CN112215253A - Lithium battery energy storage system monitoring method and system based on noise reduction self-coding - Google Patents
Lithium battery energy storage system monitoring method and system based on noise reduction self-coding Download PDFInfo
- Publication number
- CN112215253A CN112215253A CN202010886479.9A CN202010886479A CN112215253A CN 112215253 A CN112215253 A CN 112215253A CN 202010886479 A CN202010886479 A CN 202010886479A CN 112215253 A CN112215253 A CN 112215253A
- Authority
- CN
- China
- Prior art keywords
- noise reduction
- lithium battery
- self
- reduction self
- encoder
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J7/00—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries
- H02J7/0047—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries with monitoring or indicating devices or circuits
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Power Engineering (AREA)
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
- Secondary Cells (AREA)
Abstract
The invention provides a lithium battery energy storage system monitoring method and system based on noise reduction self-coding, wherein the method comprises the following steps: calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively; constructing a data sample based on the mutual information value; constructing a stack noise reduction self-coding network based on the data samples; establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack denoising self-coding network based on a regression method; calculating a predicted value of the terminal voltage of a lithium battery in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model; and judging whether to give an alarm or not according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system. The invention constructs a stack noise reduction self-coding network by utilizing the self-monitoring learning process of the self-encoder, excavates deep-level relation among characteristic variables, performs dimension reduction processing, and removes noise and redundant characteristic variables.
Description
Technical Field
The invention relates to the technical field of monitoring of battery energy storage systems, in particular to a lithium battery energy storage system monitoring method and system based on noise reduction self-coding.
Background
A battery energy storage system is an electrochemical energy storage method. The technical application of lithium batteries in energy storage mainly surrounds the fields of power grid energy storage, base station standby power supplies, electric vehicle light storage type charging stations and the like.
The working principle of energy storage of the lithium ion battery is as follows: when the battery is charged, lithium ions are generated on the positive electrode of the battery, and the generated lithium ions move to the negative electrode through the electrolyte. And the carbon in the negative electrode is of a layered structure and has a plurality of micropores, so that lithium ions reaching the negative electrode are inserted into the micropores of the carbon layer, and the more the lithium ions are inserted, the higher the charge capacity is. Similarly, when the battery is discharged, lithium ions embedded in the negative carbon layer are extracted and move back to the positive electrode. The more lithium ions that return to the positive electrode, the higher the discharge capacity.
The lithium battery has many advantages in the field of energy storage, long cycle life, relatively high energy density, low self-discharge rate, environmental protection and the like. In recent years, with the increase of the demand of electric power, the scale of an electric power system is gradually enlarged, the complexity of an electric network is also gradually increased, meanwhile, the demand of the state on an energy storage technology is also gradually increased due to the fact that the state develops a smart electric network, renewable energy sources and distributed energy sources, and the development of the state supports the energy storage greatly, so that the research and development and application of technologies such as accelerating large-scale energy storage are provided, and the lithium battery energy storage system plays an increasingly important role. However, the battery is damaged due to overcharge or overdischarge and other reasons, the safety of the battery is problematic due to factors such as poor health condition of the lithium battery, unbalanced charging of the single lithium battery and the like, and the monitoring system can find out the problems and apply control through a remote end, so that the safety and reliability of the energy storage system are improved.
Disclosure of Invention
Based on the above, the invention aims to provide a lithium battery energy storage system monitoring method and system based on noise reduction self-coding so as to realize real-time monitoring of the lithium battery energy storage system.
In order to achieve the above object, the present invention provides a method for monitoring a lithium battery energy storage system based on noise reduction self-coding, wherein the method comprises:
step S1: calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively;
step S2: constructing a data sample based on the mutual information value;
step S3: constructing a stack noise reduction self-coding network based on the data samples;
step S4: establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack denoising self-coding network based on a regression method;
step S5: calculating a predicted value of the terminal voltage of a lithium battery in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model;
step S6: and judging whether to give an alarm or not according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system.
Optionally, the constructing a stack denoising self-coding network based on data samples specifically includes:
step S31: training a first noise-reducing self-encoder based on the data samples;
step S32: training a second noise reduction self-encoder by taking the output of the trained hidden layer of the first noise reduction self-encoder as the input of an input layer of the second noise reduction self-encoder;
step S33: and stacking the trained first noise reduction self-encoder and the trained second noise reduction self-encoder to form a stack noise reduction self-encoding network.
