model;
s312: acquiring charging characteristic information from the charging characteristic data sets under different working conditions as secondary training data;
s313: training an optimization model through the secondary training data to adjust parameters of a full-connection output layer of the optimization model;
s314: calculating corresponding training loss based on the predicted thermal runaway risk value and the real thermal runaway risk value (which can be predefined through experiments) output by the optimization model, and optimizing model parameters through the training loss;
s315: steps S312 to S314 are repeated until the risk prediction model converges.
According to the risk prediction model, the deep neural network is used for constructing and training twice, the first training is performed through the charging characteristic information under a specific working condition to obtain the pre-training model, and the pre-training model with higher precision can be obtained through training by aiming at batteries working under the same type and the same working condition through smaller-scale historical data (but the pre-training model is poor in universality and cannot be used for accurately predicting thermal runaway risks under other chemical components or working conditions); the second training takes the pre-training model as a source model, freezes the parameters of a nuclear network layer of the pre-training model through a micro-tuning technology of transfer learning, invokes the parameters of the model (the thermal runaway risk prediction among different working conditions of the battery has a certain commonality), and adjusts the parameters of a fully-connected output layer of the risk prediction model through charging characteristic information under different working conditions, so that the model can adapt to various working conditions by utilizing smaller-scale historical data, namely the risk prediction model can be suitable for the thermal runaway risk prediction under various working conditions without large-scale training data, the training difficulty and cost of the model can be reduced, the application range of the model is improved, and the management effect on the battery of the electric automobile can be further improved.
In the specific implementation process, the loss function of the risk prediction model adopts the following mean square error loss function;
wherein: y is i A thermal runaway risk target value indicating the time i;a thermal runaway risk prediction value indicating the time i; n represents the number of samples.
The optimizer of the risk prediction model adopts the following adaptive moment estimation algorithm;
m t ＝β 1 m t-1 +(1-β 1 )g t
wherein: m is m t And m t-1 Representing smoothed sliding means of t times and t-1 times of iteration; t represents the iteration number; beta 1 And beta 2 Representing a smoothing constant; g t Representing an objective function gradient; v (V) t A sliding average representing the square of the gradient; w (w) t And w t+1 Update parameters respectively representing t times and t+1 times of iteration; alpha t The learning rate of t iterations is represented; epsilon=10 -8 Indicating that the avoidance divisor is 0;and->The power t of the smoothing constant is shown.
According to the invention, a mean square error loss function is adopted as a loss function, a self-adaptive moment estimation algorithm is adopted as an optimizer, and the output of the gate cycle neural network is connected to the full-connection layer, so that a model with higher precision can be obtained through training by using small-scale historical data, and partial parameters of the model under different working conditions can be adjusted by using charging characteristic information under different working conditions, namely, the model is suitable for thermal runaway risk prediction under various working conditions without large-scale training data, thereby reducing training difficulty and cost of the model and improving application range of the model.
Finally, it should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the technical solution, and those skilled in the art should understand that modifications and equivalents may be made to the technical solution of the present invention without departing from the spirit and scope of the present invention, and all such modifications and equivalents are included in the scope of the claims.

Claims (10)

1. The electric automobile battery management method based on data analysis is characterized by comprising the following steps of:
s1: acquiring charging voltage data of a battery pack in the process of charging the electric automobile by a user;
s2: extracting charging characteristic information of each single battery from charging voltage data of the battery pack through a continuous variable mode decomposition algorithm;
s3: inputting the charging characteristic information of the single battery into a risk prediction model constructed based on a deep learning model to obtain a corresponding predicted thermal runaway risk value;
s4: taking a single battery with a predicted thermal runaway risk value exceeding a risk threshold as a battery to be monitored;
s5: and acquiring data of the charge and discharge process of the battery to be monitored and feeding the data back to a user of the electric automobile so that the user knows the charge and discharge state of the battery to be monitored.
2. The method for managing batteries of an electric vehicle based on data analysis according to claim 1, wherein: in step S2, discharge data of the single battery is extracted by:
s201: dividing charging voltage data of each single battery from charging voltage data of the battery pack;
s202: decomposing the charging voltage data of the single battery by a continuous variable mode decomposition algorithm to obtain a plurality of IMF components;
s203: calculating kurtosis values of the IMF components and selecting the IMF component with the largest kurtosis value as a target IMF;
s204: and carrying out envelope demodulation on the target IMF, and further extracting and obtaining corresponding charging characteristic information.
3. The method for managing batteries of an electric vehicle based on data analysis according to claim 2, wherein: in step S202, the optimal balance parameters of a continuous variable modal decomposition algorithm are adaptively searched through a whale optimization algorithm;
the method specifically comprises the following steps:
s2021: determining an fitness function of a whale optimization algorithm and initializing whale populations, wherein each whale represents an optimal balance parameter of a continuous variable modal decomposition algorithm;
s2022: calculating the fitness value of each whale through a fitness function, and selecting the optimal whale;
s2023: performing surrounding prey, bubble net attack and searching for prey based on the optimal whale location, updating whale location;
s2024: updating the fitness value based on the position of the whale, and recording the optimal balance parameter corresponding to the current optimal whale;
s2025: judging whether a termination condition is satisfied: if yes, outputting the optimal balance parameters; otherwise, the process returns to step S2023.
4. The method for managing batteries of an electric vehicle based on data analysis according to claim 3, wherein: in step S2023, the formula surrounding the prey is described as follows:
wherein: t represents the current iteration;and->Representing the coefficient vector; />A position vector representing the best solution currently obtained; />A position vector representing a current solution; the absolute value is represented by; />Representing the distance between whale and prey;
the vector is calculated by the following formulaAnd->
Wherein:the value of (2) decreases linearly from 2 to 0; />Is [0,1]]Is a random vector in (a).
5. The method for managing batteries of an electric vehicle based on data analysis according to claim 4, wherein: in step S2023, the formula of the bubble attack is described as follows:
wherein:representing the distance from the ith whale to the prey; b represents a constant; l represents [ -1,1]Random numbers in (a);
the circle is contracted while the whale is swimming along the spiral path, assuming a 0.5 probability of choosing between a contracted envelope mechanism or spiral model to optimize the position of the updating whale;
the formula is described as follows:
wherein: p represents a random number in [0,1 ].
6. The method for managing batteries of an electric vehicle based on data analysis according to claim 5, wherein: in step S2023, the formula for searching for a prey is described as follows:
wherein:representing a random position vector selected from the current population.
7. The method for managing batteries of an electric vehicle based on data analysis according to claim 2, wherein: in step S203, the kurtosis value of the IMF component is calculated by the following formula:
wherein: k represents the kurtosis value of the IMF component; x is x i Representing an ith value in the IMF component;representing the average of the values in the IMF component and n representing the number of samples in the IMF component.
8. The method for managing batteries of an electric vehicle based on data analysis according to claim 1, wherein: in step S3, the risk prediction model is trained twice;
training for the first time: training a risk prediction model through charging characteristic information of a battery under a certain working condition to obtain a pre-training model;
training for the second time: freezing a core neural network layer of the pre-training model and calling model parameters thereof to obtain an optimized model; and then training an optimization model through the charging characteristic information of the battery under other different working conditions to reconstruct a full-connection output layer of the battery, so as to obtain a trained risk prediction model.
9. The method for managing batteries of an electric vehicle based on data analysis according to claim 8, wherein: the first training is performed by the following steps:
s301: performing battery charge-discharge cycle experiments under different working conditions to obtain corresponding charge voltage data; then extracting charging characteristic information from the charging voltage data to construct a charging characteristic data set under various working conditions;
s302: acquiring charging characteristic information from a charging characteristic data set under a certain working condition as primary training data;
s303: training a risk prediction model through primary training data;
s304: calculating corresponding training loss based on the predicted thermal runaway risk value and the real thermal runaway risk value output by the risk prediction model, and optimizing model parameters through the training loss;
s305: steps S302 to S304 are repeated until the risk prediction model converges.
10. The method for managing batteries of an electric vehicle based on data analysis according to claim 9, wherein: the second training is performed by the following steps:
s311: freezing nuclear network layer parameters of the pre-training model through a transfer learning fine tuning technology, calling model parameters of the pre-training model, connecting output of the pre-training model to a full-connection layer to form a full-connection output layer, and performing linear change on hidden layer neuron activation to obtain an optimized model;
s312: acquiring charging characteristic information from the charging characteristic data sets under different working conditions as secondary training data;
s313: training an optimization model through the secondary training data to adjust parameters of a full-connection output layer of the optimization model;
s314: calculating corresponding training loss based on the predicted thermal runaway risk value and the real thermal runaway risk value output by the optimization model, and optimizing model parameters through the training loss;
s315: steps S312 to S314 are repeated until the risk prediction model converges.
CN202311106776.7A 2023-08-30 2023-08-30 Electric automobile battery management method based on data analysis Pending CN117131687A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311106776.7A CN117131687A (en) 2023-08-30 2023-08-30 Electric automobile battery management method based on data analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311106776.7A CN117131687A (en) 2023-08-30 2023-08-30 Electric automobile battery management method based on data analysis

