WO2024045567A1 - 电池寿命预测方法、系统、终端设备及计算机可读介质 - Google Patents

电池寿命预测方法、系统、终端设备及计算机可读介质 Download PDF

Info

Publication number
WO2024045567A1
WO2024045567A1 PCT/CN2023/082425 CN2023082425W WO2024045567A1 WO 2024045567 A1 WO2024045567 A1 WO 2024045567A1 CN 2023082425 W CN2023082425 W CN 2023082425W WO 2024045567 A1 WO2024045567 A1 WO 2024045567A1
Authority
WO
WIPO (PCT)
Prior art keywords
working condition
battery
battery life
condition parameter
model
Prior art date
Application number
PCT/CN2023/082425
Other languages
English (en)
French (fr)
Inventor
李爱霞
余海军
谢英豪
李长东
Original Assignee
广东邦普循环科技有限公司
湖南邦普循环科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 广东邦普循环科技有限公司, 湖南邦普循环科技有限公司 filed Critical 广东邦普循环科技有限公司
Publication of WO2024045567A1 publication Critical patent/WO2024045567A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing

Definitions

  • This application relates to the technical field of battery life prediction, for example, to a battery life prediction method, system, terminal equipment and computer-readable medium.
  • Lithium battery is currently the battery system with the best comprehensive performance. It has the characteristics of high specific energy, high cycle life, small size, light weight, no memory effect, and no pollution. It has rapidly developed into a new generation of energy storage power supply and is an important part of information technology, Powerful support for electric vehicles, hybrid vehicles and aerospace. Among them, the battery life will gradually decrease with use. Improper use may even cause direct damage to the battery, thus posing safety risks to electrical equipment. Therefore, how to effectively predict battery life and grasp battery usage information is very critical.
  • a data-driven method is usually used to predict the life of lithium batteries. This method is based on data prediction algorithms such as machine learning and neural networks to predict the capacity attenuation of lithium-ion batteries.
  • data prediction algorithms such as machine learning and neural networks to predict the capacity attenuation of lithium-ion batteries.
  • LSTM Long Short-Term Memory
  • Methods in related technologies often only roughly consider the factors that may affect battery life, and collect real data corresponding to these factors to directly train the model. However, they do not consider the difference in the impact of multiple factors on battery life, and lack pertinence. Direct training with a large amount of data will directly increase the difficulty of model training and greatly reduce the efficiency of model training.
  • related technical methods can usually only provide prediction results of battery life, but do not provide any scientifically valid suggestions on how this result can guide users' battery use behavior.
  • This application provides a battery life prediction method, system, terminal equipment and computer-readable medium, which at least solves the problem of inaccurate battery life prediction results, low efficiency and inability to provide scientific guidance in related technologies to help users use batteries correctly, thereby extending battery life.
  • One of the issues such as lifespan.
  • This application provides a battery life prediction method, including:
  • the first working condition parameter includes a combination of multiple indicators that affect battery life according to different weights;
  • the second working condition parameter is an indicator that characterizes battery life;
  • the hybrid neural network model is trained with the historical data of the first working condition parameters as input and the historical data of the second working condition parameters as output, until the model converges, and a target prediction model is generated;
  • the battery life level in the current prediction result is determined to match the corresponding usage recommendations.
  • This application also provides a battery life prediction system, including:
  • the parameter acquisition unit is configured to acquire the first working condition parameter and the second working condition parameter of the battery; wherein the first working condition parameter includes a combination of multiple indicators that affect battery life according to different weights; the second working condition parameter is Indicators that characterize battery life;
  • the model training unit is configured to be based on a hybrid neural network model, using historical data of the first operating condition parameters as input and historical data of the second operating condition parameters as output to train the hybrid neural network model until the model converges, generating a target prediction.
  • Model
  • the life prediction unit is configured to input the battery’s data to be predicted into the target prediction model and generate a prediction test results
  • the use suggestion matching unit is configured to determine the battery life level in the current prediction result based on the pre-divided battery life level to match the corresponding usage recommendations.
  • This application also provides a terminal device, including:
  • a memory coupled to the processor and configured to store at least one program
  • the at least one processor When the at least one program is executed by the at least one processor, the at least one processor is caused to implement the battery life prediction method as described in any one of the above items.
  • the present application also provides a computer-readable medium on which a computer program is stored.
  • the computer program is executed by a processor, the battery life prediction method as described in any one of the above items is implemented.
  • Figure 1 is a schematic flowchart of a battery life prediction method provided by an embodiment of the present application
  • FIG. 2 is a schematic flowchart of the sub-steps of step S40 in Figure 1;
  • Figure 3 is a schematic structural diagram of a battery life prediction system provided by an embodiment of the present application.
  • Figure 4 is a schematic structural diagram of a terminal device provided by an embodiment of the present application.
  • the battery life prediction method includes steps S10 to S40:
  • the first working condition parameter includes a combination of multiple indicators that affect the battery life according to different weights; the second working condition parameter is an indicator that characterizes the battery life. index.
  • the first working condition parameters mainly include multiple influencing factors (indicators) related to battery life, and are represented by a weighted combination without weighting.
  • the second working condition parameter is directly set to represent battery life.
  • the impact indicators related to battery life include battery state of charge (State of Charge, SOC), battery health state (State of Health, SOH), battery power state (State of Power, SOP), battery total voltage, battery total Current, battery cell voltage, battery temperature, maximum battery temperature, minimum battery temperature, battery alarm level, battery alarm category, cumulative charge and discharge capacity, cumulative charge and discharge times, rated battery capacity, relay status, battery balancing type and status , the usage status of electrical equipment and the environment (temperature, humidity) where the electrical equipment is located, etc.
  • SOC battery state of charge
  • SOH battery health state
  • SOP battery power state
  • battery total voltage battery total Current
  • battery cell voltage battery temperature
  • maximum battery temperature minimum battery temperature
  • battery alarm level battery alarm category
  • cumulative charge and discharge capacity cumulative charge and discharge times
  • rated battery capacity relay status
  • battery balancing type and status the usage status of electrical equipment and the environment (temperature, humidity) where the electrical equipment is located, etc.
  • the target model and this method is not conducive to studying the correlation between different indicators.
  • Each indicator is used as an isolated element to train the model.
  • the dimensions are many and divergent, and the learning effect of the final model is difficult to achieve the ideal state. Therefore, in practical applications, we usually focus more on the impact of one or several factors on battery life, considering the correlation between them and how they individually and jointly affect battery life.