Optionally, the training of the first denoising autoencoder based on the data samples specifically includes:
step S311: the initialization weight, the neuron number and the activation function of the input layer of the first noise reduction self-encoder are given;
step S312: selecting a plurality of characteristic variables from the data sample;
step S313: training a first noise reduction self-encoder based on the initialization weights of the first noise reduction self-encoder input layer and a plurality of the feature variables;
step S314: reversely solving two weight variable quantities of the first noise reduction self-encoder by using an error function and a BP algorithm;
step S315: calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method;
step S316: judging whether the iteration times are larger than or equal to the maximum iteration times; if the number of iterations is greater than or equal to the maximum number of iterations, "step S32" is performed; and if the iteration number is less than the maximum iteration number, taking the weight required by the next training as the initialization weight of the input layer of the first noise reduction self-encoder, and returning to the step S312.
Optionally, the establishing a regression prediction model of the lithium battery terminal voltage according to the stack denoising self-coding network based on the regression method specifically includes:
step S41: based on a regression method, inputting the data sample into the stack denoising self-coding network to obtain a vector of an output layer of the stack denoising self-coding network;
step S42: extracting vectors of a stack denoising self-coding network output layer through matrix transformation to obtain parameters related to the end voltage of the lithium battery; the mutual information values are strongly correlated within the range of 0.6-1;
step S43: and establishing a regression prediction model of the lithium battery terminal voltage according to the lithium battery terminal voltage and the parameters related to the lithium battery terminal voltage.
Optionally, the determining whether to alarm according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system specifically includes:
step S61: comparing the predicted value of the terminal voltage of the lithium battery with the terminal voltage of the lithium battery to obtain an error;
step S62: judging whether the error is within a set range; if the error is within the set range, indicating that the lithium battery in the battery energy storage system works normally; and if the error is not in the set range, the monitoring system gives an alarm to the error.
The invention also provides a lithium battery energy storage system monitoring system based on noise reduction self-coding, which comprises:
the mutual information value calculating module is used for calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively;
the data sample construction module is used for constructing a data sample based on the mutual information value;
the stack denoising self-coding network construction module is used for constructing a stack denoising self-coding network based on the data samples;
the regression prediction model determining module is used for establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack noise reduction self-coding network based on a regression method;
the lithium battery terminal voltage predicted value determining module is used for calculating a lithium battery terminal voltage predicted value in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model;
and the judging and alarming module is used for judging whether to alarm or not according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system.
Optionally, the stack denoising self-coding network constructing module specifically includes:
a first training unit for training a first noise reduction self-encoder based on the data samples;
the second training unit is used for taking the output of the trained hidden layer of the first noise reduction self-encoder as the input of the input layer of the second noise reduction self-encoder and training the second noise reduction self-encoder;
and the stacking unit is used for stacking the trained first noise reduction self-encoder and the trained second noise reduction self-encoder to form a stack noise reduction self-encoding network.
Optionally, the first training unit specifically includes:
a first given subunit for giving an initialization weight, a neuron number and an activation function of the first noise reduction self-encoder input layer;
a selecting subunit, configured to select a plurality of characteristic variables from the data sample;
a first training subunit training a first noise reduction self-encoder based on the initialization weight of the first noise reduction self-encoder input layer and the plurality of feature variables;
the first weight variation determining subunit is used for reversely solving two weight variations of the first noise reduction self-encoder by utilizing an error function and a BP algorithm;
the first weight determining subunit is used for calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method;
the first judgment subunit is used for judging whether the iteration times are greater than or equal to the maximum iteration times; if the iteration number is greater than or equal to the maximum iteration number, executing a second training unit; and if the iteration times are less than the maximum iteration times, taking the weight required by the next training as the initialization weight of the input layer of the first noise reduction self-encoder, and returning to the 'selection subunit'.
Optionally, the regression prediction model determining module specifically includes:
the vector determining unit is used for inputting the data sample into the stack denoising self-coding network based on a regression method to obtain a vector of an output layer of the stack denoising self-coding network;
the extraction unit is used for extracting the vectors of the stack denoising self-coding network output layer through matrix transformation to obtain parameters related to the lithium battery end voltage intensity; the mutual information values are strongly correlated within the range of 0.6-1;
and the regression prediction model determining unit is used for establishing a regression prediction model of the lithium battery terminal voltage according to the lithium battery terminal voltage and the parameter related to the lithium battery terminal voltage.
Optionally, the alarm determining module specifically includes:
the comparison unit is used for comparing the predicted value of the terminal voltage of the lithium battery with the terminal voltage of the lithium battery to obtain an error;
the judging unit is used for judging whether the error is in a set range or not; if the error is within the set range, indicating that the lithium battery in the battery energy storage system works normally; and if the error is not in the set range, the monitoring system gives an alarm to the error.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a lithium battery energy storage system monitoring method and system based on noise reduction self-coding, wherein the method comprises the following steps: calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively; constructing a data sample based on the mutual information value; constructing a stack noise reduction self-coding network based on the data samples; establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack denoising self-coding network based on a regression method; calculating a predicted value of the terminal voltage of a lithium battery in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model; and judging whether to give an alarm or not according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system. The invention constructs a stack noise reduction self-coding network by utilizing the self-monitoring learning process of the self-encoder, excavates deep-level relation among characteristic variables, performs dimension reduction processing, and removes noise and redundant characteristic variables.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a lithium battery energy storage system monitoring method based on noise reduction self-coding according to an embodiment of the present invention;
FIG. 2 is a structural diagram of a lithium battery energy storage system monitoring system based on noise reduction self-coding according to an embodiment of the present invention;
fig. 3 is a data diagram of mutual information values of the lithium battery energy storage system in the embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a lithium battery energy storage system monitoring method and system based on noise reduction self-coding so as to realize real-time monitoring of a lithium battery energy storage system.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
As shown in fig. 1, the invention provides a method for monitoring a lithium battery energy storage system based on noise reduction self-coding, which comprises the following steps:
step S1: and (3) calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively.
Step S2: and constructing a data sample based on the mutual information value.
Step S3: and constructing a stack denoising self-coding network based on the data samples.
Step S4: and establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack denoising self-coding network based on a regression method.
Step S5: and calculating the predicted value of the terminal voltage of the lithium battery in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model.
Step S6: and judging whether to give an alarm or not according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system.
The individual steps are discussed in detail below:
step S1: calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively, and specifically comprising the following steps:
step S11: and acquiring the terminal voltage, the measurement parameter set and the physical parameter set of the lithium battery in the lithium battery energy storage system. The measurement parameter group comprises charge-discharge efficiency, charge multiplying power, discharge multiplying power, charge state and battery health state; the physical parameter set comprises battery capacity, rated voltage, maximum discharge rate, discharge cut-off voltage, charge and discharge current, standard working temperature, battery temperature, cycle life, consistency, maximum charge current and maximum discharge current.
Step S12: and calculating mutual information values between the terminal voltage of the lithium battery and the measurement parameter group and/or the physical parameter group respectively by using a mutual information variable selection method.
And setting the battery terminal voltage as X, and the measurement parameter group or the physical parameter group as Y. If X and Y are two discrete random variables, the mutual information value between X and Y is defined according to the following formula:
wherein I (X, Y) represents a mutual information value between the battery terminal voltage X and the measurement parameter set or the physical parameter set Y, β represents one parameter of the measurement parameter set or the physical parameter set Y, α represents one parameter of the battery terminal voltage X, P (α, β) represents a joint probability distribution function of α and β, P (α) represents an edge probability distribution function of α, and P (β) represents an edge probability distribution function of β.
The above equation applies to the case where the learner is a classification model. When the learner is a regression model, i.e., X and Y are continuous random variables, the sum number in the above equation needs to be replaced by the integral number:
step S2: constructing a data sample based on the mutual information value, specifically comprising:
step S21: judging whether the mutual information value is larger than a set threshold value or not; and if the mutual information value is larger than a set threshold value, taking the terminal voltage of the lithium battery with the mutual information value larger than the set threshold value and each parameter in the measurement parameter group and/or the physical parameter group as characteristic variables.
Step S22: and taking a plurality of feature variables obtained in multiple times as data samples.
Step S3: the method for constructing the stack denoising self-coding network based on the data samples specifically comprises the following steps:
step S31: training a first noise reduction self-encoder based on data samples, specifically comprising:
step S311: the initialization weights for the first noise reduction self-encoder input layer are given.
Step S312: a plurality of feature variables are selected from the data samples.
Step S313: training a first noise-reducing self-encoder based on the initialization weights of the first noise-reducing self-encoder input layer and the plurality of feature variables.
Step S314: and reversely solving two weight variable quantities of the first noise reduction self-encoder by using an error function and a BP algorithm.
Step S315: and calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method.
Step S316: judging whether the iteration times are larger than or equal to the maximum iteration times; if the number of iterations is greater than or equal to the maximum number of iterations, "step S32" is performed; and if the iteration number is less than the maximum iteration number, taking the weight required by the next training as the initialization weight of the input layer of the first noise reduction self-encoder, and returning to the step S312.
