Battery capacity decay curve generation method based on conditional generative adversarial network
By expanding the lithium battery capacity degradation data set by the use of conditions to generate the confrontation network (CGAN), the problem of the lack of lithium battery capacity data set and the poor data enhancement method have been solved, and the accuracy and robust of the battery capacity attenuation curve predict sex.
Patent Information
- Application Number
- CN202310772148.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-28
- Publication Date
- 2025-01-24
- Estimated Expiration
- 2043-06-28
AI Technical Summary
In the prior art, lithium battery capacity data sets are small, and traditional data enhancement methods are difficult and the generalization of the results is weak, resulting in insufficient accuracy and robustness of the prediction of the battery capacity attenuation curve.
The conditional generative adversarial network (CGAN) is used to expand the lithium battery capacity degradation data set, and through mutual confrontation training of the generator and the discriminator, a diverse battery capacity decay curve is generated.
The number of samples and diversity of the battery capacity data set is increased, the predictive analysis accuracy and robustness of the deep learning network model is enhanced, and the research and development of the degradation of lithium battery capacity degradation trajectory is promoted.
Smart Images
Figure CN116660756B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of lithium battery attenuation analysis, and in particular, relates to a method for generating a battery capacity attenuation curve based on a conditional generative adversarial network. Background Art
[0002] As an important rechargeable battery, lithium-ion batteries have been widely used in electric vehicles, mobile devices, energy storage and other fields. The performance of lithium-ion batteries is often related to factors such as the number of charge and discharge cycles and the charge and discharge rate. Therefore, accurately predicting the battery capacity curve is of great significance for battery performance evaluation, equipment design and control.
[0003] In recent years, data-driven methods have made great achievements in battery capacity degradation trajectory prediction, and deep learning technology has stood out with its accurate, stable, and efficient prediction analysis. However, deep learning-related algorithms usually require a large number of data sets for model training, and the data sets involved in network model training need to have a certain diversity to improve the accuracy and generalization of model prediction analysis. However, in practical applications, traditional data collection methods require a lot of time and cost, and the collection of battery degradation data is affected by many factors such as battery working environment, load, and collection method. Battery degradation data is not as numerous as financial and image data sets, and the data sets available in research are extremely precious. In order to obtain more and richer battery degradation data sets and further expand research results and application scenarios, this study uses data generation technology to expand the existing battery degradation data sets, thereby increasing the number and diversity of data sets, and improving the accuracy and robustness of deep learning network model prediction analysis, thereby promoting the research and development of battery capacity degradation trajectory prediction.
[0004] At present, in the generation of battery capacity data, the mainstream methods include data generation methods based on traditional models: they generally explore the law of battery capacity degradation through the establishment of mathematical or physical models, so as to simulate and generate data. Although this method generates high quantity and quality of data, the generated data is relatively simple and has poor generalization ability; based on statistical methods: this method mainly uses statistical methods to analyze and extract the law of battery capacity changes over time, and generates data according to the extracted law. This method can simulate the change of battery capacity more accurately, but it requires a lot of experimental data support, and has the same disadvantages as traditional model data generation. Therefore, it is urgent to seek a data enhancement method that can generate large quantities, strong diversity, and high fidelity of generated data. Since the birth of deep learning, a large number of excellent models have emerged, such as generative adversarial networks (GANs), recurrent neural networks (RNNs), and variational autoencoders (VAEs). These models have also achieved very good results in the field of data enhancement.
