Battery SOC (State of Charge) estimation method, system and equipment based on variational Bayes and multi-kernel entropy UKF (Unscented Kalman Filter)

Abstract

The invention belongs to the technical field of battery detection, and relates to a variational Bayes and multi-kernel entropy UKF-based battery SOC estimation method, system and device. According to the method, the state-space equation is constructed according to the internal electrochemical reaction mechanism of the battery, so that the relationship between the internal state variables of the battery is described, and a necessary mathematical model is provided for subsequent filtering estimation. Constructing an unscented Kalman filter based on a variational Bayesian approximation method and a maximum majority kernel correlation entropy criterion to obtain a newly constructed unscented Kalman filter; and the precision and the real-time performance of SOC estimation can be improved. And estimating the state of the battery according to the state-space equation in combination with the newly constructed unscented Kalman filter. A new unscented Kalman filter is constructed by introducing a variational Bayesian approximation method and a maximum majority kernel correlation entropy criterion, and battery state estimation is performed based on the filter, so that high-precision and robustness estimation of the SOC of the battery is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery detection, and relates to a battery SOC estimation method, system and device based on variational Bayes and multi-core entropy UKF. Background Art

[0002] Traditional coal, oil, natural gas and other traditional energy sources have always been the main energy sources of contemporary society. However, with the continuous increase in global energy demand, traditional energy resources are facing an increasingly serious risk of depletion, which poses a severe challenge to the sustainable development of human society. In order to alleviate the pressure of energy shortage and promote the development of society in a more sustainable direction, it is particularly urgent to actively develop energy-saving and emission-reduction technologies and continue to explore diversified energy options.

[0003] With the widespread promotion of new energy around the world, lithium-ion battery-related technologies and industries are constantly developing. Due to the advantages of lithium-ion batteries such as high output voltage, long cycle life, high energy density, low self-discharge rate, long life and high operating voltage, they have been widely used in portable electronic devices, electric vehicles and energy storage.

[0004] Despite the rapid development of battery technology, it still faces problems such as overcharging and over-discharging, which may reduce battery life and cause fire and explosion. Therefore, it is crucial to optimize the performance of the battery management system to ensure battery safety, reliability and high cost-effectiveness. The state of charge is one of the basic components of the battery management system. It reflects the relationship between the available charge stored in the battery and its maximum capacity, and is crucial for monitoring the operating status of the battery system. Accurate battery nuclear power state estimation helps to achieve efficient energy management of electric vehicles and ensure the long life and safety of the basic battery system. In the battery electrochemical model, the battery state of charge is directly determined by the surface lithium ion concentration of the electrode particles, which is defined as follows:

[0005] in, Represents the average lithium ion concentration of the positive or negative electrode particles, Indicates the ion concentration when the battery state of charge is 100%, Indicates the ion concentration when the battery state of charge is 0%. Represents the negative and positive terminals of the battery.

[0006] In recent years, the introduction of a series of methods in the field of SOC estimation has greatly promoted the progress of battery detection technology. However, due to the complex electrochemical kinetics and multi-physics coupling characteristics of lithium-ion batteries, accurate SOC estimation remains a major challenge. Existing SOC estimation methods can be mainly divided into the following categories: direct methods and data-driven methods.

[0007] Direct methods use measured physical quantities such as current, voltage, and impedance to estimate SOC by constructing mathematical equations or relationships between SOC and these measurable physical quantities. However, due to the deviation in the known initial SOC and the noise interference in the measured current, the reliability of this method will be affected, resulting in inaccurate SOC estimation, thus limiting its application in actual batteries.

[0008] Data-driven methods based on machine learning, including support vector machines, convolutional neural networks, and long short-term memory neural networks, have shown excellent performance in battery SOC estimation. By training machine learning algorithms offline, these methods establish a mapping relationship between external battery data and SOC, and achieve SOC estimation without explicitly considering the internal battery mechanism. However, since the resulting mapping relationship is independent of the complex electrochemical reactions within the battery, the accuracy of SOC estimation is reduced.

[0009] In summary, the existing battery SOC estimation method has the problem of low accuracy. Summary of the invention

[0010] The purpose of the present invention is to provide a battery SOC estimation method, system and device based on variational Bayes and multi-core entropy UKF to solve the technical problem of low accuracy of existing SOC estimation methods. The present invention can achieve accurate estimation of battery SOC.

