CN115659814A - Method, system, device, automobile and medium for predicting remaining life probability distribution of battery - Google Patents

Abstract

The application provides a method for predicting the probability distribution of the remaining life of a battery based on MCMC sampling, wherein a battery capacity decline model is established based on the discharge capacity and the number of discharge cycles to fit the discharge capacity degradation trend of a battery system; obtaining battery degradation model parameters posterior PDF based on the discharge capacity observation sequence; describing the similarity between the fitting value and the observed value of the battery capacity decline model through a likelihood function; acquiring PDF of Gaussian distribution as a likelihood function; MCMC samples to obtain a sampling value of a battery capacity decline model parameter; and substituting all sampling values into a battery capacity fading model to extrapolate to a failure threshold value, and obtaining the probability distribution of the prediction result of the RUL. The probability distribution provides quantitative prediction uncertainty, and a maintenance strategy can make a reasonable decision based on the uncertainty, so that unexpected faults and losses caused by battery degradation are avoided, the battery predictive maintenance and pushing strategy are more convenient, and the method is easy to deploy and implement at the cloud.

Description

Method, system, device, automobile and medium for predicting remaining life probability distribution of battery

Technical Field

The application relates to the technical field of new energy automobile battery safety monitoring, in particular to a method for determining probability distribution of a battery residual life prediction result based on MCMC sampling.

Background

At present, the traditional battery safety monitoring is mainly based on a battery thermal runaway algorithm, and predictive judgment is difficult to be given to the current vehicle battery state and the future battery degradation trend. The long-time cyclic charge and discharge of the lithium ion battery can cause the capacity of the lithium ion battery to decline. Capacity degradation is also relatively slow during early use of the battery; when the capacity is degraded to a certain degree, the battery can be degraded rapidly, the phenomenon of capacity 'water jump' occurs, and the capacity is degraded to the battery failure in a short time. The residual service Life RUL (Remaining Useful Life) of the battery is predicted, so that the prediction of the future failure time point of the battery is facilitated, the operation risk of the battery is sensed and reduced in advance, and the vehicle using experience of a user is improved.

In the prior art, a battery residual life prediction model is mostly established based on a machine learning method. The machine learning method may use optimization methods such as observation data and least square method to obtain parameter values of the model, but may not give the distribution of the model parameters, so that a Probability Density Function (PDF) of the predicted residual life value cannot be derived, and a confidence interval of the predicted result cannot be obtained. However, the probability distribution of the remaining life prediction results is crucial for risk analysis and maintenance decisions of battery failure. The probability distribution provides a quantitative uncertainty for the prediction result, and the maintenance strategy can make a reasonable decision based on the uncertainty, so that unexpected faults and losses caused by battery degradation are avoided. As disclosed in publication No.: the publication No. CN202010568146.1 discloses a Chinese invention patent of 'a method for predicting the residual life of a lithium battery of an electric forklift based on multi-neural-network coupling'.

The publication number CN110442941B is named as a battery state and RUL prediction method based on particle filtering and process noise fusion, an empirical state equation and measurement information are updated by using an optimal tracking model identification parameter, new proposed distribution is generated by guidance, a battery residual service life prediction model is established by using a standard PF algorithm, SIS (simulated information system) is revised and proposed by combining state tracking optimal measurement information to update particle weight calculation, an MCMC-based updated improved PF algorithm model is established, a final state estimation value of battery capacity is obtained, and whether a battery failure threshold value is reached is judged.

However, the model parameters used to describe the degradation process are fixed. In practical situations, similar batteries are affected by the consistency difference of the manufacturing process, the following degradation rule is fluctuated, but the fixed model parameters cannot describe the fluctuation in the degradation model. Therefore, when describing the battery degradation process, the uncertainty of the degradation process model parameters should be considered and finally reflected in the probability distribution of the residual life prediction result.

Disclosure of Invention

In view of the above, the present application is directed to the problem that the existing machine learning method cannot provide the probability distribution of the remaining life prediction result and the problem that the closed analytic solution of the posterior probability density of the battery degradation model parameter is difficult to obtain, in order to obtain the probability distribution of the remaining life prediction result of the battery, the posterior probability distribution of the battery capacity degradation model parameter is obtained first, and the closed analytic solution of the posterior probability distribution of the degradation model parameter is difficult to obtain because the battery capacity degradation model is often highly nonlinear. The posterior probability distribution of the battery capacity degradation model parameters is sampled through MCMC sampling, the numerical solution of the posterior probability distribution of the battery capacity degradation model parameters is obtained, the model parameters are substituted into the degradation model, the degradation track of the battery capacity is extrapolated to a failure threshold value, the posterior probability distribution of the model parameters is obtained, and then the probability function of the residual life prediction result is obtained.

According to one aspect of the application, a method for predicting the probability distribution of the remaining life of a battery based on MCMC sampling is provided, and a battery capacity decline model is established based on a battery discharge capacity curve and the number of discharge cycles to fit the discharge capacity degradation trend of a battery system; obtaining a posterior probability density function of battery degradation model parameters based on the discharge capacity observation sequence; describing the similarity between the fitting value and the observed value of the battery capacity decline model through a likelihood function; obtaining probability density function PDF of Gaussian distribution as a likelihood function; setting MCMC sampling parameters according to the degradation model, and carrying out posterior probability distribution p (theta | Y) on the battery capacity degradation model parameters _{1:T} ) Sampling, determining a battery degradation model parameter at the current discharge cycle moment according to the acceptance probability of the sampling parameter, and calculating an estimated value of the battery capacity degradation model parameter; substituting the parameter estimation value into the regression model to extrapolate to a failure threshold value to obtain a predicted value of the residual life RUL of the battery; and substituting the parameter estimation values of all MCMC sampling values into a battery capacity fading model to extrapolate to a failure threshold value, and obtaining the probability distribution of the prediction result of the RUL.

