Joint estimation method and device for health state and aging mode of battery for electric vertical take-off and landing aircraft

Abstract

The invention discloses a combined estimation method and device for the health state and the aging mode of a battery for an electric vertical take-off and landing aircraft. The method comprises the following steps: calculating the evolution trend of the health state and the aging mode of a vehicle-mounted battery; correlation analysis and relation consistency evaluation are carried out, and an input parameter matrix is selected; normalizing the input and output parameter set, and constructing a joint estimation model by adopting a residual neural network algorithm with a bottleneck structure and storing the joint estimation model; recording and storing measurement data in a specific flight phase; normalizing the recorded data and substituting the data into the joint estimation model; and carrying out reverse normalization processing to obtain a battery health state and an aging mode state quantity. According to the method, the health state and the aging mode of the battery are comprehensively considered, correlation analysis and correlation relation consistency evaluation are combined, overall measurement data in a specific flight stage of the aircraft are used as model input, a residual neural network algorithm with a bottleneck structure is adopted, and the method has high training stability, convergence and model generalization ability.

Description

Technical Field

[0001] The present invention belongs to the field of battery technology and new energy aircraft, and in particular relates to a method and related device for jointly estimating the health status and aging mode of a battery for an electric vertical take-off and landing aircraft. Background Art

[0002] Aircraft electrification, as a key approach to addressing climate change and achieving sustainable development through low-carbon aviation transformation, has received widespread attention and has seen significant development. Electric Vertical Take-Off and Landing aircraft (eVTOL) are becoming an important trend in the development of next-generation low-altitude aircraft due to their low noise and zero emissions characteristics. The onboard battery system, as the core power source for eVTOL, directly determines the safety and reliability of the aircraft. Due to their advantages in energy density, power density, self-discharge rate, and cycle life, lithium-ion batteries have gradually been used in eVTOL. However, the unique flight characteristics and extreme environmental constraints of eVTOL pose significant challenges to the development of onboard battery management systems. Summary of the Invention

[0003] In order to solve the above problems, the present invention provides a joint estimation method and related devices for the health status and aging pattern of batteries for electric vertical take-off and landing aircraft. The proposed estimation method is based on the battery measurement data set corresponding to the specific flight phase of the electric vertical take-off and landing aircraft, and adopts a residual neural network algorithm with a bottleneck structure to comprehensively and accurately estimate the health status and aging pattern of the battery.

[0004] The present invention is achieved through the following technical solutions.

[0005] In a first aspect, the present invention provides a method for jointly estimating the state of health and aging pattern of a battery for an electric vertical take-off and landing aircraft, the method comprising the following steps:

[0006] S1, based on the characteristic test data of the power battery for electric vertical take-off and landing aircraft after different aging cycles, calculate the health status and aging mode evolution trend of the onboard battery throughout its life cycle under different aging conditions, which serves as the output parameter matrix of the battery health status and aging mode joint estimation model;

[0007] S2, conducts correlation analysis and relationship consistency evaluation on flight condition data during aging cycle testing of power batteries for electric vertical take-off and landing aircraft, and selects corresponding measurement data sets of appropriate flight phases as the input parameter matrix for the joint estimation model of battery health status and aging mode;

[0008] S3, normalizing the input parameter set and the output parameter set data obtained in step S1 and step S2 as a training set of the battery health state and aging mode joint estimation model, and using a residual neural network algorithm with a bottleneck structure to construct a battery health state and aging mode joint estimation model based on the measurement data of the specific flight phase of the electric vertical take-off and landing aircraft and store it;

[0009] S4, recording and storing the time-measurement sequence determined in step S2 corresponding to the specific flight phase of the electric vertical take-off and landing aircraft determined in step S2;

[0010] S5, after the specific flight phase of the electric vertical take-off and landing aircraft determined in step S2 ends, normalizing the time-measurement sequence data recorded in step S4 to form a normalized input parameter set, and substituting the normalized input parameter set into the battery health state and aging mode joint estimation model obtained in step S3 to obtain a model output parameter set;

[0011] S6, reverse normalizing the model output parameter set obtained in step S5 to obtain the battery health state and aging mode state quantity.