The output formula of the output layer of the first noise reduction self-encoder after training is as follows:
wherein z represents the output of the output layer of the first noise-reducing self-encoder after training, f (-) represents a sigmoid function, as1t1Representing initialization weights of said first noise reduction self-encoder input layer, b1Representing the first noise-reduced self-encodingBias vector of device, as2t2Representing the trained hidden layer weights of said first noise-reducing self-encoder, b2Representing a trained bias vector of the first noise-reducing self-encoder,
representing the random noise adding processing to obtain the characteristic variable.
The error function is:
wherein L represents an error of the first noise-reducing auto-encoder, i represents a feature variable number, sj represents a plurality of input feature variables, and z represents an output of an output layer of the first noise-reducing auto-encoder after training.
Step S32: and training a second noise reduction self-encoder by taking the trained output of the hidden layer of the first noise reduction self-encoder as the input of the input layer of the second noise reduction self-encoder.
Step S321: and taking the weight required by the next training obtained by the last training of the first noise reduction self-encoder as the initialization weight of the input layer of the second noise reduction self-encoder, and giving the neuron number and the activation function of the second noise reduction self-encoder.
Step S322: and taking the trained output of the hidden layer of the first noise reduction self-encoder as the input of the input layer of the second noise reduction self-encoder.
Step S323: training a second noise-reducing self-encoder based on the initialization weight of the second noise-reducing self-encoder and the trained output of the hidden layer of the first noise-reducing self-encoder.
Step S324: reversely solving two weight variable quantities of a second noise reduction self-encoder by using an error function and a BP algorithm;
step S325: calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method;
step S326: judging whether the iteration times are larger than or equal to the maximum iteration times; if the number of iterations is greater than or equal to the maximum number of iterations, "step S33" is performed; and if the iteration number is less than the maximum iteration number, taking the weight required by the next training as the initialization weight of the input layer of the second noise reduction self-encoder, and returning to the step S323.
The output formula of the output layer of the second noise reduction self-encoder after training is the same as the solving method of the output formula of the output layer of the first noise reduction self-encoder after training, and the error function of the second noise reduction self-encoder is the same as the solving method of the error function of the first noise reduction self-encoder, and is not repeated here.
The maximum number of iterations in the above embodiment is 50, i.e., the training is stopped until the 50 th iteration.
Step S33: and stacking the trained first noise reduction self-encoder and the trained second noise reduction self-encoder to form a stack noise reduction self-encoding network. Specifically, the input layer and the hidden layer of the first noise reduction self-encoder after training are reserved, and the hidden layer of the second noise reduction self-encoder after training is used as an output layer to form a stack noise reduction self-encoding network.
Step S4: based on a regression method, establishing a regression prediction model of the terminal voltage of the lithium battery according to a stack denoising self-coding network, which specifically comprises the following steps:
step S41: and based on a regression method, inputting the data sample into the stack noise reduction self-coding network to obtain a vector of an output layer of the stack noise reduction self-coding network.
Step S42: extracting vectors of a stack denoising self-coding network output layer through matrix transformation to obtain parameters related to the end voltage of the lithium battery; the mutual information value ranges between 0 and 1, and the invention divides the mutual information value into three degrees, namely weak correlation from 0 to 0.3, medium correlation from 0.3 to 0.6 and strong correlation from 0.6 to 1.
Step S43: and establishing a regression prediction model of the lithium battery terminal voltage according to the lithium battery terminal voltage and the parameters related to the lithium battery terminal voltage.
The regression method is a support vector method or a least square method.
Step S5: and calculating a predicted value of the terminal voltage of the lithium battery in the battery energy storage system according to the stack denoising self-coding network and the regression prediction model, and specifically calculating the predicted value of the terminal voltage of the lithium battery in the battery energy storage system according to the characteristic variable, the stack denoising self-coding network and the regression prediction model.
Step S6: judging whether to give an alarm according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system, and specifically comprising the following steps:
step S61: comparing the predicted value of the terminal voltage of the lithium battery with the terminal voltage of the lithium battery to obtain an error; the terminal voltage of the lithium battery is the actual terminal voltage value of the lithium battery.
Step S62: judging whether the error is within a set range; if the error is within the set range, indicating that the lithium battery in the battery energy storage system works normally; and if the error is not within the set range, the monitoring system gives an alarm to the error so as to realize the real-time monitoring of the lithium battery in the battery energy storage system.
As shown in fig. 2, the present invention further provides a noise reduction self-coding based monitoring system for a lithium battery energy storage system, where the system includes:
and the mutual information value calculating module 1 is used for calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively.