Further preferably, the obtaining of the battery degradation model parameter at the current discharge cycle time includes: sampling each independent parameter in the battery capacity decline model until ns times of sampling are completed to obtain ns groups of sampling values, calculating the acceptance probability of each sampling value, and taking the sampling value with the acceptance probability meeting the requirement as the battery degradation model parameter at the current discharge cycle time.

Preferably, all ns model parameter sampling values are substituted into the degradation model to form ns degradation track curves, the current time is subtracted from the time when the track reaches the failure threshold value to obtain the residual service life RUL value of the model parameter, ns RUL values are obtained through the ns degradation track curves to form RUL distribution of the current time, and a posterior probability distribution numerical solution of the residual service life prediction result is constructed.

Further preferably, the battery discharge capacity f (t) at the time t in the discharge capacity degradation trend is a function of a model parameter theta and time, and the posterior probability distribution of the model parameter theta is a given observation sequence Y _{1:T} Conditional probability of time theta p (theta | Y) _{1:T} ) After sampling)And (5) testing the probability distribution to obtain a numerical solution of the posterior probability distribution.

Preferably, the step of establishing the battery capacity decline model comprises the step of establishing the discharge capacity decline model by describing the battery degradation process by using an exponential function according to the discharge cycle number and the discharge capacity curve

And calculating a battery discharge capacity fitting value f (t) of the current discharge cycle time t, wherein a, b and c are mutually independent degradation model parameters which are subjected to Gaussian distribution in a priori.

Further preferably, the determining of the battery degradation model parameter at the current discharging cycle time according to the acceptance probability of the sampling parameter includes suggesting a distribution q (θ) from the ith time of the battery degradation discharging cycle number ⁽ⁱ⁺¹⁾ |θ ⁽ⁱ⁾ ) Sampling out any independent parameter theta ⁱ⁺¹ According to the prior probability distribution p (theta) of each independent parameter at the ith moment of the discharge cycle ⁽ⁱ⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁾ ) Posterior probability distribution p (theta) ⁱ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁾ |θ ⁽ⁱ⁺¹⁾ ) The prior probability distribution p (θ) of each independent parameter at the i +1 th time ⁽ⁱ⁺¹⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁺¹⁾ ) Posterior probability distribution p (theta) ⁱ⁺¹ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁺¹⁾ |θ ⁱ ) Calling a formula:

calculating any independent model parameter theta at the i +1 th moment of the discharge cycle number in the battery degradation model ⁽ⁱ⁺¹⁾ Acceptance probability A (θ) ⁽ⁱ⁺¹⁾ |θ ⁱ ) Generating a random number U, U-U (0, 1) from the uniform distribution if U<A(θ ⁽ⁱ⁺¹⁾ |θ ⁱ ) Receiving the parameter theta at the i +1 th moment of the discharge cycle number ⁱ⁺¹ As the model parameter at the current moment, otherwise, keeping the model parameter at the current moment as the parameter theta at the ith moment ⁱ The value of (c).

Further preferably, the sampled values of all MCMCMCMCMC areSubstituting the parameter estimation value into the battery degradation model and extrapolating to the failure threshold value to obtain the prediction probability distribution of the RUL further comprises: all n obtained by sampling _s Substituting the group model parameters into the degradation model to form n _s A degenerated track curve is obtained, the time when the track reaches the failure threshold value is subtracted from the current time to obtain the residual service life RUL value of the model parameter, and the residual service life RUL value is obtained according to n _s The RUL values form the RUL distribution at the current moment, and a posterior probability distribution numerical solution of the residual life prediction result is constructed.

Further preferably, the step of substituting the parameter estimation values of all the MCMC sampling values into the battery degradation model to extrapolate to the failure threshold to obtain the predicted probability distribution of RUL further comprises: estimating an expectation of a posterior probability distribution of a degradation model parameter based on the majority theorem

I.e. the parameter theta according to any time i of the discharge cycle ⁱ Calling a formula:

calculating expectation of posterior probability distribution of degradation model parameter theta

Will be provided with

Substituting the estimated value into the degradation model and extrapolating to a failure threshold value to obtain an RUL predicted value, wherein n _s The number of discharge cycles.

Further preferably, according to the formula:

calculating a fitting value f (t) of a degradation model at the discharge cycle time t and the discharge capacity y _t Has a likelihood function of p (y) _t Theta), collecting a series of discharge capacity observation sequences Y _{1:T} (Y _{1:T} ＝y ₁ ，y _t+1 ......y _T ) According to the formula:

calculating an observation sequence Y _{1:T} Overall likelihood function p (Y) _{1:T} θ), according to the formula:

obtaining the posterior PDF p (theta | Y) of the parameters of the degeneration model _{1:T} )

According to another aspect of the present application, a system for predicting a probability distribution of remaining life of a battery based on MCMC sampling is provided, which includes: the battery capacity degradation model comprises an MCMC sampling unit, a counting unit, an algorithm module, a battery capacity degradation model and a battery capacity degradation model, wherein the discharging capacity degradation trend of a fitting battery system is established based on discharging capacity and discharging cycle number, the posterior PDF of battery degradation model parameters is obtained based on a discharging capacity observation sequence, the similarity between a fitting value and an observation value of the battery capacity degradation model is described through a likelihood function, and the Gaussian-distributed PDF is obtained to be used as the likelihood function; MCMC sampling unit for setting MCMC sampling parameters according to degradation model parameters and obtaining posterior probability distribution p (theta | Y) of model parameters _{1:T} ) Sampling parameters, and determining the parameters of the battery degradation model at the current discharge cycle time according to the acceptance probability of the sampled parameters; the algorithm module is used for calculating the estimation value of the current model parameter according to the battery degradation model parameter at the current discharge cycle time, substituting all the parameter estimation values into the battery capacity degradation model extrapolation value failure threshold value and obtaining the probability distribution of the prediction result of the RUL; a counting unit for counting the number n of discharge cycles _s Counting, when the number of MCMC samples reaches n _s And stopping sampling.

Preferably, the battery capacity decline model is constructed by describing the battery degradation process by using an exponential function according to the discharge cycle number and the discharge capacity curve

And calculating a battery discharge capacity fitting value f (t) at the current discharge cycle time t, wherein a, b and c are mutually independent degradation model parameters and are subjected to Gaussian distribution a priori.

Further preferably, the determining of the parameters of the battery degradation model at the current discharging cycle time according to the acceptance probability of the sampling parameters includes suggesting a distribution q (θ) from the ith time of the number of battery degradation discharging cycles ⁽ⁱ⁺¹⁾ |θ ⁽ⁱ⁾ ) Sampling out any independent parameter theta ⁱ⁺¹ According to the prior probability distribution p (theta) of each independent parameter at the ith moment of the discharge cycle ⁽ⁱ⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁾ ) Posterior probability distribution p (theta) ⁱ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁾ |θ ⁽ⁱ⁺¹⁾ ) The prior probability distribution p (θ) of each independent parameter at the i +1 th time ⁽ⁱ⁺¹⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁺¹⁾ ) Posterior probability distribution p (theta) ⁱ⁺¹ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁺¹⁾ |θ ⁱ ) Calling a formula:

calculating any independent model parameter theta at the i +1 th moment of the discharge cycle number in the battery degradation model ⁽ⁱ⁺¹⁾ Is (a) of ⁽ⁱ⁺¹⁾ |θ ⁱ ) Generating a random number U, U-U (0, 1) from the uniform distribution if U<A(θ ⁽ⁱ⁺¹⁾ |θ ⁱ ) Receiving the parameter theta at the i +1 th moment of the discharge cycle number ⁱ⁺¹ As the model parameter at the current moment, otherwise, keeping the model parameter at the current moment as the parameter theta at the ith moment ⁱ The value of (c).

More preferably, the discharge capacity y at the discharge cycle time t is obtained _t Setting the standard deviation sigma _s Obtaining the PDF of Gaussian distribution as a likelihood function according to a formula:

calculating a fitting value f (t) of a degradation model at the discharge cycle time t and the discharge capacity y _t Likelihood function p (y) _t | θ), a series of discharge capacity observation sequences Y are collected _{1:T} (Y _{1:T} ＝y ₁ ，y _t+1 ......y _T ) According to the formula:

the calculation is based on the total observed sequence Y _{1:T} Overall likelihood function p (Y) _{1:T} θ), according to the formula:

obtaining the posterior PDF p (theta | Y) of the parameters of the degeneration model _{1:T} )。

According to another aspect of the application, an electronic device is proposed, comprising: a processor; and a memory storing a program, wherein the program comprises instructions that when executed by the processor cause the processor to perform the method of predicting a probability distribution of remaining life of a battery based on MCMC sampling described above.

According to another aspect of the present application, a non-transitory computer-readable storage medium is provided that stores computer instructions for causing a computer to perform a method of predicting a remaining life probability distribution of a battery according to the above-described MCMC sampling.

According to another aspect of the present application, there is provided an automobile comprising a system for predicting a probability distribution of remaining life of a battery based on MCMC sampling as described above.

The method for obtaining the probability distribution of the residual life prediction result provides the probability distribution for the residual life prediction result of the battery, the probability distribution provides quantitative prediction uncertainty for the prediction result, and a maintenance strategy can make a reasonable decision based on the uncertainty, so that unexpected faults and losses caused by battery degradation are avoided. Under the condition that most of the existing battery residual life prediction models cannot provide uncertainty of a prediction result, analytical solutions of model parameter posterior probability distribution do not need to be deduced again due to the change of a capacity degradation model, sampling is directly carried out from the model parameter posterior probability distribution, and a solution for quantitatively determining the uncertainty of the battery residual life prediction result is provided; the method provides quantitative uncertain measurement for the prediction result of the remaining life of the battery, is more convenient for predictive maintenance of the battery and formulation of a push strategy, and is easy to deploy and implement at the cloud.

Drawings

Further details, features and advantages of the present application are disclosed in the following description of exemplary embodiments with reference to the attached drawings.

FIG. 1 is a flow chart illustrating a process for determining a probability distribution of a remaining life prediction result according to an exemplary embodiment of the present application;

FIG. 2 is a graph of discharge capacity degradation of a cell in a data set determined according to an exemplary embodiment of the present application;

fig. 3 is a schematic diagram illustrating PDF distribution of failure time obtained by determining a cell based on MCMC sampling after the 50 th discharge cycle according to an exemplary embodiment of the present application;

fig. 4 is a schematic diagram for determining a predicted value of remaining life and a PDF distribution thereof according to an exemplary embodiment of the present application.

