Energy storage battery cycle life prediction method and device, equipment and storage medium
By acquiring sample datasets of energy storage batteries, identifying data points with high variability as key features, constructing a convolutional neural network model, and performing preprocessing, the problem of low accuracy in predicting the cycle life of energy storage batteries was solved, achieving higher prediction accuracy and efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2026-03-13
AI Technical Summary
Existing technologies struggle to accurately obtain key characteristic parameters that affect the cycle life of energy storage batteries, resulting in low accuracy in predicting the cycle life of energy storage batteries.
By acquiring a sample dataset of energy storage batteries, the degree of variation of each data measurement point is determined. Data measurement points with a large degree of variation are selected as key features. A prediction model based on a convolutional neural network is constructed, and the sample data is numerically normalized and sparsified for training and prediction.
It improves the accuracy and efficiency of predicting the cycle life of energy storage batteries, reduces the risk of overfitting, and enhances the applicability of the prediction model.
Smart Images
Figure CN121656840A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of energy storage battery technology, and in particular to a method, apparatus, device and storage medium for predicting the cycle life of an energy storage battery. Background Technology
[0002] Energy storage batteries are the core component of energy storage technology and have been widely used in power, industrial, and commercial sectors. While energy storage batteries offer numerous advantages such as rapid response and convenient adjustment, they also face challenges related to cost, safety, and cycle life. After prolonged and repeated charge-discharge cycles, battery performance may deteriorate rapidly, leading to reduced overall efficiency and even safety risks. Accurate prediction of energy storage battery cycle life facilitates precise assessment and early warning of battery performance, improving overall operational efficiency and reducing replacement and maintenance costs. Therefore, predicting the cycle life of energy storage batteries is of great significance.
[0003] Currently, the prediction of the cycle life of energy storage batteries is mainly based on characteristic parameters. However, due to the large number of characteristic parameters that affect the cycle life of energy storage batteries, existing methods have difficulty in accurately obtaining the key characteristic parameters that affect the cycle life of energy storage batteries, resulting in low accuracy in the prediction of the cycle life of energy storage batteries. Summary of the Invention
[0004] This disclosure provides a method, apparatus, device, and storage medium for predicting the cycle life of energy storage batteries, in order to solve the problem of low accuracy in predicting the cycle life of existing energy storage batteries.
[0005] In a first aspect, this disclosure provides a method for predicting the cycle life of an energy storage battery, including:
[0006] Obtain a sample dataset of energy storage batteries. Each sample in the sample dataset includes multiple data measurement points.
[0007] Determine the degree of variation of each data measurement point. The degree of variation is used to measure the impact of the data measurement point on the cycle life.
[0008] A preset number of data points are selected as key features in descending order of the degree of variation to obtain a key feature sample dataset of the energy storage battery. Each sample data in the key feature sample dataset includes a preset number of key features.
[0009] The prediction model is trained using a key feature sample dataset to obtain a trained prediction model, which is used to predict the cycle life of energy storage batteries.
[0010] The cycle life of energy storage batteries is predicted periodically using a trained prediction model.
[0011] In some embodiments, determining the degree of variation of each data measurement point includes:
[0012] The working condition with the longest time span is selected from the sample dataset as the feature evaluation working condition;
[0013] The earliest data sample in the characteristic evaluation working condition is determined as the basic data;
[0014] The latest data sample in the feature evaluation condition is determined as the reference data;
[0015] The degree of variation of each data measurement point is determined according to the following expression:
[0016]
[0017] Where, δ i x represents the degree of variation of the i-th data point. i Let x' represent the i-th data point in the basic data. i Cov(x) represents the i-th data point in the reference data. i ,x i ') represents x i ,x i The covariance of ', var(x) i ),var(x i ') represent x respectively i ,x i The variance of '.
[0018] In some embodiments, before training the prediction model using a key feature sample dataset, the method further includes:
[0019] The key feature sample dataset is preprocessed to obtain the preprocessed key feature sample dataset. The preprocessing includes numerical normalization and sparsification. The sparsification process involves sampling the key feature sample dataset according to a preset period.