[0012] In step S1, the battery health state is comprehensively characterized by the battery capacity and the battery impedance; wherein the battery capacity is obtained by using the ampere-hour integral method on the time-current sequence data in the constant current discharge stage of the characteristic test, and the battery impedance is obtained by the pulse charge / discharge test in the characteristic test, and the calculation expression is:

[0013]

[0014] Wherein, V pulse and V0 are the battery voltages at the end of the pulse and at the beginning of the pulse, respectively, I pulse and I0 are the currents at the end of the pulse and at the beginning of the pulse, respectively.

[0015] The battery aging mode state is comprehensively characterized by the loss of lithium inventory (Loss of Lithium Inventory, LLI), the loss of active material in the negative electrode (Loss of Active Material in the Negative Electrode, LAM NE ), and the loss of active material in the positive electrode (Loss of Active Material in the Positive Electrode, LAM PE ); wherein LLI, LAM NE , and LAM PECalculated by Differential Voltage (DV) curve; wherein, the DV curve includes the following steps:

[0016] S1.1, using linear interpolation to eliminate voltage spikes;

[0017] S1.2, calculating DV curve; wherein, the DV calculation expression under constant current condition is:

[0018]

[0019] Wherein, is the DV value of the kth sampling point, V k and V k-1 are the battery voltages of the kth and k-1th sampling points, t k and t k-1 are the time values of the kth and k-1th sampling points, and I is the constant current amplitude.

[0020] S3.3, using filtering algorithm to denoise and smooth the DV curve.

[0021] In step S2, the Spearman rank correlation coefficient is used for correlation analysis, and the expression of the Spearman rank correlation coefficient is:

[0022]

[0023] Wherein, x p,i,j is the ith measurement value in the pth flight phase in the jth cycle, and the measurement value includes battery current, terminal voltage, temperature and current integral value, etc., ρ(x p,i , y) is the Spearman rank correlation coefficient between the ith measurement value in the pth flight phase and the battery health state or aging mode state, n is the cycle number, d p,i,j is the difference between the corresponding ranks of the ith measurement value in the pth flight phase in the jth cycle and the battery health state or aging mode state in the jth cycle, and p includes take-off, cruising and landing three phases. l is the length of the measurement data corresponding to each flight phase, which satisfies l=t p / T s , wherein, t p is the duration of each flight phase, and T s is the sampling period.

[0024] In step S2, the measurement value satisfying |ρ(x p,i , y)|>0.8 is preliminarily extracted as the candidate input parameter of the battery health state and aging mode joint estimation model.

[0025] The step S2 further screens the input parameters of the battery health state and aging mode joint estimation model by using a correlation relationship consistency evaluation method based on Euclidean distance, and the Euclidean distance calculation expression is:

[0026]

[0027] wherein m is the battery number, x p,m is a two-dimensional feature vector of the correlation relationship between the pth flight phase measurement value of the mth battery and the battery health state or the aging mode state, including the average value (x p,m,avg ) and the standard deviation (x p,m,std ) of the correlation relationship, i.e., x p,m =[x p,m,avg , x p,m,std ], θ p is the cluster center, and D p,m is the Euclidean distance between x p,m and θ p .

[0028] The correlation relationship consistency evaluation method based on Euclidean distance in the step S2 includes the following steps:

[0029] S2.1, selecting the minimum change range of the test battery health state or the aging mode state, and performing normalization processing on the corresponding candidate input parameters in the minimum change range;

[0030] S2.2, using a specified number of interpolation points to interpolate the correlation relationship between the measurement value in the minimum change range determined in the step S2.1 and the battery health state or the aging mode state;

[0031] S2.3, calculating the standard deviation and the average value of the correlation relationship between each interpolated measurement value and the battery health state or the aging mode state, and obtaining the cluster center (θ p ) of the correlation relationship between the measurement value of each flight phase and the battery health state or the aging mode state by using a K-means clustering algorithm, x p,m,avg and x p,m,std , and the calculation expression is:

[0032]

[0033] wherein N int is the number of interpolation points, l p is the length of the measurement data corresponding to the pth flight phase, x ij,p,m is the jth interpolation point of the correlation relationship between the ith measurement value in the pth flight phase of the mth battery and the battery health state or the aging mode state, x p,m,avg and x p,m,stdthe average and standard deviation of the correlation between the battery in the mth section and the battery state of health or the aging pattern state in the pth flight phase;

[0034] S2.4, selecting the flight phase measurement value corresponding to the lowest average Euclidean distance as the input parameter matrix of the joint estimation model of the battery state of health and the aging pattern.