Detailed Description

Embodiments of the present application will be described in more detail below with reference to the accompanying drawings. While certain embodiments of the present application are shown in the drawings, it should be understood that the present application may be embodied in various forms and should not be construed as limited to the embodiments set forth herein, but rather are provided for a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are for illustration purposes only and are not intended to limit the scope of the present application.

It should be understood that the various steps recited in the method embodiments of the present application may be performed in a different order and/or in parallel. Moreover, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the present application is not limited in this respect.

The term "including" and variations thereof as used herein is intended to be open-ended, i.e., "including but not limited to". The term "based on" is "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Relevant definitions for other terms will be given in the following description. It should be noted that the terms "first", "second", and the like in the present application are only used for distinguishing different devices, modules or units, and are not used for limiting the order or interdependence relationship of the functions performed by the devices, modules or units.

It is noted that references to "a" or "an" modification in this application are intended to be illustrative rather than limiting, and those skilled in the art will appreciate that references to "one or more" are intended to be exemplary unless the context clearly indicates otherwise.

The names of messages or information exchanged between a plurality of devices in the embodiments of the present application are for illustrative purposes only, and are not intended to limit the scope of the messages or information.

According to the embodiment of the invention, the posterior probability distribution of the model parameters is obtained by sampling the posterior probability distribution of the battery capacity degradation model parameters, and then the probability distribution of the residual life prediction result is obtained.

In order to obtain the probability distribution of the prediction result of the remaining life of the battery, the posterior probability distribution of the battery capacity degradation model parameters is obtained first. Because the battery capacity degradation model is often highly nonlinear, it is difficult to obtain a closed analytic solution of posterior probability distribution of degradation model parameters, and a numerical solution of the model parameter probability distribution needs to be obtained in a sampling manner, so as to obtain the probability distribution of the residual life prediction result, namely a probability density function.

In order to consider the uncertainty of model parameters in the degradation process, the invention obtains posterior estimation samples of the degradation model parameters based on MCMC sampling, and extrapolates the posterior estimation samples to the failure threshold value to obtain the residual life values of the number of the samples, and counts the residual life values to obtain the probability distribution of the residual life prediction result. And reflecting the uncertainty of the model parameters into the probability distribution of the residual life prediction result.

FIG. 1 shows a schematic flow diagram for determining a remaining life prediction result probability distribution according to an exemplary embodiment of the present application.

The method comprises the following steps: establishing a battery capacity degradation model, setting prior probability density function PDF (probability density function) of degradation model parameters, and acquiring the posterior probability density function PDF of the degradation model parameters based on a discharge capacity observation sequence. Initializing a degradation model parameter, setting an MCMC sampling parameter, and acquiring a preset sampling value of the degradation model parameter by MCMC sampling; calculating an estimated value of a degradation model parameter based on the sampling value; substituting the parameter estimation value of the degradation model into an extrapolation value failure threshold value of the degradation model to obtain a predicted value of the residual life RULremaininguseful life) of the battery; and substituting all MCMC sampling values into a battery capacity degradation model to extrapolate to a failure threshold value, and obtaining the probability distribution of the prediction result of the RUL. The method specifically comprises the following steps:

establishing a battery capacity degradation model based on the discharge capacity and the discharge cycle number cycle; setting a prior PDF (p (theta)) of the degradation model parameter based on the prior obeying Gaussian distribution of the degradation model parameter theta; obtaining model parameter posterior PDF based on the discharge capacity observation sequence; describing the similarity between the fitting value and the observed value of the battery capacity decline model through a likelihood function; acquiring PDF of Gaussian distribution as a likelihood function; posterior probability distribution of degradation model parameter theta for given observation sequence Y _{1:T} Conditional probability of time theta, i.e., (theta | Y) _{1:T} ) ); and obtaining a numerical solution of the posterior probability distribution by sampling the posterior probability distribution.

And establishing a capacity degradation model based on the discharge capacity and the discharge cycle number cycle. Because the battery capacity is always attenuated in the cyclic charge and discharge process, the degradation process of the battery capacity can be generally described by using a power function model, a polynomial model, an exponential model and the like, and the capacity degradation model is fitted to the battery discharge capacity f (t) at the moment t of the current discharge cycle number according to the current discharge cycle number of the battery. the battery discharge capacity f (t) at the time t is a function of the model parameters and time t, the capacity degradation model is constructed by adopting the following formula, the prior PDF of the degradation model parameters is set, the degradation model parameters are assumed to be mutually independent parameters a, b and c, the battery discharge capacity f (t) at the time t, which is fitted by the degradation model, is determined according to the current discharge cycle number time t of the battery,

the prior of the degradation model parameters (a, b and c) obeys Gaussian distribution, for the convenience of description parameters a, b and c later are uniformly represented by theta, namely theta epsilon (a, b and c), if the distribution of the model parameters theta obeys normal distribution with the mean value of mu theta and the standard deviation of sigma theta, theta-N (mu theta, sigma theta) can be obtained, and the prior uncertainty of the model parameters is represented.

Wherein the Gaussian distribution parameters are specified according to a rule such as the mean value μ of the Gaussian distribution _θ Has higher confidence, then the standard deviation sigma _θ To take a smaller value, conversely, a larger value should be assigned to ensure that the mean value μ is distributed _θ Depends mainly on the observed data and not on the prior distribution. Thus, according to the formula:

the overall likelihood function p (Y) can be obtained based on the discharge capacity observation sequence _{1:T} Theta), obtaining model parameter posterior PDF p (theta Y) _{1:T} )。

A method of calculating an overall likelihood function of an exemplary embodiment of the present application.