[0005] This study uses the conditional generative adversarial network (CGAN) in the GAN family to enhance the battery capacity degradation dataset. In the present invention, CGAN is a more appropriate choice rather than RNN, VAE or other models. First, RNN can only generate sequences similar to historical data, and when generating the battery capacity decay curve, conditional information must be considered, and RNN cannot effectively use this conditional information to generate a capacity decay curve that meets the conditions; secondly, VAE can generate data similar to the training data, but it cannot control the characteristics of the generated data. In the task of generating the battery capacity decay curve, we need to generate a capacity decay curve with specific requirements based on different conditional information, and VAE cannot effectively achieve this. In contrast, CGAN is specifically used to generate models with conditional restrictions, so it is suitable for generating battery capacity decay curves. In addition, compared with the standard GAN, the biggest advantage of CGAN is the introduction of conditional vectors. By inputting noise and conditional information, the battery capacity decay curve can be generated in a multi-dimensional space, making the generated curve more diverse. Summary of the invention
[0006] In view of this, in order to solve the above-mentioned problems existing in the prior art, the purpose of the present invention is to provide a battery capacity attenuation curve generation method based on a conditional generative adversarial network to solve the cost and time limitations faced by lithium battery capacity data due to traditional data acquisition methods, thereby effectively generating a large amount of data for subsequent research in battery capacity curve prediction, battery performance evaluation, etc.
[0007] The technical solution adopted by the present invention is: a method for generating a battery capacity decay curve based on a conditional generative adversarial network, the generation method comprising the following steps:
[0008] S1: Obtain the real capacity decay curve of the lithium battery;
[0009] S2: normalize the real capacity decay curve of the lithium battery, and divide the processed data set into a training data set and a test data set;
[0010] S3: Build a CGAN network model, which includes a generator and a discriminator;
[0011] S4: Input the training data set into the CGAN network model for training to obtain the optimal generator;
[0012] S5: Generate predicted capacity decay curves of lithium batteries in batches through an optimal generator, and select the most representative predicted capacity decay curves using a generated curve selection method with the greatest diversity;
[0013] S6: Calculate the similarity score between the test data set and the most representative predicted capacity decay curve to evaluate the performance of the generator. If the performance of the generator is good, a good set of CGAN network models is obtained; if the performance of the generator is not good, return to S4 to continue training.
[0014] Furthermore, in S1, the method for obtaining the real capacity decay curve data of the lithium battery is:
[0015] S101: Lithium batteries are divided into three categories according to their lifespan, namely short life, medium life and long life. Short life means less than 700 cycles, medium life means 700 to 1500 cycles, and long life means more than 1500 cycles.
[0016] S102: performing clustering using a k-means clustering algorithm to obtain a cluster center curve corresponding to a true capacity decay curve of each type of lithium battery.
[0017] Further, in said S2, it includes:
[0018] S201: The one-hot codes defining various types of real capacity decay curves are 000, 010, and 100 respectively;
[0019] S202: reading cluster center curve data of various real capacity attenuation curves, adding an artificially controlled noise n1 to the x-axis, and adding an artificially controlled noise n2 to the y-axis;
[0020] S203: Expand each cluster center curve and its corresponding one-hot code by m times and store them into a collection object;
[0021] S204: normalizing the acquired collection object;
[0022] S205: Encapsulate the normalized collection object into a data loader as a training data set.
[0023] Furthermore, after removing the cluster center curve from each type of real capacity decay curve, a data piece is randomly selected as a test data set.
[0024] Furthermore, in S3, the constructed CGAN network model includes: a generator and a discriminator;
[0025] The generator includes a BatchNorm1d function layer and five linear layers. The input of the generator is random noise and one-hot encoding of each type of real capacity decay curve. The output is a predicted capacity decay curve generated after processing by five linear layers.
[0026] The discriminator includes 4 linear layers. The input of the discriminator is the real capacity decay curve, the one-hot encoding of each type of real capacity decay curve and the output of the generator. The output is a predicted capacity decay curve with a higher degree of authenticity and the probability of judging that the category to which the current predicted capacity decay curve belongs is true.