[0035] In step S3, the input parameter set and the output parameter set data combination form a training set, the number of training set samples is the sum of the aging cycles of each battery in the data set, and the expression for normalizing the training set parameters is:

[0036]

[0037] wherein x i is the training set parameter, x i,norm is the normalized training set parameter, x i,min and x i,max are the minimum value and the maximum value of the corresponding same type of data set in the training set, respectively.

[0038] In step S3, the specific structure of the residual neural network with a bottleneck structure includes:

[0039] S3.1, the input layer receives the normalized input parameter set;

[0040] S3.2, the input layer first enters a group of convolution modules, wherein the main path of the convolution module is in turn a convolution layer containing 32 convolution kernels with a size of 3x3 and using the same padding method, batch normalization, and an activation function based on a linear rectifier function;

[0041] S3.3, the batch normalization output in the convolution module is connected in series to a plurality of residual modules with a bottleneck structure, wherein the main path of each residual module with a bottleneck structure is in turn a convolution layer containing 8 convolution kernels with a size of 1x1 and using the same padding method, batch normalization (BN), an activation function based on a linear rectifier function (ReLU), a convolution layer containing 8 convolution kernels with a size of 3x3 and using the same padding method, batch normalization, an activation function based on a linear rectifier function, a convolution layer containing 32 convolution kernels with a size of 1x1 and using the same padding method, and batch normalization, and the residual connection is to add the output of the main path to the output of the convolution module in step 3.2;

[0042] S3.4, the residual module outputs to a plurality of groups of full connection modules connected in series, wherein the main path of each group of full connection modules is in turn a full connection layer containing a certain number of neurons, batch normalization, and an activation function based on a linear rectifier function;

[0043] S3.5, the plurality of groups of full connection modules output to a regression layer containing a certain number of neurons, outputting the model estimated value, wherein the number of neurons in the regression layer is the number of estimated battery state of health and aging patterns.

[0044] The reverse normalization processing expression of the model output parameter in step S6 is:

[0045] y i,anti-norm = y i (y i,max -y i,min )+y i,min

[0046] Wherein y i,anti-norm is the reverse normalized model output parameter set, y i is the model output parameter set, y i,min and y i,max are the minimum and maximum values of the model output parameter set corresponding to the same data set.

[0047] In a second aspect, the present application provides a device for jointly estimating the state of health and aging patterns of a battery for an electric vertical take-off and landing aircraft, characterized in that the device comprises a measurement module, a storage module and an operation module;

[0048] The measurement module is used to measure the time, current, voltage and temperature of the battery in real time during the operation of the electric vertical take-off and landing aircraft, to preliminarily calculate the integrated current value, the differential current value, the integrated voltage value and the differential voltage value, and to transmit the measurement values and the preliminary calculation values to the storage module;

[0049] The storage module is used to store the measurement data corresponding to the appropriate flight phase determined in step S2, to store the code of the joint estimation model of the state of health and aging patterns of the battery based on the measurement data of the specific flight phase of the electric vertical take-off and landing aircraft constructed in step S3, and to transmit the measurement data and the model code to the operation module;

[0050] The operation module is used to input the measurement data in the storage module into the model code in the storage module, to execute the steps of the joint estimation method of the state of health and aging patterns of the battery for the electric vertical take-off and landing aircraft according to any of the embodiments in the first aspect, and to output the state of health and aging patterns of the battery.

[0051] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0052] 1.The battery health state and aging mode joint estimation method for the electric vertical take-off and landing aircraft according to the application comprehensively considers the battery capacity, impedance, LLI, LAM PE and LAM NE , and can more comprehensively and accurately represent the real-time life of the battery system for the electric vertical take-off and landing aircraft.

[0053] 2.The battery health state and aging mode joint estimation method for the electric vertical take-off and landing aircraft according to the application combines the correlation analysis method based on the Spearman rank correlation coefficient and the correlation relationship consistency evaluation method based on the Euclidean distance, can more accurately mine the battery measurement data with strong correlation with the battery health state and the aging mode state, and the correlation relationship has high consistency as the input of the battery health state and the aging mode joint estimation model.