As the number of discharge cycles increases, the observed value of the discharge capacity also increases. In order to continuously fuse newly added observed value information, the similarity degree between the capacity decline model fitting value and the observed value is described through a likelihood function. After the preset discharge cycle number is reached, the discharge capacity y at the current discharge cycle time t is obtained _t Set the standard deviation to be σ _s And acquiring the PDFs of the Gaussian distribution as likelihood functions. I.e. according to the formula:

calculating a fitting value f (t) and a discharge capacity y of the degradation model at the time t _t Has a likelihood function of p (y) _t| θ)。

Suppose that T discharge cycles are currently completedLooping to obtain a series of discharge capacity observation sequences Y _{1:T} (Y _{1:T} ＝y _t ，y _t+1 ......y _T ) Calculating a likelihood function for each moment of the discharge cycle according to the above formula, thereby obtaining an overall likelihood function p (Y) _{1:T} |θ)。

The present embodiment may determine the overall likelihood function as the product of the likelihood functions at each time instant, i.e. according to the formula:

calculating an observation sequence Y _{1:T} Overall likelihood function p (Y) _{1:T} |θ)。

The posterior probability distribution of any parameter theta in independent parameters of the degradation model is given to the observation sequence Y _{1:T} The conditional probability of the time parameter, the battery discharge capacity f (t) at time t is a function of the model parameter theta and time,

thus, the posterior probability distribution p (θ | Y) of the model parameters can be constructed by Bayesian inference _{1:T} ) I.e. by

The integral term in the denominator of the above equation can be regarded as a normalization constant, and when there are n parameters to be estimated in the independent parameter θ, it is n-fold integral, and it is known that p (θ | Y {1 t }) is difficult to find an analytical solution.

Due to p (theta | Y) _{1:T} ) It is difficult to obtain an analytic solution, so MCMC sampling is adopted to obtain a numerical solution of the posterior probability distribution of the model parameters. Posterior probability distribution p (theta | Y) from model parameters _{1:T} ) Sampling to obtain n _s Value n of _s The distribution of values describes the uncertainty of the degradation model parameters.

In the Markov Chain Monte Carlo method MCMC (Markov Chain Monte Carlo) sampling, the integral term is eliminated by the acceptance criterion of the Metropolis-Hastings algorithm, therefore, a large number of samples of each independent model parameter can be sampled by sampling the posterior probability distribution, and the numerical solution of the posterior probability distribution is obtained by the sampling value.

Initializing degradation model parameters, and setting MCMC sampling parameters: initializing the degradation model, and initializing each independent model parameter theta, for example, setting the MCMC sampling number as n _s Will suggest a distribution q (theta) ⁽ⁱ⁺¹⁾ |θ ⁽ⁱ⁾ ) Arranged to be evenly distributed. Number of samples n _s Generally, the calculation is set according to experience, and mainly considering computer performance overhead, the more the number of samples is, the better, the suggestion distribution can be set arbitrarily, such as uniform distribution, gaussian distribution, dirichlet distribution, and the like, and the present embodiment sets the suggestion distribution as uniform distribution for calculating the following acceptance probability a (θ) in the calculation ⁽ⁱ⁺¹⁾ |θ ⁽ⁱ⁾ ))。

Distribution q (theta) is proposed from the initial state at the i-th moment of the number of discharge cycles in battery degradation ⁽ⁱ⁺¹⁾ |θ ⁽ⁱ⁾ ) Sampling out the current arbitrary independent parameter theta ⁱ⁺¹ The proposed distribution is set to be uniformly distributed when setting MCMC sampling parameters, e.g. obeying [ theta ] ⁽ⁱ⁾ -0.5,θ ⁽ⁱ⁾ +0.5]The function is directly called to carry out uniform distribution sampling, and the parameter value theta obtained at the moment ⁱ⁺¹ For calculating the acceptance probability.

Calculating the acceptance probability A (theta) ⁱ⁺¹ |θ ⁽ⁱ⁾ ). According to the MCMC algorithm Metropolis-Hastings, the prior probability distribution p (theta) of each independent parameter at the ith moment of the discharge cycle number in the battery degradation model is calculated ⁽ⁱ⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁾ ) Posterior probability distribution p (theta) ⁱ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁾ |θ ⁽ⁱ⁺¹⁾ ) The prior probability distribution p (θ) of each independent parameter at the i +1 th time ⁽ⁱ⁺¹⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁺¹⁾ ) Posterior probability distribution p (theta) ⁱ⁺¹ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁺¹⁾ |θ ⁱ ) Calling a formula:

calculating battery withdrawalAny independent model parameter theta at the i +1 th moment of the discharge cycle number in the model ⁽ⁱ⁺¹⁾ Acceptance probability A (θ) ⁽ⁱ⁺¹⁾ |θ ⁱ ) The convergence speed is increased by deforming the original acceptance probability.

And determining the current discharge cycle number time parameter of the battery degradation model according to the receiving probability.

One way of doing this is to generate a random number U, U-U (0, 1) from a uniform distribution, U obeying [0,1 ]]Are evenly distributed in between. If u<A(θ ⁽ⁱ⁺¹⁾ |θ ⁱ ) Then the parameter theta at the time of the i +1 th discharge cycle number in the battery degradation model is accepted ⁱ⁺¹ As the current model parameter, the model parameter θ at the current time ⁱ Using predicted model parameter theta at time i +1 ⁱ⁺¹ Replacing; otherwise, rejecting the sampling, and keeping the model parameter at the current moment to be theta ⁱ The value of (a) is not changed. The battery degradation model parameter θ in all discharge cycles is obtained. Obtaining posterior probability distribution p (theta | Y) of battery degradation model parameter theta through MCMC algorithm sampling _{1:T} )。

Sampling each independent model parameter until n is completed _s Sub-sampling to obtain n _s Calculating the acceptance probability of each sampling value, wherein the sampling values meeting the requirement of the acceptance probability are used as the parameters of the battery degradation model at the moment of the current discharge cycle number; and substituting all sampling values meeting the acceptance probability into the battery degradation model to extrapolate to a failure threshold value, and obtaining the prediction probability distribution of the RUL.