[0020] In some embodiments, the method further includes:
[0021] Aggregate and generalize the preprocessed key feature sample dataset based on the number of iterations.
[0022] In some embodiments, the prediction model includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a fully connected layer, and an output layer connected in sequence.
[0023] The input layer has 12×12 nodes, and the number of channels is equal to the number of key feature items;
[0024] The first convolutional layer has a kernel size of 3×3, a depth of 32, a stride of 1, and uses the ReLU function as its activation function.
[0025] The filter size of the first pooling layer is 2×2, the step size is 2, and the maximum pooling rule is adopted.
[0026] The second convolutional layer has a kernel size of 3×3, a depth of 64, a stride of 1, and uses the ReLU function as its activation function.
[0027] The filter size of the second pooling layer is 2×2 with a step size of 2, and the maximum pooling rule is adopted.
[0028] The third convolutional layer has a kernel size of 2×2, a depth of 120, a stride of 1, and uses the ReLU function as its activation function.
[0029] The fully connected layer has 1024 nodes and uses the Sigmoid activation function.
[0030] The output layer has 600 nodes, and the activation function is the Softmax function.
[0031] In some embodiments, the data measurement points include voltage, current, temperature, charge, state of charge, and depth of charge / discharge.
[0032] Secondly, this disclosure provides a device for predicting the cycle life of an energy storage battery, comprising:
[0033] The acquisition module is used to acquire a sample dataset of energy storage batteries. Each sample in the sample dataset includes multiple data measurement points.
[0034] The processing module is used to determine the degree of variation of each data measurement point, and the degree of variation is used to measure the impact of the data measurement point on the cycle life.
[0035] The selection module is used to select a preset number of data points as key features in descending order of the degree of variation, so as to obtain a key feature sample dataset of the energy storage battery. Each sample data in the key feature sample dataset includes a preset number of key features.
[0036] The training module is used to train the prediction model using a dataset of key feature samples to obtain a trained prediction model, which is used to predict the cycle life of energy storage batteries.
[0037] The prediction module is used to periodically predict the cycle life of energy storage batteries using a trained prediction model.
[0038] Thirdly, this disclosure provides a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the foregoing aspects.
[0039] Fourthly, this disclosure provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in the above aspects.
[0040] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the methods described in the above aspects.
[0041] This disclosure provides a method, apparatus, device, and storage medium for predicting the cycle life of energy storage batteries. The method involves acquiring a sample dataset of energy storage batteries, where each sample data point includes multiple data measurement points; determining the degree of variation for each data measurement point, which measures the impact of the data measurement point on the cycle life; selecting a predetermined number of data measurement points as key features in descending order of variation, resulting in a key feature sample dataset for the energy storage battery, where each sample data point includes a predetermined number of key features; training a prediction model using the key feature sample dataset to obtain a trained prediction model, which is then used to predict the cycle life of the energy storage battery; and periodically predicting the cycle life of the energy storage battery using the trained prediction model. Selecting key features based on the degree of variation ensures the correlation between the model's input features and the cycle life of the energy storage battery, thus improving the accuracy of the energy storage battery cycle life prediction. Attached Figure Description
[0042] The present disclosure will be described in more detail below based on embodiments and with reference to the accompanying drawings:
[0043] Figure 1 A flowchart illustrating a method for predicting the cycle life of an energy storage battery, provided in an embodiment of this disclosure;
[0044] Figure 2 A schematic diagram of the architecture of a prediction model provided in an embodiment of this disclosure;
[0045] Figure 3 A schematic diagram illustrating the convergence effect of a training prediction model provided in an embodiment of this disclosure;
[0046] Figure 4 This is a schematic diagram illustrating the cycle life prediction effect of an energy storage battery according to an embodiment of the present disclosure.
[0047] Figure 5 This is a schematic diagram of the structure of an energy storage battery cycle life prediction device provided in an embodiment of this disclosure.
