CN113872322A - Micro-grid cloud edge cooperative control method based on deep learning - Google Patents

Micro-grid cloud edge cooperative control method based on deep learning Download PDF

Info

Publication number
CN113872322A
CN113872322A CN202111036998.7A CN202111036998A CN113872322A CN 113872322 A CN113872322 A CN 113872322A CN 202111036998 A CN202111036998 A CN 202111036998A CN 113872322 A CN113872322 A CN 113872322A
Authority
CN
China
Prior art keywords
neural network
network model
edge
characteristic image
label
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
Application number
CN202111036998.7A
Other languages
Chinese (zh)
Inventor
阮诗雅
吴宁
陈卫东
韩帅
肖静
吴晓锐
谭志广
伊然
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Guigang Power Supply Bureau of Guangxi Power Grid Co Ltd
Original Assignee
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Guigang Power Supply Bureau of Guangxi Power Grid Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Electric Power Research Institute of Guangxi Power Grid Co Ltd, Guigang Power Supply Bureau of Guangxi Power Grid Co Ltd filed Critical Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority to CN202111036998.7A priority Critical patent/CN113872322A/en
Publication of CN113872322A publication Critical patent/CN113872322A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
    • H02J13/00002—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by monitoring
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
    • H02J13/00004—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by the power network being locally controlled
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/28—Arrangements for balancing of the load in a network by storage of energy
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
    • H02J2300/20—The dispersed energy generation being of renewable origin
    • H02J2300/22—The renewable source being solar energy
    • H02J2300/24—The renewable source being solar energy of photovoltaic origin
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
    • H02J2300/20—The dispersed energy generation being of renewable origin
    • H02J2300/28—The renewable source being wind energy
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
    • Y02B90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02B90/20—Smart grids as enabling technology in buildings sector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/50—Photovoltaic [PV] energy
    • Y02E10/56—Power conversion systems, e.g. maximum power point trackers
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/12—Monitoring or controlling equipment for energy generation units, e.g. distributed energy generation [DER] or load-side generation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S20/00—Management or operation of end-user stationary applications or the last stages of power distribution; Controlling, monitoring or operating thereof
    • Y04S20/12—Energy storage units, uninterruptible power supply [UPS] systems or standby or emergency generators, e.g. in the last power distribution stages

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Power Engineering (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a micro-grid cloud edge cooperative control method based on deep learning, which comprises the steps of collecting operation state data at an edge controller to generate a characteristic image, receiving model parameters of a cloud server and judging a threshold value, identifying the label type of the characteristic image, directly calling or finely adjusting a source load storage regulation and control strategy issued by the cloud server to serve as a micro-grid regulation and control strategy, and uploading the operation state data and the characteristic image to the cloud server. The cloud server generates various source load storage regulation and control strategies according to the running state data uploaded by the edge controller, issues the regulation and control strategies to the edge controller, constructs a neural network model, obtains model parameters and judgment thresholds according to the characteristic images uploaded by the edge controller, and issues the model parameters and the judgment thresholds to the edge controller. Through the linkage calculation of the edge controller and the cloud server, the simplification and the precision of the calculation of the micro-grid regulation strategy are realized, the calculation efficiency is improved, and the micro-grid regulation strategy is prevented from falling into a local optimal solution, so that the current increasingly complex micro-grid operation real-time requirement is met.

