CN114154558A - Distributed energy power generation load prediction system and method based on graph neural network - Google Patents
Distributed energy power generation load prediction system and method based on graph neural network Download PDFInfo
- Publication number
- CN114154558A CN114154558A CN202111336833.1A CN202111336833A CN114154558A CN 114154558 A CN114154558 A CN 114154558A CN 202111336833 A CN202111336833 A CN 202111336833A CN 114154558 A CN114154558 A CN 114154558A
- Authority
- CN
- China
- Prior art keywords
- power
- region
- time
- graph
- clustering
- 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.)
- Granted
Links
- 238000010248 power generation Methods 0.000 title claims abstract description 80
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 28
- 239000013598 vector Substances 0.000 claims abstract description 141
- 238000000605 extraction Methods 0.000 claims abstract description 30
- 238000009826 distribution Methods 0.000 claims abstract description 17
- 238000003860 storage Methods 0.000 claims abstract description 13
- 230000005540 biological transmission Effects 0.000 claims abstract description 12
- 238000005516 engineering process Methods 0.000 claims abstract description 7
- 238000004146 energy storage Methods 0.000 claims description 28
- 238000000547 structure data Methods 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 230000008901 benefit Effects 0.000 claims description 6
- 230000005611 electricity Effects 0.000 claims description 6
- 238000004519 manufacturing process Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000004220 aggregation Methods 0.000 claims description 4
- 230000002776 aggregation Effects 0.000 claims description 4
- 239000002803 fossil fuel Substances 0.000 claims description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 3
- 238000013499 data model Methods 0.000 claims 2
- 238000010801 machine learning Methods 0.000 abstract description 6
- 238000013135 deep learning Methods 0.000 abstract description 5
- 238000011161 development Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Water Supply & Treatment (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Public Health (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a distributed energy power generation load prediction system and a method based on a graph neural network, belonging to the technical field of graph neural networks, machine learning and digital energy, aiming at solving the technical problem of accurately predicting the power generation and consumption of the power storage of a power grid source network by effectively utilizing machine learning and deep learning technologies such as clustering, the graph neural network and the like and combining with the transmission and distribution service of a power system, and adopting the technical scheme that: the system comprises an electric power system participation unit feature extraction encoder, an electric power region division clustering device, a region vector generator, a selection device for selecting the optimal region division and a future power generation load predictor. The method comprises the following specific steps: s1, participating in power system unit feature extraction and clustering; s2, obtaining the optimal area division; and S3, predicting the future power generation load.
Description
Technical Field
The invention relates to the technical field of a graph neural network, machine learning and digital energy, in particular to a distributed energy power generation load prediction system and method based on the graph neural network.
Background
With the development of machine learning and deep learning technologies and the support of mass data and high-efficiency computing power in the times of internet and cloud computing, a large-scale neural network similar to a human brain structure is obtained through training and construction, breakthrough progress is made in the fields of computer vision, voice recognition, natural language understanding and the like, and subversive changes are being brought to the whole society.
Graph Neural Network (GNN) is a method that can perform deep learning on Graph data, and includes models applied to graphs by various Neural networks. The graph is a graph formed by a plurality of nodes and edges (edges) connecting the two nodes, and is used for depicting the relationship between different nodes. The graph data is a kind of non-european space data, and is gradually receiving attention due to its ubiquitous nature. Graph Convolutional neural Network (GCN) is a type of neural Network using Graph convolution, and has shown advantages in the field of data analysis as an important branch in Graph neural networks.
The new energy sources play a role in improving energy structures and saving energy and reducing emission, and the energy system gradually moves from a centralized mode to a distributed mode. Distributed energy is an important carrier and a propulsion means of an energy revolution and is an important component of a future energy system, distributed photovoltaic, distributed wind power, biological power generation, natural gas distributed energy, energy internet, micro-grid and multi-energy complementary projects are developed in batches, the energy industry is confronted with new development opportunities and simultaneously brings new challenges to the safe operation of the power grid, and how to effectively utilize machine learning and deep learning technologies such as clustering, graph neural network and the like and combine with the power system for transmission and distribution, and accurate prediction of power grid source grid load storage power production and consumption becomes an urgent problem to be solved.
Disclosure of Invention
The invention aims to provide a distributed energy power generation load prediction system and method based on a graph neural network, and solves the problem of how to effectively utilize machine learning and deep learning technologies such as clustering, the graph neural network and the like and combine with power system generation, transmission and distribution services to accurately predict the power generation and consumption of the power grid source network and the stored power.
The technical task of the invention is realized in the following way, a distributed energy generation load prediction system based on a graph neural network comprises,
the system comprises an electric power system participation unit feature extraction Encoder (Encoder) and a time T participation electric power system unit feature extraction method, wherein the electric power system participation unit feature extraction Encoder (Encoder) is used for extracting features according to the generated energy and electricity consumption sequence of the participation electric power system unit based on a self-attention network structure, and forming a feature vector of the time T participation electric power system unit by combining the geographical position, standby electricity consumption, output proportion, energy storage capacity and grid-connected capacity features of the participation electric power system unit at the time T;
the power region dividing and clustering device (Cluster) is used for clustering based on the characteristic vectors of all the participating power system units according to the position regions and in combination with the characteristics of source network load storage to form a clustering region category;
the region vector generator (Gen-Vg) is used for calculating the generated energy, the power consumption, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units in the clustered regions divided after clustering in all time periods to form a region vector group;
selecting an optimal region division selector (SelDist) for traversing all region division modes, calculating comprehensive evaluation points of each division mode, calculating the advantages and disadvantages of an evaluation region vector group based on the region vector group, and selecting an optimal region clustering result;
and a future generation load predictor (Pred) for forming the region vector group into a graph structure and predicting and outputting the next time region data by utilizing graph convolution operation.
Preferably, the future generation load predictor (Pred) includes,
a generation base map constructor (BG-Gen) for generating an inter-region relationship map structure based on the optimal region clustering result;
an update graph constructor (updG) for updating the inter-region relationship graph based on the data of the region vector group time T;
a graph convolution operator (GCN) for performing graph convolution operation on graph structure data formed by the region vector group by using a Chebyshev convolution algorithm;
and a predicted next time zone data modeler (Wadd) for predicting data of the electric quantity and the electric power consumption of the next time zone by inputting a result of the graph convolution operation (GCN) module and a zone vector group of the current time T in combination with the historical time input zone vector group data based on the LSTM sequence modeling structure.
More preferably, the working process of the future generation load predictor (Pred) is as follows:
(1) generating an inter-region relation graph structure by using a generation base graph constructor (BG-Gen) according to the dividing mode of the optimal region clustering result;
(2) adopting an optimal region clustering result division mode, calculating all historical data of the division mode through a region vector generator (Gen-Vg) to form a vector group sequence from time 1 to time T, and continuously inputting the vector group sequence to a future power generation load prediction model device (Pred);
(3) generating an inter-region relational graph structure based on a generated basic graph constructor (BG-Gen), updating inter-region relational graph structure data by using an updating graph constructor (updG), calculating an intermediate result through a graph convolution operator (GCN) and a next time prediction region data modeler (Wadd), and finally outputting a region vector group prediction result of time T + 1;
(4) and continuously inputting the future time prediction result into a future power generation load prediction model device (Pred) and outputting the prediction result of the region vector group of the next time.
Preferably, the feature vector of the participating power system unit is divided into two channels of power generation amount and power consumption amount, the T time period is a day unit, and the components of the feature vector record the power generation amount and the power consumption amount according to 15 minutes or 30 minutes.
Preferably, the region vector group acquisition process is specifically as follows:
(1) assembling a characteristic vector vm from time 1 to current time T for each participating power system unit m at each time point T on the basis of historical data of all participating power system units to form a power generation and power consumption dual-channel vector sequence (vm1, vm2.. vmt);
(2) processing the vector sequence (vm1, vm2.. vmt) by utilizing a power system participation unit feature extraction Encoder (Encoder) to form an encoding feature vector Vsmt;
(3) extracting other characteristic vectors Vsmt 'of the geographical position, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of the power system participating unit m at the time point t by using a power system participating unit characteristic extraction Encoder (Encoder), and combining the other characteristic vectors Vsmt' and the encoding characteristic vector Vsmt to form Vptm;
(4) clustering Vpt1, Vpt2.. Vptm by using a power region division clustering device (Cluster) to form region division clusters for each time point T from time 1 to current time T;
(5) merging the same clustering result modes in the T types of region division modes, and calculating the generated energy, the power consumption, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units in time T according to the divided regions 1 to N for each region division mode to obtain N vectors to form a vector group; from time 1 to time T, T vector groups are finally formed.
Preferably, the participating power system unit comprises a power generation source, a power transmission and distribution network, a power load and an energy storage;
the power generation unit comprises a distributed power supply for wind power generation, photovoltaic power generation and water conservancy power generation, fossil fuel power generation of traditional thermal power generation and nuclear power;
the power transmission and distribution network is used for ensuring the transmission of electric power; the power generation power supply, the power load and the energy storage are connected into a main power grid or independently form a power micro-grid system;
the electric load means consumption of electric power including industrial load, residential load and commercial load;
the stored energy is used to enable storage and output of electrical energy sources, participating in the power system as loads and sources.
