CN116562008A - EMD-LSTM-based lithium battery temperature field space-time modeling method - Google Patents
EMD-LSTM-based lithium battery temperature field space-time modeling method Download PDFInfo
- Publication number
- CN116562008A CN116562008A CN202310494038.8A CN202310494038A CN116562008A CN 116562008 A CN116562008 A CN 116562008A CN 202310494038 A CN202310494038 A CN 202310494038A CN 116562008 A CN116562008 A CN 116562008A
- Authority
- CN
- China
- Prior art keywords
- space
- time
- vector
- emd
- signal sequence
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 title claims abstract description 31
- 229910052744 lithium Inorganic materials 0.000 title claims abstract description 31
- 230000009467 reduction Effects 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 6
- 108010076504 Protein Sorting Signals Proteins 0.000 claims description 39
- 238000004364 calculation method Methods 0.000 claims description 33
- 239000011159 matrix material Substances 0.000 claims description 33
- 230000008569 process Effects 0.000 claims description 13
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 6
- 229910001416 lithium ion Inorganic materials 0.000 claims description 6
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 2
- 238000004880 explosion Methods 0.000 abstract description 6
- 230000006870 function Effects 0.000 description 36
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 230000015654 memory Effects 0.000 description 3
- 230000008033 biological extinction Effects 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- 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
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- Y—GENERAL 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/10—Energy storage using batteries
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The invention discloses a lithium battery temperature field space-time modeling method based on EMD-LSTM, which comprises the steps of converting high-dimensional space-time data into low-dimensional space-time data by using a KWLTSA dimension reduction method, and obtaining a space basis function; projecting the low-dimensional space-time data to a space basis function to obtain a corresponding low-order time coefficient; decomposing the low-order time coefficient by using an EMD algorithm to obtain an IMF component and a residual error component; respectively performing LSTM training on the IMF component and the residual component to obtain a prediction time coefficient; and carrying out time-space reconstruction on the predicted time coefficient and the space basis function to obtain a lithium battery temperature field time-space modeling model. The invention solves the problems that the traditional space-time modeling method is difficult to be applied to a distributed parameter system with strong nonlinearity and instability, and gradient vanishing and gradient explosion in a long-time sequence.
Description
Technical Field
The invention relates to the technical field of space-time modeling of a distributed parameter system, in particular to a space-time modeling method of a lithium battery temperature field based on EMD-LSTM.
Background
In recent years, many space-time modeling methods have been studied, and among them, space-time separation-based modeling methods have been widely used and proved to be viable and effective Distributed Parameter System (DPSs) modeling methods. The main idea of the spatio-temporal separation method is that the spatio-temporal variable can be decomposed into a series of spatial basis functions and corresponding time coefficients based on fourier transforms. Thus, spatial basis functions and temporal coefficients may be learned separately using a mechanism-based or data-based algorithm.
The traditional space-time modeling method has satisfactory modeling performance for industrial thermal processes, but the traditional space-time modeling method still has the problems that the traditional space-time modeling method is difficult to be applied to a distributed parameter system with strong nonlinearity and instability, and gradient vanishes and gradient explosions in a long-time sequence.
Disclosure of Invention
Aiming at the defects, the invention provides a lithium battery temperature field space-time modeling method based on EMD-LSTM, which aims to solve the problems that the traditional space-time modeling method is difficult to be applied to a distributed parameter system with strong nonlinearity and instability, and gradient vanishes and gradient explodes in a long-time sequence.
To achieve the purpose, the invention adopts the following technical scheme:
a lithium battery temperature field space-time modeling method based on EMD-LSTM comprises the following steps:
step S1: converting the high-dimensional space-time data into low-dimensional space-time data by using a KWLTSA dimension reduction method, and obtaining a space basis function;
step S2: projecting the low-dimensional space-time data to a space basis function to obtain a corresponding low-order time coefficient;
step S3: decomposing the low-order time coefficient by using an EMD algorithm to obtain an IMF component and a residual error component;
step S4: respectively performing LSTM training on the IMF component and the residual component to obtain a prediction time coefficient;
step S5: and carrying out time-space reconstruction on the predicted time coefficient and the space basis function to obtain a lithium battery temperature field time-space modeling model.
Preferably, in step S1, the KWLTSA dimension reduction method specifically includes the following steps:
step S11: constructing a local neighborhood;
step S12: fitting local coordinates;
step S13: and globally arranging the local coordinates.
Preferably, in step S3, the EMD algorithm specifically comprises the following sub-steps:
step S31: input of the original signal sequence a i (t) calculating a i All maximum points and minimum points of (t), and interpolating all maximum points by interpolation to construct an upper envelope a imax (t) and interpolating all minima points to construct the lower envelope a imin (t);
Step S32: solving an upper envelope curve a imax (t) and lower envelope a imin Mean curve m of (t) i (t) the specific calculation formula is as follows: m is m i (t)=[a imax (t)+a imin (t)]/2;
Step S33: sequence a of original signals i (t) and m i (t) taking the difference to obtain a new signal sequence h i (t) the specific calculation formula is as follows: h is a i (t)=a i (t)-m i (t);
Step S34: determining a new signal sequence h i (t) whether the following two conditions are satisfied:
the first condition is in a new signal sequence h i In (t), the number of added upper and lower extreme points in the new signal sequence is equal to or at most one different from the number of curves passing through the x-axis;
the second condition is for a new signal sequence h i (t) the mean of the upper and lower envelopes must be equal to zero;
if both conditions are met, a new signal sequence h i (t) assignment is the first IMF component, l 1 ;
Step S35: according to formula a i2 (t)=h i (t)-l 1 Obtaining a second decomposed signal sequence a i2 (t), and for a i2 (t) looping steps S31 to S34 to generate a second IMF component l 2 ;
Step S36: for the second IMF component l 2 Step S31 to step S35 are looped to generate a third IMF component l 3 Determine the third IMF component l 3 Whether or not to meet a threshold valueScreening conditions, if satisfied, according to the third IMF component l 3 The resulting signal sequence is taken as the residual term r (t) of the original signal sequence.
Preferably, in step S3, a threshold function SD is used k Stopping EMD algorithm process, threshold function SD k The following are provided:
wherein k is an IMF decomposition number, and k is a positive integer greater than 1; t is the period number, t=1, 2, …, n, n is the sample volume; h is a k (t) is represented by a new signal sequence; h is a k-1 (t) the old signal sequence before decomposition; SD (secure digital memory card) k In the range of 0.2-0.3.
Preferably, in step S4, the following substeps are specifically included:
step S41: component of IMF at t-th stage l i (t) and t-1 th stage hidden layer output vector h t-1 After entering a forgetting gate, multiplying the forgetting gate by the weight matrix of each, adding the forgetting gate bias vector, and activating by a sigmoid function to obtain a forgetting gate signal vector f in the t period t The specific calculation formula is as follows:
f t =σ(W FH h t-1 +W FI l i (t)+b F )
wherein ,a hidden layer-to-forget gate weight matrix; />A left gate weight matrix is input for the layer; />A forgetting door offset vector; sigma is a sigmoid function; n (N) h The number of the output signals at the previous moment; n (N) i The number of IMF components;
step S42: calculation to decide what information to keepWeight parameter i of (2) t The specific calculation formula is as follows:
i t =σ(W GH h t-1 +W GI l i (t)+b G )
wherein ,a hidden layer to input gate weight matrix; />Inputting a gate weight matrix for an input layer; />A bias vector is input to the gate;
step S43: calculation of t-phase candidate cell State Signal vectorThe specific calculation formula is as follows:
wherein ,updating the cell state weight matrix for the hidden layer to the input gate; />Updating the cell state weight matrix for the input layer; />A bias vector is input to the gate; tanh is a hyperbolic tangent function;
step S44: calculation of t-phase cell State vector C t T-stage cell state vector C t The specific calculation formula for storing the history information is as follows:
wherein ,Ct Is the t-stage cell state vector;the sign is Hadamard product of the matrix; f (f) t A forgetting gate signal vector of the t-th period; c (C) t-1 Is the t-1 stage cell state vector; i.e t Weight parameters reserved for deciding what information; />Is a t-phase candidate cell state signal vector;
step S45: component of IMF at t-th stage l i (t) and t-1 th stage hidden layer output vector h t-1 Activated by the output gate activation function to generate output information o t The specific calculation formula is as follows:
o t =σ(W OH h t-1 +W OI l i (t)+b O )
wherein ,WOH A weight matrix for outputting information; w (W) OI A weight matrix for IMF component information; v O The weight of the paranoid at this moment;
step S46: will output information o t Vector C of cell state at t stage activated by hidden layer activation function t Hadamard product operation is carried out to obtain a t-stage hidden layer output vector h t The specific calculation formula is as follows:
wherein ,the sign is Hadamard product of the matrix; tanh is a hyperbolic tangent function.
