Working performance detection method for power storage battery of new energy automobile

Abstract

The invention relates to the technical field of new energy batteries, in particular to a working performance detection method for a power storage battery of a new energy automobile, and the method comprises the steps: obtaining a voltage residual sequence at a current sampling moment according to terminal voltage collection values and terminal voltage prediction values at a preset number of historical sampling moments before the current sampling moment; obtaining a first optimization factor and a second optimization factor for optimizing a process noise covariance matrix in an unscented Kalman filtering algorithm according to the voltage residual sequence; obtaining an optimized process noise covariance matrix by using the first optimization factor and the second optimization factor; according to the terminal voltage collection value at the current sampling moment and the optimized process noise covariance matrix, the SOC estimation value of the power storage battery at the current sampling moment is obtained and used for detecting the working performance of the power storage battery, and the estimation accuracy of the SOC value of the power storage battery is achieved by adjusting the process noise covariance matrix online.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy batteries, and in particular to a method for detecting the working performance of a power battery for a new energy vehicle. Background Art

[0002] As the global energy crisis and environmental pollution become increasingly serious, new energy vehicles have become an inevitable trend in the development of the automobile industry due to their energy-saving and environmentally friendly characteristics. Lithium iron phosphate batteries have been widely used in the field of new energy vehicles due to their high safety, long cycle life, and low cost. SOC (state of charge, which refers to the ratio of the remaining capacity of a battery to its capacity when it is fully charged) is an important parameter reflecting the remaining power of the battery and is crucial to the safe and efficient operation of the battery. Accurately estimating the SOC value of the battery helps optimize the battery management strategy, prevent overcharging and over-discharging, extend the battery life, and improve the accuracy of the range estimation of electric vehicles.

[0003] At present, the main methods for estimating the SOC value of batteries include the ampere-hour integration method, the open circuit voltage method, the neural network method, the Kalman filter method, etc. Among them, the ampere-hour integration method is simple and easy to use, but there are problems of cumulative error and initial value dependence; the open circuit voltage method requires a long period of static state and is not suitable for online estimation; the neural network method requires a large amount of data for training and the model generalization ability is limited; the Kalman filter (KF) and its extended algorithms (extended Kalman filter EKF, unscented Kalman filter UKF) can effectively handle nonlinear systems and have certain noise resistance, so they have been widely used in the field of battery SOC estimation.

[0004] The Unscented Kalman Filter (UKF) algorithm processes nonlinear systems through the Unscented Transformation (UT), avoiding the process of linearizing nonlinear functions (calculating the Jacobian matrix) in the EKF algorithm. Therefore, it has higher estimation accuracy and computational efficiency. The UKF algorithm has become one of the research hotspots in the field of battery SOC estimation. Although the UKF algorithm has shown advantages in battery SOC estimation, its performance is highly dependent on the settings of the process noise covariance matrix Q and the measurement noise covariance matrix R.

[0005] In practical applications, the parameters of the battery model (such as internal resistance and capacitance) will change with factors such as battery aging, temperature changes, and charge and discharge status, showing obvious time-varying characteristics. In addition, the statistical characteristics of the measurement noise are not constant, but show non-stationarity over time. However, the traditional UKF algorithm usually sets Q and R to fixed values, which cannot adapt to the time-varying battery model parameters and the non-stationarity of measurement noise. When the battery model parameters or noise characteristics do not match the preset values, the fixed Q value will cause the state estimation process of the UKF algorithm to fail to accurately reflect the prediction uncertainty of the battery model, which in turn causes the battery SOC estimation result to deviate from the true value, resulting in a non-negligible estimation bias.

[0006] Therefore, how to adaptively adjust the process noise covariance matrix Q online to improve the SOC estimation accuracy of lithium iron phosphate batteries based on the UKF algorithm has become an urgent problem to be solved. Summary of the invention

[0007] In view of this, an embodiment of the present invention provides a method for detecting the working performance of a power battery for new energy vehicles, so as to solve the problem of how to perform online adaptive adjustment on the process noise covariance matrix Q to improve the accuracy of SOC estimation of lithium iron phosphate batteries based on the UKF algorithm.