And the data sample construction module 2 constructs a data sample based on the mutual information value.
And the stack denoising self-coding network constructing module 3 is used for constructing a stack denoising self-coding network based on the data samples.
And the regression prediction model determining module 4 is used for establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack noise reduction self-coding network based on a regression method.
And the lithium battery terminal voltage predicted value determining module 5 is used for calculating the lithium battery terminal voltage predicted value in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model.
And the judgment warning module 6 is used for judging whether to warn according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system.
As an implementation manner, the mutual information value calculating module 1 of the present invention specifically includes:
the acquisition unit is used for acquiring the terminal voltage of the lithium battery, the measurement parameter set and the physical parameter set in the lithium battery energy storage system; the measurement parameter group comprises charge-discharge efficiency, charge multiplying power, discharge multiplying power, charge state and battery health state; the physical parameter set comprises battery capacity, rated voltage, maximum discharge rate, discharge cut-off voltage, charge and discharge current, standard working temperature, battery temperature, cycle life, consistency, maximum charge current and maximum discharge current.
And the mutual information value calculation unit is used for calculating mutual information values between the terminal voltage of the lithium battery and the measurement parameter group and/or the physical parameter group respectively by using a mutual information variable selection method.
As an embodiment, the data sample construction module 2 of the present invention specifically includes:
the screening unit is used for judging whether the mutual information value is larger than a set threshold value or not; and if the mutual information value is larger than a set threshold value, taking the terminal voltage of the lithium battery with the mutual information value larger than the set threshold value and each parameter in the measurement parameter group and/or the physical parameter group as characteristic variables.
And the data sample determining unit is used for taking a plurality of feature variables obtained by a plurality of times as data samples.
As an implementation manner, the stack denoising self-coding network constructing module 3 of the present invention specifically includes:
and the first training unit trains the first noise reduction self-encoder based on the data samples.
And the second training unit is used for taking the trained output of the hidden layer of the first noise reduction self-encoder as the input of the input layer of the second noise reduction self-encoder to train the second noise reduction self-encoder.
And the stacking unit is used for stacking the trained first noise reduction self-encoder and the trained second noise reduction self-encoder to form a stack noise reduction self-encoding network.
As an embodiment, the first training unit of the present invention specifically includes:
a first given subunit for giving the initialization weight, the number of neurons, and the activation function of the first noise reduction self-encoder input layer.
And the selecting subunit is used for selecting a plurality of characteristic variables from the data sample.
And the first training subunit trains the first noise reduction self-encoder based on the initialization weight of the input layer of the first noise reduction self-encoder and a plurality of characteristic variables.
And the first weight variation determining subunit is used for reversely solving the two weight variations of the first noise reduction self-encoder by using an error function and a BP algorithm.
And the first weight determining subunit is used for calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method.
The first judgment subunit is used for judging whether the iteration times are greater than or equal to the maximum iteration times; if the iteration number is greater than or equal to the maximum iteration number, executing a second training unit; and if the iteration times are less than the maximum iteration times, taking the weight required by the next training as the initialization weight of the input layer of the first noise reduction self-encoder, and returning to the 'selection subunit'.
As an embodiment, the second training unit of the present invention specifically includes:
and the second given subunit is used for taking the weight required by the next training obtained by the last training of the first noise reduction self-encoder as the initialization weight of the input layer of the second noise reduction self-encoder, and giving the number of neurons and the activation function of the second noise reduction self-encoder.
And the evaluation subunit is used for taking the trained output of the hidden layer of the first noise reduction self-encoder as the input of the input layer of the second noise reduction self-encoder.
And the second training subunit trains the second noise reduction self-encoder based on the initialization weight of the second noise reduction self-encoder and the trained output of the hidden layer of the first noise reduction self-encoder.
And the second weight variation determining subunit is used for reversely solving the two weight variations of the second noise reduction self-encoder by using the error function and the BP algorithm.
And the second weight determining subunit is used for calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method.
The second judgment subunit is used for judging whether the iteration times are greater than or equal to the maximum iteration times; if the iteration number is greater than or equal to the maximum iteration number, executing a 'stacking unit'; and if the iteration times are less than the maximum iteration times, taking the weight required by the next training as the initialization weight of the input layer of the second noise reduction self-encoder, and returning to the second training subunit.
As an embodiment, the regression prediction model determining module 4 of the present invention specifically includes:
and the vector determining unit is used for inputting the data sample into the stack denoising self-coding network based on a regression method to obtain a vector of an output layer of the stack denoising self-coding network.
