A battery soc anti-interference estimation method, medium and device

Abstract

The application discloses a battery SOC anti-interference estimation method, medium and equipment; the method steps are as follows: a plurality of filtering models are selected to match different charging and discharging conditions of the battery; under the interactive multiple model architecture, input interaction is carried out based on the state estimation of each model at the previous moment and time updating is completed; according to the predicted parameters of each model after time updating and the actual measurement value at the current moment, global measurement noise is estimated; the global noise estimation value is transmitted to each model to update the predicted parameters, the state estimation of each model at the current moment is obtained, and the probability of each model is updated; the state estimation of each model is output interacted based on the updated probability, and is used for estimating the battery SOC at the current moment. The equipment and the medium are used for implementing the above method. The application has the beneficial effects that all the filtering models share a set of global measurement noise estimation, and only a single fusion calculation is needed to serve multiple filtering models at the same time, thereby significantly reducing the calculation complexity of the system.

Description

Technical Field

[0001] This application relates to the field of power electronics technology, and in particular to a method, medium and device for estimating battery SOC immunity. Background Technology

[0002] In actual operation, especially under complex conditions and during aging, the internal resistance, capacitance, and other parameters of the equivalent circuit model of a battery dynamically change. Traditional single-model filtering algorithms such as EKF and UKF cannot accurately describe these changes when estimating the battery's SOC (State of Charge). However, the Interactive Multiple Model (IMM) algorithm can effectively describe the system uncertainties caused by dynamic changes in model parameters and supports flexible adjustment of the number of models, demonstrating strong adaptive variable structure capabilities. Therefore, the combination of IMM and Kalman filtering has been applied to battery SOC estimation and can effectively improve the accuracy and robustness of SOC estimation.

[0003] The effectiveness of the standard IMM combined with Kalman filtering relies on prior and known noise statistics. However, in real-world operating environments, measurement noise often exhibits uncertainty, meaning its statistical characteristics are not prior to knowledge. This interference leads to decreased filtering accuracy, causing estimation bias in IMM applications and consequently affecting the accurate estimation of SOC. To address this issue, existing research improves the filtering process of each model within the IMM into a robust filtering form. Specifically, it introduces an adaptive estimation mechanism for measurement noise for each sub-filtering process within the IMM, performing online correction of the noise covariance to enhance SOC estimation accuracy.

[0004] However, existing robust IMM algorithms require a robust estimation process for each sub-filter, significantly increasing the computational burden and runtime, thus limiting the real-time performance of battery SOC estimation. Furthermore, during noise robust adjustment, existing methods fail to adequately consider the interrelationships between the estimation results of different models, resulting in a lack of coordination in the robust process and affecting the overall fusion effect. Summary of the Invention

[0005] One objective of this application is to provide a battery SOC immunity estimation method that can solve at least one of the defects in the above-mentioned background art.

[0006] Another object of this application is to provide a computer-readable storage medium capable of implementing a battery SOC immunity estimation method that addresses at least one of the deficiencies in the aforementioned background art.

[0007] Another object of this application is to provide an electronic device capable of implementing a battery SOC immunity estimation method that addresses at least one of the deficiencies in the aforementioned background art.

[0008] To achieve at least one of the above objectives, one aspect of this application provides a battery SOC disturbance immunity estimation method, comprising the following steps: selecting multiple filtering models to match the battery state under different operating conditions during charging and discharging; in an interactive multi-model architecture, inputting and interacting with the first state parameters output by the multiple filtering models at the previous time step to update the multiple filtering models over time; estimating global measurement noise based on the predicted parameters of each filtering model after the time update and the actual measurement value of the battery at the current time step; passing the obtained global measurement noise estimate to the multiple filtering models respectively to update the predicted parameters, so that the multiple filtering models respectively output the second state parameters at the current time step and perform corresponding probability updates; based on the updated probability of the filtering models, outputting and interacting with multiple sets of second state parameters to obtain a third state parameter used to estimate the battery SOC at the current time step.