[0027] Furthermore, in S4, the optimal generator is obtained by the following method:
[0028] S401: After obtaining the training data set, randomly generate a batch of random noises of the same size that conform to the 0-1 normal distribution, and input the random noises and the one-hot encoding of various real capacity decay curves in the current training data set into the generator for forward propagation;
[0029] S402: Inputting the output of the generator, the category to which the current real capacity decay curve belongs, and the real capacity decay curve into the discriminator;
[0030] S403: The loss value is calculated through the output of the discriminator and the true capacity attenuation curve, back-propagated and fed back to the generator. The generator and the discriminator compete with each other to obtain the optimal generator.
[0031] Furthermore, in S5, the method for selecting the most representative predicted capacity decay curve is as follows:
[0032] S501: Use the trained optimal generator to batch generate predicted capacity decay curves and define them as a set C;
[0033] S502: Define an empty set C * , and then randomly select a curve C from the generated set C r ;
[0034] S503: Find a distance curve C in set C r The largest curve c, then put curve c into set C * , and remove curve c from set C;
[0035] Repeat S501 to S503 until set C * When the number of curves in is equal to the input selection number n, the loop is terminated, that is, the n most representative curves in set C are obtained.
[0036] Furthermore, in S6, the performance of the generator is evaluated using the following method:
[0037] S601: Calculate the DTW distance from the curve in the test data set to each predicted capacity decay curve in set C in sequence;
[0038] S602: Obtain the predicted capacity decay curve c1 with the shortest distance in the set C, and the predicted capacity decay curve has the data with the greatest similarity;
[0039] S603: Output the minimum predicted capacity decay curve c1 and its corresponding distance index value.
[0040] The beneficial effects of the present invention are:
[0041] 1. The battery capacity decay curve generation method based on conditional generative adversarial network provided by the present invention solves the problems of small battery capacity decay curve data sets in the existing battery capacity decay curve prediction field, high difficulty of existing data enhancement methods in this field, and weak generalization of results.
[0042] 2. The battery capacity attenuation curve generation method based on conditional generative adversarial network provided by the present invention expands the existing battery degradation data set through the CGAN network model, which can increase the number and diversity of samples in the data set, improve the accuracy and robustness of the prediction analysis of the deep learning network model, and to a certain extent promote the research and development of lithium battery capacity degradation trajectory prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] Figure 1 It is a flow chart of a method for generating a battery capacity decay curve based on a conditional generative adversarial network provided by the present invention;
[0044] Figure 2 It is a CGAN network model structure diagram of a battery capacity decay curve generation method based on a conditional generative adversarial network provided by the present invention;
[0045] Figure 3 It is a predicted capacity decay curve result graph generated in the battery capacity decay curve generation method based on the conditional generative adversarial network provided by the present invention;
[0046] Figure 4 It is a flow chart of a method for selecting a maximum diversity generation curve in a method for generating a battery capacity decay curve based on a conditional generation adversarial network provided by the present invention;
[0047] Figure 5 It is a result diagram of the maximum diversity generation curve selection algorithm in the battery capacity decay curve generation method based on the conditional generation adversarial network provided by the present invention;
[0048] Figure 6 It is a flow chart of a generator evaluation algorithm designed in the method for generating a battery capacity decay curve based on a conditional generative adversarial network provided by the present invention;
[0049] Figure 7 It is a result diagram of the generator evaluation algorithm in the battery capacity attenuation curve generation method based on the conditional generative adversarial network provided by the present invention. DETAILED DESCRIPTION
[0050] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar modules or modules with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be construed as limitations on the present application. On the contrary, the embodiments of the present application include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.
[0051] Example 1
[0052] In the embodiment, a method for generating a battery capacity decay curve based on a conditional generative adversarial network is specifically provided. Figure 1 As shown, the generation method includes the following steps:
[0053] S1: Obtain the real capacity decay curve of the lithium battery; In S1, the method for obtaining the real capacity decay curve data of the lithium battery is:
[0054] S101: The lithium battery data set is divided into three categories according to the life span, namely short life, medium life and long life, wherein the short life is less than 700 cycles, the medium life is 700-1500 cycles, and the long life is more than 1500 cycles;
[0055] S102: Use k-means clustering algorithm to perform clustering, and obtain the cluster center curve corresponding to the real capacity decay curve of each type of lithium battery. The k-means clustering algorithm is an iterative clustering analysis algorithm, and its steps are: pre-dividing the data into K groups, then randomly selecting K objects as initial cluster centers, and then calculating the distance between each object and each seed cluster center, and assigning each object to the cluster center closest to it.
[0056] S2: normalizing the actual capacity attenuation curve of the lithium battery, and dividing the processed data set into a training data set and a test data set; in S2, the method for obtaining the training data set includes:
[0057] S201: The one-hot codes defining various types of real capacity decay curves are 000, 010, and 100 respectively;
[0058] S202: reading cluster center curve data of various real capacity attenuation curves, adding an artificially controlled noise n1 to the x-axis, and adding an artificially controlled noise n2 to the y-axis;
[0059] S203: Expand each cluster center curve and its corresponding one-hot code by m times and store them into a collection object;
[0060] S204: normalizing the acquired collection object;
[0061] S205: Encapsulate the normalized collection object into a data loader as a training data set, and the training data set is used for subsequent CGAN network model model training.
[0062] In S2, after removing the cluster center curve from each type of real capacity decay curve, n pieces of data are randomly selected as a test data set through a random function. The test data set is used for model evaluation of the subsequent CGAN network model.
[0063] S3: Determine the hyperparameters required for the CGAN network model, including batch size B, learning rate LR, momentum parameter B1, and momentum parameter B2.
[0064] Construct a CGAN network model, such as Figure 2 As shown in the figure, the CGAN network model includes a generator and a discriminator. The optimizers of the generator and the discriminator both use the Adam optimizer, and the loss function uses the MSE loss.
[0065] The generator includes a BatchNorm1d function layer and five linear layers. The activation function is LeakyReLU. Finally, a Tanh activation function layer is added to convert the linear output into a value between -1 and 1. The input of the generator is random noise and one-hot encoding of each type of real capacity decay curve. The output is the predicted capacity decay curve generated after processing by five linear layers.
[0066] The discriminator includes 4 linear layers, and the activation function is LeakyReLU. At the same time, in order to prevent the discriminator from overfitting, the intermediate linear layer is followed by a Dropout layer to discard data at a fixed ratio to achieve data regularization. The input of the discriminator is the real capacity decay curve, the one-hot encoding of each type of real capacity decay curve, and the output of the generator. The output is a predicted capacity decay curve with a higher degree of authenticity and the probability of judging the category to which the current predicted capacity decay curve belongs to as true.
[0067] The specific structure of each layer in the CGAN network model can be further explained using the following table.
[0068] Table 1 Specific structure of each layer of CGAN model
[0069]
[0070]
[0071] Note: point_num represents the number of sampling points of the input battery capacity curve, and n_classes represents the number of categories of the battery capacity curve.
[0072] S4: Input the training data set into the CGAN network model for training to obtain the optimal generator; in S4, the optimal generator is obtained by the following method:
[0073] S401: After obtaining the training data set, randomly generate a batch of random noises of the same size that conform to the 0-1 normal distribution, and input the random noises and the one-hot encoding of various real capacity decay curves in the current training data set into the generator for forward propagation;
[0074] S402: Inputting the output of the generator, the category to which the current real capacity decay curve belongs, and the real capacity decay curve into the discriminator;
[0075] S403: The loss value loss is calculated through the output of the discriminator and the true capacity attenuation curve, and then fed back to the generator. The generator then continuously adjusts its model parameters according to the feedback, that is, the parameter weights of each item in the generator network model, etc. Then, the generator and the discriminator continuously compete with each other to produce the optimal generator, so as to finally obtain the optimal generator.
