Prediction method and apparatus for remaining useful life of battery, and electronic device

Abstract

A prediction method and apparatus for the remaining useful life of a battery, and an electronic device (40). The prediction method for the remaining useful life of a battery comprises: acquiring battery data of a target battery within a preset time period, and dividing the battery data, so as to obtain first battery data and second battery data (S101); performing clustering processing on the first battery data, so as to obtain a joint distribution feature, and performing scale decomposition on the second battery data, so as to obtain a frequency component (S102); inputting the joint distribution feature into a feedforward neural network model, so as to output first capacity prediction data, and inputting frequency component data into a time sequence prediction model, so as to output second capacity prediction data (S103); and using the first capacity prediction data and the second capacity prediction data to determine target capacity prediction data, and using the target capacity prediction data to determine the remaining useful life of the target battery (S104).

Description

Methods, devices, and electronic devices for predicting battery lifespan

[0001] This application claims priority to Chinese Patent Application No. 202411647935.9, filed with the Chinese Patent Office on November 18, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of energy technology, such as a method, apparatus, and electronic device for predicting battery life. Background Technology

[0003] With the ever-increasing global demand for sustainable energy, energy storage technology plays an increasingly important role in balancing electricity supply and demand, stabilizing the power grid, and promoting energy transition. As a core component of energy storage and conversion, energy storage batteries are widely used in residential and commercial energy management systems, effectively capturing, storing, and releasing energy. Furthermore, energy storage batteries can provide functions such as frequency regulation, peak load management, and emergency backup power in large-scale power grids. However, due to the intermittent and unpredictable nature of renewable energy, accurate prediction of battery performance degradation and remaining lifespan is crucial not only for battery safety but also for the overall system's economy and reliability. Therefore, efficient and accurate performance prediction of batteries is becoming increasingly important.

[0004] Battery health, including remaining lifespan, is a key parameter that needs to be monitored in energy storage system management. The accuracy of battery performance degradation rate and lifespan prediction is crucial for optimizing battery use, preventing failures, reducing maintenance costs, and achieving efficient energy dispatch. However, during the charging and discharging process, electrochemical reactions, temperature changes, cycle count, and operating environment all significantly impact battery lifespan and performance. For example, the charging and discharging process involves complex electrochemical reactions that are affected by various factors such as current, voltage, temperature, charge / discharge rate, and usage frequency, making battery lifespan prediction a challenging task.

[0005] Current methods for predicting the lifespan of energy storage batteries are mainly divided into two categories: physical models and data-driven models. Physical models rely on electrochemical principles and predict battery life by establishing mathematical models of the internal reaction mechanisms of the battery. These models can theoretically provide a deep understanding of electrochemical processes. However, they are insufficient in dealing with the complexity and nonlinear characteristics of battery performance changes. In practice, they face problems such as high computational complexity, difficulty in parameter adjustment, excessively high requirements for accurate models of the battery's internal structure and materials, and sensitivity to environmental variables. In particular, energy storage batteries are affected by various factors (such as current, voltage, temperature, and cycle number) during charging and discharging, and their electrochemical reaction processes are highly nonlinear. The prediction accuracy and reliability of physical models are significantly limited, resulting in high computational costs, large parameter uncertainties, many simplification assumptions, and insufficient dynamic response.

[0006] Data-driven models use historical and battery operating data as input to predict battery life by learning patterns and regularities hidden within the data. These methods perform well with large-scale datasets, automatically extracting features and reducing reliance on internal battery mechanisms. However, data-driven models still have limitations in capturing nonlinear and dynamic changes in battery performance, are prone to getting trapped in local optima, and lack global search capabilities, thus limiting prediction accuracy. They also have limitations in handling nonlinear and dynamic characteristics of battery performance. Furthermore, these data-driven models typically require large amounts of training data and fine-tuning of parameters, increasing the complexity of deployment and maintenance.

[0007] There is currently no effective solution to the problem of low prediction accuracy when predicting battery capacity in related technologies. Summary of the Invention

[0008] This application provides a method, apparatus, and electronic device for predicting battery usage time, in order to solve the problem of low prediction accuracy in the prediction of battery capacity in related technologies.

[0009] Firstly, this application provides a method for predicting the service life of a battery. The method includes: acquiring battery data of a target battery over a preset time period; dividing the battery data into first battery data and second battery data, wherein the first battery data exhibits nonlinear characteristics, and the second battery data exhibits multi-scale characteristics; performing clustering processing on the first battery data to obtain joint distribution features; performing scale decomposition on the second battery data to obtain frequency components; inputting the joint distribution features into a feedforward neural network model to output first capacity prediction data; inputting the frequency component data into a time series prediction model to output second capacity prediction data, wherein the first capacity prediction data refers to the predicted data obtained by the target battery's capacity being affected by operating conditions, and the second capacity prediction data is used to characterize the trend fluctuation of the target battery's capacity; using the first and second capacity prediction data to determine target capacity prediction data; and using the target capacity prediction data to determine the service life of the target battery.

[0010] Furthermore, clustering the first battery data to obtain joint distribution features includes: extracting features from the first battery data to obtain M battery features, wherein the M battery features include at least the current feature, voltage feature, and battery temperature feature of the target battery, and M is a positive integer; filtering each battery feature using a target filter to obtain M component features, wherein each component feature includes trend features and fluctuation features; inputting the M component features into a clustering algorithm, and processing the M component features using the clustering algorithm to obtain joint distribution features.