[0008] In an embodiment of the present invention, a method for detecting the working performance of a power battery for a new energy vehicle is provided, the method comprising the following steps: Obtain the terminal voltage collection value of the power storage battery at the current sampling moment, and obtain the voltage residual sequence at the current sampling moment based on the terminal voltage collection values ​​and terminal voltage prediction values ​​at a preset number of historical sampling moments before the current sampling moment; According to the mean value and the cumulative amount of changes of the voltage residual sequence, a first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained; According to the residual fluctuation change of the voltage residual sequence, a second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained; Using the first optimization factor and the second optimization factor, the process noise covariance matrix at the previous sampling moment of the current sampling moment is updated and optimized to obtain the optimized process noise covariance matrix at the current sampling moment; According to the terminal voltage acquisition value at the current sampling moment and the optimized process noise covariance matrix, the unscented Kalman filter algorithm is used to obtain the SOC estimation value of the power battery at the current sampling moment, which is used to detect the working performance of the power battery.

[0009] Preferably, the step of obtaining the voltage residual sequence at the current sampling moment according to the terminal voltage collected values ​​and the terminal voltage predicted values ​​at a preset number of historical sampling moments before the current sampling moment includes: For any historical sampling moment, the previous historical sampling moment of the said historical sampling moment is obtained, and the difference between the terminal voltage acquisition value of the said historical sampling moment and the terminal voltage prediction value of the said previous historical sampling moment is calculated as the voltage residual of the said historical sampling moment; the voltage residual of each historical sampling moment is obtained to form a voltage residual sequence of the current sampling moment.

[0010] Preferably, obtaining a first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm according to the mean and cumulative variation of the voltage residual sequence includes: According to the mean of the voltage residual sequence and the difference of adjacent residuals, a voltage prediction deviation index is obtained; the nominal voltage of the power battery is obtained, the voltage prediction deviation index is normalized by using the nominal voltage to obtain a normalized value, the normalized value is nonlinearly mapped by using a hyperbolic tangent function to obtain a corresponding first mapping value, and according to a constant 1 and the sum of the first mapping value, a first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained.

[0011] Preferably, obtaining the voltage prediction deviation index according to the mean of the voltage residual sequence and the difference between adjacent residuals includes: The mean of the voltage residual sequence is calculated, recorded as the voltage residual mean, the absolute value of the difference between every two adjacent voltage residuals in the voltage residual sequence is calculated, and the accumulated value of the absolute value of the difference is obtained, recorded as the voltage change accumulation amount, and the voltage prediction deviation index is obtained according to the sum of the absolute value of the voltage residual mean and the absolute value of the voltage change accumulation amount.

[0012] Preferably, the step of obtaining a second optimization factor for optimizing a process noise covariance matrix in an unscented Kalman filter algorithm according to the residual fluctuation change of the voltage residual sequence includes: Calculating the standard deviation of the voltage residual sequence to obtain a first ratio between the standard deviation and the nominal voltage; obtaining an adaptive autocorrelation coefficient of the voltage residual sequence; obtaining a second ratio between the absolute value of the voltage residual mean and the nominal voltage; Calculate the sum of constant 1, the absolute value of the adaptive autocorrelation coefficient and the second ratio to obtain the product of the sum and the first ratio, use a hyperbolic tangent function to perform nonlinear mapping on the product to obtain a corresponding second mapping value, and obtain a second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm based on the sum of constant 1 and the second mapping value.

[0013] Preferably, the step of obtaining the adaptive autocorrelation coefficient of the voltage residual sequence includes: According to a preset lag order, the lagged voltage residual corresponding to each voltage residual in the voltage residual sequence is obtained respectively, the product between each voltage residual in the voltage residual sequence and its corresponding lagged voltage residual is accumulated to obtain a first accumulated value, the square of each voltage residual in the voltage residual sequence is accumulated to obtain a second accumulated value, and the adaptive autocorrelation coefficient of the voltage residual sequence is obtained according to the ratio of the first accumulated value to the second accumulated value.

[0014] Preferably, the updating and optimizing of the process noise covariance matrix at the previous sampling moment of the current sampling moment by using the first optimization factor and the second optimization factor to obtain the optimized process noise covariance matrix at the current sampling moment includes: The first optimization factor and the second optimization factor are weightedly summed to obtain an adjustment coefficient, and the product of the adjustment coefficient and the process noise covariance matrix at a previous sampling moment of the current sampling moment is used as the optimized process noise covariance matrix at the current sampling moment.