A distributed energy power generation load prediction method based on a graph neural network is characterized in that a clustering technology is utilized to carry out preliminary aggregation by combining the distribution situation and the service situation of source network charge storage of power system units, and according to the connection between the aggregated power system units, the graph neural network is used for analyzing to predict the power production and consumption of a power grid; the method comprises the following specific steps:
s1, participating in power system unit feature extraction and clustering: according to the characteristics of distributed energy power generation and consumption units, based on geographical distribution, power grid laying and service conditions, selecting and extracting the characteristics of distributed power sources, loads and energy storage participating power system units through an electric power system participating unit characteristic extraction Encoder (Encoder), and reasonably clustering by using a power region division clustering device (Cluster);
s2, obtaining the optimal region division: traversing all the area division modes by using a SelDist (SelDist) selector, calculating the comprehensive evaluation score of each division mode, finding out the optimal power area division unit, and selecting the optimal area clustering result;
s3, predicting future power generation load: each power region is divided into units to form a graph structure, and the future generation load prediction modeler (Pred) predicts the power generation and power consumption of the power region units at the future time by using a graph neural network in combination with historical data and power planning data.
Preferably, the future generation load predictor (Pred) includes,
a generation base map constructor (BG-Gen) for generating an inter-region relationship map structure based on the optimal region clustering result;
an update graph constructor (updG) for updating the inter-region relationship graph based on the data of the region vector group time T;
a graph convolution operator (GCN) for performing graph convolution operation on graph structure data formed by the region vector group by using a Chebyshev convolution algorithm;
and a predicted next time zone data modeler (Wadd) for predicting data of the electric quantity and the electric power consumption of the next time zone by inputting a result of the graph convolution operation (GCN) module and a zone vector group of the current time T in combination with the historical time input zone vector group data based on the LSTM sequence modeling structure.
Preferably, the participating power system unit feature extraction and clustering in step S1 are specifically as follows:
s101, assembling a characteristic vector vm from time 1 to current time T for each participated power system unit m at each time point T based on historical data of all participated power system units to form a power generation and power consumption dual-channel vector sequence (vm1, vm2.. vmt);
s102, processing the vector sequence (vm1, vm2.. vmt) by using a power system participation unit feature extraction Encoder (Encoder) to form an encoding feature vector Vsmt;
s103, extracting other feature vectors Vsmt 'of the geographical position, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of the power system participating unit m at the time point t by using a power system participating unit feature extraction Encoder (Encoder), and combining the other feature vectors Vsmt' and the encoding feature vector Vsmt to form Vptm;
s104, clustering the Vpt1 and the Vpt2.. Vptm by using an electric power region division clustering device (Cluster) to form a region division Cluster for each time point T from the time 1 to the current time T;
s105, merging the same clustering result modes in the T types of region division modes, and calculating the generated energy, the electric power consumption, the standby electric power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units at the time T according to the divided regions 1 to N for each region division mode to obtain N vectors to form a vector group; from time 1 to time T, T vector groups are finally formed.
More preferably, the predicted future power generation load in step S3 is specifically as follows:
s301, generating an inter-region relation graph structure by using a generation base graph constructor (BG-Gen) according to the dividing mode of the optimal region clustering result;
s302, adopting an optimal region clustering result division mode, calculating all historical data of the division mode through a region vector generator (Gen-Vg) to form a vector group sequence from time 1 to time T, and continuously inputting the vector group sequence to a future power generation load prediction model device (Pred);
s303, generating an inter-region relational graph structure based on a generated basic graph constructor (BG-Gen), updating inter-region relational graph structure data by using an updating graph constructor (updG), calculating an intermediate result through a graph convolution operator (GCN) and a next time prediction region data modeler (Wadd), and finally outputting a region vector group prediction result of time T + 1;
s304, continuously inputting the future time prediction result into a future power generation load prediction model device (Pred), and outputting the prediction result of the region vector group of the next time.
The distributed energy power generation load prediction system and method based on the graph neural network have the following advantages:
the method comprises the steps of firstly, carrying out initial aggregation on source network load storage distribution conditions and service conditions participating in the power system by utilizing clustering, fully considering the relation among aggregated units, analyzing through a graph neural network model, predicting the power production and consumption of a power grid, further discovering potential safety hazards caused by unbalanced production and consumption, supporting the establishment of an effective power scheduling strategy, and ensuring the operation safety of the power system; meanwhile, the consumption of distributed clean energy is improved, and the maximization of the overall economic benefit is realized;
the method comprises the following steps of (1) designing the characteristics of distributed energy power generation and power consumption units by combining the distribution condition and the service condition of source network charge storage participating in the power system, selecting and extracting the characteristics of distributed power supply, load, energy storage and other participating power system units, reasonably clustering, finding out the optimal power region division unit, simplifying the structural complexity of the participating power system units and improving the processing efficiency;
thirdly, the invention combines the characteristics of distributed energy resources, takes the day as a unit, adopts different time period units, can better reflect the periodic relation and the change of data, and adapts to and meets the real service scene;
according to the power generation and power consumption prediction amount of the power supply, more effective division is formed by adjusting the power area units, the complexity of a power dispatching plan is reduced, and the dispatching efficiency is improved.