[0011] Further, the second battery data is scaled to obtain frequency components, including: extracting features from the second battery data to obtain a battery capacity signal; performing mode decomposition processing on the battery capacity signal using a variational mode decomposition algorithm to obtain N initial frequency components, wherein each initial frequency component includes at least one of the following: a center frequency component, an amplitude component, and a phase component, where N is a positive integer; obtaining a center frequency threshold; dividing the N initial frequency components according to the center frequency threshold; determining the initial frequency components less than the center frequency threshold as the first frequency component; determining the initial frequency components greater than or equal to the center frequency threshold as the second frequency component; and forming a frequency component by the first frequency component and the second frequency component.

[0012] Furthermore, when the time series prediction model is a Long Short-Term Memory (LSTM) network model, the frequency components include a first frequency component and a second frequency component. The LSM network model is used to process the first frequency component. Inputting the frequency component data into the time series prediction model and outputting the second capacity prediction data includes: obtaining the first model parameters associated with the forget gate and obtaining the historical frequency components of the target battery in a historical time period, where the historical time period refers to the time period preceding the preset time period; calculating the forget gate activation value using the first model parameters, the historical frequency components, and the first frequency component; updating the first frequency component using the forget gate activation value to obtain the updated frequency component; inputting the updated frequency component into the output gate and processing the updated frequency component using the output gate to obtain the output frequency component; processing the updated frequency component and the output frequency component using the hyperbolic tangent function to obtain the hidden state component; obtaining the second model parameters and processing the hidden state component using the second model parameters to obtain the second capacity prediction data.

[0013] Furthermore, when the time series prediction model is an autoregressive moving average model, the frequency components include a first frequency component and a second frequency component. The autoregressive moving average model is used to process the second frequency component. Inputting the frequency component data into the time series prediction model and outputting the second capacity prediction data includes: acquiring time window data; dividing the preset time period according to the time window data to obtain P preset time points, where P is a positive integer; using the P preset time points to decompose and represent the second frequency component to obtain an autoregressive component representation, where the autoregressive component representation includes a preset constant, a preset error term, and sub-frequency components associated with the P preset time points, each sub-frequency component being associated with a model weight parameter; constructing an autoregressive moving average equation using the autoregressive component representation; and using the autoregressive moving average equation to determine the second capacity prediction data.

[0014] Furthermore, the feedforward neural network model includes an input layer, a hidden layer, and an output layer. Inputting the joint distribution features into the feedforward neural network model and outputting the first capacity prediction data involves: inputting the joint distribution features into the input layer, segmenting the joint distribution features using the input layer to obtain a segmentation feature vector; inputting the segmentation feature vector into the hidden layer, fitting and strengthening the segmentation feature vector using a non-linear activation function in the hidden layer to obtain a hidden layer output vector; obtaining the bias term and output layer weights of the output layer, calculating the product of the hidden layer output vector and the output layer weights to obtain a vector product, and calculating the sum of the vector product and the bias term to obtain a vector sum; and processing the vector sum using the activation function of the output layer to obtain the first capacity prediction data.

[0015] Further, determining the target capacity prediction data using the first capacity prediction data and the second capacity prediction data includes: determining the first capacity weight based on the data characteristics of the feedforward neural network model, and determining the second capacity weight based on the data characteristics of the time series prediction model; calculating the product of the first capacity weight and the first capacity prediction data to obtain the first prediction product; calculating the product of the second capacity weight and the second capacity prediction data to obtain the second prediction product; and calculating the sum of the first prediction product and the second prediction product to obtain the target capacity prediction data.

[0016] Secondly, this application provides a device for predicting the service life of a battery. The device includes: an acquisition unit configured to acquire battery data of a target battery over a preset time period, divide the battery data to obtain first battery data and second battery data, wherein the first battery data exhibits nonlinear characteristics and the second battery data exhibits multi-scale characteristics; a processing unit configured to perform clustering processing on the first battery data to obtain joint distribution features, and perform scale decomposition on the second battery data to obtain frequency components; an input unit configured to input the joint distribution features into a feedforward neural network model and output first capacity prediction data, and input the frequency component data into a time series prediction model and output second capacity prediction data, wherein the first capacity prediction data refers to the prediction data obtained by the target battery's capacity being affected by operating conditions, and the second capacity prediction data is used to characterize the trend fluctuation of the target battery's capacity; and a determination unit configured to determine target capacity prediction data using the first and second capacity prediction data, and determine the service life of the target battery using the target capacity prediction data.

[0017] Thirdly, this application also provides a computer storage medium for storing a program, wherein the program, when running, controls the device where the computer storage medium is located to execute a method for predicting battery life.

[0018] Fourthly, this application also provides an electronic device comprising one or more processors and a memory; the memory stores computer-readable instructions, and the processor is used to execute the computer-readable instructions, wherein the computer-readable instructions execute a method for predicting battery life.

[0019] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs a method for predicting battery life. Attached Figure Description

[0020] Figure 1 is a flowchart of a method for predicting battery usage time according to an embodiment of this application;

[0021] Figure 2 is a schematic diagram of a battery usage time prediction method provided according to an embodiment of this application;

[0022] Figure 3 is a schematic diagram of a battery usage time prediction device provided according to an embodiment of this application;

[0023] Figure 4 is a schematic diagram of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0024] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application 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 application 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.

[0027] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this disclosure are information and data authorized by the user or fully authorized by all parties. For example, this system has an interface with the relevant user or organization. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving consent from the aforementioned user or organization.

[0028] It should be noted that the information collected in this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize use or refuse use.

[0029] The present application will now be described with reference to optional implementation steps. Figure 1 is a flowchart of a method for predicting battery usage time according to an embodiment of the present application. As shown in Figure 1, the method includes the following steps:

[0030] Step S101: Obtain battery data of the target battery within a preset time period, divide the battery data into first battery data and second battery data, wherein the data characteristics of the first battery data are nonlinear and the data characteristics of the second battery data are multi-scale.

[0031] Specifically, the target battery can refer to an energy storage battery. Battery data can include data such as current, voltage, and temperature of the target battery during the discharge process. In order to predict the remaining capacity of the energy storage battery, it is first necessary to obtain the battery data during the use of the energy storage battery. In addition, if the battery data exhibits obvious nonlinear and multi-scale characteristics, that is, there are both long-term trends and short-term fluctuations in the data, a single model may find it difficult to capture and accurately predict information at both levels simultaneously. For example, the battery capacity will gradually decrease with the increase of the number of cycles, which represents a long-term trend; at the same time, there are also small fluctuations in the battery capacity in each charge-discharge cycle, which is a short-term change. Therefore, the data can be divided into two types of battery data based on the data characteristics, that is, the first battery data and the second battery data are obtained.

[0032] Step S102: Cluster the first battery data to obtain joint distribution features, and scale decompose the second battery data to obtain frequency components.

[0033] Specifically, after obtaining the first battery data and the second battery data after division, in order to improve the prediction accuracy, preprocessing operations such as data cleaning can be performed on these battery data. Since the first battery data represents the removal of outliers and noise in the data, the preprocessed first battery data is clustered to obtain joint distribution features.

[0034] Furthermore, multi-scale decomposition can decompose complex battery data, thereby improving the robustness of the prediction model and reducing prediction errors. Therefore, the preprocessed second battery data can be subjected to mode decomposition to obtain frequency components, which include a first frequency component representing low-frequency components and a second frequency component representing high-frequency components. Each frequency component can be analyzed using a different model.

[0035] Step S103: Input the joint distribution features into the feedforward neural network model and output the first capacity prediction data. Input the frequency component data into the time series prediction model and output the second capacity prediction data. The first capacity prediction data refers to the prediction data obtained by the target battery capacity being affected by the operating conditions. The second capacity prediction data is used to characterize the trend fluctuation of the target battery capacity.

[0036] Specifically, since battery data can be divided into joint distribution features and frequency components, different types of models can be used to process these data when making predictions. Feedforward neural network models can be used to enhance the fitting ability of joint distribution features, thereby outputting the first capacity prediction data. In addition, time series prediction models can be used to process the frequency components. Since the frequency components include high-frequency and low-frequency components, different actual sequence prediction models can be selected for capacity prediction. Among them, the feedforward neural network model can be a multilayer perceptron (MLP) model, and the time series prediction model can be a long short-term memory network (LSTM) model or an auto-regressive integrated moving average (ARIMA) model. The LSTM model can identify low-frequency components, and the ARIMA model can identify high-frequency components.

[0037] Step S104: Use the first capacity prediction data and the second capacity prediction data to determine the target capacity prediction data, and use the target capacity prediction data to determine the usage period of the target battery.

[0038] Specifically, after outputting capacity prediction data based on the different models mentioned above, these data can be weighted and calculated to obtain the final prediction result, that is, the target capacity prediction data. Finally, the usage period of the target battery is determined based on the target capacity prediction data, so that the overall prediction system can comprehensively and accurately capture the changing trends and fluctuations of the battery.

[0039] The battery usage time prediction method provided in this application embodiment acquires battery data of the target battery over a preset time period, divides the battery data into first battery data and second battery data, wherein the first battery data has nonlinear characteristics and the second battery data has multi-scale characteristics; the first battery data is clustered to obtain joint distribution features, and the second battery data is scaled to obtain frequency components; the joint distribution features are input into a feedforward neural network model to output first capacity prediction data, and the frequency component data is input into a time series prediction model to output second capacity prediction data. The first capacity prediction data refers to the predicted data obtained by the target battery's capacity being affected by usage conditions, and the second capacity prediction data... The measured data is used to characterize the trend fluctuation of the target battery's capacity. The target capacity prediction data is determined using the first and second capacity prediction data, and the usage period of the target battery is determined using the target capacity prediction data. This solves the problem of low prediction accuracy in related technologies when predicting battery capacity. By acquiring battery data of the target battery within a preset time period, the battery data is divided into first battery data and second battery data. The two battery data are processed and then input into a feedforward neural network model and a time series prediction model, respectively, and then output two capacity prediction results. Finally, the target capacity prediction data and the usage period of the target battery are determined based on the two capacity prediction results, thereby improving the accuracy of battery capacity prediction.

[0040] The processing of battery data can include various methods. Optionally, in the battery usage time prediction method provided in this application embodiment, clustering the first battery data to obtain joint distribution features includes: extracting features from the first battery data to obtain M battery features, wherein the M battery features include at least the current feature, voltage feature, and battery temperature feature of the target battery, and M is a positive integer; filtering each battery feature using a target filter to obtain M component features, wherein each component feature includes trend features and fluctuation features; inputting the M component features into a clustering algorithm, and processing the M component features using the clustering algorithm to obtain joint distribution features.

