CN114528772B - Charger charging prediction method in electromechanical converter control system - Google Patents

Charger charging prediction method in electromechanical converter control system Download PDF

Info

Publication number
CN114528772B
CN114528772B CN202210415875.2A CN202210415875A CN114528772B CN 114528772 B CN114528772 B CN 114528772B CN 202210415875 A CN202210415875 A CN 202210415875A CN 114528772 B CN114528772 B CN 114528772B
Authority
CN
China
Prior art keywords
charger
data information
charging
control system
layer
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
Application number
CN202210415875.2A
Other languages
Chinese (zh)
Other versions
CN114528772A (en
Inventor
郭昌京
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Simsukian Electronics Technology Co ltd
Original Assignee
Shenzhen Simsukian Electronics Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shenzhen Simsukian Electronics Technology Co ltd filed Critical Shenzhen Simsukian Electronics Technology Co ltd
Priority to CN202210415875.2A priority Critical patent/CN114528772B/en
Publication of CN114528772A publication Critical patent/CN114528772A/en
Application granted granted Critical
Publication of CN114528772B publication Critical patent/CN114528772B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • 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
    • 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
    • H02J7/0048—Detection of remaining charge capacity or state of charge [SOC]
    • 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
    • H02J7/007—Regulation of charging or discharging current or voltage
    • 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
    • Y02T10/7072—Electromobility specific charging systems or methods for batteries, ultracapacitors, supercapacitors or double-layer capacitors

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Power Engineering (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

The invention discloses a charger charging prediction method in an electromechanical converter control system, which comprises the following steps: step 1, acquiring historical charging data information and current data information of a charger in a control system of an electromechanical transducer; step 2, constructing a convolutional neural network model, and converting various data information in the step 1 into a neural network structure; calculating charging data information through the constructed graph convolution neural network; step 3, adding a Volterra model into the convolutional neural network model; step 4, predicting charging data of the charger; outputting a judgment result; and 5, through repeated iterative calculation, realizing charging data information prediction through a fault diagnosis method. The invention can realize the charging prediction of the charger and display the current power utilization condition, and can realize early warning reminding when the electric quantity is insufficient, thereby greatly improving the charging prediction capability of the charger.