The extraction unit is used for extracting the vectors of the stack denoising self-coding network output layer through matrix transformation to obtain parameters related to the lithium battery end voltage intensity; the mutual information values are strongly correlated in the range of 0.6-1.
And the regression prediction model determining unit is used for establishing a regression prediction model of the lithium battery terminal voltage according to the lithium battery terminal voltage and the parameter related to the lithium battery terminal voltage.
As an implementation manner, the judgment and alarm module 6 of the present invention specifically includes:
and the comparison unit is used for comparing the predicted value of the terminal voltage of the lithium battery with the terminal voltage of the lithium battery to obtain an error.
The judging unit is used for judging whether the error is in a set range or not; if the error is within the set range, indicating that the lithium battery in the battery energy storage system works normally; and if the error is not in the set range, the monitoring system gives an alarm to the error.
Specific examples are:
as shown in fig. 3, it can be seen that the mutual information values of several features such as capacity and state of charge are relatively large, where the threshold value is set to be 0.3, and a few features having mutual information values higher than 0.3 are screened out, and are considered as features that contribute high to the tag information amount. Finally, the characteristic variables obtained by screening are as follows: capacity, standard operating temperature, discharge rate, state of charge, battery health, and cycle life. And testing 100 characteristic variables to form a data sample by using the SCADA.
The invention adopts a stack noise reduction self-coding network structure consisting of two noise reduction self-coders. The self-encoder that makes an uproar falls is exactly on the basis of automatic encoder, adds the noise to the input data of input layer in order to prevent the overfitting problem that produces, can let the self-encoding improve the ability of eliminating the noise when studying like this, and the viability is stronger under the abnormal condition. The noise reduction self-encoder is regarded as being composed of an encoder composed of an input layer and a hidden layer and a decoder composed of a hidden layer and an output layer.
The self-coding network is a shallow neural network, the purpose of the self-coding network is to ensure that the input and the output are consistent as much as possible, the parameters of the stack noise reduction self-coding network are required to be initialized, the input layer is 6-dimensional data, the hidden layer is set to be 4-dimensional data, the output layer is 6-dimensional data, and then the network structure of the first DAE is 6-4-6. The network structure of the second DAE is set to 4-3-4, then the stacked SDAE network structure is 6-4-3.
Based on a regression method, data are written into the input of the SDAE, vectors output by a second output layer can be obtained through the same steps as those in training, and the characteristics in the output layer in the second network are extracted through matrix transformation, namely the influence factors which are required by the patent and strongly related to the voltage V at the lithium battery terminal. And after the vectors are extracted and converted into the influence factors related to the voltage intensity of the lithium battery terminal through matrix transformation, establishing a regression prediction model of the voltage of the lithium battery terminal according to the voltage v of the lithium battery terminal and the strongly related influence factors thereof.
And calculating a predicted value a of the terminal voltage of the lithium battery based on the characteristic variable, the stack noise reduction self-coding network and the regression prediction model.
The predicted value a of the terminal voltage of the lithium battery and the actual value b of the terminal voltage of the lithium battery are counted and compared to obtain an error value c, and if the error value of the two values is within a specified error range of a system, normal work can be considered; if the numerical errors of the two are not within the error range specified by the system, the monitoring system gives an alarm to the error, and then the monitoring system is overhauled by the staff, so that the real-time monitoring of the terminal voltage of the lithium battery is realized.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A monitoring method of a lithium battery energy storage system based on noise reduction self-coding is characterized by comprising the following steps:
step S1: calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively;
step S2: constructing a data sample based on the mutual information value;
step S3: constructing a stack noise reduction self-coding network based on the data samples;
step S4: establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack denoising self-coding network based on a regression method;
step S5: calculating a predicted value of the terminal voltage of a lithium battery in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model;
step S6: and judging whether to give an alarm or not according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system.
2. The method for monitoring the lithium battery energy storage system based on the noise reduction self-coding as claimed in claim 1, wherein the constructing of the stack noise reduction self-coding network based on the data samples specifically comprises:
step S31: training a first noise-reducing self-encoder based on the data samples;
step S32: training a second noise reduction self-encoder by taking the output of the trained hidden layer of the first noise reduction self-encoder as the input of an input layer of the second noise reduction self-encoder;
step S33: and stacking the trained first noise reduction self-encoder and the trained second noise reduction self-encoder to form a stack noise reduction self-encoding network.
3. The method for monitoring the lithium battery energy storage system based on the noise reduction self-coding as claimed in claim 2, wherein the training of the first noise reduction self-coder based on the data sample specifically comprises:
step S311: the initialization weight, the neuron number and the activation function of the input layer of the first noise reduction self-encoder are given;
step S312: selecting a plurality of characteristic variables from the data sample;
step S313: training a first noise reduction self-encoder based on the initialization weights of the first noise reduction self-encoder input layer and a plurality of the feature variables;
step S314: reversely solving two weight variable quantities of the first noise reduction self-encoder by using an error function and a BP algorithm;
step S315: calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method;
step S316: judging whether the iteration times are larger than or equal to the maximum iteration times; if the number of iterations is greater than or equal to the maximum number of iterations, "step S32" is performed; and if the iteration number is less than the maximum iteration number, taking the weight required by the next training as the initialization weight of the input layer of the first noise reduction self-encoder, and returning to the step S312.
4. The method for monitoring the lithium battery energy storage system based on the noise reduction self-coding as claimed in claim 1, wherein the establishing of the regression prediction model of the lithium battery terminal voltage according to the stack noise reduction self-coding network based on the regression method specifically comprises:
step S41: based on a regression method, inputting the data sample into the stack denoising self-coding network to obtain a vector of an output layer of the stack denoising self-coding network;
step S42: extracting vectors of a stack denoising self-coding network output layer through matrix transformation to obtain parameters related to the end voltage of the lithium battery; the mutual information values are strongly correlated within the range of 0.6-1;
step S43: and establishing a regression prediction model of the lithium battery terminal voltage according to the lithium battery terminal voltage and the parameters related to the lithium battery terminal voltage.
5. The method for monitoring the lithium battery energy storage system based on the noise reduction self-coding as claimed in claim 1, wherein the step of judging whether to alarm according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system specifically comprises the steps of:
step S61: comparing the predicted value of the terminal voltage of the lithium battery with the terminal voltage of the lithium battery to obtain an error;
step S62: judging whether the error is within a set range; if the error is within the set range, indicating that the lithium battery in the battery energy storage system works normally; and if the error is not in the set range, the monitoring system gives an alarm to the error.
6. A lithium battery energy storage system monitoring system based on noise reduction self-coding is characterized in that the system comprises:
the mutual information value calculating module is used for calculating mutual information values between the terminal voltage of the lithium battery and each parameter in the measurement parameter group and/or the physical parameter group respectively;
the data sample construction module is used for constructing a data sample based on the mutual information value;
the stack denoising self-coding network construction module is used for constructing a stack denoising self-coding network based on the data samples;
the regression prediction model determining module is used for establishing a regression prediction model of the terminal voltage of the lithium battery according to the stack noise reduction self-coding network based on a regression method;
the lithium battery terminal voltage predicted value determining module is used for calculating a lithium battery terminal voltage predicted value in the battery energy storage system according to the stack noise reduction self-coding network and the regression prediction model;
and the judging and alarming module is used for judging whether to alarm or not according to the predicted value of the terminal voltage of the lithium battery in the battery energy storage system.
7. The lithium battery energy storage system monitoring system based on noise reduction and self coding of claim 6, wherein the stack noise reduction and self coding network building module specifically comprises:
a first training unit for training a first noise reduction self-encoder based on the data samples;
the second training unit is used for taking the output of the trained hidden layer of the first noise reduction self-encoder as the input of the input layer of the second noise reduction self-encoder and training the second noise reduction self-encoder;
and the stacking unit is used for stacking the trained first noise reduction self-encoder and the trained second noise reduction self-encoder to form a stack noise reduction self-encoding network.
8. The lithium battery energy storage system monitoring system based on noise reduction self-coding of claim 7, wherein the first training unit specifically comprises:
a first given subunit for giving an initialization weight, a neuron number and an activation function of the first noise reduction self-encoder input layer;
a selecting subunit, configured to select a plurality of characteristic variables from the data sample;
a first training subunit training a first noise reduction self-encoder based on the initialization weight of the first noise reduction self-encoder input layer and the plurality of feature variables;
the first weight variation determining subunit is used for reversely solving two weight variations of the first noise reduction self-encoder by utilizing an error function and a BP algorithm;
the first weight determining subunit is used for calculating the weight required by the next training according to the two weight variable quantities by adopting a gradient descent method;
the first judgment subunit is used for judging whether the iteration times are greater than or equal to the maximum iteration times; if the iteration number is greater than or equal to the maximum iteration number, executing a second training unit; and if the iteration times are less than the maximum iteration times, taking the weight required by the next training as the initialization weight of the input layer of the first noise reduction self-encoder, and returning to the 'selection subunit'.