[0048] In the accompanying drawings, the same parts are referred to by the same reference numerals, and the drawings are not drawn to scale. Detailed Implementation
[0049] To enable those skilled in the art to better understand the technical solutions of this disclosure, and to fully understand and implement the process of how this disclosure applies technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. The embodiments of this disclosure and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort should fall within the protection scope of this disclosure.
[0050] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0051] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.
[0052] Example 1
[0053] Figure 1 This is a flowchart illustrating a method for predicting the cycle life of an energy storage battery, as provided in an embodiment of this disclosure.
[0054] like Figure 1 As shown, the energy storage battery cycle life prediction method provided in this embodiment may include:
[0055] S101. Obtain the sample dataset of the energy storage battery. Each sample data in the sample dataset includes multiple data measurement points.
[0056] In this embodiment, the sample data in the energy storage battery sample dataset includes energy storage battery operation data and experimental data, and each sample data in the sample dataset includes multiple data measurement points. In some optional implementations, the data measurement points include, but are not limited to, voltage, current, temperature, energy, state of charge (SOC), and depth of charge / discharge.
[0057] S102. Determine the degree of variation of each data measurement point. The degree of variation is used to measure the impact of the data measurement point on the cycle life.
[0058] In order to select the data points that have a significant impact on cycle life from the numerous data points that affect the cycle life of energy storage batteries, this embodiment uses the degree of variation to measure the degree of influence of the data points on cycle life.
[0059] In some optional implementations, determining the degree of variation at each data measurement point may specifically include:
[0060] The working condition with the longest time span is selected from the sample dataset as the feature evaluation working condition;
[0061] The earliest data sample in the feature evaluation condition is determined as the basic data, denoted as X = [x1, x2, ... x]. i ,...x n ], where x i This represents the measurement point of the i-th data item in the basic data;
[0062] The latest data sample in the feature evaluation conditions is determined as the reference data, denoted as X'=[x'1,x'2,...x']. i ',...x' n ], where x' i This represents the measurement point of the i-th data item in the reference data;
[0063] The degree of variation of each data point in the sample data is calculated and denoted as: δ=[δ1, δ2, ... δ i ...δ n ], where the degree of variation of the i-th data measurement point δ i Determined based on the following expression:
[0064]
[0065] Where, δ i x represents the degree of variation of the i-th data point. i Let x' represent the i-th data point in the basic data. i Cov(x) represents the i-th data point in the reference data. i ,x i ') represents xi ,x i The covariance of ', var(x) i ),var(x i ') represent x respectively i ,x i The variance of '.
[0066] S103. Select a preset number of data points as key features in descending order of the degree of variation to obtain a key feature sample dataset of the energy storage battery. Each sample data in the key feature sample dataset includes a preset number of key features.
[0067] After obtaining the degree of variation of each data point, the data points are sorted in descending order of variation. It is understood that data points ranked higher have a greater impact on cycle life. Therefore, in order to accurately extract the key features affecting the cycle life of the energy storage battery, this embodiment selects a predetermined number of data points with the highest degree of variation as key features; for example, the top 5 data points with the highest degree of variation can be selected as key features.
[0068] It should be noted that by selecting key features, not only can the accuracy of energy storage battery cycle life prediction be improved, but the amount of data processing is also reduced, which helps to improve the efficiency of energy storage battery cycle life prediction.
[0069] S104. The prediction model is trained using a key feature sample dataset to obtain a trained prediction model, which is used to predict the cycle life of energy storage batteries.
[0070] In this embodiment, after obtaining the key feature sample dataset of the energy storage battery, the dataset can be used to train a prediction model for predicting the cycle life of the energy storage battery. The prediction model in this embodiment is built based on a convolutional neural network and uses the key feature sample data as input data, which helps improve prediction accuracy and efficiency.
[0071] S105. Periodically predict the cycle life of the energy storage battery using a trained prediction model.
[0072] In this embodiment, after obtaining the trained prediction model, the cycle life of the energy storage battery can be predicted periodically using the trained prediction model. For example, predictions can be made on a period of hours, days, or weeks. The specific prediction period can be set according to actual needs.
