CN117473396B - New energy automobile intelligent battery management system based on deep learning - Google Patents
New energy automobile intelligent battery management system based on deep learning Download PDFInfo
- Publication number
- CN117473396B CN117473396B CN202311572198.6A CN202311572198A CN117473396B CN 117473396 B CN117473396 B CN 117473396B CN 202311572198 A CN202311572198 A CN 202311572198A CN 117473396 B CN117473396 B CN 117473396B
- Authority
- CN
- China
- Prior art keywords
- battery
- gate
- sample
- data
- vector
- 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.)
- Active
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 12
- 238000000605 extraction Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 230000006870 function Effects 0.000 claims description 48
- 230000004913 activation Effects 0.000 claims description 30
- 239000011159 matrix material Substances 0.000 claims description 27
- 230000002159 abnormal effect Effects 0.000 claims description 22
- 238000000034 method Methods 0.000 claims description 19
- 230000008859 change Effects 0.000 claims description 15
- 230000015654 memory Effects 0.000 claims description 12
- 230000007787 long-term memory Effects 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 3
- 238000004880 explosion Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
- B60L58/00—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles
- B60L58/10—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries
- B60L58/16—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries responding to battery ageing, e.g. to the number of charging cycles or the state of health [SoH]
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/382—Arrangements for monitoring battery or accumulator variables, e.g. SoC
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/60—Other road transportation technologies with climate change mitigation effect
- Y02T10/70—Energy storage systems for electromobility, e.g. batteries
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Mechanical Engineering (AREA)
- Transportation (AREA)
- Power Engineering (AREA)
- Sustainable Energy (AREA)
- Sustainable Development (AREA)
- Secondary Cells (AREA)
Abstract
The invention discloses a new energy automobile intelligent battery management system based on deep learning, which comprises a data acquisition module, a data preprocessing module, a feature extraction module, a battery remaining life prediction model building module and a battery remaining life prediction module. The invention relates to the technical field of intelligent battery management, in particular to a new energy automobile intelligent battery management system based on deep learning.
Description
Technical Field
The invention relates to the technical field of intelligent battery management, in particular to a new energy automobile intelligent battery management system based on deep learning.
Background
An intelligent battery management system is a system that utilizes advanced technology to monitor, control and optimize the performance of a battery system by using sensors, circuits and algorithms to monitor the state and performance of the battery. The existing intelligent battery management system for the new energy automobile based on deep learning has the problems that the data dimension is different, partial characteristics have great influence on the result, and the acquired data contains abnormal values, so that the subsequent data processing is distorted; the long-term and short-term memory network prediction model has overlarge gradient values when the weight and the threshold value are updated, so that model parameters are updated too much, and further the problems of poor stability, difficult convergence and gradient explosion of the model are caused.
Disclosure of Invention
Aiming at the problems that the data dimension is different, partial characteristics have larger influence on the result, and the acquired data contains abnormal values so as to cause distortion of subsequent data processing, the invention adopts an abnormal value detection and data normalization algorithm, calculates a sample mean value vector, a covariance matrix and a mahalanobis distance between the observation value vector and the sample of the sample mean value vector by constructing the observation value vector, detects and removes the abnormal values, normalizes the data, unifies the data of different dimensions into a specific range, eliminates the dimensional difference among the different dimensions, and prevents overlarge influence of certain characteristics on the result; aiming at the problems that the model parameter is updated excessively large when the long-short-term memory network prediction model updates the weight and the threshold value, so that the model parameter is updated excessively large, and the model stability is poor, convergence is difficult and gradient explosion is caused.
The technical scheme adopted by the invention is as follows: the invention provides a new energy automobile intelligent battery management system based on deep learning, which comprises a data acquisition module, a data preprocessing module, a feature extraction module, a battery remaining life prediction model building module and a battery remaining life prediction module;
the data acquisition module is used for collecting historical data signals of the new energy automobile, wherein the historical data are historical charge and discharge cycle number data signals, historical battery capacity data signals and historical temperature data signals of the battery;
The data preprocessing module is used for sampling the historical battery data set to obtain a constructed sample data set, constructing an observation value vector, calculating the inter-sample mahalanobis distance between the sample mean value vector and the observation value vector, deleting abnormal values larger than the inter-sample mahalanobis distance, and carrying out normalization processing on the abnormal value removed data set to obtain a normalized battery data set;
The characteristic extraction module is used for extracting characteristics of data in the normalized battery data set to obtain a battery characteristic data set;
the model weight and the threshold value are updated by constructing an input gate, a forgetting gate and an output gate, the norms of the model parameters and all gradients are calculated by a loss function when the weight and the threshold value are updated, and the gradients are cut by comparing the norms of the gradients with preset threshold values;
The battery remaining life prediction module is used for outputting a battery remaining life prediction result by inputting a real-time battery data set into a battery remaining life prediction model to conduct prediction.