Drawings
The invention is further described below with reference to the accompanying drawings.
FIG. 1 is a schematic flow chart of the process of participating in the extraction and clustering of the unit features of the power system;
FIG. 2 is a schematic flow chart of obtaining optimal region division;
FIG. 3 is a schematic flow diagram of a method for predicting future power generation load.
Detailed Description
The distributed energy generation load prediction system and method based on the graph neural network of the invention are described in detail below with reference to the drawings and the specific embodiments of the specification.
Example 1:
the invention discloses a distributed energy generation load prediction system based on a graph neural network, which comprises,
the system comprises an electric power system participation unit feature extraction Encoder (Encoder) and a time T participation electric power system unit feature extraction method, wherein the electric power system participation unit feature extraction Encoder (Encoder) is used for extracting features according to the generated energy and electricity consumption sequence of the participation electric power system unit based on a self-attention network structure, and forming a feature vector of the time T participation electric power system unit by combining the geographical position, standby electricity consumption, output proportion, energy storage capacity and grid-connected capacity features of the participation electric power system unit at the time T;
the power region dividing and clustering device (Cluster) is used for clustering based on the characteristic vectors of all the participating power system units according to the position regions and in combination with the characteristics of source network load storage to form a clustering region category;
the region vector generator (Gen-Vg) is used for calculating the generated energy, the power consumption, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units in the clustered regions divided after clustering in all time periods to form a region vector group;
selecting an optimal region division selector (SelDist) for traversing all region division modes, calculating comprehensive evaluation points of each division mode, calculating the advantages and disadvantages of an evaluation region vector group based on the region vector group, and selecting an optimal region clustering result;
and a future generation load predictor (Pred) for forming the region vector group into a graph structure and predicting and outputting the next time region data by utilizing graph convolution operation.
The future generation load predictor (Pred) in the present embodiment includes,
a generation base map constructor (BG-Gen) for generating an inter-region relationship map structure based on the optimal region clustering result;
an update graph constructor (updG) for updating the inter-region relationship graph based on the data of the region vector group time T;
a graph convolution operator (GCN) for performing graph convolution operation on graph structure data formed by the region vector group by using a Chebyshev convolution algorithm;
and a predicted next time zone data modeler (Wadd) for predicting data of the electric quantity and the electric power consumption of the next time zone by inputting a result of the graph convolution operation (GCN) module and a zone vector group of the current time T in combination with the historical time input zone vector group data based on the LSTM sequence modeling structure.
The working process of the future generation load predictor (Pred) in this embodiment is specifically as follows:
(1) generating an inter-region relation graph structure by using a generation base graph constructor (BG-Gen) according to the dividing mode of the optimal region clustering result;
(2) adopting an optimal region clustering result division mode, calculating all historical data of the division mode through a region vector generator (Gen-Vg) to form a vector group sequence from time 1 to time T, and continuously inputting the vector group sequence to a future power generation load prediction model device (Pred);
(3) generating an inter-region relational graph structure based on a generated basic graph constructor (BG-Gen), updating inter-region relational graph structure data by using an updating graph constructor (updG), calculating an intermediate result through a graph convolution operator (GCN) and a next time prediction region data modeler (Wadd), and finally outputting a region vector group prediction result of time T + 1;
(4) and continuously inputting the future time prediction result into a future power generation load prediction model device (Pred) and outputting the prediction result of the region vector group of the next time.
The eigenvector of the unit participating in the power system in the embodiment is divided into two channels of the power generation amount and the power consumption amount, the T time period is a unit of one day, and the component of the eigenvector records the power generation amount and the power consumption amount according to 15 minutes or 30 minutes.
The process of obtaining the region vector group in this embodiment is specifically as follows:
(1) assembling a characteristic vector vm from time 1 to current time T for each participating power system unit m at each time point T on the basis of historical data of all participating power system units to form a power generation and power consumption dual-channel vector sequence (vm1, vm2.. vmt);
(2) processing the vector sequence (vm1, vm2.. vmt) by utilizing a power system participation unit feature extraction Encoder (Encoder) to form an encoding feature vector Vsmt;
(3) extracting other characteristic vectors Vsmt 'of the geographical position, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of the power system participating unit m at the time point t by using a power system participating unit characteristic extraction Encoder (Encoder), and combining the other characteristic vectors Vsmt' and the encoding characteristic vector Vsmt to form Vptm;
(4) clustering Vpt1, Vpt2.. Vptm by using a power region division clustering device (Cluster) to form region division clusters for each time point T from time 1 to current time T;
(5) merging the same clustering result modes in the T types of region division modes, and calculating the generated energy, the power consumption, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units in time T according to the divided regions 1 to N for each region division mode to obtain N vectors to form a vector group; from time 1 to time T, T vector groups are finally formed.
The power system participating units in the embodiment comprise power generation sources, power transmission and distribution networks, power loads and energy storage;
the power generation unit comprises a distributed power supply for wind power generation, photovoltaic power generation and water conservancy power generation, fossil fuel power generation of traditional thermal power generation and nuclear power;
the power transmission and distribution network is used for ensuring the transmission of electric power; the power generation power supply, the power load and the energy storage are connected into a main power grid or independently form a power micro-grid system;
the electric load means consumption of electric power including industrial load, residential load and commercial load;
the stored energy is used to enable storage and output of electrical energy sources, participating in the power system as loads and sources.
Example 2:
the invention relates to a distributed energy power generation load prediction method based on a graph neural network, which is characterized in that a clustering technology is utilized to carry out preliminary aggregation by combining the distribution condition and the service condition of source network charge storage of power system units, and the power production and the power consumption of a power grid are predicted by analyzing the graph neural network according to the connection between the aggregated power system units; the method comprises the following specific steps:
s1, participating in power system unit feature extraction and clustering: according to the characteristics of distributed energy power generation and consumption units, based on geographical distribution, power grid laying and service conditions, selecting and extracting the characteristics of distributed power sources, loads and energy storage participating power system units through an electric power system participating unit characteristic extraction Encoder (Encoder), and reasonably clustering by using a power region division clustering device (Cluster);
s2, obtaining the optimal region division: traversing all the area division modes by using a SelDist (SelDist) selector, calculating the comprehensive evaluation score of each division mode, finding out the optimal power area division unit, and selecting the optimal area clustering result, as shown in figure 2;
s3, predicting future power generation load: each power region is divided into units to form a graph structure, and the future generation load prediction modeler (Pred) predicts the power generation and power consumption of the power region units at the future time by using a graph neural network in combination with historical data and power planning data.
The future generation load predictor (Pred) in the present embodiment includes,
a generation base map constructor (BG-Gen) for generating an inter-region relationship map structure based on the optimal region clustering result;
an update graph constructor (updG) for updating the inter-region relationship graph based on the data of the region vector group time T;
a graph convolution operator (GCN) for performing graph convolution operation on graph structure data formed by the region vector group by using a Chebyshev convolution algorithm;
and a predicted next time zone data modeler (Wadd) for predicting data of the electric quantity and the electric power consumption of the next time zone by inputting a result of the graph convolution operation (GCN) module and a zone vector group of the current time T in combination with the historical time input zone vector group data based on the LSTM sequence modeling structure.
As shown in fig. 1, the participating power system unit feature extraction and clustering in step S1 of the present embodiment are as follows:
s101, assembling a characteristic vector vm from time 1 to current time T for each participated power system unit m at each time point T based on historical data of all participated power system units to form a power generation and power consumption dual-channel vector sequence (vm1, vm2.. vmt);
s102, processing the vector sequence (vm1, vm2.. vmt) by using a power system participation unit feature extraction Encoder (Encoder) to form an encoding feature vector Vsmt;
s103, extracting other feature vectors Vsmt 'of the geographical position, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of the power system participating unit m at the time point t by using a power system participating unit feature extraction Encoder (Encoder), and combining the other feature vectors Vsmt' and the encoding feature vector Vsmt to form Vptm;
s104, clustering the Vpt1 and the Vpt2.. Vptm by using an electric power region division clustering device (Cluster) to form a region division Cluster for each time point T from the time 1 to the current time T;
s105, merging the same clustering result modes in the T types of region division modes, and calculating the generated energy, the electric power consumption, the standby electric power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units at the time T according to the divided regions 1 to N for each region division mode to obtain N vectors to form a vector group; from time 1 to time T, T vector groups are finally formed.
As shown in fig. 3, the predicted future power generation load in step S3 of the present embodiment is specifically as follows:
s301, generating an inter-region relation graph structure by using a generation base graph constructor (BG-Gen) according to the dividing mode of the optimal region clustering result;
s302, adopting an optimal region clustering result division mode, calculating all historical data of the division mode through a region vector generator (Gen-Vg) to form a vector group sequence from time 1 to time T, and continuously inputting the vector group sequence to a future power generation load prediction model device (Pred);
s303, generating an inter-region relational graph structure based on a generated basic graph constructor (BG-Gen), updating inter-region relational graph structure data by using an updating graph constructor (updG), calculating an intermediate result through a graph convolution operator (GCN) and a next time prediction region data modeler (Wadd), and finally outputting a region vector group prediction result of time T + 1;
s304, continuously inputting the future time prediction result into a future power generation load prediction model device (Pred), and outputting the prediction result of the region vector group of the next time.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A distributed energy resource power generation load prediction system based on a graph neural network is characterized by comprising,
the electric power system participation unit feature extraction encoder is used for extracting features according to the generated energy and electricity consumption sequence of the participation electric power system unit based on the self-attention network structure, and forming a feature vector of the time T participation electric power system unit by combining the geographical position, the standby electricity consumption, the output proportion, the energy storage capacity and the grid-connected capacity feature of the time T participation electric power system unit;
the power region dividing and clustering device is used for clustering based on the characteristic vectors of all the participating power system units according to the position regions and in combination with the characteristics of source network load storage to form a clustering region category;
the region vector generator is used for calculating the generated energy, the power consumption, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units in the clustered regions divided after clustering in all time periods to form a region vector group;
selecting an optimal region division selector, traversing all region division modes, calculating comprehensive evaluation points of each division mode, calculating the advantages and disadvantages of an evaluation region vector group based on the region vector group, and selecting an optimal region clustering result;
and the future power generation load predictor is used for forming a graph structure by combining the region vectors and predicting and outputting the next time region data by utilizing graph convolution operation.
2. The distributed energy generation load prediction system based on graph neural network of claim 1, wherein the future generation load predictor comprises,
a generation base graph constructor for generating an inter-region relation graph structure based on the optimal region clustering result;
an update graph constructor for updating the inter-region relationship graph based on the data of the region vector group time T;
the graph convolution operator is used for performing graph convolution operation on graph structure data formed by the region vector group by adopting a Chebyshev convolution algorithm;
and the data modeler for predicting the next time zone is used for predicting the data of the electric quantity and the electric power consumption of the next time zone by inputting the result of a graph convolution operation (GCN) module and the zone vector group of the current time T and combining the historical time input zone vector group data based on the LSTM sequence modeling structure.
3. The distributed energy generation load prediction system based on the graph neural network according to claim 1 or 2, wherein the future generation load predictor specifically works as follows:
(1) for the dividing mode of the optimal region clustering result, generating a relationship graph structure between regions by using a generated basic graph structure device;
(2) adopting an optimal regional clustering result division mode, calculating all historical data of the division mode through a regional vector generator to form a vector group sequence from time 1 to time T, and continuously inputting the vector group sequence to a future power generation load prediction model device;
(3) generating an inter-region relational graph structure based on a generated basic graph constructor, updating inter-region relational graph structure data by using an updated graph constructor, calculating an intermediate result through a graph convolution operator and a next time prediction region data model, and finally outputting a region vector group prediction result of time T + 1;
(4) and continuously inputting the future time prediction result into a future power generation load prediction model device, and outputting the region vector group prediction result of the next time.
4. The distributed energy generation load prediction system based on the graph neural network according to claim 1, characterized in that the eigenvectors of the participating power system units are divided into two channels of power generation amount and power consumption amount, the T time period is a unit of one day, and the components of the eigenvectors record the power generation amount and the power consumption amount in 15 minutes or 30 minutes.
5. The distributed energy generation load prediction system based on the graph neural network according to claim 1 or 4, wherein the region vector group obtaining process is specifically as follows:
(1) assembling a characteristic vector vm from time 1 to current time T for each participated power system unit m and each time point T on the basis of historical data of all participated power system units by taking a set time period as a unit to form a two-channel vector sequence (vm) of power generation amount and power consumption amount1、vm2...vmt)；
(2) Extracting encoder pair vector sequence (vm) by using power system participating unit characteristics1、vm2...vmt) Processing to form a coded feature vector Vsmt；
(3) Extracting other feature vectors Vsm of the geographical position, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of the power system participating unit m at the time point t by using the power system participating unit feature extraction encodert', to apply other feature vectors Vsmt' and coding feature vector VsmtAre combined to form Vptm；
(4) From time 1 to current time T, for each time point T, using a pair of power domain partitionizers Vpt1、Vpt2...VptmClustering to form regional division clusters;
(5) merging the same clustering result modes in the T types of region division modes, and calculating the generated energy, the power consumption, the standby power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units in time T according to the divided regions 1 to N for each region division mode to obtain N vectors to form a vector group; from time 1 to time T, T vector groups are finally formed.
6. The graph neural network-based distributed energy generation load prediction system of claim 1, wherein the participating power system units comprise power generation sources, power transmission and distribution grids, power loads, and energy storage;
the power generation unit comprises a distributed power supply for wind power generation, photovoltaic power generation and water conservancy power generation, fossil fuel power generation of traditional thermal power generation and nuclear power;
the power transmission and distribution network is used for ensuring the transmission of electric power; the power generation power supply, the power load and the energy storage are connected into a main power grid or independently form a power micro-grid system;
the electric load means consumption of electric power including industrial load, residential load and commercial load;
the stored energy is used to enable storage and output of electrical energy sources, participating in the power system as loads and sources.
7. A distributed energy power generation load prediction method based on a graph neural network is characterized in that the method is characterized in that a clustering technology is utilized to carry out preliminary aggregation by combining distribution conditions and service conditions of source network charge storage of power system units, and according to the connection between the aggregated power system units, the graph neural network is used for analyzing to predict the power production and consumption of a power grid; the method comprises the following specific steps:
s1, participating in power system unit feature extraction and clustering: according to the characteristics of distributed energy power generation and consumption units, based on geographical distribution, power grid laying and service conditions, selecting and extracting the characteristics of distributed power sources, loads and energy storage participating power system units through a power system participating unit characteristic extraction encoder, and reasonably clustering by using a power region division clustering device;
s2, obtaining the optimal region division: traversing all the area division modes by using the optimal area division selector, calculating the comprehensive evaluation score of each division mode, finding out the optimal power area division unit, and selecting the optimal area clustering result;
s3, predicting future power generation load: and dividing each power area into units to form a graph structure, and predicting the power generation and power consumption of the power area units at the future time by a future power generation load prediction model through a graph neural network in combination with historical data and power planning data.
8. The method of claim 7, wherein the future power generation load predictor comprises,
a generation base graph constructor for generating an inter-region relation graph structure based on the optimal region clustering result;
an update graph constructor for updating the inter-region relationship graph based on the data of the region vector group time T;
the graph convolution operator is used for performing graph convolution operation on graph structure data formed by the region vector group by adopting a Chebyshev convolution algorithm;
and the data modeler for predicting the next time zone is used for predicting the data of the electric quantity and the electric power consumption of the next time zone by inputting the result of a graph convolution operation (GCN) module and the zone vector group of the current time T and combining the historical time input zone vector group data based on the LSTM sequence modeling structure.
9. The method for predicting distributed energy generation load based on the graph neural network according to claim 7 or 8, wherein the participating power system unit feature extraction and clustering in step S1 are specifically as follows:
s101, assembling a characteristic vector vm from time 1 to current time T for each participated power system unit m and each time point T based on historical data of all participated power system units by taking a set time period as a unit to form a two-channel vector sequence (vm) of power generation amount and power consumption amount1、vm2...vmt)；
S102, extracting a coder pair vector sequence (vm) by using the power system participating in unit feature extraction1、vm2...vmt) Processing to form a coded feature vector Vsmt；
S103, extracting the unit m participating in the power system by utilizing the power system participating unit feature extraction encoderGeographical position, standby power consumption, output proportion, energy storage capacity and other characteristic vectors Vsm of grid-connected capacity at time point tt', to apply other feature vectors Vsmt' and coding feature vector VsmtAre combined to form Vptm；
S104, from time 1 to current time T, for each time point T, using power region division clustering device to Vpt1、Vpt2...VptmClustering to form regional division clusters;
s105, merging the same clustering result modes in the T types of region division modes, and calculating the generated energy, the electric power consumption, the standby electric power consumption, the output proportion, the energy storage capacity and the grid-connected capacity of all the participating power system units at the time T according to the divided regions 1 to N for each region division mode to obtain N vectors to form a vector group; from time 1 to time T, T vector groups are finally formed.
10. The method for predicting distributed energy generation load based on neural network of claim 9, wherein the predicted future generation load in step S3 is specifically as follows:
s301, generating a relationship graph structure among the regions by utilizing a basic graph generator for the dividing mode of the optimal region clustering result;
s302, adopting an optimal region clustering result division mode, calculating all historical data of the division mode through a region vector generator to form a vector group sequence from time 1 to time T, and continuously inputting the vector group sequence to a future power generation load prediction model device;
s303, generating an inter-region relational graph structure based on a generated basic graph constructor, updating inter-region relational graph structure data by using an updated graph constructor, calculating an intermediate result through a graph convolution operator and a region data model for predicting the next time, and finally outputting a region vector group prediction result of time T + 1;
and S304, continuously inputting the future time prediction result into a future power generation load prediction model, and outputting the region vector group prediction result of the next time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111336833.1A CN114154558B (en) | 2021-11-12 | 2021-11-12 | Distributed energy power generation load prediction system and method based on graph neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111336833.1A CN114154558B (en) | 2021-11-12 | 2021-11-12 | Distributed energy power generation load prediction system and method based on graph neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114154558A true CN114154558A (en) | 2022-03-08 |
| CN114154558B CN114154558B (en) | 2024-05-21 |
Family
ID=80459672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111336833.1A Active CN114154558B (en) | 2021-11-12 | 2021-11-12 | Distributed energy power generation load prediction system and method based on graph neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114154558B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114818483A (en) * | 2022-04-14 | 2022-07-29 | 东南大学溧阳研究院 | Electromechanical disturbance positioning and propagation prediction method based on graph neural network |
| CN115313377A (en) * | 2022-08-26 | 2022-11-08 | 中国长江三峡集团有限公司 | Power load prediction method and system |
| CN115689069A (en) * | 2023-01-03 | 2023-02-03 | 东北电力大学 | Power grid dispatching control method and system based on artificial intelligence |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111461392A (en) * | 2020-01-23 | 2020-07-28 | 华中科技大学 | Power failure prediction method and system based on graph neural network |
| US20210192358A1 (en) * | 2018-05-18 | 2021-06-24 | Deepmind Technologies Limited | Graph neural network systems for behavior prediction and reinforcement learning in multple agent environments |
| CN113392781A (en) * | 2021-06-18 | 2021-09-14 | 山东浪潮科学研究院有限公司 | Video emotion semantic analysis method based on graph neural network |
| CN113505458A (en) * | 2021-07-26 | 2021-10-15 | 中国电力科学研究院有限公司 | Cascading failure key trigger branch prediction method, system, equipment and storage medium |
-
2021
- 2021-11-12 CN CN202111336833.1A patent/CN114154558B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210192358A1 (en) * | 2018-05-18 | 2021-06-24 | Deepmind Technologies Limited | Graph neural network systems for behavior prediction and reinforcement learning in multple agent environments |
| CN111461392A (en) * | 2020-01-23 | 2020-07-28 | 华中科技大学 | Power failure prediction method and system based on graph neural network |
| CN113392781A (en) * | 2021-06-18 | 2021-09-14 | 山东浪潮科学研究院有限公司 | Video emotion semantic analysis method based on graph neural network |
| CN113505458A (en) * | 2021-07-26 | 2021-10-15 | 中国电力科学研究院有限公司 | Cascading failure key trigger branch prediction method, system, equipment and storage medium |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114818483A (en) * | 2022-04-14 | 2022-07-29 | 东南大学溧阳研究院 | Electromechanical disturbance positioning and propagation prediction method based on graph neural network |
| CN115313377A (en) * | 2022-08-26 | 2022-11-08 | 中国长江三峡集团有限公司 | Power load prediction method and system |
| CN115313377B (en) * | 2022-08-26 | 2023-08-08 | 中国长江三峡集团有限公司 | Power load prediction method and system |
| CN115689069A (en) * | 2023-01-03 | 2023-02-03 | 东北电力大学 | Power grid dispatching control method and system based on artificial intelligence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114154558B (en) | 2024-05-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114154558B (en) | Distributed energy power generation load prediction system and method based on graph neural network | |
| CN110298138B (en) | Comprehensive energy system optimization method, device, equipment and readable storage medium | |
| CN106503839B (en) | Hierarchical planning method for offshore wind farm annular current collection network | |
| Sharma et al. | A critical and comparative review of energy management strategies for microgrids | |
| Bakken et al. | Grip-grids with intelligent periphery: Control architectures for grid2050 π | |
| CN104600713A (en) | Device and method for generating day-ahead reactive power dispatch of power distribution network containing wind/photovoltaic power generation | |
| CN103455729B (en) | A kind of method for determining photovoltaic-and-enegrid-connected grid-connected combined power generation dispatch value | |
| Hong et al. | An integrated approach for multi-objective optimisation and MCDM of energy internet under uncertainty | |
| Huang et al. | Smart energy management system based on reconfigurable AI chip and electrical vehicles | |
| CN108075471B (en) | Multi-objective constraint optimization power grid scheduling strategy based on stochastic power output prediction | |
| Hu et al. | Probabilistic electric vehicle charging demand forecast based on deep learning and machine theory of mind | |
| CN116402210A (en) | Multi-objective optimization method, system, equipment and medium for comprehensive energy system | |
| CN116050754A (en) | Digital twinning-based virtual power plant collaborative management and control method and system | |
| Abedinia et al. | Synergizing efficient optimal energy hub design for multiple smart energy system players and electric vehicles | |
| CN111105096A (en) | New energy grid-connected layout planning method and system | |
| Mirjalili et al. | A comparative study of machine learning and deep learning methods for energy balance prediction in a hybrid building-renewable energy system | |
| Zhang et al. | Joint forecasting of regional wind and solar power based on attention neural network | |
| CN111325368A (en) | Photovoltaic power prediction method and device for light storage type electric vehicle charging station | |
| Cheng et al. | The power system multi-objective optimization dispatching containing virtual power plant | |
| Guo | A Study of Smart Grid Program Optimization Based on K-Mean Algorithm | |
| CN115358509A (en) | User comprehensive energy consumption optimization control method and system | |
| KR102562761B1 (en) | Method for intelligent day-ahead energy sharing scheduling of the P2P prosumer community in smart grid | |
| Lv et al. | Photovoltaic Power Forecasting in Dunhuang based on Recurrent Neural Network | |
| CN115313377B (en) | Power load prediction method and system | |
| Guo et al. | Prediction Model of Energy Traceability Efficiency of Distributed Power Network System Based on Genetic Algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |