CN117272183A - Energy storage system performance prediction method and device based on combination feature selection - Google Patents
Energy storage system performance prediction method and device based on combination feature selection Download PDFInfo
- Publication number
- CN117272183A CN117272183A CN202311155658.5A CN202311155658A CN117272183A CN 117272183 A CN117272183 A CN 117272183A CN 202311155658 A CN202311155658 A CN 202311155658A CN 117272183 A CN117272183 A CN 117272183A
- Authority
- CN
- China
- Prior art keywords
- target
- performance
- subsystem
- current
- historical
- 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
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000004146 energy storage Methods 0.000 title claims abstract description 42
- 238000000605 extraction Methods 0.000 claims abstract description 92
- 238000012549 training Methods 0.000 claims description 89
- 238000004422 calculation algorithm Methods 0.000 claims description 61
- 238000013136 deep learning model Methods 0.000 claims description 60
- 238000012360 testing method Methods 0.000 claims description 39
- 238000003066 decision tree Methods 0.000 claims description 32
- 238000003860 storage Methods 0.000 claims description 26
- 238000013507 mapping Methods 0.000 claims description 25
- 238000004590 computer program Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 11
- 238000000546 chi-square test Methods 0.000 claims description 7
- 238000012935 Averaging Methods 0.000 claims description 5
- 239000000523 sample Substances 0.000 description 103
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 70
- 239000001257 hydrogen Substances 0.000 description 68
- 229910052739 hydrogen Inorganic materials 0.000 description 68
- 239000000446 fuel Substances 0.000 description 34
- 238000001816 cooling Methods 0.000 description 14
- 238000004364 calculation method Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 9
- 230000007613 environmental effect Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000012216 screening Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 239000002826 coolant Substances 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005265 energy consumption Methods 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 230000000737 periodic effect Effects 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- -1 hydrogen ions Chemical class 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 239000012528 membrane Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- BASFCYQUMIYNBI-UHFFFAOYSA-N platinum Chemical compound [Pt] BASFCYQUMIYNBI-UHFFFAOYSA-N 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- HCHKCACWOHOZIP-UHFFFAOYSA-N Zinc Chemical compound [Zn] HCHKCACWOHOZIP-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003054 catalyst Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005485 electric heating Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 125000004435 hydrogen atom Chemical group [H]* 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 125000004430 oxygen atom Chemical group O* 0.000 description 1
- 229910052697 platinum Inorganic materials 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000011701 zinc Substances 0.000 description 1
- 229910052725 zinc Inorganic materials 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
- H01—ELECTRIC ELEMENTS
- H01M—PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY
- H01M8/00—Fuel cells; Manufacture thereof
- H01M8/04—Auxiliary arrangements, e.g. for control of pressure or for circulation of fluids
- H01M8/04298—Processes for controlling fuel cells or fuel cell systems
- H01M8/04992—Processes for controlling fuel cells or fuel cell systems characterised by the implementation of mathematical or computational algorithms, e.g. feedback control loops, fuzzy logic, neural networks or artificial intelligence
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Manufacturing & Machinery (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Automation & Control Theory (AREA)
- Fuzzy Systems (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Sustainable Development (AREA)
- Sustainable Energy (AREA)
- Chemical & Material Sciences (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Electrochemistry (AREA)
- General Chemical & Material Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides an energy storage system performance prediction method and device based on combination feature selection, wherein the method comprises the following steps: acquiring current performance parameters of all target subsystems in a target battery system and system types of all target subsystems; determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the belonging system category; selecting the current performance parameter combination characteristics based on the corresponding characteristic extraction strategy to obtain current target characteristics corresponding to each target subsystem; inputting the corresponding current target characteristics into a corresponding target performance prediction model to obtain current performance prediction values of all target subsystems; and determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems, and adaptively selecting a corresponding feature extraction strategy and a corresponding target performance prediction model for each target subsystem to realize real-time effective performance prediction of the target battery system and improve prediction accuracy.
Description
Technical Field
The invention relates to the technical field of battery management, in particular to an energy storage system performance prediction method and device based on combination feature selection.
Background
The hydrogen fuel cell has the characteristics of wide fuel adaptability, high energy conversion efficiency, full solid state, modularized assembly, zero pollution and the like, and has wide application prospect when being used as a mobile power source such as a ship power source, a traffic vehicle power source and the like. Hydrogen fuel cells are used in distributed power generation, and distributed power stations are becoming an important component of energy supply due to their low cost, high maintainability, and the like.
The system performance average value of the hydrogen fuel cell represents the cell operation efficiency and is a key index of the service life of the cell. The system performance average prediction has correlation with a series of variables such as operating conditions, environmental factors, control software versions and the like in the hydrogen fuel cell system. These variables are coupled to each other and affect each other, making the evaluation of the fuel cell performance and related parameters for optimal output performance very difficult. In the prior art, the performance prediction of the hydrogen fuel cell system is usually realized through manual periodic detection, but the prediction mode is not only poor in real-time performance, but also difficult to ensure in prediction accuracy.
Disclosure of Invention
The invention provides a method and a device for predicting the performance of an energy storage system based on combination feature selection, which are used for solving the defects that in the prior art, the real-time performance is poor and the prediction accuracy is difficult to ensure in a prediction mode for realizing the performance prediction of a hydrogen fuel cell system through manual periodic detection, and enhancing the real-time performance and the prediction accuracy of the performance prediction of the hydrogen fuel cell system.
The invention provides an energy storage system performance prediction method based on combination feature selection, which comprises the following steps:
acquiring current performance parameters of each target subsystem in a target battery system and the class of the system to which each target subsystem belongs; the current performance parameters include multi-dimensional performance characteristics;
determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs;
based on the feature extraction strategy corresponding to each target subsystem, carrying out combination feature selection on the multi-dimensional performance features to obtain current target features corresponding to each target subsystem;
inputting the current target characteristics corresponding to each target subsystem into a target performance prediction model corresponding to each target subsystem to obtain a current performance prediction value of each target subsystem;
Determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems;
the target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
According to the energy storage system performance prediction method based on combination feature selection provided by the invention, the method further comprises the following steps:
the following steps are performed for each of the target subsystems:
acquiring historical performance parameters and performance labels of a sample subsystem under a system category to which a current target subsystem belongs at each historical moment;
calculating a first correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a chi-square test algorithm;
calculating a second correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a spearman algorithm;
determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the first correlation coefficient and the second correlation coefficient;
Establishing a mapping relation between the system category to which the current target subsystem belongs and the feature extraction strategy;
based on the system category to which each target subsystem belongs, determining a feature extraction strategy corresponding to each target subsystem comprises:
and acquiring a feature extraction strategy corresponding to each target subsystem according to the mapping relation and the system category to which each target subsystem belongs.
According to the energy storage system performance prediction method based on combined feature selection provided by the invention, the feature extraction strategy corresponding to the system category to which the current target subsystem belongs is determined according to the first correlation coefficient and the second correlation coefficient, and the method comprises the following steps:
comparing the first correlation coefficient with a first preset value;
according to the comparison result, at least one performance characteristic of which the first correlation coefficient is larger than the first preset value is determined in the historical performance parameters;
determining a performance characteristic of which the second correlation coefficient is larger than a second preset value from the at least one performance characteristic;
and determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the performance feature of which the second correlation coefficient is larger than a second preset value.
According to the energy storage system performance prediction method based on combination feature selection provided by the invention, the training steps of the target performance prediction model corresponding to each target subsystem comprise the following steps:
acquiring the historical performance parameters of sample subsystems and performance labels of the sample subsystems under the system category to which each target subsystem belongs;
based on the feature extraction strategy corresponding to each sample subsystem, carrying out combination feature selection on the multidimensional performance features of the sample subsystem to obtain historical target features corresponding to the sample subsystem;
constructing a sample data set according to the historical target characteristics and the performance labels;
dividing the sample data set into a training set and a testing set;
training the deep learning model according to the training set and the decision tree algorithm;
according to the test set, testing the model performance of the trained deep learning model, and adjusting the model structure and/or model parameters of the trained deep learning model under the condition that the trained deep learning model fails the test;
and training the adjusted deep learning model according to the training set and the decision tree algorithm until the trained deep learning model passes the test to obtain a target performance prediction model corresponding to each target subsystem.
According to the energy storage system performance prediction method based on combined feature selection provided by the invention, before the combined feature selection is performed on the multidimensional performance features of the sample subsystem based on the feature extraction strategy corresponding to each sample subsystem to obtain the historical target features corresponding to the sample subsystem, the method further comprises:
determining whether unsteady performance characteristics exist in historical performance parameters of the sample subsystem at each historical moment based on a steady-state processing algorithm;
and deleting the historical performance parameter and the performance label at any historical moment when the unsteady performance characteristic exists in the historical performance parameter at any historical moment.
According to the energy storage system performance prediction method based on combination feature selection provided by the invention, the current performance prediction value of the target battery system is determined according to the current performance prediction values of all the target subsystems, and the method comprises the following steps:
and carrying out weighted addition or averaging on the current performance predicted values of all the target subsystems to obtain the current performance predicted value of the target battery system.
The invention also provides an energy storage system performance prediction device based on combination feature selection, which comprises:
the system comprises an acquisition module, a control module and a control module, wherein the acquisition module is used for acquiring current performance parameters of all target subsystems in a target battery system and system categories to which all the target subsystems belong; the current performance parameters include multi-dimensional performance characteristics;
the first determining module is used for determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs;
the selection module is used for carrying out combination feature selection on the multidimensional performance features based on feature extraction strategies corresponding to the target subsystems to obtain current target features corresponding to the target subsystems;
the prediction module is used for inputting the current target characteristics corresponding to each target subsystem into the target performance prediction model corresponding to each target subsystem to obtain the current performance prediction value of each target subsystem;
the second determining module is used for determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems;
the target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the energy storage system performance prediction method based on the combination characteristic selection.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method of energy storage system performance prediction based on combined feature selection as described in any of the above.
The invention also provides a computer program product comprising a computer program which when executed by a processor implements the energy storage system performance prediction method based on combined feature selection as described in any one of the above.
According to the energy storage system performance prediction method and device based on combination feature selection, the current performance parameters of all target subsystems in the target battery system and the system category to which all target subsystems belong are firstly obtained; the system category of each target subsystem is acquired, and a feature extraction strategy and a target performance prediction model corresponding to each target subsystem are determined according to the acquired system category; then, according to the corresponding feature extraction strategy, carrying out combined feature selection on the current performance parameters of each target subsystem to obtain the current target features corresponding to each target subsystem; then, the current target characteristics corresponding to each target subsystem are input into the corresponding target performance prediction model, the output result is used as the current performance prediction value of each target subsystem, then the current performance prediction value of each target subsystem is integrated and determined, a set of complete full-automatic battery performance prediction flow is established, the corresponding characteristic extraction strategy and the corresponding target performance prediction model are adaptively selected for each target subsystem, and the prediction accuracy is improved while the real-time effective performance prediction of the target battery system is realized.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for predicting performance of an energy storage system based on combined feature selection provided by the invention;
FIG. 2 is a schematic diagram of a target subsystem provided by the present invention;
FIG. 3 is a schematic representation of the correlation between performance characteristics provided by the present invention;
FIG. 4 is a schematic flow chart of determining a feature extraction strategy provided by the present invention;
FIG. 5 is a schematic flow chart of a training process of a target energy consumption prediction model provided by the invention;
FIG. 6 is a schematic diagram of a flow chart for model training using decision tree algorithm provided by the present invention;
FIG. 7 is a schematic diagram showing a comparison of a predicted value of system performance and an actual value of system performance according to the present invention;
FIG. 8 is a schematic flow chart of a training process and a prediction process of a target performance prediction model corresponding to a target subsystem provided by the invention;
FIG. 9 is a schematic structural diagram of an energy storage system performance prediction device based on combined feature selection according to the present invention;
fig. 10 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that, the method and the device for predicting the performance of the energy storage system based on the combination feature selection provided in the present embodiment may be applied to performance prediction of various battery systems, such as a hydrogen fuel cell, a zinc air cell, etc., which are not particularly limited in this embodiment, and the hydrogen fuel cell is taken as an example to describe the present embodiment, and other batteries may be replaced adaptively.
The hydrogen fuel cell is a power generation device for directly converting chemical energy of hydrogen and oxygen into electric energy, and the basic principle is that the hydrogen is sent to an anode plate (negative electrode) of the fuel cell, one electron in hydrogen atoms is separated out through the action of a catalyst (platinum), hydrogen ions (protons) losing the electrons pass through a proton exchange membrane to reach a cathode plate (positive electrode) of the fuel cell, and the electrons cannot pass through the proton exchange membrane and can only reach the cathode plate of the fuel cell through an external circuit, so that current is generated in the external circuit. After reaching the cathode plate, the hydrogen ions are recombined with oxygen atoms and electrons to form water. The hydrogen fuel cell system mainly comprises a cell stack, a hydrogen gas path, an air path, a cooling path and an electric path. The main function of the cell stack is to convert hydrogen and air into electrical energy; the main function of the air channel is to convey air to the cell stack and discharge water generated by the cell stack; the main function of the hydrogen gas path is to convey hydrogen gas to the cell stack; the main function of the cooling circuit is to adjust the operating temperature of the battery stack; the main function of the electric circuit is to utilize the electric energy generated by the battery stack.
The hydrogen fuel cell has the characteristics of wide fuel adaptability, high energy conversion efficiency, full solid state, modularized assembly, zero pollution and the like, and has wide application prospect when being used as a mobile power supply of a ship power supply, a traffic vehicle power supply, space navigation and the like. Hydrogen fuel cells are used in distributed power generation, and distributed power stations are becoming an important component of energy supply due to their low cost, high maintainability, and the like. The system performance average value of the hydrogen fuel cell represents the cell operation efficiency and is a key index of the service life of the cell. The system performance average prediction has correlation with a series of variables such as operating conditions, environmental factors, control software versions and the like in the hydrogen fuel cell system. These variables are coupled to each other and affect each other, making the evaluation of the fuel cell performance and related parameters for optimal output performance very difficult. In the prior art, the performance prediction of the energy storage system based on the combination characteristic selection of the hydrogen fuel is usually realized through manual periodic detection, but the prediction mode is poor in instantaneity and the detection precision is difficult to ensure.
Aiming at the problems, the application provides an energy storage system performance prediction method based on combination feature selection, an algorithm model taking the current performance prediction value as a core is constructed by analyzing dynamic changes between variables and the current performance prediction value of a target battery system, a more scientific and effective hydrogen fuel cell performance calculation and analysis method is provided, and factors influencing the performance of the fuel cell system can be analyzed while the current performance level is accurately and in real time calculated. The digital twin simulation means can be provided for the performance analysis, design optimization and operation improvement of the hydrogen fuel cell system, and the assistance is provided for the design, manufacture, operation and maintenance of the hydrogen fuel cell.
The method may be performed by an electronic device, a component in an electronic device, an integrated circuit, or a chip. The electronic device may be a mobile electronic device or a non-mobile electronic device. By way of example, the mobile electronic device may be a cell phone, tablet computer, notebook computer, palm computer, vehicle mounted electronic device, wearable device, ultra-mobile personal computer (ultra-mobile personal computer, UMPC), netbook or personal digital assistant (personal digital assistant, PDA), etc., and the non-mobile electronic device may be a server, network attached storage (Network Attached Storage, NAS), personal computer (personal computer, PC), television (TV), teller machine or self-service machine, etc., without limitation of the present invention.
The energy storage system performance prediction method based on the combined feature selection of the present invention is described below with reference to fig. 1-8.
Fig. 1 is a flow chart of a method for predicting performance of an energy storage system based on combination feature selection according to the present embodiment, as shown in fig. 1, the method includes:
step 101, obtaining current performance parameters of each target subsystem in a current target battery system and system categories to which each target subsystem belongs; the current performance parameters include multi-dimensional performance characteristics;
The present embodiment is described below with the target cell system as a hydrogen fuel cell system. The target subsystems are the same or different types of subsystems in the current target battery system, which are set according to factors such as different positions, environments and the like.
Fig. 2 is a schematic structural diagram of a target subsystem provided in this embodiment, and as shown in fig. 2, the target subsystem may include a plurality of key elements, such as a stack, an electrical circuit, a cooling circuit, a hydrogen circuit, and an air circuit. Wherein the stack may be a stack comprising one or more hydrogen fuel cells; the electric circuit may be a circuit including a DCDC (Direct Current-Direct Current) and a motor; the cooling circuit may include a radiator fan, an electric heating device, a thermostat, and a cooling pump; the hydrogen path may include a hydrogen pump; the air path may include an air compressor and a humidifier.
The performance parameters of each target subsystem are parameter data generated by each key element during the current period operation of the battery, and the set parameters comprise set parameters and actual parameters, wherein the set parameters can be parameter values preset according to the rated operation condition of the target subsystem; the actual parameter may be a loop actual parameter obtained in a loop of the target subsystem from the feedback information.
The performance parameters of each target subsystem in this embodiment may include stack parameters, circuit parameters, cold gas circuit parameters, hydrogen gas circuit parameters, air circuit parameters, and environmental parameters.
The stack parameters may include a control program version number parameter, a system state parameter, a system warning parameter, a maximum voltage node number parameter, a performance tag parameter, a stack feedback parameter, a power limit flag bit parameter, a low temperature flag bit parameter, a highest warning level parameter, a hydrogen concentration parameter, an emergency shutdown flag bit parameter, a warning shutdown flag bit parameter, a Fuel Cell temperature probe total number parameter, an FCU (Fuel Cell Unit) heartbeat signal parameter, a conductivity parameter, a stack standard deviation parameter, a component maintenance number parameter, and the like; wherein, the maximum voltage node parameter can be a maximum voltage node parameter, a minimum voltage node parameter and the like; the stack feedback parameters may be parameters including one or more controllable stack feedback data; the component maintenance count parameter may be one or more pieces of component maintenance count data.
The circuit parameters may include circuit current parameters, DCDC parameters, buck DCDC parameters, and circuit set parameters, among others. Wherein the electrical circuit current parameter may be a feedback data comprising one or more electrical circuit currents; the DCDC parameters may be parameters including DCDC warning parameters, DCDC temperature parameters, DCDC heartbeat parameters, DCDC status parameters, and the like; the DCDC parameters may include a buck DCDC status sink, a buck DCDC output current parameter, a buck DCDC output voltage parameter, a buck DCDC warning parameter, a buck DCDC temperature parameter, a buck DCDC heartbeat parameter, and the like; the electric circuit setting parameter may be a parameter value preset for rated operation of one or more of the above-mentioned electric circuit current parameter, DCDC parameter, and step-down DCDC parameter.
The cold air path parameters may include a cooling path feedback parameter, a water pump feedback parameter, a PTC (Pulse-Tube refrigerator) parameter, a PTC feedback parameter, a coolant parameter, a cooling path setting parameter, and the like. Wherein the cooling circuit feedback parameters may be one or more of cooling circuit controllable feedback data and one or more of cooling circuit uncontrollable feedback data; the water pump parameters can include water pump warning parameters, water pump heartbeat parameters and the like; the water pump feedback parameter may be a feedback parameter including one or more water pump feedback data; the PTC parameters can include PTC warning instruction parameters, PTC working state parameters, PTC heartbeat parameters and the like; the PTC feedback parameters may be data comprising one or more PTC feedback data; the coolant parameters may include coolant level parameters, etc.; the cooling path setting parameter may be a parameter value preset for rated operation of one or more of the above parameters including the cooling path feedback parameter, the water pump feedback parameter, the PTC feedback parameter, and the coolant parameter.
The hydrogen path parameters may include hydrogen circulation pump parameters, hydrogen circulation pump feedback parameters, hydrogen storage pressure parameters, hydrogen storage and transportation FCU heartbeat parameters, FCU hydrogen storage and heartbeat parameters, hydrogen storage parameters, hydrogen path feedback parameters, hydrogen path setting parameters, and the like. The hydrogen circulating pump parameters can comprise hydrogen circulating pump state parameters, hydrogen circulating pump heartbeat parameters and the like; the hydrogen circulation pump feedback parameter may be feedback data including one or more hydrogen circulation pumps; the hydrogen pump feedback parameters may be data including one or more hydrogen pump feedback; the hydrogen storage pressure parameter may be feedback data including one or more hydrogen storage pressure values; the hydrogen storage parameter may include a hydrogen storage highest concentration parameter, a hydrogen storage highest concentration number parameter, a hydrogen storage highest temperature parameter, a hydrogen storage state parameter, a hydrogen storage warning level parameter, a hydrogen storage warning code parameter, and the like; the hydrogen path feedback parameters may be one or more hydrogen path feedback data; the hydrogen path setting parameter may be a parameter value preset for the rated operation condition of one or more parameters of the above parameters including the hydrogen circulation pump parameter, the hydrogen circulation pump feedback parameter, the hydrogen storage pressure parameter, the hydrogen storage and delivery FCU heartbeat parameter, the FCU hydrogen delivery and storage heartbeat parameter, the hydrogen storage parameter, and the hydrogen path feedback parameter.
The air path parameters can comprise air path feedback parameters, air quantity adjustment times, air compressor parameters, air compressor feedback parameters, air compressor controller feedback parameters, air path setting parameters and the like. Wherein the air path feedback parameters may be one or more of air path controllable feedback data and one or more of air path uncontrollable feedback data; the air compressor parameters can comprise air compressor heartbeat parameters, air compressor software version parameters, air compressor warning sign parameters and the like; the air compressor feedback parameters may include one or more pieces of air compressor feedback data; the feedback parameters of the air compressor controller can comprise one or more pieces of feedback data of the air compressor controller; the air path setting parameter may be a parameter value preset for rated operation conditions of one or more parameters of the air path feedback parameter, the air volume adjustment frequency, the air compressor parameter, the air compressor feedback parameter, the air compressor controller feedback parameter, the air path setting parameter and the like.
The environmental parameters may include environmental feedback parameters, vehicle set parameters, vehicle parameters, accumulated driving range parameters, power bus parameters, current limit parameters, and the like. Wherein the environmental feedback parameter may be a parameter comprising one or more items of environmental feedback data; the vehicle feedback parameters may include one or more vehicle feedback data; the whole vehicle setting parameter can be a parameter value preset for one parameter or rated operation conditions of a plurality of parameters in the whole vehicle feedback parameters; the vehicle parameters may include vehicle operating mode parameters, vehicle state of charge parameters, and the like; the parameters of the power connection bus can be parameters including current parameters of the power connection bus, voltage parameters of the power connection bus and the like; the current limit parameter may include a recharging current limit parameter, a discharging current limit parameter, and the like, which is not particularly limited in this embodiment.
The current performance parameters of each target subsystem comprise data of the performance parameters of each target subsystem at the current time or comprise data of the performance parameters of each target subsystem at the current time and data of the performance parameters of each target subsystem at a plurality of historical times before the current time.
Firstly, each target subsystem in a target battery system is scanned, and the current performance parameters of each target subsystem and the system category to which each target subsystem belongs are obtained according to the scanning result.
Optionally, determining the system category to which each target subsystem belongs may be first obtaining a subsystem number corresponding to each target subsystem, and then obtaining the system category to which each target subsystem belongs according to a mapping relationship between a preset subsystem number and the system category.
102, determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs;
then, according to the preset mapping relation between each system category and the feature extraction policy and the system category to which each target subsystem belongs, the feature extraction policy corresponding to the target subsystem is obtained; similarly, according to the preset mapping relation between each system type and the target performance prediction model and the system type to which each target subsystem belongs, the target performance prediction model corresponding to the target subsystem is obtained.
The target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
Alternatively, for the establishment of the target performance prediction model, a plurality of target performance prediction models may be established for different system classes one-to-one respectively, and each system class and each performance prediction model may be stored in a mapping relationship.
Before executing step 102, it is necessary to establish a corresponding performance prediction model for different system classes one-to-one respectively, and establish a one-to-one mapping relationship between the different system classes and the corresponding performance prediction models.
For each system class corresponding performance prediction model, the construction steps include:
first, the historical performance parameters and performance labels of the sample subsystem under the system category at each historical time of the full life cycle are obtained.