Further, the data preprocessing module specifically includes the following steps:
Constructing a sample data set, specifically comprising:
Obtaining a historical battery data set P A based on the data set, randomly extracting n sample data to obtain a sample data set R= { x 1,x2,…,xn };
an observation vector is constructed, wherein the observation vector comprises m elements, and the following formula is used:
Xi＝(xi1,xi2,...,xim)T；
Wherein X i represents an observation value vector, T represents a transposition operation, m represents m elements in the observation value vector, and i represents an index of sample data;
Calculating a sample mean vector, in particular by carrying out mean operation on the observation value vector to obtain the sample mean vector The formula used is as follows:
In the method, in the process of the invention, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Calculating a sample covariance matrix, specifically, calculating covariance of each dimension in a sample to obtain a sample covariance matrix Q, wherein the formula is as follows:
where Q represents the sample covariance matrix, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Calculating an inverse matrix of the sample covariance matrix to obtain an inverse matrix Q -1 of the sample covariance matrix;
and obtaining the mahalanobis distance W between the samples by carrying out distance operation on the observation value vector and the sample mean value vector, wherein the formula is as follows:
where W represents the inter-sample Markov distance, Q represents the sample covariance matrix, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Removing abnormal values, namely comparing the value of the mahalanobis distance W between samples with the value of the threshold T by setting the threshold T, and deleting sample data larger than the threshold T to obtain an abnormal value removing data set V;
The normalized battery dataset P C is obtained by performing data normalization processing on the outlier-removed dataset V, and the following formula is used:
Where P C denotes the normalized battery dataset, V denotes the outlier dataset, max denotes the maximum value in the outlier dataset, and min denotes the minimum value in the outlier dataset.
Further, the feature extraction module specifically performs feature extraction operation on the normalized battery data set to obtain a feature value and a corresponding tag, and specifically includes the following steps:
The rate of decay of the battery capacity is calculated, taking as an example a fixed decay of the battery capacity over each time interval, as follows:
The capacity fade of the battery was calculated using the following formula:
ΔC＝C(t1)-C(t1-1)；
Wherein ΔC represents the capacity reduction amount, C (t 1) represents the battery capacity at the current time, and C (t 1-1) represents the battery capacity at the previous time;
The rate of decay of the battery capacity is calculated using the following formula:
wherein Y represents the rate of decay of the battery capacity, ΔC represents the amount of capacity decay, and Δt1 represents the time interval;
calculating the change trend of the battery temperature, and obtaining the average value of the temperature change by constructing a sliding window method, wherein the average value is specifically as follows:
The average value of the temperature was calculated using the following formula:
Wherein D represents an average value of temperatures, F 1,F2,…,Fn represents temperature data in a window, and P represents a window size;
A series of temperature average values are obtained through window sliding, and the change trend of the temperature average values is analyzed to obtain the change trend of the battery temperature;
presetting a complete charge-discharge cycle process of a battery, and recording the times of charge-discharge cycles of the battery;
and constructing a battery characteristic data set, wherein the battery characteristic data set comprises characteristic values of a battery and corresponding labels, and the corresponding labels are attenuation rate of battery capacity, change trend of battery temperature and battery charge and discharge cycle times.