Description

Charger charging prediction method in electromechanical converter control system
Technical Field
The invention relates to the technical field of electromechanics, in particular to a charger charging prediction method in an electromechanical converter control system.
Background
In the electromechanical technology field, a converter is a device for converting information sent by a signal source according to a certain purpose, a converter control system can realize the control of the change of the information of the converter, in the specific application process, a controller is usually used for controlling the switching operation of two switching elements, and in the working process of the electromechanical converter control system, under the voltage control mode, the voltage is increased to the required output so as to charge a battery. In the current control mode, the energy is set to a level at which the charger is to be powered. The power supply part usually adopts a charger to realize the charging of data information. Once the charger is not powered or is insufficiently charged, the normal operation of the electromechanical converter control system is directly influenced. The converter control system is mostly powered by direct current, and how to realize the charger condition prediction is a key related to whether the electromechanical converter control system can normally work, current and voltage in the charging and discharging process are generally tracked in the prior art as a basis, but the specific state of a battery cannot be displayed, and during early warning, the normal working state of the electromechanical converter control system is influenced, the early warning precision is not high, and the charging capacity of the charger is delayed.
Disclosure of Invention
Aiming at the defects of the technology, the invention discloses a charger charging prediction method in an electromechanical converter control system, which can realize charger charging prediction and display the current power utilization condition, can realize early warning and reminding when the electric quantity is insufficient, and greatly improves the charger charging prediction capability.
In order to achieve the technical effects, the invention adopts the following technical scheme:
a charger charging prediction method in an electromechanical converter control system comprises the following steps:
step 1, acquiring historical charging data information and current charging data information of a charger in a control system of an electromechanical transducer;
the historical charging data information at least comprises charger performance parameters, daily charging cycle days, weekly charging cycle days, annual charging cycle days and fault data periods;
the current charging data information at least comprises charger performance parameters, charging time, used time, historical fault data information, the service life of the charger and the service life of the charger;
step 2, constructing a convolutional neural network model, and converting various data information in the step 1 into a neural network structure; calculating charging data information through the constructed graph convolution neural network;
step 3, adding a Volterra model into the convolutional neural network model to improve the calculation speed of the convolutional neural network model and improve the capability of searching the optimal weight and threshold of the convolutional neural network;
step 4, charger charging data prediction, namely setting a threshold value of charger charging data in the electromechanical transducer control system, judging a data value of the current charging amount, and comparing the data value of the current charging amount with the set threshold value of the charger charging data; outputting a judgment result;
and 5, through repeated iterative calculation, realizing the charging data information prediction through a fault diagnosis method.
As a further technical solution of the present invention, in step 1, the method for obtaining historical charging data information of a charger in an electromechanical transducer control system refers to historical database data information; the method for acquiring the current charging data information of the charger in the electromechanical transducer control system is realized through a state detection circuit, wherein the state detection circuit adopts a detection circuit of an LM239D four-way voltage comparator.
As a further technical solution of the present invention, a method for detecting current charging data information by a state detection circuit includes:
the LM239D four-way voltage comparator is set to three-way data information, and the comparison threshold values of the three-way voltage comparator are respectively set to +10V, +7V and + 5V; the logic values of the three outputs can be combined to determine the corresponding charging state, and an MMBD41 4148SE type switching diode is added at the input end of the comparator to clamp the voltage range of the input CP signal at-12V to + 12V; after the voltage comparator outputs a signal every time, a TLP121 type photoelectric coupler is also adopted to isolate the CP signal and output a detection result, data detection information at least comprises voltage, current and power, and detection states at least comprise a healthy working state, a fatigue working state and a strain working state.
As a further technical solution of the present invention, the CNN neural network model includes an input layer, a convolutional layer, a linear rectifying layer, a pooling layer, a full-link layer, and a Volterra model, wherein an output end of the input layer is connected to an input end of the convolutional layer, an output end of the convolutional layer is connected to an input end of the linear rectifying layer, an output end of the linear rectifying layer is connected to an input end of the pooling layer, an output end of the pooling layer is connected to an input end of the full-link layer, and the Volterra model is an output end of the CNN neural network model.
As a further technical scheme of the invention, the working method of the CNN neural network model comprises the following steps:
inputting data information of charger and operation state data information of electromechanical converter control system at input layer, and feature matrix input at previous layer of CNN neural network model
Figure 357867DEST_PATH_IMAGE001
Convolution kernel with learning
Figure DEST_PATH_IMAGE002
Performing two-dimensional convolution, outputting characteristic matrix after the data information of the two data information is convoluted through an activation function
Figure 498998DEST_PATH_IMAGE003
And the dimension between the three satisfies
Figure DEST_PATH_IMAGE004
Then the input layer outputs the function as:
Figure 62222DEST_PATH_IMAGE005
（1）
in the formula (1), the reaction mixture is,
Figure DEST_PATH_IMAGE006
is the number of convolutional sub-network layers of the CNN neural network model,
Figure 896186DEST_PATH_IMAGE007
in the form of a convolution kernel, the kernel is,
Figure DEST_PATH_IMAGE008