[0009] Preferably, the prediction parameters of the filtering model include the measurement prediction value and the theoretical innovation covariance; wherein, the theoretical innovation covariance is the sum of the measurement prediction covariance and the mixed measurement noise covariance; the estimation of global measurement noise includes the following process: based on the fusion of the measurement prediction values ​​corresponding to multiple filtering models after time update, a global multi-model collaborative mixed measurement prediction value is obtained; the difference between the mixed measurement prediction value and the actual measurement value of the battery is taken as the actual consistency innovation, and the actual consistency innovation covariance is calculated; based on the theoretical innovation covariance of each filtering model, the theoretical consistency innovation covariance is calculated; based on the error between the obtained theoretical consistency innovation covariance and the actual consistency innovation covariance, a robust adjustment rule is designed to adjust the mixed measurement noise covariance to obtain the global measurement noise estimate.

[0010] Preferably, when obtaining the mixed measurement prediction value, the probabilities corresponding to the multiple filtering models at the previous time step are used as weights, and the measurement prediction values ​​corresponding to each filtering model are weighted and averaged to obtain the required mixed measurement prediction value.

[0011] The preferred expression for calculating the theoretical consistency information covariance is: ; In the formula, Let N represent the theoretical consistency information covariance at time k, and let N represent the total number of filtering models. Let represent the probability of the j-th filter model at time k-1. This represents the theoretical innovation covariance of the j-th filter model at time k. This represents the measurement prediction value of the j-th filter model at time k. This represents the predicted value of the mixed measurement at time k.

[0012] Preferably, the process of adjusting the mixed measurement noise covariance based on the robustness adjustment rule is as follows: when the actual consistency information covariance is greater than the theoretical consistency information covariance, the mixed measurement noise covariance is increased; when the actual consistency information covariance is less than the theoretical consistency information covariance, the mixed measurement noise covariance is decreased.

[0013] Preferably, the specific adjustment expression for the mixed measurement noise covariance is as follows: ; In the formula, and Let k and k-1 represent the mixed measurement noise covariance, respectively. Indicates the step size factor. This represents the difference between the actual consistent information covariance and the theoretical consistent information covariance. Indicates adjusting the step size. This represents the predicted covariance of mixed measurements.

[0014] Preferably, for the step size factor The value of is suitable for adaptive adjustment based on the difference between the actual consistent information covariance and the theoretical consistent information covariance. The specific expression is as follows: ; In the formula, and These represent the minimum and maximum values ​​of the step size factor, respectively. This represents the confidence level, and is the difference between the actual consistency covariance and the theoretical consistency covariance. The absolute values ​​of are negatively correlated, and the range of values ​​is [0, 1].

[0015] Preferred confidence level The expression is as follows: ; In the formula, This represents the difference between the actual consistent information covariance and the theoretical consistent information covariance. Let represent the theoretical consistency information covariance at time k.

[0016] Another aspect of this application provides a computer-readable storage medium storing a computer program; when the computer program is executed by a processor, it implements the above-described battery SOC immunity estimation method.

[0017] Another aspect of this application provides an electronic device including a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-described battery SOC immunity estimation method.

[0018] Compared with the prior art, the beneficial effects of this application are as follows: In this application, all filtering models share a common set of global measurement noise estimation. Only a single fusion calculation is needed to serve multiple filtering models simultaneously. There is no need to design an adaptive noise estimator for each filtering model independently, which greatly reduces the amount of online computation and thus significantly reduces the computational complexity of the system. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the equivalent circuit of the filter model used to match the charge and discharge states of the battery in this application. Figure 2 This is a schematic diagram of the overall working steps of this application; Figure 3 This is a schematic diagram illustrating the specific workflow of this application; Figure 4 This is a flowchart illustrating the global measurement noise estimation process for this application. Detailed Implementation

[0020] The present application will now be further described in conjunction with specific embodiments. It should be noted that, in the description of this specification, the use of terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicates that the specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms should not be construed as necessarily referring to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0021] In the description of this application, it should be noted that the terms "center", "lateral", "longitudinal", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc., which indicate the orientation and positional relationship based on the orientation or positional relationship shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and should not be construed as limiting the specific protection scope of this application.