Further, the battery remaining life prediction model building module specifically comprises the following steps:
Presetting the input at the current moment as x t and the output at the moment immediately before the current moment as h t-1;
the input gate is updated using the following formula:
pt＝tanh(ωsxt+usht-1+bs；
Where p t denotes an input gate, tanh () denotes a hyperbolic tangent activation function, ω s denotes a weight of an input gate state, x t denotes an input at the current time, u s denotes a weight of a state immediately before the input gate, h t-1 denotes an output immediately before the current time, and b s denotes an offset of the input gate state;
the input gate cell state is updated using the following formula:
it＝σ(wixt+uiht-1+bi)；
Wherein i t represents the state of the input gate cell, σ represents the sigmoid activation function, w i represents the weight of the input gate, u i represents the weight of the input gate at the previous time, and b i represents the bias of the input gate;
the forgetting gate is updated using the following formula:
ft＝σ(ωfxt+ufht-1+bf)；
wherein f t represents a forgetting gate, σ represents a sigmoid activation function, ω f represents a weight of the forgetting gate, x t represents an input at a current time, u f represents a weight of the forgetting gate at a time immediately before, h t-1 represents an output at a time immediately before the current time, and b f represents a bias amount of the forgetting gate;
the amnestic portal cell state is updated using the following formula:
wherein s t represents the state of cells in the forgetting gate, i t represents the input gate, Representing a convolution operation, p t representing the state of the input gate, f t representing the forget gate, s t-1 representing the previous state of the forget gate;
the output gate is calculated using the following formula:
ot＝σ(ωoxt+uoht-1+bo)；
Where o t denotes an output gate, σ denotes a sigmoid activation function, ω o denotes a weight of the output gate, x t denotes an input at the current time, u o denotes a weight of the output gate immediately before, h t-1 denotes an output immediately before the current time, and b o denotes a bias amount of the output gate;
the output door cell state was calculated using the following formula:
Wherein h t represents the output door cell state, Representing a convolution operation, tanh () represents a hyperbolic tangent activation function, and s t represents the state of a forgetting gate;
The mean square error loss function is calculated using the following formula:
Wherein M represents a mean square error loss function, h t represents the state of an output gate, and y t represents the true value of the output;
Updating weights and thresholds through a gradient descent algorithm of a back propagation algorithm to obtain optimal weights and thresholds, wherein the method specifically comprises the following steps of:
the gradient of the output gate is updated using the following formula:
Where M represents a mean square error loss function, o t represents an output gate, h t represents a state of the output gate, tanh () represents a hyperbolic tangent activation function, s t represents a state of a forgetting gate, and σ' represents a derivative of a sigmoid activation function;
the gradient of the forgetting gate is updated using the following formula:
Wherein M represents a mean square error loss function, f t represents a forgetting gate, h t-1 represents an output immediately before the current time, σ' represents a derivative of a sigmoid activation function, and s t-1 represents a previous state of the forgetting gate;
the gradient of the input gate is updated using the following formula:
Wherein M represents a mean square error loss function, i t represents an input gate, p t represents a state of the input gate, tanh () represents a hyperbolic tangent activation function, s t represents a state of a forgetting gate, and σ' represents a derivative of a sigmoid activation function;
Setting a gradient vector g and a gradient threshold t of long-short term memory network prediction model parameters, and calculating norms of long-short term memory network prediction model gradients by using the following formula:
wherein R represents the norm of the gradient of the long-term memory network model, and g i represents the ith element of the gradient vector g;
when the norm of the long-term memory network prediction model gradient exceeds a threshold value, scaling the long-term memory network prediction model gradient, wherein the formula is as follows:
where g' represents the scaled gradient vector, t2 represents the gradient threshold, R represents the norm of the long-short term memory network prediction model gradient, and g represents the gradient vector.
Further, the battery remaining life prediction module specifically predicts by inputting a real-time battery data set into a battery remaining life prediction model, and outputs a battery remaining life prediction result.
Compared with the prior art, the invention has the beneficial effects that:
(1) Aiming at the problems that the data dimension has a large influence on the result due to partial characteristics and the acquired data contains abnormal values, so that the subsequent data processing is distorted, the abnormal value detection and data normalization algorithm is adopted, the sample mean value vector, the covariance matrix and the inter-sample mahalanobis distance between the observation value vector and the sample mean value vector are calculated through constructing the observation value vector, the abnormal values are detected and removed, the data is normalized, the data of different dimensions are unified into a specific range, the dimensional differences among the different dimensions are eliminated, and the influence of certain characteristics on the result is prevented from being too large.
(2) Aiming at the problems that the model parameter is updated excessively large when the long-short-term memory network prediction model updates the weight and the threshold value, so that the model parameter is updated excessively large, and the model stability is poor, convergence is difficult and gradient explosion is caused.
Drawings
FIG. 1 is a system block diagram of a new energy automobile intelligent battery management system based on deep learning provided by the invention;
FIG. 2 is a schematic flow diagram of a data preprocessing module;
fig. 3 is a flow chart of the feature extraction module.
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the invention; 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.
In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "top," "bottom," "inner," "outer," and the like indicate orientation or positional relationships based on those shown in the drawings, merely to facilitate description of the invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus should not be construed as limiting the invention.
Referring to fig. 1, the intelligent battery management system for a new energy automobile provided by the invention comprises a data acquisition module, a data preprocessing module, a feature extraction module, a battery remaining life prediction model building module and a battery remaining life prediction module;
the data acquisition module is used for collecting historical data signals of the new energy automobile, wherein the historical data are historical charge and discharge cycle number data signals, historical battery capacity data signals and historical temperature data signals of the battery;
the data preprocessing module is used for sampling the historical battery data set to obtain a sample data set, constructing an observation value vector, calculating the inter-sample mahalanobis distance between the sample mean value vector and the observation value vector, deleting abnormal values larger than the inter-sample mahalanobis distance, and carrying out normalization processing on the abnormal value removal data set to obtain a normalized battery data set;
The characteristic extraction module is used for extracting characteristics of data in the normalized battery data set to obtain a battery characteristic data set;
the model weight and the threshold value are updated by constructing an input gate, a forgetting gate and an output gate, the norms of the model parameters and all gradients are calculated by a loss function when the weight and the threshold value are updated, and the gradients are cut by comparing the norms of the gradients with preset threshold values;
The battery remaining life prediction module is used for outputting a battery remaining life prediction result by inputting a real-time battery data set into a battery remaining life prediction model to conduct prediction.
Referring to fig. 1, the second embodiment is based on the foregoing embodiment, and the data acquisition module specifically collects a historical data signal of the new energy automobile, where the historical data is a historical charge-discharge cycle number data signal, a historical battery capacity data signal, and a historical temperature data signal of the battery.
An embodiment III, referring to FIG. 1 and FIG. 2, is based on the foregoing embodiment, and the data preprocessing module specifically includes the following steps:
Constructing a sample data set, specifically comprising:
Obtaining a historical battery data set P A based on the data set, randomly extracting n sample data to obtain a sample data set R= { x 1,x2,…,xn };
an observation vector is constructed, wherein the observation vector comprises m elements, and the following formula is used:
Xi＝(xi1,xi2,...,xim)T；
Wherein X i represents an observation value vector, T represents a transposition operation, m represents m elements in the observation value vector, and i represents an index of sample data;
Calculating a sample mean vector, in particular by carrying out mean operation on the observation value vector to obtain the sample mean vector The formula used is as follows:
In the method, in the process of the invention, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Calculating a sample covariance matrix, specifically, calculating covariance of each dimension in a sample to obtain a sample covariance matrix Q, wherein the formula is as follows:
where Q represents the sample covariance matrix, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Calculating an inverse matrix of the sample covariance matrix to obtain an inverse matrix Q -1 of the sample covariance matrix;
and obtaining the mahalanobis distance W between the samples by carrying out distance operation on the observation value vector and the sample mean value vector, wherein the formula is as follows:
where W represents the inter-sample Markov distance, Q represents the sample covariance matrix, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Removing abnormal values, namely comparing the value of the mahalanobis distance W between samples with the value of the threshold T by setting the threshold T, and deleting sample data larger than the threshold T to obtain an abnormal value removing data set V;
The normalized battery dataset P C is obtained by performing data normalization processing on the outlier-removed dataset V, and the following formula is used:
Where P C denotes the normalized battery dataset, V denotes the outlier dataset, max denotes the maximum value in the outlier dataset, and min denotes the minimum value in the outlier dataset.
By executing the operations, aiming at the problems that the data dimension has a large influence on the result due to partial characteristics and the acquired data contains abnormal values, so that the subsequent data processing is distorted, the abnormal value detection and data normalization algorithm is adopted, the abnormal values are detected and removed by constructing the observed value vector, calculating the sample mean value vector, the covariance matrix and the inter-sample mahalanobis distance between the observed value vector and the sample mean value vector, and normalizing the data, so that the data of different dimensions are unified in a specific range, the dimensional difference among the different dimensions is eliminated, and the influence of certain characteristics on the result is prevented from being excessively large.
In a fourth embodiment, referring to fig. 1 and 3, the embodiment is based on the above embodiment, and the feature extraction module specifically obtains a feature value and a corresponding tag by performing feature extraction operation on a normalized battery data set, and specifically includes the following steps:
The rate of decay of the battery capacity is calculated, taking as an example a fixed decay of the battery capacity over each time interval, as follows:
The capacity fade of the battery was calculated using the following formula:
ΔC＝C(t1)-C(t1-1)；
Wherein ΔC represents the capacity reduction amount, C (t 1) represents the battery capacity at the current time, and C (t 1-1) represents the battery capacity at the previous time;
The rate of decay of the battery capacity is calculated using the following formula:
wherein Y represents the rate of decay of the battery capacity, ΔC represents the amount of capacity decay, and Δt1 represents the time interval;
calculating the change trend of the battery temperature, and obtaining the average value of the temperature change by constructing a sliding window method, wherein the average value is specifically as follows:
The average value of the temperature was calculated using the following formula:
Wherein D represents an average value of temperatures, F 1,F2,…,Fn represents temperature data in a window, and P represents a window size;
A series of temperature average values are obtained through window sliding, and the change trend of the temperature average values is analyzed to obtain the change trend of the battery temperature;
presetting a complete charge-discharge cycle process of a battery, and recording the times of charge-discharge cycles of the battery;
and constructing a battery characteristic data set, wherein the battery characteristic data set comprises characteristic values of a battery and corresponding labels, and the corresponding labels are attenuation rate of battery capacity, change trend of battery temperature and battery charge and discharge cycle times.
Fifth embodiment, referring to fig. 1, the present embodiment is based on the above embodiment, and the constructing a battery remaining life prediction model module specifically includes the following steps:
Presetting the input at the current moment as x t and the output at the moment immediately before the current moment as h t-1;
the input gate is updated using the following formula:
pt＝tanh(ωsxt+usht-1+bs；
Where p t denotes an input gate, tanh () denotes a hyperbolic tangent activation function, ω s denotes a weight of an input gate state, x t denotes an input at the current time, u s denotes a weight of a state immediately before the input gate, h t-1 denotes an output immediately before the current time, and b s denotes an offset of the input gate state;
the input gate cell state is updated using the following formula:
it＝σ(wixt+uiht-1+bi)；
Wherein i t represents the state of the input gate cell, σ represents the sigmoid activation function, w i represents the weight of the input gate, u i represents the weight of the input gate at the previous time, and b i represents the bias of the input gate;
the forgetting gate is updated using the following formula:
ft＝σ(ωfxt+ufht-1+bf)；
wherein f t represents a forgetting gate, σ represents a sigmoid activation function, ω f represents a weight of the forgetting gate, x t represents an input at a current time, u f represents a weight of the forgetting gate at a time immediately before, h t-1 represents an output at a time immediately before the current time, and b f represents a bias amount of the forgetting gate;
the amnestic portal cell state is updated using the following formula:
wherein s t represents the state of cells in the forgetting gate, i t represents the input gate, Representing a convolution operation, p t representing the state of the input gate, f t representing the forget gate, s t-1 representing the previous state of the forget gate;
the output gate is calculated using the following formula:
ot＝σ(ωoxt+uoht-1+bo)；
Where o t denotes an output gate, σ denotes a sigmoid activation function, ω o denotes a weight of the output gate, x t denotes an input at the current time, u o denotes a weight of the output gate immediately before, h t-1 denotes an output immediately before the current time, and b o denotes a bias amount of the output gate;
the output door cell state was calculated using the following formula:
Wherein h t represents the output door cell state, Representing a convolution operation, tanh () represents a hyperbolic tangent activation function, and s t represents the state of a forgetting gate;
The mean square error loss function is calculated using the following formula:
Wherein M represents a mean square error loss function, h t represents the state of an output gate, and y t represents the true value of the output;
Updating weights and thresholds through a gradient descent algorithm of a back propagation algorithm to obtain optimal weights and thresholds, wherein the method specifically comprises the following steps of:
the gradient of the output gate is updated using the following formula:
Where M represents a mean square error loss function, o t represents an output gate, h t represents a state of the output gate, tanh () represents a hyperbolic tangent activation function, s t represents a state of a forgetting gate, and σ' represents a derivative of a sigmoid activation function;
the gradient of the forgetting gate is updated using the following formula:
Wherein M represents a mean square error loss function, f t represents a forgetting gate, h t-1 represents an output immediately before the current time, σ' represents a derivative of a sigmoid activation function, and s t-1 represents a previous state of the forgetting gate;
the gradient of the input gate is updated using the following formula:
Wherein M represents a mean square error loss function, i t represents an input gate, p t represents a state of the input gate, tanh () represents a hyperbolic tangent activation function, s t represents a state of a forgetting gate, and σ' represents a derivative of a sigmoid activation function;
Setting a gradient vector g and a gradient threshold t of long-short term memory network prediction model parameters, and calculating norms of long-short term memory network prediction model gradients by using the following formula:
wherein R represents the norm of the gradient of the long-term memory network model, and g i represents the ith element of the gradient vector g;
when the norm of the long-term memory network prediction model gradient exceeds a threshold value, scaling the long-term memory network prediction model gradient, wherein the formula is as follows:
where g' represents the scaled gradient vector, t2 represents the gradient threshold, R represents the norm of the long-short term memory network prediction model gradient, and g represents the gradient vector.
By executing the operations, aiming at the problems that model parameters are updated excessively large when a long-short-term memory network prediction model updates weights and thresholds, so that model stability is poor, convergence is difficult and gradient explosion is caused.
In a sixth embodiment, referring to fig. 1, the present embodiment is based on the foregoing embodiment, and the battery remaining life prediction module specifically inputs the real-time battery data set into the battery remaining life prediction model to predict, and outputs a battery remaining life prediction result.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
The invention and its embodiments have been described above with no limitation, and the actual construction is not limited to the embodiments of the invention as shown in the drawings. In summary, if one of ordinary skill in the art is informed by this disclosure, a structural manner and an embodiment similar to the technical solution should not be creatively devised without departing from the gist of the present invention.
Claims (5)
1. New energy automobile intelligent battery management system based on degree of depth study, its characterized in that: the device comprises a data acquisition module, a data preprocessing module, a characteristic extraction module, a battery residual life prediction model building module and a battery residual life prediction module;
the data acquisition module is used for collecting historical data signals of the new energy automobile, wherein the historical data are historical charge and discharge cycle number data signals, historical battery capacity data signals and historical temperature data signals of the battery;
The data preprocessing module is used for sampling the historical battery data set to obtain a constructed sample data set, constructing an observation value vector, calculating the inter-sample mahalanobis distance between the sample mean value vector and the observation value vector, deleting abnormal values larger than the inter-sample mahalanobis distance, and carrying out normalization processing on the abnormal value removed data set to obtain a normalized battery data set;
The characteristic extraction module is used for extracting characteristics of data in the normalized battery data set to obtain a battery characteristic data set;
the model weight and the threshold value are updated by constructing an input gate, a forgetting gate and an output gate, the norms of the model parameters and all gradients are calculated by a loss function when the weight and the threshold value are updated, and the gradients are cut by comparing the norms of the gradients with preset threshold values;
The battery remaining life prediction module is used for inputting a real-time battery data set into a battery remaining life prediction model to predict, and outputting a battery remaining life prediction result; the battery remaining life prediction model building module specifically comprises the following steps:
Presetting the input at the current moment as x t and the output at the moment immediately before the current moment as h t-1;
the input gate is updated using the following formula:
pt＝tanh(ωsxt+usht-1+bs）；
Where p t denotes an input gate, tanh () denotes a hyperbolic tangent activation function, ω s denotes a weight of an input gate state, x t denotes an input at the current time, u s denotes a weight of a state immediately before the input gate, h t-1 denotes an output immediately before the current time, and b s denotes an offset of the input gate state;
the input gate cell state is updated using the following formula:
it＝σ(wixt+uiht-1+bi)；
Wherein i t represents the state of the input gate cell, σ represents the sigmoid activation function, w i represents the weight of the input gate, u i represents the weight of the input gate at the previous time, and b i represents the bias of the input gate;
the forgetting gate is updated using the following formula:
ft＝σ(ωfxt+ufht-1+bf)；
wherein f t represents a forgetting gate, σ represents a sigmoid activation function, ω f represents a weight of the forgetting gate, x t represents an input at a current time, u f represents a weight of the forgetting gate at a time immediately before, h t-1 represents an output at a time immediately before the current time, and b f represents a bias amount of the forgetting gate;
the amnestic portal cell state is updated using the following formula:
wherein s t represents the state of cells in the forgetting gate, i t represents the input gate, Representing a convolution operation, p t representing the state of the input gate, f t representing the forget gate, s t-1 representing the previous state of the forget gate;
the output gate is calculated using the following formula:
ot＝σ(ωoxt+uoht-1+bo)；
Where o t denotes an output gate, σ denotes a sigmoid activation function, ω o denotes a weight of the output gate, x t denotes an input at the current time, u o denotes a weight of the output gate immediately before, h t-1 denotes an output immediately before the current time, and b o denotes a bias amount of the output gate;
the output door cell state was calculated using the following formula:
Wherein h t represents the output door cell state, Representing a convolution operation, tanh () represents a hyperbolic tangent activation function, and s t represents the state of a forgetting gate;
The mean square error loss function is calculated using the following formula:
Wherein M represents a mean square error loss function, h t represents the state of an output gate, and y t represents the true value of the output;
Updating weights and thresholds through a gradient descent algorithm of a back propagation algorithm to obtain optimal weights and thresholds, wherein the method specifically comprises the following steps of:
the gradient of the output gate is updated using the following formula:
Where M represents a mean square error loss function, o t represents an output gate, h t represents a state of the output gate, tanh () represents a hyperbolic tangent activation function, s t represents a state of a forgetting gate, and σ' represents a derivative of a sigmoid activation function;
the gradient of the forgetting gate is updated using the following formula:
Wherein M represents a mean square error loss function, f t represents a forgetting gate, h t-1 represents an output immediately before the current time, σ' represents a derivative of a sigmoid activation function, and s t-1 represents a previous state of the forgetting gate;
the gradient of the input gate is updated using the following formula:
Wherein M represents a mean square error loss function, i t represents an input gate, p t represents a state of the input gate, tanh () represents a hyperbolic tangent activation function, s t represents a state of a forgetting gate, and σ' represents a derivative of a sigmoid activation function;
Setting a gradient vector g and a gradient threshold t of long-short term memory network prediction model parameters, and calculating norms of long-short term memory network prediction model gradients by using the following formula:
wherein R represents the norm of the gradient of the long-term memory network model, and g i represents the ith element of the gradient vector g;
when the norm of the long-term memory network prediction model gradient exceeds a threshold value, scaling the long-term memory network prediction model gradient, wherein the formula is as follows:
where g' represents the scaled gradient vector, t2 represents the gradient threshold, R represents the norm of the long-short term memory network prediction model gradient, and g represents the gradient vector.
