Knowledge-data fusion driven power battery available energy prediction method and system

Through the knowledge-data fusion-driven method, a power battery mechanism and characteristic prediction model is established, and combined with the data-driven model, the problem of poor prediction accuracy of power battery available energy is solved, the prediction accuracy and scene coverage are improved, and the prediction accuracy is improved. The mileage range of new energy vehicles is estimated.

CN118962456AActive Publication Date: 2024-11-15BEIHANG UNIV +1
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411046506.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-15
Estimated Expiration
2044-07-31

AI Technical Summary

Technical Problem

The available energy prediction accuracy of existing power batteries is poor, it is difficult to balance the calculation quantity and calculation accuracy, and there is a lack of effective adaptive correction methods, especially when the vehicle end operation conditions are complex.

Method used

Using the knowledge-data fusion-driven method, establish a mechanism model and battery characteristic prediction model that conforms to the basic principles of power batteries, combine the data-driven model to predict and correct available energy, analyze vehicle-side data through cloud platform, reduce dimensionality characteristics, and generate and use In operating conditions, establish a battery characteristic prediction model of metaheuristic algorithms and physical information neural networks, and predict through offline detection and correction model.

Benefits of technology

Improves the prediction accuracy of power battery availability and scene coverage, reduces the uninterpretation of the data-driven model, improves the poor open-loop prediction accuracy of the mechanism model, and has important application value in the mileage estimation of new energy vehicles. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118962456A_ABST
    Figure CN118962456A_ABST
Patent Text Reader

Abstract

The invention provides a knowledge-data fusion driven power battery available energy prediction method and system, and the method comprises the steps: sequentially building a battery mechanism model, a battery characteristic prediction model and a data driving model of credible data which accord with the basic principle of a power battery; the available energy prediction of the battery in an open-loop mode is realized through a mechanism-based model and a battery characteristic prediction model, and the available energy correction under a credible data condition is realized through a data driving model, so that the available energy prediction of the power battery based on the integration of a knowledge driving method and a data driving method is realized. According to the method and the system, the available energy prediction precision and the scene coverage degree of the battery can be improved, and the uninterpretability of a data-driven model is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of power battery testing, and in particular to a method and system for predicting available energy of a power battery driven by knowledge-data fusion. Background Art

[0002] As the production, sales and ownership of new energy vehicles continue to increase, the consumer market's requirements for new energy vehicles are constantly increasing. As a key component of new energy vehicles, the performance and control of power batteries directly affect the user experience of this new energy vehicle customer group.

[0003] However, the state estimation and energy consumption prediction of vehicle-mounted power batteries often lack methods that can be widely used because their models are difficult to balance the amount of calculation and calculation accuracy. The existing open-loop prediction accuracy of available energy of power batteries is poor. Therefore, it is necessary to propose new solutions to address the complex operating conditions of vehicle-side power batteries and the lack of continuous adaptive correction methods. Summary of the invention

[0004] In order to solve the problems of poor accuracy in the existing prediction of available energy of power batteries, the present invention proposes a method for predicting available energy of power batteries driven by knowledge-data fusion, which is a data-driven model that establishes a battery mechanistic model, a battery characteristic prediction model and reliable data that conform to the basic principles of power batteries, realizes the prediction of battery available energy in an open-loop mode through the mechanistic model and the battery characteristic prediction model, and realizes the correction of available energy under reliable data conditions through the data-driven model, thereby integrating the knowledge-driven method and the data-driven method to predict available energy of power batteries. This method can improve the accuracy of battery available energy prediction and the coverage of scenarios, and reduce the unexplainability of the data-driven model. The present invention also relates to a power battery available energy prediction system driven by knowledge-data fusion.

[0005] The technical solution of the present invention is as follows:

[0006] A method for predicting available energy of a power battery driven by knowledge-data fusion, characterized by comprising the following steps;

[0007] S1. Obtain the cloud data uploaded to the cloud platform by the vehicle-side power battery system, extract the power battery usage condition characteristics by performing data analysis on the acquired cloud data on the cloud platform, use the data dimension reduction algorithm to reduce the dimension of the high-dimensional characteristics of the power battery usage condition, perform correlation analysis on the reduced-dimensional characteristics based on physical knowledge, and then use the data-driven working condition feature classification method to classify, calculate the probability map of the power battery jumping between different working conditions, and use the working condition generation algorithm to establish the power battery usage characteristic working condition;

[0008] S2. On the cloud platform, a power battery mechanical model is established based on the power battery usage principle. The input conditions of the power battery mechanical model are the charge and discharge current / power, temperature, and SOC initial value parameters. The output conditions of the power battery mechanical model are the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery. Based on the power battery mechanical model, a battery characteristic prediction model integrating a metaheuristic algorithm and a physical information neural network is established. The battery characteristic prediction model simulates the input power battery usage characteristic working conditions to predict the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery under given input working conditions. On the cloud platform, the predicted positive and negative electrode ion concentrations and the calibrated positive and negative electrode ion concentrations of the power battery are used to predict the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery. The SOC of the positive and negative electrodes is calculated by the maximum ion concentration of the negative electrode, and then the positive electrode potential and the negative electrode potential are calculated according to the calculated SOC of the positive and negative electrodes and the calibrated positive and negative electrode SOC-OCV curves, and the electrolyte potential is calculated using the predicted electrolyte ion concentration, and then the negative electrode potential, the electrolyte potential, and the ohmic voltage drop obtained according to the characteristic working conditions of the power battery are respectively subtracted from the positive electrode potential to obtain the power battery terminal voltage, and the power battery power is obtained by the power battery terminal voltage and the current obtained according to the characteristic working conditions of the power battery, and the power battery available energy is calculated by the power-time integral when the power battery is continuously virtually discharged to reach the cut-off voltage, as a model to predict the available energy;

[0009] S3. Conduct offline testing experiments, including capacity testing and dynamic charge and discharge testing, to measure the actual test available energy of the power battery under the same initial conditions of charge and discharge current / power, temperature, and SOC initial value parameters;

[0010] S4. A data-driven approach is adopted on the cloud platform. The model predicted available energy obtained in step S2 is used as input and the actual test available energy obtained in step S3 is used as output. A training set and a test set are jointly established to construct a data-driven model. The data-driven model establishes a mapping relationship from the model predicted available energy to the actual test available energy through the data-driven approach, thereby realizing the correction of the model predicted available energy.

[0011] Preferably, it is characterized in that step S1 obtains cloud data uploaded to the cloud platform by the vehicle-side power battery system, including voltage, current, time, temperature, SOC, charge / discharge status, battery output power, battery fault status, and vehicle speed information; extracts the operating condition characteristics of the power battery by performing data analysis on the acquired cloud data on the cloud platform, including charging frequency, charge and discharge depth, acceleration time, deceleration time, average current, average voltage, average power, maximum power, minimum power, and power distribution curve characteristics; and adopts an operating condition generation algorithm to establish a power battery operating characteristic operating condition, wherein the battery operating characteristic operating condition includes a current-time operating condition curve, and the curve length selects a characteristic operating condition of less than 1800s.

[0012] Preferably, the data dimensionality reduction algorithm adopted in step S1 includes principal component analysis or partial principal component analysis or singular value decomposition or autoencoder algorithm; the data-driven operating condition feature classification method adopted includes K-means clustering algorithm, random forest algorithm, support vector machine algorithm; the operating condition generation algorithm adopted includes Monte Carlo simulation method, decision tree algorithm, decision forest algorithm.

[0013] Preferably, in step S2, a power battery mechanistic model including an equivalent circuit model, an electrochemical model or a heterogeneous model is established on a cloud platform according to the principle of power battery use. The equivalent circuit model is a first-order equivalent circuit model, a second-order equivalent circuit model, a fractional-order equivalent circuit model or other electrochemically reduced-order equivalent circuit model. The electrochemical model is a pseudo-two-dimensional electrochemical model, a single-particle model or a three-dimensional electrochemical model. The pseudo-two-dimensional electrochemical model integrates solid-phase lithium ion concentration calculation, liquid-phase lithium ion concentration calculation, solid-phase potential calculation, liquid-phase potential calculation and lithium deintercalation.

[0014] Preferably, step S2 establishes a battery characteristic prediction model that integrates a metaheuristic algorithm and a physical information neural network based on a power battery mechanistic model, introduces a prediction deviation of the power battery mechanistic model into the penalty function of the physical information neural network algorithm, and assigns a weight greater than a weight threshold, and then uses a metaheuristic algorithm to perform parameter optimization during the training process; the metaheuristic algorithm used is a particle swarm optimization-grey wolf optimization coupling algorithm, a chaos search algorithm, and a particle swarm optimization-simulated annealing coupling algorithm; the physical information neural network algorithm is a long short-term memory neural network algorithm, a graph neural network algorithm, or a bidirectional long short-term memory neural network algorithm with an attention mechanism.

[0015] Preferably, step S3 carries out offline detection experiments to measure the actual test available energy of the power battery under the same initial conditions of charge and discharge current / power, temperature, and SOC initial value parameters, including the actual test available energy of the power battery system under NEDC or CLTC test conditions, or the actual test available energy of the power battery system under characteristic usage conditions, or the actual test available energy of the power battery cell under DST and FUDS test conditions.

[0016] Preferably, the data-driven model constructed in step S4 establishes a mapping relationship from model-predicted available energy to actual tested available energy through a data-driven method, and the data-driven method adopts an extreme learning machine algorithm, a BP neural network, a recurrent neural network, an ELMAN neural network, or an LSTM neural network method.

[0017] A power battery available energy prediction system driven by knowledge-data fusion, characterized in that it comprises a power battery use characteristic working condition establishment module, a model prediction available energy calculation module, an actual test available energy test module and a data-driven model construction module, all of which are located on a cloud platform, wherein the power battery use characteristic working condition establishment module is connected to the model prediction available energy calculation module, and the model prediction available energy calculation module and the actual test available energy test module are both connected to the data-driven model construction module;

[0018] The power battery usage characteristic working condition establishment module obtains the cloud data uploaded to the cloud platform by the vehicle-side power battery system, extracts the usage working condition characteristics of the power battery by performing data analysis on the acquired cloud data, reduces the dimension of the high-dimensional features of the power battery usage working condition by using a data dimensionality reduction algorithm, performs correlation analysis on the reduced-dimensional features based on physical knowledge, and then classifies them by using a data-driven working condition feature classification method, calculates the probability map of the power battery jumping between different working conditions, and uses a working condition generation algorithm to establish the power battery usage characteristic working condition;

[0019] The model prediction available energy calculation module includes a power battery mechanism model construction module, a battery characteristic prediction model construction module and a model prediction available energy calculation processing module connected in sequence. The power battery mechanism model construction module establishes a power battery mechanism model according to the power battery use principle. The input conditions of the power battery mechanism model are the charge and discharge current / power, temperature, and SOC initial value parameters. The output conditions of the power battery mechanism model are the positive and negative electrode ion concentrations and the electrolyte ion concentrations of the power battery. The battery characteristic prediction model construction module establishes a battery characteristic prediction model that integrates a metaheuristic algorithm and a physical information neural network based on the power battery mechanism model. The battery characteristic prediction model simulates the input power battery use characteristic working conditions to predict the positive and negative electrode ions of the power battery under given input working conditions. concentration and electrolyte ion concentration; the model predicted available energy calculation processing module calculates the SOC of the positive and negative electrodes according to the predicted positive and negative electrode ion concentrations and the calibrated maximum ion concentrations of the positive and negative electrodes of the power battery, and then calculates the positive electrode potential and the negative electrode potential according to the calculated SOC of the positive and negative electrodes and the calibrated positive and negative electrode SOC-OCV curves, calculates the electrolyte potential using the predicted electrolyte ion concentration, and then subtracts the negative electrode potential, the electrolyte potential, and the ohmic voltage drop obtained according to the characteristic working conditions of the power battery from the positive electrode potential, thereby obtaining the power battery terminal voltage, and obtains the power battery power through the power battery terminal voltage and the current obtained according to the characteristic working conditions of the power battery, and calculates the power battery available energy through the power-time integral when the power battery is continuously virtually discharged to reach the cut-off voltage, as the model predicted available energy;

[0020] The actual test available energy test module carries out offline detection experiments, including capacity test and dynamic charge and discharge test, to measure the actual test available energy of the power battery under the initial conditions of the same charge and discharge current / power, temperature, and SOC initial value parameters;

[0021] The data-driven model construction module adopts a data-driven method, takes the model predicted available energy as input and the actual tested available energy as output, jointly establishes a training set and a test set, and constructs a data-driven model. The data-driven model establishes a mapping relationship from the model predicted available energy to the actual tested available energy through the data-driven method, thereby realizing the correction of the model predicted available energy.

[0022] Preferably, the power battery usage characteristic operating condition establishment module obtains cloud data uploaded to the cloud platform by the vehicle-side power battery system, including voltage, current, time, temperature, SOC, charge / discharge status, battery output power, battery fault status, and vehicle speed information; extracts the power battery usage operating condition characteristics by performing data analysis on the acquired cloud data, including charging frequency, charge and discharge depth, acceleration time, deceleration time, average current, average voltage, average power, maximum power, minimum power, and power distribution curve characteristics; the data dimensionality reduction algorithm adopted includes principal component analysis or partial principal component analysis or singular value decomposition or autoencoder algorithm; the data-driven operating condition feature classification method adopted includes K-means clustering algorithm, random forest algorithm, and support vector machine algorithm; the operating condition generation algorithm is adopted to establish the power battery usage characteristic operating condition, and the operating condition generation algorithm adopted includes Monte Carlo simulation method, decision tree algorithm, and decision forest algorithm. The battery usage characteristic operating condition includes current-time operating condition curve, and the curve length selects characteristic operating conditions of less than 1800s.