[0073] The energy storage battery cycle life prediction method provided in this embodiment obtains a sample dataset of energy storage batteries, where each sample data point includes multiple data measurement points. It determines the degree of variation of each data measurement point, which measures its impact on cycle life. A predetermined number of data measurement points are selected as key features in descending order of variation, resulting in a key feature sample dataset for the energy storage battery. Each sample data point in this dataset includes a predetermined number of key features. The prediction model is trained using this key feature sample dataset to obtain a trained prediction model, which is then used to predict the cycle life of the energy storage battery. The trained prediction model is used to periodically predict the cycle life of the energy storage battery. By selecting key features based on the degree of variation, the correlation between the model's input features and the cycle life of the energy storage battery is ensured, improving the accuracy of energy storage battery cycle life prediction.
[0074] Example 2
[0075] Based on the above embodiments, in order to further improve the prediction efficiency, the energy storage battery cycle life prediction method provided in this embodiment needs to preprocess the key feature sample dataset before training the prediction model using the key feature sample dataset to obtain the preprocessed key feature sample dataset. The preprocessing includes numerical normalization processing and sparsification processing. The sparsification processing is to sample the key feature sample dataset according to a preset period.
[0076] In this embodiment, the numerical normalization of the key feature sample dataset can be performed using dimensionless normalization, such as the Min-Max normalization method. The sparsity processing of the key feature sample dataset in this embodiment refers to sampling the energy storage battery charging or discharging process at preset periodic intervals t, for example, t = int(T / 144), where int represents rounding down and T represents the duration of one charging or discharging cycle. It is understood that through numerical normalization and sparsity processing, the amount of data processing is effectively reduced, which helps to improve the efficiency of energy storage battery cycle life prediction.
[0077] Furthermore, to reduce the risk of overfitting and improve applicability, the energy storage battery cycle life prediction method provided in this embodiment also needs to aggregate and generalize the preprocessed key feature sample dataset based on the number of cycles. In this embodiment, data samples with similar cycle counts are aggregated and generalized; for example, every 10 cycles are aggregated into one cycle life state, resulting in 600 states. It can be understood that through sample generalization, the number of neurons is reduced, the risk of overfitting is lowered, and efficiency and applicability are improved.
[0078] Example 3
[0079] Based on any of the above embodiments, in order to further improve the accuracy of energy storage battery cycle life prediction, this embodiment constructs a prediction model that matches the energy storage battery cycle life prediction based on a convolutional neural network. The specific architecture of the model can be found in [reference needed]. Figure 2 .like Figure 2 As shown, the prediction model in this embodiment may include an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a fully connected layer, and an output layer connected in sequence.
[0080] The input layer has 12×12 nodes, and the number of channels is equal to the number of key feature terms. 144 sets of continuous, equally time-period t-sampled key feature data are selected as unit inputs. In this embodiment, the top 5 features with the highest correlation are preferred as input key feature terms.
[0081] The first convolutional layer has a kernel size of 3×3, a depth of 32, a stride of 1, and uses the ReLU activation function. After the first convolutional layer, 32 10×10 feature matrices are obtained.
[0082] The first pooling layer has a filter size of 2×2 and a stride of 2, using a maximum pooling rule. After pooling, 32 5×5 eigenvalue matrices are obtained.
[0083] The second convolutional layer has a 3×3 kernel size, a depth of 64, a stride of 1, and uses the ReLU activation function. After the second convolutional layer, 64 3×3 feature matrices are obtained.
[0084] The second pooling layer has a filter size of 2×2 and a step size of 2. It adopts the maximum pooling rule, padding with zeros for non-uniform features, and obtains 64 2×2 eigenvalue matrices after pooling.
[0085] The third convolutional layer has a kernel size of 2×2, a depth of 120, a stride of 1, and uses the ReLU activation function. After the third convolutional layer, a feature vector of length 120 is obtained.
[0086] The fully connected layer has 1024 nodes, and the activation function is the Sigmoid function.
[0087] The output layer has 600 nodes, and the activation function is the Softmax function.
[0088] When using key feature sample datasets Figure 2 When the prediction model shown is trained, its convergence performance can be referenced. Figure 3 As shown. Figure 3The horizontal axis represents the number of training iterations, and the vertical axis represents the loss entropy / accuracy. It can be seen that as the number of training iterations increases, the accuracy of the prediction model continuously increases, approaching 1 infinitely; the loss of the prediction model continuously decreases, also approaching 1 infinitely. When the trained prediction model is used to periodically predict the cycle life of energy storage batteries, the prediction effect is as follows: Figure 4 As shown.
[0089] In summary, the energy storage battery cycle life prediction method provided in this application extracts key features based on the degree of variation, ensuring the correlation between input features and cycle life and improving prediction accuracy. It constructs a prediction model matching the energy storage battery cycle life prediction based on a convolutional neural network framework, using key feature data as sample data and performing sparsification and aggregation generalization processing on the samples, reducing the number of neurons in the algorithm, lowering the risk of overfitting, and improving efficiency and applicability.
[0090] Example 4
[0091] Figure 5 This is a schematic diagram of a battery cycle life prediction device provided in an embodiment of this disclosure. Figure 5 As shown, the energy storage battery cycle life prediction device 50 provided in this embodiment may include: an acquisition module 501, a processing module 502, a selection module 503, a training module 504, and a prediction module 505.
[0092] The acquisition module 501 is used to acquire a sample dataset of energy storage batteries. Each sample data in the sample dataset includes multiple data measurement points.
[0093] Processing module 502 is used to determine the degree of variation of each data measurement point. The degree of variation is used to measure the impact of the data measurement points on the cycle life.
[0094] The selection module 503 is used to select a preset number of data points as key features in descending order of the degree of variation, so as to obtain a key feature sample dataset of the energy storage battery. Each sample data in the key feature sample dataset includes a preset number of key features.
[0095] Training module 504 is used to train the prediction model using a key feature sample dataset to obtain a trained prediction model, which is used to predict the cycle life of energy storage batteries.
[0096] Prediction module 505 is used to periodically predict the cycle life of energy storage batteries using a trained prediction model.
[0097] The apparatus of this embodiment can be used to perform Figure 1 The technical solutions of the method embodiments shown are similar in principle and in effect, and will not be described again here.
[0098] In some embodiments, the processing module 502 is used to determine the degree of variation of each data measurement point, which may specifically include:
[0099] The working condition with the longest time span is selected from the sample dataset as the feature evaluation working condition;
[0100] The earliest data sample in the characteristic evaluation working condition is determined as the basic data;
[0101] The latest data sample in the feature evaluation condition is determined as the reference data;
[0102] The degree of variation of each data measurement point is determined according to the following expression:
[0103]
[0104] Where, δ i x represents the degree of variation of the i-th data point. i Let x' represent the i-th data point in the basic data. i Cov(x) represents the i-th data point in the reference data. i ,x i ') represents x i ,x i The covariance of ', var(x) i ),var(x i ') represent x respectively i ,x i The variance of '.
[0105] In some embodiments, the energy storage battery cycle life prediction device 50 may further include a preprocessing module (not shown in the figure) for preprocessing the key feature sample dataset before training the prediction model with the key feature sample dataset to obtain a preprocessed key feature sample dataset. The preprocessing includes numerical normalization processing and sparsification processing. The sparsification processing is to sample the key feature sample dataset according to a preset period.
[0106] In some embodiments, the preprocessing module is further configured to perform aggregation generalization on the preprocessed key feature sample dataset based on the number of iterations.
[0107] In some embodiments, the prediction model includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a fully connected layer, and an output layer connected in sequence.
[0108] The input layer has 12×12 nodes, and the number of channels is equal to the number of key feature items;
[0109] The first convolutional layer has a kernel size of 3×3, a depth of 32, a stride of 1, and uses the ReLU function as its activation function.
[0110] The filter size of the first pooling layer is 2×2, the step size is 2, and the maximum pooling rule is adopted.
[0111] The second convolutional layer has a kernel size of 3×3, a depth of 64, a stride of 1, and uses the ReLU function as its activation function.
[0112] The filter size of the second pooling layer is 2×2 with a step size of 2, and the maximum pooling rule is adopted.
[0113] The third convolutional layer has a kernel size of 2×2, a depth of 120, a stride of 1, and uses the ReLU function as its activation function.
[0114] The fully connected layer has 1024 nodes and uses the Sigmoid activation function.
[0115] The output layer has 600 nodes, and the activation function is the Softmax function.
[0116] In some embodiments, the data measurement points include voltage, current, temperature, charge, state of charge, and depth of charge / discharge.
[0117] Example 5
[0118] Based on the above embodiments, this embodiment provides an application example.
[0119] This invention provides a method for predicting the cycle life of energy storage batteries to solve the problems of existing technologies.
[0120] To achieve the above objectives, the present invention provides the following technical solution:
[0121] This invention relates to a method for predicting the cycle life of energy storage batteries, the main process of which includes:
[0122] Step 1: Obtain sample data of energy storage batteries.
[0123] Step 2: Extract key features from the energy storage battery sample data using feature extraction methods.
[0124] Step 3: Preprocess the key characteristic data of the energy storage battery.
[0125] Step 4: Generalize the preprocessed sample data using sample generalization rules.
[0126] Step 5: Train the optimal prediction model using the algorithm module.
[0127] Step 6: Periodically predict the cycle life of the energy storage battery using the preferred prediction model.
[0128] The sample data in step 1 includes energy storage battery operation data and experimental data. The data content includes, but is not limited to, voltage, current, temperature, charge, state of charge (SOC), depth of charge and discharge, etc.
[0129] The feature extraction method in step 2 consists of the following steps:
[0130] 2.1: Select the operating condition with the longest time span from the energy storage battery samples as the feature evaluation condition;
[0131] 2.2: Select the earliest data sample in the feature evaluation condition as the basic data, denoted as X = [x1, x2, ... x i ,...x n The latest data sample in the feature evaluation condition is selected as the reference data, denoted as X'=[x'1,x'2,...x']. i ',...x' n ]. Where, x i ,x i ' represents the i-th data point in the data sample, and n represents the total number of data points in the data sample.
[0132] 2.3: Calculate the degree of variation at each measuring point in the data sample, denoted as: δ=[δ1, δ2, ... δ i ...δ n ], where the degree of variation of the i-th measuring point is δ i satisfy:
[0133]
[0134] In the formula, cov(x) i ,x i ') represents x i ,x i The covariance of ', var(x) i ),var(x i ') represent x respectively i ,x i The variance of '.
[0135] 2.4: Arrange the elements in δ in descending order and select the data points with the highest degree of variation as key feature items. In this example, the top 5 data points with the highest degree of variation are selected as key feature items.
[0136] In step 3, the key characteristic data of the energy storage battery are preprocessed, including numerical normalization and sparsification. The sparsification process refers to sampling the energy storage charging or discharging process at equal time intervals t. In the embodiment, t = int(T / 144) is preferred, where int means rounding down and T means the duration of one charging or discharging cycle.
[0137] The sample generalization rule in step 4 refers to aggregating and generalizing data samples with similar number of iterations. In this embodiment, it is preferred that every 10 iterations are aggregated into one cycle lifetime state, for a total of 600 states.
[0138] The algorithm module in step 5 is preferably a convolutional neural network algorithm, and the specific algorithm architecture includes:
[0139] 5.1 The input layer has 12×12=144 nodes, and the number of channels is K, which is the key feature item of the sample data. 144 sets of continuous equal time period t sampled key feature data are selected as unit input. In this embodiment, the top 5 features with the highest correlation are preferred as input key feature items.
[0140] 5.2 Convolutional layer I has a kernel size of 3×3, a depth of 32, a stride of 1, and an activation function of ReLU. After passing through convolutional layer I, 32 10×10 feature matrices are obtained.
[0141] 5.3 Pooling layer I, with a filter size of 2×2 and a step size of 2, adopts the maximum pooling rule, resulting in 32 5×5 eigenvalue matrices after pooling.
[0142] 5.4 Convolutional Layer II, with a kernel size of 3×3, a depth of 64, a stride of 1, and an activation function of ReLU, yields 64 3×3 feature matrices.
[0143] 5.5 Pooling layer II, with a filter size of 2×2 and a step size of 2, adopts the maximum pooling rule, padding with zeros for non-uniformity, and obtains 64 2×2 eigenvalue matrices after pooling.
[0144] 5.6 Convolutional layer III, with a kernel size of 2×2, a depth of 120, a stride of 1, and the activation function ReLU, yields a feature vector of length 120 after convolution.
[0145] 5.7 Fully connected layer with 1024 nodes and sigmoid activation function.
[0146] 5.8 Output layer, with 600 nodes and softmax activation function.
[0147] The preferred prediction model in step 5 refers to a model with an accuracy of over 95% on both the training and test sets.
[0148] In step 6, predictions are made periodically, preferably once a day.
[0149] This application discloses a method for predicting the cycle life of energy storage batteries based on convolutional neural networks. It proposes a key feature extraction method based on the degree of variation to ensure the correlation between input features and cycle life. It constructs a convolutional neural network algorithm framework that matches the actual problem, uses key feature data as data samples, and performs sparsification and generalization processing on the samples to reduce the number of neurons in the algorithm, reduce the risk of overfitting, and improve the efficiency and applicability of the algorithm.
[0150] Specifically, considering the numerous parameters and features affecting the cycle life of energy storage batteries, effective key feature extraction and preprocessing methods are essential for accurately characterizing their cycle life. This application utilizes a feature extraction and preprocessing module to extract key features of battery cycle life and sparsifies the input features, thereby improving the algorithm efficiency and prediction accuracy for energy storage battery cycle life prediction. Since energy storage batteries cycle through thousands of times and have numerous calibration states, a reasonable algorithm structure is needed to ensure convergence. This application employs a multi-layer convolutional neural network algorithm to divide the cycle life states of energy storage batteries into multiple predetermined states, reducing training time and improving efficiency and robustness. Furthermore, considering the small differences in operating features when energy storage batteries have similar cycle lives, sample generalization is necessary to reduce the probability of overfitting. A sample feature generalization module is used to partition and generalize the data samples, reducing the number of neurons in the algorithm, lowering the risk of overfitting, and improving efficiency and applicability.
[0151] Example 6
[0152] Based on the above embodiments, this embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the above embodiments.
[0153] In some embodiments of this example, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the above embodiments.
[0154] In some embodiments of this example, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described in the above embodiments.
[0155] The processor may include, but is not limited to, one or more processors or microprocessors. Each processor may be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic component, for executing the methods in the above embodiments.
[0156] Computer-readable storage media can be implemented by any type of volatile or non-volatile storage device or a combination thereof. Computer-readable storage media may include, but are not limited to, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, and computer storage media (e.g., hard disks, floppy disks, solid-state drives, removable disks, CD-ROMs, DVD-ROMs, Blu-ray discs, etc.).
[0157] Computer-readable storage media may also store at least one computer-executable program / instruction, such as computer-readable instructions. Computer-readable storage media include, but are not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Computer-readable storage media may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, a non-transitory computer-readable storage medium may be connected to a computing device such as a computer, and then, when the computing device executes the computer-readable instructions stored on the computer-readable storage medium, the various methods described above can be performed.
[0158] In addition, the computer device may include (but is not limited to) a data bus, an input / output (I / O) bus, a display, and input / output devices (e.g., keyboard, mouse, speakers, etc.).
[0159] The processor can communicate with external devices via the I / O bus through wired or wireless networks.
[0160] In one embodiment, the at least one computer-executable instruction may also be compiled into or comprise a software product / computer program product, wherein one or more computer-executable instructions are executed by a processor to perform the steps of the various functions and / or methods in the embodiments described herein.
[0161] In the embodiments provided in this disclosure, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0162] It should be noted that, in this disclosure, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element limited by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0163] While the embodiments disclosed herein are as described above, the foregoing content is merely for the purpose of facilitating understanding of this disclosure and is not intended to limit this disclosure. Any person skilled in the art to which this disclosure pertains may make any modifications and changes in form and detail of the implementation without departing from the spirit and scope of this disclosure; however, the scope of patent protection of this disclosure shall still be determined by the scope defined in the appended claims.
Claims
1. A method for predicting the cycle life of an energy storage battery, characterized in that, include: Obtain a sample dataset of energy storage batteries, wherein each sample data in the sample dataset includes multiple data measurement points; Determine the degree of variation of each data measurement point, wherein the degree of variation is used to measure the impact of the data measurement point on the cycle life; A preset number of data points are selected as key features in descending order of the degree of variation to obtain a key feature sample dataset of the energy storage battery. Each sample data in the key feature sample dataset includes a preset number of key features. The prediction model is trained using the key feature sample dataset to obtain a trained prediction model, which is used to predict the cycle life of energy storage batteries. The trained prediction model is used to periodically predict the cycle life of the energy storage battery.
2. The method according to claim 1, characterized in that, Determining the degree of variation of each data measurement point includes: The working condition with the longest time span is selected from the sample dataset as the feature evaluation working condition; The earliest data sample in the aforementioned feature evaluation conditions is determined as the basic data; The latest data sample in the aforementioned feature evaluation conditions is determined as the reference data; The degree of variation of each data measurement point is determined according to the following expression: Where, δ i x represents the degree of variation of the i-th data point. i Let x' represent the i-th data point in the basic data. i Cov(x) represents the i-th data point in the reference data. i ,x i ') represents x i ,x i The covariance of ', var(x) i ),var(x i ') represent x respectively i ,x i The variance of '.
3. The method according to claim 1, characterized in that, Before training the prediction model using the key feature sample dataset, the method further includes: The key feature sample dataset is preprocessed to obtain a preprocessed key feature sample dataset. The preprocessing includes numerical normalization and sparsification. The sparsification process involves sampling the key feature sample dataset according to a preset period.
4. The method according to claim 3, characterized in that, The method further includes: The preprocessed key feature sample dataset is aggregated and generalized based on the number of iterations.
5. The method according to claim 1, characterized in that, The prediction model includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a fully connected layer, and an output layer connected in sequence. The input layer nodes are 12×12, and the number of channels is equal to the number of items in the key features; The first convolutional layer has a kernel size of 3×3, a depth of 32, a stride of 1, and an activation function of ReLU. The filter size of the first pooling layer is 2×2, the step size is 2, and the maximum pooling rule is adopted. The second convolutional layer has a kernel size of 3×3, a depth of 64, a stride of 1, and uses the ReLU activation function. The filter size of the second pooling layer is 2×2, the step size is 2, and the maximum pooling rule is adopted; The third convolutional layer has a kernel size of 2×2, a depth of 120, a stride of 1, and an activation function of ReLU. The fully connected layer has 1024 nodes and uses the Sigmod function for activation. The output layer has 600 nodes and uses the Softmax function as its activation function.
6. The method according to any one of claims 1-5, characterized in that, The data measurement points include voltage, current, temperature, electrical charge, state of charge, and depth of charge / discharge.
7. A device for predicting the cycle life of an energy storage battery, characterized in that, include: The acquisition module is used to acquire a sample dataset of energy storage batteries, wherein each sample data in the sample dataset includes multiple data measurement points; The processing module is used to determine the degree of variation of each data measurement point, and the degree of variation is used to measure the impact of the data measurement points on the cycle life. The selection module is used to select a preset number of data points as key features in descending order of the degree of variation, so as to obtain a key feature sample dataset of the energy storage battery. Each sample data in the key feature sample dataset includes a preset number of key features. The training module is used to train the prediction model using the key feature sample dataset to obtain a trained prediction model, which is used to predict the cycle life of the energy storage battery. The prediction module is used to periodically predict the cycle life of the energy storage battery using the trained prediction model.
8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 6.