A method for predicting the remaining useful life of a lithium battery based on a long short-term memory network

Abstract

The present invention discloses a method for predicting the remaining useful life of a lithium battery based on a long short-term memory network. The method includes the following steps: S1, measuring the discharge capacity data sequence of the lithium battery along with the charge and discharge cycles; S2, based on the discharge capacity data sequence of the lithium battery, applying a long short-term memory network to establish a prediction model for the remaining useful life of the lithium battery, and adding a mapping layer between the input layer and the hidden layer; S3, based on the established prediction model for the remaining useful life of the lithium battery, predicting the maximum available capacity sequence of the lithium battery in future charge and discharge cycles; S4, based on the maximum available capacity sequence of the lithium battery in future charge and discharge cycles, and combining with the lithium battery failure threshold, calculating the remaining life of the lithium battery. The present invention adopts a new model structure, and for the first time adds a mapping layer between the input layer and the hidden layer of the traditional long short-term memory network, having the advantages of simple operation, small error, and high prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of batteries, and in particular to a method for predicting the remaining service life of a lithium battery based on a long short-term memory network. Background Art

[0002] With the continuous innovation and breakthrough of technologies, in the face of the increasingly severe pressure of energy conservation and emission reduction, the development of electric vehicle technology has become an alternative means of transportation to fuel vehicles. In the future, pure electric vehicles will become the mainstream of newly sold vehicles, and vehicles used in the public domain will be fully electrified. Electric vehicles will surely become the main means of transportation for the future urban traffic development.

[0003] As the demand for electric vehicles gradually increases, some key technical problems of electric vehicles urgently need to be studied and broken through. As the main energy source of pure electric vehicles, the performance of power batteries directly determines the use experience of pure electric vehicles. As the power battery is used, its performance will continuously decline. In order to increase the cruising range of electric vehicles and extend the service life of the battery pack, the prediction of the remaining service life of the power battery has become one of the research keys.

[0004] The long short-term memory neural network is a kind of recurrent neural network, which has the ability to process long-term information and, to a certain extent, overcomes the problem of gradient disappearance in the processing of time series by traditional recurrent neural networks. Therefore, it is widely used in dealing with problems related to time series. However, for time series with too large time spans, there are still problems such as gradient disappearance and time-consuming calculations. The discharge capacity of a lithium battery continuously declines as the charge and discharge cycles of the lithium battery extend. Predicting the future discharge capacity of a lithium battery based on the data sequence of the discharge capacity of each cycle of the lithium battery in the past, and then calculating the remaining service life of the lithium battery is exactly the processing of a typical time series problem. By adding a mapping layer between the input layer and the hidden layer of the traditional long short-term memory network, building a long short-term memory network structure, and further establishing a prediction model for the remaining service life of the lithium battery, the accuracy, efficiency, and general applicability of the prediction of the remaining service life of the lithium battery are effectively improved. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for predicting the remaining service life of a lithium battery based on a long short-term memory network, which can accurately and efficiently predict the degradation of the discharge capacity of the lithium battery, and then predict the remaining service life of the lithium battery, aiming at the defects in the prior art.

[0006] The technical solution adopted by the present invention to solve its technical problems is as follows:

[0007] The present invention provides a method for predicting the remaining service life of a lithium battery based on a long short-term memory network, and the method includes the following steps:

[0008] S1. Measuring the data sequence of the discharge capacity of the lithium battery along with the charge and discharge cycles;

[0009] S2. Based on the lithium battery discharge capacity data sequence, establish a prediction model for the remaining useful life of the lithium battery using a long short-term memory network, and add a mapping layer between the input layer and the hidden layer;

[0010] S3. Based on the established prediction model for the remaining useful life of the lithium battery, predict the maximum available capacity sequence of the lithium battery for future charge and discharge cycles;

[0011] S4. Based on the maximum available capacity sequence of the lithium battery for future charge and discharge cycles, and combined with the lithium battery failure threshold, calculate the remaining life of the lithium battery.

[0012] Further, the specific method of step S1 of the present invention is as follows:

[0013] The measured lithium battery discharge capacity data sequence with the charge and discharge cycles is [C1,..., C i ,..., C n , where C i is the discharge capacity of the lithium battery pack in the i-th charge and discharge cycle, i = 1, 2,..., n, and n is the number of charge and discharge cycles.

[0014] Further, the specific method for establishing the model in step S2 of the present invention is as follows:

[0015] The long short-term memory network refers to adding a mapping layer between the input layer and the hidden layer of the traditional long short-term memory network. The mapping layer construction rules are as follows:

[0016] (a) The input layer X = [I1,..., I i ,..., I n T , and the input layer neurons I1,..., I i ,..., I n correspond one-to-one with the discharge capacity data sequence C1,..., C i ,...C n , that is, the lithium battery discharge capacity C i in the i-th charge and discharge cycle is transmitted to the i-th neuron of the input layer. Through k times of non-linear transformation rules δ(·), the input signal [C1,..., C i ,..., C n generates k different mapping elements Z k = [Z1,..., Z j ,..., Z k , where,

[0017] Z j = δ(XW zj + β zj ), j = 1, 2,..., k,​

[0018] Z j denotes the mapped signal after the j-th non-linear transformation, δ is a non-linear transformation rule that can be independently selected, and W zj is the randomly initialized weight parameter for the j-th non-linearity, and β zj is the randomly initialized perturbation parameter for the j-th non-linear transformation;

[0019] (c) Based on the mapping element Z k = [Z1,..., Z j ,..., Z k , m different eigen-elements H m = [H1,..., H p ,... H m are generated through the m-th deep transformation rule ζ(·), where

[0020] H p = ζ(Z k W hp + β hp ), p = 1, 2,..., m,

[0021] H p denotes the eigen-element obtained after the p-th deep transformation, ζ is a deep transformation rule that can be independently selected, W jp is the weight parameter for the p-th deep transformation, and β jp is the perturbation parameter for the p-th deep transformation;

[0022] (d) Cascade the mapping elements and eigen-elements in series to form a mapping layer [Z1,..., Z j ,..., Z k , H1,..., H p ,... H m .

[0023] Furthermore, in step S2 of the present invention, a long short-term memory network is applied to establish a lithium battery remaining service life prediction model, and the specific steps are as follows:

[0024] Input the lithium battery discharge capacity data sequence X = [C1,..., C i ,..., C n T , and use the lithium battery discharge capacity data [C1,..., C i-1 of the previous i charge and discharge cycles T as training samples, and the following discharge capacity data [C2,..., C i of the next charge and discharge cycle T as the training target, and apply the established long short-term memory network to establish a lithium battery remaining service life model.​

[0025] Further, the specific method of step S3 of the present invention is as follows:

[0026] Based on the established prediction model of the remaining service life of the lithium battery, predict the maximum available capacity sequence of the future charge and discharge cycles of the lithium battery. The specific steps are as follows:

[0027] Take the lithium battery discharge capacity data [C i ,..., C n-1 from the i-th charge and discharge cycle to the (n - 1)-th charge and discharge cycle T as the prediction sample, input it into the established prediction model of the remaining service life of the lithium battery, and predict the lithium battery discharge capacity of the (i + 1)-th to the n-th charge and discharge cycles.

[0028] Further, the specific method of step S4 of the present invention is as follows:

[0029] The remaining life refers to the number of remaining charge and discharge cycles before the lithium battery fails, RUL = |T true - T predicted |, where RUL is the remaining life of the lithium battery, T true is the number of charge and discharge cycles corresponding to when the measured lithium battery discharge capacity reaches the failure threshold, and T predicted is the number of charge and discharge cycles corresponding to when the predicted lithium battery discharge capacity of the future charge and discharge cycles reaches the failure threshold.

[0030] Further, the capacity threshold for the failure of the lithium battery pack in step S4 of the present invention refers to that the lithium battery discharge capacity drops to 70% of the nominal capacity.

[0031] The beneficial effects of the present invention are as follows: The method for predicting the remaining service life of a lithium battery based on a long short-term memory network proposed in the present invention provides a long short-term memory network structure, and for the first time adds a mapping layer between the input layer and the hidden layer of the traditional long short-term memory network; using the built long short-term memory network, based on the lithium battery discharge capacity data sequence, predict the remaining service life of the lithium battery, with simple operation and high prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present invention will be further described below in conjunction with the drawings and embodiments. In the drawings:

[0033] Figure 1 is a schematic flowchart of a method for predicting the remaining service life of a lithium battery based on a long short-term memory network provided by an embodiment of the present invention;

[0034] Figure 2 is a diagram showing the lithium battery discharge capacity degradation data sequence provided by an embodiment of the present invention;

[0035] Figure 3It is a comparison chart of the prediction results of the remaining service life of a lithium battery based on a long short-term memory network provided by an embodiment of the present invention and other commonly used methods. Detailed implementation manners

[0036] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0037] As Figure 1 shown is a schematic flowchart of a method for predicting the remaining service life of a lithium battery based on a long short-term memory network provided by an embodiment of the present invention. In Figure 1 the method shown, the following steps are included:

[0038] S1. Measure the discharge capacity data sequence of the lithium battery along with the charge and discharge cycles.

[0039] In an embodiment of the present invention, in step S1, the measured discharge capacity data sequence of the lithium battery along with the charge and discharge cycles is [C1,..., C i ,..., C n , where C i is the discharge capacity of the lithium battery pack in the i-th (i = 1, 2,..., n) charge and discharge cycle, and n is the number of charge and discharge cycles.

[0040] S2. Based on the discharge capacity data sequence of the lithium battery, establish a prediction model for the remaining service life of the lithium battery by applying a long short-term memory network.

[0041] In an embodiment of the present invention, in step S2, the long short-term memory network refers to adding a mapping layer between the input layer and the hidden layer of the traditional long short-term memory network. The construction rule of the mapping layer is:

[0042] (a) The input layer X = [I1,..., I i ,..., I n T , the input layer neurons I1,..., I i ,..., I n correspond one-to-one with the discharge capacity data sequence C1,..., C i ,... C n , that is, the discharge capacity C i of the lithium battery in the i-th charge and discharge cycle is transmitted to the i-th neuron of the input layer. Through the k-th nonlinear transformation rule δ(·), the input signal [C1,..., C i ,..., C​n Generate k different mapping elements Z k =[Z1,...,Z j ,...,Z k , where

[0043] Z j =δ(XW zj +β zj ), j = 1, 2,..., k,

[0044] Z j represents the mapping signal after the j-th non-linear transformation, δ is a self-selectable non-linear transformation rule, W zj is the randomly initialized weight parameter for the j-th non-linearity, and β zj is the randomly initialized perturbation parameter for the j-th non-linear transformation;

[0045] (c) Based on the mapping elements Z k =[Z1,...,Z j ,...,Z k , generate m different eigen elements H m =[H1,...,H p ,...H m through m deep transformation rules ζ(·), where

[0046] H p =ζ(Z k W hp +β hp ), p = 1, 2,..., m,

[0047] H p represents the eigen element obtained after the p-th deep transformation, ζ is a self-selectable deep transformation rule, W jp is the weight parameter for the p-th deep transformation, and β jp is the perturbation parameter for the p-th deep transformation;

[0048] (d) Cascade the mapping elements and eigen elements in series to form a mapping layer [Z1,...,Z j ,...,Z k ,H1,...,H p ,...H m .

[0049] Apply a long short-term memory network to establish a prediction model for the remaining useful life of a lithium battery. The specific steps are as follows:

[0050] Input the lithium battery discharge capacity data sequence X = [C1,...,C i ,...,C n T ​, the lithium battery discharge capacity data of the previous i charge and discharge cycles [C1,...,C i-1 ] T As training samples, the corresponding discharge capacity data of the following charge and discharge cycle [C2,...,C i ] T As a training target, the long short-term memory network built by the application is used to establish a lithium battery remaining service life model.

[0051] S3. Based on the established lithium battery remaining service life prediction model, the maximum available capacity sequence of the lithium battery in future charge and discharge cycles is predicted.

[0052] In an embodiment of the present invention, in step S3, based on the established lithium battery remaining service life prediction model, the maximum available capacity sequence of the lithium battery in future charge and discharge cycles is predicted. The specific steps are:

[0053] The lithium battery discharge capacity data from the i-th charge and discharge cycle to the n-1-th charge and discharge cycle [C i ,...,C n-1 ] T As a prediction sample, the established lithium battery remaining service life prediction model is input to predict the lithium battery discharge capacity from the i+1th to the nth charge and discharge cycle.

[0054] S4. Calculate the remaining life of the lithium battery based on the lithium battery failure threshold.

[0055] In the embodiment of the present invention, in step S4,

[0056] The capacity threshold for failure of the lithium battery pack refers to the discharge capacity of the lithium battery dropping to 70% of the nominal capacity.

[0057] Remaining life refers to the number of charge and discharge cycles remaining before the lithium battery fails, RUL=|T true -T predicted |, where RUL is the remaining life of the lithium battery, T true T is the number of charge and discharge cycles corresponding to when the measured lithium battery discharge capacity reaches the failure threshold, predicted The number of charge and discharge cycles corresponding to when the discharge capacity of the lithium battery in the future charge and discharge cycles reaches the failure threshold is predicted.

[0058] In order to demonstrate the process and estimation performance of the lithium battery remaining service life prediction method based on long short-term memory network proposed in the present invention, an example is used here to illustrate.

[0059] Taking the aging data of lithium battery 5 measured by the NASA laboratory as an example, a method for predicting the remaining useful life of lithium batteries based on long short-term memory network is verified. Lithium battery 5 undergoes an accelerated aging experiment at 25°C. First, it is charged at a constant current of 1.5 A until the terminal voltage of the lithium battery reaches 4.2 V, and then it is charged at a constant voltage of 4.2 V until the charging current drops to 20 mA. Then, it is discharged at a constant current of 2 A until the terminal voltage of the lithium battery drops to 2.7 V and stops. The charge and discharge of the lithium battery are repeated to obtain the accelerated aging data of lithium battery 5. Figure 2 The degradation process of the discharge capacity of the lithium battery with the number of charge and discharge cycles is shown. Applying the degradation data of the discharge capacity of lithium battery 5 with the number of charge and discharge cycles, based on the proposed long short-term memory network, the remaining life of the lithium battery is predicted. The specific operation steps are as follows:

[0060] (1) Obtain the aging data of lithium battery 5 measured by the NASA laboratory, and divide the aging data into a training set and a test set according to a ratio of 1:1.

[0061] (2) Based on the training set of the lithium battery discharge capacity data, use the long short-term memory network to establish a prediction model for the remaining useful life of the lithium battery.

[0062] (3) Based on the established prediction model for the remaining useful life of the lithium battery, input the test set to predict the maximum available capacity sequence of the lithium battery in future charge and discharge cycles.

[0063] (4) Take 70% of the rated capacity of the lithium battery as the failure threshold to calculate the remaining life of the lithium battery.

[0064] At the same time, to verify the superiority of the method proposed in the present invention, a comparative experiment is conducted using the traditional long short-term memory neural network and the method proposed in the present invention. Figure 3 The comparison chart of the prediction results. Table 1 shows the mean absolute percentage error and root mean square error of the prediction results of the proposed method and the traditional long short-term memory network for the remaining life of the lithium battery.

[0065] Table 1

[0066]

[0067] It can be seen from the comparison chart of the results of predicting the remaining life of lithium batteries and the error table that the discharge capacity degradation data of the future charge and discharge cycles calculated by the method for predicting the remaining service life of lithium batteries based on the long short-term memory network proposed by the present invention is more consistent with the true discharge capacity degradation data. The same conclusion can be drawn from Table 1. The mean absolute percentage error and the root mean square error of the method for predicting the remaining life of lithium batteries proposed by the present invention are both lower than the prediction results of the traditional long short-term memory network. At the same time, for the prediction of the lithium battery failure threshold, the error period of the hybrid prediction method proposed by the present invention is 1, while the error period of the traditional long short-term memory network is 5. Obviously, the accuracy of the life prediction method proposed by the present invention is higher. In summary, the method for predicting the remaining service life of lithium batteries based on the long short-term memory network proposed by the present invention has the advantages of small error and high efficiency.

[0068] It should be noted that according to the needs of implementation, each step / component described in the present application can be split into more steps / components, or two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0069] It should be understood that those of ordinary skill in the art can make improvements or changes according to the above description, and all such improvements and changes should fall within the protection scope of the appended claims of the present invention.

Claims

1. A method for predicting the remaining useful life of a lithium battery based on a long short-term memory network, characterized in that The method includes the following steps: S1. Measure the discharge capacity data sequence of the lithium battery along with the charge-discharge cycles; S2. Based on the discharge capacity data sequence of the lithium battery, establish a prediction model for the remaining useful life of the lithium battery by applying a long short-term memory network, and add a mapping layer between the input layer and the hidden layer; S3. Based on the established prediction model for the remaining useful life of the lithium battery, predict the maximum available capacity sequence of the lithium battery for future charge-discharge cycles; S4. Based on the maximum available capacity sequence of the lithium battery for future charge-discharge cycles, and in combination with the lithium battery failure threshold, calculate the remaining life of the lithium battery; The specific method for establishing the model in step S2 is as follows: The long short-term memory network refers to adding a mapping layer between the input layer and the hidden layer of the traditional long short-term memory network, and the mapping layer construction rule is: (a) Input layer X = [I1,..., I i ,..., I n T , the input layer neurons I1,..., I i ,..., I n correspond one-to-one with the discharge capacity data sequences C1,..., C i ,... C n , that is, the discharge capacity C of the lithium battery in the i-th charge and discharge cycle i is transmitted to the i-th neuron in the input layer, and k different mapping elements Z i ,..., C n are generated by applying the non-linear transformation rule δ(·) k times to the input signal [C1,..., C k = [Z1,..., Z j ,..., Z k , where​ Z j = δ(XW zj + β zj ), j = 1, 2, ..., k, Z j represents the mapped signal after the j-th non-linear transformation, δ is a non-linear transformation rule that can be freely selected, and W zj is the randomly initialized weight parameter for the j-th non-linearity, and β zj is the randomly initialized perturbation parameter for the j-th non-linear transformation; (c) Based on the mapping element Z k =[Z1,...,Z j ,...,Z k , m different eigen elements H are generated through m times of deep transformation rules ζ(·) m =[H1,...,H p ,...H m , where H p = ζ(Z k W hp + β hp ), p = 1, 2,..., m, H p represents the eigen - element obtained after the p - th deep transformation, ζ is a deep - transformation rule that can be freely selected, and W hp is the weight parameter of the p - th deep transformation, and β hp is the perturbation parameter of the p - th deep transformation; (d) Cascade the mapping elements and the eigen-elements in series to form a mapping layer [Z1,..., Z j ,..., Z k , H1,..., H p ,... H m .

2. The method for predicting the remaining service life of a lithium battery based on a long short-term memory network according to claim 1, wherein The specific method for step S1 is as follows: The measured discharge capacity data sequence of the lithium battery with the charge-discharge cycles is [C1,..., C i ,..., C n , where C i is the discharge capacity of the lithium battery pack in the i-th charge-discharge cycle, i = 1, 2,..., n, and n is the number of charge-discharge cycles.

3. The method for predicting the remaining useful life of a lithium battery based on a long short-term memory network according to claim 1, wherein The specific steps for applying the long short-term memory network to establish a prediction model for the remaining useful life of the lithium battery in step S2 are as follows: Input the lithium battery discharge capacity data sequence X = [C1,..., C i ,..., C n T , taking the lithium battery discharge capacity data [C1,..., C i-1 of the previous i - 1 charge - discharge cycles T as the training samples, and the corresponding discharge capacity data [C2,..., C i of the next charge - discharge cycle T as the training target, and apply the constructed long - short - term memory network to establish a lithium battery remaining service life model.​ 4. The method for predicting the remaining service life of a lithium battery based on a long short-term memory network according to claim 3, characterized in that, The specific method for step S3 is as follows: Based on the established prediction model for the remaining useful life of the lithium battery, predict the maximum available capacity sequence of the lithium battery for future charge-discharge cycles, and the specific steps are as follows: The lithium battery discharge capacity data from the i-th charge-discharge cycle to the (n-1)-th charge-discharge cycle [C i ,...,C n-1 T is used as a prediction sample and input into the established lithium battery remaining service life prediction model to predict the lithium battery discharge capacity from the (i + 1)-th to the n-th charge-discharge cycle.​ 5. The method for predicting the remaining service life of a lithium battery based on a long short-term memory network according to claim 1, wherein The specific method for step S4 is as follows: The remaining useful life refers to the number of charge-discharge cycles remaining before the lithium battery fails, RUL = |T true - T predicted |, where RUL is the remaining useful life of the lithium battery, T true is the number of charge-discharge cycles corresponding to when the measured discharge capacity of the lithium battery reaches the failure threshold, and T predicted is the number of charge-discharge cycles corresponding to when the predicted discharge capacity of the future charge-discharge cycles of the lithium battery reaches the failure threshold.

6. The method for predicting the remaining service life of a lithium battery based on a long short-term memory network according to claim 1, characterized in that The capacity threshold for the failure of the lithium battery in step S4 refers to the discharge capacity of the lithium battery dropping to 70% of the nominal capacity.

Images