in order to be offset,
Figure 971458DEST_PATH_IMAGE009
is composed of
Figure 447439DEST_PATH_IMAGE006
The output of the layer is carried out,
Figure DEST_PATH_IMAGE010
is composed of
Figure 862240DEST_PATH_IMAGE006
Inputting layers;
Figure 273629DEST_PATH_IMAGE011
expressed in CNN neural network model
Figure 305039DEST_PATH_IMAGE006
The layer has a convolution kernel of
Figure DEST_PATH_IMAGE012
Subscript of
Figure 787973DEST_PATH_IMAGE013
Denotes the firstiThe initial values of the number of convolutional layers,
Figure 729384DEST_PATH_IMAGE011
upper label of
Figure 433380DEST_PATH_IMAGE006
The number of layers of the convolutional sub-network is shown,
Figure DEST_PATH_IMAGE014
and then, performing dimension reduction processing on the input data information of the CNN neural network model, wherein the function of the convolutional layer output data information is as follows:
Figure 420927DEST_PATH_IMAGE015
（2）
actual parameter data information output by charger in operation process of electromechanical transducer control system
Figure DEST_PATH_IMAGE016
Comprises the following steps:
Figure 707552DEST_PATH_IMAGE017
（3）
in the formula (3), the reaction mixture is,
Figure DEST_PATH_IMAGE018
for the parametric data information output through the first layer convolution,
Figure 831366DEST_PATH_IMAGE019
for the parametric data information output by the second layer convolution,
Figure 100002_DEST_PATH_IMAGE020
for the parametric data information output by the nth layer convolution,
Figure 178033DEST_PATH_IMAGE021
to process information separately by convolution of the first layer volume integral output,
Figure 100002_DEST_PATH_IMAGE022
to process information separately by convolution of the second layer convolution split output,
Figure 324981DEST_PATH_IMAGE023
to process information separately by convolution of the nth layer convolution output,
Figure DEST_PATH_IMAGE024
for all charger predicted data information values that are output separately by the first layer of convolution, the error function output is:
Figure 415297DEST_PATH_IMAGE025
（4）
in the formula (4), the actual output data information outputted from the charger
Figure DEST_PATH_IMAGE026
And data information output under ideal state
Figure 659196DEST_PATH_IMAGE027
Difference in (2)
Figure 100002_DEST_PATH_IMAGE028
；
Figure 645607DEST_PATH_IMAGE029
The theoretical error output value is obtained by the method,
Figure DEST_PATH_IMAGE030
representing an actual output value;
then, the weight value is adjusted, the weight value matrix is adjusted according to the method of minimizing the error, and when the subsequent data is output, when EPWhen the value of (A) is between 0 and 1, it indicates that the charger is in a healthy state, and when E is in a healthy statePWhen the value of (A) is greater than 1, it indicates that the charger is in use and remains powered, when EPWhen the value of (D) is greater than 2, it indicates that the charger is imminent to charge, and when E is greater than 2PWhen the value of (D) is greater than 3, it means that the charger is not charged and charging is required, and when E is greater than 3PWhen the value of (d) is greater than 4, it indicates that the charger is in a strained state. Through the identification and calculation in the mode, the application capacity of the charger is greatly improved.
As a further technical scheme of the invention, the Volterra model is constructed as follows:
the charger charging information is defined as an initial discrete function as:
Figure 876256DEST_PATH_IMAGE031
(5)
in the formula (5)
Figure DEST_PATH_IMAGE032
Indicating the amount of electrical load generated in the operating state of the charger,
Figure 239104DEST_PATH_IMAGE033
representing the running state of the electromechanical transducer control system when the electric load is generated,
Figure DEST_PATH_IMAGE034
the electric load quantity of the charger in the next time interval is output under the running state of the electromechanical transducer control system;
the operation data output by the electromechanical transducer control system in different time periods are changed into:
Figure 603089DEST_PATH_IMAGE035
(6)
wherein
Figure DEST_PATH_IMAGE036
The prediction data of the electromechanical transducer control system after the charger outputs the electric load in the next time period are represented, and under the known running state of the current electromechanical transducer control system, the numerical function of the electric load predicted by the charger is as follows:
Figure 25980DEST_PATH_IMAGE037
(7)
in the formula
Figure DEST_PATH_IMAGE038
Representing a predicted value of electrical load data generated during a next period of operation of the electromechanical transducer control system,
the Volterra model is then written as:
Figure 272154DEST_PATH_IMAGE039
(8)
in the formula (8), the reaction mixture is,
Figure DEST_PATH_IMAGE040
the adaptive prediction coefficient of the charger is shown when the electromechanical transducer control system is in a normal working state,
Figure 704272DEST_PATH_IMAGE041
a prediction function reference value representing a prediction of a charging process of the charger,
Figure DEST_PATH_IMAGE042
representing a system electrical load input value in operation of the electromechanical transducer control system,
Figure 657185DEST_PATH_IMAGE043
the charger is predicted under the subsequent working state of the electromechanical transducer control systemThe polymorphic electrical load value of (a);
under the normal working state of the electromechanical transducer control system, the current prediction function of the charger is as follows:
Figure DEST_PATH_IMAGE044
（9）
wherein the difference between the true predicted value and the ideal predicted value:
Figure 516556DEST_PATH_IMAGE045
(10)。
a charger charging prediction device in an electromechanical converter control system comprises the following modules:
the FPGA control module is used for controlling the charger charging prediction in the electric converter control system;
the data acquisition module is used for acquiring historical charging data information and current charging data information of a charger in the electromechanical transducer control system;
the calculation module is used for constructing a convolutional neural network model and converting various data information acquired by the data acquisition module into a neural network structure; calculating charging data information through the constructed graph convolution neural network; adding a Volterra model into the convolutional neural network model to improve the calculation speed of the convolutional neural network model and improve the capability of searching the optimal weight and threshold of the convolutional neural network;
the prediction module is used for realizing charger charging data prediction, setting a threshold value of charger charging data in the electromechanical transducer control system, judging a data value of the current charging amount, and comparing the data value of the current charging amount with the set threshold value of the charger charging data; outputting a judgment result;
and the data output module is used for realizing charging data information prediction through a fault diagnosis method through repeated iterative calculation. Wherein:
the FPGA control module is respectively connected with the calculation module and the data output module, the output end of the data acquisition module is connected with the calculation module, and the output end of the calculation module is connected with the input end of the prediction module.
Positive and advantageous effects
According to the invention, the historical charging data information and the current charging data information of the charger in the electromechanical transducer control system are obtained, the data calculation is realized by applying a mode of improving a convolutional neural network model to the data information, the fault diagnosis capability of the convolutional neural network model is improved through a Volterra model, and the analysis capability of the historical charging data information and the current charging data information of the charger in the electromechanical transducer control system is greatly improved. The invention also detects the current charging data information through the state detection circuit, thereby improving the acquisition capability of the charger data information, detecting the data information such as voltage, current, power and the like, analyzing various data information such as the health working state, the fatigue working state, the strain working state and the like of the charger, and greatly improving the application capability of the charger.
Drawings
FIG. 1 is a schematic diagram of a charger prediction method according to the present invention;
FIG. 2 is a schematic diagram of a state detection hardware circuit according to the present invention;
FIG. 3 is a schematic diagram of the overall architecture of a convolution component calculation model according to the present invention;
FIG. 4 is a schematic diagram of a single model architecture for convolution score calculation in the present invention;
FIG. 5 is a diagram illustrating a charger prediction hardware configuration according to the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, and it should be understood that the embodiments described herein are merely for purposes of illustration and explanation, and are not intended to limit the present invention.
A charger charging prediction method in an electromechanical converter control system comprises the following steps:
step 1, acquiring historical charging data information and current charging data information of a charger in a control system of an electromechanical transducer;
the historical charging data information at least comprises the performance parameters of the charger, the days of a daily charging cycle, the days of a weekly charging cycle, the days of an annual charging cycle and a fault data cycle;
the current charging data information at least comprises charger performance parameters, charging time, used time, historical fault data information, the service life of the charger and the service life of the charger;
step 2, constructing a convolutional neural network model, and converting various data information in the step 1 into a neural network structure; calculating charging data information through the constructed graph convolution neural network;
step 3, adding a Volterra model into the convolutional neural network model to improve the calculation speed of the convolutional neural network model and improve the capability of searching the optimal weight and threshold of the convolutional neural network;
step 4, charger charging data prediction, namely setting a threshold value of charger charging data in the electromechanical transducer control system, judging a data value of the current charging amount, and comparing the data value of the current charging amount with the set threshold value of the charger charging data; outputting a judgment result;
and 5, through repeated iterative calculation, realizing charging data information prediction through a fault diagnosis method.
In step 1, the method for acquiring historical charging data information of a charger in the electromechanical transducer control system refers to historical database data information; the method for acquiring the current charging data information of the charger in the electromechanical transducer control system is realized through a state detection circuit, wherein the state detection circuit adopts a detection circuit of an LM239D four-way voltage comparator.
In the above step, the method for detecting the current charging data information by the state detection circuit includes:
the LM239D four-way voltage comparator is set to three-way data information, and the comparison threshold values of the three-way voltage comparator are respectively set to +10V, +7V and + 5V; the logic values of the three outputs can be combined to determine the corresponding charging state, and an MMBD41 4148SE type switching diode is added at the input end of the comparator to clamp the voltage range of the input CP signal at-12V to + 12V; after the voltage comparator outputs a signal every time, a TLP121 type photoelectric coupler is adopted to isolate the CP signal and output a detection result, data detection information at least comprises voltage, current and power, and detection states at least comprise a healthy working state, a fatigue working state and a strain working state.
In specific application, when electric energy data information is collected, data abnormity and data loss information which occur in the transmission and storage processes of data can be included, the heating energy of the charger is used for evaluation, when factors such as faults occur in the working process of the charger, for example, the deviation of actually measured power data is large, if data is not processed and the error is directly predicted, the loss power of the charger can be calculated.
In a particular application, a Convolutional Neural Network (CNN) is a feed-forward Neural Network whose artificial neurons can respond to a portion of the coverage of surrounding cells. Very similar to common neural networks, they are composed of neurons with learnable weights and bias constants (biases). Each neuron receives some input and makes some dot product calculations, and the output is a score for each class.
In the above embodiment, the CNN neural network model includes an input layer, a convolutional layer, a linear rectifying layer, a pooling layer, a full-link layer, and a Volterra model, where an output end of the input layer is connected to an input end of the convolutional layer, an output end of the convolutional layer is connected to an input end of the linear rectifying layer, an output end of the linear rectifying layer is connected to an input end of the pooling layer, an output end of the pooling layer is connected to an input end of the full-link layer, and the Volterra model is an output end of the CNN neural network model.
In the above embodiment, the Convolutional layer (Convolutional layer) is formed by each Convolutional layer in the Convolutional neural network and composed of several Convolutional units, and the parameters of each Convolutional unit are optimized by a back propagation algorithm. The convolution operation aims to extract different input features, the first layer of convolution layer can only extract some low-level features such as edges, lines, angles and other levels, and more layers of networks can iteratively extract more complex features from the low-level features. The Activation function (Activation function) of the Linear rectification layer (Rectified Linear Units, ReLU) uses Linear rectification. Pooling layers (Pooling layers) typically result in very large-dimensional features after the layers are convolved, and the features are cut into regions, the maximum or average of which is taken to obtain new, smaller-dimensional features. The Fully-Connected layer combines all local features into a global feature that is used to calculate the score for each final class.
In the above embodiment, the working method of the CNN neural network model is as follows:
inputting data information of charger and operation state data information of electromechanical converter control system at input layer, and feature matrix input at previous layer of CNN neural network model
Figure 594234DEST_PATH_IMAGE001
Convolution kernel with learning
Figure 358272DEST_PATH_IMAGE002
Performing two-dimensional convolution, outputting characteristic matrix after the data information of the two after the convolution passes through an activation function
Figure 837794DEST_PATH_IMAGE003
And the dimension among the three satisfies
Figure 399226DEST_PATH_IMAGE004
Then the input layer outputs the function as:
Figure 964199DEST_PATH_IMAGE005
（1）
in the formula (1), the reaction mixture is,
Figure 269279DEST_PATH_IMAGE006
is the number of convolutional sub-network layers of the CNN neural network model,
Figure 603308DEST_PATH_IMAGE007
in the form of a convolution kernel, the kernel is,
Figure 335641DEST_PATH_IMAGE008
in order to be offset,
Figure 387910DEST_PATH_IMAGE009
is composed of
Figure 231102DEST_PATH_IMAGE006
The output of the layer is carried out,
Figure 685217DEST_PATH_IMAGE010
is composed of
Figure 791713DEST_PATH_IMAGE006
Inputting layers;
Figure 455912DEST_PATH_IMAGE011
expressed in CNN neural network model
Figure 978161DEST_PATH_IMAGE006
The layer has a convolution kernel of
Figure 83520DEST_PATH_IMAGE012
Subscript of
Figure 157655DEST_PATH_IMAGE013
Is shown asiThe initial values of the plurality of convolution layers,
Figure 918938DEST_PATH_IMAGE011
upper label of
Figure 372440DEST_PATH_IMAGE006
Indicating the number of convolutional packet network layers,
Figure 535568DEST_PATH_IMAGE014
and then, performing dimension reduction processing on the input data information of the CNN neural network model, wherein the function of the convolutional layer output data information is as follows:
Figure 780605DEST_PATH_IMAGE015
（2）
actual parameter data information output by charger in operation process of electromechanical converter control system
Figure 294763DEST_PATH_IMAGE016
Comprises the following steps:
Figure 283448DEST_PATH_IMAGE017
（3）
in the formula (3), the reaction mixture is,
Figure 301082DEST_PATH_IMAGE018
for the parametric data information output through the first layer convolution,
Figure 982599DEST_PATH_IMAGE019
for the parametric data information output by the second layer convolution,
Figure 718474DEST_PATH_IMAGE020
for the parametric data information output by the nth layer convolution,
Figure 510849DEST_PATH_IMAGE021
to process information separately by convolution of the first layer volume integral output,
Figure 382991DEST_PATH_IMAGE022
for convolution separately processing information output through the second layer convolution,
Figure 235409DEST_PATH_IMAGE023
to process information separately by convolution of the nth layer convolution output,
Figure 458580DEST_PATH_IMAGE024
for all charger predicted data information values that are output separately by the first layer of convolution, the error function output is:
Figure 992329DEST_PATH_IMAGE025
（4）
in the formula (4), the actual output data information outputted from the charger
Figure 109190DEST_PATH_IMAGE026
Data information output under ideal state
Figure 7876DEST_PATH_IMAGE027
Difference of (2)
Figure 574468DEST_PATH_IMAGE028
；
Figure 584012DEST_PATH_IMAGE029
The theoretical error output value is obtained by the method,
Figure 820958DEST_PATH_IMAGE030
representing an actual output value;
then, the weight is adjusted, the weight matrix is adjusted according to the method of minimizing the error, and when E is output in the following dataPWhen the value of (A) is between 0 and 1, it indicates that the charger is in a healthy state, and when E is in a healthy statePWhen the value of (A) is greater than 1, it indicates that the charger is in use and remains powered, when EPWhen the value of (A) is greater than 2, it indicates that the charger is imminent to charge, when EPWhen the value of (D) is greater than 3, it means that the charger is not charged and charging is required, and when E is greater than 3PWhen the value of (d) is greater than 4, it indicates that the charger is in a strained state. Through the identification and calculation in the mode, the application capacity of the charger is greatly improved.
In the above example, the Volterra model was constructed as follows:
the charger charging information is defined as an initial discrete function as:
Figure 421704DEST_PATH_IMAGE031
(5)
in the formula (5)
Figure 619467DEST_PATH_IMAGE032
Indicating the amount of electrical load generated in the operating state of the charger,
Figure 291757DEST_PATH_IMAGE033
representing the running state of the electromechanical transducer control system when the electric load is generated,
Figure 258576DEST_PATH_IMAGE034
the electric load quantity of the charger in the next time interval is output under the running state of the electromechanical transducer control system;
in order to improve the prediction capability of the prediction algorithm, a feedback structure is added on the basis of prediction, so that the prediction capability of the system is greatly improved,
the operation data output by the electromechanical transducer control system in different time periods are changed into:
Figure 358119DEST_PATH_IMAGE035
(6)
wherein
Figure 43178DEST_PATH_IMAGE036
The prediction data of the electromechanical transducer control system after the charger outputs the electric load in the next time period are represented, and under the known running state of the current electromechanical transducer control system, the numerical function of the electric load predicted by the charger is as follows:
Figure 519159DEST_PATH_IMAGE037
(7)
in the formula
Figure 340484DEST_PATH_IMAGE038
Representing a predicted value of electrical load data generated during a next time period of operation of the electromechanical transducer control system,
for the long-term prediction of the combined supply system, a feedback structure needs to be built, and the predicted value is ensured not to deviate from the reality too much;
if the prediction algorithm cannot fully simulate the running state of the combined supply system, the long-term prediction accuracy of the prediction algorithm is reduced, and if the constructed feedback structure is unstable, the accuracy of the long-term prediction data of the prediction algorithm is also influenced.
The Volterra model is then written as:
Figure 876508DEST_PATH_IMAGE039
(8)
in the formula (8), the reaction mixture is,
Figure 783284DEST_PATH_IMAGE040
the adaptive prediction coefficient of the charger is shown when the electromechanical transducer control system is in a normal working state,
Figure 797376DEST_PATH_IMAGE041
a prediction function reference value representing a prediction of a charging process of the charger,
Figure 473208DEST_PATH_IMAGE042
representing a system electrical load input value in operation of the electromechanical transducer control system,
Figure 914554DEST_PATH_IMAGE043
the electric load value is a polymorphic electric load value obtained by predicting the charger in the subsequent working state of the electromechanical transducer control system;
in the feedback system designed in this study, neglecting the error generated by the feedback calculation, the output signal generated for the feedback structure is expressed as:
Figure DEST_PATH_IMAGE046
(9)
wherein the content of the first and second substances,
Figure 905031DEST_PATH_IMAGE047
representing variations of feedback structure inputAmount of the compound (A). Under normal conditions, the feedback signal is a one-step predicted value of a previous point, real data is obtained from experiments, and a real input signal vector is written as follows:
Figure DEST_PATH_IMAGE048
(10)
therefore, under the normal working state of the electromechanical converter control system, the current prediction function of the charger is as follows:
Figure 191656DEST_PATH_IMAGE044
（11）
wherein the difference between the true predicted value and the ideal predicted value:
Figure 987574DEST_PATH_IMAGE045
(12)
due to the previous predicted value
Figure 865400DEST_PATH_IMAGE049
When the basis function and the feedback function are derived, and the error is predicted
Figure DEST_PATH_IMAGE050
Also smaller, are:
Figure 340244DEST_PATH_IMAGE051
(13)
the interference of cold and hot loads on electric loads is shielded, and prediction errors are analyzed to obtain:
Figure DEST_PATH_IMAGE052
(14)
wherein the content of the first and second substances,
Figure 837084DEST_PATH_IMAGE053
is the component that contains the feedback signal. Thus, the feedback prediction error:
Figure 100002_DEST_PATH_IMAGE054
(15)
in the system prediction polymorphic electric load value, the fixed point is
Figure 80984DEST_PATH_IMAGE055
When, if
Figure DEST_PATH_IMAGE056
Amplifying the prediction error of the previous step to generate larger deviation of the system prediction value; on the contrary, if
Figure 598553DEST_PATH_IMAGE057
The prediction error of the feedback is compressed.
The state of the electromechanical converter control system in normal operation and the use condition of the charger can be organically combined through the prediction model, and the working state of the charger is associated with the electric load of the charger in an induction and sorting mode. Extracting various factor data of the operation of the charger, predicting various data information predicted values under the working state of the electromechanical transducer control system, fitting and analyzing polymorphic electric load data by acquiring various data forms of the charger in the working process, and predicting a long-term value according to the actual operation electric load value of the electromechanical transducer control system; interference factors in charger prediction were analyzed by correlation test. In the application, the charger prediction capability is greatly improved.
A charger charging prediction device in an electromechanical converter control system comprises the following modules:
the FPGA control module is used for controlling the charger charging prediction in the electric converter control system;
the data acquisition module is used for acquiring historical charging data information and current charging data information of a charger in the electromechanical transducer control system;
the calculation module is used for constructing a convolutional neural network model and converting various data information acquired by the data acquisition module into a neural network structure; calculating charging data information through the constructed graph convolution neural network; adding a Volterra model into the convolutional neural network model to improve the calculation speed of the convolutional neural network model and improve the capability of searching the optimal weight and threshold of the convolutional neural network;
the prediction module is used for realizing charger charging data prediction, setting a threshold value of charger charging data in the electromechanical transducer control system, judging a data value of the current charging amount, and comparing the data value of the current charging amount with the set threshold value of the charger charging data; outputting a judgment result;
and the data output module is used for realizing charging data information prediction through a fault diagnosis method through repeated iterative calculation.
In a specific application, the whole hardware integrates a quad-core ARM Cortex-A53 Processing System (PS) and FPGA Programmable Logic (PL). The XCZU7EV has a 16nm manufacturing technology and abundant super memory, and can relieve the buffering and storage requirements of insufficient storage resources on an FPGA chip. The ARM processor realizes data transmission of instructions and convolution results between the FPGA and the software PE and realizes full connection layer and softmax functions. The FPGA is composed of a Direct Memory Access (DMA), a controller, an input buffer, an output buffer, and a hardware PE. The hardware PE is responsible for convolutional layers, pool layers, and nonlinear functions. An on-chip buffer, including an input buffer and an output buffer, prepares data for hardware computation and stores results. Wherein: the FPGA control module is respectively connected with the calculation module and the data output module, the output end of the data acquisition module is connected with the calculation module, and the output end of the calculation module is connected with the input end of the prediction module.
Although specific embodiments of the present invention have been described above, it will be understood by those skilled in the art that these specific embodiments are merely illustrative and that various omissions, substitutions and changes in the form of the detail of the methods and systems described above may be made by those skilled in the art without departing from the spirit and scope of the invention. For example, it is within the scope of the present invention to combine the steps of the above-described methods to perform substantially the same function in substantially the same way to achieve substantially the same result. Accordingly, the scope of the invention is to be limited only by the following claims.