[0041] When processing battery data exhibiting nonlinear characteristics, clustering algorithms, such as K-means clustering, can be used. Specifically, feature extraction can first be performed on the battery data to extract battery features such as output current, output voltage, and operating temperature. Then, data filtering is used to output component features. For example, when the battery feature is voltage, extracting trend and fluctuation features using a filter can yield trend features, such as the slope and rate of change of the trend curve, and fluctuation features, such as peak values ​​and waveform periods. It can be obtained from the following formula: The fluctuation characteristics can be obtained by the following formula: ,in, and This refers to the filter parameters. Expressed as the rate of change of voltage. This represents the trend characteristics over a historical period. It should be noted that the filter can be an SG filter (also known as a Savitzky-Golay filter). This filter is a commonly used signal processing filter that uses least-squares fitting to smooth signals, filtering them by fitting a polynomial function to remove noise and extract signal features, effectively improving signal quality and accuracy.

[0042] Furthermore, the component features of each component are input into a clustering algorithm, which processes the component features to obtain joint distribution features. This embodiment utilizes the first battery data to obtain joint distribution features, thereby reflecting the correlation between different battery features and the combination patterns between features, providing more accurate information for subsequent battery state assessment and prediction.

[0043] The processing of battery data can include various methods. Optionally, in the battery usage time prediction method provided in this application embodiment, the scaling of the second battery data to obtain frequency components includes: extracting features from the second battery data to obtain a battery capacity signal; performing mode decomposition processing on the battery capacity signal using a variational mode decomposition algorithm to obtain N initial frequency components, wherein each initial frequency component includes at least one of the following: a center frequency component, an amplitude component, and a phase component, where N is a positive integer; obtaining a center frequency threshold; dividing the N initial frequency components according to the center frequency threshold; determining the initial frequency components less than the center frequency threshold as the first frequency component; determining the initial frequency components greater than or equal to the center frequency threshold as the second frequency component; and constituting the frequency components from the first frequency component and the second frequency component.

[0044] Specifically, while processing the first battery data, it is also necessary to process the second battery data, which exhibits multi-scale characteristics. Variational Mode Decomposition (VMD), a time-frequency analysis method for signal processing, can estimate the intrinsic patterns of a signal through an optimization problem, decomposing complex signals into several oscillation modes with different center frequencies, amplitudes, and modulation characteristics, making these modes localized in the time-frequency domain. Therefore, after obtaining the second battery data and performing feature extraction processing, the processed battery capacity signal is input into the aforementioned variational mode decomposition algorithm. This algorithm decomposes the battery capacity signal into components of different scales, thereby obtaining initial frequency components, each of which can characterize more detailed frequency information.

[0045] It should be noted that the above battery capacity signal can be obtained by decomposing it into N modes, and can be expressed by the following formula: In other words, after performing mode decomposition using the variational principle, the battery capacity signal can be decomposed into amplitude components and phase components.

[0046] The variational principle can be expressed as: The amplitude component can be expressed as: The phase component can be expressed as: ,in, The signal represents the battery capacity, and N represents the number of modes. Therefore, the battery capacity can be represented as the superposition of N modes. Each mode i consists of two parts, namely, an amplitude component and a phase component. At this time, each mode has specific center frequency data, amplitude data and phase data.

[0047] Furthermore, by comparing the center frequency differences between different initial frequency components, the initial frequency components are further divided into a first frequency component representing low-frequency components and a second frequency component representing high-frequency components based on a center frequency threshold. This embodiment extracts features and performs mode decomposition on the battery capacity signal to obtain the frequency components, thereby providing a more comprehensive understanding of the battery's operating state and performance characteristics. This lays the foundation for accurately predicting the battery's lifespan and performance, and also provides more effective support and guidance for battery use and management.

[0048] There are various methods for predicting battery capacity. Optionally, in the battery usage time prediction method provided in this application embodiment, when the time series prediction model is a Long Short-Term Memory (LSTM) network model, the frequency components include a first frequency component and a second frequency component. The LSM network model is used to process the first frequency component. Inputting the frequency component data into the time series prediction model and outputting the second capacity prediction data includes: obtaining the first model parameters associated with the forget gate and obtaining the historical frequency components of the target battery in a historical time period, wherein the historical time period refers to the time period preceding the preset time period; calculating the forget gate activation value using the first model parameters, the historical frequency components, and the first frequency components; updating the first frequency component using the forget gate activation value to obtain the updated frequency component; inputting the updated frequency component into the output gate and processing the updated frequency component using the output gate to obtain the output frequency component; processing the updated frequency component and the output frequency component using the hyperbolic tangent function to obtain the hidden state component; obtaining the second model parameters and processing the hidden state component using the second model parameters to obtain the second capacity prediction data.

[0049] It should be noted that the Long Short-Term Memory (LSTM) network model, as a recurrent neural network model, consists of three parts: a forget gate, an input gate, and an output gate. The coordinated operation of these gates can capture the long-term dependencies between frequency components associated with battery capacity degradation and construct a capacity predictor. Therefore, when using the LSM network model to process frequency components, it is first necessary to obtain the first model parameters associated with the forget gate and the historical frequency components of the energy storage battery over a historical time period. Then, by combining the first model parameters, historical frequency components, and the first frequency component, the output of the forget gate (i.e., the forget gate activation value) is determined. ): This activation value helps determine which information needs to be retained and which needs to be forgotten. This represents the hidden state of frequency components within a historical time period. Indicates the first frequency component. and These represent the first model parameters, specifically the weights and biases of the forget gate. This represents the activation function.

[0050] Furthermore, the first frequency component is updated using the activation value of the forget gate, resulting in the updated frequency component, which can be determined by the following formula: ,in, Historical frequency components representing historical time periods This represents the activation value of the forget gate. The updated frequency components are then sequentially input into the output gate, and the updated frequency components and the output frequency components of the output gate are processed using the hyperbolic tangent function and the second model parameters to obtain the second capacity prediction data. ,in, This represents the sigmoid activation function. and These represent the second model parameters, specifically the weights and biases of the output layer. This embodiment calculates the second capacity prediction data, which helps to better understand the performance and characteristics of the target battery, thus providing a reference for future decision-making and planning.

[0051] To predict high-frequency components, optionally, in the battery usage time prediction method provided in this application embodiment, when the time series prediction model is an autoregressive moving average model, the frequency components include a first frequency component and a second frequency component. The autoregressive moving average model is used to process the second frequency component. Inputting the frequency component data into the time series prediction model and outputting the second capacity prediction data includes: acquiring time window data; dividing a preset time period according to the time window data to obtain P preset time points, where P is a positive integer; using the P preset time points to decompose and represent the second frequency component to obtain an autoregressive component representation, where the autoregressive component representation includes a preset constant, a preset error term, and sub-frequency components associated with the P preset time points, each sub-frequency component being associated with a model weight parameter; constructing an autoregressive moving average equation using the autoregressive component representation; and determining the second capacity prediction data using the autoregressive moving average equation.

[0052] Specifically, the autoregressive moving average model, a statistical model widely used in time series forecasting, models the linear characteristics of data by combining three components: autoregression (AR), differencing (I), and moving average (MA). It can handle linear time series data, thereby predicting short-term changes in target battery performance. When using this model for capacity prediction, the second frequency component is first decomposed and represented, resulting in multiple autoregressive component representations associated with preset time points.

[0053] ;

[0054] in, For the autoregressive component representation, c is a preset constant. , … This represents the model weight parameters, where p is the order of the autoregressive term. This is a preset error term: Then, using the above autoregressive component representation, an autoregressive moving average equation is constructed. Finally, based on the autoregressive moving average equation, the second capacity prediction data is determined. The autoregressive moving average equation can be obtained through the following formula:

[0055] ;

[0056] The second capacity prediction data was determined using the above methods: ;in, This refers to the estimated value of the constant term. This embodiment utilizes three parts—autoregression, differencing, and moving average—to achieve data modeling, and then determines the second capacity prediction data based on the equation. This also allows for a better understanding of the target battery's performance and characteristics, providing a reference for future decision-making and planning.

[0057] Optionally, in the battery usage time prediction method provided in this application embodiment, the feedforward neural network model includes an input layer, a hidden layer, and an output layer. Inputting joint distribution features into the feedforward neural network model and outputting first capacity prediction data includes: inputting the joint distribution features into the input layer, segmenting the joint distribution features using the input layer to obtain a segmentation feature vector; inputting the segmentation feature vector into the hidden layer, fitting and strengthening the segmentation feature vector using a nonlinear activation function in the hidden layer to obtain a hidden layer output vector; obtaining the bias term and output layer weights of the output layer, calculating the product of the hidden layer output vector and the output layer weights to obtain a vector product, and calculating the sum of the vector product and the bias term to obtain a vector sum; and processing the vector sum using the activation function of the output layer to obtain the first capacity prediction data.

[0058] Specifically, since feedforward neural network models can utilize nonlinear activation functions to enhance their ability to fit complex nonlinear relationships, they can be used to predict the capacity of joint distribution features. After inputting these features into the feedforward neural network model, the input layer is used to segment the joint distribution features, resulting in a segmented feature vector. For example, when the segmentation feature vector is When this vector is used, it represents the three-dimensional joint distribution characteristics of current, voltage, and temperature.

[0059] Furthermore, the segmentation feature vector is input into the hidden layer, and the non-linear activation function in the hidden layer is used to fit and enhance the segmentation feature vector to obtain the hidden layer output vector: ,in, This represents the bias term of the hidden layer. This represents the weight of the i-th feature in the j-th hidden layer. The output vector of this hidden layer is then input into the output layer, which outputs the first capacity prediction data. ,in, It is the activation function of the output layer. It is the bias term of the output layer. Z represents the output layer weights, and Z represents the hidden layer output vector. This embodiment utilizes a neural network model to learn and predict joint distribution features, thereby achieving accurate modeling and prediction capabilities for the data.

[0060] Optionally, in the battery usage time prediction method provided in this application embodiment, determining the target capacity prediction data using the first capacity prediction data and the second capacity prediction data includes: determining the first capacity weight based on the data characteristics of the feedforward neural network model, and determining the second capacity weight based on the data characteristics of the time series prediction model; calculating the product of the first capacity weight and the first capacity prediction data to obtain the first prediction product; calculating the product of the second capacity weight and the second capacity prediction data to obtain the second prediction product; and calculating the sum of the first prediction product and the second prediction product to obtain the target capacity prediction data.

[0061] To determine the target capacity prediction data, since the LSTM model is used to predict the long-term degradation trend of the battery, the ARIMA model is used to predict the short-term fluctuations of the battery, and the MLP model is used to predict the short-term operating conditions of the battery, the prediction results at different scales can be fused to improve the accuracy and robustness of the prediction. Specifically, different weights can be assigned to the output of each model according to the prediction performance of different models and the characteristics of the data. That is, the feedforward neural network model mainly considers the nonlinear relationship and complexity of the data, and this model can correspond to the first capacity weight (i.e., Time series forecasting models place greater emphasis on the trend of data, and this model can correspond to a second capacity weight (i.e., and Finally, by using different weights and different capacity prediction data, a weighted sum is calculated to obtain the target capacity prediction data:

[0062] ;in, This represents the first capacity prediction data. This represents the second capacity prediction data output by the Long Short-Term Memory network model. This represents the second capacity prediction data output by the autoregressive moving average model. The target capacity prediction data comprehensively considers the nonlinear relationships, complexity, and time-series characteristics of the data, making it more accurate. This embodiment improves the accuracy and stability of the prediction model by assigning different weights to models with different data characteristics, providing a more reliable reference for decision-making.

[0063] 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.

[0064] This application embodiment also provides a method for predicting battery usage time. Figure 2 is a schematic diagram of an optional method for predicting battery usage time according to an embodiment of this application. As shown in Figure 2, the method includes:

[0065] To predict the remaining capacity of an energy storage battery, it is first necessary to obtain battery data during the battery's use. This data can include the target battery's charging and discharging current, voltage, temperature, etc. To improve prediction accuracy, these battery data can be preprocessed, such as data cleaning, and then divided according to the data characteristics of the preprocessed battery data to obtain first battery data and second battery data.

[0066] Furthermore, K-means clustering is used to cluster the first battery data to obtain joint distribution features. Simultaneously, variational mode decomposition and SG filters are used to scale the second battery data to obtain frequency components. Since battery data can be divided into joint distribution features and frequency components, different types of models can be used to process these data for prediction. A feedforward neural network model (i.e., MLP (Multilayer Perceptron) model) can be used to process the joint distribution features, thus outputting the first capacity prediction data. A Long Short-Term Memory (LSTM) model can be used to predict the low-frequency components, achieving trend prediction. An Auto-Regressive Integrated Moving Average (ARIMA) model can be used to predict the high-frequency components, achieving fluctuation prediction. The second capacity prediction data is then determined from these two prediction data. Finally, the prediction data from the three models are fused, and a weighted average method is used to obtain the final target capacity prediction data.

[0067] It should be noted that before the above three models are used for capacity prediction, the models need to be trained, and the predictive performance of the fusion model should be evaluated using metrics such as root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²) to verify the effectiveness of model fusion.

[0068] This embodiment acquires battery data of the target battery within a preset time period, divides the battery data into first battery data and second battery data, processes the two battery data and inputs them into a feedforward neural network model and a time series prediction model respectively, and then outputs two capacity prediction results. Finally, based on the two capacity prediction results, the target capacity prediction data and the usage time limit of the target battery are determined, thereby improving the accuracy of battery capacity prediction.

[0069] This application also provides a battery usage time prediction device. It should be noted that the battery usage time prediction device of this application can be used to execute the battery usage time prediction method provided in this application. The following describes the battery usage time prediction device provided in this application.

[0070] Figure 3 is a schematic diagram of a battery usage time prediction device provided according to an embodiment of this application. As shown in Figure 3, the device includes: an acquisition unit 30, a processing unit 31, an input unit 32, and a determination unit 33.

[0071] The acquisition unit 30 is configured to acquire battery data of the target battery within a preset time period, divide the battery data to obtain first battery data and second battery data, wherein the data characteristics of the first battery data are nonlinear characteristics, and the data characteristics of the second battery data are multi-scale characteristics.

[0072] Processing unit 31 is configured to perform clustering processing on the first battery data to obtain joint distribution features, and to perform scale decomposition on the second battery data to obtain frequency components;

[0073] Input unit 32 is configured to input the joint distribution features into a feedforward neural network model and output first capacity prediction data, and input the frequency component data into a time series prediction model and output second capacity prediction data. The first capacity prediction data refers to the prediction data obtained by the target battery capacity being affected by the operating conditions, and the second capacity prediction data is used to characterize the trend fluctuation of the target battery capacity.

[0074] The determining unit 33 is configured to determine target capacity prediction data using the first capacity prediction data and the second capacity prediction data, and to determine the usage period of the target battery using the target capacity prediction data.

[0075] The battery usage time prediction device provided in this application embodiment acquires battery data of the target battery over a preset time period through an acquisition unit 30, divides the battery data to obtain first battery data and second battery data, wherein the data characteristics of the first battery data are nonlinear and the data characteristics of the second battery data are multi-scale; a processing unit 31 performs clustering processing on the first battery data to obtain joint distribution features, and performs scale decomposition on the second battery data to obtain frequency components; an input unit 32 inputs the joint distribution features into a feedforward neural network model and outputs first capacity prediction data, and inputs the frequency component data into a time series prediction model and outputs second capacity prediction data, wherein the first capacity prediction data refers to the predicted capacity of the target battery affected by the operating conditions. The first capacity prediction data and the second capacity prediction data are used to characterize the trend fluctuation of the target battery's capacity. The determining unit 33 uses the first capacity prediction data and the second capacity prediction data to determine the target capacity prediction data, and uses the target capacity prediction data to determine the usage period of the target battery. This solves the problem of low prediction accuracy when predicting battery capacity in related technologies. By acquiring battery data of the target battery in a preset time period, the battery data is divided into first battery data and second battery data. The two battery data are processed and then input into a feedforward neural network model and a time series prediction model, respectively, and then output two capacity prediction results. Finally, the target capacity prediction data and the usage period of the target battery are determined based on the two capacity prediction results, thereby improving the accuracy of battery capacity prediction.

[0076] Optionally, in the battery usage time prediction device provided in this application embodiment, the processing unit 31 includes: a first extraction module, configured to extract features from the first battery data to obtain M battery features, wherein the M battery features include at least the current feature, voltage feature, and battery temperature feature of the target battery, and M is a positive integer; a first processing module, configured to filter each battery feature using a target filter to obtain M component features, wherein each component feature includes trend features and fluctuation features; and a first input module, configured to input the M component features into a clustering algorithm and process the M component features using the clustering algorithm to obtain the joint distribution features.

[0077] Optionally, in the battery usage time prediction device provided in this application embodiment, the processing unit 31 includes: a second extraction module, configured to extract features from the second battery data to obtain a battery capacity signal; a second processing module, configured to perform mode decomposition processing on the battery capacity signal using a variational mode decomposition algorithm to obtain N initial frequency components, wherein each initial frequency component includes at least one of the following: a center frequency component, an amplitude component, and a phase component, and N is a positive integer; and a first acquisition module, configured to acquire a center frequency threshold, divide the N initial frequency components according to the center frequency threshold, determine the initial frequency components less than the center frequency threshold as first frequency components, determine the initial frequency components greater than or equal to the center frequency threshold as second frequency components, and constitute the frequency component by the first frequency component and the second frequency component.

[0078] Optionally, in the battery usage time prediction device provided in this application embodiment, the input unit 32 includes: a second acquisition module, configured to acquire first model parameters associated with the forget gate and acquire the historical frequency components of the target battery in a historical time period, wherein the historical time period refers to the time period preceding the preset time period; a first calculation module, configured to calculate using the first model parameters, the historical frequency components, and the first frequency components to obtain a forget gate activation value; an update module, configured to update the first frequency components using the forget gate activation value to obtain an updated frequency component; a second input module, configured to input the updated frequency component into an output gate and process the updated frequency component using the output gate to obtain an output frequency component; a third processing module, configured to process the updated frequency component and the output frequency component using a hyperbolic tangent function to obtain a hidden state component; and a third acquisition module, configured to acquire second model parameters and process the hidden state component using the second model parameters to obtain the second capacity prediction data.

[0079] Optionally, in the battery usage time prediction device provided in this application embodiment, the input unit 32 includes: a fourth acquisition module, configured to acquire time window data, and divide the preset time period according to the time window data to obtain P preset times, where P is a positive integer; a splitting module, configured to use the P preset times to split the second frequency component to obtain an autoregressive component representation, wherein the autoregressive component representation includes a preset constant, a preset error term, and sub-frequency components associated with the P preset times, and each sub-frequency component is associated with a model weight parameter; and a construction module, configured to use the autoregressive component representation to construct an autoregressive moving average equation, and use the autoregressive moving average equation to determine the second capacity prediction data.

[0080] Optionally, in the battery usage time prediction device provided in this application embodiment, the input unit 32 includes: a third input module, configured to input the joint distribution features into the input layer, and use the input layer to segment the joint distribution features to obtain a segmentation feature vector; a fourth input module, configured to input the segmentation feature vector into the hidden layer, and use the nonlinear activation function in the hidden layer to fit and strengthen the segmentation feature vector to obtain a hidden layer output vector; a fifth acquisition module, configured to acquire the bias term and output layer weights of the output layer, calculate the product of the hidden layer output vector and the output layer weights to obtain a vector product, and calculate the sum of the vector product and the bias term to obtain a vector sum; and a fourth processing module, configured to process the vector sum using the activation function of the output layer to obtain the first capacity prediction data.

[0081] Optionally, in the battery usage time prediction device provided in this application embodiment, the determining unit 33 includes: a determining module, configured to determine a first capacity weight based on the data characteristics of the feedforward neural network model, and to determine a second capacity weight based on the data characteristics of the time series prediction model; a second calculation module, configured to calculate the product of the first capacity weight and the first capacity prediction data to obtain a first prediction product, and to calculate the product of the second capacity weight and the second capacity prediction data to obtain a second prediction product; and a third calculation module, configured to calculate the sum of the first prediction product and the second prediction product to obtain the target capacity prediction data.

[0082] The aforementioned battery usage time prediction device includes a processor and a memory. The aforementioned acquisition unit 30, processing unit 31, input unit 32, determination unit 33, etc., are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0083] The processor contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured, and adjusting kernel parameters can address the low prediction accuracy issue in related technologies when predicting battery capacity.

[0084] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0085] This application provides a computer storage medium for storing a program, wherein the program, when running, controls the device where the computer storage medium is located to execute a method for predicting battery life.

[0086] Figure 4 is a schematic diagram of an electronic device provided according to an embodiment of this application. As shown in Figure 4, this application provides an electronic device 40, which includes a processor, a memory, and a program stored in the memory and executable on the processor. The processor is used to execute computer-readable instructions, wherein the computer-readable instructions execute a method for predicting battery life. The device in this document may be a server, PC, PAD, mobile phone, etc.

[0087] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of a battery usage time prediction method according to various embodiments of this application.

[0088] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0090] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0091] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0092] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0093] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0094] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

Claims

1. A method for predicting battery lifespan, comprising: The battery data of the target battery during a preset time period is obtained, and the battery data is divided to obtain first battery data and second battery data. The data characteristics of the first battery data are nonlinear, and the data characteristics of the second battery data are multi-scale. Clustering is performed on the first battery data to obtain joint distribution features, and scale decomposition is performed on the second battery data to obtain frequency components; The joint distribution features are input into a feedforward neural network model to output first capacity prediction data. The frequency component data are input into a time series prediction model to output second capacity prediction data. The first capacity prediction data refers to the prediction data obtained by the target battery capacity being affected by the operating conditions. The second capacity prediction data is used to characterize the trend fluctuation of the target battery capacity. The target capacity prediction data is determined using the first capacity prediction data and the second capacity prediction data, and the usage period of the target battery is determined using the target capacity prediction data.

2. The method of claim 1, wherein, Clustering is performed on the first battery data to obtain joint distribution features, including: The first battery data is subjected to feature extraction to obtain M battery features, wherein the M battery features include at least the current feature, voltage feature and battery temperature feature of the target battery, and M is a positive integer; Each battery feature is filtered using a target filter to obtain M component quantitative features, where each component quantitative feature includes trend features and fluctuation features; The M-component quantitative features are input into a clustering algorithm, and the clustering algorithm is used to process the M-component quantitative features to obtain the joint distribution features.

3. The method of claim 1, wherein, The second battery data is scaled to obtain frequency components, including: The second battery data is used to extract features to obtain the battery capacity signal; The battery capacity signal is processed by variational mode decomposition algorithm to obtain N initial frequency components, wherein each initial frequency component includes at least one of the following: center frequency component, amplitude component and phase component, and N is a positive integer; A center frequency threshold is obtained, and the N initial frequency components are divided according to the center frequency threshold. The initial frequency components that are less than the center frequency threshold are determined as the first frequency components, and the initial frequency components that are greater than or equal to the center frequency threshold are determined as the second frequency components. The first frequency components and the second frequency components constitute the frequency component.

4. The method of claim 1, wherein, When the time series prediction model is a Long Short-Term Memory (LSTM) network model, the frequency components include a first frequency component and a second frequency component. The LTM network model processes the first frequency component, inputs the frequency component data into the time series prediction model, and outputs second capacity prediction data, including: Obtain the first model parameters associated with the forget gate, and obtain the historical frequency components of the target battery in the historical time period, wherein the historical time period refers to the time period preceding the preset time period; The forget gate activation value is obtained by calculating using the first model parameters, the historical frequency components, and the first frequency components. The first frequency component is updated using the forget gate activation value to obtain the updated frequency component; The updated frequency component is input into the output gate, and the updated frequency component is processed by the output gate to obtain the output frequency component. The updated frequency components and the output frequency components are processed using the hyperbolic tangent function to obtain the hidden state components; Obtain the second model parameters, and use the second model parameters to process the hidden state components to obtain the second capacity prediction data.

5. The method of claim 1, wherein, When the time series forecasting model is an autoregressive moving average model, the frequency components include a first frequency component and a second frequency component. The autoregressive moving average model is used to process the second frequency component, inputting the frequency component data into the time series forecasting model and outputting second capacity forecast data, including: Obtain time window data, and divide the preset time period according to the time window data to obtain P preset time moments, where P is a positive integer; The second frequency component is decomposed and represented using the P preset time points to obtain an autoregressive component representation, wherein the autoregressive component representation includes a preset constant, a preset error term, and sub-frequency components associated with the P preset time points, and each sub-frequency component is associated with a model weight parameter. The autoregressive component representation is used to construct an autoregressive moving average equation, and the second capacity prediction data is determined using the autoregressive moving average equation.

6. The method of claim 1, wherein, The feedforward neural network model includes an input layer, a hidden layer, and an output layer. The joint distribution features are input into the feedforward neural network model, and the first capacity prediction data is output, including: The joint distribution features are input into the input layer, and the input layer is used to segment the joint distribution features to obtain a segmentation feature vector; The segmentation feature vector is input into the hidden layer, and the segmentation feature vector is fitted and enhanced using the non-linear activation function in the hidden layer to obtain the output vector of the hidden layer. Obtain the bias term and output layer weights of the output layer, calculate the product of the hidden layer output vector and the output layer weights to obtain the vector product, and calculate the sum of the vector product and the bias term to obtain the vector sum; The first capacity prediction data is obtained by processing the vector sum using the activation function of the output layer.

7. The method of claim 1, wherein, Determining target capacity prediction data using the first capacity prediction data and the second capacity prediction data includes: The first capacity weight is determined based on the data characteristics of the feedforward neural network model, and the second capacity weight is determined based on the data characteristics of the time series prediction model. Calculate the product of the first capacity weight and the first capacity prediction data to obtain the first prediction product; calculate the product of the second capacity weight and the second capacity prediction data to obtain the second prediction product. The sum of the first prediction product and the second prediction product is calculated to obtain the target capacity prediction data.

8. A device for predicting battery life, comprising: The acquisition unit is configured to acquire battery data of the target battery within a preset time period, divide the battery data to obtain first battery data and second battery data, wherein the data characteristics of the first battery data are nonlinear characteristics, and the data characteristics of the second battery data are multi-scale characteristics. The processing unit is configured to perform clustering processing on the first battery data to obtain joint distribution features, and to perform scale decomposition on the second battery data to obtain frequency components; The input unit is configured to input the joint distribution features into a feedforward neural network model and output first capacity prediction data, and input the frequency component data into a time series prediction model and output second capacity prediction data. The first capacity prediction data refers to the prediction data obtained by the target battery capacity being affected by the operating conditions, and the second capacity prediction data is used to characterize the trend fluctuation of the target battery capacity. The determining unit is configured to determine target capacity prediction data using the first capacity prediction data and the second capacity prediction data, and to determine the usage period of the target battery using the target capacity prediction data.

9. A computer program product comprising a computer program that, when executed by a processor, implements the method for predicting battery life as described in any one of claims 1 to 7.

10. An electronic device, comprising one or more processors and memory storing one or more programs, wherein, When the one or more programs are executed by the one or more processors, the one or more processors implement the battery life prediction method according to any one of claims 1 to 7.

Images