[0022] It should be noted that the terms "first," "second," etc., in the specification and claims of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0023] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., should be interpreted broadly. For example, they can refer to a connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0024] In this application, unless otherwise expressly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature being directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0025] The terms “comprising” and “having”, and any variations thereof, in the specification and claims of this application are intended to cover non-exclusive inclusion, 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 explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or device.

[0026] To facilitate understanding of the technical solution of this application, the specific architecture of the filter model used for matching battery charge and discharge states will be described below. It should be noted that there are various types of filter models that can be used for matching battery charge and discharge states; a second-order RC equivalent circuit model will be used as an example for detailed explanation below.

[0027] like Figure 1 The diagram shown is an equivalent circuit diagram of a battery; in the diagram, V OCV R0 represents the open-circuit voltage of the battery; R1 and R2 both represent the resistance of the polarization effect, and their voltages V1 and V2 are the voltages of the polarization effect; C1 and C2 both represent the capacitance of the polarization effect. t Indicates the battery's terminal voltage; I L This indicates the battery charging current or discharging current. It is a positive value when charging and a negative value when discharging.

[0028] Based on the above equivalent circuit, the discretized state equation and measurement equation of the battery are as follows: .

[0029] .

[0030] Where, x k Represents the battery state vector at time k, SOC k and SOC k-1 V represents the battery state of charge at time k and time k-1, respectively. 1,k and V 2,k Let represent the voltage of the polarization effect at time k, e represent the base of the natural logarithm, and t represent the measurement sampling interval. and Both represent time constants. , V 1,k-1 and V 2,k-1 Let Q represent the voltage at time k-1, η represent the coulomb efficiency, and Q represent the voltage at time k-1. r Indicates the battery's rated capacity, I L,k w represents the battery charging / discharging current at time k. 1,k w 2,k w 3,k Both represent process noise, v k This indicates measurement noise; both process noise and measurement noise are Gaussian white noise and are uncorrelated. k V represents the measurement vector of battery terminal voltage. OCV,k Let x represent the battery open-circuit voltage at time k; where x is the battery state vector. k The initial value is determined by the actual situation.

[0031] Understandably, during the actual charging and discharging process of a battery, its internal parameters, such as internal resistance, polarization resistance, and capacitance, will change significantly with operating conditions, such as temperature and current rate. To more accurately describe this time-varying characteristic, a set of filtering models with different parameters can be designed to match the dynamic behavior of the battery under different operating conditions. The model set can be designed according to the actual battery model, which will not be elaborated here. Subsequently, these filtering models can be applied to the filtering algorithm based on an interactive multi-model architecture to achieve robust estimation of the battery's SOC; for ease of understanding, a detailed description will be provided below.

[0032] One aspect of this application provides a method for estimating battery SOC immunity, such as... Figure 2 and Figure 3 As shown, one preferred embodiment includes the following steps: First, N (N>1) filter models are selected to match the battery's state under different operating conditions during charging and discharging. The total number of filter models N can be chosen based on the number of operating conditions the battery may face during actual charging and discharging. For example, if the battery may face 10 different operating conditions during actual charging and discharging, then the total number of filter models N can be 10. The filter models can be the aforementioned second-order RC equivalent circuit model, or other known equivalent circuit models, depending on the specific needs of those skilled in the art.

[0033] Then, under the Interactive Multi-Model (IMM) architecture, the N filter models are updated in time based on the first state parameters output by the N filter models at the previous time step. The N filter models can be labeled as model #1 to model #N. The first state parameters of the filter models mainly include the estimated state vector x and its error covariance P. Assuming the current time step is k, the state vector x corresponding to the N filter models at the previous time step (k-1) can be represented as x 1,k-1 x 2,k-1 ... x N,k-1 The corresponding error covariance P can be expressed as P 1,k-1 P 2,k-1 ... P N,k-1 After the first state parameter is input and interacted with, the initial mixed state vector x0 and its error covariance P0 can be obtained; then the initial mixed state vector x0 corresponding to each of the N filtering models can be expressed as x 01,k-1 x 02,k-1 ... x 0N,k-1 The corresponding error covariance P can be expressed as P 01,k-1 P 02,k-1 ... P 0N,k-1 The time update of the filtering model mainly refers to extrapolating the state estimate of the filtering model from the previous moment to the current moment. The basis for the extrapolation is the passage of time, rather than the actual measurement value of the battery at the current moment.