Publications (1)

Publication Number Publication Date
CN117131687A true CN117131687A (en) 2023-11-28

Family

ID=88850517

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311106776.7A Pending CN117131687A (en) 2023-08-30 2023-08-30 Electric automobile battery management method based on data analysis

Country Status (1)

Country Link
CN (1) CN117131687A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117799498A (en) * 2024-03-01 2024-04-02 湘潭南方电机车制造有限公司 Comprehensive protection system for explosion-proof storage battery electric locomotive

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117799498A (en) * 2024-03-01 2024-04-02 湘潭南方电机车制造有限公司 Comprehensive protection system for explosion-proof storage battery electric locomotive

Similar Documents

Publication Publication Date Title
US11544917B2 (en) Power electronic circuit fault diagnosis method based on optimizing deep belief network
KR102439041B1 (en) Method and apparatus for diagnosing defect of battery cell based on neural network
KR20200140093A (en) Prediction Method and Prediction System for predicting Capacity Change according to Charging / Discharging Cycle of Battery
CN114372417A (en) Electric vehicle battery health state and remaining life evaluation method based on charging network
CN110806541B (en) AD-BAS-based lithium battery model parameter identification method
CN111680848A (en) Battery life prediction method based on prediction model fusion and storage medium
CN117131687A (en) Electric automobile battery management method based on data analysis
CN112215434A (en) LSTM model generation method, charging duration prediction method and medium
CN112881914B (en) Lithium battery health state prediction method
CN116449218B (en) Lithium battery health state estimation method
CN112731183A (en) Lithium ion battery life prediction method based on improved ELM
CN117031310A (en) Method for predicting residual service life of power battery of electric automobile
CN113791351B (en) Lithium battery life prediction method based on transfer learning and difference probability distribution
CN113538037A (en) Method, system, equipment and storage medium for monitoring charging event of battery car
WO2022144542A1 (en) Method for predicting condition parameter degradation of a cell
Pei et al. The real‐time state identification of the electricity‐heat system based on Borderline‐SMOTE and XGBoost
Zhang et al. SOH estimation and RUL prediction of lithium batteries based on multidomain feature fusion and CatBoost model
CN116796821A (en) Efficient neural network architecture searching method and device for 3D target detection algorithm
CN116632834A (en) Short-term power load prediction method based on SSA-BiGRU-Attention
Zhang et al. An Energy Management Strategy Based on DDPG With Improved Exploration for Battery/Supercapacitor Hybrid Electric Vehicle
CN116579789A (en) Power battery performance analysis-based secondary vehicle estimation method and system
CN113687237B (en) Lithium battery residual charging time prediction method for guaranteeing electrical safety
KR20230142121A (en) Method and apparatus for predicting usage of battery based on neural network
JP7314822B2 (en) Battery deterioration determination device, battery deterioration determination method, and battery deterioration determination program
CN117394409B (en) Intelligent assessment method and system for equipment state of energy storage power station

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