CN117706406A - Lithium battery health state monitoring model, method, system and storage medium - Google Patents
Lithium battery health state monitoring model, method, system and storage medium Download PDFInfo
- Publication number
- CN117706406A CN117706406A CN202410163216.3A CN202410163216A CN117706406A CN 117706406 A CN117706406 A CN 117706406A CN 202410163216 A CN202410163216 A CN 202410163216A CN 117706406 A CN117706406 A CN 117706406A
- Authority
- CN
- China
- Prior art keywords
- lithium battery
- output
- input
- layer
- time
- 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
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 title claims abstract description 128
- 229910052744 lithium Inorganic materials 0.000 title claims abstract description 128
- 230000036541 health Effects 0.000 title claims abstract description 58
- 238000012544 monitoring process Methods 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000004927 fusion Effects 0.000 claims abstract description 38
- 210000002569 neuron Anatomy 0.000 claims description 40
- 230000001537 neural effect Effects 0.000 claims description 21
- 230000000946 synaptic effect Effects 0.000 claims description 12
- 230000007704 transition Effects 0.000 claims description 12
- 238000005316 response function Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 8
- 230000003862 health status Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 101100182248 Caenorhabditis elegans lat-2 gene Proteins 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000007599 discharging Methods 0.000 claims description 6
- 230000005484 gravity Effects 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 4
- 210000000170 cell membrane Anatomy 0.000 claims description 3
- 210000005215 presynaptic neuron Anatomy 0.000 claims description 3
- 101100134058 Caenorhabditis elegans nth-1 gene Proteins 0.000 claims description 2
- 238000004146 energy storage Methods 0.000 abstract description 3
- 238000011156 evaluation Methods 0.000 abstract description 2
- 238000012549 training Methods 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000008859 change 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
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000003487 electrochemical reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Landscapes
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
- Secondary Cells (AREA)
Abstract
The invention relates to the technical field of performance evaluation of energy storage lithium batteries, in particular to a lithium battery health state monitoring model, a lithium battery health state monitoring method, a lithium battery health state monitoring system and a storage medium. The invention provides a method for constructing a lithium battery health state monitoring model, which constructs the lithium battery health state monitoring model by taking the fusion characteristic of a lithium battery as input. Therefore, in the invention, the working condition data of the lithium battery is preprocessed through feature fusion, so that the input of the lithium battery health state monitoring model is simplified, and the model prediction efficiency is improved. According to the invention, the characteristic fusion is used for comprehensively considering the working condition data of the lithium battery, so that the accuracy of the health state assessment is improved, and an efficient and accurate lithium battery health state monitoring model is realized.
Description
Technical Field
The invention relates to the technical field of performance evaluation of energy storage lithium batteries, in particular to a lithium battery state of health (SOH) monitoring model, a method, a system and a storage medium.
Background
The lithium battery has the advantages of high energy density, long service life and the like, and is widely applied to the fields of new energy automobiles, large-scale power grid energy storage stations, aerospace military industry and the like due to good performance. However, as the battery is continuously used and aged, the performance and health of the battery are deteriorated, and even safety accidents are caused in serious cases. Inaccurate estimation of battery state of health may lead to premature failure of the battery, reduce the battery's original life, and also reduce system performance. Therefore, accurate estimation of the SOH of the lithium battery has important significance for practical application of the lithium battery in the fields of industry and the like and performance of the battery.
Methods for lithium battery SOH estimation mainly include two main categories: model-based methods and data-driven methods.
Model-based methods describe the degradation behavior of a battery mainly by constructing an equivalent circuit model (Equivalent Circuit Model, ECM) and an electrochemical model. The ECM method predicts SOH of a battery by creating a simplified circuit model using electrochemical characteristics of the battery and actual test data of the battery. The ECM method can estimate SOH on line, but the error of parameter identification is increased continuously, and the prediction performance is greatly influenced by the model structure. The main idea of the electrochemical model approach is to describe the dynamic behavior of the cell using mathematical and physical equations and infer SOH of the cell by comparing the model with actual observed data. The electrochemical model method has high accuracy, but the parameter identification of the model is difficult, and the degradation mechanism inside the battery cannot be accurately described.
Data driving is a method of firstly establishing a black box model and then adjusting model parameters through a large amount of collected data so that the model can learn the mapping rule between the external parameters of the battery and SOH. The data driving method ignores electrochemical reaction and failure mechanism inside the battery, has lower requirement on knowledge in the related field and lower realization difficulty, but the quality of using a data set can directly influence the estimation effect, and the method needs charging or discharging data in the full SOH range, but most of the situations in the practical application of the battery only can obtain a few fragments, so the application range is narrow.
Disclosure of Invention
In order to overcome the defect that the prior art lacks a high-precision and wide-application range lithium battery health state assessment method, the invention provides a construction method of a lithium battery health state monitoring model, and the lithium battery health state monitoring model with wide application range and high assessment precision can be realized.
The invention provides a method for constructing a lithium battery health state monitoring model, which comprises the following steps:
s1, constructing a learning sample and a basic model; the learning sample is fusion characteristics Cor (f 1, f2, f 3) of the lithium battery marked with the residual life of the lithium battery, the input of the basic model is fusion characteristics Cor (f 1, f2, f 3) of the lithium battery, and the output of the basic model is a residual life predicted value of the lithium battery;
Cor(f1,f2,f3)=f1·f2/f3
f1=α×S(v)/I(a)
f2=β×I(a)/t(c)
f3=W(N)/P(d)
wherein f1, f2 and f3 are all fusion characteristics, alpha and beta are set fusion coefficients, S (v) is the discharge voltage integral of the lithium battery, and I (a) is the discharge current average value of the lithium battery; t (c) is the charging part time of the lithium battery, W (N) is the net discharge electric energy of the lithium battery, and P (d) is the discharge specific gravity of the lithium battery;
s2, the basic model learns the learning sample to train basic model parameters, and the trained basic model is obtained to be used as a lithium battery health state monitoring model, namely a health monitoring model for short.
Preferably, the calculation formula of the net discharge power W (N) of the lithium battery is as follows:
W(N)=a2×P(d) 2 +a1×P(d)+a0；
wherein a2, a1, a0 are fitting parameters.