[0011] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a battery SOC estimation method based on variational Bayes and multi-core entropy UKF, comprising the following steps: Construct the state space equation according to the internal electrochemical reaction mechanism of the battery; An unscented Kalman filter is constructed based on a variational Bayesian approximation method and a maximum multi-core correlation entropy criterion, and a newly constructed unscented Kalman filter is obtained; Battery state estimation is performed based on the state-space equation combined with a newly constructed unscented Kalman filter.

[0012] In a second aspect, the present invention provides a battery SOC estimation system based on variational Bayes and multi-core entropy UKF, comprising: State space equation acquisition module: used to construct the state space equation according to the internal electrochemical reaction mechanism of the battery; Filter construction module: used to construct an unscented Kalman filter based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion to obtain a newly constructed unscented Kalman filter; Battery state estimation module: used to estimate the battery state based on the state space equation combined with the newly constructed unscented Kalman filter.

[0013] In a third aspect, the present invention provides an electronic device, comprising: a processor; a memory for storing computer program instructions; and steps for implementing a battery SOC estimation method based on variational Bayes and multi-core entropy UKF when executing the computer program.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention constructs a state-space equation based on the internal electrochemical reaction mechanism of the battery, thereby describing the relationship between the internal state variables of the battery and providing the necessary mathematical model for subsequent filtering estimation. An unscented Kalman filter is constructed based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion to obtain a newly constructed unscented Kalman filter; this is beneficial to improving the accuracy and real-time performance of SOC estimation. Battery state estimation is performed based on the state-space equation combined with the newly constructed unscented Kalman filter. A new unscented Kalman filter is constructed by introducing the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion, and battery state estimation is performed based on the filter, thereby achieving high-precision and robust estimation of the battery SOC.

[0015] 2. The system of the present invention includes: a state space equation acquisition module, a filter construction module and a battery state estimation module. The state space equation acquisition module is used to construct the state space equation according to the internal electrochemical reaction mechanism of the battery; the filter construction module is used to construct an unscented Kalman filter based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion to obtain a newly constructed unscented Kalman filter; the battery state estimation module is used to estimate the battery state according to the state space equation combined with the newly constructed unscented Kalman filter. The various modules cooperate with each other to achieve accurate estimation of the battery SOC.

[0016] 3. The electronic device of the present invention can also achieve accurate estimation of battery SOC. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is an overall flow chart of an embodiment of the present invention; Figure 2 is a flow chart of a method according to an embodiment of the present invention; Figure 3 is a flow chart of the method of the present invention; Figure 4 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0019] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] The present invention is further described in detail below in conjunction with the accompanying drawings: See also Figure 3 The present invention discloses a battery SOC estimation method based on variational Bayes and multi-core entropy UKF, comprising the following steps: The state space equation is constructed according to the internal electrochemical reaction mechanism of the battery to describe the relationship between the internal state variables of the battery and provide the necessary mathematical model for subsequent filtering estimation.

[0021] An unscented Kalman filter is constructed based on a variational Bayesian approximation method and a maximum multi-core correlation entropy criterion to obtain a newly constructed unscented Kalman filter; the variational Bayesian method of the present invention can make the filter adaptive; the maximum multi-core correlation entropy criterion can improve the robustness of the filter, which is beneficial to improving the accuracy and real-time performance of SOC estimation.

[0022] The battery state is estimated based on the state space equation combined with a newly constructed unscented Kalman filter. A new unscented Kalman filter is constructed by introducing the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion, and the battery state is estimated based on the filter, thereby achieving high-precision and robust estimation of the battery SOC.

[0023] Embodiment 1: See also Figure 2 and Figure 3This embodiment discloses a battery SOC estimation method based on variational Bayes and multi-core entropy UKF, comprising the following steps: S1, construct the state space equation according to the internal electrochemical reaction mechanism of the battery, as follows: The battery electrochemical reaction mechanism equation is constructed based on the average lithium ion concentration of the solid phase, the lithium ion concentration on the solid phase surface, the pore wall flow, the solid phase particle radius and the diffusion coefficient of lithium ions in the solid phase particles; The state space equation is constructed according to the battery electrochemical reaction mechanism equation.