  • the first working condition parameter in step S10 can be selected from at least two indicators among ambient temperature, state of charge, depth of discharge, total battery voltage, total battery current and cumulative charge and discharge capacity.
  • a combination of indicators composed of weights.
  • the index combination corresponding to the first working condition parameter in this embodiment can be arbitrarily selected from at least two of the above parameters according to actual needs, for example, the two indicators of ambient temperature and state of charge are selected.
  • the indicator combination is obtained by weighting according to their respective weights. Then what is studied at this time is the impact of the two factors of ambient temperature and state of charge (SOC) on battery life; or the three indicators of ambient temperature, state of charge and discharge depth are selected. , obtain the indicator combination according to their respective weights, and then train the model. In this case, the model learns the impact of these three factors on battery life.
  • SOC state of charge
  • index types of the first working condition parameters pointed out in this embodiment include the above-mentioned categories of ambient temperature, state of charge, depth of discharge, total battery voltage, total battery current and cumulative charge and discharge capacity, but are not limited to these working conditions. parameter. In practical applications, other influencing factors related to battery life can also be obtained and weighted to obtain an indicator combination. Therefore, this embodiment does not place any restrictions on the types and quantities of indicators in the indicator combination.
  • determining the weights of multiple indicators in the first working condition parameters includes:
  • an indicator combination hierarchical model about the first working condition parameters is constructed, and the analytic hierarchy process is used to determine the weight corresponding to each indicator in the indicator combination hierarchical model.
  • the implementation principle of the analytic hierarchy process is to decompose the problem into different component factors according to the nature of the problem and the overall goal to be achieved, and to aggregate and combine the factors at different levels according to the interrelated influences and affiliations between the factors to form A multi-level analytical structure model, which ultimately boils down the problem to the determination of the relative importance of the lowest level (programs, measures, etc. for decision-making) relative to the highest level (overall goal) or the arrangement of relative priorities.
  • the weight of each indicator can be effectively determined through the analytic hierarchy process, and finally a weighted indicator combination as input to the model is obtained.
  • the second working condition parameter in step S10 includes the remaining capacity of the battery or the internal resistance of the battery.
  • the remaining capacity of the battery and the internal resistance of the battery can be selected, where Either one can be used as an indicator to evaluate battery life, or two indicators can be combined to evaluate battery life with the combined indicator.
  • the data can also be preprocessed, including data cleaning, removing outliers and Duplicate values, fill missing values, etc.
  • the data can also be normalized and data of the same dimension can be used for model training. It should be noted that the order of data normalization and data cleaning is not limited here.
  • the sample data obtained after preprocessing and normalizing the acquired historical data of the first working condition parameter and the second working condition parameter can be divided into a training set and a test set according to a preset proportion.
  • Sets for example, allocate them in a ratio of 8:2, use the training set to train the model, and use the test set to detect the prediction effect of the model.
  • 8:2 is just an optional distribution ratio. In actual applications, other ratios can be selected as needed, and this embodiment does not impose any restrictions.
  • the algorithms used for battery life prediction are usually a single neural network model, such as LSTM model or Recurrent Neural Network (RNN) model or Convolutional Neural Networks (CNN) model, but these models have their own
  • LSTM model Recurrent Neural Network
  • CNN Convolutional Neural Networks
  • one-dimensional convolutional neural network is often used to deal with timing problems.
  • ANN artificial neural network
  • the LSTM network continuously updates the hidden layer state of the network through the self-circulating memory unit inside the cell, so it has better timing processing capabilities and generalization capabilities.
  • this embodiment uses a hybrid neural network model, such as an LSTM-CNN hybrid model.
  • the LSTM-CNN hybrid model used may be a sequence model composed of 8 layers.
  • the first two layers of the model are composed of LSTM, each LSTM has 32 neurons, and the excitation The live function is ReLU.
  • the LSTM output dimension needs to be changed at the connection between the LSTM layer and the CNN layer, because the output of LSTM has 3 dimensions (number of samples, time step, input), while CNN requires 4 dimensional input (number of samples, 1, time step) ,enter).
  • the first CNN layer has 64 neurons
  • the second CNN layer has 128 neurons.
  • a maximum pooling layer can be set up to perform the downsampling operation; then there is a Global Average Pooling (GAP) layer, which converts the multi-D feature map into 1 -D feature vector, because no parameters are needed in this layer, it will reduce the global model parameters; then there is the batch normalization (BN) layer, which helps the convergence of the model; the last layer is the output of the model layer, this output layer is just a fully connected layer of 6 neurons with a SoftMax classifier layer that represents the probability of the current class.
  • GAP Global Average Pooling
  • BN batch normalization
  • the initial network parameters are set before training, including the following:
  • the optimizer can use Adam, whose optimization effect is better than the stochastic gradient descent method (Stochastic Gradient Descent, SGD);
  • the number of iterations can be set to 1000 by default. When the accuracy of the training set and the accuracy of the test set are not much different, the current number of iterations can be considered reasonable. When the accuracy of the training set and the accuracy of the test set are significantly different, the number of iterations can be increased. Continue training as many times as you want.
  • batch_size can default to 1. Generally speaking, the smaller the batch_size, the higher the accuracy
  • dropout module to prevent over-fitting of the model.
  • dropout can be selected between 0.4-0.5, which has better performance. The position only needs to be before the last layer of softmax;
  • Loss function Select regression loss, using Mean Absolute Error (MAE) or Mean Square Error (MSE).
  • the mean absolute error measures the average error margin of the distance between the predicted value and the true value, and its range is from 0 to positive infinity. It is characterized by fast convergence speed and ability to The gradient gives appropriate penalty weights instead of "treating everyone equally" so that the direction of the gradient update can be more accurate.
  • the disadvantage is that it is very sensitive to outliers, the direction of gradient update is easily dominated by outliers, and it is not robust.
  • the mean square error (MSE) measures the sum of squares of the distance between the predicted value and the true value, and its range is from 0 to positive infinity. The characteristic is that it is more robust to outliers or outliers.
  • the disadvantage is that the derivative at 0 point is discontinuous, which makes the solution inefficient and leads to slow convergence.
  • the gradient is as large as the gradient of other interval loss values, so it is not conducive to network learning.
  • MAE mean absolute error
  • the model prediction can be considered If the accuracy meets the requirements, the corresponding target prediction model is generated.
  • the data to be predicted is obtained, which can be real-time data under the first working condition parameters to predict battery life.
  • step S40 includes the following sub-steps:
  • three battery life levels are usually preset, namely level one, level two, or level three.
  • level one if the remaining battery capacity is selected as the indicator for evaluating battery life, then two capacity thresholds can be set to divide Lifetime rating, for example:
  • the battery life level is rated as Level 1;
  • the battery life level is rated as Level 2;
  • the battery life level is rated as level three.
  • the battery life level is level one, it means that the battery is in relatively good condition.
  • the historical data of the third working condition parameters can be obtained, that is, the battery temperature under historical working conditions, and then the battery temperature under historical working conditions can be calculated. Is it outside the temperature range under normal working conditions? If there is a situation where the battery temperature is outside the temperature range under normal working conditions or the battery temperature is too high for more than a certain number of times, then the user can be reminded to pay attention when using the battery. Detect the battery temperature in a timely manner. Once the temperature exceeds the standard, the user can be reminded to suspend use. After the battery returns to normal temperature, it can be recommended to continue using it.
  • the battery life level is Level 2
  • you can also obtain historical data of the fourth working condition parameters including but not limited to charging time, Discharge time, charge and discharge times and charge and discharge rate. After obtaining these parameters, determine whether the battery's historical charging time, discharge time, number of charge and discharge times, and charge and discharge rate are outside the rated range. If so, reasonable suggestions can be provided to the user. For example, the charging time should not exceed 5 hours at a time. It can be charged when 20% of the battery is left. Do not charge after complete discharge. The number of charging and discharging should not exceed 5 times in a week. The charging and discharging current should not be too large and should match the battery power, etc.
  • An embodiment of the present application also provides a battery life prediction system, including:
  • the parameter acquisition unit 01 is configured to acquire the first working condition parameter and the second working condition parameter of the battery; wherein the first working condition parameter includes an index combination composed of multiple indicators that affect battery life according to different weights; the second Working condition parameters are indicators that characterize battery life;
  • the model training unit 02 is set to be based on a hybrid neural network model, using the historical data of the first working condition parameters as input and the historical data of the second working condition parameters as the output to train the hybrid neural network model until the model converges, generating a target predictive models;
  • the life prediction unit 03 is configured to input the battery's data to be predicted into the target prediction model and generate prediction results;
  • the usage suggestion matching unit 04 is configured to determine the battery life grade in the current prediction result based on the pre-divided battery life grade to match the corresponding usage suggestion.
  • the battery life prediction system provided by this embodiment is configured to perform the battery life prediction method as described in any of the above embodiments and achieve the same effect, which will not be described again here.
  • An embodiment of the present application also provides a terminal device, including:
  • a memory coupled to the processor and configured to store at least one program
  • the at least one processor When the at least one program is executed by the at least one processor, the at least one processor implements the battery life prediction method as described above.
  • the processor is configured to control the overall operation of the terminal device to complete all or part of the steps of the above battery life prediction method.
  • the memory is configured to store various types of data to support operations on the terminal device. These data may include, for example, instructions for any application program or method operating on the terminal device, as well as application program-related data.
  • the memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory, ROM), magnetic memory, flash memory, magnetic disk or optical disk.
  • SRAM Static Random Access Memory
  • EEPROM Electrically Erasable Programmable Read-Only Memory
  • EPROM Erasable Programmable Read-Only Memory
  • PROM Programmable Read-Only Memory
  • ROM Read-
  • the terminal device may be configured with at least one application specific integrated circuit (Application Specific Integrated Circuit, AS1C), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (Field Programmable Gate Array, FPGA), controller, microcontroller, microprocessor or other electronic components are implemented, configured to perform the battery life prediction method as described in any of the above embodiments, and achieve consistency with the above method. technical effects.
  • AS1C Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • DSPD Digital Signal Processing Device
  • PLD Programmable Logic Device
  • FPGA Field Programmable Gate Array
  • controller microcontroller, microprocessor or other electronic components
  • a computer-readable medium including a computer program is also provided.
  • the steps of the battery life prediction method as described in any of the above embodiments are implemented.
  • the computer-readable medium can be the above-mentioned memory including a computer program.
  • the above-mentioned computer program can be executed by a processor of the terminal device to complete the battery life prediction method as described in any of the above embodiments, and achieve the same results as the above method. technical effects.
  • the computer-readable medium in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two.
  • Examples of computer readable storage media include at least (non-exhaustive list) the following: electrical connection with at least one wiring (electronic device), portable computer disk case (magnetic device), random access memory (Random Access Memory, RAM) , read-only memory (Read-Only Memory, ROM), erasable programmable read-only memory (EPROM) or flash memory, fiber optic devices, and portable read-only memory (Compact Disc Read-Only Memory, CDROM).
  • the computer-readable storage medium may even be paper or other suitable medium on which the program may be printed, as the program may be printed, for example, by optical scanning of the paper or other medium, followed by editing, interpretation, or in other suitable manner if necessary Processing is performed to obtain a program electronically and store it in computer memory.
  • This application is trained based on the LSTM-CNN hybrid neural network model, which has better prediction results than a single LSTM model or CNN convolutional network model.
  • This application considers the impact of different indicators on battery life, combines the analytic hierarchy process to determine the weight of the impact indicators, and uses the weighted results of indicators and weights as model input to train the model. This improves the accuracy of model prediction and is more targeted, avoiding the situation of using a large number of isolated indicators to increase the difficulty of model training, and improving the efficiency of model training.
  • This application classifies battery life in advance, determines the current battery life class after obtaining the prediction results, and matches different usage suggestions according to different levels, which can provide scientific guidance for users to use the battery correctly and help extend battery life. , improving the safety and stability of battery use.
  • This application uses a hybrid neural network model and uses the weight combination of multiple impact indicators as input to train the model, which improves the prediction accuracy of the model and provides different usage suggestions for different life situations, thereby scientifically guiding users to use the battery and helping to extend the battery life. service life.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Hardware Design (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Geometry (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