All n are _s Substituting the group model parameters into the degradation model can form n _s A degenerated track curve is obtained, the time when the track reaches the failure threshold value is subtracted by the current time to obtain the residual life RUL value of the model parameter, and the residual life RUL value is obtained through n _s N can be obtained by a strip degeneration track curve _s A RUL value according to n _s Each RUL value forms the RUL distribution at the current time. To obtain n _s And constructing a posterior probability distribution numerical solution of the residual life prediction result according to the residual life prediction value.

Exemplary embodiments of the present application for obtaining the RUL value further include calculating estimated values of the model parameters based on the sampled values. Due to MCMC miningThe parameter values of the samples are subjected to a posterior probability distribution p (θ | Y {1 t }), and the expectation of the posterior probability distribution of the degradation model parameters can be estimated based on the majority theorem

Parameter theta at any ith time according to discharge cycle number ⁱ Calling a formula:

calculating expectation of posterior probability distribution of degradation model parameter theta

Expectation of posterior probability distribution of degradation model parameters

Substituting the estimated value into the degradation model to extrapolate to the failure threshold value to obtain the RUL predicted value and obtain the predicted value of the residual life.

In order to make the technical means, the achievement objects and the functions of the present invention easy to understand, the method for determining the probability distribution of the battery remaining life prediction result according to the present invention is specifically described below with reference to the specific embodiments.

The battery charge and discharge data used in the examples are from the NASA battery test public data set, the failure threshold of the battery capacity is set to 1.25Ah, the degradation model parameters are determined according to the MCMC sampling, and the discharge cycle-discharge capacity curve of the battery from operation to failure is shown in fig. 2.

In the embodiment, the exponential function is taken as an example to construct the battery capacity degradation model, and the exponential function can better fit the discharge capacity degradation trend of the battery system. Describing the degradation process of the battery by adopting an exponential function according to a discharge cycle-discharge capacity curve, constructing a discharge capacity degradation model, and according to a formula:

calculating outBattery discharge capacity f (t) at time t fitted by the degradation model. In the formula, t represents the current discharge cycle number of the battery, and f (t) represents the battery discharge capacity at time t. Three unknown parameters θ = [ a, b, c ] contained in the formula]For controlling the tendency of the capacity degradation trajectory. According to the MCMC sampling method, ns group model parameters [ a, b, c ] are obtained]And substituting the ns model parameters into the degradation model to form ns degradation track curves, wherein the residual service life RUL is the moment when the track reaches the failure threshold value-the current moment, so that ns RUL values can be obtained, and RUL distribution at the current moment can be formed according to the ns RUL values.

Fig. 3 is a schematic diagram showing a PDF of the failure time obtained based on MCMC sampling after the 50 th discharge cycle of the embodiment of the present invention, where the abscissa is the discharge cycle and the ordinate is the discharge capacity (AL). The real capacity degradation value curve is positioned above the failure threshold value, and n corresponds to the number of discharge cycles _s And (4) obtaining the real failure time by the degradation track, and obtaining the failure time distribution. According to the method provided by the invention, after the battery finishes the 50 th discharge cycle, the battery obtains n based on MCMC sampling _s The bar degradation trajectory and the failure time distribution PDF, and the probability distribution of the remaining life prediction result at the current time can be determined by subtracting the number of discharge cycles at the current time from the failure time distribution.

In the actual process of predicting the residual life at the 50 th discharge cycle, the model parameters are sampled to obtain n _s A sample value, i.e. n _s Group model parameters, n _s Substituting the group model parameters into the degradation model to form n _s A degenerated track curve is obtained, the time when the track curve reaches a failure threshold minus the current time is the residual service life RUL, and n is obtained _s A RUL value according to n _s Each RUL value forms the RUL distribution at the current time.

Fig. 4 is a schematic diagram showing a PDF relationship between a predicted remaining life value and a failure time thereof according to an embodiment of the present invention. And comparing the RUL predicted value with the RUL true value under the current cycle of the cyclic discharge, wherein the RUL predicted value is compared with the RUL true value, the RUL predicted result is in a probability density function relationship, and the curve is the probability density function of the RUL predicted result, wherein the broken line represents the RUL predicted value, and the RUL true value is represented. Fig. 4 shows predicted values of remaining life and their PDFs of the battery after 50 th, 75 th, 100 th and 125 th discharge cycles according to the method of the present invention, and shows probability distributions of multiple predicted results of remaining life of the battery, where the probability distribution at the 50 th discharge cycle in fig. 4 corresponds to the probability distribution of predicted remaining life at the 50 th discharge cycle in fig. 3.

The embodiment of the application also provides an automobile, which is a new energy automobile using the rechargeable lithium battery, and the automobile comprises an automobile battery and an automobile central control device. A system for predicting a probability distribution of remaining life of a battery based on MCMC sampling.

An exemplary embodiment of the present application also provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor. The memory stores a computer program executable by the at least one processor, the computer program, when executed by the at least one processor, is for causing the electronic device to perform a method according to an embodiment of the application.

The exemplary embodiments of this application also provide a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is adapted to cause the computer to perform a method according to embodiments of this application.

The exemplary embodiments of this application also provide a computer program product comprising a computer program, wherein the computer program is adapted to cause a computer to perform the method according to an embodiment of this application when executed by a processor of the computer.

Electronic device is intended to represent various forms of digital electronic computer devices, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the applications described and/or claimed herein.

Program code for implementing the methods of the present application may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program code, when executed by the processor or controller, causes the functions/acts specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this application, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and/or device (e.g., magnetic discs, optical disks, memory, programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user may provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

Claims (16)

1. The method for predicting the probability distribution of the remaining life of the battery based on MCMC sampling is characterized in that a battery capacity decline model is established based on a battery discharge capacity curve and the number of discharge cycles to fit a battery systemA discharge capacity degradation tendency; obtaining a posterior probability density function of battery degradation model parameters based on the discharge capacity observation sequence; describing the similarity between the fitting value and the observed value of the battery capacity decline model through a likelihood function; acquiring probability density function PDF of Gaussian distribution as a likelihood function; setting MCMC sampling parameters according to the degradation model, and obtaining posterior probability distribution p (theta | Y) of model parameters _{1:T} ) Sampling, determining a battery degradation model parameter at the current discharge cycle moment according to the acceptance probability of the sampling parameter, and calculating an estimated value of the battery capacity degradation model parameter; substituting the parameter estimation value into the regression model to extrapolate to a failure threshold value to obtain a predicted value of the residual life RUL of the battery; and substituting the parameter estimation values of all MCMC sampling values into a battery capacity fading model to extrapolate to a failure threshold value, and obtaining the probability distribution of the prediction result of the RUL.

2. The method of claim 1, wherein obtaining battery degradation model parameters at a current discharge cycle time comprises: sampling each independent parameter in the battery capacity degradation model until ns times of sampling are completed to obtain ns groups of sampling values, calculating the acceptance probability of each sampling value, and taking the sampling value with the acceptance probability meeting the requirement as the battery degradation model parameter at the current discharge cycle time.

3. The method of claim 2, wherein all ns model parameter sampling values are substituted into the degradation model to form ns degradation trajectory curves, the current time is subtracted from the time when the trajectory reaches the failure threshold to obtain the residual life RUL value of the model parameter, the ns RUL values are obtained through the ns degradation trajectory curves to form RUL distribution at the current time, and a posterior probability distribution numerical solution of the residual life prediction result is constructed.

4. A method according to any one of claims 1-3, characterized in that the battery discharge capacity f (t) at time t in said discharge capacity degradation trend is a function of a model parameter θ and time, the a posteriori probability distribution of the model parameter θ being given an observation sequence Y _{1:T} Conditional probability of time thetap(θ|Y _{1:T} ) And obtaining a numerical solution of the posterior probability distribution by sampling the posterior probability distribution.

5. The method according to any one of claims 1 to 3, wherein the step of establishing the battery capacity degradation model comprises the step of establishing the discharge capacity degradation model by describing the degradation process of the battery by using an exponential function according to the discharge cycle number and the discharge capacity curve

And calculating a battery discharge capacity fitting value f (t) of the current discharge cycle time t, wherein a, b and c are mutually independent degradation model parameters which are subjected to Gaussian distribution in a priori.

6. The method of any one of claims 1-3, wherein determining the battery degradation model parameter at the current discharge cycle based on the acceptance probability of the sampled parameter comprises suggesting a distribution q (θ) from the ith time of the number of battery degradation discharge cycles ⁽ⁱ⁺¹⁾ |θ ⁽ⁱ⁾ ) Sampling out any independent parameter theta ⁱ⁺¹ According to the prior probability distribution p (theta) of each independent parameter at the ith moment of the discharge cycle ⁽ⁱ⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁾ ) Posterior probability distribution p (theta) ⁱ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁾ |θ ⁽ⁱ⁺¹⁾ ) The prior probability distribution p (θ) of each independent parameter at the i +1 th time ⁽ⁱ⁺¹⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁺¹⁾ ) Posterior probability distribution p (theta) ⁱ⁺¹ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁺¹⁾ |θ ⁱ ) Calling a formula:

calculating any independent model parameter theta at the i +1 th moment of the discharge cycle number in the battery degradation model ⁽ⁱ⁺¹⁾ Is (a) of ⁽ⁱ⁺¹⁾ |θ ⁱ ) Generating a random number U, U-U (0, 1) from the uniform distribution if U<A(θ ⁽ⁱ⁺¹⁾ |θ ⁱ ) Receiving the parameter theta at the i +1 th moment of the discharge cycle number ⁱ⁺¹ As the model parameter at the current moment, otherwise, keeping the model parameter at the current moment as the parameter theta at the ith moment ⁱ The value of (c).

7. The method of claim 6, wherein substituting the parameter estimates for all MCMC sample values into a battery degradation model to extrapolate to a failure threshold to obtain a predicted probability distribution for RUL further comprises: all n obtained by sampling _s Substituting the group model parameters into the degradation model to form n _s A degenerated track curve is obtained, the time when the track reaches the failure threshold value is subtracted from the current time to obtain the residual service life RUL value of the model parameter, and the residual service life RUL value is obtained according to n _s The respective RUL values form the RUL distribution at the current moment, and a posterior probability distribution numerical solution of the residual life prediction result is constructed.

8. The method of claim 6, wherein substituting the parameter estimates for all MCMC sample values into a battery degradation model to extrapolate to a failure threshold to obtain a predicted probability distribution for RUL further comprises: estimating an expectation of a posterior probability distribution of a degradation model parameter based on the majority theorem

I.e. the parameter theta according to any time i of the discharge cycle ⁱ Calling a formula:

calculating expectation of posterior probability distribution of degradation model parameter theta

Will be provided with

Substituting the estimated value into the degradation model to extrapolate to a failure threshold value to obtain a RUL predicted value, wherein n _s The number of discharge cycles.

9. Method according to one of claims 1-3, 7, 8, characterized in that according to the formula:

calculating a fitting value f (t) of a degradation model at the discharge cycle time t and the discharge capacity y _t Has a likelihood function of p (y) _t | θ), a series of discharge capacity observation sequences Y are collected _{1:T} (Y _{1 :T}＝y ₁ ，y _t+1 ......y _T ) According to the formula:

calculating an observation sequence Y _{1:T} Overall likelihood function p (Y) _{1:T} θ), according to the formula:

obtaining the posterior PDF p (theta | Y) of the parameters of the degeneration model _{1:T} )。

10. A system for predicting the probability distribution of the remaining life of a battery based on MCMC sampling is characterized by comprising the following steps: the battery capacity degradation model comprises an MCMC sampling unit, a counting unit, an algorithm module, a battery capacity degradation model and a battery capacity degradation model, wherein the discharging capacity degradation trend of a fitting battery system is established based on discharging capacity and discharging cycle number, the posterior PDF of battery degradation model parameters is obtained based on a discharging capacity observation sequence, the similarity between a fitting value and an observation value of the battery capacity degradation model is described through a likelihood function, and the Gaussian-distributed PDF is obtained to be used as the likelihood function; MCMC sampling unit for setting MCMC sampling parameters according to degradation model parameters and obtaining posterior probability distribution p (theta | Y) of model parameters _{1:T} ) Parameter sampling is carried out, and battery degradation model parameters at the current discharge cycle moment are determined according to the acceptance probability of the sampling parameters; the algorithm module is used for calculating the estimation value of the current model parameter according to the battery degradation model parameter at the current discharge cycle time, substituting all the parameter estimation values into the battery capacity degradation model extrapolation value failure threshold value and obtaining the probability distribution of the prediction result of the RUL; counting sheetBased on the number of discharge cycles n _s Counting, when the number of MCMC samples reaches n _s And stopping sampling.

11. The system according to claim 10, wherein the battery capacity degradation model is constructed by describing a degradation process of the battery by using an exponential function according to a discharge cycle number and a discharge capacity curve

And calculating a battery discharge capacity fitting value f (t) at the current discharge cycle time t, wherein a, b and c are mutually independent degradation model parameters and are subjected to Gaussian distribution a priori.

12. The system of claim 10 or 11, wherein determining the battery degradation model parameter at the current discharge cycle time based on the acceptance probability of the sampled parameter comprises suggesting a distribution q (θ) from the ith time of the number of battery degradation discharge cycles ⁽ⁱ⁺¹⁾ |θ ⁽ⁱ⁾ ) Sampling out any independent parameter theta ⁱ⁺¹ According to the prior probability distribution p (theta) of each independent parameter at the ith moment of the discharge cycle ⁽ⁱ⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁾ ) Posterior probability distribution p (theta) ⁱ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁾ |θ ⁽ⁱ⁺¹⁾ ) The prior probability distribution p (θ) of each independent parameter at the i +1 th time ⁽ⁱ⁺¹⁾ ) Likelihood function p (Y) _{1:T} |θ ⁽ⁱ⁺¹⁾ ) Posterior probability distribution p (theta) ⁱ⁺¹ |Y _{1:T} ) Proposed distribution q (theta) ⁽ⁱ⁺¹⁾ |θ ⁱ ) Calling a formula:

calculating any independent model parameter theta at the i +1 th moment of the discharge cycle number in the battery degradation model ⁽ⁱ⁺¹⁾ Acceptance probability A (θ) ⁽ⁱ⁺¹⁾ |θ ⁱ ) From the uniform distribution, random numbers U, U-U (0, 1) are generated if U<A(θ ⁽ⁱ⁺¹⁾ |θ ⁱ ) Then receive and releaseParameter theta at the i +1 th moment of the number of electrical cycles ⁱ⁺¹ As the model parameter at the current moment, otherwise, keeping the model parameter at the current moment as the parameter theta at the ith moment ⁱ The value of (c).

13. System according to claim 10 or 11, characterized in that the discharge capacity y at the moment t of the discharge cycle is obtained _t Setting the standard deviation sigma _s Acquiring PDF of Gaussian distribution as a likelihood function, according to a formula:

calculating a fitting value f (t) of a degradation model at the discharge cycle time t and the discharge capacity y _t Likelihood function p (y) _t | θ), a series of discharge capacity observation sequences Y are collected _{1:T} (Y _{1:T} ＝y ₁ ，y _t+1 ......y _T ) According to the formula:

the calculation is based on the total observed sequence Y _{1:T} Overall likelihood function p (Y) _{1:T} | θ), according to the formula:

obtaining the posterior PDF p (theta | Y) of the parameters of the degeneration model _{1:T} )。

14. An electronic device, comprising: a processor; and a memory storing a program, wherein the program comprises instructions that when executed by the processor cause the processor to perform the method of predicting a probability distribution of remaining life of a battery based on MCMC sampling of any of claims 1-9.

15. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method of predicting a remaining life probability distribution of a battery based on MCMC sampling according to any one of claims 1 to 9.

16. A vehicle comprising a system for predicting a probability distribution of remaining life of a battery based on MCMC sampling according to any of claims 10 to 13.

Images