[0024] Preferably, the state space equation is as follows:

[0025]

[0026] in, is the state quantity; is the battery charge state; , , and is the intermediate variable corresponding to the positive and negative poles; is the matrix transpose symbol; is the voltage, i.e. the observed quantity; is the positive reference voltage; is the negative reference voltage; for The positive electrode ion concentration proportional coefficient at the moment; for The negative electrode ion concentration proportional coefficient at the moment; For time; is a function of current.

[0027] S2, based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion, an unscented Kalman filter is constructed to obtain a newly constructed unscented Kalman filter, as follows: Get the estimated value of the state quantity at the current moment; Perform an untraceable transformation on the estimated value of the state quantity at the current moment to obtain a set of sampling points corresponding to the state quantity at the current moment; According to a set of sampling points corresponding to the state quantity at the current moment, obtain the prior estimate of the state quantity at the next moment; Obtain a priori estimated value of the observation quantity according to the priori estimated value of the state quantity, and obtain the observation noise covariance matrix according to the priori estimated value of the observation quantity; The observation noise covariance matrix is ​​iteratively updated according to the variational Bayesian approximation method, as follows:

[0028] in, is the observation noise covariance matrix, is a free parameter, is the dimension of the state quantity, is the inverse scaling matrix, is the number of identifiers, is the current time step.

[0029] The fixed-point iterative algorithm is used to iteratively update the prior estimate of the state quantity to obtain the posterior estimate of the state quantity, as follows: Obtaining the covariance matrix of system prediction according to the sampling point of the current untraceable transformation, the state quantity at the current moment, and the prior estimation value of the observation quantity at the current moment; Obtaining a measurement slope matrix based on the current a priori covariance matrix and the system predicted covariance matrix; Construct a batch regression model based on the current state, observations, posterior estimates of observations, measurement slope matrix, joint noise matrix, and n-order identity matrix; A regression equation is obtained according to the batch regression model; Obtain the posterior estimate of the state quantity based on the regression equation.

[0030] The posterior covariance matrix is ​​obtained according to the posterior estimate of the state quantity combined with the observation noise covariance matrix, and the process parameters of the variational Bayesian method are updated to obtain the newly constructed unscented Kalman filter, as follows: Obtaining a posterior covariance matrix according to the measurement slope matrix, the observation noise covariance matrix and the prior covariance matrix; The process parameters of the variational Bayes method are updated based on the prior inverse scaling matrix, the prior estimates of the observations, and the actual observations.

[0031] S3, the battery state is estimated according to the state space equation combined with the newly constructed unscented Kalman filter, as follows: Combine actual battery data and state-space equations to obtain a priori estimation results of the battery state of charge; The battery state of charge estimation value is obtained by processing the battery's prior estimation results through the newly constructed unscented Kalman filter.

[0032] The variational Bayesian method of the present invention can make the filter adaptive. Adaptability means that the noise covariance matrix can be adaptively estimated. When the working environment of the battery is very complex and is affected by the noise whose covariance changes over time, this property can improve the accuracy of the estimation. The maximum multi-core correlation entropy criterion can improve the robustness of the filter. Robustness means that when there are some abnormally large values ​​among the observed values, the filter can be unaffected by the abnormal values ​​and give a more accurate SOC estimation value.

[0033] Embodiment 2: See also Figure 1 and Figure 2 This embodiment discloses a battery SOC estimation method based on variational Bayes and multi-core entropy UKF, comprising the following steps: S1. Establishing a state-space equation according to the internal electrochemical reaction mechanism of the battery. The state-space equation is constructed based on the electrochemical reaction mechanism inside the lithium-ion battery, including the simplification of the solid-phase diffusion process and the liquid-phase diffusion process inside the battery.

[0034] The battery state of charge calculation formula is as follows:

[0035] in, represents the average lithium ion concentration of the electrode solid phase particles, Indicates the ion concentration when the battery state of charge is 100%, Indicates the ion concentration when the battery state of charge is 0%. Represents the negative and positive poles of the battery. For batteries State of charge at the moment.

[0036] Inside the battery particles, the solid phase lithium ion concentration According to Fick's second law:

[0037] in, Indicates time, Indicates the battery thickness, represents the solid phase particle radius, is the diffusion coefficient of lithium ions in the solid particles, is the solid phase lithium ion concentration.

[0038] The relevant parameters in the state-space equation construction process are first converted into dimensionless form to ensure that the model is applicable to batteries with different materials and different production processes.

[0039] Due to the battery's state of charge and terminal voltage and the average lithium ion concentration in the battery solid phase and solid surface lithium ion concentration And the following formula can be derived:

[0040]

[0041]

[0042] in, is the time step in dimensionless form, is the difference between the average lithium ion concentration in the solid phase and the lithium ion concentration on the solid surface in dimensionless form, is the dimensionless form of the hole wall flow, is the average lithium ion concentration in the solid phase in dimensionless form, is the solid surface lithium ion concentration in dimensionless form; Therefore, the equation based on the battery electrochemical reaction mechanism can be constructed as follows:

[0043]

[0044]

[0045]

[0046] in, is the hole wall flow; is the radius of the solid phase particle; is the time interval, usually 1s; , , and It is the parameter of the battery electrochemical reaction mechanism equation, which can be calculated by the optimization algorithm; and are all intermediate variables. For the time steps, For the time steps, is the diffusion coefficient of lithium ions in the solid particles, is the base of natural logarithm, is the lithium ion concentration on the solid surface.

[0047] According to the above results, the state quantity and observation quantity are as follows:

[0048]

[0049]

[0050]

[0051] in, is the positive reference voltage, is the negative reference voltage, for The positive electrode ion concentration proportional coefficient at the moment, for The negative electrode ion concentration proportional coefficient at the moment, For time, is a function of current, is the state quantity, is the battery charge state, , , and is the intermediate variable corresponding to the positive and negative poles, is the matrix transpose symbol, is the voltage, i.e. the observed quantity, is the lithium ion concentration on the negative electrode surface; is the maximum lithium ion concentration of the positive electrode; is the lithium ion concentration on the positive electrode surface; is the maximum lithium ion concentration at the negative electrode.

[0052] S2, based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion, an unscented Kalman filter based on variational Bayesian and multi-core maximum correlation entropy is constructed, as follows: S21, perform an untraceable transformation on the estimated value of the state quantity at the current moment to obtain a set of sampling points corresponding to the state quantity at the current moment. The untraceable transformation process is as follows: When performing an unscented transformation, the sampling points are first selected as follows:

[0053]

[0054]

[0055] in, is the state quantity at the previous moment, is the state dimension, is a free parameter, is the posterior covariance matrix of the previous moment, For the The sampling points of the unscented transformation are is the sampling point of the 0th unscented transform, Refers to the moment. The sampling points of the unscented transformation are from 0 to , that is sampling points.

[0056] S22, calculate the prior estimates of the state quantity, observation quantity and covariance matrix at the next moment according to the sampling points, as follows:

[0057]

[0058]

[0059]

[0060]

[0061] in, is the nonlinear governing equation, and is the weight value of the unscented Kalman filter, is the state dimension, is a free parameter, is the scaling parameter, is the state distribution parameter, is the process noise covariance matrix, , , and are the weight values ​​of the unscented Kalman filter with the identification of the sampling points added. is the matrix transpose symbol, is the current moment posterior covariance matrix, is the sampling point of the untraceable transformation at the current moment, is the state quantity of the current moment.

[0062] S23, according to the prior estimation value of the state quantity, the prior estimation of the observation quantity and the corresponding covariance matrix of the state quantity and the observation quantity are calculated, and the calculated prior estimation of the observation quantity will be used to calculate the covariance matrix of the state quantity and the observation quantity, and the covariance matrix of the state quantity and the observation quantity is the observation noise covariance matrix; the observation noise covariance matrix will be used to calculate the measurement slope matrix mentioned later, as follows:

[0063]

[0064]

[0065]

[0066]

[0067] In this step, sampling points are first selected, and then the prior estimate of the observed quantity is calculated through untraceable transformation, and the relevant covariance matrix is ​​calculated based on the prior estimate results of the state quantity and the observed quantity.