Description

Micro-grid cloud edge cooperative control method based on deep learning
Technical Field
The invention relates to the field of micro-grid cooperative control, in particular to a micro-grid cloud edge cooperative control method based on deep learning.
Background
With the continuous improvement of the permeability of renewable energy sources, the microgrid system has a large and wide trend in construction time and region space, aiming at the control problem of the microgrid, in the aspect of construction of a control system, the existing microgrid engineering project generally adopts a two-layer control structure of an on-site unit layer and a monitoring layer, the monitoring layer is a decision layer for monitoring control and safe and economic operation of the microgrid system, and a decision is made according to operation information uploaded by each unit of a stratum and is issued to a corresponding physical element of the stratum for execution, so that the scheduling control of each power generation, energy storage and load unit in the microgrid is realized; in the aspect of establishing a micro-grid regulation and control strategy, most of the existing researches are based on physical mechanism analysis, and then an optimization regulation and control model is established, and an optimization algorithm is adopted to establish the regulation and control strategy. In the conventional microgrid control and regulation method, under the current microgrid development situation that the point is multi-faceted and wide and the regulation and control mode is increasingly complex, the problems of complex solution process redundancy, dimensionality disaster, low optimization calculation efficiency, easy falling into local optimal solution and the like exist, and the current increasingly complex microgrid operation real-time requirement is difficult to meet.
Disclosure of Invention
In order to solve the problems, the invention provides a micro-grid cloud edge cooperative control method based on deep learning. And generating a source load storage regulation and control strategy and issuing the source load storage regulation and control strategy to the edge controller at the cloud server according to the running state data uploaded by the edge controller, constructing a neural network model, obtaining model parameters and confirmation judgment thresholds of all label types according to the characteristic images uploaded by the edge controller, and issuing the model parameters and the confirmation judgment thresholds to the edge controller. Through the linkage calculation of the edge controller and the cloud server, the simplification and the precision of the calculation of the micro-grid regulation strategy are realized, the optimized calculation efficiency is improved, and the micro-grid regulation strategy is prevented from falling into a local optimal solution, so that the current increasingly complex micro-grid operation real-time requirement is met.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a micro-grid cloud edge cooperative control method based on deep learning comprises the following steps of an edge controller:
s1: collecting operation state data;
s2: generating a characteristic image according to the running state data;
s3: receiving model parameters and a tag type confirmation judgment threshold value of the cloud server, and identifying the tag type and the probability value of the tag type to which the characteristic image belongs by adopting a model;
s4: comparing the highest probability value of the probability values with a confirmation judgment threshold value of the corresponding label category;
if the highest probability value is larger than a determination threshold value, calling a source load storage regulation and control strategy corresponding to the label category issued by the cloud server, and uploading the characteristic image and the running state data to the cloud server;
if the highest probability value is smaller than the confirmation judgment threshold value, fine-tuning a source load storage regulation and control strategy corresponding to the label category to form a new source load storage regulation and control strategy, and uploading the characteristic image and the running state data to the cloud server;
s5: and decomposing and issuing the source load storage regulation and control strategy to the corresponding distributed sensing units for execution.
Further, the operation state data comprises the source load storage operation state data of the micro-grid wind power, photovoltaic, energy storage and load units.
Further, the characteristic image is a source load running state comprehensive characteristic image.
Further, the model adopts a CNN convolutional neural network model.
A micro-grid cloud edge cooperative control method based on deep learning comprises the following steps of a cloud server:
a1: constructing an optimization model, inputting running state data uploaded by the edge controller, generating a source load storage regulation and control strategy under each label category, and issuing the source load storage regulation and control strategy to the edge controller;
a2: and constructing a neural network model, inputting the characteristic image uploaded by the edge controller for training and testing, obtaining parameters of the neural network model and the confirmation judgment threshold value of each label category, and sending the parameters and the confirmation judgment threshold value to the edge controller.
Further, the constructing an optimization model and inputting the running state data uploaded by the edge controller, generating a source load storage regulation and control strategy under each label category and issuing the edge controller comprises the following steps:
a11: receiving the running state data uploaded by the edge controller and setting corresponding label types;
a12: generating corresponding typical operation data aiming at the operation state data of each label category;
a13: constructing an optimization model and inputting typical operation data to obtain a corresponding source load storage regulation strategy;
a14: and issuing the source load storage regulation and control strategy and the corresponding label category to the edge controller.
Further, the constructing a neural network model and inputting the characteristic image uploaded by the edge controller for training and testing to obtain the neural network model parameters and the determination threshold of each label category, and issuing the neural network model parameters and the determination threshold to the edge controller includes the following steps:
a21: receiving the characteristic image uploaded by the edge controller and setting a corresponding label type;
a22: dividing the characteristic image data under each label category into a training set and a test set, wherein the proportion of the training set is greater than that of the test set;
a23: constructing a neural network model, and training the neural network model by using a training set;
a24: testing the trained neural network model by using a test set;
a25: and issuing the successfully tested neural network model parameters and the confirmation judgment threshold values of all label types to the edge controller.
Further, the training of the neural network model by using the training set includes two parts, namely a forward propagation process and a backward propagation process, and the forward propagation process includes the following steps:
a231: initializing a neural network model;
a232: inputting the characteristic images of the training set into a neural network model;
a233: calculating by a neural network model to obtain the label category and the probability value thereof;
a234: calculating the deviation between the target value and the probability value, and judging whether the deviation is within an allowable range; if the weight parameter is within the allowable range, confirming and judging the threshold value by the weight parameter of the fixed neural network model and the label type; if the deviation is not within the allowable range, performing optimization calculation on the weight parameters in a back propagation process until the deviation is within the allowable range;
the back propagation process comprises the following steps:
a235: finishing training and storing the training model;
a236: solving the correction quantity of the connection weight of the input layer and the hidden layer and the correction quantity of the connection weight of the hidden layer and the output layer;
a237: the connection weights of the input layer, the hidden layer, and the output layer are updated, respectively, and step a233 is performed.
Further, the testing the trained neural network model by using the test set includes the following steps:
a241: inputting the characteristic images of the test set into a neural network model one by one to calculate the probability value of each label category;
a242: judging whether the label category corresponding to the highest probability value is the label category to which the characteristic image belongs, if not, returning to the training process to train the neural network model again, and if so, taking the highest probability value as the probability value of the label category to which the characteristic image belongs;
a243: after obtaining the probability values of the label categories of all the images in the test set, judging whether all the probability values are larger than a preset threshold value; if yes, fixing the neural network parameters and sending the parameters to the edge controller; if not, returning to the training process to train the neural network model again.
Further, the operation state data comprises source load storage operation state data of the micro-grid wind power, photovoltaic, energy storage and load units; the characteristic image is a source load storage running state comprehensive characteristic image; the neural network model adopts a CNN convolutional neural network model.
The invention provides a micro-grid cloud edge cooperative control method based on deep learning, which is characterized in that an edge controller collects operation state data to generate a characteristic image, a model parameter of a cloud server is received, a label type to which the characteristic image belongs is identified after a judgment threshold value is confirmed, a source load storage regulation and control strategy issued by the cloud server is directly called or finely adjusted according to the label type to serve as a micro-grid regulation and control strategy, and the operation state data and the characteristic image are uploaded to the cloud server. And generating a load storage regulation and control strategy of each label type source at the cloud server according to the running state data uploaded by the edge controller, issuing the load storage regulation and control strategy to the edge controller, constructing a neural network model, training and testing according to the characteristic image uploaded by the edge controller, obtaining model parameters and a confirmation judgment threshold value of each label type, and issuing the model parameters and the confirmation judgment threshold value to the edge controller. Through the linkage calculation of the edge controller and the cloud server, the simplification and the precision of the calculation of the micro-grid regulation strategy are realized, the optimized calculation efficiency is improved, and the micro-grid regulation strategy is prevented from falling into a local optimal solution, so that the current increasingly complex micro-grid operation real-time requirement is met.
Drawings
Fig. 1 is a flow chart of an edge controller of a micro-grid cloud edge cooperative control method based on deep learning;
fig. 2 is a flow chart of a cloud server of a micro-grid cloud edge cooperative control method based on deep learning;
FIG. 3 is a flow chart of CNN convolutional neural network model training and testing;
FIG. 4 is a schematic structural diagram of a CNN convolutional neural network model;
FIG. 5 is a schematic diagram of a neuron structure.
Detailed Description
The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
The embodiments of the present disclosure are described below with specific examples, and other advantages and effects of the present disclosure will be readily apparent to those skilled in the art from the disclosure in the specification. It is to be understood that the described embodiments are merely illustrative of some, and not restrictive, of the embodiments of the disclosure. The disclosure may be embodied or carried out in various other specific embodiments, and various modifications and changes may be made in the details within the description without departing from the spirit of the disclosure. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
Example one
As shown in fig. 1, an embodiment schematic diagram of a deep learning-based microgrid cloud-edge cooperative control method is described in detail in this embodiment, where an edge controller is used as a main body, and the method includes the following steps:
s1: collecting running state data of units such as wind power, photovoltaic, energy storage and load of a micro-grid;
s2: generating a comprehensive characteristic image of the micro-grid source load storage and transportation state of a new cycle of regulation and control period according to the running state data;
s3: receiving CNN convolutional neural network model parameters sent by a cloud server and a confirmation judgment threshold value corresponding to each label category; adopting a CNN convolutional neural network model to identify the label category to which the comprehensive characteristic image of the micro-grid source load storage and transportation state belongs in the regulation and control period, and calculating the probability value of belonging to each label category;
s4: comparing the highest probability value of the probability values with a confirmation judgment threshold value of the label category corresponding to the highest probability value;
if the highest probability value is larger than the judgment threshold value, the tag type is confirmed successfully, a source load storage regulation and control strategy under the tag type corresponding to the highest probability value is called in the microgrid regulation and control strategy base to serve as the microgrid regulation and control strategy of the regulation and control period, meanwhile, the corresponding microgrid source load storage and transmission state comprehensive characteristic image is transmitted to a cloud server to be stored in an image sub-base corresponding to the tag type, and corresponding source load storage operation data is stored in a source load storage operation data sub-base corresponding to the tag type of the cloud server;