[0023] Preferably, the battery characteristic prediction model construction module in the model prediction available energy calculation module establishes a battery characteristic prediction model that integrates a metaheuristic algorithm and a physical information neural network based on a power battery mechanistic model, introduces a prediction deviation of the power battery mechanistic model into the penalty function of the physical information neural network algorithm, and assigns a weight greater than a weight threshold, and then uses a metaheuristic algorithm to perform parameter optimization during the training process; the metaheuristic algorithm used is a particle swarm optimization-grey wolf optimization coupling algorithm, a chaos search algorithm, and a particle swarm optimization-simulated annealing coupling algorithm; the physical information neural network algorithm is a long short-term memory neural network algorithm, a graph neural network algorithm, or a bidirectional long short-term memory neural network algorithm with an attention mechanism.

[0024] The beneficial effects of the present invention are:

[0025] The present invention proposes a method for predicting the available energy of a power battery driven by knowledge-data fusion. First, cloud data uploaded to the cloud platform by the vehicle-side power battery system is obtained, such as voltage, current, time and other information. Based on the cloud platform, the battery usage condition characteristics are extracted by analyzing the data, and the high-dimensional characteristics of the power battery usage condition are reduced by a data dimension reduction algorithm. The reduced characteristics are analyzed for correlation according to physical knowledge, and data-driven methods such as cluster analysis and support vector machines are used for classification. The probability map of the battery jumping between different working conditions is calculated, and then the Monte Carlo simulation method is used to establish the power battery usage characteristic working condition. Secondly, according to the power battery usage principle, a power battery mechanistic model such as an equivalent circuit or an electrochemical model is established, and then a battery characteristic prediction model integrating a meta-heuristic algorithm and a physical information neural network is constructed to predict the positive and negative ions of the power battery under a given input condition. The power battery terminal voltage is obtained by a series of logical calculations based on the positive and negative electrode ion concentrations and the electrolyte ion concentrations, combined with the calibrated maximum ion concentrations of the positive and negative electrodes of the power battery, the positive and negative electrode SOC-OCV curves and other parameters, and the simulated predicted available energy is calculated, combined with the real available energy of the power battery under the same initial conditions measured by offline experiments (or called the actual test available energy); data-driven methods such as neural networks and regression analysis can be used to construct a data-driven model, and a mapping relationship from the model-predicted available energy to the actual test available energy is established through data-driven methods such as neural networks; in this way, in the actual use process, the established battery usage characteristic conditions are used, and the available energy (model-predicted available energy) under the battery state is predicted through the mechanistic model and the battery characteristic prediction model, and the model-predicted available energy is corrected using the data-driven method to achieve accurate available energy prediction. The present invention is a method for predicting available energy of a power battery that establishes a battery mechanistic model, a battery characteristic prediction model and a data-driven model of trusted data that conform to the basic principles of power batteries. The mechanistic model and the battery characteristic prediction model are used to predict the available energy of the battery in an open-loop mode. The data-driven model is used to correct the available energy under the condition of trusted data, thereby integrating the knowledge-driven method and the data-driven method to predict the available energy of the power battery. The method can improve the accuracy of the prediction of the available energy of the battery and the coverage of scenarios, and improve the interpretability of the data-driven model, thereby improving the problem of poor open-loop prediction accuracy of the mechanistic model, and has important application value in improving the accuracy of the estimation of the driving range of new energy vehicles.

[0026] The present invention also relates to a knowledge-data fusion driven power battery available energy prediction system, which corresponds to the above-mentioned knowledge-data fusion driven power battery available energy prediction method, and can be understood as a system that realizes the above-mentioned knowledge-data fusion driven power battery available energy prediction method. The system includes a power battery usage characteristic operating condition establishment module, a model prediction available energy calculation module, an actual test available energy test module and a data-driven model construction module, all of which are located on a cloud platform. Each module works in a systematic manner with each other, adopts data dimension reduction algorithm, physical knowledge drive, correlation analysis, data-driven operating condition feature classification method, operating condition generation algorithm, power battery mechanistic model modeling, battery characteristic prediction model modeling and model prediction available energy calculation processing, combined with the construction of a data-driven model, the available energy correction under the condition of reliable data is realized through a data-driven method, thereby integrating the knowledge-driven method and the data-driven method, improving the battery available energy prediction accuracy and scenario coverage, and reducing the unexplainability of the data-driven model. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the power battery available energy prediction method driven by knowledge-data fusion of the present invention.

[0028] Figure 2 It is a preferred flow chart of the method for predicting available energy of power batteries driven by inventive knowledge-data fusion.

[0029] Figure 3 This is the characteristic operating condition of the power battery in the embodiment of the present invention.

[0030] Figure 4 It is a flow chart of establishing a battery characteristic prediction model of the present invention.

[0031] Figure 5 It is a flow chart of the calculation process of available energy predicted by the model of the present invention.

[0032] Figure 6 This is a diagram of the offline detection of a power battery operating condition according to an embodiment of the present invention.

[0033] Figure 7 It is a comparison diagram before and after the neural network correction of the embodiment of the present invention.

[0034] Figure 8 It is a structural block diagram of the power battery available energy prediction system driven by knowledge-data fusion of the present invention. DETAILED DESCRIPTION

[0035] The present invention will be described below in conjunction with the accompanying drawings.

[0036] The present invention relates to a method for predicting available energy of a power battery driven by knowledge-data fusion. The principle of the method is as follows: Figure 1As shown, a method is to obtain cloud data of power batteries, establish characteristic operating conditions, and then establish a battery model, or a battery mechanism model and a battery characteristic prediction model that conforms to the basic principles of power batteries, to achieve battery available energy prediction (i.e., model prediction of available energy, or model simulation prediction of available energy), and obtain real available energy through offline detection experiments. The model simulation predicted available energy is used as input, and the real available energy obtained through experiments is used as output. Machine learning training is performed, and a machine learning training model is constructed (or a data-driven model is constructed) to achieve correction of the model simulation predicted available energy, thereby integrating knowledge-driven methods and data-driven methods to achieve a power battery available energy prediction method.

[0037] Specifically, the flowchart of the power battery available energy prediction method driven by knowledge-data fusion is as follows: Figure 2 As shown, the following steps are included;

[0038] S1: Obtain cloud data and establish characteristic operating conditions, which can also be called the step of establishing characteristic operating conditions for power battery use.

[0039] Obtain the cloud data uploaded to the cloud platform by the vehicle-side power battery system, extract the power battery usage condition characteristics by performing data analysis on the acquired cloud data on the cloud platform, use the data dimension reduction algorithm to reduce the dimension of the high-dimensional characteristics of the power battery usage condition, perform correlation analysis on the reduced characteristics based on physical knowledge, and then use the data-driven condition feature classification method to classify them, calculate the probability map of the power battery jumping between different conditions, and use the condition generation algorithm to establish the power battery usage characteristic condition.

[0040] Furthermore, cloud data such as voltage, current, time, temperature, SOC, charge / discharge status, battery output power, battery fault status, vehicle speed information, etc. of the power battery are obtained through the cloud platform as real application data, and initial values ​​are input. The above-mentioned data characteristics are used as characteristic data to form the battery operating condition, and then data analysis is performed to extract the power battery's operating condition characteristics, including charging frequency, charge and discharge depth, acceleration time, deceleration time, average current, average voltage, average power, maximum power, minimum power, and power distribution curve characteristics. A data dimension reduction algorithm is used to reduce the dimension of high-dimensional features of the power battery usage conditions. The data dimension reduction algorithm may include principal component analysis, partial principal component analysis, singular value decomposition, or autoencoder algorithm, etc.; the reduced features are subjected to correlation analysis based on physical knowledge, and then classified using a data-driven operating condition feature classification method. The data-driven operating condition feature classification method may include K-means clustering algorithm, DBSCAN density clustering, OPTICS density clustering, random forest algorithm, support vector machine algorithm, etc.; a probability graph of the power battery jumping between different operating conditions is calculated, and a condition generation algorithm is used to establish a power battery usage characteristic condition. The condition generation algorithm may include Monte Carlo simulation method, decision tree algorithm, decision forest algorithm, etc. The established power battery usage characteristic condition includes a current-time condition curve. The curve length may be selected from characteristic conditions of less than 1800s, or longer time segments, etc.

[0041] In this embodiment, a support vector machine is used as a classifier. After the training of the support vector machine is completed, most of the training samples do not need to be retained, and the final model is only related to the support vector. The input training set is T = {(x1, y1), (x2, y2), ..., (x N ,y N )}, where x i = {U i ,I i ,T i ,K i ,G i}, i = 1, 2, ..., N takes voltage, current, time and temperature as continuous features, y i∈{+1,-1},i=1,2,...N, and using the working condition as the classification feature, the working condition of the data under the above voltage, current, temperature and time conditions can be obtained quickly, and a probability process with the working condition as a random variable can be established. Then, the Monte Carlo random process is used to perform distributed sampling from the above probability process to establish the battery operating condition within the required time period. The process of establishing the working condition using Monte Carlo does not require discretization of the problem of continuous available energy of the battery. At the same time, for a large number of cloud data samples, the Monte Carlo working condition simulation results will be approximate to the optimal solution. The Monte Carlo random process will establish the battery usage condition characteristics, that is, the current-time working condition curve and other usage condition characteristics will be extracted, such as Figure 3 shown.

[0042] S2. Establishing a power battery model (or battery knowledge model), the model predicts available energy, which can also be called the model predicts available energy calculation step.

[0043] On the cloud platform, based on the principle of power battery use, real application data is used as the input initial value to establish a power battery mechanical model. The input conditions of the power battery mechanical model are the charge and discharge current / power, temperature, and SOC initial value parameters. The output conditions of the power battery mechanical model are the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery. Based on the power battery mechanical model, a battery characteristic prediction model integrating a metaheuristic algorithm and a physical information neural network is established. The battery characteristic prediction model simulates the input power battery use characteristic working conditions to predict the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery under given input working conditions. On the cloud platform, the predicted positive and negative electrode ion concentrations and the calibrated The SOC of the positive and negative electrodes of the power battery is calculated by the maximum ion concentration of the positive and negative electrodes, and then the positive electrode potential and the negative electrode potential are calculated according to the calculated SOC of the positive and negative electrodes and the calibrated positive and negative electrode SOC-OCV curves, and the electrolyte potential is calculated using the predicted electrolyte ion concentration. Then, on the basis of the positive electrode potential, the negative electrode potential, the electrolyte potential, and the ohmic voltage drop obtained according to the characteristic operating conditions of the power battery are subtracted respectively to obtain the power battery terminal voltage, and the power battery power is obtained by the power battery terminal voltage and the current obtained according to the characteristic operating conditions of the power battery. When the power battery is continuously virtually discharged to reach the cut-off voltage, the power battery available energy is calculated by the power-time integral, which is used as the model to predict the available energy.

[0044] S2.1. Establishing a mechanical model of power batteries

[0045] On the cloud platform, based on the principle of power battery use, a power battery mechanism model including an equivalent circuit model, an electrochemical model or a heterogeneous model is established. Among them, the equivalent circuit model can be a first-order equivalent circuit model, a second-order equivalent circuit model, a fractional-order equivalent circuit model or other electrochemically reduced-order equivalent circuit model, and the electrochemical model can be a pseudo-two-dimensional electrochemical model, a single particle model or a three-dimensional electrochemical model.

[0046] In this embodiment, a pseudo two-dimensional electrochemical model of a battery is established as a mechanistic model of a power battery. The pseudo two-dimensional electrochemical model integrates solid phase lithium ion concentration calculation, liquid phase lithium ion concentration calculation, solid phase potential calculation, liquid phase potential calculation, and lithium insertion and extraction. Specifically, the pseudo two-dimensional electrochemical model formula is as follows:

[0047] 1) Calculation of solid phase lithium ion concentration

[0048] The spherical coordinate system is established with the center of the spherical particles of the positive and negative electrode active materials as the origin. According to Fick's law, the solid phase lithium ion concentration expression at coordinate r is shown in formula (1):

[0049]

[0050] The subscript s indicates the solid phase region, y = n or p, corresponding to the negative electrode particles and the positive electrode particles, respectively; C s,y (r, t) is the solid phase lithium ion concentration; D s,y is the lithium ion diffusion coefficient of the positive and negative electrodes in the solid phase region; t is time.

[0051] 2) Calculation of liquid phase lithium ion concentration

[0052] The movement of lithium ions in the electrolyte is affected by two factors. One factor is the lithium ion concentration gradient in the liquid phase, which leads to the diffusion of lithium ions. The other factor is the migration of lithium ions under the electric field. The liquid phase lithium ion concentration satisfies formula (2):

[0053]

[0054] Where x is the integral along the electrode thickness direction; subscript e represents the liquid phase region; c e (x, t) is the liquid phase lithium ion concentration; ε e is the volume fraction of liquid electrolyte; D eff is the effective diffusion coefficient of lithium ions in the liquid phase; F is the Faraday constant; t + is the migration number of lithium ions; j tot (t) is the total current density.

[0055] 3) Solid phase potential calculation

[0056] The solid phase potential distribution in the positive and negative electrode materials is shown in formula (3):

[0057]

[0058] Among them, φ s,y (x, t) is the solid phase potential distribution, the subscript y = n or y = p, n corresponds to the negative electrode, p corresponds to the positive electrode; σ eff,y is the equivalent conductivity in the solid phase material, j tot (t) is the total current density.

[0059] 4) Calculation of liquid phase potential

[0060] The liquid phase potential distribution of lithium-ion batteries in the P2D model satisfies Ohm's law and concentrated solution theory, and its expression is shown in formula (4):

[0061]

[0062] Among them, φ e (x, t) is the liquid phase potential; c e (x, t) is the liquid ion concentration; R is the ideal gas constant; T is the temperature of the single cell; κ eff is the effective conductivity of the electrolyte; f is the ion activity coefficient of the electrolyte. tot (t) is the total current density in the electrode material.

[0063] 5) Lithium removal and lithium insertion

[0064] Lithium insertion and extraction occurs at the electrolyte / electrode interface. The Butler-Volmer equation can be used to calculate the current density generated by the lithium insertion and extraction reaction on the particle surface, as shown in formula (5):

[0065]

[0066] where i 0,int is the exchange current density of the electrode reaction during the lithium insertion and extraction process, η int,y is the reaction overpotential during lithium insertion and extraction. a,int is the anode transfer coefficient, α c,int is the cathode transfer coefficient.

[0067] i 0,int The expression of is shown in formula (6):

[0068]

[0069] Among them, k int is the reaction rate parameter of lithium removal or lithium insertion process; c e is the liquid phase ion concentration; c e,ref is the reference concentration of lithium ions in the electrolyte; cs,y,max is the maximum lithium insertion concentration of the positive electrode or negative electrode material; c s,y,surf It is the concentration of lithium ions on the surface of spherical particles of positive electrode or negative electrode material.

[0070] S2.2. Establishing a battery characteristics prediction model

[0071] Based on the power battery mechanism model of S2.1, a battery characteristic prediction model integrating metaheuristic algorithm and physical information neural network is established. The battery characteristic prediction model simulates the input power battery usage characteristic working condition to predict the positive and negative electrode ion concentration and electrolyte ion concentration of the power battery under given input working condition. Specifically, the prediction deviation of the power battery mechanism model is introduced into the penalty function of the physical information neural network algorithm, and a weight greater than the weight threshold is assigned, and then the metaheuristic algorithm is used for parameter optimization during the training process; the metaheuristic algorithm used is the particle swarm optimization-grey wolf optimization coupling algorithm, the chaos search algorithm, the particle swarm optimization-simulated annealing coupling algorithm, etc.; the physical information neural network algorithm is the long short-term memory neural network algorithm, the graph neural network algorithm, or the bidirectional long short-term memory neural network algorithm with attention mechanism, etc.

[0072] In this embodiment, Figure 4 As shown in the figure, a battery characteristic prediction model of particle swarm optimization-simulated annealing-gated recurrent neural network with attention mechanism is established based on the mechanical model of power battery. The prediction deviation of the mechanical model of power battery (the error between the model prediction and the actual sampling voltage) is introduced into the penalty function of the bidirectional recurrent neural network and assigned a weight of 75%. Then, particle swarm optimization-simulated annealing is used to optimize the training during the training process. Then, the characteristic working conditions of power battery usage are simulated for the trained model to predict the positive and negative electrodes and electrolyte ion concentrations of the power battery under given input working conditions.

[0073] The training parameters include: number of neural network layers, learning rate, gradient threshold, learning rate, slice size, and number of iterations.

[0074] The penalty function includes the prediction deviation of the power battery mechanism model and the residual of the neural network training, and adopts the following formula:

[0075] Γ total =ω F Γ F +ω D Γ D (7)

[0076] Where: total is the total penalty function, ω F is the weight of the neural network training residual, Γ F is the neural network training residual, ωD is the weight of the prediction deviation of the power battery mechanism rational model, Γ D It is the prediction deviation of the rational model of the power battery mechanism.

[0077] S2.3. Model prediction of available energy calculation and processing

[0078] like Figure 5 As shown, the SOC of the positive and negative electrodes is calculated on the cloud platform according to the predicted positive and negative electrode ion concentrations and the calibrated maximum ion concentrations of the positive and negative electrodes of the power battery, and then the positive electrode potential and the negative electrode potential are calculated according to the calculated SOC of the positive and negative electrodes and the calibrated positive and negative electrode SOC-OCV curves, and the electrolyte potential is calculated using the predicted electrolyte ion concentration. Then, on the basis of the positive electrode potential, the negative electrode potential, the electrolyte potential, and the ohmic voltage drop obtained according to the characteristic operating conditions of the power battery are subtracted respectively, so as to obtain the power battery terminal voltage, and the power battery power is obtained by the power battery terminal voltage and the current obtained according to the characteristic operating conditions of the power battery. When the power battery is continuously virtually discharged to reach the cut-off voltage, the power battery available energy is calculated by the power-time integral, which is used as the model to predict the available energy.

[0079] S3. Conduct offline testing experiments

[0080] Offline detection experiments may include capacity testing and dynamic charge and discharge testing to measure the actual test available energy of the power battery under the same initial conditions of charge and discharge current / power, temperature, and SOC initial value parameters (referred to as actual available energy, also known as real available energy). It may include the actual available energy of the battery system under given test conditions (such as NEDC, CLTC, etc.), or the actual available energy of the battery system under characteristic operating conditions, or the actual available energy of the battery cell under given test conditions (such as DST, FUDS, etc.).

[0081] In this embodiment, the battery operating condition established in step S1 is used as the test condition, and the available energy within a period of time is obtained through the battery testing equipment to measure the actual available energy of the battery under the same test conditions. The offline test adopts the NEDC condition, and the power battery condition is as follows: Figure 6 shown.

[0082] S4: Establish a data-driven model, or in other words, establish a correction model for estimating the available energy of power batteries.

[0083] A data-driven method is used on the cloud platform, with the model predicted available energy obtained in step S2 as input and the actual test available energy obtained in step S3 as output, to jointly establish a training set and a test set, and to construct a data-driven model. The data-driven model uses a data-driven method to establish a mapping relationship from the model predicted available energy to the actual test available energy, thereby correcting the model predicted available energy and achieving accurate available energy prediction. The data-driven model is a correction model for power battery available energy estimation. The data-driven method can use an extreme learning machine algorithm, a BP neural network, a recurrent neural network, an ELMAN neural network, or an LSTM neural network method, etc.

[0084] In this embodiment, the ELMAN neural network is used to establish a correction model for the available energy of the power battery. The ELMAN neural network is very sensitive to the data of the historical state. The internal feedback network can increase the processing capability of dynamic information and has strong generalization ability. The nonlinear state space expression of the Elman neural network is:

[0085]

[0086] Among them, y is the m-dimensional output node vector; x is the n-dimensional intermediate layer node unit vector; u is the r-dimensional input vector; x c is the n-dimensional feedback state vector, w 3 is the connection weight from the intermediate layer to the output layer; w 2 is the link weight from the input layer to the middle layer; w 1 is the connection weight from the receiving layer to the middle layer, g(*) is the transfer function of the output neuron, and f(*) is the transfer function of the middle layer neuron. The battery available energy data in steps S2 and S3 are used as feature data, where the input feature is: predicted available energy, and the output feature is: real available energy. The two available energies are used as column vectors to construct a sample matrix, and the training set and the sample set are divided. The training set is X%, and the sample set is 1-X%. In this embodiment, the training set is divided into 70% and the sample set is 30%. The ELMAN neural network is trained using the training set, and the output sum of squared errors is used as the indicator function. The BP algorithm is used to correct the weights. The correction model for power battery available energy estimation (i.e., the trained machine learning correction model-data-driven model) is obtained through the above training data:

[0087]

[0088] where y k (w) is the target observation vector, is the target input vector. In this embodiment, the Elman neural network is implemented in the algorithm engineering.

[0089] The results show that the correction model has an obvious correction effect. The model trained under the working condition of cloud data extraction is still effective under actual working conditions. The maximum error before the neural network correction is 2.7%. After the neural network correction, the average error is <1% and the maximum error is 1.6%. The calculation results are as follows Figure 7 shown.

[0090] The present invention also relates to a power battery available energy prediction system driven by knowledge-data fusion, which corresponds to the power battery available energy prediction method driven by knowledge-data fusion, and can be understood as a system for realizing the power battery available energy prediction method driven by knowledge-data fusion. The structure of the system is as follows: Figure 8 As shown, it includes a power battery usage characteristic condition establishment module, a model prediction available energy calculation module, an actual test available energy test module and a data-driven model construction module, all of which are located on the cloud platform. The power battery usage characteristic condition establishment module is connected to the model prediction available energy calculation module, and the model prediction available energy calculation module and the actual test available energy test module are both connected to the data-driven model construction module. Among them,

[0091] The power battery usage characteristic working condition establishment module obtains the cloud data uploaded to the cloud platform by the vehicle-side power battery system, extracts the usage working condition characteristics of the power battery by performing data analysis on the acquired cloud data, reduces the dimension of the high-dimensional features of the power battery usage working condition by using a data dimensionality reduction algorithm, performs correlation analysis on the reduced-dimensional features based on physical knowledge, and then classifies them by using a data-driven working condition feature classification method, calculates the probability map of the power battery jumping between different working conditions, and uses a working condition generation algorithm to establish the power battery usage characteristic working condition;

[0092] The model prediction available energy calculation module includes a power battery mechanism model construction module, a battery characteristic prediction model construction module and a model prediction available energy calculation processing module connected in sequence. The power battery mechanism model construction module establishes a power battery mechanism model according to the power battery use principle. The input conditions of the power battery mechanism model are the charge and discharge current / power, temperature, and SOC initial value parameters. The output conditions of the power battery mechanism model are the positive and negative electrode ion concentrations and the electrolyte ion concentrations of the power battery. The battery characteristic prediction model construction module establishes a battery characteristic prediction model that integrates a metaheuristic algorithm and a physical information neural network based on the power battery mechanism model. The battery characteristic prediction model simulates the input power battery use characteristic working conditions to predict the positive and negative electrode ions of the power battery under given input working conditions. concentration and electrolyte ion concentration; the model predicted available energy calculation processing module calculates the SOC of the positive and negative electrodes according to the predicted positive and negative electrode ion concentrations and the calibrated maximum ion concentrations of the positive and negative electrodes of the power battery, and then calculates the positive electrode potential and the negative electrode potential according to the calculated SOC of the positive and negative electrodes and the calibrated positive and negative electrode SOC-OCV curves, calculates the electrolyte potential using the predicted electrolyte ion concentration, and then subtracts the negative electrode potential, the electrolyte potential, and the ohmic voltage drop obtained according to the characteristic working conditions of the power battery from the positive electrode potential, thereby obtaining the power battery terminal voltage, and obtains the power battery power through the power battery terminal voltage and the current obtained according to the characteristic working conditions of the power battery, and calculates the power battery available energy through the power-time integral when the power battery is continuously virtually discharged to reach the cut-off voltage, as the model predicted available energy;

[0093] The actual test available energy test module carries out offline detection experiments, including capacity test and dynamic charge and discharge test, to measure the actual test available energy of the power battery under the initial conditions of the same charge and discharge current / power, temperature, and SOC initial value parameters;

[0094] The data-driven model construction module adopts a data-driven method, takes the model predicted available energy as input and the actual tested available energy as output, jointly establishes a training set and a test set, and constructs a data-driven model. The data-driven model establishes a mapping relationship from the model predicted available energy to the actual tested available energy through the data-driven method, thereby realizing the correction of the model predicted available energy.

[0095] Furthermore, the power battery usage characteristic operating condition establishment module obtains cloud data uploaded to the cloud platform by the vehicle-side power battery system, including voltage, current, time, temperature, SOC, charge / discharge status, battery output power, battery fault status, and vehicle speed information; extracts the power battery usage operating condition characteristics by performing data analysis on the acquired cloud data, including charging frequency, charge and discharge depth, acceleration time, deceleration time, average current, average voltage, average power, maximum power, minimum power, and power distribution curve characteristics; the data dimensionality reduction algorithm used includes principal component analysis or partial principal component analysis or singular value decomposition or autoencoder algorithm; the data-driven operating condition feature classification method used includes K-means clustering algorithm, random forest algorithm, and support vector machine algorithm; the operating condition generation algorithm is used to establish the power battery usage characteristic operating condition, and the operating condition generation algorithm used includes Monte Carlo simulation method, decision tree algorithm, and decision forest algorithm. The battery usage characteristic operating condition includes a current-time operating condition curve, and the curve length selects a characteristic operating condition of less than 1800s.

[0096] Furthermore, the battery characteristic prediction model construction module in the model prediction available energy calculation module establishes a battery characteristic prediction model that integrates a metaheuristic algorithm and a physical information neural network based on a power battery mechanistic model. The prediction deviation of the power battery mechanistic model is introduced into the penalty function of the physical information neural network algorithm, and a weight greater than a weight threshold is assigned, and then a metaheuristic algorithm is used for parameter optimization during the training process; the metaheuristic algorithm used is a particle swarm optimization-grey wolf optimization coupling algorithm, a chaos search algorithm, and a particle swarm optimization-simulated annealing coupling algorithm; the physical information neural network algorithm is a long short-term memory neural network algorithm, a graph neural network algorithm, or a bidirectional long short-term memory neural network algorithm with an attention mechanism.

[0097] The present invention relates to a method and system for predicting available energy of a power battery driven by knowledge-data fusion, which successively establishes a battery mechanistic model, a battery characteristic prediction model and a data-driven model of trusted data that conform to the basic principles of power batteries, realizes the prediction of battery available energy in an open-loop mode through the mechanistic model and the battery characteristic prediction model, and realizes the correction of available energy under the condition of trusted data through the data-driven model, thereby realizing the prediction of available energy of a power battery by integrating the knowledge-driven method and the data-driven method. After testing, the method and system of the present invention can improve the accuracy of battery available energy prediction and scenario coverage, and reduce the unexplainability of the data-driven model, thereby improving the problem of poor open-loop prediction accuracy of the mechanistic model, and has important application value in improving the accuracy of mileage estimation of new energy vehicles.

[0098] It should be noted that the above-described specific implementations can enable those skilled in the art to more fully understand the invention, but do not limit the invention in any way. Therefore, although this specification has described the invention in detail with reference to the drawings and embodiments, those skilled in the art should understand that the invention can still be modified or replaced by equivalents. In short, all technical solutions and improvements that do not deviate from the spirit and scope of the invention should be included in the protection scope of the patent for the invention.

Claims

1. A method for predicting available energy of power batteries driven by knowledge-data fusion, characterized in that: The steps include: S1. Obtain the cloud data uploaded to the cloud platform by the vehicle-side power battery system, extract the power battery usage condition characteristics by performing data analysis on the acquired cloud data on the cloud platform, use the data dimension reduction algorithm to reduce the dimension of the high-dimensional characteristics of the power battery usage condition, perform correlation analysis on the reduced-dimensional characteristics based on physical knowledge, and then use the data-driven working condition feature classification method to classify, calculate the probability map of the power battery jumping between different working conditions, and use the working condition generation algorithm to establish the power battery usage characteristic working condition; S2. On the cloud platform, a power battery mechanical model is established based on the power battery usage principle. The input conditions of the power battery mechanical model are the charge and discharge current / power, temperature, and SOC initial value parameters. The output conditions of the power battery mechanical model are the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery. Based on the power battery mechanical model, a battery characteristic prediction model integrating a metaheuristic algorithm and a physical information neural network is established. The battery characteristic prediction model simulates the input power battery usage characteristic working conditions to predict the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery under given input working conditions. On the cloud platform, the predicted positive and negative electrode ion concentrations and the calibrated positive and negative electrode ion concentrations of the power battery are used to predict the positive and negative electrode ion concentrations and electrolyte ion concentrations of the power battery. The SOC of the positive and negative electrodes is calculated by the maximum ion concentration of the negative electrode, and then the positive electrode potential and the negative electrode potential are calculated according to the calculated SOC of the positive and negative electrodes and the calibrated positive and negative electrode SOC-OCV curves, and the electrolyte potential is calculated using the predicted electrolyte ion concentration, and then the negative electrode potential, the electrolyte potential, and the ohmic voltage drop obtained according to the characteristic working conditions of the power battery are respectively subtracted from the positive electrode potential to obtain the power battery terminal voltage, and the power battery power is obtained by the power battery terminal voltage and the current obtained according to the characteristic working conditions of the power battery, and the power battery available energy is calculated by the power-time integral when the power battery is continuously virtually discharged to reach the cut-off voltage, as a model to predict the available energy; S3. Conduct offline testing experiments, including capacity testing and dynamic charge and discharge testing, to measure the actual test available energy of the power battery under the same initial conditions of charge and discharge current / power, temperature, and SOC initial value parameters; S4. A data-driven approach is adopted on the cloud platform. The model predicted available energy obtained in step S2 is used as input and the actual test available energy obtained in step S3 is used as output. A training set and a test set are jointly established to construct a data-driven model. The data-driven model establishes a mapping relationship from the model predicted available energy to the actual test available energy through the data-driven approach, thereby realizing the correction of the model predicted available energy.

2. The method for predicting available energy of power batteries driven by knowledge-data fusion according to claim 1 is characterized in that: Step S1 obtains the cloud data uploaded to the cloud platform by the vehicle-side power battery system, including voltage, current, time, temperature, SOC, charge / discharge status, battery output power, battery fault status, and vehicle speed information; extracts the power battery usage operating condition characteristics by analyzing the acquired cloud data on the cloud platform, including charging frequency, charge and discharge depth, acceleration time, deceleration time, average current, average voltage, average power, maximum power, minimum power, and power distribution curve characteristics; uses a condition generation algorithm to establish a power battery usage characteristic condition, wherein the battery usage characteristic condition includes a current-time condition curve, and the curve length selects a characteristic condition of less than 1800s.

3. The method for predicting available energy of power batteries driven by knowledge-data fusion according to claim 2 is characterized in that: The data dimension reduction algorithms used in step S1 include principal component analysis or partial principal component analysis or singular value decomposition or autoencoder algorithm; the data-driven working condition feature classification methods used include K-means clustering algorithm, random forest algorithm, support vector machine algorithm; the working condition generation algorithms used include Monte Carlo simulation method, decision tree algorithm, and decision forest algorithm.

4. The method for predicting available energy of a power battery driven by knowledge-data fusion according to any one of claims 1 to 3, characterized in that: In step S2, a power battery mechanistic model including an equivalent circuit model, an electrochemical model or a heterogeneous model is established on a cloud platform according to the use principle of the power battery. The equivalent circuit model is a first-order equivalent circuit model, a second-order equivalent circuit model, a fractional-order equivalent circuit model or other electrochemically reduced-order equivalent circuit model. The electrochemical model is a pseudo-two-dimensional electrochemical model, a single particle model or a three-dimensional electrochemical model. The pseudo-two-dimensional electrochemical model integrates solid-phase lithium ion concentration calculation, liquid-phase lithium ion concentration calculation, solid-phase potential calculation, liquid-phase potential calculation and lithium deintercalation.

5. The method for predicting available energy of power batteries driven by knowledge-data fusion according to claim 4 is characterized in that: Step S2 establishes a battery characteristic prediction model that integrates a metaheuristic algorithm and a physical information neural network based on a power battery mechanism model. The prediction deviation of the power battery mechanism model is introduced into the penalty function of the physical information neural network algorithm, and a weight greater than the weight threshold is assigned. Then, a metaheuristic algorithm is used for parameter optimization during training. The metaheuristic algorithms used are a particle swarm optimization-grey wolf optimization coupling algorithm, a chaos search algorithm, and a particle swarm optimization-simulated annealing coupling algorithm. The physical information neural network algorithm is a long short-term memory neural network algorithm, a graph neural network algorithm, or a bidirectional long short-term memory neural network algorithm with an attention mechanism.

6. The method for predicting available energy of power batteries driven by knowledge-data fusion according to claim 5 is characterized in that: In step S3, offline detection experiments are carried out to measure the actual test available energy of the power battery under the same initial conditions of charge and discharge current / power, temperature, and SOC initial value parameters, including the actual test available energy of the power battery system under NEDC or CLTC test conditions, or the actual test available energy of the power battery system under characteristic operating conditions, or the actual test available energy of the power battery cell under DST and FUDS test conditions.

7. The method for predicting available energy of a power battery driven by knowledge-data fusion according to any one of claims 1 to 3, characterized in that: The data-driven model constructed in step S4 establishes a mapping relationship from model-predicted available energy to actual tested available energy through a data-driven method, wherein the data-driven method adopts an extreme learning machine algorithm, a BP neural network, a recurrent neural network, an ELMAN neural network, or an LSTM neural network method.

8. A power battery available energy prediction system driven by knowledge-data fusion, characterized in that: It includes a power battery use characteristic working condition establishment module, a model prediction available energy calculation module, an actual test available energy test module and a data-driven model construction module, all of which are located on the cloud platform. The power battery use characteristic working condition establishment module is connected to the model prediction available energy calculation module, and the model prediction available energy calculation module and the actual test available energy test module are both connected to the data-driven model construction module; The power battery usage characteristic working condition establishment module obtains the cloud data uploaded to the cloud platform by the vehicle-side power battery system, extracts the usage working condition characteristics of the power battery by performing data analysis on the obtained cloud data, reduces the dimension of the high-dimensional features of the power battery usage working condition by using a data dimensionality reduction algorithm, performs correlation analysis on the reduced-dimensional features based on physical knowledge, and then classifies them by using a data-driven working condition feature classification method, calculates the probability map of the power battery jumping between different working conditions, and uses a working condition generation algorithm to establish the power battery usage characteristic working condition; The model prediction available energy calculation module includes a power battery mechanism model construction module, a battery characteristic prediction model construction module and a model prediction available energy calculation processing module connected in sequence. The power battery mechanism model construction module establishes a power battery mechanism model according to the power battery use principle. The input conditions of the power battery mechanism model are the charge and discharge current / power, temperature, and SOC initial value parameters. The output conditions of the power battery mechanism model are the positive and negative electrode ion concentrations and the electrolyte ion concentrations of the power battery. The battery characteristic prediction model construction module establishes a battery characteristic prediction model that integrates a metaheuristic algorithm and a physical information neural network based on the power battery mechanism model. The battery characteristic prediction model simulates the input power battery use characteristic working conditions to predict the positive and negative electrode ions of the power battery under the given input working conditions. concentration and electrolyte ion concentration; the model predicted available energy calculation processing module calculates the SOC of the positive and negative electrodes according to the predicted positive and negative electrode ion concentrations and the calibrated maximum ion concentrations of the positive and negative electrodes of the power battery, and then calculates the positive electrode potential and the negative electrode potential according to the calculated SOC of the positive and negative electrodes and the calibrated positive and negative electrode SOC-OCV curves, calculates the electrolyte potential using the predicted electrolyte ion concentration, and then subtracts the negative electrode potential, the electrolyte potential, and the ohmic voltage drop obtained according to the characteristic working conditions of the power battery from the positive electrode potential, thereby obtaining the power battery terminal voltage, and obtains the power battery power through the power battery terminal voltage and the current obtained according to the characteristic working conditions of the power battery, and calculates the power battery available energy through the power-time integral when the power battery is continuously virtually discharged to reach the cut-off voltage, as the model predicted available energy; The actual test available energy test module carries out offline detection experiments, including capacity test and dynamic charge and discharge test, to measure the actual test available energy of the power battery under the initial conditions of the same charge and discharge current / power, temperature, and SOC initial value parameters; The data-driven model construction module adopts a data-driven method, takes the model predicted available energy as input and the actual tested available energy as output, jointly establishes a training set and a test set, and constructs a data-driven model. The data-driven model establishes a mapping relationship from the model predicted available energy to the actual tested available energy through the data-driven method, thereby realizing the correction of the model predicted available energy.

9. The power battery available energy prediction system driven by knowledge-data fusion according to claim 8 is characterized in that: The power battery usage characteristic operating condition establishment module obtains cloud data uploaded to the cloud platform by the vehicle-side power battery system, including voltage, current, time, temperature, SOC, charge / discharge status, battery output power, battery fault status, and vehicle speed information; extracts the power battery usage operating condition characteristics by performing data analysis on the acquired cloud data, including charging frequency, charge and discharge depth, acceleration time, deceleration time, average current, average voltage, average power, maximum power, minimum power, and power distribution curve characteristics; the data dimensionality reduction algorithm adopted includes principal component analysis or partial principal component analysis or singular value decomposition or autoencoder algorithm; the data-driven operating condition feature classification method adopted includes K-means clustering algorithm, random forest algorithm, and support vector machine algorithm; the operating condition generation algorithm is adopted to establish the power battery usage characteristic operating condition, and the operating condition generation algorithm adopted includes Monte Carlo simulation method, decision tree algorithm, and decision forest algorithm. The battery usage characteristic operating condition includes current-time operating condition curve, and the curve length selects characteristic operating conditions of less than 1800s.

10. The power battery available energy prediction system driven by knowledge-data fusion according to claim 8 or 9, characterized in that: The battery characteristic prediction model building module in the model prediction available energy calculation module establishes a battery characteristic prediction model that integrates a meta-heuristic algorithm and a physical information neural network based on a power battery mechanism model. In the penalty function of the physical information neural network algorithm, a prediction deviation of the power battery mechanism model is introduced, and a weight greater than a weight threshold is assigned, and then a meta-heuristic algorithm is used for parameter optimization during training; the meta-heuristic algorithm used is a particle swarm optimization-grey wolf optimization coupling algorithm, a chaos search algorithm, and a particle swarm optimization-simulated annealing coupling algorithm; the physical information neural network algorithm is a long short-term memory neural network algorithm, a graph neural network algorithm, or a bidirectional long short-term memory neural network algorithm with an attention mechanism.

Citation Information

Patent Citations

  • Electric vehicle lithium ion battery lithium precipitation diagnostic method with data drive

    CN109143083A

  • Management method of power battery module based on cloud control technology

    CN110492186A

  • Battery health state estimation method fusing mechanism and data driving model

    CN116643196A

  • Method and system for calculating maximum available energy of lithium ion battery

    CN116956667A

  • Method and apparatus for predicting the available energy of a battery

    US20020193953A1