2. The deep learning-based intelligent battery management system for a new energy automobile according to claim 1, wherein the intelligent battery management system is characterized in that: the data preprocessing module specifically comprises the following steps:
Constructing a sample data set, specifically comprising:
Obtaining a historical battery data set P A based on the data set, randomly extracting n sample data to obtain a sample data set R= { x 1,x2,…,xn };
an observation vector is constructed, wherein the observation vector comprises m elements, and the following formula is used:
Xi＝(xi1,xi2,...,xim)T；
Wherein X i represents an observation value vector, T represents a transposition operation, m represents m elements in the observation value vector, and i represents an index of sample data;
Calculating a sample mean vector, in particular by carrying out mean operation on the observation value vector to obtain the sample mean vector The formula used is as follows:
In the method, in the process of the invention, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Calculating a sample covariance matrix, specifically, calculating covariance of each dimension in a sample to obtain a sample covariance matrix Q, wherein the formula is as follows:
where Q represents the sample covariance matrix, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Calculating an inverse matrix of the sample covariance matrix to obtain an inverse matrix Q -1 of the sample covariance matrix;
and obtaining the mahalanobis distance W between the samples by carrying out distance operation on the observation value vector and the sample mean value vector, wherein the formula is as follows:
where W represents the inter-sample Markov distance, Q represents the sample covariance matrix, Representing a sample mean vector, X i representing an observation vector, i representing an index of sample data, n representing an amount of sample data;
Removing abnormal values, namely comparing the value of the mahalanobis distance W between samples with the value of the threshold T by setting the threshold T, and deleting sample data larger than the threshold T to obtain an abnormal value removing data set V;
The normalized battery dataset P C is obtained by performing data normalization processing on the outlier-removed dataset V, and the following formula is used:
Where P C denotes the normalized battery dataset, V denotes the outlier dataset, max denotes the maximum value in the outlier dataset, and min denotes the minimum value in the outlier dataset.
3. The deep learning-based intelligent battery management system for a new energy automobile according to claim 1, wherein the intelligent battery management system is characterized in that: the feature extraction module specifically performs feature extraction operation on the normalized battery data set to obtain a feature value and a corresponding tag, and specifically includes the following steps:
The rate of decay of the battery capacity is calculated, taking as an example a fixed decay of the battery capacity over each time interval, as follows:
The capacity fade of the battery was calculated using the following formula:
ΔC＝C(t1)-C(t1-1)；
Wherein ΔC represents the capacity reduction amount, C (t 1) represents the battery capacity at the current time, and C (t 1-1) represents the battery capacity at the previous time;
The rate of decay of the battery capacity is calculated using the following formula:
wherein Y represents the rate of decay of the battery capacity, ΔC represents the amount of capacity decay, and Δt1 represents the time interval;
calculating the change trend of the battery temperature, and obtaining the average value of the temperature change by constructing a sliding window method, wherein the average value is specifically as follows:
The average value of the temperature was calculated using the following formula:
Wherein D represents an average value of temperatures, F 1,F2,…,Fn represents temperature data in a window, and P represents a window size;
A series of temperature average values are obtained through window sliding, and the change trend of the temperature average values is analyzed to obtain the change trend of the battery temperature;
presetting a complete charge-discharge cycle process of a battery, and recording the times of charge-discharge cycles of the battery;
and constructing a battery characteristic data set, wherein the battery characteristic data set comprises characteristic values of a battery and corresponding labels, and the corresponding labels are attenuation rate of battery capacity, change trend of battery temperature and battery charge and discharge cycle times.
4. The deep learning-based intelligent battery management system for a new energy automobile according to claim 1, wherein the intelligent battery management system is characterized in that: the data acquisition module is used for collecting historical data signals of the new energy automobile, wherein the historical data are historical charge and discharge cycle number data signals, historical battery capacity data signals and historical temperature data signals of the battery.
5. The deep learning-based intelligent battery management system for a new energy automobile according to claim 1, wherein the intelligent battery management system is characterized in that: the battery remaining life prediction module is used for outputting a battery remaining life prediction result by inputting a real-time battery data set into a battery remaining life prediction model to conduct prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311572198.6A CN117473396B (en) | 2023-11-23 | 2023-11-23 | New energy automobile intelligent battery management system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311572198.6A CN117473396B (en) | 2023-11-23 | 2023-11-23 | New energy automobile intelligent battery management system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117473396A CN117473396A (en) | 2024-01-30 |
| CN117473396B true CN117473396B (en) | 2024-08-02 |
Family
ID=89629042
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311572198.6A Active CN117473396B (en) | 2023-11-23 | 2023-11-23 | New energy automobile intelligent battery management system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117473396B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112241608A (en) * | 2020-10-13 | 2021-01-19 | 国网湖北省电力有限公司电力科学研究院 | Lithium battery life prediction method based on LSTM network and transfer learning |
| CN116627194A (en) * | 2023-06-06 | 2023-08-22 | 常州大学 | WOA-LSTM temperature control method based on OLED thermal experiment |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106372745A (en) * | 2016-08-27 | 2017-02-01 | 山东电力工程咨询院有限公司 | Photovoltaic generating capacity prediction method based on fuzzy EBF (Elliptical Basis Function) network |
| US11131713B2 (en) * | 2018-02-21 | 2021-09-28 | Nec Corporation | Deep learning approach for battery aging model |
| CN111127251A (en) * | 2019-12-23 | 2020-05-08 | 南方电网科学研究院有限责任公司 | Attack identification method based on LSTM neural network and grid-connected interface device |
| US20230236251A1 (en) * | 2020-07-08 | 2023-07-27 | Panasonic Intellectual Property Management Co., Ltd. | Arithmetic system, battery inspection method, and battery inspection program |
| CN113361692B (en) * | 2021-06-28 | 2023-05-23 | 福建师范大学 | Lithium battery remaining life combined prediction method |
| CN115730525A (en) * | 2022-11-30 | 2023-03-03 | 上海应用技术大学 | Rail transit UPS storage battery health state prediction method |
| CN116894475A (en) * | 2023-06-09 | 2023-10-17 | 中国船舶集团有限公司系统工程研究院 | Gas turbine state early warning method based on neural network |
-
2023
- 2023-11-23 CN CN202311572198.6A patent/CN117473396B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112241608A (en) * | 2020-10-13 | 2021-01-19 | 国网湖北省电力有限公司电力科学研究院 | Lithium battery life prediction method based on LSTM network and transfer learning |
| CN116627194A (en) * | 2023-06-06 | 2023-08-22 | 常州大学 | WOA-LSTM temperature control method based on OLED thermal experiment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117473396A (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110263846B (en) | Fault diagnosis method based on fault data deep mining and learning | |
| CN107480440B (en) | Residual life prediction method based on two-stage random degradation modeling | |
| Giannakis et al. | A unifying maximum-likelihood view of cumulant and polyspectral measures for non-Gaussian signal classification and estimation | |
| CN111030889B (en) | Network traffic prediction method based on GRU model | |
| CN111999648A (en) | Lithium battery residual life prediction method based on long-term and short-term memory network | |
| CN117170294B (en) | Intelligent control method of satellite thermal control system based on space thermal environment prediction | |
| CN115758290A (en) | Fan gearbox high-speed shaft temperature trend early warning method based on LSTM | |
| CN115902647A (en) | Intelligent battery state monitoring method | |
| CN117851815B (en) | Real-time early warning method and system for safety state of switch cabinet | |
| CN113392507A (en) | Method for predicting residual life of lithium ion power battery | |
| CN116559667A (en) | Model training method and device, battery detection method and device, equipment and medium | |
| CN113610147A (en) | Multi-potential subspace information fusion earthquake short-term prediction method based on LSTM | |
| CN114818827A (en) | Non-invasive load decomposition method based on seq2point network | |
| CN116298984A (en) | Lithium ion battery capacity jump point and battery attenuation degree identification method | |
| CN115719294A (en) | Indoor pedestrian flow evacuation control method and system, electronic device and medium | |
| CN116106761A (en) | Real-time lithium ion battery electric quantity estimation method based on typical correlation analysis | |
| CN117473396B (en) | New energy automobile intelligent battery management system based on deep learning | |
| CN103616680B (en) | Mobile dim target tracking-before-detecting method based on discrete variable rate sampling | |
| CN110887652B (en) | Interactive multi-model detection method for vibration detection and displacement extraction of accelerometer | |
| CN117315585A (en) | Intelligent control system and method for electrolytic tank | |
| CN104933052A (en) | Data true value estimation method and data true value estimation device | |
| Fischer et al. | Increasing the performance of a training algorithm for local model networks | |
| CN111007418B (en) | Lithium battery residual life prediction method | |
| CN113357138A (en) | Method and device for predicting remaining service life of hydraulic pump and terminal equipment | |
| CN112668713A (en) | Intelligent energy consumption prediction method and system |
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 |