9. The system for monitoring the lithium battery energy storage system based on the noise reduction self-coding as claimed in claim 6, wherein the regression prediction model determining module specifically comprises:
the vector determining unit is used for inputting the data sample into the stack denoising self-coding network based on a regression method to obtain a vector of an output layer of the stack denoising self-coding network;
the extraction unit is used for extracting the vectors of the stack denoising self-coding network output layer through matrix transformation to obtain parameters related to the lithium battery end voltage intensity; the mutual information values are strongly correlated within the range of 0.6-1;
and the regression prediction model determining unit is used for establishing a regression prediction model of the lithium battery terminal voltage according to the lithium battery terminal voltage and the parameter related to the lithium battery terminal voltage.
10. The lithium battery energy storage system monitoring system based on noise reduction self-coding of claim 6, wherein the judgment warning module specifically comprises:
the comparison unit is used for comparing the predicted value of the terminal voltage of the lithium battery with the terminal voltage of the lithium battery to obtain an error;
the judging unit is used for judging whether the error is in a set range or not; if the error is within the set range, indicating that the lithium battery in the battery energy storage system works normally; and if the error is not in the set range, the monitoring system gives an alarm to the error.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010886479.9A CN112215253A (en) | 2020-08-28 | 2020-08-28 | Lithium battery energy storage system monitoring method and system based on noise reduction self-coding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010886479.9A CN112215253A (en) | 2020-08-28 | 2020-08-28 | Lithium battery energy storage system monitoring method and system based on noise reduction self-coding |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112215253A true CN112215253A (en) | 2021-01-12 |
Family
ID=74059043
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010886479.9A Pending CN112215253A (en) | 2020-08-28 | 2020-08-28 | Lithium battery energy storage system monitoring method and system based on noise reduction self-coding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112215253A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113076689A (en) * | 2021-03-25 | 2021-07-06 | 华中科技大学 | Battery state evaluation method based on automatic encoder |
-
2020
- 2020-08-28 CN CN202010886479.9A patent/CN112215253A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113076689A (en) * | 2021-03-25 | 2021-07-06 | 华中科技大学 | Battery state evaluation method based on automatic encoder |
| CN113076689B (en) * | 2021-03-25 | 2024-03-19 | 华中科技大学 | Battery state evaluation method based on automatic encoder |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111584952B (en) | Method and system for online evaluation of electrochemical cells of energy storage power station | |
| CN112858919B (en) | Battery system online fault diagnosis method and system based on cluster analysis | |
| CN107422266B (en) | Fault diagnosis method and device for high-capacity battery energy storage system | |
| CN109375116B (en) | Battery system abnormal battery identification method based on self-encoder | |
| CN103954913A (en) | Predication method of electric vehicle power battery service life | |
| CN108960321B (en) | Battery fault prediction method for large lithium battery energy storage power station | |
| CN111736084A (en) | Valve-regulated lead-acid storage battery health state prediction method based on improved LSTM neural network | |
| Guo et al. | A novel radial basis function neural network principal component analysis scheme for PMU-based wide-area power system monitoring | |
| CN111487532B (en) | Retired battery screening method and system based on analytic hierarchy process and entropy method | |
| CN112666479A (en) | Battery service life prediction method based on charging cycle fusion | |
| De Sousa et al. | Comparison of different approaches for lead acid battery state of health estimation based on artificial neural networks algorithms | |
| CN111460656A (en) | Method and system for evaluating operation life of communication power supply of electric power machine room | |
| CN112215253A (en) | Lithium battery energy storage system monitoring method and system based on noise reduction self-coding | |
| CN114578234A (en) | Lithium ion battery degradation and capacity prediction model considering causality characteristics | |
| CN114460481A (en) | Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism | |
| Chen et al. | State of health estimation of lithium-ion batteries based on equivalent circuit model and data-driven method | |
| CN104569838B (en) | The evaluation method of container energy storage device core component based on remote monitoring | |
| CN116643178B (en) | SOC estimation method and related device of battery management system | |
| CN117289167A (en) | Battery remaining life prediction method, device and medium based on multiple neural network | |
| CN113538037A (en) | Method, system, equipment and storage medium for monitoring charging event of battery car | |
| CN117310533A (en) | Service life acceleration test method and system for proton exchange membrane fuel cell | |
| CN116736172A (en) | Lithium battery pack health state evaluation method based on generation type countermeasure network | |
| CN117169761A (en) | Battery state evaluation method, apparatus, device, storage medium, and program product | |
| França et al. | A new approach to estimate SoH of lead-acid batteries used in off-grid PV system | |
| KR102292748B1 (en) | Solar cell power plant and control system, and its method of operating thereof |
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 |