[0015] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The present invention uses the mean value and cumulative change amount of the voltage residual sequence to evaluate the average difference between the predicted value and the measured value of the battery model and the persistence of the difference, so as to obtain the first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm according to the mean value and cumulative change amount of the voltage residual sequence. At the same time, in order to reduce the random fluctuations existing in the battery model prediction, the second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained according to the fluctuation amplitude, periodicity and trend of the voltage residual sequence. Then, the process noise covariance matrix in the operation process of the unscented Kalman filter algorithm is optimized in real time by combining the first optimization factor and the second optimization factor, so as to effectively reduce the influence of the time-varying parameters of the battery model and the non-stationarity of the measurement noise on the SOC estimation accuracy by adjusting the process noise covariance matrix (that is, the optimized process noise covariance matrix) online, thereby achieving the estimation accuracy of the SOC value of the power battery, and thus reducing the error of the working performance detection of the power battery based on the SOC estimation value. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0017] Figure 1 It is a method flow chart of a method for detecting the working performance of a power battery for a new energy vehicle provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0018] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0019] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0020] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0021] See also Figure 1 , is a method flow chart of a method for detecting the working performance of a power battery for a new energy vehicle provided in Embodiment 1 of the present invention, such as Figure 1 As shown, the method may include: Step S101, obtaining a terminal voltage acquisition value of the power storage battery at a current sampling moment, and obtaining a voltage residual sequence at the current sampling moment according to terminal voltage acquisition values ​​and terminal voltage prediction values ​​at a preset number of historical sampling moments before the current sampling moment.

[0022] In the battery state of charge (SOC) estimation process based on the unscented Kalman filter algorithm (UKF algorithm), the operation of the UKF algorithm depends on an accurate battery model, which is used to describe the relationship between the battery voltage, current, SOC value and temperature. The parameters of the battery model include: the relationship between the open circuit voltage OCV and SOC (that is, the OCV-SOC curve), ohmic internal resistance (R0), polarization internal resistance (Rₚ) and polarization capacitance (Cₚ), which are all functions of SOC and temperature. Among them, the battery model commonly used is the Thevenin equivalent circuit model, which belongs to the prior art and will not be described in detail here.

[0023] It is worth noting that the process noise covariance matrix Q in the UKF algorithm directly affects the trust weight between the UKF algorithm's predicted value and the measured value of the battery model. If Q is set to a small value, the UKF algorithm is more inclined to trust the predicted result of the battery model; if Q is set to a large value, the UKF algorithm is more inclined to trust the actual measured value. The traditional UKF algorithm usually sets Q to a fixed value, and this approach is based on the assumption that the parameters of the battery model are accurate enough and the measurement noise is stable. However, in practical applications, these two assumptions are often difficult to meet. The parameters of the battery model will change with the changes in factors such as the aging of the battery, ambient temperature, and charge and discharge rate, showing obvious time-varying characteristics; the measurement noise (voltage sensor noise) may also be affected by various factors and show non-stationarity.

[0024] Therefore, if the parameters of the battery model change, if a fixed Q value is still used, the UKF algorithm will not be able to accurately evaluate the uncertainty of the battery model prediction. Specifically: if there is a deviation between the parameters of the battery model and the actual value, there will be a systematic deviation between the battery terminal voltage predicted by the battery model and the actual measured value. This systematic deviation will not only be reflected in the mean of the residual (i.e., the measured value minus the predicted value), but also in the changing trend of the residual sequence. If only the residual mean is relied upon to determine whether there is a systematic deviation, missed judgments or misjudgments may occur. For example, when the parameters of the battery model change rapidly or oscillate, the residual sequence may alternate rapidly between positive and negative, causing the residual mean to be close to zero, masking the deviation of the battery model prediction. In order to capture the systematic deviation of the battery model prediction more comprehensively and accurately, an embodiment of the present invention provides a Q adaptive adjustment strategy based on the voltage residual sequence, so that the estimation result of the battery SOC value based on the UKF algorithm is more in line with the actual usage scenario.

[0025] Specifically, in the embodiment of the present invention, during the operation of the power battery, a high-precision voltage sensor is used to collect the terminal voltage of the power battery at each sampling moment, which is recorded as the terminal voltage collection value, wherein the sampling frequency of the voltage sensor is set to 10Hz to meet the monitoring requirements for the dynamic characteristics of the battery. The moment when the SOC value of the power battery needs to be analyzed in real time is taken as the current sampling moment, recorded as moment t, and the terminal voltage collection value at the current sampling moment is obtained. At the same time, the terminal voltage acquisition values ​​and terminal voltage prediction values ​​of N historical sampling moments before the current sampling moment are obtained, so as to obtain the voltage residual sequence of the current sampling moment according to the terminal voltage acquisition values ​​and terminal voltage prediction values ​​of a preset number of historical sampling moments before the current sampling moment, and to obtain the adaptive Q value corresponding to the current sampling moment when estimating the SOC value of the power battery at the current sampling moment by using the UKF algorithm. Considering that the voltage residual sequence is too long, its data is redundant, the effective information is less, and the calculation amount is large, while the voltage residual sequence is too short and the systematic deviation of the battery model prediction cannot be observed, therefore, N is set to 300, and no restriction is made here; the terminal voltage prediction value of each historical sampling moment is obtained by the battery model prediction, which belongs to the prior art and will not be described in detail here.

[0026] Among them, the method for obtaining the voltage residual sequence at the current sampling moment based on the terminal voltage collection values ​​and the terminal voltage prediction values ​​of a preset number of historical sampling moments before the current sampling moment is: for any historical sampling moment, obtain the previous historical sampling moment of any historical sampling moment, calculate the difference between the terminal voltage collection value of any historical sampling moment and the terminal voltage prediction value of the previous historical sampling moment, as the voltage residual of any historical sampling moment; obtain the voltage residual of each historical sampling moment to form the voltage residual sequence at the current sampling moment.

[0027] At this point, based on the difference between the model predicted value and the measured value at the historical sampling time before the current sampling time, the voltage residual sequence within a period of time before the current sampling time is obtained.

[0028] Step S102: obtaining a first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm according to the mean value and the cumulative amount of changes of the voltage residual sequence.

[0029] The voltage residual is a key variable in the UKF algorithm, which reflects the difference between the predicted value and the measured value of the battery model. Ideally, when the battery model is completely accurate and there is no measurement noise, the voltage residual should be 0. However, in practical applications, due to the existence of factors such as model errors and measurement noise, the voltage residual is usually not 0. Therefore, the voltage residual sequence at the current sampling moment includes the prediction error information of the battery model, which is also an important basis for evaluating the prediction accuracy of the battery model. Therefore, in the embodiment of the present invention, based on the voltage residual sequence at the current sampling moment, the two indicators of the voltage residual mean and the voltage cumulative change are introduced at the same time to obtain the first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm, which is used to adaptively realize the adaptive adjustment optimization of the process noise covariance matrix Q to cope with the systematic deviation caused by the time-varying parameters of the battery model.

[0030] Specifically, the mean of the voltage residual sequence is calculated and recorded as the voltage residual mean , used to characterize the systematic deviation of the battery model prediction, if the voltage residual mean Close to 0, indicating that the battery model prediction has no obvious deviation in the average sense, that is, the errors between the predicted value and the measured value are positive and negative, offsetting each other, showing the characteristics of random fluctuations. If the voltage residual mean is significantly greater than 0, it means that the predicted value of the battery model is continuously low (the residual is continuously positive), and there is a negative systematic deviation. It is significantly less than 0, indicating that the predicted value of the battery model continues to be high (the residual continues to be negative) and there is a positive systematic deviation.

[0031] Calculate the absolute value of the difference between every two adjacent voltage residuals in the voltage residual sequence, and obtain the accumulated value of the absolute value of the difference, which is recorded as the voltage change accumulation , which is used to reflect the severity and persistence of the change in the voltage residual sequence. Close to 0, the voltage residual sequence may also have large fluctuations. If the parameters of the battery model change rapidly or oscillate, the voltage residual sequence may quickly alternate between positive and negative, resulting in the voltage residual mean Close to 0, but the change in voltage residual will be larger, and the cumulative voltage change It can capture this changing trend and make up for the voltage residual mean Deficiencies.

[0032] If the mean voltage residual is close to 0, and the accumulated voltage change is also small, it means that the battery model prediction has no obvious systematic deviation, and Q can be kept unchanged at this time; if the mean voltage residual deviates significantly from zero, or the accumulated voltage change is large, it means that the battery model prediction has a systematic deviation, and Q should be increased at this time to make the UKF algorithm rely more on the measured value for correction, thereby reducing the SOC estimation deviation. Therefore, according to the sum of the absolute value of the mean voltage residual and the absolute value of the accumulated voltage change, the voltage prediction deviation index is obtained to reflect the degree of systematic deviation in the battery model prediction. Preferably, the sum of the absolute value of the mean voltage residual and the absolute value of the accumulated voltage change is used as the voltage prediction deviation index. , which can ensure that the battery model's prediction is either overestimated or underestimated. As long as there is a systematic deviation or a large cumulative voltage change, Q needs to be increased.

[0033] Furthermore, after obtaining the voltage prediction deviation index, the nominal voltage of the power battery is obtained to normalize the voltage prediction deviation index to make it a dimensionless value, and then the voltage prediction deviation index is normalized using the nominal voltage to obtain a normalized value, eliminating the influence of different battery types and different voltage levels. Then, the normalized value is nonlinearly mapped using a hyperbolic tangent function to obtain a corresponding first mapping value, and the first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained based on the constant 1 and the sum of the first mapping value.

[0034] In one embodiment, the calculation expression of the first optimization factor is:

[0035] in, represents the first optimization factor, 1 represents a constant, represents the hyperbolic tangent function, represents the voltage prediction deviation index, Indicates the nominal voltage.

[0036] It should be noted that represents the normalized value after normalization by the nominal voltage, so that the first optimization factor has a wider applicability; and in order to further enhance the sensitivity of the first optimization factor to the voltage prediction deviation index and make it have nonlinear adjustment characteristics, the hyperbolic tangent function tanh() is introduced, and the domain of the tanh() function is all real numbers, and the value range is (-1, 1). The normalized value is used as the input of the hyperbolic tangent function, and the normalized value can be mapped to the (-1, 1) interval, and then through The operation will The value of is limited to the range of (1, 2).

[0037] At this point, the first optimization factor corresponding to the current sampling moment is obtained, which is used to optimize the process noise covariance matrix in the unscented Kalman filter algorithm. The first optimization factor realizes the sensitivity capture of the systematic deviation of the battery model prediction. When the battery model prediction has no obvious systematic deviation and the residual changes smoothly, Close to , Q remains basically unchanged; when the battery model prediction has obvious systematic deviations or the residuals change dramatically, Close to , Q increases. The nonlinear characteristics of the function make Ability to respond quickly to small deviations in battery model predictions and avoid over-adjustment when deviations are large.

[0038] Step S103: obtaining a second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm according to the residual fluctuation change of the voltage residual sequence.

[0039] The first optimization factor The correction of the systematic deviation of the battery model prediction is realized. The first optimization factor focuses on the mean and cumulative change of the voltage residual sequence, which is used to evaluate the average difference between the predicted value and the measured value of the battery model and the persistence of this difference. However, even if the prediction of the battery model has no obvious systematic deviation, that is, the mean of the voltage residual is close to zero, the prediction results of the battery model may still have random fluctuations.

[0040] This random fluctuation comes from many aspects: first, the measurement noise itself is random. Even if the measurement equipment (voltage sensor) is very accurate, its measurement results will inevitably contain a certain degree of random error. Secondly, the battery model is usually a simplification and approximation of the actual battery system, and cannot fully capture the complex electrochemical reactions and dynamic characteristics inside the battery. These factors not considered by the battery model, such as: microstructure changes inside the battery, small fluctuations in electrolyte concentration, etc., will also cause random fluctuations in the prediction results of the battery model; in addition, the working environment of the battery (such as temperature, load) also has random changes, which in turn affects the accuracy of the battery model prediction.

[0041] If the process noise covariance matrix Q is set too small, the UKF algorithm will over-trust the model prediction with random fluctuations, resulting in unnecessary oscillations in the SOC estimation results, reducing the smoothness and stability of the estimation results. To solve this problem, the embodiment of the present invention further proposes a second optimization factor , which is used to adjust Q according to the fluctuation pattern of the voltage residual sequence.

[0042] Specifically, first, the standard deviation of the voltage residual sequence is calculated , obtain a first ratio between the standard deviation and the nominal voltage , where the standard deviation represents the dispersion of the residuals in a period of time before the current sampling moment. The larger it is, the greater the fluctuation amplitude of the voltage residual in the voltage residual sequence is, and the greater the random error predicted by the battery model is; The smaller it is, the smaller the fluctuation amplitude of the voltage residual in the voltage residual sequence is, and the smaller the random error predicted by the battery model is.