[0076] S5: Generate predicted capacity decay curves of lithium batteries in batches through an optimal generator, and select the most representative predicted capacity decay curves using a maximum diversity generation curve selection method; in S5, if Figure 4 As shown, the method to select the most representative predicted capacity decay curve is as follows:
[0077] S501: Use the trained optimal generator to batch generate predicted capacity decay curves and define them as a set C. The generated curve set is as follows: Figure 3 As shown; input set C and determine the number of extractions to be n;
[0078] S502: Define an empty set C * , and then randomly select a curve C from the generated set C r ; Initialize set C * =0 and k=1;
[0079] S503: Find a distance curve C in set C r The largest curve c, then put curve c into set C * , and remove curve c from set C;
[0080] S504: Determine whether k is equal to n;
[0081] S505: If they are equal, obtain the selected set C * and end; if they are not equal, find a distance curve C in set C r The largest curve c', then put the curve c' into the set C * , and remove the curve c' from the set C, and after k+1, return to S504;
[0082] Repeat this process until set C * When the number of curves in is equal to the input selection number n, the loop is terminated, that is, the most representative n curves in the set C are obtained, such as Figure 5 shown.
[0083] S6: Calculate the similarity score between the test data set and the most representative predicted capacity decay curve to evaluate the performance of the generator. If the performance of the generator is good, a good CGAN network model is obtained; if the performance of the generator is not good, return to S4 to continue training and adjust the model parameters. In S6, Figure 6 As shown, the performance of the generator is evaluated as follows:
[0084] S601: Calculate the DTW distance from the curve in the test data set to each predicted capacity decay curve in set C in sequence;
[0085] S602: Obtain the predicted capacity decay curve c1 with the shortest distance in the set C, and the predicted capacity decay curve has the data with the greatest similarity, such as Figure 7 As shown;
[0086] S603: Output the minimum predicted capacity decay curve c1 and its corresponding distance index value. In order to measure the difference between the generated predicted capacity decay curve and the actual capacity decay curve, this embodiment uses mean square error (MSE) as the evaluation index.
[0087]
[0088] Where k is the total number of test sets, y k and They are the generated curve capacity value and the real curve capacity value respectively.
[0089] The performance of the generator trained in this embodiment can be further illustrated by the following table.
[0090] Table 2 Error results of the real voltage sequence and the generated curve voltage sequence
[0091]
[0092] From the data in Table 2, it can be seen that the real capacity decay curve and the predicted capacity decay curve generated by the CGAN network model are close, with less loss, so the performance of the generator is better.
[0093] The battery capacity decay curve generation method based on conditional generative adversarial network provided in this embodiment solves the problems that there are few battery capacity decay curve data sets in the existing battery capacity decay curve prediction field, and the existing data enhancement methods in this field are difficult and the results are weak in generalization. This study expands the existing battery degradation data set through the CGAN network model, which can increase the number and diversity of data sets, improve the accuracy and robustness of deep learning network model prediction analysis, and to a certain extent promote the research and development of lithium battery capacity degradation trajectory prediction.
[0094] It should be noted that any process or method description in the flowchart or otherwise described herein may be understood to represent a module, fragment or portion of a code comprising one or more executable instructions for implementing steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations, in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.
[0095] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0096] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0097] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0098] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.
[0099] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A method for generating a battery capacity attenuation curve based on a conditional generative adversarial network, characterized in that: The generation method comprises the following steps: S1: Obtain the real capacity decay curve of the lithium battery; S2: normalize the real capacity decay curve of the lithium battery, and divide the processed data set into a training data set and a test data set; S3: Build a CGAN network model, which includes a generator and a discriminator; S4: Input the training data set into the CGAN network model for training to obtain the optimal generator; S5: Generate predicted capacity decay curves of lithium batteries in batches through an optimal generator, and select the most representative predicted capacity decay curves using a generated curve selection method with the greatest diversity; S6: Calculate the similarity score between the test data set and the most representative predicted capacity decay curve to evaluate the performance of the generator. If the performance of the generator is good, a good set of CGAN network models is obtained; if the performance of the generator is not good, return to S4 to continue training.