Preferably, the basic model adopts a pulse neural network, and comprises an input layer, an input layer feedforward network, an intermediate layer feedforward network, an intermediate layer lateral network, an output layer feedforward network, an output layer lateral network and an output layer feedback network which are sequentially connected;
the input of the input layer is used as the input of the basic model, the neurons of the input layer are divided into a plurality of neural clusters, each neural cluster converts the input fusion characteristic into input current, and the output of the input layer is the input current converted by all the neural clusters; the input layer feedforward network is used for initializing the input current output by the input layer into a Gaussian distribution array w (Feed 1) conforming to the set first Gaussian distribution;
the middle layer side network is used for converting the pulse current array output by the middle layer into a Gaussian distribution array w (Lat 1) conforming to a set second Gaussian distribution; the output layer feedback network converts the pulse current array output by the output layer into a Gaussian distribution array w (back) conforming to a set third Gaussian distribution;
the middle layer converts a Gaussian distribution array w (Feed 1) output by the Feed-forward network of the input layer, a Gaussian distribution array w (Lat 1) output by the lateral network of the middle layer and a Gaussian distribution array w (back) output by the feedback network of the output layer into a pulse current array and outputs the pulse current array;
the intermediate layer feedforward network is used for converting the pulse current array output by the intermediate layer into a Gaussian distribution array w (Feed 2) conforming to a first Gaussian distribution; the output layer side network is used for converting the pulse current array output by the output layer into a Gaussian distribution array w (Lat 2) conforming to a second Gaussian distribution;
the output layer converts a Gaussian distribution array w (Feed 2) output by the intermediate layer feedforward network and a Gaussian distribution array w (Lat 2) output by the output layer side network into a pulse current array; the output layer feedforward network converts the pulse current array output by the output layer into a Gaussian distribution array w (Out) conforming to a fourth Gaussian distribution;
the output of the base model is the sum of the values in the gaussian distribution array w (Out) output by the output layer feed forward network.
Preferably, the n 1-th neural cluster in the input layer converts the input fusion feature Cor (f 1, f2, f 3) into the input current I (n 1), and the formula is as follows:
I(n1)=β1×exp[-(x(n1)-x(0,n1)) 2 /σ(n1) 2 ]
1≤n1≤N1
x (n 1) is the feature quantity encoded by the nth 1 neural cluster; x (0, n 1) is the set of feature quantities in the dimension of the input fusion feature Cor (f 1, f2, f 3) that gives rise to the optimal response of the neurons in the neural cluster n 1; sigma (n 1) is the standard deviation of values in each dimension in x (0, n 1), and beta 1 is the set neuron input current intensity coefficient; n1 is the number of neural clusters of the input layer; exp represents an exponential function based on a natural number e.
Preferably, the bias ∂ of the relative error of the output discharge time t (o, i) of the base modelE/∂ t (o, i) has the following formula:
∂E/∂t(o,i)=∑ U [∂E/∂t(j)][∂t(j)/∂t(o,i)]+A1
A1=∑ t (y(t)-y'(t))[w(o,s1)[∂ε s1 /∂t(o,i)]+w(o,s2)[∂ε s2 /∂t(o,i)]]
wherein, the set of neurons of the U-based model is ∂E/∂ t (j) represents the partial derivative of the relative error of the self-discharge time of the neuron, ∂ t (j)/∂ t (o, i) represents the bias of the relative error of the self-discharge time t (j) of the neuron with respect to the self-output discharge time ∂ t (o, i); y (t) is a basic model, and is a predicted value of the residual life of the lithium battery at the moment t, wherein y' (t) is a marked value of the residual life of the lithium battery at the moment t, and the predicted value of the residual life is output by fusion characteristics Cor (f 1, f2 and f 3) of the lithium battery at the moment t; w (o, s 1) is the weight of the output layer feedforward network neuron, and w (o, s 2) is the weight of the output layer neuron; w (o, s 1) and w (o, s 2) are hyper-parameters to be trained;
ε s1 a sudden electric current response function at t (i) time, wherein t (i) is a designated reference time; epsilon s2 At t(j) A sudden electric current response function at a moment; ∂ epsilon s1 ∂ t (o, i) represents the inrush current response function ε s1 (t-t (i)) relative to the output discharge time ∂ t (o, i), ∂ ε s2 ∂ t (o, i) represents the inrush current response function ε s2 (t-t (i)) relative to the output discharge time ∂ t (o, i); a1 is a transition parameter;
ε s1 (t-t(i))=c×(τ(m)-τ(s1))/c(s1,m)+c×(τ(s1)-τ1)/c(s1,1)-c×(τ(s1)-τ1)/c(s1)
ε s2 (t-t(i))=c×(τ(m)-τ(s2))/c(s2,m)+c×(τ(s2)-τ1)/c(s2,1)-c×(τ(s2)-τ1)/c(s2)
c=τ(m)-τ1
c(s1,m)=τ(m)e -(t-t(i))/τ(m) ；c(s2,m)=τ(m)e -(t-t(j))/τ(m) ；
c(s1,1)=τ1e -(t-t(i))/τ1 ；c(s2,1)=τ1e -(t-t(j))/τ1 ；
c(s1)=τ(s1)e -(t-t(i))/τ(s1) ；c(s2)=τ(s2)e -(t-t(j))/τ(s2) ；
where τ (m) is the time constant of the cell membrane, τ1 is the second order time constant of the synaptic current, s1 is the synaptic current intermediate variable at time t (i), τ (s 1) is the time constant of s 1; s2 is an abrupt current intermediate variable at a time t (j), and τ (s 2) is a time constant of s 2; c. c (s 1, m), c (s 1, 1), c (s 2, m), c (s 2, 1), c (s 2) are all transition parameters.
Preferably, the error gradient ∂ of the weights w (o, s 1) of the output layer feedforward network neuronsE/∂ w (o, s 1) has the following formula:
∂E/∂w(o,s1)=∑ t (y(t)-y'(t))ε s1 (t)
ε s1 (t)=c'×(τ(m)-τ(s1))/c'(s1,m)+c'×(τ(s1)-τ1)/c'(s1,1)-c'×(τ(s1)-τ1)/c'(s1)
c'=τ(m)-τ1；c'(s1,m)=τ(m)e -t/τ(m) ；c'(s1,1)=τ1e -t/τ1 ；c'(s1)=τ(s1)e -t/τ(s1) ；
wherein c ', c' (s 1, m), c '(s 1, 1) and c' (s 1) are all transition parameters; t is time.
Preferably, output layer feedbackPartial derivatives ∂ of relative errors of neurons in a network at time t (i)E/∂ t (i) has the following formula:
∂E/∂t(i)=∑ U [∂E/∂t(j)][∂t(j)/∂t(i)]
∂t(j)/∂t(i)=t(i){w(i,s1)[∂ε s1 (t)/∂t]+w(i,s2)[∂ε s2 (t)/∂t]}/[∂v(t)/∂t]
v(t)=∑ U w(i,s1)ε s1 (t-t(i))+∑ U w(i,s2)ε s2 (t-t(i))+∑ U η(t-t(j))
η(t-t(j))=(t-t(j))τ(m)e -(t-t(j))/τ(m) /[V(reset)-V(threshold)]
therein, ∂E/∂ t (j) represents the partial derivative of the relative error of t (j), ∂ t (j)/∂ t (i) represents the bias of the relative error of t (j) with respect to t (i);
w (i, s 1) is the synaptic weight at time t (i), and w (i, s 2) is the synaptic weight at time t (j); ∂ epsilon s1 (t)/∂ t represents ε s1 (t) deviation of relative error of time t, ∂ ε s2 (t)/∂ t represents ε s2 (t) a bias to the relative error of time t, ∂ v (t)/∂ t representing a bias to the relative error of time t by the transition term v (t);
u is a set of neurons; t is the current moment, t (i) is the input discharge moment of the presynaptic neuron, and t (j) is the self discharge moment of the neuron; η is a neuron self-discharging reset function;
v (reset) is a set rest voltage, and V (threshold) is a set threshold voltage. .
The invention provides a lithium battery health state monitoring method, which comprises the steps of firstly, acquiring a health monitoring model by adopting the construction method of the lithium battery health state monitoring model; acquiring a discharge voltage integral S (v) of a lithium battery to be evaluated, a discharge current average value I (a) of the lithium battery, a charging part time t (c) of the lithium battery, a net discharge electric energy W (N) of the lithium battery and a discharge specific gravity P (d) of the lithium battery, and calculating fusion characteristics Cor (f 1, f2 and f 3) of the lithium battery to be evaluated; and then inputting the fusion characteristics Cor (f 1, f2, f 3) of the lithium battery to be evaluated into a health monitoring model, and outputting the predicted value of the residual life of the lithium battery to be evaluated by the health monitoring model.
The invention provides a lithium battery health state monitoring system, which comprises a memory and a processor, wherein a computer program and a health monitoring model are stored in the memory, the processor is connected with the memory, and the processor is used for executing the computer program so as to realize the lithium battery health state monitoring method.
The invention also provides a storage medium, which stores a computer program and a health monitoring model, wherein the computer program is used for realizing the lithium battery health state monitoring method when being executed.
The invention has the advantages that:
(1) The invention provides a method for constructing a lithium battery health state monitoring model, which constructs the lithium battery health state monitoring model by taking the fusion characteristic of a lithium battery as input. Therefore, in the invention, the working condition data of the lithium battery is preprocessed through feature fusion, so that the input of the lithium battery health state monitoring model is simplified, and the model prediction efficiency is improved. According to the invention, the characteristic fusion is used for comprehensively considering the working condition data of the lithium battery, so that the accuracy of the health state assessment is improved, and an efficient and accurate lithium battery health state monitoring model is realized.
(2) According to the invention, the aging characteristic and the working condition characteristic extracted during discharging under the dynamic working condition of the lithium battery are utilized and then fused, so that the fused characteristic is used as model input, and compared with the traditional input quantity, the input quantity fully considers the coupling relation among all factors, and the robustness of soh prediction can be improved.
(3) According to the invention, the mapping function of the net discharge electric energy W (N) of the lithium battery and the discharge specific gravity P (d) of the lithium battery is obtained through parameter fitting, so that simple and efficient calculation of the net discharge electric energy W (N) of the lithium battery is realized, the overall calculation efficiency of the lithium battery health state assessment is improved, and the acquisition process of the lithium battery working condition data is simplified.
(4) The invention introduces the pulse neural network, has the characteristics of noise resistance, low power consumption and the like, predicts the health State (SOH) of the lithium battery, and can improve the prediction precision when being applied to the prediction of the health state of the lithium battery under some complex working condition environments.
(5) According to the invention, on the traditional impulse neural network, the error back-propagation calculation method and the network structure of the impulse time are optimized, so that the stable and flexible learning of the multi-layer network structure with complex lateral connection and the continuous network activity model can be ensured, the impulse neural network can update network parameters more quickly, and the model convergence speed and the model prediction precision can be improved.
Drawings
FIG. 1 is a flow chart of a method for constructing a lithium battery health status monitoring model;
FIG. 2 is a schematic diagram of network connections of a health monitoring model;
FIG. 3 is a graph comparing MAE error curves of the method, the CNN model and the LSTM model during training;
FIG. 4 is a graph showing SOH predicted and actual values of the methods, CNN models and LSTM models according to the embodiments of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a method for constructing a lithium battery health status monitoring model according to the present embodiment includes the following steps S1-S2.
S1, constructing a learning sample and a basic model, wherein the learning sample is recorded as { Cor (f 1, f2, f 3) }, the residual life of the lithium battery, the input of the basic model is fusion characteristics Cor (f 1, f2, f 3) of the lithium battery, and the output of the basic model is a residual life predicted value of the lithium battery;
Cor(f1,f2,f3)=f1·f2/f3
f1=α×S(v)/I(a)
f2=β×I(a)/t(c)
f3=W(N)/P(d)
wherein f1, f2 and f3 are all fusion characteristics, alpha and beta are set fusion coefficients, S (v) is the discharge voltage integral of the lithium battery, and I (a) is the discharge current average value of the lithium battery; t (c) is the charging part time of the lithium battery, W (N) is the net discharge electric energy of the lithium battery, and P (d) is the discharge specific gravity of the lithium battery;
in specific implementation, the following formulas can be used to calculate S (v), W (N) and P (d);
S(v)=∫ 0 t(end) U(t)dt
u (t) is the discharge voltage of the lithium battery; t is time, t (end) is discharge end time;
P(d)=Q(d)/[Q(c)+Q(d)]
W(N)=a2×P(d) 2 +a1×P(d)+a0
wherein Q (d) is the discharge amount in the discharge process of the lithium battery under the dynamic working condition; q (c) is the charge amount in the discharging process of the dynamic working condition of the lithium battery; a2, a1 and a0 are fitting parameters; specifically, the parameters of a2, a1 and a0 can be solved by carrying out parameter fitting on a sample { P (d), W (N) }, and the calculation formula of W (N) in the sample is as follows:
W(N)=∫U(d)I(d)dt(d)-∫U(c)I(c)dt(c)
i (d) is the discharge current of the lithium battery, U (d) is the voltage corresponding to the I (d), t (d) is the discharge part time of the lithium battery, I (c) is the charge current of the lithium battery, U (c) is the voltage corresponding to the I (c), and t (c) is the charge part time of the lithium battery.
S2, the basic model learns the learning sample to train basic model parameters, and the trained basic model is obtained to be used as a lithium battery health state monitoring model, namely a health monitoring model for short.
In specific implementation, the mean square error can be used as a loss function in the basic model training process.
Referring to fig. 2, the basic model adopts a pulsed neural network model, which includes an input layer, an input layer feedforward network, an intermediate layer feedforward network, an intermediate layer side network, an output layer feedforward network, an output layer side network, and an output layer feedback network, which are sequentially connected.
The input of the input layer is taken as the input of the basic model, the input of the input layer is fusion characteristics Cor (f 1, f2, f 3) of the lithium battery, and each nerve cluster of the input layer encodes the input fusion characteristics Cor (f 1, f2, f 3) respectively to generate input current;
the process of converting the input fusion feature Cor (f 1, f2, f 3) into the input current I (n 1) by the n 1-th neural cluster in the input layer is expressed as follows:
I(n1)=β1×exp[-(x(n1)-x(0,n1)) 2 /σ(n1) 2 ]
1≤n1≤N1
x (n 1) is the feature quantity encoded by the neural cluster n1, namely, the data of the neural cluster n1 dimensionally sampled from the input fusion features Cor (f 1, f2, f 3); x (0, n 1) is the set of feature quantities in the dimension of the input fusion feature Cor (f 1, f2, f 3) that gives rise to the optimal response of the neurons in the neural cluster n 1; sigma (n 1) is the standard deviation of values in each dimension in x (0, n 1), and beta 1 is the set neuron input current intensity coefficient; n1 is the number of neural clusters of the input layer. In implementation, the input layer may be configured with 800 neurons, and divided into 20 neural clusters, each having 40 neurons, i.e., n1=20.
Let the array w (aa) conforming to the Gaussian distribution N (w ', σ') be denoted as w (aa) to N (w ', σ').
The input layer feedforward network is used for outputting an input current { I (n 1) output by the input layer; 1.ltoreq.n1.ltoreq.n1 } initializing to a Gaussian distribution array w (Feed 1) to N (w 0, sigma (w)), that is, w (Feed 1) obeys a Gaussian distribution N (w 0, sigma (w)) with a mean value of w0 and a standard deviation of sigma (w); w0 and σ (w) are both set values.
The middle layer side network is used for converting the pulse current array output by the middle layer into a Gaussian distribution array w (Lat 1) -N (w (Lat), sigma (Lat)), namely, the w (Lat 1) obeys Gaussian distribution N (w (Lat), sigma (Lat)) with the average value of w (Lat) and the standard deviation of sigma (Lat); w (lat) and sigma (lat) are set values.
The output layer feedback network converts the pulse current array output by the output layer into Gaussian distribution arrays w (back) to N (0, sigma (w)), namely, w (back) obeys Gaussian distribution N (0, sigma (w)) with the mean value of 0 and the standard deviation of sigma (w).
The middle layer converts the Gaussian distribution array w (Feed 1) output by the Feed-forward network of the input layer, the Gaussian distribution array w (Lat 1) output by the lateral network of the middle layer and the Gaussian distribution array w (back) output by the feedback network of the output layer into a pulse current array and outputs the pulse current array.
The intermediate layer feedforward network is used for converting the pulse current array output by the intermediate layer into a Gaussian distribution array w (Feed 2) to N (w 0, sigma (w)), namely, the w (Feed 2) obeys a Gaussian distribution N (w 0, sigma (w)) with the average value of w0 and the standard deviation of sigma (w).
The output layer side network is used for converting the pulse current array output by the output layer into a Gaussian distribution array w (Lat 2) -N (w (Lat), sigma (Lat)), namely, the w (Lat 2) obeys Gaussian distribution N (w (Lat), sigma (Lat)) with the mean value of w (Lat) and the standard deviation of sigma (Lat).
The output layer converts the Gaussian distribution array w (Feed 2) output by the intermediate layer feedforward network and the Gaussian distribution array w (Lat 2) output by the output layer side network into a pulse current array.
The output layer feedforward network converts the pulse current array output by the output layer into a Gaussian distribution array w (Out) -N (0, sigma (Out)), namely w (Out) obeys Gaussian distribution N (0, sigma (Out)) with the average value of 0 and the standard deviation of sigma (Out); sigma (out) is a set value.
The output of the base model is the sum of the values in the gaussian distribution array w (Out) output by the output layer feed forward network.
In the present embodiment, the bias ∂ of the relative error of the output discharge time t (o, i) of the basic model is further calculatedE/∂ t (o, i), error gradient ∂ of weights w (o, s 1) of output layer feedforward network neuronsE/∂ w (o, s 1), bias conductance ∂ of relative error of neurons in output layer feedback network at time t (i)E/∂ t (i) and the like.
∂E/∂t(o,i)=∑ U [∂E/∂t(j)][∂t(j)/∂t(o,i)]+A1
A1=∑ t (y(t)-y'(t))[w(o,s1)[∂ε s1 /∂t(o,i)]+w(o,s2)[∂ε s2 /∂t(o,i)]]
Wherein, the set of neurons of the U-based model is ∂E/∂ t (j) represents the partial derivative of the relative error of the self-discharge time of the neuron, ∂ t (j)/∂ t (o, i) represents the bias of the relative error of the self-discharge time t (j) of the neuron with respect to the self-output discharge time ∂ t (o, i); y (t) is the fundamental modeThe predicted value of the residual life of the lithium battery output aiming at the fusion characteristics Cor (f 1, f2, f 3) of the lithium battery at the moment t, wherein y' (t) is the marked value of the residual life of the lithium battery at the moment t; w (o, s 1) is the weight of the output layer feedforward network neuron, and w (o, s 2) is the weight of the output layer neuron; w (o, s 1) and w (o, s 2) are hyper-parameters to be trained;
ε s1 a sudden electric current response function at t (i) time, wherein t (i) is a designated reference time; epsilon s2 A sudden electric current response function at the time t (j); ∂ epsilon s1 ∂ t (o, i) represents the inrush current response function ε s1 (t-t (i)) relative to the output discharge time ∂ t (o, i), ∂ ε s2 ∂ t (o, i) represents the inrush current response function ε s2 (t-t (i)) relative to the output discharge time ∂ t (o, i); a1 is a transition parameter.
ε s1 (t-t(i))=c×(τ(m)-τ(s1))/c(s1,m)+c×(τ(s1)-τ1)/c(s1,1)-c×(τ(s1)-τ1)/c(s1)
ε s2 (t-t(i))=c×(τ(m)-τ(s2))/c(s2,m)+c×(τ(s2)-τ1)/c(s2,1)-c×(τ(s2)-τ1)/c(s2)
c=τ(m)-τ1
c(s1,m)=τ(m)e -(t-t(i))/τ(m) ；c(s2,m)=τ(m)e -(t-t(j))/τ(m) ；
c(s1,1)=τ1e -(t-t(i))/τ1 ；c(s2,1)=τ1e -(t-t(j))/τ1 ；
c(s1)=τ(s1)e -(t-t(i))/τ(s1) ；c(s2)=τ(s2)e -(t-t(j))/τ(s2) ；
Where τ (m) is the time constant of the cell membrane, τ1 is the second order time constant of the synaptic current, s1 is the synaptic current intermediate variable at time t (i), τ (s 1) is the time constant of s 1; s2 is an abrupt current intermediate variable at a time t (j), and τ (s 2) is a time constant of s 2; c. c (s 1, m), c (s 1, 1), c (s 2, m), c (s 2, 1), c (s 2) are all transition parameters.
∂E/∂w(o,s1)=∑ t (y(t)-y'(t))ε s1 (t)
ε s1 (t)=c'×(τ(m)-τ(s1))/c'(s1,m)+c'×(τ(s1)-τ1)/c'(s1,1)-c'×(τ(s1)-τ1)/c'(s1)
c'=τ(m)-τ1；c'(s1,m)=τ(m)e -t/τ(m) ；c'(s1,1)=τ1e -t/τ1 ；c'(s1)=τ(s1)e -t/τ(s1) ；
Wherein c ', c' (s 1, m), c '(s 1, 1) and c' (s 1) are all transition parameters; t is time.
∂E/∂t(i)=∑ U [∂E/∂t(j)][∂t(j)/∂t(i)]
∂E/∂ t (j) represents the partial derivative of the relative error of t (j), ∂ t (j)/∂ t (i) represents the bias of the relative error of t (j) with respect to t (i);
∂t(j)/∂t(i)=t(i){w(i,s1)[∂ε s1 (t)/∂t]+w(i,s2)[∂ε s2 (t)/∂t]}/[∂v(t)/∂t]
wherein w (i, s 1) is the synaptic weight at the time of t (i), and w (i, s 2) is the synaptic weight at the time of t (j); ∂ epsilon s1 (t)/∂ t represents ε s1 (t) deviation of relative error of time t, ∂ ε s2 (t)/∂ t represents ε s2 (t) a bias to the relative error of time t, ∂ v (t)/∂ t representing a bias to the relative error of time t by the transition term v (t);
v(t)=∑ U w(i,s1)ε s1 (t-t(i))+∑ U w(i,s2)ε s2 (t-t(i))+∑ U η(t-t(j))
wherein U is a set of neurons; t is the current moment, t (i) is the input discharge moment of the presynaptic neuron, and t (j) is the self discharge moment of the neuron; η is a neuron self-discharging reset function;
η(t-t(j))=(t-t(j))τ(m)e -(t-t(j))/τ(m) /[V(reset)-V(threshold)]
v (reset) is a set rest voltage, and V (threshold) is a set threshold voltage.
The following describes a lithium battery health status monitoring model (abbreviated as health monitoring model) according to the present invention with reference to specific embodiments.
In this embodiment, the fusion coefficient α=0.5, β=0.2, β1=0.5 is substituted into the formula to calculate the fusion characteristic Cor (f 1, f2, f 3) of the lithium battery, and the above method for constructing the health monitoring model of the lithium battery is first adopted to construct the health monitoring model. Specifically, in the basic model of the present embodiment, 800 neurons are set in the input layer and divided into 20 neural clusters, that is, n1=20; the middle layer and the output layer are provided with 500 neurons, respectively.
In this embodiment, a CNN model and an LSTM model are used as the comparison models, the input of the comparison models is also set as the fusion characteristics Cor (f 1, f2, f 3) of the lithium battery, and the output is the remaining life of the lithium battery.
In this embodiment, a training set is constructed based on the same learning sample { Cor (f 1, f2, f 3), remaining life } to train parameters of three models.
In the training process, the mean square error (MAE) change process of each model is shown in fig. 3, and it can be seen that the lithium battery health state monitoring model provided by the invention has the highest convergence speed and the smallest error, and the model has higher accuracy.
In this embodiment, the accuracy of the three models after training is further tested on the training set, and the pair of the predicted value of the remaining life of the lithium battery and the actual value GroundTruth at each stage of the life cycle of each model is shown in fig. 4.
It will be understood by those skilled in the art that the present invention is not limited to the details of the foregoing exemplary embodiments, but includes other specific forms of the same or similar structures that may be embodied without departing from the spirit or essential characteristics thereof. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.
Furthermore, it should be understood that although the present disclosure describes embodiments, not every embodiment is provided with a separate embodiment, and that this description is provided for clarity only, and that the disclosure is not limited to the embodiments described in detail below, and that the embodiments described in the examples may be combined as appropriate to form other embodiments that will be apparent to those skilled in the art.
The technology, shape, and construction parts of the present invention, which are not described in detail, are known in the art.
Claims (10)
1. The construction method of the lithium battery health state monitoring model is characterized by comprising the following steps of:
s1, constructing a learning sample and a basic model; the learning sample is fusion characteristics Cor (f 1, f2, f 3) of the lithium battery marked with the residual life of the lithium battery, the input of the basic model is fusion characteristics Cor (f 1, f2, f 3) of the lithium battery, and the output of the basic model is a residual life predicted value of the lithium battery;
Cor(f1,f2,f3)=f1·f2/f3
f1=α×S(v)/I(a)
f2=β×I(a)/t(c)
f3=W(N)/P(d)
wherein f1, f2 and f3 are all fusion characteristics, alpha and beta are set fusion coefficients, S (v) is the discharge voltage integral of the lithium battery, and I (a) is the discharge current average value of the lithium battery; t (c) is the charging part time of the lithium battery, W (N) is the net discharge electric energy of the lithium battery, and P (d) is the discharge specific gravity of the lithium battery;
s2, the basic model learns the learning sample to train basic model parameters, and the trained basic model is obtained to be used as a lithium battery health state monitoring model, namely a health monitoring model for short.
2. The method for constructing a lithium battery health monitoring model according to claim 1, wherein the calculation formula of the net discharge electric energy W (N) of the lithium battery is as follows:
W(N)=a2×P(d) 2 +a1×P(d)+a0；
wherein a2, a1, a0 are fitting parameters.
3. The method for constructing a lithium battery health status monitoring model according to claim 2, wherein the basic model adopts a pulse neural network, and comprises an input layer, an input layer feedforward network, an intermediate layer feedforward network, an intermediate layer lateral network, an output layer feedforward network, an output layer lateral network and an output layer feedback network which are sequentially connected;
the input of the input layer is used as the input of the basic model, the neurons of the input layer are divided into a plurality of neural clusters, each neural cluster converts the input fusion characteristic into input current, and the output of the input layer is the input current converted by all the neural clusters; the input layer feedforward network is used for initializing the input current output by the input layer into a Gaussian distribution array w (Feed 1) conforming to the set first Gaussian distribution;
the middle layer side network is used for converting the pulse current array output by the middle layer into a Gaussian distribution array w (Lat 1) conforming to a set second Gaussian distribution; the output layer feedback network converts the pulse current array output by the output layer into a Gaussian distribution array w (back) conforming to a set third Gaussian distribution;
the middle layer converts a Gaussian distribution array w (Feed 1) output by the Feed-forward network of the input layer, a Gaussian distribution array w (Lat 1) output by the lateral network of the middle layer and a Gaussian distribution array w (back) output by the feedback network of the output layer into a pulse current array and outputs the pulse current array;
the intermediate layer feedforward network is used for converting the pulse current array output by the intermediate layer into a Gaussian distribution array w (Feed 2) conforming to a first Gaussian distribution; the output layer side network is used for converting the pulse current array output by the output layer into a Gaussian distribution array w (Lat 2) conforming to a second Gaussian distribution;
the output layer converts a Gaussian distribution array w (Feed 2) output by the intermediate layer feedforward network and a Gaussian distribution array w (Lat 2) output by the output layer side network into a pulse current array; the output layer feedforward network converts the pulse current array output by the output layer into a Gaussian distribution array w (Out) conforming to a fourth Gaussian distribution;
the output of the base model is the sum of the values in the gaussian distribution array w (Out) output by the output layer feed forward network.
4. A method for constructing a lithium battery health monitoring model according to claim 3, wherein the process of converting the input fusion characteristic Cor (f 1, f2, f 3) into the input current I (n 1) by the n 1-th neural cluster in the input layer is expressed as follows:
I(n1)=β1×exp[-(x(n1)-x(0,n1)) 2 /σ(n1) 2 ]
1≤n1≤N1
x (n 1) is the feature quantity encoded by the nth 1 neural cluster; x (0, n 1) is the set of feature quantities in the dimension of the input fusion feature Cor (f 1, f2, f 3) that gives rise to the optimal response of the neurons in the neural cluster n 1; sigma (n 1) is the standard deviation of values in each dimension in x (0, n 1), and beta 1 is the set neuron input current intensity coefficient; n1 is the number of neural clusters of the input layer; exp represents an exponential function based on a natural number e.
5. The method for constructing a model for monitoring the health of a lithium battery as set forth in claim 3, wherein the bias ∂ of the relative error of the output discharge time t (o, i) of the basic modelE/∂ t (o, i) has the following formula:
∂E/∂t(o,i)=∑ U [∂E/∂t(j)][∂t(j)/∂t(o,i)]+A1
A1=∑ t (y(t)-y'(t))[w(o,s1)[∂ε s1 /∂t(o,i)]+w(o,s2)[∂ε s2 /∂t(o,i)]]
wherein, the set of neurons of the U-based model is ∂E/∂ t (j) represents the partial derivative of the relative error of the self-discharge time of the neuron, ∂ t (j)/∂ t (o, i) represents the bias of the relative error of the self-discharge time t (j) of the neuron with respect to the self-output discharge time ∂ t (o, i); y (t) is a basic model, and is a predicted value of the residual life of the lithium battery at the moment t, wherein y' (t) is a marked value of the residual life of the lithium battery at the moment t, and the predicted value of the residual life is output by fusion characteristics Cor (f 1, f2 and f 3) of the lithium battery at the moment t; w (o, s 1) is the weight of the output layer feedforward network neuron, and w (o, s 2) is the weight of the output layer neuron; w (o, s 1) and w (o, s 2) are hyper-parameters to be trained;
ε s1 abrupt current response at time t (i)A function, t (i) is a specified reference time; epsilon s2 A sudden electric current response function at the time t (j); ∂ epsilon s1 ∂ t (o, i) represents the inrush current response function ε s1 (t-t (i)) relative to the output discharge time ∂ t (o, i), ∂ ε s2 ∂ t (o, i) represents the inrush current response function ε s2 (t-t (i)) relative to the output discharge time ∂ t (o, i); a1 is a transition parameter;
ε s1 (t-t(i))=c×(τ(m)-τ(s1))/c(s1,m)+c×(τ(s1)-τ1)/c(s1,1)-c×(τ(s1)-τ1)/c(s1)
ε s2 (t-t(i))=c×(τ(m)-τ(s2))/c(s2,m)+c×(τ(s2)-τ1)/c(s2,1)-c×(τ(s2)-τ1)/c(s2)
c=τ(m)-τ1
c(s1,m)=τ(m)e -(t-t(i))/τ(m) ；c(s2,m)=τ(m)e -(t-t(j))/τ(m) ；
c(s1,1)=τ1e -(t-t(i))/τ1 ；c(s2,1)=τ1e -(t-t(j))/τ1 ；
c(s1)=τ(s1)e -(t-t(i))/τ(s1) ；c(s2)=τ(s2)e -(t-t(j))/τ(s2) ；
where τ (m) is the time constant of the cell membrane, τ1 is the second order time constant of the synaptic current, s1 is the synaptic current intermediate variable at time t (i), τ (s 1) is the time constant of s 1; s2 is an abrupt current intermediate variable at a time t (j), and τ (s 2) is a time constant of s 2; c. c (s 1, m), c (s 1, 1), c (s 2, m), c (s 2, 1), c (s 2) are all transition parameters.
6. The method for building a health monitoring model of a lithium battery according to claim 5, wherein the error gradient ∂ of the weights w (o, s 1) of the neurons of the feed-forward network of the output layerE/∂ w (o, s 1) has the following formula:
∂E/∂w(o,s1)=∑ t (y(t)-y'(t))ε s1 (t)
ε s1 (t)=c'×(τ(m)-τ(s1))/c'(s1,m)+c'×(τ(s1)-τ1)/c'(s1,1)-c'×(τ(s1)-τ1)/c'(s1)
c'=τ(m)-τ1；c'(s1,m)=τ(m)e -t/τ(m) ；c'(s1,1)=τ1e -t/τ1 ；c'(s1)=τ(s1)e -t/τ(s1) ；
wherein c ', c' (s 1, m), c '(s 1, 1) and c' (s 1) are all transition parameters; t is time.
7. The method for building a lithium battery health status monitoring model according to claim 5, wherein the bias guide ∂ of the relative error of the neurons in the output layer feedback network at the time t (i)E/∂ t (i) has the following formula:
∂E/∂t(i)=∑ U [∂E/∂t(j)][∂t(j)/∂t(i)]
∂t(j)/∂t(i)=t(i){w(i,s1)[∂ε s1 (t)/∂t]+w(i,s2)[∂ε s2 (t)/∂t]}/[∂v(t)/∂t]
v(t)=∑ U w(i,s1)ε s1 (t-t(i))+∑ U w(i,s2)ε s2 (t-t(i))+∑ U η(t-t(j))
η(t-t(j))=(t-t(j))τ(m)e -(t-t(j))/τ(m) /[V(reset)-V(threshold)]
therein, ∂E/∂ t (j) represents the partial derivative of the relative error of t (j), ∂ t (j)/∂ t (i) represents the bias of the relative error of t (j) with respect to t (i);
w (i, s 1) is the synaptic weight at time t (i), and w (i, s 2) is the synaptic weight at time t (j); ∂ epsilon s1 (t)/∂ t represents ε s1 (t) deviation of relative error of time t, ∂ ε s2 (t)/∂ t represents ε s2 (t) a bias to the relative error of time t, ∂ v (t)/∂ t representing a bias to the relative error of time t by the transition term v (t);
u is a set of neurons; t is the current moment, t (i) is the input discharge moment of the presynaptic neuron, and t (j) is the self discharge moment of the neuron; η is a neuron self-discharging reset function;
v (reset) is a set rest voltage, and V (threshold) is a set threshold voltage.
8. A method for monitoring the health state of a lithium battery, which is characterized in that a health monitoring model is firstly obtained by adopting the method for constructing the health state monitoring model of the lithium battery according to any one of claims 1 to 7; acquiring a discharge voltage integral S (v) of a lithium battery to be evaluated, a discharge current average value I (a) of the lithium battery, a charging part time t (c) of the lithium battery, a net discharge electric energy W (N) of the lithium battery and a discharge specific gravity P (d) of the lithium battery, and calculating fusion characteristics Cor (f 1, f2 and f 3) of the lithium battery to be evaluated; and then inputting the fusion characteristics Cor (f 1, f2, f 3) of the lithium battery to be evaluated into a health monitoring model, and outputting the predicted value of the residual life of the lithium battery to be evaluated by the health monitoring model.
9. A lithium battery health status monitoring system comprising a memory and a processor, wherein the memory stores a computer program and a health monitoring model, the processor is connected to the memory, and the processor is configured to execute the computer program to implement the lithium battery health status monitoring method of claim 8.
10. A storage medium, characterized in that a computer program and a health monitoring model are stored, which computer program, when executed, is adapted to implement the lithium battery health status monitoring method according to claim 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410163216.3A CN117706406B (en) | 2024-02-05 | 2024-02-05 | Lithium battery health state monitoring model, method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410163216.3A CN117706406B (en) | 2024-02-05 | 2024-02-05 | Lithium battery health state monitoring model, method, system and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117706406A true CN117706406A (en) | 2024-03-15 |
| CN117706406B CN117706406B (en) | 2024-04-16 |
Family
ID=90162855
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410163216.3A Active CN117706406B (en) | 2024-02-05 | 2024-02-05 | Lithium battery health state monitoring model, method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117706406B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110346734A (en) * | 2019-06-19 | 2019-10-18 | 江苏大学 | A kind of lithium-ion-power cell health status evaluation method based on machine learning |
| CN114742203A (en) * | 2022-03-31 | 2022-07-12 | 武汉理工大学 | Lithium battery state prediction method, system and medium based on hybrid machine learning method |
| CN114994541A (en) * | 2022-06-02 | 2022-09-02 | 中国计量大学 | Lithium ion battery SOH estimation method based on multi-strategy fusion |
| CN115684971A (en) * | 2022-10-18 | 2023-02-03 | 国网福建省电力有限公司 | Lithium ion battery health state estimation method based on fragment multi-charging feature fusion |
| EP4145157A1 (en) * | 2021-09-06 | 2023-03-08 | Dukosi Limited | Battery system state of health monitoring system |
| US20230187710A1 (en) * | 2021-12-13 | 2023-06-15 | Enevate Corporation | State-of-health models for lithium-silicon batteries |
| CN116387661A (en) * | 2023-01-18 | 2023-07-04 | 国网浙江省电力有限公司衢州供电公司 | Lithium battery safety operation and maintenance management system and lithium battery health state assessment method |
| CN116643174A (en) * | 2023-05-09 | 2023-08-25 | 大连理工大学 | Battery remaining life prediction method |
| CN116680983A (en) * | 2023-06-06 | 2023-09-01 | 重庆大学 | Lithium ion residual life prediction method based on improved particle filter model |
| CN116794543A (en) * | 2023-06-07 | 2023-09-22 | 广东工业大学 | Lithium battery performance prediction method based on GBLS boost multitask learning model |
-
2024
- 2024-02-05 CN CN202410163216.3A patent/CN117706406B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110346734A (en) * | 2019-06-19 | 2019-10-18 | 江苏大学 | A kind of lithium-ion-power cell health status evaluation method based on machine learning |
| EP4145157A1 (en) * | 2021-09-06 | 2023-03-08 | Dukosi Limited | Battery system state of health monitoring system |
| US20230187710A1 (en) * | 2021-12-13 | 2023-06-15 | Enevate Corporation | State-of-health models for lithium-silicon batteries |
| CN114742203A (en) * | 2022-03-31 | 2022-07-12 | 武汉理工大学 | Lithium battery state prediction method, system and medium based on hybrid machine learning method |
| CN114994541A (en) * | 2022-06-02 | 2022-09-02 | 中国计量大学 | Lithium ion battery SOH estimation method based on multi-strategy fusion |
| CN115684971A (en) * | 2022-10-18 | 2023-02-03 | 国网福建省电力有限公司 | Lithium ion battery health state estimation method based on fragment multi-charging feature fusion |
| CN116387661A (en) * | 2023-01-18 | 2023-07-04 | 国网浙江省电力有限公司衢州供电公司 | Lithium battery safety operation and maintenance management system and lithium battery health state assessment method |
| CN116643174A (en) * | 2023-05-09 | 2023-08-25 | 大连理工大学 | Battery remaining life prediction method |
| CN116680983A (en) * | 2023-06-06 | 2023-09-01 | 重庆大学 | Lithium ion residual life prediction method based on improved particle filter model |
| CN116794543A (en) * | 2023-06-07 | 2023-09-22 | 广东工业大学 | Lithium battery performance prediction method based on GBLS boost multitask learning model |
Non-Patent Citations (1)
| Title |
|---|
| 高静怡 等: "充放电多特征融合的锂电池寿命预测方法", 重庆理工大学学报（自然科学）, vol. 37, no. 2, 31 December 2023 (2023-12-31) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117706406B (en) | 2024-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112241608B (en) | Lithium battery life prediction method based on LSTM network and transfer learning | |
| KR102354112B1 (en) | Apparatus and method for estimating status of battery based on artificial intelligence | |
| CN110221225B (en) | Spacecraft lithium ion battery cycle life prediction method | |
| CN111736084B (en) | Valve-regulated lead-acid storage battery health state prediction method based on improved LSTM neural network | |
| CN111856287B (en) | Lithium battery health state detection method based on stacked residual causal convolutional neural network | |
| CN111680848A (en) | Battery life prediction method based on prediction model fusion and storage medium | |
| Yamazaki et al. | Estimation of the residual capacity of sealed lead-acid batteries by neural network | |
| CN110689171A (en) | Turbine health state prediction method based on E-LSTM | |
| CN107436409B (en) | Intelligent SOC prediction device for power battery of electric automobile | |
| CN115632179B (en) | Intelligent quick charging method and system for lithium ion battery | |
| KR20220021973A (en) | Method and apparatus for diagnosing defect of battery cell based on neural network | |
| CN112630659A (en) | Lithium battery SOC estimation method based on improved BP-EKF algorithm | |
| CN110082682A (en) | A kind of lithium battery charge state estimation method | |
| CN113917334A (en) | Battery health state estimation method based on evolution LSTM self-encoder | |
| CN113554200B (en) | Method, system and equipment for predicting voltage inconsistency of power battery | |
| CN110232432B (en) | Lithium battery pack SOC prediction method based on artificial life model | |
| CN114839542A (en) | Sample data set generation method and SOC estimation method for power lithium battery | |
| CN113791351B (en) | Lithium battery life prediction method based on transfer learning and difference probability distribution | |
| CN113408138B (en) | Lithium battery SOH estimation method and system based on secondary fusion | |
| KR20230043466A (en) | A System and Method for Batteries State of Health Estimation using Multiple channel charging profile | |
| CN117706406B (en) | Lithium battery health state monitoring model, method, system and storage medium | |
| CN117022048A (en) | Electric automobile battery state of charge evaluation method | |
| CN114895206B (en) | Lithium ion battery SOH estimation method based on RBF neural network of improved gray wolf optimization algorithm | |
| CN115372852A (en) | Storage battery health state prediction method based on feedforward neural network | |
| CN114624602A (en) | Energy storage battery system parallel branch current estimation value correction 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |