A Prediction Method for the Remaining Useful Life of Lithium-Ion Batteries Based on Correlation Analysis and VMD-LSTM

Abstract

The present invention discloses a method for predicting the remaining useful life of a lithium-ion battery based on correlation analysis and VMD-LSTM. It belongs to the technical field of lithium-ion battery capacity detection. The specific steps are as follows: The discharge power of the lithium-ion battery, the constant current charging time, the average discharge temperature, the discharge cut-off voltage, and the ratio of the constant current charging time to the constant voltage charging time are used as battery capacity characteristic parameters, and the correlation coefficients between these characteristic parameters and the battery capacity are calculated. The characteristic parameter with the strongest correlation is selected as the HI for predicting the RUL of the lithium-ion battery. At the same time, VMD is used to decompose the selected HI into three modal components: global attenuation, local regeneration, and other noises, and these three modal components are used as HI for RUL prediction, effectively avoiding variable information overlap, and the decomposition process has strong robustness. The decomposed modal components are respectively subjected to preliminary RUL prediction by LSTM, and finally the prediction results of the three modal components are accumulated to achieve accurate prediction of the RUL of the lithium-ion battery.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium-ion battery capacity detection, and in particular to a method for predicting the remaining life of a lithium-ion battery based on correlation analysis and VMD-LSTM. Background Art

[0002] Lithium batteries have characteristics such as high energy density, low self-discharge rate, long cycle life, wide operating temperature range, and no pollution, and are widely used in intelligent manufacturing fields such as computing engineering, logistics, aerospace, electric vehicles, and electronic devices. The degradation of lithium batteries will cause battery failure, resulting in shortened battery life and even serious accidents. RUL (Remaining Useful Life) is usually defined as: under real-time working conditions, the number of cycle periods experienced when the real-time capacity of the battery decays to 70% of the rated capacity. It is generally considered that when the battery capacity drops to 70% of the rated capacity, the battery reaches the end of its service life. Therefore, a high accuracy of RUL of lithium-ion batteries is one of the important measures for protecting lithium batteries and is of great significance.

[0003] Since the direct measurement method of lithium-ion battery capacity needs to penetrate into the battery interior and will damage the internal structure of the battery, the RUL of lithium-ion batteries often needs to be predicted indirectly. At present, through a large number of studies, it is found that the change trends of parameters such as the discharge power, constant current charging time, voltage at the battery terminal during the constant current charging stage, and average temperature during discharge of lithium-ion batteries have a large correlation with the degradation trend of the battery capacity, and can be used as HI (Health Indicator) for RUL prediction. HI solves the difficulty of online obtaining the capacity, but there are still the following deficiencies: 1. The HI of some lithium-ion batteries has a low correlation with the battery capacity, which easily leads to uneven expression capabilities for the attenuation of the battery capacity. 2. The modal components generated by the existing signal decomposition methods are too many, which easily leads to modal repetition or the generation of additional noise.

[0004] In summary, in the field of lithium-ion battery RUL prediction, the existing methods have problems such as low prediction accuracy, poor generalization, and weak robustness of lithium-ion battery RUL prediction due to characteristics such as non-linearity, multi-modal, and multi-noise. Summary of the Invention

[0005] In view of the problems existing in the above technical background, the present invention proposes a method for predicting the RUL of lithium-ion batteries based on correlation analysis and VMD-LSTM. The correlation analysis between the characteristic parameters of lithium-ion batteries and the capacity is carried out to obtain the correlation coefficient between the characteristic parameters and the capacity, and the characteristic parameter with the strongest correlation is selected as the HI for predicting the RUL of lithium-ion batteries. Therefore, the selected HI can represent the RUL of lithium-ion batteries, and then indirectly predict the RUL, overcoming the problem of low quality of existing HIs. At the same time, VMD (Variational Mode Decomposition) is used to decompose the selected HI into three components: global attenuation, local regeneration, and other noises, and these three components are used as HIs for RUL prediction, effectively avoiding variable information overlap, and the decomposition process has strong robustness. The decomposed modal components are respectively predicted for RUL by LSTM (Long Short Term Memory), and finally the prediction results of each modal component are accumulated to achieve accurate prediction of the RUL of lithium-ion batteries.

[0006] The technical solution adopted by the present invention includes the following steps:

[0007] Step 1, since the discharge power of lithium-ion batteries, constant current charging time, average discharge temperature, discharge cut-off voltage, and the ratio of constant current charging time to constant voltage charging time have strong correlations with the battery capacity and can be used as HIs for predicting the RUL of lithium-ion batteries. The correlation analysis between the characteristic parameters of lithium-ion batteries and the capacity is carried out to obtain the correlation coefficient between the characteristic parameters and the capacity, and the characteristic parameter with the strongest correlation is selected as the HI for predicting the RUL of lithium-ion batteries.

[0008] Step 2, considering that the capacity degradation trend of lithium-ion batteries can be divided into three parts: global attenuation, local regeneration, and other noises. Therefore, VMD is used to perform multi-scale decomposition on the selected HI to obtain three groups of modal components.

[0009] Step 3, use the trained long short-term memory network model to predict the RUL of lithium-ion batteries for the three groups of modal components, and finally accumulate the three groups of prediction results to achieve accurate prediction of the RUL of lithium-ion batteries.

[0010] Step 1 is specifically implemented according to the following steps:

[0011] Obtain the discharge power of lithium-ion batteries, constant current charging time, average discharge temperature, discharge cut-off voltage, and the ratio of constant current charging time to constant voltage charging time through the publicly available dataset, and calculate the Person coefficient between the above characteristic parameters and the capacity.

[0012] Step 1.1: By calculation, it is found that the correlation between the constant current charging time and the capacity is the strongest, which can be used as HI to indirectly predict the RUL of lithium-ion batteries.

[0013] Step 2 is specifically implemented according to the following steps:

[0014] Step 2.1: The VMD sets the original signal f to be composed of a finite number of modal components u k (t). Calculate the single-sided spectrum of each u k (t) through the Hilbert transform, and then modulate the base bandwidth of the spectrum of u k by the mixed center frequency w k . Finally, solve the variational model with constraints that minimizes the sum of u k (t), as shown in formula (1):

[0015]

[0016] In formula (1), δ(t) is the Dirac function, is the gradient operation, and * represents the convolution operation.

[0017] The solution idea of the above variational model is as follows:

[0018] 1. Introduce the quadratic penalty factor α and the Lagrange operator λ(t) to transform the constrained problem into an unconstrained problem:

[0019]

[0020] In formula (2), <> represents the inner product operation.

[0021] Introduce the multiplicative operator alternating direction method to iteratively update u k , w k , and λ, and find the saddle point of formula (2), which is the optimal solution of the constrained variational equation. Use the Parseval / Plancherel Fourier theory to obtain The expression in the frequency domain n represents the number of iterations.

[0022] Initialization and Then update according to formulas (3) - (6) and

[0023]

[0024]

[0025]

[0026] In formula (5), τ represents the bandwidth.

[0027] The iteration termination condition is

[0028] In formula (6), ε represents the discrimination accuracy and is greater than 0.

[0029] Let the data of HI with the strongest correlation be H(t), and its VMD decomposition result is

[0030] In formula (7), N is the number of modal components. Considering that the battery capacity degradation trend consists of three parts: global attenuation, local regeneration, and other noises, N is set to 3.

[0031] After VMD decomposition, the main trend data u1(t), local fluctuation data u2(t), and u3(t) are obtained.

[0032] Step 3 is specifically implemented according to the following steps:

[0033] Step 3.1: Input the main trend data u1(t), local fluctuation data u2(t), and u3(t) obtained after VMD decomposition into the trained LSTM model.

[0034] Step 3.2: Accumulate the obtained prediction results, to obtain the final prediction result of the lithium-ion battery life.

[0035] The advantages and beneficial effects of the present invention are:

[0036] 1. It overcomes the problem of low quality of existing HI. After correlation analysis, HI can well represent the battery capacity for indirect prediction.

[0037] 2. VMD performs multi-scale analysis on HI, decomposing the original signal into three modal components, avoiding modal repetition or generation of additional noises caused by too many modal components. Brief Description of the Drawings

[0038] Figure 1 It is a schematic flowchart of a method for predicting the remaining life of a lithium-ion battery based on correlation analysis and VMD-LSTM disclosed by the present invention.

[0039] Figure 2 It is the battery capacity degradation curve diagram of B5, B6, B7, and B18.

[0040] Figure 3 It is the curve diagram of the constant current charging time and battery capacity changing with the battery cycle number.

[0041] Figure 4 It is the schematic diagram of the LSTM network structure. Specific Embodiments

[0042] In order to more clearly express the implementation objectives, implementation schemes and advantages of the present invention, the following will further explain and illustrate the present invention in conjunction with the accompanying drawings. However, it should be noted that the described implementation process is a part of the overall implementation process of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0043] See Figure 1 , Figure 1 which is the structural flowchart of the method for predicting the remaining useful life of a lithium-ion battery based on correlation analysis and VMD-LSTM according to an embodiment of the present invention, and includes the following steps:

[0044] Step 1. Since there is a strong correlation between the discharge power, constant current charging time, average discharge temperature, cut-off discharge voltage of the lithium-ion battery, and the ratio of constant current charging time to constant voltage charging time and the battery capacity, they can be used as HIs for predicting the RUL of the lithium-ion battery. Perform a correlation analysis between the characteristic parameters of the lithium-ion battery and the capacity to obtain the correlation coefficient between the characteristic parameters and the capacity, and select the characteristic parameter with the strongest correlation as the HI for predicting the RUL of the lithium-ion battery.

[0045] Step 1 is specifically implemented according to the following steps:

[0046] Step 1.1. The experimental data is from the public dataset provided by NASA's PCoE. Select a set of charge and discharge data of 18650-type lithium-ion batteries (B5, B6, B7, B18) collected under the same experimental environment, and the rated capacity of the battery is 2 A·h.

[0047] Figure 2 shows the capacity degradation curves of batteries B5, B6, B7, and B18

[0048] Through correlation analysis, the correlation coefficients between the battery discharge power P, constant current charging time L, average discharge temperature T c , cut-off discharge voltage V, and the ratio K of constant current charging time to constant voltage charging time and the battery capacity are obtained. As shown in Table 1:

[0049] Table 1 Correlation analysis table of characteristic parameters and capacity (taking battery B5 as an example)

[0050]

[0051] Step 1.2. It is calculated that the correlation between the constant current charging time and the capacity is the strongest and can be used as an HI to indirectly predict the RUL of the lithium-ion battery.

[0052] Figure 3 It is a curve graph showing the constant current charging time and battery capacity varying with the battery cycle number.

[0053] Step 2 is specifically implemented according to the following steps:

[0054] Step 2.1: The original signal f in VMD is set as a composite of a finite number of modal components u k (t). Calculate the single-sided spectrum of each u k (t) through the Hilbert transform, and then modulate the base bandwidth of the spectrum of u k through the hybrid center frequency w k (t). Finally, solve the variational model with constraints to minimize the sum of u k (t), as shown in formula (1):

[0055]

[0056] In formula (1), δ(t) is the Dirac function, is the gradient operation, and * represents the convolution operation.

[0057] The solution idea of the above variational model is as follows:

[0058] 1. Introduce the quadratic penalty factor α and the Lagrange operator λ(t) to transform the constrained problem into an unconstrained problem:

[0059]

[0060] In formula (2), <> represents the inner product operation.

[0061] Introduce the multiplicative operator alternating direction method to iteratively update u k , w k , and λ, and find the saddle point of formula (2), which is the optimal solution of the constrained variational equation. Use the Parseval / Plancherel Fourier theory to obtain The expression in the frequency domain n represents the number of iterations.

[0062] Initialize and Then update according to formulas (3) to (6) and

[0063]

[0064]

[0065]

[0066] In Equation (5), τ represents the bandwidth.

[0067] The iteration termination condition is

[0068] In Equation (6), ε represents the discrimination accuracy and is greater than 0.

[0069] Let the data of HI with the strongest correlation be H(t), and its VMD decomposition result is

[0070] In Equation (7), N is the number of modal components. Considering that the battery capacity degradation trend consists of three parts: global attenuation, local regeneration, and other noises, N is set to 3.

[0071] After VMD decomposition, the main trend data u1(t), local fluctuation data u2(t), and u3(t) are obtained.

[0072] Step 3 is specifically implemented according to the following steps:

[0073] Step 3.1: Input the main trend data u1(t), local fluctuation data u2(t), and u3(t) obtained after VMD decomposition into the trained LSTM model.

[0074] Figure 4 The schematic diagram of the LSTM network structure is given

[0075] Step 3.2: Accumulate the obtained prediction results, to obtain the final prediction result of the lithium-ion battery life.

Claims

1. A method for predicting the remaining life of a lithium-ion battery based on correlation analysis and VMD-LSTM, characterized in that It includes the following steps: (1) Since the discharge power, constant current charging time, average discharge temperature, cut-off discharge voltage of the lithium-ion battery, and the ratio of constant current charging time to constant voltage charging time have a strong correlation with the battery capacity, as the HI for predicting the RUL of the lithium-ion battery, perform a correlation analysis between the characteristic parameters of the lithium-ion battery and the capacity to obtain the correlation coefficient between the characteristic parameters and the capacity, and select the characteristic parameter with the strongest correlation as the HI for predicting the RUL of the lithium-ion battery; (2) Considering that the capacity degradation trend of the lithium-ion battery is divided into three parts: global attenuation, local regeneration, and other noises, therefore, use VMD to perform multi-scale decomposition on the selected HI to obtain three groups of modal components; (3) Use the trained LSTM to predict the RUL of the lithium-ion battery for the three groups of modal components, and finally add up the three groups of prediction results to achieve accurate prediction of the RUL of the lithium-ion battery.

2. The method for predicting the remaining life of a lithium-ion battery based on correlation analysis and VMD-LSTM according to claim 1, wherein The specific steps of step (1) include: Obtain the discharge power, constant current charging time, average discharge temperature, cut-off discharge voltage of the lithium-ion battery, and the ratio of constant current charging time to constant voltage charging time through the publicly available dataset, and calculate the Person coefficient between the above characteristic parameters and the capacity.

3. The method for predicting the remaining life of a lithium-ion battery based on correlation analysis and VMD-LSTM according to claim 1, wherein The specific steps of step (2) include: The VMD sets the original signal f as a composite of a finite number of mode components u k (t), calculates the single-sided spectrum of each u k (t) through the Hilbert transform, and then, through the mixing center frequency w k , modulates the spectral base bandwidth of u k (t), and finally solves the variational model with constraints that minimizes the sum of u k (t), as shown in formula (1): In formula (1), δ(t) is the Dirac function, δ(t) is the gradient operation, and * represents the convolution operation; The solution idea of the above variational model is: Introduce the quadratic penalty factor α and the Lagrangian operator λ(t), and transform the constrained problem into an unconstrained problem: In formula (2), <> represents the inner product operation; Introduce the multiplicative operator alternating direction method to iteratively update \(u\) k , \(w\) k and \(\lambda\), find the saddle point of equation (2), which is the optimal solution of the constrained variational equation, and obtain the expression in the frequency domain where \(n\) represents the number of iterations; Initialization Then update according to equations (3) to (6) and In formula (5), τ represents the bandwidth; The iteration termination condition is In formula (6), ε represents the discrimination accuracy and is greater than 0; Let the data of the HI with the strongest correlation be H(t), and its VMD decomposition result is: In formula (7), N is the number of modal components. Considering that the battery capacity degradation trend consists of three parts: global attenuation, local regeneration, and other noises, therefore, set N to 3. After VMD decomposition, the main trend data u1(t) and local fluctuation data u2(t), u3(t) are obtained.

4. The method for predicting the remaining life of a lithium-ion battery based on correlation analysis and VMD-LSTM according to claim 1, wherein The specific steps of step (3) include: (4.1) Input the main trend data u1(t) and local fluctuation data u2(t), u3(t) obtained after VMD decomposition into the trained LSTM model; (4.2) Accumulate the obtained prediction results, to obtain the final prediction result of the lithium-ion battery life.

Images