[0054] 3.Different from the battery health state estimation method based on the voltage feature points in the flight process, the battery health state and aging mode joint estimation method for the electric vertical take-off and landing aircraft according to the application takes the corresponding measurement data set of the electric vertical take-off and landing aircraft in a specific flight stage as the input parameter matrix of the battery health state and aging mode joint estimation model, can more comprehensively and accurately mine the battery aging information, and improves the accuracy and generalization of the trained model.

[0055] 4.The battery health state and aging mode joint estimation method for the electric vertical take-off and landing aircraft according to the application adopts the residual neural network algorithm with a bottleneck structure, the battery health state and aging mode joint estimation model based on the measurement data of the electric vertical take-off and landing aircraft in a specific flight stage has fewer model parameter quantities and calculation amounts, can relieve the gradient disappearance and gradient explosion problems in the deep network training, has high training stability, and has high convergence performance and model generalization ability. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The flowchart of the battery health state and aging mode joint estimation method for the electric vertical take-off and landing aircraft according to the application.

[0057] Figure 2 The flowchart of the DV curve calculation method according to the application.

[0058] Figure 3 The flowchart of the correlation relationship consistency evaluation method based on the Euclidean distance according to the application.

[0059] Figure 4 The schematic diagram of the residual neural network with a bottleneck structure according to the application.

[0060] Figure 5 The schematic diagram of the battery health state and aging mode joint estimation device for the electric vertical take-off and landing aircraft according to the application.

[0061] Figure 6The estimated values of the battery health state and the aging mode state estimated by using the application are compared with the actually measured values. DETAILED DESCRIPTION

[0062] Exemplary embodiments of the present application will be described in greater detail below with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it is understood that the present application can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

[0063] The embodiment of the present application provides a joint estimation method and related device for battery health state and aging mode of an electric vertical take-off and landing aircraft, and realizes efficient and safe management of an on-board power battery.

[0064] A joint estimation method for battery health state and aging mode of an electric vertical take-off and landing aircraft is shown in the flowchart of Figure 1 The method mainly includes two parts: the first part is an offline training part, and the second part is an online estimation part. The two parts will be further described below.

[0065] The offline training part includes the following steps:

[0066] 1) According to the characteristic test data of the power battery of the electric vertical take-off and landing aircraft after different aging cycles, the health state and the aging mode evolution trend of the on-board battery in the whole life cycle under different aging conditions are calculated as the output parameter matrix of the joint estimation model of the battery health state and the aging mode.

[0067] The battery health state is comprehensively characterized by the battery capacity and the battery impedance. The battery capacity is obtained by using the ampere-hour integration method on the time-current sequence data in the constant current discharge stage of the characteristic test, and the battery impedance is obtained by the pulse charge / discharge test in the characteristic test. The calculation expression is:

[0068]

[0069] Wherein, V pulse and V0 are the battery voltages at the end of the pulse and at the beginning of the pulse, respectively, I pulse and I0 are the currents at the end of the pulse and at the beginning of the pulse, respectively.

[0070] The battery aging mode state is characterized by the loss of lithium inventory (Loss of Lithium Inventory, LLI), the loss of active material in the negative electrode (Loss of Active Material in the Negative Electrode, LAMNE ) and the loss of active material in the positive electrode (LAM PE ) are comprehensively characterized; wherein, LLI, LAM NE and LAM PE are calculated by a differential voltage (DV) curve; wherein, a flow chart for calculating the curve is shown in Figure 2 , and specifically includes the following steps:

[0071] 1.1) using linear interpolation to eliminate voltage spikes;

[0072] 1.2) calculating the DV curve; wherein, the DV calculation expression under constant current condition is:

[0073]

[0074] wherein, is the DV value of the kth sampling point, V k and V k-1 are the battery voltages of the kth and (k-1)th sampling points, t k and t k-1 are the time values of the kth and (k-1)th sampling points, and I is the constant current amplitude;

[0075] 1.3) using a filtering algorithm to denoise and smooth the DV curve.

[0076] 2) performing correlation analysis and relationship consistency evaluation on the flight condition data in the aging cycle test of the power battery for the electric vertical take-off and landing aircraft, selecting the corresponding measurement data set of the appropriate flight stage as the input parameter matrix of the battery state of health and aging mode joint estimation model;

[0077] wherein, the Spearman rank correlation coefficient is used for correlation analysis, and the expression of the Spearman rank correlation coefficient is:

[0078]

[0079] wherein, x p,i,j is the ith measurement value in the pth flight stage in the jth cycle, the measurement values include battery current, terminal voltage, temperature and current integral value, etc., p(x p,i , y) is the Spearman rank correlation coefficient between the ith measurement value in the pth flight stage and the state of health or aging mode of the battery, n is the cycle number, d p,i,jThe difference between the i-th measurement value in the p-th flight phase in the j-th cycle and the corresponding level between the battery state of health or the aging mode state in the j-th cycle, p includes take-off, cruise and landing three phases; i e Nl+, l is the length of the measurement data corresponding to each flight phase, satisfying l = t p / T s , wherein t p is the duration of each flight phase, T s is the sampling period;

[0080] In the step 2), the measurement values satisfying | p (x p,i , y) | > 0.8 are preliminarily extracted as the candidate input parameters of the joint estimation model of the battery state of health and the aging mode;

[0081] In the step 2), the correlation relationship consistency evaluation method based on the Euclidean distance is used to further screen the input parameters of the joint estimation model of the battery state of health and the aging mode, and the Euclidean distance calculation expression is:

[0082]

[0083] Wherein, m is the battery number, x p,m is a two-dimensional feature vector of the correlation relationship between the measurement value of the p-th flight phase of the m-th battery and the battery state of health or the aging mode state, including the average value (x p,m,avg ) and the standard deviation (x p,m,std ) of the correlation relationship, that is, x p,m = [x p,m,avg , x p,m,std ], θ p is the cluster center, D p,m is the Euclidean distance between x p,m and θ p ;

[0084] The flow chart of the correlation relationship consistency evaluation method based on the Euclidean distance in the step 2) is shown in Figure 3 , and specifically includes the following steps:

[0085] 2.1) Select the minimum change range of the test battery state of health or the aging mode state, and normalize the corresponding candidate input parameters in the minimum change range;

[0086] 2.2) Use a specified number of interpolation points to interpolate the correlation relationship between the measurement values in the minimum change range determined in step 2.1) and the battery state of health or the aging mode state;

[0087] 2.3) Calculate the standard deviation and mean value of the correlation between each interpolated measurement value and the battery health state or aging mode state, and obtain the cluster center (θ p ), x p,m,avg and x p,m,std The calculation expression is:

[0088]

[0089] Where N int is the number of interpolation points, l p is the length of the measurement data corresponding to the pth flight phase, x ij,p,m is the jth interpolation point of the correlation between the ith measurement value of the mth battery in the pth flight phase and the battery health state or aging mode state, x p,m,avg and x p,m,std are the mean value and standard deviation of the correlation between the pth flight phase and the battery health state or aging mode state of the mth battery;

[0090] 2.4) Select the flight phase measurement value corresponding to the lowest average Euclidean distance as the input parameter matrix of the battery health state and aging mode joint estimation model.

[0091] 3) Normalize the input parameter set and output parameter set data obtained in steps 1) and 2) as the training set of the battery health state and aging mode joint estimation model, and use a residual neural network algorithm with a bottleneck structure to construct and store the battery health state and aging mode joint estimation model based on the specific flight phase measurement data of the electric vertical take-off and landing aircraft;

[0092] Wherein, in step 3), the input parameter set and the output parameter set data are combined to form a training set, the number of training set samples is the sum of the aging cycles of each battery in the data set, and the expression for normalizing the training set parameters is:

[0093]

[0094] Where x i is the training set parameter, x i,norm is the normalized training set parameter, x i,min and x i,max are the minimum and maximum values of the corresponding same type of data set in the training set, respectively;

[0095] Wherein, in step 3), the schematic diagram of the residual neural network with a bottleneck structure is as shown in Figure 4 The specific structure includes:

[0096] 3.1) The input layer receives the normalized input parameter set;

[0097] 3.2) The input layer first enters a set of convolution modules, wherein the main path of the convolution module is in turn a convolution layer containing 32 convolution kernels with a size of 3x3 and using the same padding method, batch normalization, and an activation function based on a linear rectifier function;

[0098] 3.3) The batch normalization output in the convolution module is connected in series to multiple sets of residual modules with a bottleneck structure, wherein the main path of each set of residual modules with a bottleneck structure is in turn a convolution layer containing 8 convolution kernels with a size of 1x1 and using the same padding method, batch normalization (BN), an activation function based on a linear rectifier function (ReLU), a convolution layer containing 8 convolution kernels with a size of 3x3 and using the same padding method, batch normalization, an activation function based on a linear rectifier function, a convolution layer containing 32 convolution kernels with a size of 1x1 and using the same padding method, and batch normalization. The residual connection is to add the output of the main path to the output of the convolution module in step 3.2;

[0099] 3.4) The residual module output is connected in series to multiple sets of fully connected modules, wherein the main path of each set of fully connected modules is in turn a fully connected layer containing a certain number of neurons, batch normalization, and an activation function based on a linear rectifier function;

[0100] 3.5) The multiple sets of fully connected modules output to a regression layer containing a certain number of neurons, outputting the model estimate, wherein the number of neurons in the regression layer is the number of estimated battery state of health and aging patterns.

[0101] The online estimation part includes the following steps:

[0102] 1) In the specific flight phase of the electric vertical take-off and landing aircraft determined in offline training step 2), record and store the corresponding time-measurement sequence determined in offline training step 2);

[0103] 2) After the specific flight phase of the electric vertical take-off and landing aircraft determined in offline training step 2) is over, normalize the time-measurement sequence data recorded in online estimation step 1) to form a normalized input parameter set, and input the normalized input parameter set into the battery state of health and aging pattern joint estimation model obtained in offline training step 3) to obtain a model output parameter set;

[0104] 3) De-normalize the model output parameter set obtained in online estimation step 2) to obtain the battery state of health and aging pattern state quantity;

[0105] wherein the inverse normalization processing expression of the model output parameter is:

[0106] y i,anti-norm = y i (y i,max -y i,min )+ y i,min

[0107] 10. wherein y i,anti-norm is the inverse normalized model output parameter set, y i is the model output parameter set, y i,min and y i,max are the minimum and maximum values of the model output parameter set corresponding to the same type of data set, respectively.

[0108] A schematic diagram of a joint estimation device for battery state of health and aging mode of an electric vertical take-off and landing aircraft is shown in Figure 5 , which comprises a measurement module, a storage module and an operation module.

[0109] The measurement module is used to measure the time, current, voltage and temperature of the battery in real time during the operation of the electric vertical take-off and landing aircraft, to preliminarily calculate the integrated current value, the differential current value, the integrated voltage value and the differential voltage value, and to transmit the measurement values and the preliminary calculation values to the storage module.

[0110] The storage module is used to store the measurement data corresponding to the appropriate flight phase determined in the offline training step 2), to store the battery state of health and aging mode joint estimation model code based on the specific flight phase measurement data of the electric vertical take-off and landing aircraft constructed in the offline training step 3), and to transmit the measurement data and the model code to the operation module.

[0111] The operation module is used to input the measurement data in the storage module into the model code in the storage module, to execute the steps of the joint estimation method for the battery state of health and aging mode of the electric vertical take-off and landing aircraft, and to output the battery state of health and aging mode state.

[0112] In this embodiment, the implementation object is a ternary lithium ion battery with a nominal capacity of 3.0 Ah, and in actual application, it is not limited thereto. The estimated value of the battery state of health and aging mode state estimated by the present application is compared with the actual measured value as shown in Figure 6 . It can be found that the estimated value can better track the measured value, and therefore, the present method can well realize comprehensive and accurate estimation of the battery state of health and aging mode.

[0113] In summary, the application discloses a battery health state and aging mode combined estimation method for an electric vertical take-off and landing aircraft and a related device, which comprises the following steps: calculating the evolution trend of the battery health state and the aging mode; performing correlation analysis and relationship consistency evaluation, and selecting an input parameter matrix; normalizing the input and output parameter sets, and constructing a combined estimation model by using a residual neural network algorithm with a bottleneck structure and storing the model; recording and storing measurement data in a specific flight phase; normalizing the recorded data and substituting the data into the combined estimation model; and performing reverse normalization to obtain the battery health state and the aging mode state quantity.

[0114] Those skilled in the art can understand that the drawings are only schematic diagrams of a preferred embodiment, and the above-mentioned embodiment serial numbers are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0115] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the exemplary description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0116] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for jointly estimating the state of health and aging pattern of batteries for electric vertical take-off and landing aircraft, characterized in that: The method comprises the following steps: S1, based on the characteristic test data of the power battery for electric vertical take-off and landing aircraft after different aging cycles, calculate the health status and aging mode evolution trend of the onboard battery throughout its life cycle under different aging conditions, which serves as the output parameter matrix of the battery health status and aging mode joint estimation model; S2, conducts correlation analysis and relationship consistency evaluation on flight condition data during aging cycle testing of power batteries for electric vertical take-off and landing aircraft, and selects corresponding measurement data sets of appropriate flight phases as the input parameter matrix for the joint estimation model of battery health status and aging mode; S3, normalizing the input parameter set and output parameter set data obtained in step S1 and step S2 to serve as a training set for a joint estimation model for the battery state of health and aging pattern, and using a residual neural network algorithm with a bottleneck structure to construct and store a joint estimation model for the battery state of health and aging pattern based on measurement data of a specific flight phase of the electric vertical take-off and landing aircraft; S4, recording and storing the corresponding time-measurement sequence determined in step S2 for the specific flight phase of the electric vertical take-off and landing aircraft determined in step S2; S5, when the specific flight phase of the electric vertical take-off and landing aircraft determined in step S2 ends, normalizing the time-measurement quantity sequence data recorded in step S4 to form a normalized input parameter set, and substituting the normalized input parameter set into the battery health state and aging pattern joint estimation model obtained in step S3 to obtain a model output parameter set; S6, denormalizing the model output parameter set obtained in step S5 to obtain the battery health status and aging mode state quantity.

2. A method according to claim 1, characterized in that In step S1, the battery health status is comprehensively characterized by the battery capacity and battery impedance; wherein the battery capacity is obtained by using the ampere-hour integration method on the time-current series data of the constant current discharge stage in the characteristic test, and the battery impedance is obtained by the pulse charge / discharge test in the characteristic test. The calculation expression is: Among them, V pulse and V0 are the battery voltages at the end and beginning of the pulse, respectively. pulse I0 and I0 are the currents corresponding to the end and beginning of the pulse, respectively; Battery aging mode status through lithium inventory loss LLI, negative electrode active material loss LAM NE and positive electrode active material loss LAM PE Comprehensive characterization; among them, LLI, LAM NE and LAM PE It is obtained by calculating the differential voltage DV curve; wherein, obtaining the DV curve includes the following steps: S1.1, use linear interpolation to eliminate voltage spikes; S1.2, calculate the DV curve; the DV calculation expression under constant current conditions is: in, is the DV value of the kth sampling point, V k and V k-1 are the battery voltages at the kth and k-1th sampling points, respectively, t k and t k-1 are the time values ​​of the kth and k-1th sampling points respectively, and I is the constant current charging current amplitude; S1.3, uses filtering algorithm to reduce noise and smooth the DV curve.

3. A method according to claim 1, characterized in that In step S2, the Spearman rank correlation coefficient is used for correlation analysis. The Spearman rank correlation coefficient expression is: Among them, x p,i,j is the i-th measurement value in the p-th flight phase in the j-th cycle, and the measurement values ​​include battery current, terminal voltage, temperature, and current integral value, etc., ρ(x p,i ,y) is the Spearman rank correlation coefficient between the i-th measurement value and the battery health status or aging mode status in the corresponding p-th flight phase, n is the number of cycles, d p,i,j is the difference between the i-th measurement value in the p-th flight phase in the j-th cycle and the corresponding level of the battery health state or aging mode state in the j-th cycle, where p includes the three phases of takeoff, cruise and landing; l is the length of the measurement data corresponding to each flight phase, satisfying l = t p / T s , where t p is the duration of each flight phase, T s is the sampling period; In the step S2, the initial extraction satisfies |ρ(x p,i ,y)|>0.8 are used as candidate input parameters for the joint estimation model of battery health status and aging pattern.

4. A method according to claim 1, characterized in that In step S2, a correlation relationship consistency evaluation method based on Euclidean distance is used to further screen the input parameters of the battery health status and aging mode joint estimation model. The Euclidean distance calculation expression is: Among them, m is the battery number, x p,m is a two-dimensional feature vector of the correlation between the measured value of the p-th flight phase of the m-th battery and the battery health status or aging mode status, including the average value of the correlation relationship (x p,m,avg ) and standard deviation (x p,m,std ), that is, x p,m =[x p,m,avg , x p,m,std ],θ p is the cluster center, D p,m is x p,m and θ p The Euclidean distance.