Preferably, in step S5, the space-time reconstruction specific calculation formula is as follows:
wherein ,representing the predicted temperature of the lithium battery in the time dimension and the space dimension; t is t k Representing the boundary condition of the lithium ion battery at the moment k; s represents the space coordinates of the battery; />Representing a spatial basis function; />Representing the predicted time coefficient.
The technical scheme provided by the embodiment of the application can comprise the following beneficial effects:
according to the scheme, a space-time modeling model of the lithium battery temperature field is obtained according to an EMD algorithm and an LSTM. The EMD-LSTM can better process the distributed parameter system with strong nonlinearity and instability, has better robustness to the distributed parameter system with strong nonlinearity and instability, and the obtained lithium battery temperature field space-time modeling model has high precision. The solution also solves the problems of gradient extinction and gradient explosion in long time sequences by means of LSTM.
Drawings
FIG. 1 is a flow chart of the steps of a method for modeling the temperature field of a lithium battery based on EMD-LSTM.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are exemplary only for explaining the present invention and are not to be construed as limiting the present invention.
A lithium battery temperature field space-time modeling method based on EMD-LSTM is characterized in that: the method comprises the following steps:
step S1: converting the high-dimensional space-time data into low-dimensional space-time data by using a KWLTSA dimension reduction method, and obtaining a space basis function;
step S2: projecting the low-dimensional space-time data to a space basis function to obtain a corresponding low-order time coefficient;
step S3: decomposing the low-order time coefficient by using an EMD algorithm to obtain an IMF component and a residual error component;
step S4: respectively performing LSTM training on the IMF component and the residual component to obtain a prediction time coefficient;
step S5: and carrying out time-space reconstruction on the predicted time coefficient and the space basis function to obtain a lithium battery temperature field time-space modeling model.
According to the lithium battery temperature field space-time modeling method based on the EMD-LSTM, as shown in fig. 1, the first step is to convert high-dimensional space-time data into low-dimensional space-time data by using a KWAT SA dimension reduction method and obtain a space basis function, and particularly, as the temperature field space-time distribution in the lithium ion battery is space-time coupling, the temperature field space-time distribution is expanded in infinite dimensions, so that the temperature field space-time distribution is difficult to calculate. The spatio-temporal variables are then decomposed into a series of spatial basis functions according to a fourier transform. The second step is to project the low-dimensional space-time data to the space basis function to obtain corresponding low-order time coefficients, and specifically, the space-time variables are decomposed into the space basis function according to Fourier transformation, and the corresponding low-order time coefficients are also obtained. The third step is to decompose the low-order time coefficient by using EMD algorithm to obtain IMF component and residual error component, in particular, because partial abnormal value may exist in the working process of the lithium ion battery and data has instability, the scheme adopts EMD noise reduction method to process nonstationary data, and obtains relatively stable solid state mode function and a residual error item by decomposing data, the EMD noise reduction method has better self-adaptability than the traditional noise reduction method, no need to select substrate or set effective decomposition layer number, greatly reduces complexity of algorithm and obtains wide application. And fourthly, respectively performing LSTM training on the IMF component and the residual component to obtain a predicted time coefficient, wherein the LSTM generally refers to a long-term and short-term memory artificial neural network, which is a time-cyclic neural network and can solve the problems of gradient disappearance and gradient explosion in a long-time sequence. The low-order time coefficient obtained through LSTM training can improve the precision of the predictive modeling model. And fifthly, performing time-space reconstruction on the predicted time coefficient and the space basis function to obtain a lithium battery temperature field time-space modeling model, and specifically, obtaining the thermodynamic characteristics of the lithium ion battery through the lithium battery temperature field time-space modeling model.
According to the scheme, a space-time modeling model of the lithium battery temperature field is obtained according to an EMD algorithm and an LSTM. The EMD-LSTM can better process the distributed parameter system with strong nonlinearity and instability, has better robustness to the distributed parameter system with strong nonlinearity and instability, and the obtained lithium battery temperature field space-time modeling model has high precision. The solution also solves the problems of gradient extinction and gradient explosion in long time sequences by means of LSTM.
Preferably, in step S1, the KWLTSA dimension reduction method specifically includes the following steps:
step S11: constructing a local neighborhood;
step S12: fitting local coordinates;
step S13: and globally arranging the local coordinates.
In this embodiment, by performing local domain construction and fitting local coordinates according to the local domain construction, global arrangement is performed on the local coordinates, so that accurate local coordinates can be obtained, and positioning of a lithium battery temperature field space-time modeling model is facilitated.
Preferably, in step S3, the EMD algorithm specifically includes the following sub-steps:
step S31: input of the original signal sequence a i (t) calculating a i All maximum points and minimum points of (t), and interpolating all maximum points by interpolation to construct an upper envelope a imax (t) and interpolating all minima points to construct the lower envelope a imin (t);
Step S32: solving an upper envelope curve a imax (t) and lower envelope a imin Mean curve m of (t) i (t) the specific calculation formula is as follows: m is m i (t)=[a imax (t)+a imin (t)]/2;
Step S33: sequence a of original signals i (t) and m i (t) taking the difference to obtain a new signal sequence h i (t) the specific calculation formula is as follows: h is a i (t)=a i (t)-m i (t);
Step S34: determining a new signal sequence h i (t) whether the following two conditions are satisfied:
the first condition is in a new signal sequence h i In (t), the number of added upper and lower extreme points in the new signal sequence is equal to or at most one different from the number of curves passing through the x-axis;
the second condition is for a new signal sequence h i (t) the mean of the upper and lower envelopes must be equal to zero;
if both conditions are met, a new signal sequence h i (t) assignment is the first IMF component, l 1 ;
Step S35: according to formula a i2 (t)=h i (t)-l 1 Obtaining a second decomposed signal sequence a i2 (t), and for a i2 (t) looping steps S31 to S34 to generate a second IMF component l 2 ;
Step S36: for the second IMF component l 2 Step S31 to step S35 are looped to generate a third IMF component l 3 Determine the third IMF component l 3 Whether a threshold screening condition is satisfied, if so, according to a third IMF component l 3 The resulting signal sequence is taken as the residual term r (t) of the original signal sequence.
In this embodiment, the EMD algorithm is used to determine the low-order time coefficients, i.e., the original signal sequenceColumn a i (t) decomposing to obtain IMF component and residual component, i.e. new signal sequence h i (t) and a residual term r (t) of the original signal sequence.
The specific decomposition formula is as follows:
wherein m is the final IMF decomposition number; a, a i (t) represents the low-order time coefficient of the t-th period, l i (t) represents the IMF component at the t-th stage; r (t) represents the residual term of the original signal sequence. And finally, carrying out noise reduction treatment on the high-frequency part of the time sequence by adopting a threshold noise reduction method.
Preferably, in step S3, a threshold function SD is used k Stopping EMD algorithm process, threshold function SD k The following are provided:
wherein k is an IMF decomposition number, and k is a positive integer greater than 1; t is the period number, t=1, 2, …, n, n is the sample volume; h is a k (t) is represented by a new signal sequence; h is a k-1 (t) the old signal sequence before decomposition; SD (secure digital memory card) k In the range of 0.2-0.3.
Specifically, since the upper and lower envelopes of the IMF component are almost piled up based on the x-axis, this process is a process of eliminating data distortion in practice. Since the upper envelope and the lower envelope are averaged, and some information is actually lost in the averaging process, stopping the EMD process requires determining whether the difference between two adjacent IMF components is less than a given value. In general, a threshold function SD is used k As a function of a threshold to stop the EMD process.
Preferably, in step S4, the method specifically comprises the following substeps:
step S41: component of IMF at t-th stage l i (t) and t-1 th stage hidden layer output vector h t-1 After entering the forgetting gate, multiply by eachAdding the weight matrix and the forgetting gate bias vector, and activating by using sigmoid function to obtain forgetting gate signal vector f of the t period t The specific calculation formula is as follows:
f t =σ(W FH h t-1 +W FI l i (t)+b F )
wherein ,a hidden layer-to-forget gate weight matrix; />A left gate weight matrix is input for the layer; />A forgetting door offset vector; sigma is a sigmoid function; n (N) h The number of the output signals at the previous moment; n (N) i The number of IMF components;
step S42: calculating weight parameter i for deciding what information is reserved t The specific calculation formula is as follows:
i t =σ(W GH h t-1 +W GI l i (t)+b G )
wherein ,a hidden layer to input gate weight matrix; />Inputting a gate weight matrix for an input layer; />A bias vector is input to the gate;
step S43: calculation of t-phase candidate cell State Signal vectorThe specific calculation formula is as follows:
wherein ,updating the cell state weight matrix for the hidden layer to the input gate; />Updating the cell state weight matrix for the input layer; />A bias vector is input to the gate; tanh is a hyperbolic tangent function;
step S44: calculation of t-phase cell State vector C t T-stage cell state vector C t The specific calculation formula for storing the history information is as follows:
wherein ,Ct Is the t-stage cell state vector;the sign is Hadamard product of the matrix; f (f) t A forgetting gate signal vector of the t-th period; c (C) t-1 Is the t-1 stage cell state vector; i.e t Weight parameters reserved for deciding what information; />Is a t-phase candidate cell state signal vector;
step S45: component of IMF at t-th stage l i (t) and t-1 th stage hidden layer output vector h t-1 Activated by the output gate activation function to generate output information o t The specific calculation formula is as follows:
o t =σ(W OH h t-1 +W OI l i (t)+b O )
wherein ,WOH A weight matrix for outputting information; w (W) OI A weight matrix for IMF component information; b O The weight of the paranoid at this moment;
step S46: will output information o t Vector C of cell state at t stage activated by hidden layer activation function t Hadamard product operation is carried out to obtain a t-stage hidden layer output vector h t The specific calculation formula is as follows:
wherein ,the sign is Hadamard product of the matrix; tanh is a hyperbolic tangent function.
In this embodiment, the LSTM is used to train the IMF component and the residual component to obtain the t-stage hidden layer output vector h t I.e. network output. LSTM is a Recurrent Neural Network (RNN) variant generated by a gating cell system consisting of an input gate, a forget gate and an output gate. The gate is a structure capable of selectively controlling information, and is realized by a sigmoid layer and a point-by-point multiplication operation, wherein the output value is between 0 and 1, 0 indicates that the information is completely not passed, and 1 indicates that the information is completely passed. The LSTM uses the internal memory unit, namely the state of the cell, to save the historical information, and uses different gates to make the network forget the historical information timely, and updates the cell state according to the new input, so as to solve the problems of gradient disappearance and gradient explosion of the cyclic neural network (RNN).
Further described, the t-th IMF component l i (t) after three gating networks, a portion of the information is retained in the t-stage cell state vector C t In the method, namely network memory, the problem that the RNN cannot learn data relations for a long time is solved through the network memory, and the gating structure of the LSTM also relieves the short-term memory problem of the RNN model to a certain extent.
Preferably, in step S5, the space-time reconstruction is specifically calculated as follows:
wherein ,representing the predicted temperature of the lithium battery in the time dimension and the space dimension; t is t k Representing the boundary condition of the lithium ion battery at the moment k; s represents the space coordinates of the battery; />Representing a spatial basis function; />Representing the predicted time coefficient.
In the embodiment, the space-time reconstruction is performed on the space basis function and the prediction time system, so that the global prediction lithium battery temperature field space-time distribution can be obtained, and the construction of a lithium battery temperature field space-time modeling model is facilitated.
Furthermore, functional units in various embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units may be integrated into one module. The integrated modules may be implemented in hardware or in software functional modules. The integrated modules may also be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations of the above embodiments may be made by those skilled in the art within the scope of the invention.
Claims (6)
1. A lithium battery temperature field space-time modeling method based on EMD-LSTM is characterized in that: the method comprises the following steps:
step S1: converting the high-dimensional space-time data into low-dimensional space-time data by using a KWLTSA dimension reduction method, and obtaining a space basis function;
step S2: projecting the low-dimensional space-time data to a space basis function to obtain a corresponding low-order time coefficient;
step S3: decomposing the low-order time coefficient by using an EMD algorithm to obtain an IMF component and a residual error component;
step S4: respectively performing LSTM training on the IMF component and the residual component to obtain a prediction time coefficient;
step S5: and carrying out time-space reconstruction on the predicted time coefficient and the space basis function to obtain a lithium battery temperature field time-space modeling model.
2. The lithium battery temperature field space-time modeling method based on EMD-LSTM as defined in claim 1, wherein the method comprises the following steps: in step S1, the KWLTSA dimension reduction method specifically includes the following steps:
step S11: constructing a local neighborhood;
step S12: fitting local coordinates;
step S13: and globally arranging the local coordinates.
3. The lithium battery temperature field space-time modeling method based on EMD-LSTM as defined in claim 1, wherein the method comprises the following steps: in step S3, the EMD algorithm specifically includes the following sub-steps:
step S31: input of the original signal sequence a i (t) calculating a i All maximum points and minimum points of (t), and interpolating all maximum points by interpolation to construct an upper envelope a imax (t) and interpolating all minima points to construct the lower envelope a imin (t);
Step S32: solving an upper envelope curve a imax (t) and lower envelope a imin Mean curve m of (t) i (t),The specific calculation formula is as follows: m is m i (t)=[a imax (t)+a imin (t)]/2;
Step S33: sequence a of original signals i (t) and m i (t) taking the difference to obtain a new signal sequence h i (t) the specific calculation formula is as follows: h is a i (t)=a i (t)-m i (t);
Step S34: determining a new signal sequence h i (t) whether the following two conditions are satisfied:
the first condition is in a new signal sequence h i In (t), the number of added upper and lower extreme points in the new signal sequence is equal to or at most one different from the number of curves passing through the x-axis;
the second condition is for a new signal sequence h i (t) the mean of the upper and lower envelopes must be equal to zero;
if both conditions are met, a new signal sequence h i (t) assignment is the first IMF component, l 1 ;
Step S35: according to formula a i2 (t)=h i (t)-l 1 Obtaining a second decomposed signal sequence a i2 (t), and for a i2 (t) looping steps S31 to S34 to generate a second IMF component l 2 ;
Step S36: for the second IMF component l 2 Step S31 to step S35 are looped to generate a third IMF component l 3 Determine the third IMF component l 3 Whether a threshold screening condition is satisfied, if so, according to a third IMF component l 3 The resulting signal sequence is taken as the residual term r (t) of the original signal sequence.
4. The lithium battery temperature field space-time modeling method based on EMD-LSTM as defined in claim 1, wherein the method comprises the following steps: in step S3, a threshold function SD is used k Stopping EMD algorithm process, threshold function SD k The following are provided:
wherein k is an IMF decomposition number, and k is a positive integer greater than 1; t is the period number, t=1, 2, …, n, n is the sample volume; h is a k (t) is represented by a new signal sequence; h is a k-1 (t) the old signal sequence before decomposition; SD (secure digital memory card) k In the range of 0.2-0.3.
5. The lithium battery temperature field space-time modeling method based on EMD-LSTM as defined in claim 1, wherein the method comprises the following steps: in step S4, the method specifically includes the following substeps:
step S41: component of IMF at t-th stage l i (t) and t-1 th stage hidden layer output vector h t-1 After entering a forgetting gate, multiplying the forgetting gate by the weight matrix of each, adding the forgetting gate bias vector, and activating by a sigmoid function to obtain a forgetting gate signal vector f in the t period t The specific calculation formula is as follows:
f t =σ(W FH h t-1 +W FI l i (t)+b F )
wherein ,a hidden layer-to-forget gate weight matrix; />A left gate weight matrix is input for the layer; />A forgetting door offset vector; sigma is a sigmoid function; n (N) h The number of the output signals at the previous moment; n (N) i The number of IMF components;
step S42: calculating weight parameter i for deciding what information is reserved t The specific calculation formula is as follows:
i t =σ(W GH h t-1 +W GI l i (t)+b G )
wherein ,a hidden layer to input gate weight matrix; />Inputting a gate weight matrix for an input layer; />A bias vector is input to the gate;
step S43: calculation of t-phase candidate cell State Signal vectorThe specific calculation formula is as follows:
wherein ,updating the cell state weight matrix for the hidden layer to the input gate; />Updating the cell state weight matrix for the input layer; />A bias vector is input to the gate; tanh is a hyperbolic tangent function;
step S44: calculation of t-phase cell State vector C t T-stage cell state vector C t The specific calculation formula for storing the history information is as follows:
wherein ,Ct Is the t-stage cell state vector;the sign is Hadamard product of the matrix; f (f) t A forgetting gate signal vector of the t-th period; c (C) t-1 Is the t-1 stage cell state vector; i.e t Weight parameters reserved for deciding what information; />Is a t-phase candidate cell state signal vector;
step S45: component of IMF at t-th stage l i (t) and t-1 th stage hidden layer output vector h t-1 Activated by the output gate activation function to generate output information o t The specific calculation formula is as follows:
o t =σ(W OH h t-1 +W OI l i (t)+b O )
wherein ,WOH A weight matrix for outputting information; w (W) OI A weight matrix for IMF component information; b O Is the bias weight at this time;
step S46: will output information o t Vector C of cell state at t stage activated by hidden layer activation function t Hadamard product operation is carried out to obtain a t-stage hidden layer output vector h t The specific calculation formula is as follows:
wherein ,the sign is Hadamard product of the matrix; tanh is a hyperbolic tangent function.
6. The lithium battery temperature field space-time modeling method based on EMD-LSTM as defined in claim 1, wherein the method comprises the following steps: in step S5, the space-time reconstruction specific calculation formula is as follows:
wherein ,representing the predicted temperature of the lithium battery in the time dimension and the space dimension; t is t k Representing the boundary condition of the lithium ion battery at the moment k; s represents the space coordinates of the battery; />Representing a spatial basis function; />Representing the predicted time coefficient.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310494038.8A CN116562008B (en) | 2023-05-04 | 2023-05-04 | EMD-LSTM-based lithium battery temperature field space-time modeling method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310494038.8A CN116562008B (en) | 2023-05-04 | 2023-05-04 | EMD-LSTM-based lithium battery temperature field space-time modeling method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116562008A true CN116562008A (en) | 2023-08-08 |
CN116562008B CN116562008B (en) | 2024-07-30 |
Family
ID=87494030
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310494038.8A Active CN116562008B (en) | 2023-05-04 | 2023-05-04 | EMD-LSTM-based lithium battery temperature field space-time modeling method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116562008B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111240268A (en) * | 2020-01-14 | 2020-06-05 | 重庆大学 | Axle system thermal error modeling method and thermal error compensation system based on SLSTM neural network |
CN112966432A (en) * | 2021-02-09 | 2021-06-15 | 东北电力大学 | Method and device for predicting remaining effective life of lithium ion battery |
CN113344475A (en) * | 2021-08-05 | 2021-09-03 | 国网江西省电力有限公司电力科学研究院 | Transformer bushing defect identification method and system based on sequence modal decomposition |
US20220045509A1 (en) * | 2020-08-05 | 2022-02-10 | Wuhan University | Method and system of predicting electric system load based on wavelet noise reduction and emd-arima |
CN114692494A (en) * | 2022-03-17 | 2022-07-01 | 广东工业大学 | Lithium battery temperature field online modeling method and system |
CN115392029A (en) * | 2022-08-29 | 2022-11-25 | 广东工业大学 | Lithium battery thermal process space-time modeling method based on KWLSAT |
-
2023
- 2023-05-04 CN CN202310494038.8A patent/CN116562008B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111240268A (en) * | 2020-01-14 | 2020-06-05 | 重庆大学 | Axle system thermal error modeling method and thermal error compensation system based on SLSTM neural network |
US20220045509A1 (en) * | 2020-08-05 | 2022-02-10 | Wuhan University | Method and system of predicting electric system load based on wavelet noise reduction and emd-arima |
CN112966432A (en) * | 2021-02-09 | 2021-06-15 | 东北电力大学 | Method and device for predicting remaining effective life of lithium ion battery |
CN113344475A (en) * | 2021-08-05 | 2021-09-03 | 国网江西省电力有限公司电力科学研究院 | Transformer bushing defect identification method and system based on sequence modal decomposition |
CN114692494A (en) * | 2022-03-17 | 2022-07-01 | 广东工业大学 | Lithium battery temperature field online modeling method and system |
CN115392029A (en) * | 2022-08-29 | 2022-11-25 | 广东工业大学 | Lithium battery thermal process space-time modeling method based on KWLSAT |
Non-Patent Citations (2)
Title |
---|
戴邵武 等: "基于EMD-LSTM的时间序列预测方法", 深圳大学学报(理工版), vol. 37, no. 3, pages 49 - 54 * |
朱贵重;胡松;: "基于LSTM-RNN的海水表面温度模型研究", 应用海洋学学报, no. 02 * |
Also Published As
Publication number | Publication date |
---|---|
CN116562008B (en) | 2024-07-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Charles et al. | Dynamic Filtering of Time-Varying Sparse Signals via $\ell _1 $ Minimization | |
CN110245801A (en) | A kind of Methods of electric load forecasting and system based on combination mining model | |
CN107102969A (en) | The Forecasting Methodology and system of a kind of time series data | |
CN112966432A (en) | Method and device for predicting remaining effective life of lithium ion battery | |
CN109559329B (en) | Particle filter tracking method based on depth denoising automatic encoder | |
CN113392961A (en) | Method for extracting mesoscale eddy track stable sequence and predicting cyclic neural network | |
CN110570035B (en) | People flow prediction system for simultaneously modeling space-time dependency and daily flow dependency | |
CN116362398A (en) | Power load prediction method based on modal decomposition and reconstruction and LSTM-MLR hybrid model | |
CN116861201B (en) | Power supply system based on artificial intelligence | |
CN114004342B (en) | Laser communication system distortion wavefront prediction method based on LSTM network | |
CN111680786A (en) | Time sequence prediction method based on improved weight gating unit | |
CN111950759A (en) | Short-term wind speed prediction method based on two-stage decomposition, LSTM and AT | |
CN118295029B (en) | Seismic data denoising method integrating self-attention and Mamba architecture | |
Shen et al. | Quantized generalized maximum correntropy criterion based kernel recursive least squares for online time series prediction | |
CN114676645B (en) | Non-stationary time sequence prediction method and system | |
CN117218837A (en) | Traffic prediction method, system, equipment and medium based on dynamic space-time diagram convolutional neural network | |
CN116227180A (en) | Data-driven-based intelligent decision-making method for unit combination | |
CN116739118A (en) | Power load prediction method for realizing error correction mechanism based on LSTM-XGBoost | |
CN111523644A (en) | Intermediate-long term runoff prediction method based on LSFL combined model | |
CN117787470A (en) | Time sequence prediction method and system based on EWT and integration method | |
CN116562008B (en) | EMD-LSTM-based lithium battery temperature field space-time modeling method | |
CN115856425B (en) | Spectrum anomaly detection method and device based on hidden space probability prediction | |
CN114547276A (en) | Three-channel diagram neural network-based session recommendation method | |
Chau et al. | An efficient particle-based method for maximum likelihood estimation in nonlinear state-space models | |
CN113556130A (en) | Stable complex field multitask Bayes compressed sensing method |
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 | ||
CB03 | Change of inventor or designer information |
Inventor after: Yang Haidong Inventor after: Xu Kangkang Inventor after: Zhu Chengjiu Inventor after: Zhuang Jiawei Inventor before: Xu Kangkang Inventor before: Zhuang Jiawei |
|
CB03 | Change of inventor or designer information | ||
GR01 | Patent grant | ||
GR01 | Patent grant |