本申请公开了一种电池寿命预测方法，包括：获取电池的第一工况参数和第二工况参数；第一工况参数包括影响电池寿命的多个指标按不同权重构成的指标组合；第二工况参数为表征电池寿命的指标；基于混合神经网络模型，以第一工况参数的历史数据为输入、第二工况参数的历史数据为输出训练混合神经网络模型，直至模型收敛时生成目标预测模型；将电池的待预测数据输入至目标预测模型生成预测结果；确定当前预测结果中的电池寿命等级，以匹配对应的使用建议。

Description

电池寿命预测方法、系统、终端设备及计算机可读介质
本申请要求在2022年8月31日提交中国专利局、申请号为202211051555.X的中国专利申请的优先权，该申请的全部内容通过引用结合在本申请中。
技术领域
本申请涉及电池寿命预测技术领域，例如涉及一种电池寿命预测方法、系统、终端设备及计算机可读介质。
背景技术
锂电池是目前综合性能最好的电池体系，具有高比能量、高循环寿命、体积小、质量轻、无记忆效应、无污染等特点，已经迅速发展成为新一代储能电源，是信息技术、电动车、混合动力车和航空航天等领域的动力支持。其中，电池寿命会随着使用过程而逐渐衰减，若使用不当甚至会导致电池直接损坏，从而为用电设备带来安全隐患。因此，如何有效对电池寿命进行预测，以掌握电池使用信息十分关键。
相关技术中通常采用数据驱动的方法预测锂电池寿命，该方法基于机器学习和神经网络等数据预测算法实现锂离子电池的容量衰减预测。然而，相关技术的方法在训练模型时，习惯于采用单一的神经网络模型，例如长短期记忆网络(Long Short-Term Memory，LSTM)模型进行预测，通常会出现数据过拟合问题，进而影响模型预测结果的准确性。相关技术的方法往往只是粗略考虑了可能影响电池寿命的因素，采集这些因素所对应的真实数据直接训练模型。但它们并没有考虑多个因素对电池寿命影响程度的区别，缺乏针对性，且用大量的数据直接训练会直接增加模型训练的困难度，大大降低了模型训练的效率。 此外，相关技术的方法通常只能给出电池寿命的预测结果，而针对这一结果如何指导用户使用电池的行为，并未提供任何科学有效的建议。
发明内容
本申请提供一种电池寿命预测方法、系统、终端设备及计算机可读介质，至少解决相关技术中电池寿命预测结果不准确、效率低下以及无法提供科学指导以帮助用户正确使用电池、从而延长电池使用寿命等问题之一。
本申请提供一种电池寿命预测方法，包括：
获取电池的第一工况参数和第二工况参数；其中，第一工况参数包括影响电池寿命的多个指标按不同权重构成的指标组合；第二工况参数为表征电池寿命的指标；
基于混合神经网络模型，以第一工况参数的历史数据为输入、第二工况参数的历史数据为输出训练所述混合神经网络模型，直至模型收敛时，生成目标预测模型；
将电池的待预测数据输入至目标预测模型，生成预测结果；
基于预先划分的电池寿命等级，确定当前预测结果中的电池寿命等级，以匹配对应的使用建议。
本申请还提供了一种电池寿命预测系统，包括：
参数获取单元，设置为获取电池的第一工况参数和第二工况参数；其中，第一工况参数包括影响电池寿命的多个指标按不同权重构成的指标组合；第二工况参数为表征电池寿命的指标；
模型训练单元，设置为基于混合神经网络模型，以第一工况参数的历史数据为输入、第二工况参数的历史数据为输出训练所述混合神经网络模型，直至模型收敛时，生成目标预测模型；
寿命预测单元，设置为将电池的待预测数据输入至目标预测模型，生成预 测结果；
使用建议匹配单元，设置为基于预先划分的电池寿命等级，确定当前预测结果中的电池寿命等级，以匹配对应的使用建议。
本申请还提供了一种终端设备，包括：
至少一个处理器；
存储器，与所述处理器耦接，设置为存储至少一个程序；
当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器实现如上任一项所述的电池寿命预测方法。
本申请还提供了一种计算机可读介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现如上任一项所述的电池寿命预测方法。
附图说明
图1是本申请一实施例提供的电池寿命预测方法的流程示意图；
图2是图1中步骤S40的子步骤的流程示意图；
图3是本申请一实施例提供的电池寿命预测系统的结构示意图；
图4是本申请一实施例提供的终端设备的结构示意图。
具体实施方式
应当理解，文中所使用的步骤编号仅是为了方便描述，不对作为对步骤执行先后顺序的限定。
应当理解，在本申请说明书中所使用的术语仅仅描述特定实施例而并不意在限制本申请。如在本申请说明书和所附权利要求书中所使用的那样，除非上下文清楚地指明其它情况，否则单数形式的“一”、“一个”及“该”意在包括复数形式。
术语“包括”和“包含”指示所描述特征、整体、步骤、操作、元素和/或组件的存在，但并不排除至少一个其它特征、整体、步骤、操作、元素、组件 和/或其集合的存在或添加。
术语“和/或”是指相关联列出的项中的至少一个的任何组合以及所有可能组合，并且包括这些组合。
请参阅图1，本申请一实施例提供一种电池寿命预测方法。如图1所示，该电池寿命预测方法包括步骤S10至步骤S40：
S10、获取电池的第一工况参数和第二工况参数；其中，第一工况参数包括影响电池寿命的多个指标按不同权重构成的指标组合；第二工况参数为表征电池寿命的指标。
本步骤中，主要用于获取电池的第一工况参数和第二工况参数。其中第一工况参数主要包含多个跟电池寿命有关的影响因素(指标)，并且按照不用权重的加权组合方式来表征。第二工况参数则直接设置为表征电池寿命。
通常，与电池寿命有关的影响指标包含电池荷电状态(State Of Charge，SOC)、电池健康状态(State Of Health，SOH)、电池电能状态(State of Power，SOP)、电池总电压、电池总电流、电池单体电压、电池温度、电池温度最高值、电池温度最低值、电池告警级别、电池告警类别、累计充放电容量、累计充放电次数、额定电池容量、继电器状态、电池均衡类型和状态、用电设备使用状态及用电设备所处环境(温度、湿度)等。然而，在进行模型训练的过程中，若直接采用上述所有参数作为训练样本，第一会因为数据过多而造成计算量过大，大大增加了模型训练的难度，也不利于得到一个预测效果好的目标模型；并且这种方式不利于研究不同指标间的关联，每个指标作为孤立的元素来训练模型，维度多而发散，最终模型的学习效果也很难达到理想状态。因此，在实际应用中，通常会更关注于一个或者几个因素对于电池寿命的影响，考虑它们之间的关联以及它们各自和共同是如何影响电池寿命的。
在一个示例性的实施例中，步骤S10中的第一工况参数可选为环境温度、荷电状态、放电深度、电池总电压、电池总电流及累计充放电容量中至少两种指标按不同权重构成的指标组合。
可以理解的是，本实施例中的第一工况参数所对应的指标组合，可以根据实际需要从上述参数中任意选取至少两个进行组合，例如选取环境温度和荷电状态这两个指标，按照各自的权重加权得到指标组合，那么此时研究的就是环境温度和荷电状态(SOC)这两个因素对于电池寿命的影响；又或选取环境温度、荷电状态和放电深度这三个指标，按照各自的权重加权得到指标组合，再对模型进行训练，那么这种情况下，模型学习的则是这三个因素对于电池寿命的影响。
此外，本实施例中所指出的第一工况参数的指标类型包括上述环境温度、荷电状态、放电深度、电池总电压、电池总电流及累计充放电容量这些类别，但不限于这些工况参数。在实际应用中，还可以获取其他与电池寿命相关的影响因素，进行加权组合得到指标组合。因此，本实施例不对指标组合中的指标类别和数量进行任何限定。
在一个实施例中，确定所述第一工况参数中多个指标的权重，包括：
在获取电池的第一工况参数和第二工况参数之后，构建关于第一工况参数的指标组合层次模型，利用层次分析法确定指标组合层次模型中每个指标对应的权重。
本实施例中，为了确定第一工况参数不同指标的权重，对于多个指标进行分层，以构造指标组合层次模型，引入了层次分析法确定指标组合层次模型中每个指标对应的权重。其中，层次分析法的实现原理是根据问题的性质和要达到的总目标，将问题分解为不同的组成因素，并按照因素间的相互关联影响以及隶属关系，将因素按不同层次聚集组合，形成一个多层次的分析结构模型，从而最终使问题归结为最低层(供决策的方案、措施等)相对于最高层(总目标)的相对重要权值的确定或相对优劣次序的排定。本实施例中通过层次分析法，能够有效地确定出每个指标的权重，最终以得到作为模型输入的加权指标组合。
在一个示例性的实施例中，步骤S10中的第二工况参数包括电池剩余容量或电池内阻阻值，在实际应用中，可以选择电池剩余容量和电池内阻阻值其中 任意一种作为评估电池寿命的指标，或者将两种指标进行组合，以组合后的指标来评估电池寿命。
为了提高训练样本数据的质量，在一个实施例中，在获取了第一工况参数和第二工况参数对应的历史数据后，还可以对数据进行预处理，包括数据清洗、去除异常值和重复值、填补缺失值等。可选地，为了减少不同量纲数据对于训练过程的影响，还可以对数据进行归一化处理，利用同一量纲的数据进行模型训练。需要说明的是，此处对于数据归一化和数据清洗的顺序不作限定。
在一个可选地实施例中，可以将获取的第一工况参数和第二工况参数的历史数据经过预处理和归一化后得到的样本数据，按照预设比例分为训练集和测试集，例如按照8：2的比例进行分配，利用训练集训练模型，利用测试集检测模型的预测效果。同样地，8：2只是一种可选地分配比例，实际应用中可根据需要选择其他比例，本实施例不作任何限制。
S20、基于混合神经网络模型，以第一工况参数的历史数据为输入、第二工况参数的历史数据为输出训练所述混合神经网络模型，直至模型收敛时，生成目标预测模型。
针对电池寿命预测采用的算法通常多为单一的神经网络模型，例如LSTM模型或循环神经网络(Recurrent Neural Network，RNN)模型或卷积神经网络(Convolutional Neural Networks，CNN)模型，但这些模型各自的预测效果均不能达到较为理想的状态。其中一维卷积神经网络常用于处理时序问题，相比于人工神经网络(Artificial Neural Network，ANN)，通过采用卷积核参数的权值共享和层间系数连接，可以减少参数数量，避免过拟合，具有高级特征提取能力。而LSTM网络通过细胞内部自循环的记忆单元，不断更新网络的隐藏层状态，因此具有更好的时序处理能力和泛化能力。本实施例为了增强模型的预测效果，选用了混合神经网络模型，例如为LSTM-CNN混合模型。
在一个实施例中，所采用的LSTM-CNN混合模型可为一个由8层组成的序列模型。模型前两层由LSTM组成，每个LSTM具有32个神经元，使用的激 活函数为ReLU。然后是用于提取空间特征的CNN层。在LSTM层与CNN层的连接处需要改变LSTM输出维度，因为LSTM的输出具有3个维度(样本数，时间步长，输入)，而CNN则需要4维输入(样本数，1，时间步长，输入)。其中，第一个CNN层具有64个神经元，第二个CNN层有128个神经元。在第一CNN层和第二CNN层之间，可设置一个最大池化层来执行下采样操作；然后是全局平均池(Global Average Pooling，GAP)层，GAP层将多D特征映射转换为1-D特征向量，因为在此层中不需要参数，所以会减少全局模型参数；然后是批标准化(Batch Normalization，BN)层，该层有助于模型的收敛性；最后一层是模型的输出层，该输出层只是具有SoftMax分类器层的6个神经元的完全连接的层，该层表示当前类的概率。
基于上述的模型结构，在训练前先对初始网络参数进行设置，包括以下内容：
1)优化器可采用Adam，其优化效果较随机梯度下降法(Stochastic Gradient Descent，SGD)更好；
2)迭代次数可以默认为1000次，当训练集精确度和测试集精确度相差不大时，可以认为当前迭代次数合理，当训练集精确度和测试集精确度相差大时，可增大迭代次数继续训练。
3)批次大小即batch_size可默认为1，一般来说batch_size越小精确度越高；
4)使用小随机数初始化网络权重，以防产生不活跃的神经元，均匀分布的效果比较好；
5)设有dropout模块，防止模型的过拟合。其中dropout可选在0.4-0.5之间，具有较好的表现，位置只需要在最后一层softmax之前即可；
6)损失函数：选择回归损失，采用平均绝对误差(Mean Absolute Error，MAE)或均方差(Mean Square Error，MSE)。
需要说明的是，平均绝对误差(MAE)，衡量的是预测值与真实值之间距离的平均误差幅度，作用范围为0到正无穷。它的特点是收敛速度快，能够对 梯度给予合适的惩罚权重，而不是“一视同仁”，使梯度更新的方向可以更加精确。缺点是对异常值十分敏感，梯度更新的方向很容易受离群点所主导，不具备鲁棒性。均方差(MSE)衡量的是预测值与真实值之间距离的平方和，作用范围同为0到正无穷。特点是对离群点(Outliers)或者异常值更具有鲁棒性。缺点是在0点处的导数不连续，使得求解效率低下，导致收敛速度慢；而对于较小的损失值，其梯度也同其他区间损失值的梯度一样大，所以不利于网络的学习。可选择平均绝对误差(MAE)作为损失函数，并设定一个损失函数的预设阈值，当利用测试集对模型输出结果进行测试时，只要损失函数的大小满足预设阈值，即可认为模型预测精度达到要求，即生成对应的目标预测模型。
S30、将电池的待预测数据输入至目标预测模型，生成预测结果。
本步骤中，获取待预测的数据，可以为第一工况参数下的实时数据，进行电池寿命预测。
S40、基于预先划分的电池寿命等级，确定当前预测结果中的电池寿命等级，以匹配对应的使用建议。
请参阅图2，图2提供了步骤S40的子步骤的流程示意图。如图2所示，步骤S40又包括以下子步骤：
S401、当所述电池寿命等级为一级时，获取电池的第三工况参数的历史数据，若第三工况参数的历史数据处于额定工作范围之外，匹配对应的使用建议；
S402、当所述电池寿命等级为二级时，获取电池的第三工况参数的历史数据和第四工况参数的历史数据，若第三工况参数的历史数据或第四工况参数的历史数据中的任一数据处于额定工作范围之外，匹配对应的使用建议；
S403、当所述电池寿命等级为三级时，提醒更换电池。
上述步骤中，通常会预设三个电池寿命等级，分别为一级、二级或三级；例如，假设选择电池剩余容量为评估电池寿命的指标时，那么可以设定两个容量阈值来划分寿命等级，例如：
当电池剩余容量大于90％时，则将电池寿命等级定为一级；
当电池剩余容量在70％-90％之间时，则将电池寿命等级定为二级；
当电池剩余容量小于70％时，则将电池寿命等级定为三级。
由于不同的电池寿命等级能够体现电池的损耗情况，因此为了更有针对性的来指导正确使用电池，在本实施例中，可根据不同情况匹配不同使用建议，包括：
1)当电池寿命等级为一级时，说明电池的使用情况相对较好，此时可以获取第三工况参数的历史数据，即历史工况下的电池温度，进而统计历史工况下电池温度是否处于正常工作状态下的温度范围之外，如果存在电池温度处于正常工作状态下的温度范围之外的情况或者电池温度过高的情况超过一定次数，那么此时可以提醒用户注意在使用电池时及时检测电池温度，一旦温度超标，可以提醒用户暂停使用，待电池回归正常温度后，则可建议继续使用。
2)当电池寿命等级为二级时，说明电池的使用情况一般，那么此时除了获取历史工况下的电池温度，还可以获取第四工况参数的历史数据，包括但不限于充电时间、放电时间、充放电次数及充放电倍率。在获得这些参数后，判断电池历史的充电时间、放电时间、充放电次数及充放电倍率是否处于额定范围之外，如果是，则可以给用户提供合理建议，例如充电时间一次不超过5h，当电量剩余20％时即可充电，不要完全放电后再充电；充放电次数在一周内最好不超过5次，充放电的电流不宜过大，要和电池功率去匹配等等。
3)当电池寿命等级为三级时，说明电池使用情况较差，继续使用可能带来安全问题，此时则可以提醒用户直接更换电池。
综上，通过事先对电池寿命划分等级，在得到预测结果后确定当前电池寿命等级，按照不同的等级以匹配不同的使用建议，能够为用户正确使用电池提供科学指导，有利于延长电池寿命，提高了电池使用的安全性和稳定性。
请参阅图3，本申请一实施例还提供一种电池寿命预测系统，包括：
参数获取单元01，设置为获取电池的第一工况参数和第二工况参数；其中，第一工况参数包括影响电池寿命的多个指标按不同权重构成的指标组合；第二 工况参数为表征电池寿命的指标；
模型训练单元02，设置为基于混合神经网络模型，以第一工况参数的历史数据为输入、第二工况参数的历史数据为输出训练所述混合神经网络模型，直至模型收敛时，生成目标预测模型；
寿命预测单元03，设置为将电池的待预测数据输入至目标预测模型，生成预测结果；
使用建议匹配单元04，设置为基于预先划分的电池寿命等级，确定当前预测结果中的电池寿命等级，以匹配对应的使用建议。
可以理解的是，本实施例提供的电池寿命预测系统设置为执行如上述任意一项实施例所述的电池寿命预测方法，并实现与其相同的效果，此处不再赘述。
请参阅图4，本申请一实施例还提供一种终端设备，包括：
至少一个处理器；
存储器，与所述处理器耦接，设置为存储至少一个程序；
当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器实现如上所述的电池寿命预测方法。
处理器设置为控制该终端设备的整体操作，以完成上述的电池寿命预测方法的全部或部分步骤。存储器设置为存储各种类型的数据以支持在该终端设备的操作，这些数据例如可以包括用于在该终端设备上操作的任何应用程序或方法的指令，以及应用程序相关的数据。该存储器可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，例如静态随机存取存储器(Static Random Access Memory，SRAM)，电可擦除可编程只读存储器(Electrically Erasable Programmable Read-Only Memory，EEPROM)，可擦除可编程只读存储器(Erasable Programmable Read-Only Memory，EPROM)，可编程只读存储器(Programmable Read-Only Memory，PROM)，只读存储器(Read-Only Memory，ROM)，磁存储器，快闪存储器，磁盘或光盘。
在一示例性实施例中，终端设备可以被至少一个应用专用集成电路 (Application Specific 1ntegrated Circuit，AS1C)、数字信号处理器(Digital Signal Processor，DSP)、数字信号处理设备(Digital Signal Processing Device，DSPD)、可编程逻辑器件(Programmable Logic Device，PLD)、现场可编程门阵列(Field Programmable Gate Array，FPGA)、控制器、微控制器、微处理器或其他电子元件实现，设置为执行如上述任一项实施例所述的电池寿命预测方法，并达到如上述方法一致的技术效果。
在另一示例性实施例中，还提供一种包括计算机程序的计算机可读介质，该计算机程序被处理器执行时实现如上述任一项实施例所述的电池寿命预测方法的步骤。例如，该计算机可读介质可以为上述包括计算机程序的存储器，上述计算机程序可由终端设备的处理器执行以完成如上述任一项实施例所述的电池寿命预测方法，并达到如上述方法一致的技术效果。其中，本申请实施例的计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质的示例至少(非穷尽性列表)包括以下：具有至少一个布线的电连接部(电子装置)，便携式计算机盘盒(磁装置)，随机存取存储器(Random Access Memory，RAM)，只读存储器(Read-Only Memory，ROM)，可擦除可编辑只读存储器(EPROM)或闪速存储器，光纤装置，以及便携式只读存储器(Compact Disc Read-Only Memory，CDROM)。另外，计算机可读存储介质甚至可以是可在其上打印程序的纸或其他合适的介质，因为可以例如通过对纸或其他介质进行光学扫描，接着进行编辑、解译或必要时以其他合适方式进行处理来以电子方式获得程序，将其存储在计算机存储器中。
相对于相关技术，本申请的效果在于：
1)本申请基于LSTM-CNN混合神经网络模型进行训练，相比于单一的LSTM模型或者CNN卷积网络模型都具有更好的预测效果。
2)本申请考虑了不同指标对于电池寿命的影响程度，结合层次分析法确定了影响指标的权重，并以指标和权重的加权结果作为模型输入以训练模型，同 样提高了模型预测的精确度，更具有针对性，避免了利用大量孤立的指标训练增大了模型训练难度的情况，提升了模型训练的效率。
3)本申请通过事先对电池寿命划分了等级，在得到预测结果后确定当前电池寿命等级，按照不同的等级以匹配不同的使用建议，能够为用户正确使用电池提供科学指导，有利于延长电池寿命，提高了电池使用的安全性和稳定性。
本申请采用混合神经网络模型，以多个影响指标的权重组合为输入训练模型，提高了模型的预测精度，同时针对不同寿命情况提供不同的使用建议，从而科学指导用户使用电池，有利于延长电池使用寿命。

Claims (10)

  1. 一种电池寿命预测方法，包括：
    获取电池的第一工况参数和第二工况参数；其中，第一工况参数包括影响电池寿命的多个指标按不同权重构成的指标组合；第二工况参数为表征电池寿命的指标；
    基于混合神经网络模型，以第一工况参数的历史数据为输入、第二工况参数的历史数据为输出训练所述混合神经网络模型，直至模型收敛时，生成目标预测模型；
    将电池的待预测数据输入至目标预测模型，生成预测结果；
    基于预先划分的电池寿命等级，确定当前预测结果中的电池寿命等级，以匹配对应的使用建议。
  2. 根据权利要求1所述的电池寿命预测方法，其中，所述第一工况参数包括：环境温度、荷电状态、放电深度、电池总电压、电池总电流及累计充放电容量中至少两种指标按不同权重构成的指标组合。
  3. 根据权利要求1所述的电池寿命预测方法，其中，所述第二工况参数包括电池剩余容量或电池内阻阻值。
  4. 根据权利要求1所述的电池寿命预测方法，其中，确定所述第一工况参数中多个指标的权重，包括：
    在所述获取电池的第一工况参数和第二工况参数之后，构建关于第一工况参数的指标组合层次模型，利用层次分析法确定指标组合层次模型中每个指标对应的权重。
  5. 根据权利要求1所述的电池寿命预测方法，其中，所述混合神经网络模型为长短期记忆网络-卷积神经网络LSTM-CNN混合模型。
  6. 根据权利要求1所述的电池寿命预测方法，其中，所述基于预先划分的电池寿命等级，确定当前预测结果中的电池寿命等级，以匹配对应的使用建议，包括：
    当所述电池寿命等级为一级时，获取电池的第三工况参数的历史数据，响 应于第三工况参数的历史数据处于额定工作范围之外，匹配对应的使用建议；
    当所述电池寿命等级为二级时，获取电池的第三工况参数的历史数据和第四工况参数的历史数据，响应于第三工况参数的历史数据或第四工况参数的历史数据中的任一数据处于额定工作范围之外，匹配对应的使用建议；
    当所述电池寿命等级为三级时，提醒更换电池。
  7. 根据权利要求6所述的电池寿命预测方法，其中，所述第三工况参数包括电池温度；所述第四工况参数包括充电时间、放电时间、充放电次数及充放电倍率。
  8. 一种电池寿命预测系统，包括：
    参数获取单元(01)，设置为获取电池的第一工况参数和第二工况参数；其中，第一工况参数包括影响电池寿命的多个指标按不同权重构成的指标组合；第二工况参数为表征电池寿命的指标；
    模型训练单元(02)，设置为基于混合神经网络模型，以第一工况参数的历史数据为输入、第二工况参数的历史数据为输出训练所述混合神经网络模型，直至模型收敛时，生成目标预测模型；
    寿命预测单元(03)，设置为将电池的待预测数据输入至目标预测模型，生成预测结果；
    使用建议匹配单元(04)，设置为基于预先划分的电池寿命等级，确定当前预测结果中的电池寿命等级，以匹配对应的使用建议。
  9. 一种终端设备，包括：
    至少一个处理器；
    存储器，与所述处理器耦接，设置为存储至少一个程序；
    当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器实现如权利要求1-7任一项所述的电池寿命预测方法。
  10. 一种计算机可读介质，所述计算机可读介质上存储有计算机程序，其中，所述计算机程序被处理器执行时实现如权利要求1-7任一项所述的电池寿命 预测方法。
PCT/CN2023/082425 2022-08-31 2023-03-20 电池寿命预测方法、系统、终端设备及计算机可读介质 WO2024045567A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211051555.X 2022-08-31
CN202211051555.XA CN115587527A (zh) 2022-08-31 2022-08-31 电池寿命预测方法、系统、终端设备及计算机可读介质

Publications (1)

Publication Number Publication Date
WO2024045567A1 true WO2024045567A1 (zh) 2024-03-07

Family

ID=84771970

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/082425 WO2024045567A1 (zh) 2022-08-31 2023-03-20 电池寿命预测方法、系统、终端设备及计算机可读介质

Country Status (2)

Country Link
CN (1) CN115587527A (zh)
WO (1) WO2024045567A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117973238A (zh) * 2024-04-02 2024-05-03 深圳和润达科技有限公司 一种电芯化成流程充放电时间的智能预测方法及装置

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115587527A (zh) * 2022-08-31 2023-01-10 广东邦普循环科技有限公司 电池寿命预测方法、系统、终端设备及计算机可读介质
CN115983135B (zh) * 2023-01-28 2023-12-26 扬州浩海蓝生海洋装备有限公司 深海采油树油管悬挂器寿命评估方法及系统
CN117592383B (zh) * 2024-01-19 2024-03-26 四川晟蔚智能科技有限公司 一种设备健康寿命预测方法、系统、设备及介质
CN117609846B (zh) * 2024-01-23 2024-06-18 宁德时代新能源科技股份有限公司 电池性能参数的获取方法、装置、电子设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109991542A (zh) * 2019-03-27 2019-07-09 东北大学 基于wde优化lstm网络的锂离子电池剩余寿命预测方法
CN112381316A (zh) * 2020-11-26 2021-02-19 华侨大学 一种基于混合神经网络模型的机电设备健康状态预测方法
WO2022047204A1 (en) * 2020-08-30 2022-03-03 Hewlett-Packard Development Company, L.P. Battery life predictions using machine learning models
CN114330130A (zh) * 2021-12-30 2022-04-12 山东浪潮科学研究院有限公司 一种电池剩余使用寿命预测方法、设备及介质
CN115587527A (zh) * 2022-08-31 2023-01-10 广东邦普循环科技有限公司 电池寿命预测方法、系统、终端设备及计算机可读介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109991542A (zh) * 2019-03-27 2019-07-09 东北大学 基于wde优化lstm网络的锂离子电池剩余寿命预测方法
WO2022047204A1 (en) * 2020-08-30 2022-03-03 Hewlett-Packard Development Company, L.P. Battery life predictions using machine learning models
CN112381316A (zh) * 2020-11-26 2021-02-19 华侨大学 一种基于混合神经网络模型的机电设备健康状态预测方法
CN114330130A (zh) * 2021-12-30 2022-04-12 山东浪潮科学研究院有限公司 一种电池剩余使用寿命预测方法、设备及介质
CN115587527A (zh) * 2022-08-31 2023-01-10 广东邦普循环科技有限公司 电池寿命预测方法、系统、终端设备及计算机可读介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117973238A (zh) * 2024-04-02 2024-05-03 深圳和润达科技有限公司 一种电芯化成流程充放电时间的智能预测方法及装置
CN117973238B (zh) * 2024-04-02 2024-06-04 深圳和润达科技有限公司 一种电芯化成流程充放电时间的智能预测方法及装置

Also Published As

Publication number Publication date
CN115587527A (zh) 2023-01-10

Similar Documents

Publication Publication Date Title
WO2024045567A1 (zh) 电池寿命预测方法、系统、终端设备及计算机可读介质
CN109462231B (zh) 居民微电网的负荷优化调度方法、系统和存储介质
US10191116B2 (en) Battery test system for predicting battery test results
WO2022253038A1 (zh) 基于弹性网的锂电池健康状态预测方法、系统、设备及介质
KR20200140093A (ko) 배터리의 충방전 사이클에 따른 용량 변화 예측방법 및 예측시스템
CN106055775B (zh) 一种粒子滤波与机理模型相结合的二次电池寿命预测方法
CN111476435B (zh) 基于密度峰值的充电桩负荷预测方法
JP2022514992A (ja) バッテリ診断システム、バッテリ診断方法、及び記憶媒体
Feng et al. State of health estimation of large-cycle lithium-ion batteries based on error compensation of autoregressive model
CN108646189A (zh) 一种面向铅酸蓄电池剩余容量预测的libsvm建模方法
EP4055399A1 (en) Battery performance prediction
CN109613440A (zh) 电池的分级方法、装置、设备和存储介质
CN113219357A (zh) 电池包健康状态计算方法、系统及电子设备
CN113762591B (zh) 一种基于gru和多核svm对抗学习的短期电量预测方法及系统
Pu et al. A novel cuckoo search particle filtering strategy for the remaining useful life prediction of the lithium‐ion batteries in hybrid electric vehicle
Cailian Life prediction of battery based on random forest optimized by genetic algorithm
Kharlamova et al. Evaluating machine-learning-based methods for modeling a digital twin of battery systems providing frequency regulation
CN113033898A (zh) 基于k均值聚类与bi-lstm神经网络的电负荷预测方法及系统
KR102159817B1 (ko) 딥러닝을 통한 에너지저장장치 운영 방법, 이를 구현하기 위한 프로그램이 저장된 기록매체 및 이를 구현하기 위해 매체에 저장된 컴퓨터프로그램
Zhang et al. State-of-charge estimation for lead-acid battery using isolation forest algorithm and long short term memory network with attention mechanism
Zhang et al. Online state-of-health estimation for the lithium-ion battery based on an LSTM neural network with attention mechanism
Tian et al. Consistency evaluation of electric vehicle battery pack: multi-feature information fusion approach
CN116008815A (zh) 一种电芯内短路检测方法、装置、设备、存储介质及车辆
Zhang et al. SOH estimation of Li-ion battery based on FA-BPNN-K-means optimization algorithm
Yan et al. SOC prediction method of wireless sensor nodes batteries based on Attention-LSTM

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23858609

Country of ref document: EP

Kind code of ref document: A1