WO2020191801A1 - 基于灰狼群优化lstm网络的锂离子电池剩余寿命预测方法 - Google Patents

基于灰狼群优化lstm网络的锂离子电池剩余寿命预测方法 Download PDF

Info

Publication number
WO2020191801A1
WO2020191801A1 PCT/CN2019/081356 CN2019081356W WO2020191801A1 WO 2020191801 A1 WO2020191801 A1 WO 2020191801A1 CN 2019081356 W CN2019081356 W CN 2019081356W WO 2020191801 A1 WO2020191801 A1 WO 2020191801A1
Authority
WO
WIPO (PCT)
Prior art keywords
battery capacity
ion battery
lithium
numtrain
data
Prior art date
Application number
PCT/CN2019/081356
Other languages
English (en)
French (fr)
Inventor
张长胜
吴琼
Original Assignee
东北大学
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 东北大学 filed Critical 东北大学
Publication of WO2020191801A1 publication Critical patent/WO2020191801A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/004Artificial life, i.e. computing arrangements simulating life
    • G06N3/006Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02EREDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E60/00Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02E60/10Energy storage using batteries

Definitions

  • the invention relates to the technical field of lithium ion batteries, and in particular to a method for predicting the remaining life of a lithium ion battery based on a gray wolf pack optimization LSTM network.
  • the remaining life of the lithium-ion battery is used to describe the number of charge-discharge cycles when the capacity of the recycled lithium-ion battery reaches a certain threshold and cannot continue to work.
  • the experience-based method mainly uses the historical data of the battery to estimate its life. It can also be called the basic statistical law method. It mainly includes the cycle number method, the ampere-hour method and the weighted ampere-hour method, and the event-oriented aging accumulation method. method.
  • These three methods can only give a rough estimate of the remaining life of the lithium-ion battery. They are based on the statistics of the monitoring data of the lithium-ion battery. They can only be applied to special conditions. Although they have a faster calculation speed, However, it is unable to give an accurate description of the physical and chemical change process inside the battery, has poor adaptability, and cannot adapt to the prediction problem under complex conditions.
  • the performance-based prediction method has strong applicability. It can use various performance models in the process of battery life prediction, while taking into account the internal degradation process and external force factors of lithium-ion batteries Impact.
  • performance-based prediction methods mainly include model-based prediction methods, data-driven prediction methods, and three prediction methods based on fusion models.
  • Lithium-ion battery capacity data can effectively reflect the remaining life of the lithium-ion battery. As the number of charge and discharge increases, the capacity of the lithium-ion battery gradually decreases. When the actual battery capacity is less than 70% of the rated battery capacity, the lithium-ion battery is considered to be out of normal use. At this time, it is necessary to consider replacing the lithium-ion battery. How to use the early lithium-ion battery capacity data to realize the remaining life prediction of lithium-ion batteries, and to rationally plan the reserves of lithium-ion batteries in industrial production is of great significance to maximize the benefits of actual industrial production.
  • the Long Short-Term Memory improves on the deficiencies of the cyclic neural network. One is to add a forget gate, an input gate and an output gate inside the hidden layer, and the other is to add an information flow. To represent long-term memory, these two improvements make the long- and short-term memory network have better long- and short-term memory capabilities, and can better solve the time series prediction problem.
  • the technical problem to be solved by the present invention is to provide a method for predicting the remaining life of a lithium-ion battery based on the gray wolf pack optimization LSTM network in view of the above-mentioned shortcomings of the prior art, so as to realize direct prediction of the remaining life of the lithium-ion battery.
  • the technical solution adopted by the present invention is: a method for predicting the remaining life of a lithium ion battery based on the gray wolf pack optimization LSTM network, which includes the following steps:
  • Step 1 Obtain the monitoring data of the lithium-ion battery, extract the lithium-ion battery capacity data from it, divide the battery capacity data into a training data set, a verification data set, and a test data set, and normalize the battery capacity data deal with;
  • Step 2 Determine the structure of the long and short-term memory network, and construct a LSTM-based prediction model for the remaining life of the lithium-ion battery;
  • the lithium-ion battery remaining life prediction model includes an input layer, an LSTM layer, a fully connected layer, a Droupout layer, a fully connected layer, a regression layer, and an output layer; each neuron in the first fully connected layer and the previous LSTM layer Fully connected to play the role of feature fusion; the Droupout layer is added to the first fully connected layer to prevent over-fitting and improve generalization ability; the Droupout layer is used during each parameter training process. The probability p discards some neurons, and the remaining neurons are retained with a probability of 1-p; at the same time, a fully connected layer with 1 neuron and a regression layer are added to the Droupout layer to ensure that the output result is a continuous predicted value;
  • Step 3 Use the gray wolf pack algorithm to optimize the key parameters in the direct prediction model of the remaining life of the lithium-ion battery, and obtain the direct prediction model based on the gray wolf pack optimization LSTM network;
  • the key parameters in the direct prediction model of the remaining life of the lithium-ion battery include the training set length numTrain, the validation set length numValidation, and the structural parameters of the LSTM network as the criterion for dividing the lithium-ion battery capacity data.
  • the number of fully connected layer nodes is numfullyConnectedLayer, Droupout layer dropout probability pro_dropoutLayer, the maximum number of training times maxEpochs and initial learning rate initialLearnRate seven parameters;
  • Step 3.1 Parameter initialization: use the above seven parameters to be optimized as the gray wolf individual position vector X in the gray wolf pack optimization algorithm, initialize the initial population with the number of population individuals N, and calculate the initial population individuals by the fitness function calculation formula The corresponding fitness value;
  • the construction method of the fitness function is:
  • the battery capacity data corresponding to the first to numTrain charge and discharge processes before reaching the failure threshold is selected to train the LSTM network; the first before reaching the failure threshold is selected
  • the battery capacity data corresponding to the charge and discharge process from numTrain+1 to numTrain+numValidation is used to verify the predictive ability of the LSTM network; during the training process of the LSTM network, the battery capacity data of the previous numTrain-1 charge and discharge cycles are used as the LTSM network in turn
  • the battery capacity data of the next charge and discharge cycle of the current charge and discharge cycle is used as the output of the LSTM network; after the LSTM network training is completed, the battery capacity data of the numTrain charge and discharge cycle is used as the input of the LSTM network to predict the next The battery capacity data of the charge and discharge cycle; then, the battery capacity data of the next charge and discharge cycle is again used as the input of the LSTM network to predict the battery capacity data corresponding to
  • Fp sec ⁇ fp numTrain+1 , fp numTrain+2 ,..., fp numTrain+numValidation ⁇ .
  • Fit 1 represents the relationship between the predicted battery capacity data and the differentially processed battery capacity data, and length(Fp sec ) is the length of the battery capacity prediction data;
  • the original battery capacity data corresponding to the charge and discharge process from numTrain+1 to numTrain+numValidation is:
  • Step 3.2 Use the gray wolf pack optimization algorithm to update the parameters of the lithium-ion battery remaining life prediction model to ensure the minimum fitness value, thereby obtaining the optimized lithium-ion battery capacity data division criterion and LSTM network structure parameters;
  • the individual closest to the prey that is, the individual with the smallest fitness value
  • the other two individuals that are closer to the prey are called fitness.
  • the other two individuals with smaller values are called assistant wolves ⁇ and ⁇ , and the remaining wolves are represented as ⁇ ; in the hunting process, the three wolves that are closer to the prey, namely ⁇ , ⁇ and ⁇ wolves are used to guide the remaining wolves
  • the group individual ⁇ searches for the prey; during the search, the individual position update formula of the gray wolf group is as follows:
  • a and C represent coefficient factors
  • t represents the number of iterations
  • X p represents the position vector of the current prey
  • X represents the position vector of the gray wolf individual
  • r 1 and r 2 are random numbers in the range of [0, 1], and the coefficient a decreases linearly from 2 to 0 as the number of iterations increases;
  • X ⁇ , X ⁇ and X ⁇ represent the positions of the leader wolf ⁇ , assistant wolves ⁇ and ⁇ respectively
  • d ⁇ , d ⁇ and d ⁇ respectively represent the approximate distances of the current wolf pack tending to the prey position, calculated as follows
  • the formula determines the distance between the current wolf pack and the position of the prey as:
  • a 1 , A 2 and A 3 are the coefficient factors that control the advance or retreat of the gray wolf group, and X(t+1) is the position of the wolf group during t+1 iteration;
  • Step 4 Use the optimized data to determine the optimal direct prediction model for the remaining life of the lithium-ion battery
  • the lithium-ion data is divided into training data sets and test data sets, and the training set samples are used as the input of the long- and short-term memory network model; then the long- and short-term memory is trained through other parameters obtained by optimization Network, the long and short-term memory network model after training is the optimal network structure;
  • Step 5 Use the optimal lithium-ion battery remaining life direct prediction model to predict the later lithium-ion battery capacity data
  • the last charge-discharge cycle data in the training sample is used as the input of the LSTM network, and the output of the LSTM network is the predicted value of the lithium-ion battery capacity data in the next charge-discharge cycle; again, the predicted value of the lithium-ion battery capacity in the next charge-discharge cycle is used as the LSTM network
  • the output of the LSTM network is obtained as the predicted value of the lithium-ion battery capacity corresponding to the subsequent charge and discharge cycle; the cycle is repeated until the predicted value of the lithium-ion battery capacity reaches the rated failure threshold.
  • the method for predicting the remaining life of lithium-ion batteries based on the gray wolf pack optimization LSTM network determines the optimal direct prediction model for the remaining life of the lithium ion battery and synthesizes the gray wolf pack optimization algorithm
  • the fast convergence ability and the accurate time series prediction ability of the long and short-term memory network use the early lithium-ion battery capacity data to predict the later lithium-ion battery capacity data when it is close to the failure threshold.
  • the direct prediction model for the remaining life of the lithium-ion battery of the present invention predicts that the corresponding charge-discharge cycle when the battery capacity data reaches the failure threshold is earlier than the corresponding charge-discharge cycle when the real battery capacity data reaches the failure threshold, which can more accurately predict the lithium ion The remaining battery life.
  • FIG. 1 is a flowchart of a method for predicting the remaining life of a lithium-ion battery based on a gray wolf pack optimized LSTM network according to an embodiment of the present invention
  • FIG. 2 is a schematic diagram of the charging, discharging and impedance measurement operation process in the process of acquiring monitoring data of the B0005 lithium ion battery sample provided by an embodiment of the present invention
  • FIG. 3 is a schematic structural diagram of a LSTM-based lithium-ion battery remaining life prediction model provided by an embodiment of the present invention
  • FIG. 4 is a diagram of the prediction result of the B0005 lithium ion battery predicted by the LSTM-based lithium ion battery remaining life prediction model provided by an embodiment of the present invention
  • FIG. 5 is a prediction result diagram of a B0005 lithium ion battery based on the prediction model of the GWO optimized BP network when the length of the training set and the verification set are unchanged according to the embodiment of the present invention
  • FIG. 6 is a prediction result diagram of a B0005 lithium ion battery based on a prediction model of the GWO optimized BP network when the length of the training set and the verification set are changed according to an embodiment of the present invention.
  • This embodiment uses the degradation data of lithium-ion batteries from the NASA Prognostic Center of Excellence (PCoE), and selects the first group of lithium-ion battery sample battery capacity data labeled B0005 as specific implementation Data used in the case.
  • the remaining life of the lithium ion battery is predicted by using the method for predicting the remaining life of the lithium ion battery based on the gray wolf pack optimization LSTM network of the present invention.
  • the method for predicting the remaining life of a lithium-ion battery based on the gray wolf pack optimization LSTM network includes the following steps:
  • Step 1 Obtain the monitoring data of lithium-ion batteries, extract the lithium-ion battery capacity data from them, divide these battery capacity data into training data sets, verification data sets and test data sets, and normalize these battery capacities at the same time :
  • the test object of the lithium ion battery degradation test is a 18650 lithium cobalt oxide battery with a rated capacity of 2Ah.
  • 36 lithium-ion batteries were divided into 9 groups, each group containing 3 or 4 lithium-ion batteries, and the lithium-ion batteries were continuously charged, discharged, and impedance measured under different ambient temperature and discharge current conditions. Steps.
  • the monitoring data of the first group of lithium-ion battery samples labeled B0005 under three test conditions of charging, discharging and impedance testing are selected to provide experimental verification for the subsequent research of the embodiment to prove the remaining lithium-ion battery of the present invention Effectiveness of life prediction schemes.
  • Figure 2 shows the specific execution of the three steps of charging, discharging and impedance measurement in the process of acquiring B0005 lithium ion battery sample monitoring data.
  • the capacity of the lithium-ion battery gradually decreases.
  • the B0005 lithium-ion battery corresponds to The charge and discharge cycle is 129 times.
  • Step 2 Determine the long and short-term memory network structure, and construct the LSTM-based residual life prediction model of the lithium-ion battery as shown in Figure 3;
  • the lithium-ion battery residual life prediction model includes an input layer, an LSTM layer, a fully connected layer, and a Droupout layer , Fully connected layer, regression layer and output layer; each neuron in the first fully connected layer is fully connected with the previous LSTM layer, which plays a role of feature fusion; add the Droupout layer to the first fully connected layer Above, it plays a role in preventing over-fitting and improving generalization ability; in each parameter training process, the Droupout layer discards some neurons with probability p, and the remaining neurons are retained with a probability of 1-p; Add a fully connected layer with 1 neuron and a regression layer on the Droupout layer to ensure that the output result is a continuous predicted value;
  • Step 3 Use the gray wolf pack algorithm to optimize the key parameters in the direct prediction model of the remaining life of the lithium-ion battery, and obtain the direct prediction model based on the gray wolf pack optimization LSTM network;
  • the key parameters in the direct prediction model of the remaining life of the lithium-ion battery include the training set length numTrain, the validation set length numValidation, and the structural parameters of the LSTM network as the criterion for dividing the lithium-ion battery capacity data.
  • the number of fully connected layer nodes is numfullyConnectedLayer, Droupout layer dropout probability pro_dropoutLayer, the maximum number of training times maxEpochs and initial learning rate initialLearnRate seven parameters;
  • Step 3.1 Parameter initialization: use the above seven parameters to be optimized as the gray wolf individual position vector X in the gray wolf pack optimization algorithm, initialize the initial population with the number of population individuals N, and calculate the initial population individuals by the fitness function calculation formula The corresponding fitness value;
  • the construction method of the fitness function is:
  • the battery capacity data corresponding to the first to numTrain charge and discharge processes before reaching the failure threshold is selected to train the LSTM network; the first before reaching the failure threshold is selected
  • the battery capacity data corresponding to the charge and discharge process from numTrain+1 to numTrain+numValidation is used to verify the predictive ability of the LSTM network; during the training process of the LSTM network, the battery capacity data of the previous numTrain-1 charge and discharge cycles are used as the LTSM network in turn
  • the battery capacity data of the next charge and discharge cycle of the current charge and discharge cycle is used as the output of the LSTM network; after the LSTM network training is completed, the battery capacity data of the numTrain charge and discharge cycle is used as the input of the LSTM network to predict the next The battery capacity data of the charge and discharge cycle; then, the battery capacity data of the next charge and discharge cycle is again used as the input of the LSTM network to predict the battery capacity data corresponding to
  • Fp sec ⁇ fp numTrain +1, fp numTrain+2 ,..., fp numTrain+numValidation ⁇ .
  • Fit 1 represents the relationship between the predicted battery capacity data and the differentially processed battery capacity data, and length(Fp sec ) is the length of the battery capacity prediction data;
  • the original battery capacity data corresponding to the charge and discharge process from numTrain+1 to numTrain+numValidation is:
  • Step 3.2 Use the gray wolf pack optimization algorithm to update the parameters of the lithium-ion battery remaining life prediction model to ensure the minimum fitness value, thereby obtaining the optimized lithium-ion battery capacity data division criterion and LSTM network structure parameters;
  • the individual closest to the prey that is, the individual with the smallest fitness value
  • the other two individuals that are closer to the prey are called fitness.
  • the other two individuals with smaller values are called assistant wolves ⁇ and ⁇ , and the remaining wolves are represented as ⁇ ; in the hunting process, the three wolves that are closer to the prey, namely ⁇ , ⁇ and ⁇ wolves are used to guide the remaining wolves
  • the group individual ⁇ searches for the prey; during the search, the individual position update formula of the gray wolf group is as follows:
  • a and C represent coefficient factors
  • t represents the number of iterations
  • X p represents the position vector of the current prey
  • X represents the position vector of the gray wolf individual
  • r 1 and r 2 are random numbers in the range of [0, 1], and the coefficient a decreases linearly from 2 to 0 as the number of iterations increases;
  • X ⁇ , X ⁇ and X ⁇ represent the positions of the leader wolf ⁇ , assistant wolves ⁇ and ⁇ respectively
  • d ⁇ , d ⁇ and d ⁇ respectively represent the approximate distances of the current wolf pack tending to the prey position, calculated as follows
  • the formula determines the distance between the current wolf pack and the prey position as:
  • a 1 , A 2 and A 3 are the coefficient factors that control the advance or retreat of the gray wolf group, and X(t+1) is the position of the wolf group during t+1 iteration;
  • Step 4 Use the optimized data to determine the optimal direct prediction model for the remaining life of the lithium-ion battery
  • the lithium-ion data is divided into training data sets and test data sets, and the training set samples are used as the input of the long- and short-term memory network model; then the long- and short-term memory is trained through other parameters obtained by optimization Network, the long and short-term memory network model after training is the optimal network structure;
  • Step 5 Use the optimal lithium-ion battery remaining life direct prediction model to predict the later lithium-ion battery capacity data
  • the last charge-discharge cycle data in the training sample is used as the input of the LSTM network, and the output of the LSTM network is the predicted value of the lithium-ion battery capacity data in the next charge-discharge cycle; again, the predicted value of the lithium-ion battery capacity in the next charge-discharge cycle is used as the LSTM network
  • the output of the LSTM network is obtained as the predicted value of the lithium-ion battery capacity corresponding to the subsequent charge and discharge cycle; the cycle is repeated until the predicted value of the lithium-ion battery capacity reaches the rated failure threshold.
  • the gray wolf pack algorithm is used to optimize the key parameters in the direct prediction model of the remaining life of the lithium ion battery to obtain the optimization results as shown in Table 1.
  • the optimization results of the model parameters listed in Table 1 are brought into the LSTM network, and the lithium-ion battery capacity prediction results predicted by the direct prediction model for the remaining life of the lithium-ion battery of the present invention are shown in Figure 4, which can be seen from the figure.
  • the lithium-ion battery capacity change trend predicted by the direct prediction model for the remaining life of the lithium-ion battery of the present invention is closer to the actual lithium-ion battery capacity change trend, the predicted battery capacity curve is relatively flat, and the predicted battery capacity data reaches the failure threshold.
  • the comparison result of the charge-discharge cycle and the actual battery capacity data reaches the failure threshold, the corresponding charge-discharge cycle is shown in Table 2. It can be seen that the predicted result is closer to the actual result.
  • the lithium-ion battery remaining life direct prediction model of the present invention It can more accurately reflect the change trend of the lithium-ion battery capacity data, and can effectively reflect the remaining life of the lithium-ion battery.
  • Lithium ion battery sample B0005 Predict the charge and discharge cycle when the data reaches the failure threshold (cycle) 125
  • this embodiment also provides the use of more classic genetic algorithm (GA) and particle swarm optimization (PSO) Replace the gray wolf pack algorithm, optimize the key parameters in the direct prediction model of the lithium-ion battery remaining life of the present invention, and compare the optimal fitness of the above two methods when reaching the maximum number of iterations of the direct prediction model of the lithium-ion battery remaining life of the present invention Value, the fitness value when reaching the final stable state and the number of iterations required to reach the final stable state (considering the time required for the LSTM network training process and the optimization algorithm optimization effect, the maximum iteration steps of the above two optimization algorithms are set to 100 times, The population size is consistent with the population size in the model mentioned in this chapter).
  • GA genetic algorithm
  • PSO particle swarm optimization
  • this embodiment also provides two comparative examples.
  • a shallow BP network with three hidden layers was selected as the predictor to predict the change of lithium-ion battery capacity data.
  • the gray wolf pack algorithm is used to optimize the above parameters in the BP network to realize the adaptive selection of parameters.
  • the training set and the validation set are consistent with the division criterion of the training set and the verification set in the direct prediction model for the remaining life of the lithium ion battery of the present invention; in the second comparison example, the length of the training set and the verification set are also used as parameters to be optimized.
  • the results of the two comparative examples are summarized below.
  • the optimization results of the gray wolf pack algorithm to optimize the BP network are shown in Table 4. It can be seen from the table that the final fitness value obtained by using the gray wolf pack to optimize the BP network prediction model is larger than the fitness value of the lithium-ion battery remaining life direct prediction model of the present invention, and the BP network has a higher prediction effect on verifying the battery capacity data. Relatively poor.
  • the BP network after parameter optimization is used to predict the change law of the battery capacity corresponding to the test set data, and the result is shown in Figure 5. It can be seen from the figure that although the time for the BP network to predict the battery capacity data to reach the failure threshold is always shorter than the time for the real lithium-ion battery capacity data to reach the failure threshold, the battery capacity data predicted by the BP network can be used to determine the remaining life of the lithium-ion battery. A more conservative assessment is made. However, because the BP network is a shallow network, it has a poor ability to predict time series, and the difference between the predicted battery capacity data and the real battery capacity data is relatively large, and its prediction effect is relative to that of the lithium battery of the present invention. The prediction effect of the direct prediction model of the remaining life of the ion battery still needs to be improved.
  • the optimization results of the gray wolf pack optimization BP network are shown in Table 5. It can be seen from the table that the final fitness value obtained by using the gray wolf pack to optimize the BP network prediction model is still larger than the fitness value of the lithium-ion battery remaining life direct prediction model of the present invention.
  • the BP network can verify the prediction effect of the battery capacity data There is still a lot of room for improvement.
  • the optimized training set and test set division criteria are substituted into the BP network to predict the change law of the lithium-ion battery capacity corresponding to the test set data. The result is shown in Figure 6.
  • the direct prediction model of the remaining life of the lithium ion battery of the present invention selects the LSTM network as the predictor to be effective in the prediction of the lithium ion battery capacity data.

Landscapes

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

Abstract

一种基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,涉及锂离子电池技术领域。该方法首先获取锂离子电池的监测数据,并从中提取出锂离子电池容量数据;确定长短期记忆网结构,构造基于LSTM的锂离子电池剩余寿命预测模型;然后利用灰狼群算法优化锂离子电池剩余寿命直接预测模型中的关键参数,得到基于灰狼群优化LSTM网络的直接预测模型;利用优化数据确定最优的锂离子电池剩余寿命直接预测模型;最后利用最优的锂离子电池剩余寿命直接预测模型预测后期锂离子电池容量数据。基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,能够较为准确的预测锂离子电池剩余寿命。

Description

基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法 技术领域
本发明涉及锂离子电池技术领域,尤其涉及一种基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法。
背景技术
锂离子电池剩余寿命是用来描述当循环使用的锂离子电池容量达到确定阈值不能继续工作时所对应的充放电循环周期次数。目前,锂离子电池寿命的预测方法大致分可以为两类:基于经验的预测方法和基于性能的预测方法。基于经验的方法主要是利用电池历史数据对其寿命进行估计,也可称为基本统计规律法,主要包括循环周期数法、安时法与加权安时法和面向事件的老化累积方法等三种方法。这三种方法只能对锂离子电池剩余寿命给出粗略估计,它们是在对锂离子电池监测数据统计的基础上进行的,只能适用于特殊的条件场合,虽然具有较快的计算速度,但是无法对电池内部的物理和化学的变化过程给出精确的描述,具有较差的适应性,无法适应复杂条件下的预测问题。
针对基于经验的预测方法的不足,基于性能的预测方法具有较强的适用性,它在电池寿命预测的过程中可以使用各种不同的性能模型,同时考虑锂离子电池内部的衰退过程和外力因素的影响。目前,基于性能的预测方法主要包括基于模型的预测方法、基于数据驱动的预测方法和基于融合模型的三种预测方法。
锂离子电池容量数据能够有效反映锂离子电池的剩余寿命情况。随着充放电次数的增加,锂离子电池容量逐渐减小,当实际电池容量小于额定电池容量的70%时,认为锂离子电池已无法正常使用,此时需考虑更换锂离子电池。如何利用早期锂离子电池容量数据,实现锂离子电池的剩余寿命预测,合理规划工业生产中的锂离子电池储量,对满足实际工业生产效益最大化具有重要意思。
长短期记忆网络(Long Short-Term Memory,即LSTM)针对循环神经网络的缺陷进行改进,一是在隐含层的内部添加了遗忘门、输入门和输出门,二是增加一条信息流,用来代表长期记忆,这两项改进使长短期记忆网络具有较好的长短期记忆能力,能够更好的解决时间序列预测问题。
发明内容
本发明要解决的技术问题是针对上述现有技术的不足,提供一种基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,实现对锂离子电池剩余寿命的直接预测。
为解决上述技术问题,本发明所采取的技术方案是:基于灰狼群优化LSTM网络的锂离 子电池剩余寿命预测方法,包括以下步骤:
步骤1、获取锂离子电池的监测数据,并从中提取出锂离子电池容量数据,将这些电池容量数据划分成训练数据集、验证数据集和测试数据集,同时对这些电池容量数据进行归一化处理;
步骤2、确定长短期记忆网结构,构造基于LSTM的锂离子电池剩余寿命预测模型;
所述锂离子电池剩余寿命预测模型包括输入层、LSTM层、全连接层、Droupout层、全连接层、回归层以及输出层;第一层全连接层中每个神经元与其前一层LSTM层进行全连接,起到特征融合的作用;将Droupout层添加到第一层全连接层之上,起到防止过拟合和提高泛化能力的作用;Droupout层在每次参数训练过程中,以概率p舍弃部分神经元,剩余神经元以1-p的概率予以率保留;同时在Droupout层上添加神经元个数为1的全连接层以及回归层,确保输出结果为连续的预测值;
步骤3:利用灰狼群算法优化锂离子电池剩余寿命直接预测模型中的关键参数,得到基于灰狼群优化LSTM网络的直接预测模型;
所述锂离子电池剩余寿命直接预测模型中的关键参数包括作为锂离子电池容量数据划分准则的训练集长度numTrain、验证集长度numValidation以及LSTM网络的结构参数LSTM网络隐含层神经元节点数numHiddenUnits、全连接层节点数numfullyConnectedLayer、Droupout层舍弃概率pro_dropoutLayer、训练过程最大训练次数maxEpochs和初始学习率initialLearnRate七个参数;
步骤3.1:参数初始化:将上述七个待优化参数作为灰狼群优化算法中灰狼个体的位置向量X,初始化产生种群个体数为N的初始化种群,并通过适应度函数计算公式计算初始种群个体对应的适应度值;
所述适应度函数的构造方法为:
(1)将锂离子电池容量数据进行一次差分处理,将其转化成LSTM网络训练过程所需平稳时间序列;
假设原始电池容量数据为F={f 1,f 2,…,f S},其中,S表示锂离子电池总的充放电周期次数,对其进行一次差分处理后,得到的时间序列如下公式所示:
Figure PCTCN2019081356-appb-000001
(2)在基于灰狼优化的LSTM网络直接预测模型中,选取到达失效阈值之前的第1次至第numTrain次充放电过程对应的电池容量数据用于训练LSTM网络;选取到达失效阈值之 前的第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量数据用于验证LSTM网络的预测能力;LSTM网络训练过程中,将前numTrain-1次充放电周期的电池容量数据依次作为LTSM网络的输入,将当前充放电周期的下一充放电周期的电池容量数据作为LSTM网络的输出;LSTM网络训练结束后,以第numTrain次充放电周期的电池容量数据作为LSTM网络的输入,预测下一充放电周期的电池容量数据;然后,以下一充放电周期的电池容量数据再次作为LSTM网络的输入,预测后续充放电周期对应的电池容量数据,上述过程不断重复直至预测充放电周期次数到达numValidation;
假设第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量预测数据为:
Fp sec={fp numTrain+1,fp numTrain+2,…,fp numTrain+numValidation}.   (2)
与之相对应的,经差分处理后第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量真实数据为:
Fr sec={fr numTrain+1,fr numTrain+2,…,fr numTrain+numValidation}.   (3)
构造如下函数表示预测后的电池容量数据与差分处理后的电池容量数据之间的关系:
Figure PCTCN2019081356-appb-000002
其中,Fit 1表示预测后的电池容量数据与差分处理后的电池容量数据之间的关系,length(Fp sec)为电池容量预测数据的长度;
(3)对式2所示锂离子电池容量预测数据进行逆差分化处理后,将预测得到的第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量数据恢复至原始电池容量数据区间,得到恢复后的电池容量数据
Figure PCTCN2019081356-appb-000003
其表达式为:
Figure PCTCN2019081356-appb-000004
其中,
Figure PCTCN2019081356-appb-000005
与之相对应的,第numTrain+1次至第numTrain+numValidation次充放电过程对应的原始电池容量数据为:
F sec={f numTrain+1,f numTrain+2,…,f numTrain+numValidation}.   (7)
构造如下函数表示恢复至原始电池容量区间的电池容量数据与原始电池容量数据之间的关系:
Figure PCTCN2019081356-appb-000006
(4)通过上述两个表述预测数据与真实数据间相近程度的关系式得到基于灰狼群优化的LSTM网络直接预测模型的适应度函数Fit direct,如下公式所示:
Fit direct=Fit 1+Fit 2.   (9)
步骤3.2:采用灰狼群优化算法对锂离子电池剩余寿命预测模型的参数进行更新,确保适应度值最小,由此得到优化后的锂离子电池容量数据划分准则以及LSTM网络结构参数;
在灰狼群算法中,为了模拟灰狼群的社会行为,将距离猎物最近的个体,即适应度值最小的个体称为首领狼α,将距离猎物较近的其他两个个体,即适应度值较小的其他两个个体称为助理狼β和δ,剩余狼群个体表示为ω;捕猎过程中,利用距离猎物较近的三只狼,即α狼、β狼和δ狼引导剩余狼群个体ω对猎物进行搜索;搜索过程中,灰狼群个体位置更新公式如下公式所示:
X(t+1)=X p(t)-A·d,   (10a)
d=|C·X p(t)-X(t)|,   (10b)
其中,A和C表示系数因子,t表示迭代次数,X p表示当前猎物的位置向量,X表示灰狼个体的位置向量;系数因子A和C的计算公式如下所示:
A=2a·r 1-a,   (11a)
C=2·r 2,   (11b)
其中,r 1和r 2是[0,1]范围内的随机数,系数a随着迭代次数的增加从2到0线性递减;
在搜索猎物过程中,由于首领狼α、助理狼β和δ距离猎物较近,剩余狼群个体ω的位置 根据处于领导阶层的首领狼α、助理狼β和δ的位置进行更新,其表示式为:
d α=|C·X α-X|,   (12a)
d β=|C·X β-X|,   (12b)
d δ=|C·X δ-X|,   (12c)
其中,X α、X β和X δ分别表示首领狼α、助理狼β和δ所处的位置,d α、d β和d δ分别表示当前狼群趋向于猎物位置的近似距离,通过如下计算公式确定当前狼群与猎物位置间的距离为:
Figure PCTCN2019081356-appb-000007
X 1=X α-A 1·d α,   (13b)
X 2=X β-A 2·d β,   (13c)
X 3=X δ-A 3·d δ,   (13d)
其中,A 1、A 2和A 3是控制灰狼群个体前进或后退的系数因子,X(t+1)为狼群t+1次迭代时所处位置;
步骤4:利用优化数据确定最优的锂离子电池剩余寿命直接预测模型;
根据优化得到的锂离子电池数据划分准则,将锂离子数据分成训练数据集和测试数据集,并将训练集样本作为长短期记忆网络模型的输入;然后再通过优化得到的其他参数训练长短期记忆网络,训练后的长短期记忆网络模型为最优网络结构;
步骤5:利用最优的锂离子电池剩余寿命直接预测模型预测后期锂离子电池容量数据;
将训练样本中最后充放电周期数据作为LSTM网络的输入,LSTM网络的输出为下一充放电周期锂离子电池容量数据的预测值;再次将下一充放电周期锂离子电池容量预测值作为LSTM网络的输入,得到LSTM网络的输出作为后续充放电周期对应的锂离子电池容量预测值;依次循环,直至锂离子电池容量预测值到达额定失效阈值。
采用上述技术方案所产生的有益效果在于:本发明提供的基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,确定最优的锂离子电池剩余寿命直接预测模型综合了灰狼群优化算法的快速收敛能力和长短期记忆网络的准确时间序列预测能力,利用早期锂离子电池容量数据预测后期接近失效阈值时的锂离子电池容量数据。利用本发明的锂离子电池剩余寿命直接预测模型预测电池容量数据达到失效阈值时对应的充放电周期均比真实电池容量数据 达到失效阈值时对应的充放电周期早一些,能够较为准确的预测锂离子电池剩余寿命。
附图说明
图1为本发明实施例提供的基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法的流程图;
图2为本发明实施例提供的B0005锂离子电池样本监测数据获取过程中的充电、放电和阻抗测量操作过程示意图;
图3为本发明实施例提供的基于LSTM的锂离子电池剩余寿命预测模型的结构示意图;
图4为本发明实施例提供的基于LSTM的锂离子电池剩余寿命预测模型对B0005锂离子电池进行预测的预测结果图;
图5为本发明实施例提供的训练集和验证集长度不变时基于GWO优化BP网络的预测模型对B0005锂离子电池进行预测的预测结果图;
图6为本发明实施例提供的训练集和验证集长度改变时基于GWO优化BP网络的预测模型对B0005锂离子电池进行预测的预测结果图。
具体实施方式
下面结合附图和实施例,对本发明的具体实施方式作进一步详细描述。以下实施例用于说明本发明,但不用来限制本发明的范围。
本实施例以来源于美国国家航空航天局卓越故障预测研究中心(NASA Prognostic Center of Excellence,PCoE)锂离子电池退化数据,选取其中第一组标号为B0005的锂离子电池样本电池容量数据作为具体实施案例中所用数据。使用本发明的基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法对该锂离子电池的剩余寿命进行预测。
基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,如图1所示,包括以下步骤:
步骤1、获取锂离子电池的监测数据,并从中提取出锂离子电池容量数据,将这些电池容量数据划分成训练数据集、验证数据集和测试数据集,同时对这些电池容量进行归一化处理:
本实施例中,锂离子电池退化试验的试验对象为18650型钴酸锂离子电池,其额定容量为2Ah。试验过程中,将36块锂离子电池分为9组,每组包含3或4块锂离子电池,分别在不同环境温度、放电电流条件下对锂离子电池不断执行充电、放电和阻抗测量等三个步骤。
本实施例选用第1组标号为B0005的锂离子电池样本在充电、放电和阻抗测试等三种测试条件下的监测数据为实施例后续研究提供试验验证,以证明本发明所提锂离子电池剩余寿命预测方案的有效性。B0005锂离子电池样本监测数据获取过程中的充电、放电和阻抗测量 三个步骤的具体执行操作如图2所示。
随着充放电次数的增加,锂离子电池容量逐渐减小,当电池容量下降至失效阈值U=1.38Ah时,则认为锂离子电池失效,无法继续正常使用,此时B0005锂离子电池样对应的充放电周期为129次。
步骤2、确定长短期记忆网结构,构造如图3所示的基于LSTM的锂离子电池剩余寿命预测模型;所述锂离子电池剩余寿命预测模型包括输入层、LSTM层、全连接层、Droupout层、全连接层、回归层以及输出层;第一层全连接层中每个神经元与其前一层LSTM层进行全连接,起到特征融合的作用;将Droupout层添加到第一层全连接层之上,起到防止过拟合和提高泛化能力的作用;Droupout层在每次参数训练过程中,以概率p舍弃部分神经元,剩余神经元以1-p的概率予以率保留;同时在Droupout层上添加神经元个数为1的全连接层以及回归层,确保输出结果为连续的预测值;
步骤3:利用灰狼群算法优化锂离子电池剩余寿命直接预测模型中的关键参数,得到基于灰狼群优化LSTM网络的直接预测模型;
所述锂离子电池剩余寿命直接预测模型中的关键参数包括作为锂离子电池容量数据划分准则的训练集长度numTrain、验证集长度numValidation以及LSTM网络的结构参数LSTM网络隐含层神经元节点数numHiddenUnits、全连接层节点数numfullyConnectedLayer、Droupout层舍弃概率pro_dropoutLayer、训练过程最大训练次数maxEpochs和初始学习率initialLearnRate七个参数;
步骤3.1:参数初始化:将上述七个待优化参数作为灰狼群优化算法中灰狼个体的位置向量X,初始化产生种群个体数为N的初始化种群,并通过适应度函数计算公式计算初始种群个体对应的适应度值;
所述适应度函数的构造方法为:
(1)将锂离子电池容量数据进行一次差分处理,将其转化成LSTM网络训练过程所需平稳时间序列;
假设原始电池容量数据为F={f 1,f 2,…,f S},其中,S表示锂离子电池总的充放电周期次数,对其进行一次差分处理后,得到的时间序列如下公式所示:
Figure PCTCN2019081356-appb-000008
(2)在基于灰狼优化的LSTM网络直接预测模型中,选取到达失效阈值之前的第1次至第numTrain次充放电过程对应的电池容量数据用于训练LSTM网络;选取到达失效阈值之 前的第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量数据用于验证LSTM网络的预测能力;LSTM网络训练过程中,将前numTrain-1次充放电周期的电池容量数据依次作为LTSM网络的输入,将当前充放电周期的下一充放电周期的电池容量数据作为LSTM网络的输出;LSTM网络训练结束后,以第numTrain次充放电周期的电池容量数据作为LSTM网络的输入,预测下一充放电周期的电池容量数据;然后,以下一充放电周期的电池容量数据再次作为LSTM网络的输入,预测后续充放电周期对应的电池容量数据,上述过程不断重复直至预测充放电周期次数到达numValidation;
假设第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量预测数据为:
Fp sec={fp numTrain+1,fp numTrain+2,…,fp numTrain+numValidation}.   (2)
与之相对应的,经差分处理后第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量真实数据为:
Fr sec={fr numTrain+1,fr numTrain+2,…,fr numTrain+numValidation}.   (3)
构造如下函数表示预测后的电池容量数据与差分处理后的电池容量数据之间的关系:
Figure PCTCN2019081356-appb-000009
其中,Fit 1表示预测后的电池容量数据与差分处理后的电池容量数据之间的关系,length(Fp sec)为电池容量预测数据的长度;
(3)对式2所示锂离子电池容量预测数据进行逆差分化处理后,将预测得到的第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量数据恢复至原始电池容量数据区间,得到恢复后的电池容量数据
Figure PCTCN2019081356-appb-000010
其表达式为:
Figure PCTCN2019081356-appb-000011
其中,
Figure PCTCN2019081356-appb-000012
与之相对应的,第numTrain+1次至第numTrain+numValidation次充放电过程对应的原始电池容量数据为:
F sec={f numTrain+1,f numTrain+2,…,f numTrain+numValidation}.   (7)
构造如下函数表示恢复至原始电池容量区间的电池容量数据与原始电池容量数据之间的关系:
Figure PCTCN2019081356-appb-000013
(4)通过上述两个表述预测数据与真实数据间相近程度的关系式得到基于灰狼群优化的LSTM网络直接预测模型的适应度函数Fit direct,如下公式所示:
Fit direct=Fit 1+Fit 2.   (9)
步骤3.2:采用灰狼群优化算法对锂离子电池剩余寿命预测模型的参数进行更新,确保适应度值最小,由此得到优化后的锂离子电池容量数据划分准则以及LSTM网络结构参数;
在灰狼群算法中,为了模拟灰狼群的社会行为,将距离猎物最近的个体,即适应度值最小的个体称为首领狼α,将距离猎物较近的其他两个个体,即适应度值较小的其他两个个体称为助理狼β和δ,剩余狼群个体表示为ω;捕猎过程中,利用距离猎物较近的三只狼,即α狼、β狼和δ狼引导剩余狼群个体ω对猎物进行搜索;搜索过程中,灰狼群个体位置更新公式如下公式所示:
X(t+1)=X p(t)-A·d,   (10a)
d=|C·X p(t)-X(t)|,   (10b)
其中,A和C表示系数因子,t表示迭代次数,X p表示当前猎物的位置向量,X表示灰狼个体的位置向量;系数因子A和C的计算公式如下所示:
A=2a·r 1-a,   (11a)
C=2·r 2,   (11b)
其中,r 1和r 2是[0,1]范围内的随机数,系数a随着迭代次数的增加从2到0线性递减;
在搜索猎物过程中,由于首领狼α、助理狼β和δ距离猎物较近,剩余狼群个体ω的位置 根据处于领导阶层的首领狼α、助理狼β和δ的位置进行更新,其表示式为:
d α=|C·X α-X|,   (12a)
d β=|C·X β-X|,   (12b)
d δ=|C·X δ-X|,   (12c)
其中,X α、X β和X δ分别表示首领狼α、助理狼β和δ所处的位置,d α、d β和d δ分别表示当前狼群趋向于猎物位置的近似距离,通过如下计算公式确定当前狼群与猎物位置间的距离为:
Figure PCTCN2019081356-appb-000014
X 1=X α-A 1·d α,   (13b)
X 2=X β-A 2·d β,   (13c)
X 3=X δ-A 3·d δ,  (13d)
其中,A 1、A 2和A 3是控制灰狼群个体前进或后退的系数因子,X(t+1)为狼群t+1次迭代时所处位置;
步骤4:利用优化数据确定最优的锂离子电池剩余寿命直接预测模型;
根据优化得到的锂离子电池数据划分准则,将锂离子数据分成训练数据集和测试数据集,并将训练集样本作为长短期记忆网络模型的输入;然后再通过优化得到的其他参数训练长短期记忆网络,训练后的长短期记忆网络模型为最优网络结构;
步骤5:利用最优的锂离子电池剩余寿命直接预测模型预测后期锂离子电池容量数据;
将训练样本中最后充放电周期数据作为LSTM网络的输入,LSTM网络的输出为下一充放电周期锂离子电池容量数据的预测值;再次将下一充放电周期锂离子电池容量预测值作为LSTM网络的输入,得到LSTM网络的输出作为后续充放电周期对应的锂离子电池容量预测值;依次循环,直至锂离子电池容量预测值到达额定失效阈值。
本实施例中,利用灰狼群算法对锂离子电池剩余寿命直接预测模型中关键参数进行优化得到优化结果如表1所示。
表1灰狼群算法优化LSTM网络寻优结果统计
锂离子电池样本 B0005
种群大小 16
默认迭代次数 40
最终适应度值 0.6364
训练集长度(numTrain) 90
验证集长度(numValidation) 7
隐含层节点数(numHiddenUnits) 10
全连接层节点数(numfullyConnectLayer) 10
Droupout层丢弃概率(pro_dropoutLayer) 0.5152
最大训练次数(maxEpochs) 1046
初始学习率(initiaLearnRate) 0.0001
将表1列出的模型参数寻优结果带入到LSTM网络中,得到本发明的锂离子电池剩余寿命直接预测模型进行预测的锂离子电池容量预测结果如图4所示,从图中可以看出,利用本发明的锂离子电池剩余寿命直接预测模型预测的锂离子电池容量变化趋势与实际锂离子电池容量变化趋势较为接近,预测电池容量曲线相对较为平缓,预测电池容量数据到达失效阈值时对应的充放电周期与实际电池容量数据达到失效阈值时对应的充放电周期对比结果如表2所示,从中可以看出,预测结果与实际结果较为接近,本发明的锂离子电池剩余寿命直接预测模型能够较为准确的反映锂离子电池容量数据的变化趋势,可有效反映锂离子电池的剩余寿命。
表2预测电池容量数据与实际电池容量数据结果对比
锂离子电池样本 B0005
预测数据达到失效阈值时的充放电周期(cycle) 125
实际数据达到失效阈值时的充放电周期(cycle) 129
为了验证本发明的锂离子电池剩余寿命直接预测模型所选用的灰狼群算法在参数优化方面的有效性,本实施例还提供了使用较为经典的遗传算法(GA)、粒子群算法(PSO)替换灰狼群算法,优化本发明的锂离子电池剩余寿命直接预测模型中的关键参数,对比上述两种方法在达到本发明的锂离子电池剩余寿命直接预测模型最大迭代次数时的最优适应度值、达到最终稳定状态时的适应度值以及达到最终稳定状态所需迭代次数(综合考虑LSTM网络训练过程所需时间以及优化算法优化效果,上述两种优化算法的最大迭代步骤设置为100次,种群大小与本章所提模型中种群大小一致)。对上述两种优化算法的寻优结果进行统计,如表3所示。从中可以看出,上述两种优化算法的搜索速度较为缓慢,当达到本发明的锂离子电池剩余寿命直接预测模型的最大迭代次数时,搜索到的最优适应度值相对于本章所提模型的搜索结果 仍然存在较大差距;当上述两种优化算法的搜索结果最终稳定不变时,其对应的最优适应度值仍与本发明的锂离子电池剩余寿命直接预测模型的最优适应度值差别较大。由此,证明了本发明所选用的灰狼群算法在参数优化方面的有效性。
表3其他优化算法优化结果统计
寻优算法 遗传算法 粒子群算法
锂离子电池样本 B0005 B0005
执行当前迭代次数后的适应度值 0.9638 0.7854
达到最终稳定状态时的适应度值 0.7125 0.8623
达到最终稳定状态所需迭代次数 73 82
为了验证本发明的锂离子电池剩余寿命直接预测模型所选用的LSTM网络在锂离子电池容量数据预测方面的有效性,本实施例还提供了两个对比实例。在两个对比实例中,均选用具有三个隐含层的浅层BP网络作为预测器,预测锂离子电池容量数据变化情况。与此同时,针对BP网络中的不同隐含层节点数以及BP网络最大训练次数等参数选取问题,利用灰狼群算法优化BP网络中的上述参数,实现参数的自适应选择。考虑到利用不同划分准则得到的训练集、验证集数据对锂离子电池剩余寿命直接预测模型的预测精度影响较大,本实施例中,在所提第一个对比实例中,训练集、验证集划分准则与本发明的锂离子电池剩余寿命直接预测模型中的训练集、验证集划分准则一致;所提第二个对比实例中,将训练集、验证集长度也作为待优化参数。下面对两个对比实例的结果分别进行概述。
(1)训练集和验证集长度参数不作为待优化变量时的寻优结果
当训练集和验证集长度参数不作为待优化变量时,灰狼群算法优化BP网络的寻优结果如表4所示。从表可知,利用灰狼群优化BP网络预测模型得到的最终适应度值相对本发明的锂离子电池剩余寿命直接预测模型的适应度值较大,BP网络对验证集电池容量数据的预测效果也相对较差。
利用参数优化后的BP网络预测测试集数据对应电池容量变化规律,结果如图5所示。从图可知,尽管BP网络预测电池容量数据到达失效阈值的时间总是比真实锂离子电池容量数据到达失效阈值的时间要短一些,依据BP网络预测后的电池容量数据可对锂离子电池剩余寿命做出较为保守的评估,但是,由于BP网络属于浅层网络,对时间序列的预测能力较差,预测电池容量数据与真实电池容量数据间的差异较大,其预测效果相对于本发明的锂离子电池剩余寿命直接预测模型的预测效果仍有待提高。
表4训练集和验证集长度不变时灰狼群算法优化BP网络寻优结果统计
Figure PCTCN2019081356-appb-000015
(2)训练集和验证集长度参数作为待优化变量时的寻优结果
当训练集和验证集长度参数作为待优化变量时,灰狼群优化BP网络的寻优结果如表5所示。从表可知,利用灰狼群优化BP网络预测模型得到的最终适应度值相对本发明的锂离子电池剩余寿命直接预测模型的适应度值仍然较大,BP网络对验证集电池容量数据的预测效果仍然存在较大提升空间。将优化后的训练集、测试集划分准则代入BP网络中,用于预测测试集数据对应的锂离子电池容量变化规律,结果如图6所示。从图可知,利用基于灰狼群优化BP神经网络的预测模型得到的预测电池容量数据与真实电池容量数据的变化趋势仍然存在较大差异。在B0005锂离子电池容量的预测结果中,预测电池容量数据相对于实际电池容量数据达到失效阈值的时间显著提前,且预测电池容量数据随着充放电周期呈直线下降趋势,并没有出现实际电池容量数据中的波动现象。由此,证明了本发明的锂离子电池剩余寿命直接预测模型选用LSTM网络作为预测器在锂离子电池容量数据预测方面的有效性。
表5训练集和验证集长度改变时灰狼群算法优化BP网络寻优结果统计
Figure PCTCN2019081356-appb-000016
Figure PCTCN2019081356-appb-000017
最后应说明的是:以上实施例仅用以说明本发明的技术方案,而非对其限制;尽管参照前述实施例对本发明进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述实施例所记载的技术方案进行修改,或者对其中部分或者全部技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本发明权利要求所限定的范围。

Claims (4)

  1. 一种基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,其特征在于:包括以下步骤:
    步骤1、获取锂离子电池的监测数据,并从中提取出锂离子电池容量数据,将这些电池容量数据划分成训练数据集、验证数据集和测试数据集,同时对这些电池容量数据进行归一化处理;
    步骤2、确定长短期记忆网结构,构造基于LSTM的锂离子电池剩余寿命预测模型;
    所述锂离子电池剩余寿命预测模型包括输入层、LSTM层、全连接层、Droupout层、全连接层、回归层以及输出层;第一层全连接层中每个神经元与其前一层LSTM层进行全连接,起到特征融合的作用;将Droupout层添加到第一层全连接层之上,起到防止过拟合和提高泛化能力的作用;Droupout层在每次参数训练过程中,以概率p舍弃部分神经元,剩余神经元以1-p的概率予以率保留;同时在Droupout层上添加神经元个数为1的全连接层以及回归层,确保输出结果为连续的预测值;
    步骤3:利用灰狼群算法优化锂离子电池剩余寿命直接预测模型中的关键参数,得到基于灰狼群优化LSTM网络的直接预测模型;
    所述锂离子电池剩余寿命直接预测模型中的关键参数包括作为锂离子电池容量数据划分准则的训练集长度numTrain、验证集长度numValidation以及LSTM网络的结构参数LSTM网络隐含层神经元节点数numHiddenUnits、全连接层节点数numfullyConnectedLayer、Droupout层舍弃概率pro_dropoutLayer、训练过程最大训练次数maxEpochs和初始学习率initialLearnRate七个参数;
    步骤4:利用优化数据确定最优的锂离子电池剩余寿命直接预测模型;
    根据优化得到的锂离子电池数据划分准则,将锂离子数据分成训练数据集和测试数据集,并将训练集样本作为长短期记忆网络模型的输入;然后再通过优化得到的其他参数训练长短期记忆网络,训练后的长短期记忆网络模型为最优网络结构;
    步骤5:利用最优的锂离子电池剩余寿命直接预测模型预测后期锂离子电池容量数据;
    将训练样本中最后充放电周期数据作为LSTM网络的输入,LSTM网络的输出为下一充放电周期锂离子电池容量数据的预测值;再次将下一充放电周期锂离子电池容量预测值作为LSTM网络的输入,得到LSTM网络的输出作为后续充放电周期对应的锂离子电池容量预测值;依次循环,直至锂离子电池容量预测值到达额定失效阈值。
  2. 根据权利要求1所述的基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,其特征在于:所述步骤3的具体方法为:
    步骤3.1:参数初始化:将上述七个待优化参数作为灰狼群优化算法中灰狼个体的位置向 量X,初始化产生种群个体数为N的初始化种群,并通过适应度函数计算公式计算初始种群个体对应的适应度值;
    步骤3.2:采用灰狼群优化算法对锂离子电池剩余寿命预测模型的参数进行更新,确保适应度值最小,由此得到优化后的锂离子电池容量数据划分准则以及LSTM网络结构参数。
  3. 根据权利要求2所述的基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,其特征在于:步骤3.1所述适应度函数的构造方法为:
    (1)将锂离子电池容量数据进行一次差分处理,将其转化成LSTM网络训练过程所需平稳时间序列;
    假设原始电池容量数据为F={f 1,f 2,…,f S},其中,S表示锂离子电池总的充放电周期次数,对其进行一次差分处理后,得到的时间序列如下公式所示:
    Figure PCTCN2019081356-appb-100001
    (2)在基于灰狼优化的LSTM网络直接预测模型中,选取到达失效阈值之前的第1次至第numTrain次充放电过程对应的电池容量数据用于训练LSTM网络;选取到达失效阈值之前的第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量数据用于验证LSTM网络的预测能力;LSTM网络训练过程中,将前numTrain-1次充放电周期的电池容量数据依次作为LTSM网络的输入,将当前充放电周期的下一充放电周期的电池容量数据作为LSTM网络的输出;LSTM网络训练结束后,以第numTrain次充放电周期的电池容量数据作为LSTM网络的输入,预测下一充放电周期的电池容量数据;然后,以下一充放电周期的电池容量数据再次作为LSTM网络的输入,预测后续充放电周期对应的电池容量数据,上述过程不断重复直至预测充放电周期次数到达numValidation;
    假设第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量预测数据为:
    Fp sec={fp numTrain+1,fp numTrain+2,…,fp numTrain+numValidation}.    (2)
    与之相对应的,经差分处理后第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量真实数据为:
    Fr sec={fr numTrain+1,fr numTrain+2,…,fr numTrain+numValidation}.    (3)
    构造如下函数表示预测后的电池容量数据与差分处理后的电池容量数据之间的关系:
    Figure PCTCN2019081356-appb-100002
    其中,Fit 1表示预测后的电池容量数据与差分处理后的电池容量数据之间的关系,length(Fp sec)为电池容量预测数据的长度;
    (3)对式2所示锂离子电池容量预测数据进行逆差分化处理后,将预测得到的第numTrain+1次至第numTrain+numValidation次充放电过程对应的电池容量数据恢复至原始电池容量数据区间,得到恢复后的电池容量数据
    Figure PCTCN2019081356-appb-100003
    其表达式为:
    Figure PCTCN2019081356-appb-100004
    其中,
    Figure PCTCN2019081356-appb-100005
    与之相对应的,第numTrain+1次至第numTrain+numValidation次充放电过程对应的原始电池容量数据为:
    F sec={f numTrain+1,f numTrain+2,…,f numTrain+numValidation}.    (7)
    构造如下函数表示恢复至原始电池容量区间的电池容量数据与原始电池容量数据之间的关系:
    Figure PCTCN2019081356-appb-100006
    (4)通过上述两个表述预测数据与真实数据间相近程度的关系式得到基于灰狼群优化的LSTM网络直接预测模型的适应度函数Fit direct,如下公式所示:
    Fit direct=Fit 1+Fit 2.    (9)。
  4. 根据权利要求2所述的基于灰狼群优化LSTM网络的锂离子电池剩余寿命预测方法,其特征在于:所述步骤3.2的具体方法为:
    在灰狼群算法中,为了模拟灰狼群的社会行为,将距离猎物最近的个体,即适应度值最 小的个体称为首领狼α,将距离猎物较近的其他两个个体,即适应度值较小的其他两个个体称为助理狼β和δ,剩余狼群个体表示为ω;捕猎过程中,利用距离猎物较近的三只狼,即α狼、β狼和δ狼引导剩余狼群个体ω对猎物进行搜索;搜索过程中,灰狼群个体位置更新公式如下公式所示:
    X(t+1)=X p(t)-A·d,    (10a)
    d=|C·X p(t)-X(t)|,    (10b)
    其中,A和C表示系数因子,t表示迭代次数,X p表示当前猎物的位置向量,X表示灰狼个体的位置向量;系数因子A和C的计算公式如下所示:
    A=2a·r 1-a,    (11a)
    C=2·r 2,    (11b)
    其中,r 1和r 2是[0,1]范围内的随机数,系数a随着迭代次数的增加从2到0线性递减;
    在搜索猎物过程中,由于首领狼α、助理狼β和δ距离猎物较近,剩余狼群个体ω的位置根据处于领导阶层的首领狼α、助理狼β和δ的位置进行更新,其表示式为:
    d α=|C·X α-X|,    (12a)
    d β=|C·X β-X|,    (12b)
    d δ=|C·X δ-X|,    (12c)
    其中,X α、X β和X δ分别表示首领狼α、助理狼β和δ所处的位置,d α、d β和d δ分别表示当前狼群趋向于猎物位置的近似距离,通过如下计算公式确定当前狼群与猎物位置间的距离为:
    Figure PCTCN2019081356-appb-100007
    X 1=X α-A 1·d α,    (13b)
    X 2=X β-A 2·d β,    (13c)
    X 3=X δ-A 3·d δ,    (13d)
    其中,A 1、A 2和A 3是控制灰狼群个体前进或后退的系数因子,X(t+1)为狼群t+1次迭代 时所处位置。
PCT/CN2019/081356 2019-03-27 2019-04-04 基于灰狼群优化lstm网络的锂离子电池剩余寿命预测方法 WO2020191801A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910238129.9 2019-03-27
CN201910238129.9A CN109993270B (zh) 2019-03-27 2019-03-27 基于灰狼群优化lstm网络的锂离子电池剩余寿命预测方法

Publications (1)

Publication Number Publication Date
WO2020191801A1 true WO2020191801A1 (zh) 2020-10-01

Family

ID=67131557

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/081356 WO2020191801A1 (zh) 2019-03-27 2019-04-04 基于灰狼群优化lstm网络的锂离子电池剩余寿命预测方法

Country Status (2)

Country Link
CN (1) CN109993270B (zh)
WO (1) WO2020191801A1 (zh)

Cited By (33)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112348271A (zh) * 2020-11-12 2021-02-09 华北电力大学 基于vmd-ipso-gru的短期光伏功率预测方法
CN112379274A (zh) * 2020-11-16 2021-02-19 河南科技大学 一种动力电池剩余寿命预测方法
CN112461919A (zh) * 2020-11-10 2021-03-09 云南电网有限责任公司保山供电局 应用多频超声技术检测变压器油理化性能的系统及方法
CN112487702A (zh) * 2020-10-26 2021-03-12 湖州师范学院 一种锂离子电池剩余使用寿命预测方法
CN112666479A (zh) * 2020-12-02 2021-04-16 西安交通大学 一种基于充电循环融合的电池寿命预测方法
CN112731183A (zh) * 2020-12-21 2021-04-30 首都师范大学 一种基于改进的elm的锂离子电池寿命预测方法
CN112784491A (zh) * 2021-01-26 2021-05-11 浙江中新电力工程建设有限公司 一种基于lstm与iqpso面向高弹性电网的城市充电网点规划方法
CN113283576A (zh) * 2021-01-18 2021-08-20 上海应用技术大学 一种基于鲸鱼算法优化lstm的频谱感知方法
CN113361777A (zh) * 2021-06-08 2021-09-07 淮阴工学院 基于vmd分解和ihho优化lstm的径流预测方法及系统
CN113536671A (zh) * 2021-07-09 2021-10-22 北京航空航天大学 一种基于lstm的锂电池寿命预测方法
CN113761795A (zh) * 2021-08-17 2021-12-07 浙江工商大学 一种飞机发动机故障检测方法和系统
CN113960482A (zh) * 2021-09-03 2022-01-21 西南科技大学 基于改进灰狼粒子滤波的锂电池荷电状态智能预测方法
CN113988399A (zh) * 2021-10-22 2022-01-28 石河子大学 综合能源调度方法、装置、电子设备及存储介质
CN114239373A (zh) * 2021-12-17 2022-03-25 上海空间电源研究所 一种基于粒子群算法的航天bdr模块剩余寿命预测方法
CN114264967A (zh) * 2021-12-14 2022-04-01 哈尔滨工业大学 基于容量损失机理的退役电池余能快速估计方法及系统
CN114354455A (zh) * 2022-01-17 2022-04-15 北京石油化工学院 一种电池浆料粒度分布在线测量的方法
CN114487848A (zh) * 2022-01-17 2022-05-13 北京和利时系统集成有限公司 一种蓄电池的状态计算方法和装置
CN114706144A (zh) * 2022-02-24 2022-07-05 南方海洋科学与工程广东省实验室(湛江) 海洋环境预报方法、装置、系统及存储介质
CN114744657A (zh) * 2022-04-19 2022-07-12 国网浙江省电力有限公司宁波供电公司 一种基于微电网的电池储能系统尺寸优化方法
CN115062537A (zh) * 2022-06-06 2022-09-16 中国人民武装警察部队工程大学 基于醉汉漫步混沌狼群算法的光伏电池模型参数辨识方法
CN115242428A (zh) * 2022-06-08 2022-10-25 大连大学 一种基于优化cw-rnn的网络安全态势预测方法
CN115276177A (zh) * 2022-08-18 2022-11-01 上海采日能源科技有限公司 储能电池充放电功率控制方法、装置及电池控制系统
CN115659844A (zh) * 2022-12-05 2023-01-31 南方电网数字电网研究院有限公司 基于风电场动态模型的仿真方法、装置和计算机设备
CN115860277A (zh) * 2023-02-27 2023-03-28 西安骏硕通信技术有限公司 一种数据中心能耗预测方法及系统
CN115954504A (zh) * 2022-12-17 2023-04-11 福州大学 一种基于cgsasa算法的pemfc模型参数优化方法
CN116298947A (zh) * 2023-03-07 2023-06-23 中国铁塔股份有限公司黑龙江省分公司 一种蓄电池核容监测装置
CN116506307A (zh) * 2023-06-21 2023-07-28 大有期货有限公司 全链路的网络延时情况分析系统
CN116774081A (zh) * 2021-12-12 2023-09-19 崔跃芹 一种采用复合寿命指标的充电电池寿命预测方法、装置、电子设备及可读存储介质
CN117130283A (zh) * 2023-10-26 2023-11-28 吉林大学 一种玉米按需施肥控制系统和土壤氮含量软测量方法
CN117408720A (zh) * 2023-11-14 2024-01-16 武汉亿量科技有限公司 一种电销外呼被投诉标记预防方法、装置、设备及介质
CN117808034A (zh) * 2024-02-29 2024-04-02 济南农智信息科技有限公司 一种基于狼鸟优化算法的农作物产量预测优化方法
CN117877028A (zh) * 2024-03-13 2024-04-12 浙江大学 基于微观图像特征的电机绝缘寿命预测方法及系统
CN114264967B (zh) * 2021-12-14 2024-05-24 哈尔滨工业大学 基于容量损失机理的退役电池余能快速估计方法及系统

Families Citing this family (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110333462B (zh) * 2019-08-08 2021-04-30 首都师范大学 一种在随机放电环境下基于dgwo-elm的锂离子电池寿命预测方法
CN110554280B (zh) * 2019-08-09 2022-02-15 上海电力大学 基于层级模型和改进灰狼优化算法的配电网故障定位方法
CN110703120A (zh) * 2019-09-29 2020-01-17 上海海事大学 基于粒子滤波和长短时记忆网络的锂离子电池寿命预测方法
CN110824293A (zh) * 2019-10-15 2020-02-21 淮阴工学院 一种基于狼群算法的多特征融合参数的电网故障诊断方法
CN111103544B (zh) * 2019-12-26 2021-12-21 江苏大学 基于长短时记忆lstm和粒子滤波pf的锂离子电池剩余使用寿命预测方法
CN111308375B (zh) * 2020-02-04 2021-12-07 浙江大学 一种基于lstm-ffnn的电动叉车锂离子电池健康状态预测方法
CN111371607B (zh) * 2020-02-28 2022-09-16 大连大学 一种基于决策灰狼算法优化lstm的网络流量预测方法
CN111443294B (zh) * 2020-04-10 2022-09-23 华东理工大学 一种锂离子电池剩余寿命间接预测方法及装置
CN111667098B (zh) * 2020-05-14 2023-04-18 湖北工业大学 一种基于多模型组合优化的风力电站输出功率预测方法
CN111948563B (zh) * 2020-06-19 2021-12-17 浙江大学 一种基于多神经网络耦合的电动叉车锂电池剩余寿命预测方法
CN112001113B (zh) * 2020-07-02 2023-12-19 浙江大学 一种基于粒子群优化长短时记忆网络的电池寿命预测方法
CN112038670B (zh) * 2020-07-31 2021-09-14 上海捷氢科技有限公司 一种质子交换膜燃料电池及其水状态监测方法和装置
CN112416913B (zh) * 2020-10-15 2023-04-25 中国人民解放军空军工程大学 一种基于gwo-bp算法的飞机燃油系统状态缺失值补充方法
CN112800659A (zh) * 2021-01-28 2021-05-14 福建师范大学 基于目标分解的改进多目标粒子群优化的锂离子电池容量预测方法
CN113486569B (zh) * 2021-06-04 2023-10-10 江苏科技大学 一种离心泵剩余寿命预测方法
CN113569460A (zh) * 2021-06-08 2021-10-29 北京科技大学 实车燃料电池系统状态多参数预测方法及装置
CN113884937B (zh) * 2021-11-22 2022-08-30 江南大学 基于分解集成策略的锂离子电池剩余寿命预测方法及系统
CN114510870B (zh) * 2022-01-07 2024-04-16 华东交通大学 一种城市轨道交通地下结构的剩余寿命预测方法及装置
CN114545280B (zh) * 2022-02-24 2022-11-15 苏州市职业大学 一种基于寻优算法的新能源汽车锂电池寿命预测方法
CN115577755A (zh) * 2022-11-28 2023-01-06 中环服(成都)科技有限公司 机器人位姿矫正方法、装置、计算机设备和存储介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107870306A (zh) * 2017-12-11 2018-04-03 重庆邮电大学 一种基于深度神经网络下的锂电池荷电状态预测算法
CN108519556A (zh) * 2018-04-13 2018-09-11 重庆邮电大学 一种基于循环神经网络的锂离子电池soc预测方法
CN109143105A (zh) * 2018-09-05 2019-01-04 上海海事大学 一种电动汽车锂离子电池的荷电状态计算方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103033761B (zh) * 2012-12-17 2014-12-10 哈尔滨工业大学 动态灰色相关向量机的锂离子电池剩余寿命预测方法
CN103399276B (zh) * 2013-07-25 2016-01-20 哈尔滨工业大学 一种锂离子电池容量估计及剩余循环寿命预测方法
CN107067121A (zh) * 2017-06-13 2017-08-18 广东工业大学 一种基于多目标的改进灰狼优化算法
CN108510074A (zh) * 2018-05-30 2018-09-07 江苏理工学院 一种改进gwo算法的实现方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107870306A (zh) * 2017-12-11 2018-04-03 重庆邮电大学 一种基于深度神经网络下的锂电池荷电状态预测算法
CN108519556A (zh) * 2018-04-13 2018-09-11 重庆邮电大学 一种基于循环神经网络的锂离子电池soc预测方法
CN109143105A (zh) * 2018-09-05 2019-01-04 上海海事大学 一种电动汽车锂离子电池的荷电状态计算方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LIU, YUEFENG ET AL.: "Lithium-ion Battery Remaining Useful Life Prediction with Long Short-term Memory Recurrent Neural Network", ANNUAL CONFERENCE OF THE PROGNOSTICS AND HEALTH MANAGEMENT SOCIETY 2017, 5 October 2017 (2017-10-05), XP055737395, DOI: 20191217154041A *
WANG, SHUQIN ET AL.: "Long Short Term Memory Networks Based on Grey Wolf Optimizer in Time Series Forecasting", CHINA SCIENCEPAPER, vol. 12, no. 20, 31 October 2017 (2017-10-31), ISSN: 2095-2783, DOI: 20191217161846A *

Cited By (55)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112487702A (zh) * 2020-10-26 2021-03-12 湖州师范学院 一种锂离子电池剩余使用寿命预测方法
CN112487702B (zh) * 2020-10-26 2024-04-05 湖州师范学院 一种锂离子电池剩余使用寿命预测方法
CN112461919A (zh) * 2020-11-10 2021-03-09 云南电网有限责任公司保山供电局 应用多频超声技术检测变压器油理化性能的系统及方法
CN112348271A (zh) * 2020-11-12 2021-02-09 华北电力大学 基于vmd-ipso-gru的短期光伏功率预测方法
CN112348271B (zh) * 2020-11-12 2024-01-30 华北电力大学 基于vmd-ipso-gru的短期光伏功率预测方法
CN112379274A (zh) * 2020-11-16 2021-02-19 河南科技大学 一种动力电池剩余寿命预测方法
CN112666479A (zh) * 2020-12-02 2021-04-16 西安交通大学 一种基于充电循环融合的电池寿命预测方法
CN112666479B (zh) * 2020-12-02 2023-05-16 西安交通大学 一种基于充电循环融合的电池寿命预测方法
CN112731183A (zh) * 2020-12-21 2021-04-30 首都师范大学 一种基于改进的elm的锂离子电池寿命预测方法
CN112731183B (zh) * 2020-12-21 2023-04-21 首都师范大学 一种基于改进的elm的锂离子电池寿命预测方法
CN113283576A (zh) * 2021-01-18 2021-08-20 上海应用技术大学 一种基于鲸鱼算法优化lstm的频谱感知方法
CN113283576B (zh) * 2021-01-18 2024-02-06 上海应用技术大学 一种基于鲸鱼算法优化lstm的频谱感知方法
CN112784491A (zh) * 2021-01-26 2021-05-11 浙江中新电力工程建设有限公司 一种基于lstm与iqpso面向高弹性电网的城市充电网点规划方法
CN112784491B (zh) * 2021-01-26 2024-04-16 浙江中新电力工程建设有限公司 一种基于lstm与iqpso面向高弹性电网的城市充电网点规划方法
CN113361777A (zh) * 2021-06-08 2021-09-07 淮阴工学院 基于vmd分解和ihho优化lstm的径流预测方法及系统
CN113361777B (zh) * 2021-06-08 2023-11-03 淮阴工学院 基于vmd分解和ihho优化lstm的径流预测方法及系统
CN113536671A (zh) * 2021-07-09 2021-10-22 北京航空航天大学 一种基于lstm的锂电池寿命预测方法
CN113536671B (zh) * 2021-07-09 2023-05-30 北京航空航天大学 一种基于lstm的锂电池寿命预测方法
CN113761795A (zh) * 2021-08-17 2021-12-07 浙江工商大学 一种飞机发动机故障检测方法和系统
CN113960482A (zh) * 2021-09-03 2022-01-21 西南科技大学 基于改进灰狼粒子滤波的锂电池荷电状态智能预测方法
CN113988399A (zh) * 2021-10-22 2022-01-28 石河子大学 综合能源调度方法、装置、电子设备及存储介质
CN116774081B (zh) * 2021-12-12 2024-02-09 崔跃芹 一种采用复合寿命指标的充电电池寿命预测方法、装置、电子设备及可读存储介质
CN116774081A (zh) * 2021-12-12 2023-09-19 崔跃芹 一种采用复合寿命指标的充电电池寿命预测方法、装置、电子设备及可读存储介质
CN114264967A (zh) * 2021-12-14 2022-04-01 哈尔滨工业大学 基于容量损失机理的退役电池余能快速估计方法及系统
CN114264967B (zh) * 2021-12-14 2024-05-24 哈尔滨工业大学 基于容量损失机理的退役电池余能快速估计方法及系统
CN114239373A (zh) * 2021-12-17 2022-03-25 上海空间电源研究所 一种基于粒子群算法的航天bdr模块剩余寿命预测方法
CN114354455A (zh) * 2022-01-17 2022-04-15 北京石油化工学院 一种电池浆料粒度分布在线测量的方法
CN114487848A (zh) * 2022-01-17 2022-05-13 北京和利时系统集成有限公司 一种蓄电池的状态计算方法和装置
CN114487848B (zh) * 2022-01-17 2024-05-03 北京和利时系统集成有限公司 一种蓄电池的状态计算方法和装置
CN114354455B (zh) * 2022-01-17 2023-12-08 北京石油化工学院 一种电池浆料粒度分布在线测量的方法
CN114706144A (zh) * 2022-02-24 2022-07-05 南方海洋科学与工程广东省实验室(湛江) 海洋环境预报方法、装置、系统及存储介质
CN114744657B (zh) * 2022-04-19 2024-05-28 国网浙江省电力有限公司宁波供电公司 一种基于微电网的电池储能系统尺寸优化方法
CN114744657A (zh) * 2022-04-19 2022-07-12 国网浙江省电力有限公司宁波供电公司 一种基于微电网的电池储能系统尺寸优化方法
CN115062537A (zh) * 2022-06-06 2022-09-16 中国人民武装警察部队工程大学 基于醉汉漫步混沌狼群算法的光伏电池模型参数辨识方法
CN115062537B (zh) * 2022-06-06 2024-04-12 中国人民武装警察部队工程大学 基于醉汉漫步混沌狼群算法的光伏电池模型参数辨识方法
CN115242428A (zh) * 2022-06-08 2022-10-25 大连大学 一种基于优化cw-rnn的网络安全态势预测方法
CN115242428B (zh) * 2022-06-08 2024-05-31 大连大学 一种基于优化cw-rnn的网络安全态势预测方法
CN115276177B (zh) * 2022-08-18 2023-09-26 上海采日能源科技有限公司 储能电池充放电功率控制方法、装置及电池控制系统
CN115276177A (zh) * 2022-08-18 2022-11-01 上海采日能源科技有限公司 储能电池充放电功率控制方法、装置及电池控制系统
CN115659844A (zh) * 2022-12-05 2023-01-31 南方电网数字电网研究院有限公司 基于风电场动态模型的仿真方法、装置和计算机设备
CN115954504A (zh) * 2022-12-17 2023-04-11 福州大学 一种基于cgsasa算法的pemfc模型参数优化方法
CN115860277B (zh) * 2023-02-27 2023-05-09 西安骏硕通信技术有限公司 一种数据中心能耗预测方法及系统
CN115860277A (zh) * 2023-02-27 2023-03-28 西安骏硕通信技术有限公司 一种数据中心能耗预测方法及系统
CN116298947B (zh) * 2023-03-07 2023-11-03 中国铁塔股份有限公司黑龙江省分公司 一种蓄电池核容监测装置
CN116298947A (zh) * 2023-03-07 2023-06-23 中国铁塔股份有限公司黑龙江省分公司 一种蓄电池核容监测装置
CN116506307B (zh) * 2023-06-21 2023-09-12 大有期货有限公司 全链路的网络延时情况分析系统
CN116506307A (zh) * 2023-06-21 2023-07-28 大有期货有限公司 全链路的网络延时情况分析系统
CN117130283B (zh) * 2023-10-26 2024-01-05 吉林大学 一种玉米按需施肥控制系统和土壤氮含量软测量方法
CN117130283A (zh) * 2023-10-26 2023-11-28 吉林大学 一种玉米按需施肥控制系统和土壤氮含量软测量方法
CN117408720A (zh) * 2023-11-14 2024-01-16 武汉亿量科技有限公司 一种电销外呼被投诉标记预防方法、装置、设备及介质
CN117408720B (zh) * 2023-11-14 2024-05-07 武汉亿量科技有限公司 一种电销外呼被投诉标记预防方法、装置、设备及介质
CN117808034A (zh) * 2024-02-29 2024-04-02 济南农智信息科技有限公司 一种基于狼鸟优化算法的农作物产量预测优化方法
CN117808034B (zh) * 2024-02-29 2024-05-10 济南农智信息科技有限公司 一种基于狼鸟优化算法的农作物产量预测优化方法
CN117877028A (zh) * 2024-03-13 2024-04-12 浙江大学 基于微观图像特征的电机绝缘寿命预测方法及系统
CN117877028B (zh) * 2024-03-13 2024-05-14 浙江大学 基于微观图像特征的电机绝缘寿命预测方法及系统

Also Published As

Publication number Publication date
CN109993270A (zh) 2019-07-09
CN109993270B (zh) 2022-11-15

Similar Documents

Publication Publication Date Title
WO2020191801A1 (zh) 基于灰狼群优化lstm网络的锂离子电池剩余寿命预测方法
WO2020191800A1 (zh) 基于wde优化lstm网络的锂离子电池剩余寿命预测方法
CN112241608B (zh) 一种基于lstm网络和迁移学习的锂电池寿命预测方法
Ma et al. A novel method for state of health estimation of lithium-ion batteries based on improved LSTM and health indicators extraction
Ma et al. A hybrid transfer learning scheme for remaining useful life prediction and cycle life test optimization of different formulation Li-ion power batteries
Che et al. Data efficient health prognostic for batteries based on sequential information-driven probabilistic neural network
CN112036084B (zh) 一种相似产品寿命迁移筛选方法和系统
KR20200119383A (ko) 인공 지능에 기반하여 배터리의 상태를 추정하는 장치 및 방법
Liu et al. Transfer learning for battery smarter state estimation and ageing prognostics: Recent progress, challenges, and prospects
CN111999648A (zh) 一种基于长短期记忆网络的锂电池剩余寿命预测方法
CN111680848A (zh) 基于预测模型融合的电池寿命预测方法及存储介质
CN110888058A (zh) 一种基于动力电池soc和soh联合估计的算法
CN111861013A (zh) 一种电力负荷预测方法及装置
Suryo et al. Improved time series prediction using LSTM neural network for smart agriculture application
Liu et al. Lithium-ion battery remaining useful life prediction with long short-term memory recurrent neural network
Li et al. Remaining useful life prediction for lithium-ion batteries with a hybrid model based on TCN-GRU-DNN and dual attention mechanism
CN112305441A (zh) 一种集成式聚类下的动力电池健康状态评估方法
CN115049115A (zh) 计及nwp风速横纵向误差的rdpg风速修正方法
CN112257348B (zh) 一种锂电池长期退化趋势预测方法
Mazzi et al. Lithium-ion battery state of health estimation using a hybrid model based on a convolutional neural network and bidirectional gated recurrent unit
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
CN116540134A (zh) 一种基于gwo-lstm的锂离子电池健康状态估计方法
CN112232557B (zh) 基于长短期记忆网络的转辙机健康度短期预测方法
CN114896865A (zh) 一种面向数字孪生的自适应演化神经网络健康状态在线预测方法
Reddy et al. Forecasting energy consumption using deep echo state networks optimized with genetic algorithm

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: 19920784

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19920784

Country of ref document: EP

Kind code of ref document: A1