5. A method according to claim 1, characterized in that In step S2, the correlation relationship consistency evaluation method based on Euclidean distance includes the following steps: S2.1, selecting a minimum variation range for testing the battery health state or aging mode state, and normalizing the candidate input parameters corresponding to the minimum variation range; S2.2, interpolating the correlation between the measured value within the minimum variation range determined in step S2.1 and the battery health state or aging mode state using a specified number of interpolation points; S2.3, calculate the standard deviation and mean of the correlation between each interpolated measurement value and the battery health state or aging mode state, and obtain the cluster center (θ) of the correlation between the measurement value and the battery health state or aging mode state corresponding to each flight phase by using the K-means clustering algorithm. p ), x p,m,avg and x p,m,std The calculation expression is: Among them, N int is the number of interpolation points, l p is the length of the measurement data corresponding to the pth flight phase, x ij,p,m is the jth interpolation point corresponding to the correlation between the ith measurement value of the mth battery in the pth flight phase and the battery health status or aging mode status, x p,m,avg and x p,m,std is the mean and standard deviation of the correlation between the mth battery and the battery health status or aging mode status at the pth flight stage; S2.4, select the flight phase measurements corresponding to the lowest average Euclidean distance as the input parameter matrix for the joint estimation model of battery health status and aging pattern.

6. A method according to claim 1, characterized in that In step S3, the input parameter set and the output parameter set data are combined to form a training set. The number of training set samples is the sum of the aging cycles of each battery in the data set. The expression for normalizing the training set parameters is: Among them, x i is the training set parameter, x i,norm is the normalized training set parameter, x i,min and x i,max are the minimum and maximum values ​​of the same type of data set in the training set, respectively.

7. A method according to claim 1, characterized in that In step S3, the specific structure of the residual neural network with a bottleneck structure includes: S3.1, the input layer receives the normalized input parameter set; In S3.2, the input layer first enters a set of convolutional modules. The main path of the convolutional modules consists of 32 convolutional layers with 3×3 kernels and the same padding, batch normalization, and activation functions based on linear rectification functions. S3.3, the batch normalization output in the convolution module is connected to multiple groups of residual modules with bottleneck structures in series, where the main path of each group of residual modules with bottleneck structures is a convolutional layer containing 8 convolution kernels of size 1×1 and using the same padding, batch normalization (BN), an activation function based on the linear rectification function (ReLU), a convolutional layer containing 8 convolution kernels of size 3×3 and using the same padding, batch normalization, an activation function based on the linear rectification function, a convolutional layer containing 32 convolution kernels of size 1×1 and using the same padding, and batch normalization. The residual connection is to add the output of the main path to the output of the convolution module in step 3.2; S3.4, the residual module outputs to multiple groups of fully connected modules connected in series, where the main path of each group of fully connected modules is a fully connected layer containing a certain number of neurons, batch normalization, and an activation function based on a linear rectification function; S3.5, multiple groups of fully connected modules are output to a regression layer containing a certain number of neurons, and output model estimation values, where the number of neurons in the regression layer is the number of estimated battery health states and aging modes.

8. A method according to claim 1, characterized in that In step S6, the denormalization expression for the model output parameters is: and i,anti-norm =and i (and i,max -and i,min )+y i,min Among them, y i,anti-norm is the model output parameter set after denormalization, y i is the model output parameter set, y i,min and y i,max They are the minimum and maximum values ​​of the model output parameter set corresponding to the same type of data set.

9. A device for jointly estimating the state of health and aging pattern of a battery for an electric vertical take-off and landing aircraft, characterized in that: The device includes: a measurement module, a storage module and a calculation module; The measurement module is used to measure the time, current, voltage and temperature of the battery in real time during the operation of the electric vertical take-off and landing aircraft, preliminarily calculate the integrated current value, the differential current value, the integrated voltage value and the differential voltage value, and transmit the measured values ​​and the preliminarily calculated values ​​to the storage module; The storage module is configured to store the measurement data corresponding to the appropriate flight phase determined in step S2, store the battery health status and aging mode joint estimation model code based on the measurement data of the specific flight phase of the electric vertical take-off and landing aircraft constructed in step S3, and transmit the measurement data and model code to the operation module; The operation module is used to input the measurement data in the storage module into the model code in the storage module, execute the steps of the joint estimation method of battery health status and aging mode for electric vertical take-off and landing aircraft according to any one of claims 1-8, and output the battery health status and aging mode status.

Images