[0068] in, is the prior estimate of the observation at the current moment, is the nonlinear observation equation, is the covariance matrix of system prediction, is the current state quantity, The sampling point of the state quantity at the current moment, For the The sampling points of the unscented transformation are Take a positive integer, is a free parameter, is the current moment posterior covariance matrix, and To add the weight value of the unscented Kalman filter with the identification of the sampling point, is the prior estimate of the observed quantity, is the matrix transpose symbol.

[0069] S24, iteratively update the observation noise covariance matrix according to the variational Bayesian approximation method, as follows:

[0070] in, is the observation noise covariance matrix, is a free parameter, is the state dimension, is the inverse scaling matrix, is the number of identifiers, is the current time step, i.e., the time step identifier; This formula is used to iteratively update the noise covariance matrix according to the variational Bayes method. In each update, the noise covariance matrix is multiplied by one and the free parameter The constant coefficient can be regarded as the prior information in the noise model, which is used to balance the influence between prior knowledge and data observation.

[0071] S25, iteratively update the state quantity using a fixed-point iteration algorithm to obtain a posterior estimation value of the state quantity, as follows: Build a batch regression model;

[0072]

[0073]

[0074]

[0075] in, is the covariance matrix of system prediction, is the prior state quantity at the current moment, is the measured slope matrix, is the joint noise matrix, is the joint noise matrix covariance, is the actual observed value, is the posterior estimate of the observed quantity, is the n-order identity matrix, is the posterior estimate of the state quantity, is the inverse of the prior covariance matrix, is the observation noise, is the process noise, is the observation noise covariance matrix, is the Cholesky decomposition value of the joint noise matrix covariance, for The transpose of is the transpose of the joint noise matrix, is the covariance matrix predicted by the system; In the above batch model, left multiplication , we can get the following regression equation:

[0076] in, and They are all intermediate process quantities. is the residual vector, is the inverse of the Cholesky decomposition of the joint noise matrix covariance; Solving the corresponding optimization problem can obtain the posterior estimate of the state quantity:

[0077] in: and A diagonal matrix with diagonal elements of,

[0078]

[0079]

[0080] Update state variables Until the following inequality holds:

[0081] in, is a very small positive number, For the The posterior estimate of the cycle, To solve the The posterior estimate of the cycle, For the The posterior estimate of the cycle, is the current time step, is the loop step index, that is, the current loop number. Step , a loop iteration solution is required, indicating that the current loop iteration solution is Second cycle, is the matrix transpose symbol, is a diagonal matrix, and Both are cycles The diagonal matrix of times, is the Gaussian kernel weight value, is the Gaussian kernel width, is the Gaussian kernel function, is the residual vector, is the Gaussian kernel weight value, for No. OK, is the first OK, In this time step, there will be another loop to iteratively calculate the posterior estimate. and is a constant; Represents the matrix diagonal elements; It means the Gaussian kernel.

[0082] S26, calculate the posterior covariance matrix according to the posterior estimated value of the state quantity, and update the process parameters of the variational Bayesian method. The process parameters of the variational Bayesian method are used to control the update of the observation noise covariance matrix, which are as follows:

[0083]

[0084]

[0085]

[0086] in, is the observation noise covariance matrix, is the actual observed value, is the posterior covariance matrix, is the measured slope matrix, is the posterior estimate of the state quantity, For the time steps, is the matrix transpose symbol, is the inverse of the prior covariance matrix, For the The inverse scale matrix obtained by loop update is is the prior inverse scaling matrix, For the The sampling points of the unscented transformation are is the prior estimate of the observed quantity, To add the weight value of the unscented Kalman filter with the identification of the sampling point, is the number of sampling points, and Both indicated that The number of tokens for the cyclic sequence within this time step, The time step is The new factor when is a diagonal matrix; are the elements inside the diagonal matrix, Take 1~ , is the observation dimension, which takes a positive integer.

[0087] S3. Use the constructed filter to estimate the battery state, as follows: The actual battery data containing information such as current and voltage is input into the battery state space equation constructed based on the electrochemical reaction mechanism. The prior estimation results are processed by the filtering algorithm to eliminate the influence of complex and variable noise, and a more accurate battery state of charge estimation value is obtained. In addition, the results of the battery state of charge estimation will be evaluated using two indicators: mean square error and average mean square error.