[0034] Then, based on the predicted parameters of each filtering model after time update and the actual measurement value of the battery at the current moment, global measurement noise is estimated. The predicted parameters of the filtering model mainly include the measurement prediction value and the theoretical innovation covariance. The measurement prediction value refers to the predicted value of the battery terminal voltage after the filtering model completes the time update; the difference between the predicted value and the actual measurement value of the battery terminal voltage is the innovation, and the theoretical innovation covariance is the theoretical covariance matrix of the innovation sequence corresponding to the filtering model, which is equal to the sum of the measurement prediction covariance and the mixed measurement noise covariance.

[0035] Then, the obtained global measurement noise estimates are passed to N filtering models to update the prediction parameters, so that each of the N filtering models outputs the second state parameters at the current time and performs the corresponding probability update. The second state parameters mainly include the model-estimated state vector x and its error covariance P; therefore, the state vector x corresponding to the N filtering models at the current time k can be represented as x 1,k x 2,k ... x N,k The corresponding error covariance P can be expressed as P 1,k P 2,k ... P N,k The probability of a filtering model mainly refers to the probability that the filtering model can correctly describe the current battery state. The sum of the probabilities of all filtering models equals 1, or 100%. As the battery charging and discharging process continues, the real-time operating conditions of the battery will change. This will cause a previously matched filtering model to become mismatched, while another previously mismatched filtering model may become a match. This will cause the probabilities of each filtering model to change. Therefore, the probabilities of the filtering models need to be updated at each time step.

[0036] Finally, based on the probabilities of the updated filtering model, the N sets of second-state parameters are output and interacted to obtain the third-state parameters used to estimate the battery SOC at the current moment. The third-state parameters mainly include the state vector x estimated by the entire IMM architecture. k and its error covariance P k That is, using the probabilities of the updated filtering model as weights, the state vectors x corresponding to the second state parameters and their error covariance P are weighted and fused to obtain the desired state vector x. k and its error covariance P k Specifically, in the IMM architecture, each filtering model only provides a state estimate based on its own model. The final battery SOC estimation is achieved by fusing the estimates from all filtering models into a single output. More specifically, the first and second state parameters each include N state vectors and their error covariances, while the third state parameter includes only one state vector and its error covariance.

[0037] It is understood that all filtering models in this application share a set of global measurement noise estimation, which can serve multiple filtering models simultaneously with a single fusion calculation. There is no need to design an adaptive noise estimator for each filtering model independently, which can significantly reduce the amount of online computation and thus significantly reduce the computational complexity of the system.

[0038] It should be noted that the input interaction based on the first state parameter in the technical solution of this application is a conventional technique used by those skilled in the art, and therefore will not be described in detail here. That is, the technical solution of this application mainly focuses on the improvement of the IMM architecture. By changing the traditional structure of N models moving in parallel with N sets of complex measurement noise estimation to N models sharing a single lightweight noise estimation structure, it can greatly reduce redundant calculations, improve the overall efficiency of the algorithm, and enhance the real-time performance in resource-constrained scenarios. Therefore, the core of the technical solution of this application lies in the estimation of global measurement noise. For ease of understanding, the estimation process of global measurement noise will be described in detail below.

[0039] In a specific embodiment, such as Figure 4 As shown, the estimation of global measurement noise includes the following process: Step 1: Based on the fusion of the measurement prediction values ​​corresponding to N filtering models after time update, a global multi-model collaborative hybrid measurement prediction value is obtained.

[0040] Understandably, after completing the time update, each filtering model will calculate the predicted value of the battery terminal voltage at the current time k based on its own model, i.e., the measurement prediction value. Since the prediction accuracy of different filtering models varies under different battery operating conditions, directly using the prediction result of a single filtering model will limit the subsequent noise estimation to the assumptions of that model. Therefore, it is necessary to fuse the measurement prediction values ​​calculated by all filtering models after the time update.

[0041] Step 2: The difference between the predicted value of the mixed measurement and the actual measurement value of the battery is taken as the actual consistency information. Based on the obtained actual consistency information, it is compared with the actual measurement value of the battery terminal voltage. The difference covariance between the prediction of the filtering model and the actual measurement can be calculated, that is, the actual consistency information covariance. For ease of understanding, the actual consistency information will be expressed by an expression below. and actual consistency of new information covariance To express.

[0042] ; .

[0043] In the formula, z k This represents the actual measured value of the battery terminal voltage at time k. This represents the predicted value of the mixed measurement after the time update (i.e., at the current time k).

[0044] Understandably, traditional adaptive filtering often uses the average covariance of the innovation sequence within a time window to estimate noise. However, the window length leads to detection lag and requires a trade-off between response speed and estimation stability. In this scheme, the instantaneous values ​​of the innovation and its covariance are calculated based on the current prediction and measurement information. This allows for the immediate capture of step changes in noise, enabling adjustments to be initiated immediately after the noise undergoes a step change. This eliminates the detection and response lag of at least one window length inherent in traditional time-window methods, achieving real-time tracking of dynamic measurement noise. Furthermore, the elimination of the need to store historical innovation sequences reduces computational load and avoids the trade-offs between window length and response speed required by time-window methods.

[0045] Step 3: Calculate the theoretical consistency innovation covariance based on the theoretical innovation covariance of each filtering model. Based on the error between the obtained theoretical consistency innovation covariance and the actual consistency innovation covariance, design a robust adjustment rule to adjust the mixed measurement noise covariance and obtain the global measurement noise estimate.

[0046] Understandably, if the theoretical consistency information covariance estimated by the filtering model matches the actual consistency information covariance, it indicates that the current filtering model is operating in an optimal state; if there is a significant difference, it indicates that there is an error in the assumption of measurement noise, and the measurement noise needs to be adjusted, that is, the mixed measurement noise covariance needs to be adjusted.

[0047] Step 4: After adjusting the mixed measurement noise covariance, this unique, real-time updated covariance is distributed to all filtering models within the IMM architecture.

[0048] Understandably, unified noise estimation ensures that all filtering models have a consistent understanding of the current measurement environment, avoiding potential conflicts caused by asynchronous noise judgments among different filtering models, thereby improving the overall coordination of the IMM architecture and its robustness under complex interference.

[0049] In a specific example, the calculation of the mixed measurement prediction value in the first step above is generally carried out by weighted fusion. The weights corresponding to each filter model can be assigned by those skilled in the art. However, considering that the probability of the filter model can intuitively reflect the confidence level of the filter model in predicting the current battery state, the probability corresponding to each filter model can be directly used as the weight to calculate the mixed measurement prediction value. That is, by using the probabilities corresponding to N filter models after time update as weights, the measurement prediction values ​​corresponding to each filter model are fused by weighted average to obtain the required mixed measurement prediction value.

[0050] Understandably, the single-time filtering model only performs one probability update, and this update occurs after the global measurement noise estimate has been distributed. Therefore, when calculating the mixed measurement prediction after the time update of the filtering model, the probability of the time-updated filtering model is the same as the probability of the filtering model at the previous time. The mixed measurement prediction obtained by probability-weighted fusion is the result of all filtering models collectively predicting the current measurement based on information from the previous time. This mixed measurement prediction not only serves as the basis for subsequent consistency information calculation but also essentially reflects the relative confidence and interaction results of each model in the model set at the current time.

[0051] In a specific example, the calculation of the theoretical consistency information covariance in step three above can be derived using the internal formula of the Kalman filter algorithm built into the filtering model, combined with the preset statistical characteristics of measurement noise. Specifically, the theoretical consistency information covariance is obtained by weighted summation of the theoretical information covariances of each filtering model, and the corresponding weights for each filtering model can still be represented by the probabilities corresponding to the filtering models. For ease of understanding, the theoretical consistency information covariance will be expressed as an expression below.

[0052] .

[0053] In the formula, Let N represent the theoretical consistency information covariance at time k, and let N represent the total number of filtering models. This represents the probability of the j-th filter model at time k-1. Since the filter model has not yet been updated after the time update, the probability of the filter model at this time is equal to the probability at time k-1. This represents the theoretical innovation covariance of the j-th filter model after the time update (current time k). This represents the measurement prediction value of the j-th filter model after the time update (at the current time k). This represents the predicted value of the mixed measurement after the time update (at the current time k).

[0054] It is understandable that the theoretical innovation covariance is equal to the sum of the measurement prediction covariance and the measurement noise covariance; the specific expression is: .

[0055] In the formula, This represents the measurement prediction covariance of the j-th filter model after the time update (current time k), used to evaluate the reliability of the measurement prediction value. Specifically, it is obtained by propagating the error covariance of the state vector through the known measurement matrix. This represents the measurement noise covariance of the filter model at time k-1.

[0056] It is important to note that since all filtering models share a single global measurement noise estimate, therefore It can also be used for the mixed measurement noise covariance of all filtering models at time k-1. Based on the formula for calculating the theoretically consistent innovation covariance, the mixed measurement noise covariance can be... Once extracted, the formula for calculating the theoretical consistency information covariance can be transformed into: .

[0057] .

[0058] In the formula, This represents the theoretical consistency information covariance at time k. This represents the predicted covariance of the mixed measurements, and N represents the total number of filtering models; Let represent the probability of the j-th filter model at time k-1. This represents the theoretical innovation covariance of the j-th filter model after the time update (current time k). This represents the measurement prediction value of the j-th filter model after the time update (at the current time k). This represents the predicted value of the mixed measurement after the time update (at the current time k).

[0059] In a specific example, based on the foregoing, the deviation between the actual consistent information covariance and the theoretical consistent information covariance directly reflects the difference between the currently assumed noise level and the actual noise level. Therefore, the process of adjusting the mixed measurement noise covariance based on the robustness adjustment rule is as follows: When the actual consistent information covariance is greater than the theoretical consistent information covariance, it indicates that the fluctuation of the information exceeds the theoretical expectation, that is, the preset mixed measurement noise covariance is too small and fails to cover the uncertainty in the actual measurement; therefore, in order to match the real environment, it is necessary to increase the mixed measurement noise covariance to improve the theoretical information covariance. When the actual consistent information covariance is less than the theoretical consistent information covariance, it indicates that the fluctuation of the information is less than the theoretical expectation, that is, the preset mixed measurement noise covariance is too large and excessively amplifies the uncertainty in the actual measurement; therefore, in order not to be too conservative and sacrifice the filtering convergence speed, it is necessary to decrease the mixed measurement noise covariance to reduce the theoretical information covariance. When the actual consistent information covariance is equal to the theoretical consistent information covariance, it indicates that the preset measurement noise covariance is reasonable and there is no need to adjust the mixed measurement noise covariance.

[0060] Understandably, when adjusting the mixed measurement noise covariance, the adjustment step size depends on the degree of deviation between the actual consistent information covariance and the theoretical consistent information covariance; that is, the greater the deviation, the longer the adjustment step size. For ease of understanding, the adjustment process for the mixed measurement noise covariance will be represented below using specific expressions.

[0061] ; In the formula, and These represent the mixed measurement noise covariance at time k and time k-1, respectively, which correspond to the mixed measurement noise covariance before and after adjustment. This represents the difference between the actual consistent information covariance and the theoretical consistent information covariance. Its value determines the adjustment direction of the mixed measurement noise covariance. At this time, it is necessary to increase the covariance of mixed measurement noise. At this time, it is necessary to reduce the covariance of mixed measurement noise; This indicates the adjustment step size, with a value of 1 or -1; that is... hour, , hour, ; Indicates the step size factor. This represents the predicted covariance of mixed measurements.

[0062] Understandable, This is a key safety constraint mechanism that determines the benchmark for adjustment magnitude. Based on this safety constraint mechanism, the adjustment amount within an estimation step can be strictly limited to the current prediction uncertainty level, i.e., not exceeding the current mixed-measurement prediction covariance. Since the actual consistent information covariance is an instantaneous value, by adding a safety constraint mechanism, the impact of a single outlier or extreme measurement on the noise estimation can be effectively constrained. This enhances the stability of the system under non-Gaussian or intermittent interference environments and avoids an excessive difference between the actual and theoretical consistent information covariance, which would lead to an excessively large adjusted mixed measurement noise covariance. This would cause the filtering model to distrust the measurement values ​​excessively for a long period afterward, resulting in performance degradation of the filtering model.

[0063] Specifically, based on the design of the security constraint mechanism, when the difference between the actual consistent information covariance and the theoretical consistent information covariance is too large, the covariance is predicted only using the current mixed measurement. To adjust the step size, make tentative adjustments. The upper limit for adjustment is the predicted covariance of the mixed measurements. Non-theoretical consistency information covariance The core principle is to ensure that the magnitude of a single-step adjustment never exceeds the prediction uncertainty level of the system's preset model at the current moment, thus providing a physically reasonable and highly effective defense against outlier shocks. If theoretical consistency information covariance is used... The mixed measurement noise covariance at the previous time step This approach avoids inappropriately amplifying the adjustment upper limit, which could lead to over-adjustment and disrupt the steady-state equilibrium of the filtering model if the difference between the actual consistent information covariance and the theoretical consistent information covariance becomes too large. This design ensures that the adjustment behavior strictly focuses on current prediction information rather than historical values, thus achieving a balance between rapid adaptation and stable smoothness, and endowing the algorithm with strong resistance to outliers.

[0064] What needs to be known is the step size factor. This value controls the smoothness and speed of the adjustment update for the mixed measurement noise covariance. A smaller value results in a more conservative and smooth adjustment update, with slower tracking of noise changes but greater stability. A larger value results in a more sensitive and faster adjustment update, but is also more susceptible to transient fluctuations. Therefore, in this embodiment, the step size factor... The value of can be adaptively adjusted based on the difference between the actual consistent information covariance and the theoretical consistent information covariance. The specific expression is as follows: .

[0065] In the formula, and These represent the minimum and maximum values ​​of the step size factor, respectively, and the specific values ​​can be set according to the actual needs of those skilled in the art. This represents the confidence level, and is the difference between the actual consistency covariance and the theoretical consistency covariance. The absolute values ​​of are negatively correlated, and the range of values ​​is [0, 1].

[0066] Specifically, when the difference When the absolute value is large, the confidence level When the value of approaches 0, it indicates that the current estimate is not trusted, and the step size factor... When the value of approaches its maximum, a larger step size factor is used for rapid correction; when the difference... When the absolute value is very small, the confidence level When the value of approaches 1, it indicates confidence in the current estimate; at this point, the step size factor... When the value of approaches its minimum, fine-tuning or maintaining it using a smaller step size factor is sufficient. Based on large step size adjustments for large errors and small step size adjustments for small errors, a balance between rapid response and steady-state smoothness can be effectively achieved.

[0067] Understandably, confidence level The value and difference It can be a linear correlation or a non-linear correlation. For a linear correlation, a difference can be set. The absolute value of the upper and lower limits, then the difference The absolute value within the upper and lower limits and the confidence level The range of values ​​corresponds linearly. However, considering that the state change of the battery is non-linear, the confidence level is not specified in this embodiment. The value and difference The relationship can preferably be nonlinear; a specific example of the calculation expression is as follows: .

[0068] In the formula, This represents the difference between the actual consistent information covariance and the theoretical consistent information covariance. Let represent the theoretical consistency information covariance at time k.

[0069] Another aspect of this application provides a computer-readable storage medium, in a preferred embodiment of which a computer program is stored on the storage medium; when the computer program is executed by a processor, it implements the above-described battery SOC immunity estimation method.

[0070] Another aspect of this application provides an electronic device, in a preferred embodiment of which includes a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-described battery SOC immunity estimation method.

[0071] The basic principles, main features, and advantages of this application have been described above. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely the principles of this application. Various changes and modifications can be made to this application without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claims. The scope of protection claimed by this application is defined by the appended claims and their equivalents.

Claims

1. A method for estimating battery SOC immunity, characterized in that, Includes the following steps: Multiple filtering models were selected to match the state of the battery under different operating conditions during the charging and discharging process; In the interactive multi-model architecture, the input interaction is based on the first state parameters output by multiple filtering models at the previous time step, and the multiple filtering models are updated in time. Based on the predicted parameters of each filtering model after time update and the actual measurement value of the battery at the current moment, the global measurement noise is estimated. The obtained global measurement noise estimate is passed to multiple filtering models to update the prediction parameters, so that the multiple filtering models output the second state parameters at the current time and perform the corresponding probability update. Based on the probability of the updated filtering model, multiple sets of second state parameters are output and interacted to obtain third state parameters used to estimate the battery SOC at the current moment.

2. The battery SOC disturbance immunity estimation method as described in claim 1, characterized in that, The prediction parameters of the filtering model include the measurement prediction values ​​and the theoretical innovation covariance; whereby the theoretical innovation covariance is the sum of the measurement prediction covariance and the mixed measurement noise covariance; the estimation of global measurement noise includes the following process: Based on the fusion of the measurement prediction values ​​corresponding to multiple filtering models after time update, a global multi-model collaborative hybrid measurement prediction value is obtained; the difference between the hybrid measurement prediction value and the actual measurement value of the battery is used as the actual consistency information, and the covariance of the actual consistency information is calculated. Theoretical consistency covariance is calculated based on the theoretical innovation covariance of each filtering model. Based on the error between the obtained theoretical consistency covariance and the actual consistency covariance, a robust adjustment rule is designed to adjust the mixed measurement noise covariance, thereby obtaining the global measurement noise estimate.

3. The battery SOC disturbance rejection estimation method as described in claim 2, characterized in that, When obtaining the mixed measurement prediction value, the probabilities corresponding to the multiple filtering models at the previous time step are used as weights, and the measurement prediction values ​​corresponding to each filtering model are weighted and averaged to obtain the required mixed measurement prediction value.

4. The battery SOC disturbance immunity estimation method as described in claim 3, characterized in that, The expression for calculating the theoretical consistency information covariance is as follows: ； In the formula, Let N represent the theoretical consistency information covariance at time k, and let N represent the total number of filtering models. Let represent the probability of the j-th filter model at time k-1. This represents the theoretical innovation covariance of the j-th filter model at time k. This represents the measurement prediction value of the j-th filter model at time k. This represents the predicted value of the mixed measurement at time k.

5. The battery SOC immunity estimation method as described in any one of claims 2-4, characterized in that, The process of adjusting the mixed measurement noise covariance based on the robust adjustment rule is as follows: when the actual consistent information covariance is greater than the theoretical consistent information covariance, the mixed measurement noise covariance is increased; when the actual consistent information covariance is less than the theoretical consistent information covariance, the mixed measurement noise covariance is decreased.

6. The battery SOC disturbance rejection estimation method as described in claim 5, characterized in that, The specific adjustment expression for the mixed measurement noise covariance is as follows: ； In the formula, and Let k and k-1 represent the mixed measurement noise covariance, respectively. Indicates the step size factor. This represents the difference between the actual consistent information covariance and the theoretical consistent information covariance. Indicates adjusting the step size. This represents the predicted covariance of mixed measurements.

7. The battery SOC immunity estimation method as described in claim 6, characterized in that, For step size factor The value of is suitable for adaptive adjustment based on the difference between the actual consistent information covariance and the theoretical consistent information covariance. The specific expression is as follows: ； In the formula, and These represent the minimum and maximum values ​​of the step size factor, respectively. This represents the confidence level, and is the difference between the actual consistency covariance and the theoretical consistency covariance. The absolute values ​​of are negatively correlated, and the range of values ​​is [0, 1].

8. The battery SOC disturbance immunity estimation method as described in claim 7, characterized in that, Confidence The expression is as follows: ； In the formula, This represents the difference between the actual consistent information covariance and the theoretical consistent information covariance. Let represent the theoretical consistency information covariance at time k.

9. An electronic device, characterized in that, It includes a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program to implement the battery SOC disturbance rejection estimation method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program; when the computer program is executed by a processor, it implements the battery SOC disturbance rejection estimation method as described in any one of claims 1-8.

Images