Claims (3)

1. A charger charging prediction method in an electromechanical converter control system is characterized in that: the method comprises the following steps:
step 1, acquiring historical charging data information and current charging data information of a charger in a control system of an electromechanical transducer;
the historical charging data information at least comprises charger performance parameters, daily charging cycle days, weekly charging cycle days, annual charging cycle days and fault data periods;
the current charging data information at least comprises charger performance parameters, charging time, used time, historical fault data information, the service life of the charger and the service life of the charger;
step 2, constructing a convolutional neural network model, and converting various data information in the step 1 into a neural network structure; calculating charging data information through the constructed graph convolution neural network;
step 3, adding a Volterra model into the convolutional neural network model to improve the calculation speed of the convolutional neural network model and improve the capability of searching the optimal weight and threshold of the convolutional neural network; wherein the convolutional neural network model is a model based on a CNN neural network model;
step 4, charger charging data prediction, namely setting a threshold value of charger charging data in the electromechanical transducer control system, judging a data value of the current charging amount, and comparing the data value of the current charging amount with the set threshold value of the charger charging data; outputting a judgment result;
step 5, through repeated iterative computation, realizing charging data information prediction through a fault diagnosis method;
the CNN neural network model comprises an input layer, a convolutional layer, a linear rectifying layer, a pooling layer, a full connection layer and a Volterra model, wherein the output end of the input layer is connected with the input end of the convolutional layer, the output end of the convolutional layer is connected with the input end of the linear rectifying layer, the output end of the linear rectifying layer is connected with the input end of the pooling layer, the output end of the pooling layer is connected with the input end of the full connection layer, and the Volterra model is the output end of the CNN neural network model;
the working method of the CNN neural network model comprises the following steps:
inputting data information of charger and operation state data information of electromechanical converter control system at input layer, and feature matrix input at previous layer of CNN neural network model
Figure DEST_PATH_IMAGE002A
Convolution kernel with learning
Figure DEST_PATH_IMAGE004A
Performing two-dimensional convolution, outputting characteristic matrix after the data information of the two after the convolution passes through an activation function
Figure DEST_PATH_IMAGE006A
And the dimension between the three satisfies
Figure DEST_PATH_IMAGE008A
Then the input layer outputs the function as:
Figure DEST_PATH_IMAGE010A
（1）
in the formula (1), the reaction mixture is,
Figure DEST_PATH_IMAGE012_5A
is the number of convolutional sub-network layers of the CNN neural network model,
Figure DEST_PATH_IMAGE014A
in the form of a convolution kernel, the kernel is,
Figure DEST_PATH_IMAGE016A
in order to be offset,
Figure DEST_PATH_IMAGE018A
is composed of
Figure DEST_PATH_IMAGE012_6A
The output of the layer is carried out,
Figure DEST_PATH_IMAGE020
is composed of
Figure DEST_PATH_IMAGE012_7A
Inputting layers;
Figure DEST_PATH_IMAGE022
expressed in CNN neural network model
Figure DEST_PATH_IMAGE012_8A
The layer has a convolution kernel of
Figure DEST_PATH_IMAGE023
Subscript of
Figure DEST_PATH_IMAGE025
Is shown as
Figure DEST_PATH_IMAGE027
The initial values of the number of convolutional layers,
Figure DEST_PATH_IMAGE022A
upper label of
Figure DEST_PATH_IMAGE028
Indicating the number of convolutional packet network layers,
Figure DEST_PATH_IMAGE030A
and then, performing dimension reduction processing on the input data information of the CNN neural network model, wherein the function of the convolutional layer output data information is as follows:
Figure DEST_PATH_IMAGE032A
（2）
actual parameter data information output by charger in operation process of electromechanical converter control system
Figure DEST_PATH_IMAGE034AA
Comprises the following steps:
Figure DEST_PATH_IMAGE036A
（3）
in the formula (3), the reaction mixture is,
Figure DEST_PATH_IMAGE038A
for the parametric data information output through the first layer convolution,
Figure DEST_PATH_IMAGE040A
for the parametric data information output by the second layer convolution,
Figure DEST_PATH_IMAGE042A
to pass through
Figure DEST_PATH_IMAGE044A
The information of the parameter data outputted by the layer convolution,
Figure DEST_PATH_IMAGE046A
to process information separately by convolution of the first layer volume integral output,
Figure DEST_PATH_IMAGE048A
to process information separately by convolution of the second layer convolution split output,
Figure DEST_PATH_IMAGE050A
to pass through
Figure DEST_PATH_IMAGE044AA
Layered convolution separationThe output convolution is used to process the information separately,
Figure DEST_PATH_IMAGE052A
for all charger predicted data information values that are output separately by the first layer of convolution, the error function output is:
Figure DEST_PATH_IMAGE054
（4）
in the formula (4), the actual output data information outputted from the charger
Figure DEST_PATH_IMAGE055A
Data information output under ideal state
Figure DEST_PATH_IMAGE057A
Difference of (2)
Figure DEST_PATH_IMAGE059A
；
Figure DEST_PATH_IMAGE061A
The theoretical error output value is obtained by the method,
Figure DEST_PATH_IMAGE063A
representing an actual output value;
then, the weight value is adjusted, the weight value matrix is adjusted according to the method of minimizing the error, and when the subsequent data is output, when EPWhen the value of (A) is between 0 and 1, it indicates that the charger is in a healthy state, and when E is in a healthy statePWhen the value of (A) is greater than 1, it indicates that the charger is in use and remains powered, when EPWhen the value of (D) is greater than 2, it indicates that the charger is imminent to charge, and when E is greater than 2PWhen the value of (D) is greater than 3, it means that the charger is not charged and charging is required, and when E is greater than 3PWhen the value of (1) is more than 4, the charger is in a strain state;
the Volterra model was constructed as follows:
the charger charging information is defined as an initial discrete function as:
Figure DEST_PATH_IMAGE065A
(5)
in the formula
Figure DEST_PATH_IMAGE066
Indicating the amount of electrical load generated in the operating state of the charger,
Figure DEST_PATH_IMAGE068
representing the running state of the electromechanical transducer control system when the electric load is generated,
Figure DEST_PATH_IMAGE070
the electric load quantity of the charger in the next time interval is output under the running state of the electromechanical transducer control system;
the operation data output by the electromechanical transducer control system in different time periods are changed into:
Figure DEST_PATH_IMAGE072
(6)
wherein
Figure DEST_PATH_IMAGE074
The prediction data of the electromechanical transducer control system after the charger outputs the electric load in the next time period are represented, and under the known running state of the current electromechanical transducer control system, the numerical function of the electric load predicted by the charger is as follows:
Figure DEST_PATH_IMAGE076
(7)
in the formula
Figure DEST_PATH_IMAGE078
Indicating next time of control system of electromechanical transducerA predicted value of electrical load data generated during operation of the segment,
the Volterra model is then written as:
Figure DEST_PATH_IMAGE080
(8)
in the formula (8), the reaction mixture is,
Figure DEST_PATH_IMAGE082
the adaptive prediction coefficient of the charger is shown when the electromechanical transducer control system is in a normal working state,
Figure DEST_PATH_IMAGE084
a prediction function reference value representing a prediction of a charging process of the charger,
Figure DEST_PATH_IMAGE086
representing a system electrical load input value in operation of the electromechanical transducer control system,
Figure DEST_PATH_IMAGE088
the electric load value is a polymorphic electric load value obtained by predicting the charger in the subsequent working state of the electromechanical transducer control system;
under the normal working state of the electromechanical transducer control system, the current prediction function of the charger is as follows:
Figure DEST_PATH_IMAGE090
（9）
wherein the difference between the true predicted value and the ideal predicted value:
Figure DEST_PATH_IMAGE092
(10)。
2. the method of claim 1, wherein the method comprises: in step 1, the method for acquiring historical charging data information of a charger in the electromechanical transducer control system refers to historical database data information; the method for acquiring the current charging data information of the charger in the electromechanical transducer control system is realized through a state detection circuit, wherein the state detection circuit adopts a detection circuit of an LM239D four-way voltage comparator.
3. The method of claim 2, wherein the method comprises: the method for detecting the current charging data information by the state detection circuit comprises the following steps:
the LM239D four-way voltage comparator is set to three-way data information, and the comparison threshold values of the three-way voltage comparator are respectively set to +10V, +7V and + 5V; the logic values of the three outputs can be combined to determine the corresponding charging state, and an MMBD41 4148SE type switching diode is added at the input end of the comparator to clamp the voltage range of the input CP signal at-12V to + 12V; after the voltage comparator outputs a signal every time, a TLP121 type photoelectric coupler is adopted to isolate the CP signal and output a detection result, data detection information at least comprises voltage, current and power, and detection states at least comprise a healthy working state, a fatigue working state and a strain working state.
CN202210415875.2A 2022-04-20 2022-04-20 Charger charging prediction method in electromechanical converter control system Active CN114528772B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210415875.2A CN114528772B (en) 2022-04-20 2022-04-20 Charger charging prediction method in electromechanical converter control system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210415875.2A CN114528772B (en) 2022-04-20 2022-04-20 Charger charging prediction method in electromechanical converter control system

Publications (2)

Publication Number Publication Date
CN114528772A CN114528772A (en) 2022-05-24
CN114528772B true CN114528772B (en) 2022-07-01

Family

ID=81627965

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210415875.2A Active CN114528772B (en) 2022-04-20 2022-04-20 Charger charging prediction method in electromechanical converter control system

Country Status (1)

Country Link
CN (1) CN114528772B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104348205A (en) * 2013-07-31 2015-02-11 杭州锂软科技有限公司 SOC-SOH (state of charge-state of health)-based distributed BMS (Battery Management System)
CN108804764A (en) * 2018-05-07 2018-11-13 电子科技大学 A kind of aging of lithium battery trend forecasting method based on extreme learning machine
CN111142027A (en) * 2019-12-31 2020-05-12 国电南瑞南京控制系统有限公司 Lithium iron phosphate battery state-of-charge monitoring and early warning method based on neural network
WO2020102733A1 (en) * 2018-11-16 2020-05-22 Nvidia Corporation Learning to generate synthetic datasets for training neural networks
JP2021180581A (en) * 2020-05-14 2021-11-18 トヨタ自動車株式会社 Soc controller for power storage device

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5802463B2 (en) * 2011-07-22 2015-10-28 株式会社東芝 Electric quantity adjusting device, electric quantity adjusting method, electric quantity adjusting program, and power supply system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104348205A (en) * 2013-07-31 2015-02-11 杭州锂软科技有限公司 SOC-SOH (state of charge-state of health)-based distributed BMS (Battery Management System)
CN108804764A (en) * 2018-05-07 2018-11-13 电子科技大学 A kind of aging of lithium battery trend forecasting method based on extreme learning machine
WO2020102733A1 (en) * 2018-11-16 2020-05-22 Nvidia Corporation Learning to generate synthetic datasets for training neural networks
CN111142027A (en) * 2019-12-31 2020-05-12 国电南瑞南京控制系统有限公司 Lithium iron phosphate battery state-of-charge monitoring and early warning method based on neural network
JP2021180581A (en) * 2020-05-14 2021-11-18 トヨタ自動車株式会社 Soc controller for power storage device

Also Published As

Publication number Publication date
CN114528772A (en) 2022-05-24

Similar Documents

Publication Publication Date Title
Wu et al. Data-driven remaining useful life prediction via multiple sensor signals and deep long short-term memory neural network
Liu et al. Single-machine-based joint optimization of predictive maintenance planning and production scheduling
Liu et al. Single-machine-based joint optimization of predictive maintenance planning and production scheduling
CN112990556A (en) User power consumption prediction method based on Prophet-LSTM model
CN112630662B (en) Power battery SOH estimation method based on data driving and multi-parameter fusion
CN110135064B (en) Method, system and controller for predicting temperature faults of rear bearing of generator
CN110837915A (en) Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning
CN114722923B (en) Lightweight electromechanical equipment fault diagnosis method
Li et al. An optimal stacking ensemble for remaining useful life estimation of systems under multi-operating conditions
Alshathri et al. Digital twin-based automated fault diagnosis in industrial IoT applications
CN114528772B (en) Charger charging prediction method in electromechanical converter control system
ul Hassan et al. Online static security assessment for cascading failure using stacked De-noising Auto-encoder
CN112232570A (en) Forward active total electric quantity prediction method and device and readable storage medium
CN115409217B (en) Multitasking predictive maintenance method based on multi-expert hybrid network
CN110991519A (en) Intelligent switch state analysis and adjustment method and system
Mehta et al. A Comprehensive study of Machine Learning Techniques used for estimating State of Charge for Li-ion Battery
CN112232598A (en) Output power combination prediction method and system for progressive photovoltaic power generation system
Wu et al. Study on the Smart Diagnosis of Gas Turbine With Distributed Computing Technique
Wan et al. A review of power electronic parametric fault diagnosis methods
Basia et al. Comparison of data driven algorithms for SoH estimation of Lithium-ion batteries
Ziryukin et al. The Program for Finding the Optimal Power Transit Separation Place of a Ring Electric Power Grid Based on Artificial Neural Networks
CN112651491B (en) Industrial robot fault prediction method based on CNN convolution model
Muthuvinayagam et al. Comprehensive Analysis on Power Electronic Systems in Relation with Machine Learning
CN115828165B (en) New energy intelligent micro-grid data processing method and system
Hajji Deep Learning Based Faults Diagnosis in Grid-Connected Photovoltaic Systems

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