[0088] In actual use, the model of the present invention is embedded in a battery state of charge estimation device, which receives data such as current, voltage and temperature from a battery sensor, and then estimates the battery state of charge value at the current moment using the model of the present invention. The present invention can realize real-time and accurate prediction of the battery state of charge.

[0089] Embodiment three: This embodiment discloses a battery SOC estimation method based on variational Bayesian and multi-core entropy UKF, such as Figure 1 and Figure 2 As shown, the following steps are included: S1. Select sampling points by performing traceless transformation based on the initial current and voltage, state quantity and initial value information of the covariance matrix:

[0090]

[0091]

[0092] in, is the state quantity at the previous moment, is the state dimension, is a free parameter, is the posterior covariance matrix of the previous moment.

[0093] Calculate the prior estimates of the state quantity and covariance matrix at the current moment:

[0094]

[0095]

[0096]

[0097]

[0098] in, is the relevant weight value, is the state dimension, is a free parameter, is the scaling parameter, is the state distribution parameter.

[0099] S2. Calculate the scale parameter and covariance matrix of variational Bayesian estimation:

[0100]

[0101] in, is the forgetting factor, the matrix Usually set to .

[0102] S3. Perform measurement updates, calculate the prior estimates of the observed quantities, the covariance matrix and variance matrix of the state quantities and the observed quantities:

[0103]

[0104]

[0105]

[0106]

[0107]

[0108] In this step, sampling points are first selected, and then the prior estimate of the observed quantity is calculated through untraceable transformation, and the relevant covariance matrix is ​​calculated based on the prior estimate results of the state quantity and the observed quantity.

[0109] S4. The noise covariance matrix is ​​estimated using the variational Bayesian method, and the state estimation is performed based on the maximum multi-core correlation entropy criterion: set up , , , and threshold parameters , perform iterative updates according to the following steps:

[0110]

[0111] Updated by fixed-point iterative algorithm :

[0112] in:

[0113]

[0114]

[0115] Update state variables Until the following inequality holds:

[0116] in, is a very small positive number.

[0117] Update the covariance matrix according to the above iterative results:

[0118]

[0119]

[0120]

[0121] in, is the actual observed value, The time step is The new factor of time.

[0122] S5. Full cycle state of charge estimation; The battery state of charge value at the next moment can be estimated through the above steps, and the battery state of charge value within the battery operation cycle can be accurately estimated by iteratively repeating the above steps.

[0123] The specific testing method of this embodiment is as follows: 1. Test environment: The test software environment for battery state of charge estimation based on variational Bayes and maximum multi-core correlation entropy unscented Kalman filtering proposed in the present invention is Matlab 2023a, the CPU is Intel core i5-13600KF 3.50 GHz, the memory is 32GB, and the hard disk storage is 1TB.

[0124] 2. Experimental data: The data used in the present invention comes from the SLPB 75106100 high-energy nickel-manganese-cobalt battery, which has a nominal voltage of 3.7 V, a nominal capacity of 7.5 Ah, and an operating voltage range of 2.7 V to 4.2 V.

[0125] 3. Benchmark method: In order to evaluate the effectiveness of the present invention, seven benchmark methods were selected for experiments, namely, unscented Kalman filter (UKF), variational Bayesian unscented Kalman filter (VBUKF), Student-T unscented Kalman filter (STUKF), maximum correlated entropy unscented Kalman filter (MCUKF), minimum error entropy unscented Kalman filter with benchmark (MEEF-UKF), maximum multi-core correlated entropy unscented Kalman filter (MMKC-UKF) and unscented Kalman filter based on variational Bayes and maximum multi-core correlated entropy (VBMMKC-UKF).

[0126] 4. Experimental setup: In order to verify and evaluate the impact of non-Gaussian noise signals on the battery state of charge estimation results, four groups of experiments were set up to verify the SOC estimation results under different noise conditions and different working conditions.

[0127] The experimental conditions include: pulse test and synthetic cycle load current (SLCB) test. The four noise conditions set include: Gaussian noise, pulse noise, bimodal Gaussian noise and bimodal Gaussian pulse noise. The root mean square error and average root mean square error are used as experimental indicators.