2. The method for generating a battery capacity attenuation curve based on a conditional generative adversarial network according to claim 1, characterized in that: In S1, the method for obtaining the real capacity decay curve data of the lithium battery is: S101: Lithium batteries are divided into three categories according to their lifespan, namely short life, medium life and long life. Short life means less than 700 cycles, medium life means 700 to 1500 cycles, and long life means more than 1500 cycles. S102: performing clustering using a k-means clustering algorithm to obtain a cluster center curve corresponding to a true capacity decay curve of each type of lithium battery.
3. The method for generating a battery capacity attenuation curve based on a conditional generative adversarial network according to claim 2, characterized in that: In said S2, it includes: S201: The one-hot codes defining various types of real capacity decay curves are 000, 010, and 100 respectively; S202: reading cluster center curve data of various real capacity attenuation curves, adding an artificially controlled noise n1 to the x-axis, and adding an artificially controlled noise n2 to the y-axis; S203: Expand each cluster center curve and its corresponding one-hot code by m times and store them into a collection object; S204: normalizing the acquired collection object; S205: Encapsulate the normalized collection object into a data loader as a training data set.
4. The method for generating a battery capacity attenuation curve based on a conditional generative adversarial network according to claim 2, characterized in that: After removing the cluster center curve from each type of real capacity decay curve, n data are randomly selected as the test data set.
5. The method for generating a battery capacity attenuation curve based on a conditional generative adversarial network according to claim 1, characterized in that: In S3, the constructed CGAN network model includes: generator and discriminator; The generator includes a BatchNorm1d function layer and five linear layers. The input of the generator is random noise and one-hot encoding of each type of real capacity decay curve. The output is a predicted capacity decay curve generated after processing by five linear layers. The discriminator includes 4 linear layers. The input of the discriminator is the real capacity decay curve, the one-hot encoding of each type of real capacity decay curve and the output of the generator. The output is a predicted capacity decay curve with a higher degree of authenticity and the probability of judging that the category to which the current predicted capacity decay curve belongs is true.
6. The method for generating a battery capacity attenuation curve based on a conditional generative adversarial network according to claim 1, characterized in that: In S4, the optimal generator is obtained by the following method: S401: After obtaining the training data set, randomly generate a batch of random noises of the same size that conform to the 0-1 normal distribution, and input the random noises and the one-hot encoding of various real capacity decay curves in the current training data set into the generator for forward propagation; S402: Inputting the output of the generator, the category to which the current real capacity decay curve belongs, and the real capacity decay curve into the discriminator; S403: The loss value is calculated through the output of the discriminator and the true capacity attenuation curve, back-propagated and fed back to the generator. The generator and the discriminator compete with each other to obtain the optimal generator.
7. The method for generating a battery capacity attenuation curve based on a conditional generative adversarial network according to claim 1, characterized in that: In S5, the method for selecting the most representative predicted capacity decay curve is as follows: S501: Use the trained optimal generator to batch generate predicted capacity decay curves and define them as a set C; S502: Define an empty set C * , and then randomly select a curve C from the generated set C r ; S503: Find a distance curve C in set C r The largest curve c, then put curve c into set C * , and remove curve c from set C; Repeat S501 to S503 until set C * When the number of curves in is equal to the input selection number n, the loop is terminated, that is, the n most representative curves in set C are obtained.
8. The method for generating a battery capacity attenuation curve based on a conditional generative adversarial network according to claim 1, characterized in that: In S6, the performance of the generator is evaluated as follows: S601: Calculate the DTW distance from the curve in the test data set to each predicted capacity decay curve in set C in sequence; S602: Obtain the predicted capacity decay curve c1 with the shortest distance in the set C, and the predicted capacity decay curve has the data with the greatest similarity; S603: Output the minimum predicted capacity decay curve c1 and its corresponding distance index value.