After the historical performance parameters and performance labels of the sample subsystem are obtained, the historical performance parameters of the sample subsystem can be subjected to combined feature selection according to a feature extraction strategy corresponding to the sample subsystem, historical target features corresponding to the sample subsystem are obtained, then a sample data set is constructed according to the historical target features and the performance labels, the sample data set is divided into a training set and a testing set, and then the deep learning model is subjected to iterative training according to the training set and a decision tree algorithm until the trained deep learning model is subjected to testing by using the testing set, so that a target performance prediction model corresponding to each target subsystem is obtained.
Step 103, based on the feature extraction strategy corresponding to each target subsystem, performing combination feature selection on the multi-dimensional performance features to obtain current target features corresponding to each target subsystem;
features extracted for different feature extraction strategies are different.
And then, according to a feature extraction strategy corresponding to the target subsystem, determining features required to be selected for the current performance parameter so as to extract corresponding current target features from the current performance parameter, and inputting the current target features into a target performance prediction model corresponding to the target subsystem so as to obtain a current performance prediction value of the target subsystem.
104, inputting the current target characteristics corresponding to each target subsystem into a target performance prediction model corresponding to each target subsystem to obtain a current performance prediction value of each target subsystem;
the current performance prediction value may be a prediction of the real-time performance value of the target subsystem by the target performance prediction model at the current time.
Optionally, the target feature is learned based on the target energy consumption prediction model, so that the energy consumption predicted value of the target vehicle at the current moment can be obtained through prediction output.
Step 105, determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems;
alternatively, the current performance predicted value of the target battery system may be obtained by superimposing (weighted addition) the current performance predicted values of the target subsystems according to the actual scenario, or may be obtained by inputting the current performance predicted values of the target subsystems into a preset scoring model for training, which is not specifically limited in this embodiment.
The method comprises the steps of firstly, obtaining current performance parameters of all target subsystems in a target battery system and system types of all target subsystems; the system category of each target subsystem is acquired, and a feature extraction strategy and a target performance prediction model corresponding to each target subsystem are determined according to the acquired system category; then, according to the corresponding feature extraction strategy, carrying out combined feature selection on the current performance parameters of each target subsystem to obtain the current target features corresponding to each target subsystem; then, the current target characteristics corresponding to each target subsystem are input into the corresponding target performance prediction model, the output result is used as the current performance prediction value of each target subsystem, then the current performance prediction value of each target subsystem is integrated and determined, a set of complete full-automatic battery performance prediction flow is established, the corresponding characteristic extraction strategy and the corresponding target performance prediction model are adaptively selected for each target subsystem, and the prediction accuracy is improved while the real-time effective performance prediction of the target battery system is realized.
In some embodiments, the method further comprises:
the following steps are performed for each of the target subsystems:
acquiring historical performance parameters and performance labels of a sample subsystem under a system category to which a current target subsystem belongs at each historical moment; calculating a first correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a chi-square test algorithm; calculating a second correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a spearman algorithm; determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the first correlation coefficient and the second correlation coefficient; establishing a mapping relation between the system category to which the current target subsystem belongs and the feature extraction strategy;
based on the system category to which each target subsystem belongs, determining a feature extraction strategy corresponding to each target subsystem comprises: and acquiring a feature extraction strategy corresponding to each target subsystem according to the mapping relation and the system category to which each target subsystem belongs.
Firstly, according to the system category of the current target subsystem, acquiring a sample subsystem corresponding to the system category, scanning the sample subsystem, and acquiring historical performance parameters and performance labels of the sample subsystem at each historical moment of the whole life cycle according to the scanning result.
Optionally, the performance label may be obtained by pre-marking according to historical performance parameters of each historical time of different sample subsystems; in other words, there is a one-to-one mapping relationship between the historical performance parameters of each historical time of the sample subsystem and the performance labels, and the performance labels of the sample subsystem can be obtained according to the historical performance parameters of each historical time of the sample subsystem.
Optionally, the sample subsystems and the data thereof can be combined and classified in advance according to the system category to which each target subsystem belongs and stored in a subsystem numbering mode, so that the original data of the battery system in which the sample subsystem is located is divided into a plurality of categories of sample subsystem data, and corresponding performance prediction model construction is carried out on different categories of sample subsystem data, so that the data volume of each performance prediction model construction can be reduced, the problem of overlarge data volume is solved at the characteristic level, the time cost is reduced, and the prediction effect is improved.
Data classification is the grouping together of data that has some common attribute or feature, and the distinction of data is made by the attribute or feature of its category. In other words, the information with the same content and the same property and the information requiring unified management are gathered together, the information which is different and needs to be managed respectively are distinguished, and then the relation among the sets is determined, so that an organized classification system is formed. The purpose of data classification is to assign new data objects to a correct class based on their attributes. The classification of data emphasizes the classification by attribute and feature according to the kind.
And then, carrying out feature selection on each sample subsystem under each system category so as to determine a feature extraction strategy corresponding to each system category. The research of high-dimensional data is very challenging, and under the premise of ensuring the prediction precision of a learning algorithm, the sample demand during training is exponentially increased along with the improvement of the feature dimension. The feature selection can remove some irrelevant and redundant features, so that the effects of reducing the dimension, the number of the features, the running time and the running time of an algorithm are achieved. The chi-square test algorithm and the Spearman (Spearman) algorithm are two relatively common feature selection methods.
As shown in fig. 3, taking a historical time of a certain sample subsystem as an example, the correlation between the performance characteristics is different, and the correlation between part of the performance characteristics is almost 0, which seriously affects the battery performance prediction.
Alternatively, a first correlation coefficient between each performance feature in the historical performance parameters and the performance label may be calculated based on a chi-square test algorithm; the first correlation coefficient may be a chi-square value calculated by a chi-square checking algorithm.
X in chi-square test 2 The value (i.e., chi-square value) describes the degree of correlation between the independent variable and the dependent variable. Wherein X is 2 The larger the difference between the actual and expected values, the smaller the independence between the two variables, and the more relevant; x is X 2 The smaller the expression actual-to-expected gap approximation, the greater the independence and the smaller the correlation, so X can be used 2 Values are used to do relevant tasks such as feature selection, etc. The specific formula for chi-square value calculation is as follows:
wherein A is the actual frequency, T is the theoretical frequency, and X 2 Is a chi square value.
Likewise, a second correlation coefficient between each performance feature in the historical performance parameters and the performance tag may be calculated based on a spearman algorithm; wherein the second correlation coefficient may be a spearman correlation coefficient calculated by a spearman algorithm.
The spearman correlation coefficient is a non-parametric indicator and the data used for calculation is rank. When some external factor is the main feature affecting the mean value of the system performance, the value is close to 1, and conversely, close to 0. The calculation expression of the spearman correlation coefficient is as follows:
wherein ρ is the spearman correlation coefficient between 2 sets of vectors; n is the number of samples; r, S are each performance feature and performance tag; r is R i 、S i The i-th parameter of each performance characteristic and the performance label vector is respectively;the average levels of the performance characteristic quantity and the performance label vector are respectively.
Then, determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the first correlation coefficient and the second correlation coefficient;
optionally, the feature extraction policy corresponding to the system class to which the target subsystem belongs may be determined according to one or more relatively more sensitive performance features in the sample subsystem, that is, one or more performance features with higher similarity, and the like; for example, a performance feature set with the first correlation coefficient and the second correlation coefficient higher than a preset threshold value is selected, and a feature extraction strategy is determined according to the performance feature set; the first correlation coefficient and the second correlation coefficient may be weighted and added to obtain a total correlation coefficient, and the feature extraction policy may be determined according to a performance feature set in which the total correlation coefficient is higher than a preset threshold, which is not specifically limited in this embodiment.
Optionally, then establishing a mapping relationship between the system category to which the current target subsystem belongs and the feature extraction policy, wherein the mapping relationship can be a mapping relationship directly existing between the system category to which the configuration current target subsystem belongs and the feature extraction policy; the mapping relationship may also be constructed by setting an identifier for the system class to which the current target subsystem belongs and associating it with the target feature extraction policy one by one, which is not specifically limited in this embodiment.
After the mapping relation between the feature extraction strategy and the system category to which each target subsystem belongs is established, the target feature extraction strategy corresponding to the system category to which the target subsystem belongs can be obtained according to the mapping relation and the system category to which the target subsystem belongs.
Accordingly, when the feature extraction policy corresponding to the system class to which the target subsystem belongs is obtained according to the mapping relationship, the feature extraction policy may be directly obtained according to the mapping relationship between the two, or may be obtained according to the mapping relationship between the identifier preset for the system class to which the target subsystem belongs and the target feature extraction policy, which is not specifically limited in this embodiment.
According to the embodiment, for each target subsystem, the sample subsystem under the system category to which the current target subsystem belongs and the historical performance parameters and performance labels of the sample subsystem at each historical moment are obtained; then, based on a Kalman test algorithm and a Szechwan algorithm, respectively calculating a first correlation coefficient and a second correlation coefficient between each performance characteristic and a performance label in the historical performance parameters; then, determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the first correlation coefficient and the second correlation coefficient, and establishing a mapping relation between the system category to which the current target subsystem belongs and the feature extraction strategy; the feature extraction strategy corresponding to each target subsystem can be obtained according to the mapping relation and the system category to which each target subsystem belongs, and the feature extraction strategy applicable to the system category to which each target subsystem belongs can be obtained more flexibly, so that the real-time effective performance prediction of the target battery system can be realized, the time cost is reduced, and the prediction effect is improved.
In some embodiments, the determining, according to the first correlation coefficient and the second correlation coefficient, a feature extraction policy corresponding to a system class to which the current target subsystem belongs includes: comparing the first correlation coefficient with a first preset value; according to the comparison result, at least one performance characteristic of which the first correlation coefficient is larger than the first preset value is determined in the historical performance parameters; determining a performance characteristic of which the second correlation coefficient is larger than a second preset value from the at least one performance characteristic; and determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the performance feature of which the second correlation coefficient is larger than a second preset value.
Fig. 4 is a flowchart illustrating a method for determining a feature extraction policy according to the present embodiment. As shown in fig. 4, determining a feature extraction policy corresponding to a system class to which a current target subsystem belongs specifically includes the following steps:
step 401, data preparation, namely obtaining historical performance parameters and performance labels of a sample subsystem under a system category to which a current target subsystem belongs at each historical moment;
step 402, after finishing data preparation, the data may be preprocessed;
Step 403, after the chi-square value calculation is completed, dividing each performance feature into a high section and a low section according to a first preset value and the calculated first correlation coefficient, and obtaining a division result;
step 404, selecting performance characteristics of the high section to form a characteristic subset according to the division result;
comparing the first correlation coefficient of each performance characteristic in the historical performance parameter with a first preset value, and determining at least one performance characteristic (namely a characteristic subset) with the first correlation coefficient larger than the first preset value in the historical performance parameter according to a comparison result.
Step 405, after the calculation of the second correlation coefficient of the feature subset based on the spearman algorithm is completed, comparing the obtained second correlation coefficient with a second preset value and obtaining a comparison result;
step 406, eliminating the features of the interval which do not meet the requirement of the second preset value, and forming a new feature subset from the rest performance features; and eliminating the performance characteristics with the second correlation number smaller than or equal to a second preset value from at least one performance characteristic, and determining the reserved performance characteristics with the second correlation coefficient larger than the second preset value.
Step 407, completing feature selection, namely taking a set formed by the performance features with the reserved second correlation number larger than a second preset value as a feature extraction strategy corresponding to the system category to which the current target subsystem belongs, and ending the current step.
The embodiment obtains a comparison result by firstly comparing the first correlation coefficient with a first preset value, and determines at least one performance characteristic of which the first correlation coefficient is larger than the first preset value in the historical performance parameters according to the comparison result; and then, determining the performance characteristics with the second correlation number larger than a second preset value from at least one performance characteristic, and determining the characteristic extraction strategy corresponding to the system category to which the current target subsystem belongs according to the performance characteristics with the second correlation number larger than the second preset value, so that the corresponding characteristic extraction strategy is adaptively selected for each target subsystem, and the target battery system is effectively and practically predicted.
Fig. 5 is a flow chart of a training process of the target energy consumption prediction model provided in this embodiment, as shown in fig. 5, in some embodiments, the training steps of the target performance prediction model corresponding to each target subsystem include:
step 501, obtaining the historical performance parameters of the sample subsystem and the performance labels of the sample subsystem under the system category to which each target subsystem belongs;
wherein the historical performance parameters include performance parameters of the sample subsystem collected at one or more historical moments prior to the current moment.
It should be noted that, for a sample subsystem at any historical time, the historical performance parameter corresponding to the sample subsystem may be a performance parameter only including the historical time; or include the performance parameters at the historical time and the performance parameters at a plurality of historical times prior to the historical time.
Step 502, performing combined feature selection on the multidimensional performance features of the sample subsystem based on feature extraction strategies corresponding to the sample subsystems to obtain historical target features corresponding to the sample subsystems;
and according to the feature extraction strategy corresponding to the sample subsystem, the combination feature selection can be carried out on the target features corresponding to the historical performance parameters so as to facilitate the training of the deep learning model.
Because the correlation of the characteristics in different sample subsystems is different, the selected characteristics are different, and therefore, for the historical performance parameters of the sample subsystems, independent deep learning models are adopted respectively during training, the historical target characteristics obtained by combining characteristic selection in the historical performance parameters of the sample subsystems are selected as the input of the models, and the effective distinction of the sample subsystems and the efficient learning training of the deep learning models are realized.
Step 503, constructing a sample data set according to the historical target characteristics and the performance labels;
storing the historical target characteristics corresponding to each sample subsystem and the performance labels of the sample subsystems in a one-to-one correspondence manner, and forming sample data sets corresponding to different system types by all the sample subsystems in different system types;
step 504, dividing the sample data set into a training set and a testing set;
step 505, training the deep learning model according to the training set and the decision tree algorithm;
model training was performed using the LightGBM (Light Gradient Boosting Machine, decision tree) algorithm. The LightGBM is an integrated strong learner model based on a distributed gradient lifting tree (Gradient Boosting Decision Tree, GBDT), which is applied to regression problems by virtue of its rapidity, low internal consumption, high accuracy, etc. The LightGBM uses a decision tree as a base learner, and its training formula can be expressed as:
wherein H is t Is the t learner; Θ is the aggregate space of all learners.
Fig. 6 is a schematic flow chart of model training using decision tree algorithm according to the present embodiment.
The LightGBM algorithm proposes a Gradient-based One-Side Sampling algorithm (GOSS) for the problem of large number of samples, and a mutually exclusive feature binding algorithm (Exclusive Feature Bundling, EFB) for the problem of large number of features. Taking a GOSS algorithm sampling as an example for the training set, as shown in fig. 6, the steps of model training by using the decision tree algorithm are as follows:
Firstly, starting model training;
then, selecting a large gradient training sample in a training set;
the LightGBM algorithm may be to screen training samples using the derived gradients. The larger the gradient should be the less learning. If the large gradient training samples can be predicted correctly, the contribution to the gain will be greater, so it is desirable to accurately divide the large gradient training samples when the nodes split, and the small gradient training samples may be erroneous. In this case, the training samples are screened while retaining the large gradient training samples in the complete training set.
Then, selecting a small gradient training sample in a training set;
then, constructing a new small gradient training sample;
and sampling the small gradient training samples in the training set under the condition of keeping the data distribution unchanged as much as possible.
Then, merging the samples;
and combining the complete large-gradient training sample and the sampled small-gradient training sample to obtain a new training set.
Then, training a weak learner;
the new training set is input into the model to train the weak learner.
Then, judging whether an iteration termination condition is reached;
if the iteration termination condition is judged to be met, model parameters in a parameter space are input into a final prediction result of the deep learning model; otherwise, the step of selecting the large gradient training sample is returned again, and the iterative training is continued.
And finally, outputting a final prediction result.
Step 506, testing the model performance of the trained deep learning model according to the test set, and adjusting the model structure and/or model parameters of the trained deep learning model when the trained deep learning model fails the test;
and step 507, training the adjusted deep learning model according to the training set and the decision tree algorithm until the trained deep learning model passes the test to obtain a target performance prediction model corresponding to each target subsystem.
After each training is completed, whether the deep learning model reaches an iteration termination condition is required to be judged, namely, a test set is used for testing the model performance of the trained deep learning model.
If the trained deep learning model fails the test, the model structure and/or model parameters of the trained deep learning model are adjusted, and then the training of the adjusted deep learning model is continued according to the training set and the decision tree algorithm until the trained deep learning model passes the test, so that a target performance prediction model corresponding to each target subsystem is obtained; and if the trained deep learning model passes the test, taking the trained deep learning model as a target performance prediction model.
By way of example, the model parameters may be a combination of one or more of parameters including num_leave (number of leaf nodes), learning_rate (learning rate), feature_fraction (selection feature number), bagging_fraction (usage data amount), max_depth (maximum depth), min_child_weight (sample weight sum), and the like. Where num_leave may be the number of leaf nodes used to control each decision tree; the learning_rate may be a speed for updating the weight of each model during gradient descent; feature_fraction may be used to control how many features are selected each time a decision tree grows; bagging_fraction may be used to control how much data the model will use in each iteration; max_depth may be the depth used to decide the decision tree, i.e., the maximum depth reached by each leaf node; min_child_weight may be the sum of all sample weights in the smallest child node, which is not specifically limited in this embodiment.
It can be understood that, after the target performance prediction model is obtained, in order to effectively evaluate the prediction effect, the present embodiment researches the following three commonly used evaluation indexes of the regression model:
the root mean square error (Root Mean Squared Error, RMSE) is the square root of the ratio of the square of the deviation of the predicted value from the true value to the number of observations, which is always limited in actual measurement and can only be replaced by the most reliable value. The calculation method is as follows:
The mean absolute error (Mean Absolute Error, MAE) is the mean absolute error between the true and predicted values, and is typically used to measure how close the prediction is to the actual result. The calculation method is as follows:
determining coefficient (R) 2 ) The prediction value is measured as compared with the case of using only the average value, and the interval is usually between (0, 1). The calculation method is as follows:
as shown in Table 1, R is used 2 The comparison between the prediction result of the decision tree algorithm (hereinafter abbreviated as LightGBM) used in this embodiment and the prediction result of the existing linear regression algorithm shows that the prediction error of LightGBM is smaller than that of linear regression, which indicates the superiority of the decision tree algorithm in this embodiment in training.
TABLE 1 summary of model Performance prediction results
| Evaluation index | LightGBM | Linear regression |
| R 2 | 0.9995 | 0.9644 |
| MAE | 0.0009 | 0.0082 |
| RMSE | 0.0012 | 0.0106 |
Fig. 7 is a schematic diagram showing a comparison between a predicted value of system performance and an actual value of system performance, and as shown in fig. 7, it can be found that the target performance prediction model of this embodiment can accurately and effectively predict the performance of the battery system, which indicates the superiority of the target performance prediction model of this embodiment.
In the embodiment, firstly, a sample subsystem under a system category to which each target subsystem belongs is obtained, and a historical performance parameter of the sample subsystem and a performance label of the sample subsystem are obtained; then, according to the feature extraction strategies corresponding to the sample subsystems, carrying out combination feature selection on the historical target features of the sample subsystems to obtain the historical target features corresponding to the sample subsystems, and integrating the historical target features with the performance labels to construct a sample data set; then dividing the obtained sample data set into a training set and a testing set; training the deep learning model according to the training set and the decision tree algorithm, testing the model performance of the trained deep learning model according to the testing set, adjusting the model structure and/or model parameters of the trained deep learning model if the trained deep learning model fails the test, and continuing training the adjusted deep learning model until the trained deep learning model passes the test, so as to finally obtain a target performance prediction model corresponding to each target subsystem, and adaptively selecting the corresponding target performance prediction model for each target subsystem, thereby being beneficial to realizing real-time effective performance prediction of the target battery system and improving prediction accuracy.
In some embodiments, before the selecting the combined feature from the multi-dimensional performance features of the sample subsystem based on the feature extraction policy corresponding to each sample subsystem to obtain the historical target feature corresponding to the sample subsystem, the method further includes: determining whether unsteady performance characteristics exist in historical performance parameters of the sample subsystem at each historical moment based on a steady-state processing algorithm; and deleting the historical performance parameter and the performance label at any historical moment when the unsteady performance characteristic exists in the historical performance parameter at any historical moment.
Judging whether the historical performance parameters of the sample subsystem at each historical moment have unsteady performance characteristics by using a steady-state processing algorithm, and deleting the historical performance parameters of the sample subsystem at the historical moment and the performance labels of the sample subsystem at the historical moment if the historical performance parameters at each historical moment have unsteady performance characteristics; if the non-steady state performance characteristics do not exist in the historical performance parameters of the historical moment, the historical performance parameters of the sample subsystem of the historical moment and the performance labels of the sample subsystem of the historical moment are reserved.
The specific steps for processing the characteristic data by using the steady-state processing algorithm are as follows:
the first step: screening and leaving data of continuous change of heartbeat parameters of the FCU sending VCU;
the continuous change means that there is a change in 60 adjacent rows, and if more than 60 rows have no change, only the first row of data is reserved.
And a second step of: screening data with the highest warning level parameter of 1 or 0;
and a third step of: screening data leaving circuit setting parameter 1 (argument) 192;
fourth step: screening and leaving data of more than 600 rows continuously as a group on the basis of the third step;
where consecutive means that none of the rows within the set of data is screened out.
Fifth step: screening leaves data that satisfies both 74.5< cooling path feedback parameter 3 (controllable) <77.5 and 74.5< cooling path feedback parameter 5 (controllable) < 77.5;
sixth step: the first 180 rows of data for each set of data are screened out.
And converting the historical performance parameters into steady-state data after the steady-state processing is completed. The steady state of the battery refers to the continuous steady running state of the battery, and the steady state performance evaluation index is screened according to a specific rule to judge the precondition of the performance of the hydrogen fuel cell system.
FIG. 8 is a flow chart of a target performance prediction model training process and a prediction process corresponding to a target subsystem. As shown in fig. 8, taking the target battery system as an example of the hydrogen fuel battery system, the overall process of training and predicting the target performance prediction model is as follows:
Firstly, starting model training;
next, historical performance parameters in the hydrogen fuel cell system are collected;
then, data grouping is carried out on the historical performance parameters, and the historical performance parameters of the sample subsystem are obtained;
then, performing steady-state processing on the historical performance parameters;
then, performing feature selection on the history performance parameters after steady-state processing to obtain history target features;
it can be appreciated that after the historical target features are obtained, the historical target features and the performance tags are integrated to construct a sample data set.
Then, splitting the sample data set into a training set and a testing set;
then training the deep learning model according to the training set and the LightGBM algorithm;
then, performing model verification on the model performance of the trained deep learning model according to the test set, and judging whether the trained deep learning model passes the test;
and under the condition that the trained deep learning model fails the test, adjusting the model structure and/or model parameters of the trained deep learning model, and continuously training the adjusted deep learning model according to the training set and the decision tree algorithm until the trained deep learning model passes the test.
Then, obtaining a target performance prediction model corresponding to each target subsystem, and inputting the current target characteristics corresponding to the target subsystem into the target performance prediction model corresponding to each target subsystem;
the current target characteristics corresponding to the target subsystems are obtained by firstly obtaining the current performance parameters of each target subsystem in the target battery system and the system category to which each target subsystem belongs; the current performance parameters comprise multidimensional performance characteristics, then a characteristic extraction strategy and a target performance prediction model corresponding to each target subsystem are determined based on the system category to which each target subsystem belongs, and then combined characteristic selection is performed on combined characteristic selection based on the characteristic extraction strategy corresponding to each target subsystem to obtain the current target characteristics corresponding to each target subsystem.
And finally, taking the output result as a current performance predicted value of the target subsystem.
According to the embodiment, whether the unsteady performance characteristics exist in the historical performance parameters of the sample subsystem at each historical moment or not is judged based on a steady-state processing algorithm, and if the unsteady performance characteristics exist in the historical performance parameters at any historical moment, the historical performance parameters and the performance labels at any historical moment are deleted, so that the historical performance parameters are converted into steady-state data, and the accuracy of pool performance prediction is improved.
In some embodiments, the determining the current performance prediction value of the target battery system according to the current performance prediction values of all the target subsystems includes: and carrying out weighted addition or averaging on the current performance predicted values of all the target subsystems to obtain the current performance predicted value of the target battery system.
Alternatively, the current performance predicted values of the target battery systems may be obtained by weighted addition or averaging of the current performance predicted values of all the target subsystems; for example, in this embodiment, the current performance prediction value of each target subsystem is calculated according to the following formula, so as to obtain the current performance prediction value of the target battery system:
wherein U is 1 Refers to the current performance prediction value of the target subsystem;refers to the current performance prediction value of the target battery system.
According to the embodiment, the current performance predicted values of the target battery systems are obtained through weighted addition or average value calculation of the current performance predicted values of all the target subsystems, so that the prediction accuracy is improved while the target battery systems are effectively predicted in real time.
The energy storage system performance prediction device based on the combination feature selection provided by the invention is described below, and the energy storage system performance prediction device based on the combination feature selection described below and the energy storage system performance prediction method based on the combination feature selection described above can be correspondingly referred to each other.
As shown in fig. 9, a schematic structural diagram of an energy storage system performance prediction apparatus based on combination feature selection according to the present invention is provided, where the apparatus includes:
an obtaining module 901, configured to obtain current performance parameters of each target subsystem in a target battery system, and a system class to which each target subsystem belongs; the current performance parameters include multi-dimensional performance characteristics;
a first determining module 902, configured to determine, based on a system class to which each of the target subsystems belongs, a feature extraction policy and a target performance prediction model corresponding to each of the target subsystems;
a selection module 903, configured to perform combined feature selection on the multi-dimensional performance features based on feature extraction policies corresponding to the target subsystems, so as to obtain current target features corresponding to the target subsystems;
the prediction module 904 is configured to input a current target feature corresponding to each target subsystem into a target performance prediction model corresponding to each target subsystem, so as to obtain a current performance prediction value of each target subsystem;
a second determining module 905, configured to determine, according to the current performance predicted values of all the target subsystems, a current performance predicted value of the target battery system;
The target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
According to the energy storage system performance prediction device based on combination feature selection, the current performance parameters of all target subsystems in the target battery system and the system category to which all target subsystems belong are firstly obtained; the system category of each target subsystem is acquired, and a feature extraction strategy and a target performance prediction model corresponding to each target subsystem are determined according to the acquired system category; then, according to the corresponding feature extraction strategy, carrying out combined feature selection on the current performance parameters of each target subsystem to obtain the current target features corresponding to each target subsystem; then, the current target characteristics corresponding to each target subsystem are input into the corresponding target performance prediction model, the output result is used as the current performance prediction value of each target subsystem, then the current performance prediction value of each target subsystem is integrated and determined, a set of complete full-automatic battery performance prediction flow is established, the corresponding characteristic extraction strategy and the corresponding target performance prediction model are adaptively selected for each target subsystem, and the prediction accuracy is improved while the real-time effective performance prediction of the target battery system is realized.
In some embodiments, the first determining module 902 is specifically configured to perform, for each of the target subsystems, the following steps: acquiring historical performance parameters and performance labels of a sample subsystem under a system category to which a current target subsystem belongs at each historical moment; calculating a first correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a chi-square test algorithm; calculating a second correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a spearman algorithm; determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the first correlation coefficient and the second correlation coefficient; establishing a mapping relation between the system category to which the current target subsystem belongs and the feature extraction strategy; based on the system category to which each target subsystem belongs, determining a feature extraction strategy corresponding to each target subsystem comprises: and acquiring a feature extraction strategy corresponding to each target subsystem according to the mapping relation and the system category to which each target subsystem belongs.
In some embodiments, the first determining module 902 is further to: comparing the first correlation coefficient with a first preset value; according to the comparison result, at least one performance characteristic of which the first correlation coefficient is larger than the first preset value is determined in the historical performance parameters; determining a performance characteristic of which the second correlation coefficient is larger than a second preset value from the at least one performance characteristic; and determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the performance feature of which the second correlation coefficient is larger than a second preset value.
In some embodiments, the energy storage system performance prediction apparatus based on the combined feature selection further comprises a training module, specifically configured to: acquiring the historical performance parameters of sample subsystems and performance labels of the sample subsystems under the system category to which each target subsystem belongs; based on the feature extraction strategy corresponding to each sample subsystem, carrying out combination feature selection on the multidimensional performance features of the sample subsystem to obtain historical target features corresponding to the sample subsystem; constructing a sample data set according to the historical target characteristics and the performance labels; dividing the sample data set into a training set and a testing set; training the deep learning model according to the training set and the decision tree algorithm; according to the test set, testing the model performance of the trained deep learning model, and adjusting the model structure and/or model parameters of the trained deep learning model under the condition that the trained deep learning model fails the test; and training the adjusted deep learning model according to the training set and the decision tree algorithm until the trained deep learning model passes the test to obtain a target performance prediction model corresponding to each target subsystem.
In some embodiments, the training module is further to: determining whether unsteady performance characteristics exist in historical performance parameters of the sample subsystem at each historical moment based on a steady-state processing algorithm;
and deleting the historical performance parameter and the performance label at any historical moment when the unsteady performance characteristic exists in the historical performance parameter at any historical moment.
In some embodiments, the second determining module 905 is specifically configured to: and carrying out weighted addition or averaging on the current performance predicted values of all the target subsystems to obtain the current performance predicted value of the target battery system.
Fig. 10 illustrates a physical structure diagram of an electronic device, as shown in fig. 10, which may include: a processor 1001, a communication interface (Communications Interface) 1002, a memory 1003, and a communication bus 1004, wherein the processor 1001, the communication interface 1002, and the memory 1003 perform communication with each other through the communication bus 1004. The processor 1001 may invoke logic instructions in the memory 1003 to perform an energy storage system performance prediction method selected based on the combined characteristics, the method comprising: acquiring current performance parameters of each target subsystem in a target battery system and the class of the system to which each target subsystem belongs; the current performance parameters include multi-dimensional performance characteristics; determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs; based on the feature extraction strategy corresponding to each target subsystem, carrying out combination feature selection on the multi-dimensional performance features to obtain current target features corresponding to each target subsystem; inputting the current target characteristics corresponding to each target subsystem into a target performance prediction model corresponding to each target subsystem to obtain a current performance prediction value of each target subsystem; determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems; the target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
The electronic device provided in the embodiment of the present invention is used for executing the above embodiments of the method, and specific flow and details refer to the above embodiments, which are not repeated herein.
Further, the logic instructions in the memory 1003 described above may be implemented in the form of software functional units and sold or used as a separate product, and may be stored in a computer readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product including a computer program, the computer program being storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of performing the energy storage system performance prediction method based on the combination feature selection provided by the methods above, the method comprising: acquiring current performance parameters of each target subsystem in a target battery system and the class of the system to which each target subsystem belongs; the current performance parameters include multi-dimensional performance characteristics; determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs; based on the feature extraction strategy corresponding to each target subsystem, carrying out combination feature selection on the multi-dimensional performance features to obtain current target features corresponding to each target subsystem; inputting the current target characteristics corresponding to each target subsystem into a target performance prediction model corresponding to each target subsystem to obtain a current performance prediction value of each target subsystem; determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems; the target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
The electronic device provided in the embodiment of the present invention is used for executing the above embodiments of the method, and specific flow and details refer to the above embodiments, which are not repeated herein.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the energy storage system performance prediction method based on combined feature selection provided by the methods above, the method comprising: acquiring current performance parameters of each target subsystem in a target battery system and the class of the system to which each target subsystem belongs; the current performance parameters include multi-dimensional performance characteristics; determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs; based on the feature extraction strategy corresponding to each target subsystem, carrying out combination feature selection on the multi-dimensional performance features to obtain current target features corresponding to each target subsystem; inputting the current target characteristics corresponding to each target subsystem into a target performance prediction model corresponding to each target subsystem to obtain a current performance prediction value of each target subsystem; determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems; the target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
The electronic device provided in the embodiment of the present invention is used for executing the above embodiments of the method, and specific flow and details refer to the above embodiments, which are not repeated herein.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. The energy storage system performance prediction method based on combination feature selection is characterized by comprising the following steps:
acquiring current performance parameters of each target subsystem in a target battery system and the class of the system to which each target subsystem belongs; the current performance parameters include multi-dimensional performance characteristics;
determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs;
based on the feature extraction strategy corresponding to each target subsystem, carrying out combination feature selection on the multi-dimensional performance features to obtain current target features corresponding to each target subsystem;
Inputting the current target characteristics corresponding to each target subsystem into a target performance prediction model corresponding to each target subsystem to obtain a current performance prediction value of each target subsystem;
determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems;
the target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
2. The energy storage system performance prediction method based on combined feature selection of claim 1, further comprising:
the following steps are performed for each of the target subsystems:
acquiring historical performance parameters and performance labels of a sample subsystem under a system category to which a current target subsystem belongs at each historical moment;
calculating a first correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a chi-square test algorithm;
Calculating a second correlation coefficient between each performance feature in the historical performance parameters and the performance label based on a spearman algorithm;
determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the first correlation coefficient and the second correlation coefficient;
establishing a mapping relation between the system category to which the current target subsystem belongs and the feature extraction strategy;
based on the system category to which each target subsystem belongs, determining a feature extraction strategy corresponding to each target subsystem comprises:
and acquiring a feature extraction strategy corresponding to each target subsystem according to the mapping relation and the system category to which each target subsystem belongs.
3. The method for predicting performance of an energy storage system based on combined feature selection of claim 2, wherein determining a feature extraction policy corresponding to a system class to which the current target subsystem belongs according to the first correlation coefficient and the second correlation coefficient comprises:
comparing the first correlation coefficient with a first preset value;
according to the comparison result, at least one performance characteristic of which the first correlation coefficient is larger than the first preset value is determined in the historical performance parameters;
Determining a performance characteristic of which the second correlation coefficient is larger than a second preset value from the at least one performance characteristic;
and determining a feature extraction strategy corresponding to the system category to which the current target subsystem belongs according to the performance feature of which the second correlation coefficient is larger than a second preset value.
4. A method of predicting performance of an energy storage system based on combined feature selection as claimed in any one of claims 1 to 3, wherein the training step of the target performance prediction model corresponding to each of the target subsystems comprises:
acquiring the historical performance parameters of sample subsystems and performance labels of the sample subsystems under the system category to which each target subsystem belongs;
based on the feature extraction strategy corresponding to each sample subsystem, carrying out combination feature selection on the multidimensional performance features of the sample subsystem to obtain historical target features corresponding to the sample subsystem;
constructing a sample data set according to the historical target characteristics and the performance labels;
dividing the sample data set into a training set and a testing set;
training the deep learning model according to the training set and the decision tree algorithm;
According to the test set, testing the model performance of the trained deep learning model, and adjusting the model structure and/or model parameters of the trained deep learning model under the condition that the trained deep learning model fails the test;
and training the adjusted deep learning model according to the training set and the decision tree algorithm until the trained deep learning model passes the test to obtain a target performance prediction model corresponding to each target subsystem.
5. The method for predicting performance of an energy storage system based on combined feature selection of claim 4, wherein before said combined feature selection is performed on said multi-dimensional performance features of said sample subsystem based on a feature extraction policy corresponding to each of said sample subsystems to obtain historical target features corresponding to said sample subsystem, said method further comprises:
determining whether unsteady performance characteristics exist in historical performance parameters of the sample subsystem at each historical moment based on a steady-state processing algorithm;
and deleting the historical performance parameter and the performance label at any historical moment when the unsteady performance characteristic exists in the historical performance parameter at any historical moment.
6. A method of predicting performance of an energy storage system based on combined feature selection as claimed in any one of claims 1 to 3, wherein said determining a current performance prediction value for said target battery system based on current performance prediction values for all of said target subsystems comprises:
and carrying out weighted addition or averaging on the current performance predicted values of all the target subsystems to obtain the current performance predicted value of the target battery system.
7. An energy storage system performance prediction device based on combined feature selection, comprising:
the system comprises an acquisition module, a control module and a control module, wherein the acquisition module is used for acquiring current performance parameters of all target subsystems in a target battery system and system categories to which all the target subsystems belong; the current performance parameters include multi-dimensional performance characteristics;
the first determining module is used for determining a feature extraction strategy and a target performance prediction model corresponding to each target subsystem based on the system category to which each target subsystem belongs;
the selection module is used for carrying out combination feature selection on the multidimensional performance features based on feature extraction strategies corresponding to the target subsystems to obtain current target features corresponding to the target subsystems;
The prediction module is used for inputting the current target characteristics corresponding to each target subsystem into the target performance prediction model corresponding to each target subsystem to obtain the current performance prediction value of each target subsystem;
the second determining module is used for determining the current performance predicted value of the target battery system according to the current performance predicted values of all the target subsystems;
the target performance prediction model corresponding to each target subsystem is obtained by training a pre-built deep learning model based on a decision tree algorithm, historical performance parameters and performance labels of sample subsystems in system categories to which each target subsystem belongs at each historical moment.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the energy storage system performance prediction method based on combined feature selection of any one of claims 1 to 6 when the program is executed by the processor.
9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the energy storage system performance prediction method based on combined feature selection of any of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the energy storage system performance prediction method based on combined feature selection of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311155658.5A CN117272183A (en) | 2023-09-07 | 2023-09-07 | Energy storage system performance prediction method and device based on combination feature selection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311155658.5A CN117272183A (en) | 2023-09-07 | 2023-09-07 | Energy storage system performance prediction method and device based on combination feature selection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117272183A true CN117272183A (en) | 2023-12-22 |
Family
ID=89215182
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311155658.5A Pending CN117272183A (en) | 2023-09-07 | 2023-09-07 | Energy storage system performance prediction method and device based on combination feature selection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117272183A (en) |
-
2023
- 2023-09-07 CN CN202311155658.5A patent/CN117272183A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ming et al. | A systematic review of machine learning methods applied to fuel cells in performance evaluation, durability prediction, and application monitoring | |
| Jiang et al. | Bayesian learning for rapid prediction of lithium-ion battery-cycling protocols | |
| CN110579709B (en) | Fault diagnosis method for proton exchange membrane fuel cell for tramcar | |
| CN109324291B (en) | Prediction method for predicting service life of proton exchange membrane fuel cell | |
| Zhou et al. | Machine learning as an online diagnostic tool for proton exchange membrane fuel cells | |
| Wu et al. | A hybrid prognostic model applied to SOFC prognostics | |
| Gong et al. | Adaptive optimization strategy of air supply for automotive polymer electrolyte membrane fuel cell in life cycle | |
| CN116449218B (en) | Lithium battery health state estimation method | |
| CN114943372A (en) | Method and device for predicting life of proton exchange membrane based on Bayesian recurrent neural network | |
| Hosseinzadeh et al. | Development and validation of a simple analytical model of the proton exchange membrane fuel cell (PEMFC) in a fork-lift truck power system | |
| Pang et al. | Intelligent health states recognition of fuel cell by cell voltage consistency under typical operating parameters | |
| Wang et al. | A fusion prognostics strategy for fuel cells operating under dynamic conditions | |
| Abd Elaziz et al. | Optimal parameters extracting of fuel cell based on Gorilla Troops Optimizer | |
| CN115146723A (en) | Electrochemical model parameter identification method based on deep learning and heuristic algorithm | |
| Kishor et al. | Fuzzy modeling of fuel cell based on mutual information between variables | |
| He et al. | A quick evaluation method for the lifetime of the fuel cell MEA with the particle filter algorithm | |
| Chang | Equivalent circuit parameters estimation for PEM fuel cell using RBF neural network and enhanced particle swarm optimization | |
| CN117272183A (en) | Energy storage system performance prediction method and device based on combination feature selection | |
| CN115716469A (en) | Output power distribution control method of hybrid power system | |
| Zuo et al. | Post-prognostics decision making strategy to manage the economic lifetime of a two-stack PEMFC system | |
| Chen et al. | A coupled and interactive influence of operational parameters for optimizing power output of cleaner energy production systems under uncertain conditions | |
| Crawford et al. | Development and validation of chemistry agnostic flow battery cost performance model and application to nonaqueous electrolyte systems | |
| Isen et al. | Improved stochastic fractal search algorithm involving design operators for solving parameter extraction problems in real-world engineering optimization problems | |
| Wu et al. | Methods for estimating the accumulated nitrogen concentration in anode of proton exchange membrane fuel cell stacks based on back propagation neural network | |
| CN110807599A (en) | Method, device, server and storage medium for deciding electrochemical energy storage scheme |
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 |