[0128] 5. Experimental results: The comparison results of the experimental effects of the existing mainstream solutions and the present invention are shown in Table 1, Table 2, Table 3 and Table 4. From the results, the prediction accuracy of the battery state of charge estimation method based on variational Bayes and maximum relevant entropy unscented Kalman filtering of the present invention is higher than that of other methods.

[0129] Table 1. Average root mean square error of SOC estimation results under Gaussian noise:

[0130] Table 2. Average root mean square error of SOC estimation results under impulse noise:

[0131] Table 3. Average root mean square error of SOC estimation results under bimodal Gaussian noise:

[0132] Table 4. Average root mean square error of SOC estimation results under bimodal Gaussian pulse noise:

[0133] Due to the complexity and variability of the actual operation process of the battery, and in order to improve the performance of the battery state of charge estimation, the present invention estimates the battery SOC through an unscented Kalman filter based on variational Bayes and maximum multi-core correlation entropy. The filter is robust and adaptive. The present invention uses the maximum multi-core entropy criterion to effectively suppress outliers to improve the robustness of the filter. At the same time, the adaptation is achieved by approximating the covariance of the measurement noise using the variational Bayes approximation method.

[0134] The present invention constructs smooth pseudo-measurements in the filtering process by introducing new factors. The constructed smoothing matrix is ​​intended to reduce the influence of measurement outliers, thereby improving the accuracy of approximating noise parameters in the variational Bayesian approximation process.

[0135] The proposed robust adaptive filtering method improves the state estimation capability in complex engineering environments, such as load and temperature changes. Since this method is both robust and adaptive, it is very suitable for display systems where noise is affected by complex interference, and can achieve high-precision estimation of battery state of charge.

[0136] The model of the present invention has a simple structure, a convenient architecture, and a wide range of applicability and promotion.

[0137] Based on the above method, the present invention discloses a battery SOC estimation system based on variational Bayes and multi-core entropy UKF, see Figure 4 ,include: State space equation acquisition module: used to construct the state space equation according to the internal electrochemical reaction mechanism of the battery; Filter construction module: used to construct an unscented Kalman filter based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion to obtain a newly constructed unscented Kalman filter; Battery state estimation module: used to estimate the battery state based on the state space equation combined with the newly constructed unscented Kalman filter.

[0138] The various modules of the system of the present invention cooperate with each other to achieve accurate estimation of the battery SOC.

[0139] An electronic device comprises: a processor; a memory for storing computer program instructions; and steps for implementing a battery SOC estimation method based on variational Bayes and multi-core entropy UKF when executing the computer program.

[0140] A storage medium stores computer program instructions. When the computer program instructions are loaded and run by a processor, the processor executes a battery SOC estimation method based on variational Bayes and multi-core entropy UKF.

[0141] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0142] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0143] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0144] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0145] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A battery SOC estimation method based on variational Bayes and multi-core entropy UKF, characterized in that: The following steps are involved: Construct the state space equation according to the internal electrochemical reaction mechanism of the battery; An unscented Kalman filter is constructed based on a variational Bayesian approximation method and a maximum multi-core correlation entropy criterion, and a newly constructed unscented Kalman filter is obtained; Battery state estimation is performed based on the state-space equation combined with a newly constructed unscented Kalman filter.

2. The battery SOC estimation method based on variational Bayes and multi-core entropy UKF according to claim 1 is characterized in that: The state space equation is constructed according to the internal electrochemical reaction mechanism of the battery, as follows: The battery electrochemical reaction mechanism equation is constructed based on the average lithium ion concentration of the solid phase, the lithium ion concentration on the solid phase surface, the pore wall flow, the solid phase particle radius and the diffusion coefficient of lithium ions in the solid phase particles; The state space equation is constructed according to the battery electrochemical reaction mechanism equation.

3. The battery SOC estimation method based on variational Bayes and multi-core entropy UKF according to claim 1 is characterized in that: The state space equation is as follows: in, is the state quantity; is the battery charge state; , , and is the intermediate variable corresponding to the positive and negative poles; is the matrix transpose symbol; is the voltage, i.e. the observed quantity; is the positive reference voltage; is the negative reference voltage; for The positive electrode ion concentration proportional coefficient at the moment; for The negative electrode ion concentration proportional coefficient at the moment; For time; is a function of current.

4. The battery SOC estimation method based on variational Bayes and multi-core entropy UKF according to claim 1 is characterized in that: The unscented Kalman filter is constructed based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion to obtain a newly constructed unscented Kalman filter, which is specifically as follows: Get the estimated value of the state quantity at the current moment; Perform an untraceable transformation on the estimated value of the state quantity at the current moment to obtain a set of sampling points corresponding to the state quantity at the current moment; According to a set of sampling points corresponding to the state quantity at the current moment, obtain the prior estimate of the state quantity at the next moment; Obtain a priori estimated value of the observation quantity according to the priori estimated value of the state quantity, and obtain the observation noise covariance matrix according to the priori estimated value of the observation quantity; Iteratively update the observation noise covariance matrix according to the variational Bayesian approximation method; A fixed-point iteration algorithm is used to iteratively update the prior estimation value of the state quantity to obtain the posterior estimation value of the state quantity; The posterior covariance matrix is ​​obtained according to the posterior estimated value of the state quantity combined with the observation noise covariance matrix, and the process parameters of the variational Bayesian method are updated to obtain a newly constructed unscented Kalman filter.

5. The battery SOC estimation method based on variational Bayes and multi-core entropy UKF according to claim 4 is characterized in that: The observation noise covariance matrix is ​​iteratively updated according to the variational Bayesian approximation method, as follows: in, is the observation noise covariance matrix, is a free parameter, is the dimension of the state quantity, is the inverse scaling matrix, is the number of identifiers, is the current time step.

6. The battery SOC estimation method based on variational Bayes and multi-core entropy UKF according to claim 4 is characterized in that: The fixed-point iteration algorithm is used to iteratively update the prior estimate of the state quantity to obtain the posterior estimate of the state quantity, as follows: Obtaining the covariance matrix of system prediction according to the sampling point of the current untraceable transformation, the state quantity at the current moment, and the prior estimation value of the observation quantity at the current moment; Obtaining a measurement slope matrix based on the current a priori covariance matrix and the system predicted covariance matrix; Construct a batch regression model based on the current state, observations, posterior estimates of observations, measurement slope matrix, joint noise matrix, and n-order identity matrix; A regression equation is obtained according to the batch regression model; Obtain the posterior estimate of the state quantity based on the regression equation.

7. The battery SOC estimation method based on variational Bayes and multi-core entropy UKF according to claim 4 is characterized in that: The posterior covariance matrix is ​​obtained according to the posterior estimate of the state quantity combined with the observation noise covariance matrix, and the process parameters of the variational Bayesian method are updated to obtain a newly constructed unscented Kalman filter, as follows: Obtaining a posterior covariance matrix according to the measurement slope matrix, the observation noise covariance matrix and the prior covariance matrix; The process parameters of the variational Bayes method are updated based on the prior inverse scaling matrix, the prior estimates of the observations, and the actual observations.

8. The battery SOC estimation method based on variational Bayes and multi-core entropy UKF according to claim 1 is characterized in that: The battery state estimation is performed according to the state space equation combined with the newly constructed unscented Kalman filter, as follows: Combine actual battery data and state-space equations to obtain a priori estimation results of the battery state of charge; The battery state of charge estimation value is obtained by processing the battery's prior estimation results through the newly constructed unscented Kalman filter.

9. A battery SOC estimation system based on variational Bayes and multi-core entropy UKF, characterized in that: include: State space equation acquisition module: used to construct the state space equation according to the internal electrochemical reaction mechanism of the battery; Filter construction module: used to construct an unscented Kalman filter based on the variational Bayesian approximation method and the maximum multi-core correlation entropy criterion to obtain a newly constructed unscented Kalman filter; Battery state estimation module: used to estimate the battery state based on the state space equation combined with the newly constructed unscented Kalman filter.

10. An electronic device, comprising: Processor; memory, electronic device used to store computer program instructions; characterized in that it is used to implement the steps of the battery SOC estimation method based on variational Bayes and multi-core entropy UKF as described in any one of claims 1-8 when executing the computer program.

Images