CN112348185A - Lithium battery residual life prediction method based on variational modal decomposition and integrated depth model - Google Patents
Lithium battery residual life prediction method based on variational modal decomposition and integrated depth model Download PDFInfo
- Publication number
- CN112348185A CN112348185A CN202011216878.0A CN202011216878A CN112348185A CN 112348185 A CN112348185 A CN 112348185A CN 202011216878 A CN202011216878 A CN 202011216878A CN 112348185 A CN112348185 A CN 112348185A
- Authority
- CN
- China
- Prior art keywords
- lithium battery
- data
- sub
- battery
- input
- Prior art date
- Legal status (The legal status 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 status listed.)
- Withdrawn
Links
Images
Classifications
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Tests Of Electric Status Of Batteries (AREA)
Abstract
The invention discloses a lithium battery residual life prediction method based on Variational Modal Decomposition (VMD) and an integrated depth model. Taking the dischargeable capacity of the battery as a performance index for measuring the residual life of the battery, firstly, performing multi-scale decomposition on the dischargeable capacity data by using a VMD (virtual matrix display), and deeply mining implicit information behind different scales of the capacity data; and then respectively selecting two sub-learners, namely a long short-term memory neural network (LSTM) and a multilayer perceptron (MLP), for training aiming at different modal component characteristics, and integrating the results of the sub-learners based on a parallel framework to predict the residual service life of the lithium battery. The method can effectively sense the regeneration and fluctuation characteristics of the battery capacity, and has high prediction precision and generalization capability in the prediction of the residual service life of the lithium battery.
Description
Technical Field
The invention relates to the technical field of lithium battery residual life prediction, in particular to a lithium battery residual life prediction method based on variational modal decomposition and an integrated depth model.
Technical Field
Lithium batteries are widely used because of their advantages, such as high energy density, light weight, stable discharge, and low price. However, with the increase of the number of charge and discharge cycles in the use process of the lithium battery, the corrosion of electrode materials, the gradual aging of internal diaphragms, high and low temperature environments, complex use conditions and other factors cause the reduction of the activity of lithium ions which can be used by the battery, the decline of capacity and power, the influence on the remaining life of the battery, the gradual shortening of the service life, the direct failure of the functions of electric equipment, even equipment failure, and the threat of life safety of personnel. Therefore, in order to improve the reliability and safety of the electric equipment, the prediction of the residual life of the lithium battery is important.
The aging process of the lithium battery is very complicated, a plurality of internal and external factors interact with each other and are coupled with each other, and the lithium battery has typical nonlinear and uncertain characteristics, for example, after the lithium battery is placed for a period of time, the available capacity of the lithium battery slightly rises, and the capacity regeneration phenomenon occurs. Therefore, the performance degradation data of the lithium battery in the actual working state not only contains the overall performance degradation information, but also includes the capacity regeneration component caused by the placement of the battery and the fluctuation amount changed along with the environmental factors, so that the performance degradation process of the battery presents nonlinearity and time-varying property, and great difficulty is brought to the life prediction of the lithium battery. The degradation characteristic information of the original capacity data of the lithium battery is extracted under a single scale, the problem of instability of the capacity data caused by a capacity local regeneration phenomenon generated in the performance degradation process of the lithium battery is not fully considered, in order to reduce the complexity and instability of the data in the prediction of the service life of the lithium battery, empirical mode decomposition, wavelet decomposition and other methods are often used for decomposing the complex capacity degradation data of the battery, but the problems that the modal aliasing phenomenon is serious, the noise is sensitive and the like exist in the empirical mode decomposition method, and the problems that the wavelet base selection is difficult and the like exist in the wavelet decomposition method are not well solved.
In recent years, deep learning has been widely used in the field of predicting the remaining life of lithium batteries because of its advantages such as strong generalization ability and capability of adaptively extracting features from data. However, the existing lithium battery life prediction method based on the deep learning method mostly adopts a single prediction model, and is difficult to accurately describe the complicated lithium battery performance degradation process, so that the problems of poor generalization performance of the model, low life prediction precision, unstable prediction and the like are caused.
In order to solve the problems, the invention provides a lithium battery residual life prediction method based on variational modal multi-scale decomposition (VMD) and an integrated depth model.
Disclosure of Invention
The invention provides a method for predicting the remaining life of a lithium battery based on variational modal decomposition and an integrated depth model, and aims to solve the problems of unstable capacity data caused by insufficient consideration of a capacity local regeneration phenomenon generated in a lithium battery performance degradation process in a single scale, poor generalization capability in a single prediction model, unstable prediction and the like.
A lithium battery residual life prediction method based on variational modal decomposition and an integrated depth model is characterized in that: the method comprises the following steps:
step 3, constructing a long-short term memory neural network (LSTM) sub-learner: the intrinsic mode component reflects capacity regeneration and random fluctuation information and presents certain periodicity, so that a long-short-term memory neural network model is selected to train the intrinsic mode component, the intrinsic mode components of a plurality of training set data obtained by decomposition are respectively input into a plurality of long-short-term memory network models to be trained independently, and a plurality of LSTM sub-learners are constructed;
step 4, constructing a multilayer perceptron (MLP) sub-learner: because the residual component reflects the overall degradation trend of the battery and presents monotonicity and stability, the multilayer perceptron is selected for training the residual component, the residual component of the training set data obtained by decomposition is input into the multilayer perceptron for model training, and an MLP sub-learner is constructed;
step 5, sub-learner integration: respectively inputting the intrinsic modal component and the residual component of the prediction set data obtained by decomposition into a well-trained LSTM sub-learner and an MLP sub-learner, integrating prediction results of the sub-learners based on a parallel integration method, and outputting a lithium battery service life prediction result at the t +1 momentAccording toAnd (3) judging whether the health state of the lithium battery reaches an end of service life (EOL) condition, namely whether the prediction result reaches 80% of the rated capacity of the lithium battery, if the health state of the lithium battery does not reach the EOL condition, repeating the processes of the steps 3-5 to finish life prediction at the next moment and judge the health state of the lithium battery, and if the health state of the lithium battery reaches the EOL condition, stopping prediction.
Further, the lithium battery residual life prediction method based on variational modal decomposition and integrated depth model is characterized in that: the step 2 of original signal variation modal decomposition comprises the following steps:
step 21, initializing IMF components and center frequency: decomposing the capacity attenuation data C of the lithium battery into k IMF components, and taking each IMF component and the central frequency thereof as initialization values;
step 22, updating IMF components and center frequency: updating the IMF component and the center frequency according to the Fourier transform theorem;
step 23, updating the Lagrange multiplier;
step 24, iteration stop judgment: and E & gt 0 is set as the discrimination precision, if the residual error is smaller than the discrimination precision, the iteration is stopped, otherwise, the process of the steps 22-24 is repeated until the iteration is stopped.
Further, the lithium battery residual life prediction method based on variational modal decomposition and integrated depth model is characterized in that: the step 3 of constructing the sub-learner LSTM includes the following steps:
step 31, calculating the hidden layer output value: the input of the network model at the current moment consists of two parts, namely the input at the current moment (a modal vector after VMD decomposition) and the output value of the hidden layer at the previous moment, the output value of the hidden layer at the current moment is obtained by calculation, and the operation is repeated until all the input is read;
step 33, calculating the input door state value: the input gate determines the number of the input units for saving the network input at the current moment, so for the data input at the moment, the input gate can store the key information to the input gate unit to a limited extent;
step 35, calculating the current state value of the memory cell;
step 36, calculating an output gate state value: obtaining a current state value result;
step 37, unit memory output.
Further, the lithium battery residual life prediction method based on variational modal decomposition and integrated depth model is characterized in that: the step 4 of constructing the sub-learner MLP comprises the following steps:
step 41, MLP forward propagation: initializing a plurality of weight matrixes and calculating output of an output layer by using a deviation coefficient;
step 42, MLP back propagation: the most appropriate linear coefficient matrix bias vector is found, the optimal solution is solved by gradient descent, and iteration updating is carried out continuously, so that the method has excellent nonlinear fitting performance.
Compared with the prior art, the invention has the following advantages:
(1) in the face of random fluctuation and capacity regeneration phenomena in the prediction precision problem, the random fluctuation state cannot be completely predicted by a single scale, the method adopts a multi-scale decomposition method to decompose complex signals under different scales, the signals of different scales have different information, the key characteristic information of data can be effectively increased, the mutual influence among the information of different scales is reduced, the service life prediction precision is improved, and the method has important practical application value.
(2) The Empirical Mode Decomposition (EMD) has the defects of serious modal aliasing phenomenon, sensitivity to noise and the like in signal processing, and the variational modal decomposition adopted by the invention is a self-adaptive, quasi-orthogonal and completely non-recursive signal processing model, overcomes the problems of endpoint effect and modal component aliasing in the EMD method, can deeply mine hidden information behind different scales, and realizes effective separation of inherent modal components of complex signals.
(3) In order to solve the problems of poor generalization performance and low stability of a single model, a depth model integration method is adopted in the text, and a plurality of sub-learners with characteristics are integrated, so that an integral model with better performance is constructed, wherein a long-term neural network (LSTM) has excellent long-term prediction performance and can well predict the regeneration and random fluctuation characteristics of the battery capacity, and a multi-layer perceptron (MLP) has the characteristics of simple algorithm, easy realization and high calculation speed and can effectively perceive the integral degradation trend of the battery capacity.
Drawings
FIG. 1 is a flow chart of a lithium battery residual life prediction method based on variational modal decomposition and an integrated depth model.
FIG. 2 is a flow chart of a variational modal decomposition algorithm.
Fig. 3B 5 is a schematic diagram of information of each scale after decomposition of the battery dischargeable capacity data through variation mode.
Fig. 4B 7 is a schematic diagram of information of each scale after decomposition of the battery dischargeable capacity data through variation mode.
Fig. 5 is a graph of residual component prediction results of MLP versus decomposition of the dischargeable capacity of B5 and B7 batteries.
Fig. 6 is a graph of the results of eigenmode component predictions for decomposition of the dischargeable capacity of the LSTM versus B5 and B7 cells.
Fig. 7B 5 shows the predicted results of the battery prediction set data.
Fig. 8B 7 shows the predicted results of the battery prediction set data.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the following detailed description of the present invention will be made for clarity and completeness. It should be understood that the described embodiments are only some, not all, and not all embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without inventive step, are within the scope of the present invention.
The invention provides a method for predicting the residual life of a lithium battery based on variational modal decomposition and an integrated depth model, which is shown as a specific flow chart in figure 1 and comprises the following specific implementation steps:
the capacity degradation data of two lithium batteries B5 and B7 used in the example are originated from a NASA research center, the experimental object is a 18650 lithium battery, the rated capacity is 2Ah, and the experimental environment comprises: the system comprises a thermostat, a PXI case data acquisition module, an experiment control module, a computer and the like. The experiment is an accelerated life experiment of the lithium battery, the discharge capacity data of each time is collected, the service life of the lithium battery is defined by combining the national standard, the threshold value is determined to be 80% of the rated capacity of the lithium battery, and if the current capacity reaches 80% of the rated capacity, the battery is considered to be invalid.
According to the step 2, the original capacity data is subjected to variation modal decomposition, and the method comprises the following steps:
step 21, initializing the IMF component and the IMF center frequency: decomposing the capacity attenuation data C of the lithium battery into k decomposed IMF components { u }kI.e. uk}={u1,u2,…uk}; each IMF has a center frequency of [ omega ]kI.e., { ωk}={ω1,ω2,…ωk};
Step 22, update { ukAnd { omega } andk}: by utilizing Fourier transform theorem to convert to frequency domain processing, non-negative frequency integration can be carried out, and finally, quadratic optimization solution is calculated to carry out ukUpdating, wherein the specific expression is as follows:
similarly, also in the frequency domain, the center frequency problem is handled, for the component center frequency ukAnd ωkUpdate with push-down progressLine:
step 23, updating the Lagrange multiplier lambda:
step 24, judging whether to stop iteration: and E & gt 0 is set as the discrimination precision, if the residual error is smaller than the discrimination precision E, the iteration is stopped, and otherwise, the steps 22-24 are repeated until the iteration is stopped.
Step 3, when constructing the LSTM sub-learners of a plurality of eigenmode components, each hidden node is an information storage node, and there are usually three gates: the Input Gate (Input Gate) is responsible for node writing; an Output Gate (Output Gate) is responsible for the node Output result; forgetting gates (Forget gates) are responsible for node information storage, and the gates set corresponding parameters and only influence the connection of each node;
constructing an LSTM sub-learner according to the step 3, wherein the specific algorithm flow is as follows:
step 31, calculating hidden layer output: if at the current time t, assume that the input of the network model is input by the input x at the current timet(Modal vector after VMD decomposition) and output value c of hidden layer at previous timet-1Two parts are used for calculating the output value y of the hidden layer at the current momenttRepeating this operation until all inputs are read, the formula can be expressed as:
yt=σ(wxcxt+wccct-1+bc)
wherein: w is axcIs a weight from the input layer to the hidden layer, wccA hidden layer to hidden layer weight, bcFor the biasing of the hidden layer, σ is the sigmoid activation function.
step 33, calculating the input door state value it: for data entry at this moment itThere is a limit to storing critical information to the cell, which has an impact:
it=σ(wxixt+whiht-1+bc)
ft=σ(wxfxt+whfht-1+bf)
Step 35, calculating the current state value c of the memory cellt:
In the formula: denotes the convolution, ct-1Is the value of the cell at the last time,for temporary memory of status information.
Step 36, calculating the output gate state value otAnd obtaining the state value result at the time:
ot=σ(wxoxt+whoht-1+bo)
step 37, outputting the cell memory ht:
ht=ot tanh(ct)
The parameters of the long-short term memory neural network are set as follows: each batch has a size of 64, the number of nodes in the first layer is 50, the number of nodes in the second layer is 100, the activation function of the neuron is Relu, the loss function is set as Mean Square Error (MSE), the optimizer uses RMSProp, the learning rate is set to 0.005, and the number of training times is set to 500.
Step 4, when constructing the MLP sub-learner of the degradation trend component, inputting residual component signals of the decomposed training set data into the MLP, performing forward propagation first, calculating layer by layer according to the network model structure, and extracting features; then, gradient descent successive iteration is adopted to obtain model parameters meeting the precision requirement, and the weight and deviation of each neuron are obtained, wherein the hidden layer can sense the signal characteristics and extract signal information;
constructing an MLP sub-learner according to the step 4, wherein the specific steps are as follows:
step 41, MLP forward propagation: a plurality of weight matrices are initialized, and a deviation factor b. Firstly, starting from an input layer, taking a residual vector x after capacitance data decomposition of a test set as an input of an MLP, and performing linear calculation according to the formula:
hidden layer network weights and deviations, denoted w and b, respectively, may have multiple hidden layers, assuming that the m-1 th layer in a network has k neurons in total, and then the jth neuron in the mth layer outputs ajThe concrete formula is as follows:
wherein h isjIs the result of the weighted sum of all the inputs to this node, g () is an activation function,represents the jth neuron weight of the mth layer.Representing the ith neuron output at layer m-1.Representing the jth neuron bias at the mth layer.
After passing through one or more hidden layers, the data is finally transmitted to an output layer, and the output value of the output layer is as follows:
hkrepresenting the sum of the output layer neuron input weights.
Step 42, MLP back propagation: the objective of algorithm training is to make the prediction result calculated by the algorithm sufficiently close to the true value, and to achieve this goal, MLP uses the back propagation algorithm to find the most appropriate linear coefficient matrix w and bias vector b, which is similar to the above back propagation algorithm and is described as follows:
firstly, loss function calculation is carried out on the output predicted value:
based on the minimum loss function, each layer w, b is obtained by adopting a gradient descent method in a reverse way, and the formula is as follows:
and further, gradient descent is adopted to solve an optimal solution, and iteration updating is carried out continuously, so that the method has excellent nonlinear fitting performance.
The model parameters of the multilayer perceptron are set as follows: the hidden layer is set to be 2 layers, the number of neurons in each layer is set to be 32 and 4 respectively, the number of neurons in the output layer is set to be 1, the activation function of the neurons in the hidden layer is Relu, the loss function is set to be Mean Square Error (MSE), the optimizer uses Adam, the learning rate is set to be 0.001, and the training times are 500.
And 5, respectively inputting the intrinsic modal component and the residual component of the prediction set decomposed by the VMD into the trained LSTM sub-learner model and the MLP sub-learner model, adding the prediction results of the sub-learners based on a parallel integration method, outputting the prediction result of the battery life at the next moment, and completing the life prediction of the lithium battery.
Integrating the prediction results of the sub-learners according to the step 5 to obtain the life prediction results of the B5 and B7 battery prediction data shown in FIGS. 7 and 8, wherein the MAE of the B5 battery prediction set reaches 0.0030, and the RMSE reaches 0.0015; the B7 battery prediction set MAE reaches 0.0013 and RMSE reaches 0.0013. Through experimental verification, the integrated deep learning model has strong fitting capability, self-adaptive ground function fitting characteristic and stronger generalization capability, has stronger adaptability to the lithium battery degradation capacity data under two different working conditions of B5 and B7, and can accurately complete the life prediction.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (4)
1. The lithium battery residual life prediction method based on the variational modal decomposition and the integrated depth model is characterized by comprising the following steps of: the method comprises the following steps:
step 1, data preprocessing: screening out capacity attenuation data and corresponding cycle life in the discharging process of the lithium battery as residual life prediction data of the lithium battery, taking the capacity data of the lithium battery at the time 1, … and t as a training set, and taking the capacity data of the lithium battery after the time t +1 as a prediction data set, wherein t represents the current time;
step 2, original signal Variation Modal Decomposition (VMD): carrying out variation modal decomposition on the capacity attenuation data of the lithium battery in the training set and the test set to obtain the data including intrinsic modal components (IMF)1,…,IMFN) And a plurality of modal components of the residual component (r (t)), which are taken as characteristics of the lithium battery capacity data under different scales, wherein the residual component represents the overall degradation trend of the battery, and the intrinsic modal component represents the characteristics of battery capacity regeneration and random fluctuation;
step 3, constructing a long-term and short-term memory neural network (LSTM) of the sub-learner: the intrinsic mode component reflects capacity regeneration and random fluctuation information and presents certain periodicity, so that a long-short-term memory neural network model is selected to train the intrinsic mode component, the intrinsic mode components of a plurality of training set data obtained by decomposition are respectively input into a plurality of long-short-term memory network models to be trained independently, and a plurality of LSTM sub-learners are constructed;
step 4, constructing a multilayer perceptron (MLP) of the sub-learner: because the residual component reflects the overall degradation trend of the battery and presents monotonicity and stability, the multilayer perceptron is selected for training the residual component, the residual component of the training set data obtained by decomposition is input into the multilayer perceptron for model training, and an MLP sub-learner is constructed;
step 5, sub-learner integration: respectively inputting the intrinsic modal component and the residual component of the prediction set data obtained by decomposition into a well-trained LSTM sub-learner and an MLP sub-learner, integrating prediction results of the sub-learners based on a parallel integration method, and outputting a lithium battery service life prediction result at the t +1 momentAccording toJudging whether the health state of the battery reaches an end of service (EOL) condition of the battery, namely the prediction result reaches 80% of the rated capacity of the lithium battery, if not, repeating the steps 3-5 to complete the life prediction of the next moment and judgeHealth status of lithium battery.
2. The lithium battery remaining life prediction method according to claim 1, characterized in that: the step 2 of original signal variation modal decomposition comprises the following steps:
step 21, initializing IMF components and center frequency: decomposing the capacity attenuation data C of the lithium battery into k IMF components, and taking each IMF component and the central frequency thereof as initialization values;
step 22, updating IMF components and center frequency: updating the IMF component and the center frequency according to the Fourier transform theorem;
step 23, updating the Lagrange multiplier;
step 24, iteration stop judgment: and E & gt 0 is set as the discrimination precision, if the residual error is smaller than the discrimination precision, the iteration is stopped, and otherwise, the steps 22 to 24 are repeated.
3. The lithium battery remaining life prediction method according to claim 1, characterized in that: the step 3 of constructing the long-term and short-term memory neural network of the sub-learner comprises the following steps:
step 31, calculating the hidden layer output value: if the input of the network model is supposed to be composed of the input (the modal vector after the variational modal decomposition) of the current moment and the output value of the hidden layer of the previous moment at the current moment, calculating to obtain the output value of the hidden layer of the current moment, and repeating the operation until all the inputs are read;
step 32, temporarily memorizing the state information: before updating the memory unit, a temporary memory unit is generated, and temporary memory state information is obtained according to the input of the current moment and the hidden layer state value of the previous moment, so that the state is further updated;
step 33, calculating the input door state value: the input gate determines how much input of the network is stored in the input unit at the current moment, so for the data input at the moment, the input gate can store key information to the input gate unit to the limit;
step 34, calculating a forgetting gate state value; the forgetting gate determines how much the unit state at the previous moment is reserved to the current moment;
step 35, calculating the current state value of the memory cell;
step 36, calculating an output gate state value: obtaining a current state value result;
step 37, unit memory output.
4. The lithium battery remaining life prediction method according to claim 1, characterized in that: the step 4 of constructing the multilayer perceptron of the sub-learner comprises the following steps:
step 41, MLP forward propagation: initializing a plurality of weight matrixes and deviation coefficients, and calculating output of an output layer;
step 42, MLP back propagation: the most proper linear coefficient matrix and the most proper bias vector are found, the optimal solution is solved by adopting gradient descent, and the iterative updating is continuously carried out, so that the nonlinear fitting performance is excellent.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011216878.0A CN112348185A (en) | 2020-11-04 | 2020-11-04 | Lithium battery residual life prediction method based on variational modal decomposition and integrated depth model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011216878.0A CN112348185A (en) | 2020-11-04 | 2020-11-04 | Lithium battery residual life prediction method based on variational modal decomposition and integrated depth model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112348185A true CN112348185A (en) | 2021-02-09 |
Family
ID=74430015
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011216878.0A Withdrawn CN112348185A (en) | 2020-11-04 | 2020-11-04 | Lithium battery residual life prediction method based on variational modal decomposition and integrated depth model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112348185A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113065283A (en) * | 2021-03-26 | 2021-07-02 | 深圳技术大学 | Battery life prediction method, system, electronic device and storage medium |
CN113057586A (en) * | 2021-03-17 | 2021-07-02 | 上海电气集团股份有限公司 | Disease early warning method, device, equipment and medium |
CN113093020A (en) * | 2021-04-02 | 2021-07-09 | 中国矿业大学 | Method for predicting remaining service life of lithium ion battery based on LSTM neural network |
CN113406525A (en) * | 2021-06-15 | 2021-09-17 | 安庆师范大学 | Lithium battery pack residual life prediction method based on optimized variational modal decomposition |
CN113569485A (en) * | 2021-07-30 | 2021-10-29 | 湖北工业大学 | Short-time abnormal battery identification method based on variational modal decomposition |
CN113850016A (en) * | 2021-08-16 | 2021-12-28 | 国网江苏省电力有限公司技能培训中心 | Simulation transformer substation storage battery residual life prediction method in intermittent working mode |
CN113884937A (en) * | 2021-11-22 | 2022-01-04 | 江南大学 | Lithium ion battery residual life prediction method and system based on decomposition integration strategy |
CN114282571A (en) * | 2021-10-27 | 2022-04-05 | 河北建投能源投资股份有限公司 | Method, system, equipment and medium for constructing multidimensional health index of bearing |
CN114646891A (en) * | 2022-03-10 | 2022-06-21 | 电子科技大学 | Residual life prediction method combining LSTM network and wiener process |
CN114936682A (en) * | 2022-05-09 | 2022-08-23 | 重庆大学 | Method for predicting remaining service life of lithium ion battery based on variational modal decomposition |
CN115017945A (en) * | 2022-05-24 | 2022-09-06 | 南京林业大学 | Mechanical fault diagnosis method and system based on enhanced convolutional neural network |
CN115508711A (en) * | 2022-10-10 | 2022-12-23 | 南京工程学院 | Electric vehicle power lithium battery residual life prediction method based on deep network model |
CN116774058A (en) * | 2023-08-18 | 2023-09-19 | 深圳凌奈智控有限公司 | Battery life prediction method, device, equipment and storage medium |
WO2024066055A1 (en) * | 2022-09-30 | 2024-04-04 | 深圳先进技术研究院 | Service life prediction and state estimation method for bidirectional lithium ion battery |
-
2020
- 2020-11-04 CN CN202011216878.0A patent/CN112348185A/en not_active Withdrawn
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113057586A (en) * | 2021-03-17 | 2021-07-02 | 上海电气集团股份有限公司 | Disease early warning method, device, equipment and medium |
CN113057586B (en) * | 2021-03-17 | 2024-03-12 | 上海电气集团股份有限公司 | Disease early warning method, device, equipment and medium |
CN113065283A (en) * | 2021-03-26 | 2021-07-02 | 深圳技术大学 | Battery life prediction method, system, electronic device and storage medium |
CN113093020B (en) * | 2021-04-02 | 2022-07-12 | 中国矿业大学 | Method for predicting remaining service life of lithium ion battery based on LSTM neural network |
CN113093020A (en) * | 2021-04-02 | 2021-07-09 | 中国矿业大学 | Method for predicting remaining service life of lithium ion battery based on LSTM neural network |
CN113406525A (en) * | 2021-06-15 | 2021-09-17 | 安庆师范大学 | Lithium battery pack residual life prediction method based on optimized variational modal decomposition |
CN113406525B (en) * | 2021-06-15 | 2023-06-02 | 金陵科技学院 | Lithium battery pack residual life prediction method based on optimization variation modal decomposition |
CN113569485A (en) * | 2021-07-30 | 2021-10-29 | 湖北工业大学 | Short-time abnormal battery identification method based on variational modal decomposition |
CN113569485B (en) * | 2021-07-30 | 2024-03-08 | 湖北工业大学 | Short-time abnormal battery identification method based on variation modal decomposition |
CN113850016A (en) * | 2021-08-16 | 2021-12-28 | 国网江苏省电力有限公司技能培训中心 | Simulation transformer substation storage battery residual life prediction method in intermittent working mode |
CN113850016B (en) * | 2021-08-16 | 2024-04-05 | 国网江苏省电力有限公司技能培训中心 | Method for predicting residual life of storage battery of simulated transformer substation in intermittent working mode |
CN114282571A (en) * | 2021-10-27 | 2022-04-05 | 河北建投能源投资股份有限公司 | Method, system, equipment and medium for constructing multidimensional health index of bearing |
CN113884937A (en) * | 2021-11-22 | 2022-01-04 | 江南大学 | Lithium ion battery residual life prediction method and system based on decomposition integration strategy |
CN114646891A (en) * | 2022-03-10 | 2022-06-21 | 电子科技大学 | Residual life prediction method combining LSTM network and wiener process |
CN114646891B (en) * | 2022-03-10 | 2023-05-30 | 电子科技大学 | Residual life prediction method combining LSTM network and wiener process |
CN114936682A (en) * | 2022-05-09 | 2022-08-23 | 重庆大学 | Method for predicting remaining service life of lithium ion battery based on variational modal decomposition |
CN114936682B (en) * | 2022-05-09 | 2024-05-17 | 重庆大学 | Lithium ion battery residual service life prediction method based on variation modal decomposition |
CN115017945A (en) * | 2022-05-24 | 2022-09-06 | 南京林业大学 | Mechanical fault diagnosis method and system based on enhanced convolutional neural network |
WO2024066055A1 (en) * | 2022-09-30 | 2024-04-04 | 深圳先进技术研究院 | Service life prediction and state estimation method for bidirectional lithium ion battery |
CN115508711B (en) * | 2022-10-10 | 2023-06-30 | 南京工程学院 | Electric vehicle power lithium battery residual life prediction method based on deep network model |
CN115508711A (en) * | 2022-10-10 | 2022-12-23 | 南京工程学院 | Electric vehicle power lithium battery residual life prediction method based on deep network model |
CN116774058A (en) * | 2023-08-18 | 2023-09-19 | 深圳凌奈智控有限公司 | Battery life prediction method, device, equipment and storage medium |
CN116774058B (en) * | 2023-08-18 | 2023-10-20 | 深圳凌奈智控有限公司 | Battery life prediction method, device, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112348185A (en) | Lithium battery residual life prediction method based on variational modal decomposition and integrated depth model | |
Huang et al. | Convolutional gated recurrent unit–recurrent neural network for state-of-charge estimation of lithium-ion batteries | |
Marino et al. | Building energy load forecasting using deep neural networks | |
US9256823B2 (en) | Apparatus and methods for efficient updates in spiking neuron network | |
CN110321603A (en) | A kind of depth calculation model for Fault Diagnosis of Aircraft Engine Gas Path | |
CN111832825B (en) | Wind power prediction method and system integrating long-term memory network and extreme learning machine | |
CN113156325A (en) | Method for estimating state of health of battery | |
CN113589189B (en) | Lithium battery health condition prediction method and device based on charging and discharging data characteristics | |
CN116316591A (en) | Short-term photovoltaic power prediction method and system based on hybrid bidirectional gating cycle | |
CN110750852B (en) | Method and device for predicting residual service life of supercapacitor and electronic equipment | |
CN111784061B (en) | Training method, device and equipment for power grid engineering cost prediction model | |
Ge et al. | Lithium ion battery health prediction via variable mode decomposition and deep learning network with self-attention mechanism | |
CN110458341B (en) | Ultra-short-term wind power prediction method and system considering meteorological characteristics | |
CN114660497A (en) | Lithium ion battery service life prediction method aiming at capacity regeneration phenomenon | |
CN116500449A (en) | Cell temperature prediction method and device, storage medium and electronic equipment | |
CN114578234A (en) | Lithium ion battery degradation and capacity prediction model considering causality characteristics | |
Lu et al. | Health state prediction of lithium ion battery based on deep learning method | |
Guo et al. | A machine learning method for prediction of remaining useful life of supercapacitors with multi-stage modification | |
CN116125279A (en) | Method, device, equipment and storage medium for determining battery health state | |
CN112685900B (en) | Power load simulation method for representing impact load power characteristics | |
CN116050601A (en) | Load prediction method based on variational modal decomposition and long-term and short-term memory mixed model | |
CN117473864A (en) | Lithium battery life prediction method based on multi-scale integrated deep learning model | |
Brink et al. | Efficient Random Vibration Analysis of Nonlinear Systems with Long Short-Term Memory Networks for Uncertainty Quantification. | |
Wu et al. | Remaining Useful Life Prediction of Lithium‐Ion Batteries Based on a Combination of Ensemble Empirical Mode Decomposition and Deep Belief Network–Long Short‐Term Memory | |
Zheng et al. | Wind Electricity Power Prediction Based on CNN-LSTM Network Model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210209 |