if the highest probability value is smaller than the judgment threshold value, the label type is not determined successfully, then a source load storage regulation and control strategy under the label type corresponding to the highest probability value is used as a basis, fine tuning is carried out according to expert experience to form a new source load storage regulation and control strategy which is used as a microgrid regulation and control strategy of the regulation and control period, the microgrid regulation and control strategy is stored in a microgrid regulation and control strategy library, meanwhile, the fact that the microgrid operation state of the regulation and control period belongs to an operation state of a new type is indicated by the fact that classification and recognition are not successful, a cloud server is informed to add the new type label, an image sub-library corresponding to the new type label and a source load storage and control data sub-library, a corresponding comprehensive characteristic image of the microgrid source load storage operation state is stored in an image sub-library corresponding to the cloud server new type label, and corresponding source load storage operation state data are stored in a source load storage operation data sub-library corresponding to the cloud server new type label.
S5: and decomposing and issuing the micro-grid regulation strategy of the regulation and control period to a corresponding distributed sensing unit for execution.
In specific implementation, the processes are periodically executed on line by the edge controller according to a preset regulation time scale, so that the on-line decision and regulation of the operation of the microgrid are realized.
In specific implementation, each label category issued by the cloud server and the source load storage regulation strategy corresponding to the label category are stored in the microgrid regulation strategy database.
Example two
As shown in fig. 2, a schematic diagram of another embodiment of a deep learning-based microgrid cloud edge cooperative control method is described in detail in this embodiment, where a cloud server is used as a main body, and the method includes the following steps:
a1: constructing a microgrid optimization model, inputting source load storage and transportation row data under each label type of the microgrid, and solving by adopting a genetic algorithm to generate a source load storage and transportation regulation strategy under each label type of the microgrid;
a2: and constructing a CNN (convolutional neural network) and inputting the CNN into a comprehensive characteristic image of the source load storage operation state under each label type of the microgrid for training and testing to obtain CNN convolutional neural network model parameters and the confirmation judgment threshold of each label type.
In specific implementation, as shown in fig. 2, the constructing of the microgrid optimization model and the inputting of the source load storage and transportation data of the microgrid under each tag category, and the solving by using the genetic algorithm to generate the source load storage and transportation regulation and control strategy under each tag category of the microgrid includes the following steps:
a11: establishing a source load storage and transportation data sub-database to store source load storage and transportation data reflecting the operation state of the micro-grid uploaded by the edge controller, and setting corresponding label types; each micro-grid running state is provided with a corresponding source load storage running data sub-database for storing corresponding source load storage running data and a corresponding label category;
a12: generating corresponding typical operation data of the micro-grid under each label category by adopting a weighted average method aiming at the source load storage and transportation row data under each label category;
a13: constructing a micro-grid operation optimization mathematical model, inputting typical micro-grid operation data under each label category into the micro-grid operation optimization mathematical model, and solving the micro-grid operation optimization mathematical model by adopting classical optimization algorithms such as a genetic algorithm and the like to obtain an optimal source load storage regulation strategy of the micro-grid operation state under the corresponding label category;
a14: and storing the optimal source load storage regulation and control strategy into a micro-grid strategy sub-library, setting corresponding label categories, and then sending the label categories to an edge controller.
In a specific implementation, as shown in fig. 2, the constructing a CNN convolutional neural network and inputting the source load storage operation state comprehensive characteristic image of each tag type of the microgrid for training and testing to obtain a CNN convolutional neural network model parameter and a determination threshold of each tag type includes the following steps:
a21: establishing a source load running state comprehensive characteristic image sub-library to store the source load running state comprehensive characteristic images uploaded by the edge controller, and setting corresponding label types; each micro-grid running state has a corresponding source load running state comprehensive characteristic image sub-library for storing a corresponding source load running state comprehensive characteristic image and has a corresponding label category;
a22: randomly dividing image data in a source load storage running state comprehensive characteristic image sub-library under each label type into a training set and a testing set, wherein the proportion of the training set is greater than that of the testing set;
a23: constructing a CNN convolutional neural network model, and training the CNN convolutional neural network model by using a training set;
a24: testing the trained CNN convolutional neural network model by using a test set;
a25: and issuing the CNN convolutional neural network model parameters which are successfully tested and the confirmation judgment threshold values of all label types to the edge controller.
In a specific implementation, as shown in fig. 3, the training of the CNN convolutional neural network model by using the training set includes a forward propagation process and a backward propagation process, where the forward propagation process includes the following steps:
a231: initializing a CNN convolutional neural network model;
a232: inputting the source load storage traffic state comprehensive characteristic image of the training set into a CNN convolution neural network model;
a233: calculating to obtain the most similar label category and probability value by using the CNN convolutional neural network model;
a234: calculating a target value and a probability value to obtain a deviation e, and judging whether e is in an allowable range; if the deviation e is within the allowable range, the CNN convolutional neural network model meets the requirements, the training is ended, and the weight parameters and the determination threshold are fixed; if the deviation e is not in the allowable range, the CNN convolutional neural network model does not meet the requirement, and then a back propagation process is carried out to carry out optimization calculation on the weight parameters until the actual output deviation e is in the allowable range;
the back propagation process comprises the following steps:
a235: finishing training and storing the training model;
a236: using a BP (Back propagation) algorithm to obtain the weight correction quantity of the input layer and the hidden layer and the weight correction quantity of the hidden layer and the output layer;
a237: the connection weights of the input layer, the hidden layer, and the output layer are updated, respectively, and step a233 is performed.
In a specific implementation, the step a236 of obtaining the correction amount of the weight of the input layer and the hidden layer and the correction amount of the weight of the hidden layer and the output layer by using a bp (back propagation) algorithm includes the following steps:
output A from the input layer to the hidden layerjComprises the following steps:
Figure BDA0003247602640000091
in the formula, WihAs weights of the input layer and the hidden layer, WihValue rangeThe value of the weight is (-1,1), and the weight is assigned in a random number mode; bihIs a bias constant of the input layer and the hidden layer, AiFor the input source load running state comprehensive characteristic image training set, k is AiThe number of (c), (d) are activation functions.
The output of the activation function is:
Hi＝f(Ai)
hidden layer output to output layer output HjComprises the following steps:
Figure BDA0003247602640000092
in the formula, WhoAs weights of the hidden layer and the output layer, bhoIs a bias constant of the hidden layer and the output layer, WhoThe values are (-1, 1).
Output Y via activation functionjComprises the following steps:
Yj＝f(Hj)
the error function E is:
Figure BDA0003247602640000093
in the formula, DjIs the desired output.
The correction amount of each input weight is:
Figure BDA0003247602640000094
wherein l is the learning rate, Δ WihIs a correction amount of the weight of the input layer and the hidden layer, Δ WhoIs the correction of the weights of the hidden layer and the output layer.
In a specific implementation, the formula for updating the connection weight in step a237 is as follows:
Figure BDA0003247602640000101
in the formula (I), the compound is shown in the specification,
Figure BDA0003247602640000102
and
Figure BDA0003247602640000103
the weights of the input layer and the hidden layer and the weights of the hidden layer and the output layer are respectively obtained after the Nth learning;
Figure BDA0003247602640000104
and
Figure BDA0003247602640000105
the weights of the input layer and the hidden layer and the weights of the hidden layer and the output layer after the (N + 1) th learning are respectively.
In a specific implementation, the CNN convolutional neural network model shown in fig. 4 includes an input layer, a hidden layer, and an output layer, where the hidden layer is connected to the input layer and the output layer, respectively, and the input layer, the hidden layer, and the output layer are all composed of a plurality of neurons shown in fig. 5 having learnable weights and paranoia variables.
In a specific implementation, the neurons are described by mathematical expressions as follows:
t＝f(w1×a1+w2×a2+...wn×an+b)
t is the output, where a1～anRepresenting the input vector, b representing the bias constant, w1～wnRepresenting the weight, f (x) is an activation function; when inputting a1～anWhen vectors are applied to the neuron, each input vector has a different weight wnThe neuron outputs a result t through an activation function under the condition that inputs of different weights and bias constants are considered.
In specific implementation, the input layer is used for receiving the source load running state comprehensive characteristic image. Usually, the computer divides the color depth into 256 levels, and in the computer, the nature of the image is actually an array formed by color information of each pixel point. The gray-scale image has only light and shade information for each pixel, and the color image has three R/G/B color channels for each pixel. When the source load running state comprehensive characteristic image is input, data with a certain area size, namely a corresponding three-dimensional matrix, is actually input into a computer, and the length and the width of the matrix represent the size of the image.
In specific implementation, the hidden layer comprises a convolution layer, a pooling layer, a full-link layer and a Softmax regression layer, the features of the source load and transport line state comprehensive characteristic image are extracted mainly through alternation of a multi-round convolution-pooling structure to obtain a feature matrix, then the extracted feature matrix is expanded into a one-dimensional matrix through the full-link layer, and finally the corresponding most similar label category and probability value are output by the Sfotmax regression layer.
In a specific implementation, the output layer outputs the probability value of the most similar label category obtained by the Softmax regression layer.
In one implementation, the convolutional layer is the most important layer in the convolutional neural network. When the source load running state comprehensive characteristic curve image pixel data is converted into a matrix through the input layer, the characteristics of the image data are mainly extracted through a convolution kernel in the convolution layer in a local connection and weight sharing mode. Describing the convolution process with a mathematical expression is:
Figure BDA0003247602640000111
in the formula, G represents a comprehensive characteristic image matrix of the storage and transportation line state of the source charge of the micro-grid after feature extraction, m and n are rows and columns, and f represents an input comprehensive characteristic image matrix of the storage and transportation line state of the source charge of the micro-grid; w represents a convolution kernel; s and t define the size of the convolution kernel.
Through convolution operation, the size of the source load storage running state comprehensive characteristic image is reduced, but the original image characteristics are still kept. In addition, convolution layers involve parameters such as convolution step size and edge fill. The convolution step length refers to the step length of each movement of a convolution kernel when the convolution kernel is used for carrying out convolution on image elements, the step length is different, the output feature graph is different, the step length is smaller, the extracted features are thinner, the features are richer, the step length is larger, the extracted features are thicker, and the features are simpler; and the edge filling is to add a certain number of rows and columns on each side of the input feature map, so that the sizes of the output feature map and the input feature map are the same, and the utilization rate of the edge information is improved to a certain extent. The calculation formula of the convolution result is:
Figure BDA0003247602640000112
Figure BDA0003247602640000113
in the formula, H1、W1Indicates the length, width, H, of the input2、W2Length, width, f, of the output feature mapu、fwDenotes the size of the length and width of the convolution kernel, s denotes the step size of the sliding window, and p denotes the boundary fill.
In specific implementation, the dimension reduction and feature extraction of the source load running state comprehensive characteristic image are completed through the convolution layer, but the dimension of the feature image is still high at the moment, and too high dimension not only increases the calculation amount, but also easily causes overfitting to reduce the accuracy of actual prediction. Therefore, the features extracted by the convolutional layer are divided into different small areas by the pooling layer, and then the convolutional features after pooling and dimension reduction are represented by the maximum value or average value features of the areas, so that the function of secondarily extracting the features is achieved. Therefore, the difference of the image positions is reduced, the quantity and the calculated amount of parameters are greatly and efficiently reduced, and the overfitting problem is controlled to a certain degree.
In a specific implementation, the pooling layer comprises maximum pooling and mean pooling; the maximum pooling is that the maximum value is taken for the feature points in the region, so that the texture extraction is better; the mean pooling means that the feature points in the region are averaged, so that the background is better reserved; according to the requirements, the invention selects the maximum pooling.
In specific implementation, the full connection layer is abbreviated as an FC layer, and neurons in the full connection layer have a connection relation with neurons of adjacent layers before and after the CNN convolutional neural network, so that useful information of the integrated upper layer can be extracted and transmitted to the next connection layer. The main purpose of the fully-connected layer is dimension transformation, wherein a convolution layer is extracted to a high-dimensional distributed feature representation to become a low-dimensional sample mark through a convolution kernel with the size of 1 x 1, and a classification evaluation value is calculated by the features extracted by the convolution and pooling layers. In order to improve the performance of the CNN convolutional neural network model, a ReLU (rectified Linear Unit) function is adopted as an excitation function of each neuron of the full connection layer. ReLU is called a linear rectification function in full, and the mathematical expression of the ReLU is as follows:
f(x)＝max(0,x)
redundant data can be removed through the ReLU function, data characteristics are reserved and mapped out, the influence of other complex activation functions is eliminated, and the calculation process is simplified.
In a specific implementation, the Softmax regression layer is also called a logistic regression analysis, and the Softmax regression algorithm can be used for multi-classification. The principle of the Softmax algorithm is as follows: carrying out weighted average on pixels of the source load storage running state comprehensive characteristic image training data containing the label categories to obtain weights C of the label categories on the pixel points; if C>0, then belong to the tag class; if C<0, it does not belong to the tag class. And the Softmax regression layer outputs the previous FC layer to perform linear regression, and the probability of each label category is obtained by performing regularization processing on the classification evaluation value obtained by the FC layer through specific data calculation. If a given source load running state comprehensive characteristic image training set { (x)(i),y(i)；i∈1,...,N,y(i)E.g., 0, K-1, where x(i)Is the ith input image block; y is(i)The calculation formula for its class label is:
Figure BDA0003247602640000131
in the formula (I), the compound is shown in the specification,
Figure BDA0003247602640000132
indicating that the ith input belongs to class jOutputting the probability;
Figure BDA0003247602640000133
indicating the output probability that the ith input belongs to the jth class.
In a specific implementation, the determination threshold is a minimum value of probability values obtained by testing a test set under a certain label category through a CNN convolutional neural network model, and is used as a determination threshold for online identification and determination of the label category.
In specific implementation, the CNN convolutional neural network model needs to be trained and tested at regular time every day, so that the corresponding micro-grid operation state is identified more accurately along with the increase of the micro-grid operation data volume and the update of the micro-grid regulation and control strategy library, the formulated regulation and control strategy is more accurate, and continuous deep learning is realized so that the control is more perfect.
In specific implementation, when the error function E is small enough or the number of learning times reaches the set upper limit N, the model stops training, and stores parameters for identifying classification, such as weights.
In a specific implementation, as shown in fig. 3, the step a24 of testing the trained CNN convolutional neural network model by using the test set includes the following steps:
a241: inputting images in the source load storage traffic state comprehensive characteristic image test set one by one, and calculating the probability corresponding to each label category by calling a trained CNN convolutional neural network model;
a242: judging whether the label type corresponding to the highest probability value is the label type to which the label type belongs, if the label type corresponding to the highest probability value is different from the label type to which the label type belongs, indicating that the test of the CNN convolutional neural network model fails, and returning to the training process to train the CNN convolutional neural network model again; if the label category corresponding to the highest probability value is the same as the label category to which the label category belongs, outputting the probability value of the label category corresponding to the label category;
a243: after all the images in the test set are tested and calculated to obtain the corresponding label class probability through the steps, judging whether the label class probability values corresponding to all the images in the test set are larger than a preset threshold value or not, if the label class probability values corresponding to all the images in the test set are judged to be successful, the model test is successful, the CNN convolutional neural network model is successfully trained, the CNN convolutional neural network parameter is fixed to form a model, and the model is issued to an edge controller; if the label category probability corresponding to any image in the test set is smaller than the preset threshold value, the CNN convolutional neural network model is failed to train, and the training process needs to be returned to train the convolutional neural network model again.
The invention provides a micro-grid cloud edge cooperative control method based on deep learning, which comprises the steps of collecting operation state data at an edge controller, generating a characteristic image according to the operation state data, identifying a label type to which the characteristic image belongs after receiving model parameters and a label type confirmation judgment threshold value of a cloud server, directly calling or finely adjusting a source load storage regulation and control strategy issued by the cloud server according to the label type to serve as a micro-grid regulation and control strategy, and uploading the operation state data and the characteristic image to the cloud server. And generating a load storage regulation and control strategy of each label type source at the cloud server according to the running state data uploaded by the edge controller, issuing the load storage regulation and control strategy to the edge controller, constructing a neural network model, training and testing according to the characteristic image uploaded by the edge controller, obtaining model parameters and a confirmation judgment threshold value of each label type, and issuing the model parameters and the confirmation judgment threshold value to the edge controller. Through respectively calculating and linking at the edge controller and the cloud server, the calculation of the micro-grid regulation and control strategy is simplified and accurate, the calculation optimization efficiency is improved, and the situation that the micro-grid regulation and control strategy is trapped in a local optimal solution is prevented, so that the current increasingly complex micro-grid operation real-time requirement is met.
The above description is for the purpose of illustrating embodiments of the invention and is not intended to limit the invention, and it will be apparent to those skilled in the art that any modification, equivalent replacement, or improvement made without departing from the spirit and principle of the invention shall fall within the protection scope of the invention.

Claims (10)

1. A micro-grid cloud edge cooperative control method based on deep learning is characterized by comprising the following steps of an edge controller:
s1: collecting operation state data;
s2: generating a characteristic image according to the running state data;
s3: receiving a model parameter and a tag type confirmation judgment threshold value of a cloud server, and identifying the tag type and the probability value of the tag type to which the characteristic image belongs by adopting a model;
s4: comparing the highest probability value of the probability values with a confirmation judgment threshold value of the corresponding label category;
if the highest probability value is larger than a determination threshold value, calling a source load storage regulation and control strategy corresponding to the label category issued by the cloud server, and uploading the characteristic image and the running state data to the cloud server;
if the highest probability value is smaller than the confirmation judgment threshold value, fine-tuning a source load storage regulation and control strategy corresponding to the label category to form a new source load storage regulation and control strategy, and uploading the characteristic image and the running state data to the cloud server;
s5: and decomposing and issuing the source load storage regulation and control strategy to the corresponding distributed sensing units for execution.
2. The deep learning-based microgrid cloud-edge cooperative control method according to claim 1, characterized in that the operating state data comprises microgrid wind power, photovoltaic, energy storage and load unit source-to-load operating state data.
3. The microgrid cloud-edge cooperative control method based on deep learning of claim 1 is characterized in that the characteristic image is a source load running state comprehensive characteristic image.
4. The microgrid cloud-edge cooperative control method based on deep learning of any one of claims 1 to 3, characterized in that the model adopts a CNN convolutional neural network model.
5. A micro-grid cloud edge cooperative control method based on deep learning is characterized by comprising the following steps of a cloud server:
a1: constructing an optimization model, inputting running state data uploaded by the edge controller, generating a source load storage regulation and control strategy under each label category, and issuing the source load storage regulation and control strategy to the edge controller;
a2: and constructing a neural network model, inputting the characteristic image uploaded by the edge controller for training and testing, obtaining parameters of the neural network model and the confirmation judgment threshold value of each label category, and sending the parameters and the confirmation judgment threshold value to the edge controller.
6. The microgrid cloud-edge cooperative control method based on deep learning of claim 5, wherein the method for constructing an optimization model, inputting running state data uploaded by the edge-end controller, generating a source load storage regulation and control strategy under each label category and issuing the edge-end controller comprises the following steps:
a11: receiving the running state data uploaded by the edge controller and setting corresponding label types;
a12: generating corresponding typical operation data aiming at the operation state data of each label category;
a13: constructing an optimization model and inputting typical operation data to obtain a corresponding source load storage regulation strategy;
a14: and issuing the source load storage regulation and control strategy and the corresponding label category to the edge controller.
7. The microgrid cloud-edge cooperative control method based on deep learning of claim 5, wherein the method for constructing the neural network model and inputting the characteristic image uploaded by the edge controller for training and testing to obtain the neural network model parameters and the confirmation judgment threshold values of the label categories and sending the confirmation judgment threshold values to the edge controller comprises the following steps:
a21: receiving the characteristic image uploaded by the edge controller and setting a corresponding label type;
a22: dividing the characteristic image data under each label category into a training set and a test set, wherein the proportion of the training set is greater than that of the test set;
a23: constructing a neural network model, and training the neural network model by using a training set;
a24: testing the trained neural network model by using a test set;
a25: and issuing the successfully tested neural network model parameters and the confirmation judgment threshold values of all label types to the edge controller.
8. The microgrid cloud-edge cooperative control method based on deep learning of claim 7, wherein the training of the neural network model by using the training set includes two parts of a forward propagation process and a backward propagation process, and the forward propagation process includes the following steps:
a231: initializing a neural network model;
a232: inputting the characteristic images of the training set into a neural network model;
a233: calculating by a neural network model to obtain the label category and the probability value thereof;
a234: calculating the deviation between the target value and the probability value, and judging whether the deviation is within an allowable range; if the weight parameter is within the allowable range, confirming and judging the threshold value by the weight parameter of the fixed neural network model and the label type; if the deviation is not within the allowable range, performing optimization calculation on the weight parameters in a back propagation process until the deviation is within the allowable range;
the back propagation process comprises the following steps:
a235: finishing training and storing the training model;
a236: solving the correction quantity of the connection weight of the input layer and the hidden layer and the correction quantity of the connection weight of the hidden layer and the output layer;
a237: the connection weights of the input layer, the hidden layer, and the output layer are updated, respectively, and step a233 is performed.
9. The microgrid cloud-edge cooperative control method based on deep learning of claim 7, wherein the testing of the trained neural network model by using the test set comprises the following steps:
a241: inputting the characteristic images of the test set into a neural network model one by one to calculate the probability value of each label category;
a242: judging whether the label category corresponding to the highest probability value is the label category to which the characteristic image belongs, if not, returning to the training process to train the neural network model again, and if so, taking the highest probability value as the probability value of the label category to which the characteristic image belongs;
a243: after obtaining the probability values of the label categories of all the images in the test set, judging whether all the probability values are larger than a preset threshold value; if yes, fixing the neural network parameters and sending the parameters to the edge controller; if not, returning to the training process to train the neural network model again.
10. The microgrid cloud-edge cooperative control method based on deep learning of any one of claims 5 to 9, wherein the operating state data comprises source-to-charge operating state data of microgrid wind power, photovoltaic, energy storage and load units; the characteristic image is a source load storage running state comprehensive characteristic image; the neural network model adopts a CNN convolutional neural network model.
CN202111036998.7A 2021-09-06 2021-09-06 Micro-grid cloud edge cooperative control method based on deep learning Pending CN113872322A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111036998.7A CN113872322A (en) 2021-09-06 2021-09-06 Micro-grid cloud edge cooperative control method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111036998.7A CN113872322A (en) 2021-09-06 2021-09-06 Micro-grid cloud edge cooperative control method based on deep learning

Publications (1)

Publication Number Publication Date
CN113872322A true CN113872322A (en) 2021-12-31

Family

ID=78989536

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111036998.7A Pending CN113872322A (en) 2021-09-06 2021-09-06 Micro-grid cloud edge cooperative control method based on deep learning

Country Status (1)

Country Link
CN (1) CN113872322A (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108053084A (en) * 2018-01-18 2018-05-18 中南大学 A kind of short-term micro-grid load Forecasting Methodology based on ABC algorithms
CN109347149A (en) * 2018-09-20 2019-02-15 国网河南省电力公司电力科学研究院 Micro-capacitance sensor energy storage dispatching method and device based on depth Q value network intensified learning
CN110649641A (en) * 2019-09-29 2020-01-03 国网山东省电力公司德州供电公司 Electric automobile quick charging station energy storage system and method based on source network charge storage cooperative service
KR20200109196A (en) * 2019-03-12 2020-09-22 한국전력공사 Apparatus for estimating power supply of microgrid and method thereof
CN112287990A (en) * 2020-10-23 2021-01-29 杭州卷积云科技有限公司 Model optimization method of edge cloud collaborative support vector machine based on online learning
CN112700057A (en) * 2021-01-07 2021-04-23 福州大学 Power load prediction method based on modular recurrent neural network

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108053084A (en) * 2018-01-18 2018-05-18 中南大学 A kind of short-term micro-grid load Forecasting Methodology based on ABC algorithms
CN109347149A (en) * 2018-09-20 2019-02-15 国网河南省电力公司电力科学研究院 Micro-capacitance sensor energy storage dispatching method and device based on depth Q value network intensified learning
KR20200109196A (en) * 2019-03-12 2020-09-22 한국전력공사 Apparatus for estimating power supply of microgrid and method thereof
CN110649641A (en) * 2019-09-29 2020-01-03 国网山东省电力公司德州供电公司 Electric automobile quick charging station energy storage system and method based on source network charge storage cooperative service
CN112287990A (en) * 2020-10-23 2021-01-29 杭州卷积云科技有限公司 Model optimization method of edge cloud collaborative support vector machine based on online learning
CN112700057A (en) * 2021-01-07 2021-04-23 福州大学 Power load prediction method based on modular recurrent neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
丁学辉等: "基于CNN特征选择与QRGRU的电力负荷概率密度预测方法", 《电力信息与通信技术》, vol. 19, no. 6, 30 June 2021 (2021-06-30) *
吕磊: "基于深度学习的储能装置容量预测模型及其优化控制研究", 《中国优秀硕士学位论文全文数据库 工程科技辑II辑》, no. 4, 30 April 2021 (2021-04-30) *

Similar Documents

Publication Publication Date Title
US5796611A (en) Weather forecast apparatus and method based on recognition of echo patterns of radar images
CN110348624B (en) Sand storm grade prediction method based on Stacking integration strategy
CN106228185A (en) A kind of general image classifying and identifying system based on neutral net and method
CN110674965A (en) Multi-time step wind power prediction method based on dynamic feature selection
CN117498400B (en) Distributed photovoltaic and energy storage data processing method and system
Wang et al. A regional pretraining-classification-selection forecasting system for wind power point forecasting and interval forecasting
CN109558898B (en) Multi-choice learning method with high confidence based on deep neural network
CN114492191A (en) Heat station equipment residual life evaluation method based on DBN-SVR
CN112381282A (en) Photovoltaic power generation power prediction method based on width learning system
CN114897129A (en) Photovoltaic power station short-term power prediction method based on similar daily clustering and Kmeans-GRA-LSTM
CN114463540A (en) Segmenting images using neural networks
CN114564651A (en) Self-supervision recommendation method combined with contrast learning method
CN115912502A (en) Intelligent power grid operation optimization method and device
CN115345297A (en) Platform area sample generation method and system based on generation countermeasure network
CN111275074A (en) Power CPS information attack identification method based on stack type self-coding network model
CN114548350A (en) Power load prediction method based on goblet sea squirt group and BP neural network
CN113033898A (en) Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network
CN117096871A (en) Wind power probability density prediction method based on space-time distribution
CN113872322A (en) Micro-grid cloud edge cooperative control method based on deep learning
CN116245259A (en) Photovoltaic power generation prediction method and device based on depth feature selection and electronic equipment
CN113191984B (en) Deep learning-based motion blurred image joint restoration and classification method and system
CN113807005A (en) Bearing residual life prediction method based on improved FPA-DBN
CN111563413A (en) Mixed dual-model-based age prediction method
CN113297791B (en) Wind power combination prediction method based on improved DBN
Wang et al. Forecasting Short-Term Solar PV Using Hierarchical Clustering and Cascade Model

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