[0043] Then, according to the main electrochemical time constant of the battery, the polarization time constant (Rₚ and Cₚ) in the Thevenin equivalent circuit model is used in the embodiment of the present invention to set it. Considering that the sampling frequency of the common polarization time constant of the lithium iron phosphate battery is 10 Hz, the lag order p is set to 100, and then according to the preset lag order, the lag voltage residual corresponding to each voltage residual in the voltage residual sequence is obtained respectively, and the product between each voltage residual in the voltage residual sequence and its corresponding lag voltage residual is accumulated to obtain a first accumulated value, and the square of each voltage residual in the voltage residual sequence is accumulated to obtain a second accumulated value. According to the ratio of the first accumulated value to the second accumulated value, the adaptive autocorrelation coefficient of the voltage residual sequence is obtained, which is used to measure the correlation between the voltage residuals separated by a specific time interval in the voltage residual sequence.

[0044] Preferably, the calculation expression of the adaptive autocorrelation coefficient of the voltage residual sequence is:

[0045] in, represents the adaptive autocorrelation coefficient of the voltage residual sequence, N represents the length of the voltage residual sequence, represents the ith voltage residual in the voltage residual sequence, represents the lagged voltage residual of the ith voltage residual in the voltage residual sequence.

[0046] It should be noted that the traditional autocorrelation coefficient usually only considers the correlation between the voltage residual sequence and its lagged one-period sequence. In order to better identify the periodic fluctuations that may exist in the voltage residual sequence, the embodiment of the present invention introduces an adaptive autocorrelation coefficient to better measure the correlation between the residual values ​​separated by a specific time interval p in the voltage residual sequence. Close to When , it means that there is no obvious correlation between the residual values ​​of the voltage residual sequence separated by time interval p; when Significantly greater than When , it indicates that there is a positive correlation between the residual values ​​of the voltage residual sequence at time intervals p, that is, the residual values ​​tend to fluctuate in the same direction with a period of p; when Significantly less than When , it indicates that there is a negative correlation between the residual values ​​of the voltage residual sequence separated by a time interval p, that is, the residual values ​​tend to fluctuate in the opposite direction with a period of p.

[0047] Secondly, obtain a second ratio between the absolute value of the voltage residual mean and the nominal voltage , which is used to reflect the average variation of the voltage residual sequence. The larger the absolute value of , the more drastic the change of the voltage residual sequence is, and there may be mutations or rapid changes. and The difference is: The focus is on the overall fluctuation pattern (trend, periodicity) of the voltage residual series, while What we are concerned about is the local variation amplitude of the voltage residual sequence. Even if the voltage residual sequence has no obvious trend or periodicity, if its local variation amplitude is large (there are mutation points), It will also be larger.

[0048] Finally, the constant 1, the absolute value of the adaptive autocorrelation coefficient and the added value of the second ratio are calculated to obtain the product of the added value and the first ratio, and the product is nonlinearly mapped using a hyperbolic tangent function to obtain a corresponding second mapping value, and the second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained based on the sum of the constant 1 and the second mapping value.

[0049] In one embodiment, the calculation expression of the second optimization factor is:

[0050] in, represents the second optimization factor, 1 represents a constant, represents the hyperbolic tangent function, || represents the absolute value sign, L1 represents the first ratio, and L2 represents the second ratio.

[0051] It should be noted that by introducing the standard deviation , It can directly reflect the fluctuation amplitude of the voltage residual sequence, so as to adjust Q accordingly. When the residual fluctuation amplitude increases, Increase, As a result, Q increases, which reduces the trust of the UKF algorithm in the battery model prediction and makes it rely more on the measured value for state estimation. , It can identify the periodic fluctuation pattern in the voltage residual sequence and adjust Q accordingly. When the voltage residual sequence shows obvious periodic fluctuation, Increase, As a result, Q increases.

[0052] At this point, the second optimization factor corresponding to the current sampling time is obtained, which is used to optimize the process noise covariance matrix in the unscented Kalman filter algorithm. The second optimization factor Comprehensively evaluate the fluctuation amplitude, trend, periodicity and mutation of the voltage residual sequence. When the voltage residual sequence fluctuates smoothly, Close to 1, Q remains basically unchanged; when the voltage residual sequence fluctuates significantly, It will be significantly greater than 1, Q increases, which reduces the UKF algorithm's trust in the battery model prediction and relies more on the measured value for state estimation, thereby improving the stability of the battery SOC estimation.

[0053] Step S104: using the first optimization factor and the second optimization factor to update and optimize the process noise covariance matrix at the previous sampling moment of the current sampling moment, to obtain the optimized process noise covariance matrix at the current sampling moment.

[0054] After obtaining the first optimization factor and the second optimization factor, the first optimization factor and the second optimization factor can be used to update and optimize the process noise covariance matrix at the previous sampling moment of the current sampling moment, and the process noise covariance matrix when the SOC value of the power battery at the current sampling moment is evaluated using the UKF algorithm, that is, the optimized process noise covariance matrix, wherein the updating and optimization method is: weighted summing the first optimization factor and the second optimization factor to obtain an adjustment coefficient, and the product of the adjustment coefficient and the process noise covariance matrix at the previous sampling moment of the current sampling moment is used as the optimized process noise covariance matrix at the current sampling moment.

[0055] In one embodiment, the calculation expression for updating optimization is:

[0056] in, represents the optimized process noise covariance matrix at the current sampling time, represents the weight of the first optimization factor, represents the weight of the second optimization factor, represents the first optimization factor, represents the second optimization factor, represents the process noise covariance matrix at time t-1, that is, the process noise covariance matrix at the previous sampling time of the current sampling time.

[0057] It should be noted that, considering that the first optimization factor and the second optimization factor have the same importance for adjusting the process noise covariance matrix Q, we set , no restriction is made here.

[0058] Step S105, according to the terminal voltage acquisition value at the current sampling moment and the optimized process noise covariance matrix, an unscented Kalman filter algorithm is used to obtain an estimated SOC value of the power battery at the current sampling moment, which is used to detect the working performance of the power battery.

[0059] Get the optimized process noise covariance matrix at the current sampling time Afterwards, according to the calculation process of the standard unscented Kalman filter algorithm, the SOC value of the power battery at the current sampling moment is estimated according to the terminal voltage collection value at the current sampling moment, and the SOC estimation value of the power battery at the current sampling moment is obtained. It is worth noting that the unscented Kalman filter algorithm includes three main steps: initialization, time update (prediction) and measurement update (correction). Used in the time update step, that is, to replace the original process noise covariance matrix Q at the current sampling time with the optimized process noise covariance matrix , while keeping other steps unchanged, using the unscented Kalman filter algorithm to estimate the SOC value of the battery belongs to the prior art and will not be described in detail here.

[0060] After determining the estimated SOC value of the power battery at the current sampling moment, the working performance of the power battery can be detected according to the estimated SOC value, which can provide key technical support for the performance evaluation, safe operation and life management of the power battery of new energy vehicles, thereby improving the overall performance and user experience of new energy vehicles, including but not limited to the following battery performance evaluation and application: (1) Battery State of Health (SOH) Assessment: Long-term monitoring of the SOC estimate of the power battery and combined with the battery capacity decay model to assess the battery state of health (SOH).

[0061] (2) Range estimation: Based on the SOC estimation value and the vehicle’s energy consumption model, the remaining range of the electric vehicle can be estimated more accurately.

[0062] (3) Charge and discharge control: The estimated SOC value is used as a feedback signal to achieve precise control of the battery charge and discharge process, avoid overcharging and over-discharging, extend battery life, and improve the safety of the battery system.

[0063] It should be noted that the focus of the present invention is on how to perform online adaptive adjustment on the process noise covariance matrix Q to improve the SOC estimation accuracy of the lithium iron phosphate battery based on the UKF algorithm, and to perform power battery working performance detection based on the battery SOC estimation value, which belongs to the prior art and will not be described in detail here.

[0064] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for detecting the working performance of a power battery for a new energy vehicle, characterized in that: The method comprises: Obtain the terminal voltage collection value of the power storage battery at the current sampling moment, and obtain the voltage residual sequence at the current sampling moment based on the terminal voltage collection values ​​and terminal voltage prediction values ​​at a preset number of historical sampling moments before the current sampling moment; According to the mean value and the cumulative amount of changes of the voltage residual sequence, a first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained; According to the residual fluctuation change of the voltage residual sequence, a second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained; Using the first optimization factor and the second optimization factor, the process noise covariance matrix at the previous sampling moment of the current sampling moment is updated and optimized to obtain the optimized process noise covariance matrix at the current sampling moment; According to the terminal voltage acquisition value at the current sampling moment and the optimized process noise covariance matrix, the unscented Kalman filter algorithm is used to obtain the SOC estimation value of the power battery at the current sampling moment, which is used to detect the working performance of the power battery.

2. A method for detecting the working performance of a power battery for a new energy vehicle according to claim 1, characterized in that: The step of obtaining a voltage residual sequence at the current sampling moment according to the terminal voltage collection values ​​and terminal voltage prediction values ​​at a preset number of historical sampling moments before the current sampling moment includes: For any historical sampling moment, the previous historical sampling moment of the said historical sampling moment is obtained, and the difference between the terminal voltage acquisition value of the said historical sampling moment and the terminal voltage prediction value of the said previous historical sampling moment is calculated as the voltage residual of the said historical sampling moment; the voltage residual of each historical sampling moment is obtained to form a voltage residual sequence of the current sampling moment.

3. A method for detecting the working performance of a power battery for a new energy vehicle according to claim 1, characterized in that: The step of obtaining a first optimization factor for optimizing a process noise covariance matrix in an unscented Kalman filter algorithm according to the mean and cumulative change of the voltage residual sequence includes: According to the mean of the voltage residual sequence and the difference of adjacent residuals, a voltage prediction deviation index is obtained; the nominal voltage of the power battery is obtained, the voltage prediction deviation index is normalized by using the nominal voltage to obtain a normalized value, the normalized value is nonlinearly mapped by using a hyperbolic tangent function to obtain a corresponding first mapping value, and according to a constant 1 and the sum of the first mapping value, a first optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm is obtained.

4. A method for detecting the working performance of a power battery for a new energy vehicle according to claim 3, characterized in that: The step of obtaining a voltage prediction deviation index according to the mean of the voltage residual sequence and the difference between adjacent residuals includes: The mean of the voltage residual sequence is calculated, recorded as the voltage residual mean, the absolute value of the difference between every two adjacent voltage residuals in the voltage residual sequence is calculated, and the accumulated value of the absolute value of the difference is obtained, recorded as the voltage change accumulation amount, and the voltage prediction deviation index is obtained according to the sum of the absolute value of the voltage residual mean and the absolute value of the voltage change accumulation amount.

5. A method for detecting the working performance of a power battery for a new energy vehicle according to claim 3, characterized in that: The step of obtaining a second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm according to the residual fluctuation change of the voltage residual sequence includes: Calculating the standard deviation of the voltage residual sequence to obtain a first ratio between the standard deviation and the nominal voltage; obtaining an adaptive autocorrelation coefficient of the voltage residual sequence; obtaining a second ratio between the absolute value of the voltage residual mean and the nominal voltage; Calculate the sum of constant 1, the absolute value of the adaptive autocorrelation coefficient and the second ratio to obtain the product of the sum and the first ratio, use a hyperbolic tangent function to perform nonlinear mapping on the product to obtain a corresponding second mapping value, and obtain a second optimization factor for optimizing the process noise covariance matrix in the unscented Kalman filter algorithm based on the sum of constant 1 and the second mapping value.

6. A method for detecting the working performance of a power battery for a new energy vehicle according to claim 5, characterized in that: The step of obtaining the adaptive autocorrelation coefficient of the voltage residual sequence includes: According to a preset lag order, the lagged voltage residual corresponding to each voltage residual in the voltage residual sequence is obtained respectively, the product between each voltage residual in the voltage residual sequence and its corresponding lagged voltage residual is accumulated to obtain a first accumulated value, the square of each voltage residual in the voltage residual sequence is accumulated to obtain a second accumulated value, and the adaptive autocorrelation coefficient of the voltage residual sequence is obtained according to the ratio of the first accumulated value to the second accumulated value.

7. A method for detecting the working performance of a power battery for a new energy vehicle according to claim 1, characterized in that: The method of updating and optimizing the process noise covariance matrix at the previous sampling moment of the current sampling moment by using the first optimization factor and the second optimization factor to obtain the optimized process noise covariance matrix at the current sampling moment includes: The first optimization factor and the second optimization factor are weightedly summed to obtain an adjustment coefficient, and the product of the adjustment coefficient and the process noise covariance matrix at a previous sampling moment of the current sampling moment is used as the optimized process noise covariance matrix at the current sampling moment.

Images