CN116565863A - Short-term photovoltaic output prediction method based on space-time correlation - Google Patents

Short-term photovoltaic output prediction method based on space-time correlation Download PDF

Info

Publication number
CN116565863A
CN116565863A CN202310837787.6A CN202310837787A CN116565863A CN 116565863 A CN116565863 A CN 116565863A CN 202310837787 A CN202310837787 A CN 202310837787A CN 116565863 A CN116565863 A CN 116565863A
Authority
CN
China
Prior art keywords
photovoltaic
matrix
photovoltaic power
data
output
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
Application number
CN202310837787.6A
Other languages
Chinese (zh)
Other versions
CN116565863B (en
Inventor
李豪
马刚
孟宇翔
李伟康
李天宇
沈静文
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.)
Nanjing Normal University
Original Assignee
Nanjing Normal University
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 Nanjing Normal University filed Critical Nanjing Normal University
Priority to CN202310837787.6A priority Critical patent/CN116565863B/en
Publication of CN116565863A publication Critical patent/CN116565863A/en
Application granted granted Critical
Publication of CN116565863B publication Critical patent/CN116565863B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/10—Pre-processing; Data cleansing
    • G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • 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
    • 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
    • H02J3/381—Dispersed generators
    • 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
    • H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
    • 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
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Molecular Biology (AREA)
  • Economics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Strategic Management (AREA)
  • Marketing (AREA)
  • Power Engineering (AREA)
  • General Business, Economics & Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Tourism & Hospitality (AREA)
  • Game Theory and Decision Science (AREA)
  • Probability & Statistics with Applications (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Primary Health Care (AREA)

Abstract

The invention discloses a short-term photovoltaic output prediction method based on space-time correlation, which comprises the following steps: acquiring historical power of a plurality of photovoltaic power stations in a certain area and historical data of power station influence factors to be tested, and establishing a characteristic database; selecting a strongly-correlated meteorological factor by pearson correlation analysis; introducing historical power generation and constructing a delay input characteristic; establishing a GCN model based on a topological structure of the regional photovoltaic power station, transversely tracking the photovoltaic spatial evolution modes of the multi-photovoltaic power station, aggregating the spatial features of adjacent photovoltaic electric fields, and outputting a spatial feature set containing the spatial evolution modes; performing modal decomposition on the input features by adopting SGMD to obtain a multi-stage modal subsequence capable of representing time sequence change features of historical data, and constructing a high-dimensional feature set; and performing feature extraction and photovoltaic power generation power prediction by adopting a CNN-BiLSTM neural network, and performing error evaluation. The invention provides a basis for energy management and optimal scheduling of the novel power system, and has higher prediction precision and feasibility.

Description

Short-term photovoltaic output prediction method based on space-time correlation
Technical Field
The invention relates to the field of new energy power generation and network access photovoltaic prediction, in particular to a short-term photovoltaic output prediction method based on space-time correlation.
Background
In recent years, with the continuous consumption of fossil fuels and the continuous increase of new energy demands, the effective utilization of renewable energy sources for power generation has become a hot topic. Among them, photovoltaic is considered as one of the most promising renewable energy sources, and is widely studied and used, with its permeability gradually increasing in electric power systems. However, due to the chaos and randomness of the weather conditions, the photovoltaic power generation presents uncertainty and randomness, and as the permeability of the photovoltaic power generation in the power system is higher and higher, a certain impact may be caused on the power system. Therefore, the accuracy of photovoltaic output prediction is improved, which is an important basis for maintaining the stable and efficient operation of the novel power system and is important for the economic dispatch of the power grid.
Disclosure of Invention
The invention aims to: the invention aims to provide a short-term photovoltaic output prediction method based on space-time correlation, so that a foundation is provided for energy management and optimal scheduling of a novel power system, daily scheduling and daily scheduling are better realized, and an important effect is played on safe and economic operation of an energy system.
The technical scheme is as follows: the invention discloses a short-term photovoltaic output prediction method based on space-time correlation, which comprises the following steps of:
(1) And acquiring historical power of adjacent photovoltaic power stations in a certain area and historical data of power station influence factors to be tested, establishing a characteristic database, removing abnormal values and filling missing values of original data, and normalizing all data by using a maximum-minimum method due to different dimensionalities of different characteristics.
(2) Selecting a strongly-correlated meteorological factor by pearson correlation analysis; and introducing historical generated power and constructing a delay input characteristic.
(3) Establishing a GCN model based on a topological structure of the regional photovoltaic power station, transversely tracking the photovoltaic spatial evolution modes of the multi-photovoltaic power station, aggregating the spatial features of adjacent photovoltaic electric fields, and outputting a spatial feature set containing the spatial evolution modes.
(4) The advantage of nonlinear unsteady sequence processing by SGMD is utilized, irradiance, temperature, humidity, rainfall, regional photovoltaic output aggregation characteristics and photovoltaic output of a power station to be tested are decomposed into a group of relatively stable subsequences, error components with larger partial volatility are removed, a multi-layer modal subsequence characteristic matrix which can reflect time sequence change characteristics of historical data is constructed, and subsequence input matrixes are formed by corresponding subsequences of different input characteristics and photovoltaic output.
(5) And performing feature extraction and photovoltaic power generation power prediction by adopting a CNN-BiLSTM neural network, obtaining a final prediction result through inverse normalization and subsequence reconstruction, and performing error evaluation on the final prediction result.
The step (1) specifically comprises the following steps:
the historical data of meteorological factors of a plurality of photovoltaic power stations in a certain area are arranged; and carrying out outlier removal and data complementation on the sample set.
And (1.1) giving a confidence probability by using a Laida criterion method, determining a confidence limit, considering that errors exceeding the confidence limit do not belong to a random error range, and eliminating abnormal values to obtain historical power of a plurality of photovoltaic power stations and weather factor historical data of the power station to be tested.
And (1.2) filling missing data in the historical output data and the meteorological factor historical data respectively by adopting a bilinear interpolation method on the basis of eliminating abnormal values.
And (1.3) carrying out normalization processing on the cleaned historical output data and the meteorological factor data of the power station to be tested, wherein the normalization formula is as follows:
in the method, in the process of the invention,Xis the data before the normalization and is used for the data,X´is the corresponding normalized data of the data set,X min andX max is the minimum and maximum in the sequence.
The step (2) specifically comprises the following steps:
(2.1) analyzing the correlation between the photovoltaic output and the meteorological factors by using the Pearson correlation coefficient so as to determine a strong correlation meteorological factor, wherein the calculation formula is as follows:
in the method, in the process of the invention,P XY for the Pearson correlation coefficient,X、Yis a correlation factor;、/>respectively the average of the two factors.
(2.2) considering that the photovoltaic output has strong time sequence continuity in the period, the moment to be measuredtHas strong correlation with the historical power, and needs to introduce the historical output as predictionIs a time delay input feature of (a); the delay step length can directly influence the prediction effect of the model, if the delay step length is too long, the prediction model is over-fitted, and if the delay step length is too short, the model cannot completely learn the time dependence of the input characteristics; and according to the selection result of the delay step length of the photovoltaic power, carrying out delay input characteristic expansion on the target photovoltaic power station, and introducing historical photovoltaic power generation power as input characteristics.
The step (3) specifically comprises the following steps:
and (3.1) the output forces of the photovoltaic power stations in the area have spatial similarity, when strong weather mutation occurs, the influence on the photovoltaic power stations in the same area also has time sequence, when the output force of one adjacent power station first fluctuates greatly, the GCN network forms the spatial characteristic of the power station to be tested by aggregating the output forces of the adjacent photovoltaic power stations, the output fluctuation caused by weather mutation is quickly captured, the photovoltaic output connectivity in space is established, and the prediction precision is improved.
(3.1.1) defining a map structure of a photovoltaic power plant in a region, wherein the spatial similarity and characteristic divergence of the output changes of a plurality of photovoltaic power plants in the same region are considered, so that one map structure data needs to be definedGDescribing their spatial dependencies. The undirected graph of a plurality of photovoltaic power stations in a certain area is expressed as:
in the formula, each photovoltaic power station is regarded as one node in the undirected graphv i ，VA group of photovoltaic power plants representing a region,Nthe number of the regional photovoltaic power stations is represented,Erepresenting a set of relationships between each node (i.e., each photovoltaic power plant).
The connection feature between the nodes uses oneN×NMatrix of dimensionsAIs referred to as an adjacency matrix, which can reflect the structural features of the entire graph. The invention adopts the absolute value of the Pearson correlation coefficient between nodes to represent the spatial correlation of adjacent photovoltaic power stations, and uses the correlation matrix between multiple photovoltaic power stations to represent the adjacent matrix. Will beThe historical power generation data of each photovoltaic power station is regarded as node attribute characteristics, so as to construct aN×FFeature matrix of dimensionX，FRepresenting the historical time series length.
Taking multiple photovoltaic power plants as an example, the elapsed time is collectedt-hTo the time to be measuredtIs used for constructing an adjacency matrix according to historical power data of the power generationAFeature matrixXIs input into the GCN to obtain spatial characteristics of a plurality of adjacent photovoltaic power plants.
(3.1.2) region photovoltaic power station information aggregation, wherein the graph convolution neural network derives information of nodes to be detected by utilizing information of other nodes, the core idea is to aggregate by utilizing edge characteristics and node characteristics to generate new node representations, and the essential purpose is to extract spatial characteristics of a graph, namely, spatial characteristics of the power stations to be detected are obtained by aggregating spatial characteristic data of other photovoltaic power stations in the same region; the graph book integration is divided into two major categories, namely a spectrum-based method and a space-based method; the present invention models spatial features using a space-based GCN.
Adjacency matrix for multi-photovoltaic power stationAFeature matrixXInput into GCN model, GCN is implemented by pairingAAndXcoding is carried out, so that the spatial correlation between the photovoltaic power station to be tested and other peripheral power station units can be obtained; the operation formula between the graph convolution layers is expressed as:
in the method, in the process of the invention,His a feature of each layer, and for the input layer,Hnamely the feature matrixX；σA nonlinear activation function;is->A degree matrix of (2);Wis a learning weight.
The step (4) specifically comprises the following steps:
(4.1) irradiance and photovoltaic power signals are highly oscillatory and unstable signals whose volatility can make it difficult to predict such signals. The invention adopts a mode decomposition method to decompose the time sequence signal into a plurality of components, so as to generate a time sequence with more stability and smaller fluctuation. These temporal sub-sequence signals may improve the prediction accuracy.
Time series data of meteorological factors and photovoltaic power generation power are set asx=[x 1 ,x 2 ,……，x n ]WhereinnFor time series signal length, builddDimensional track matrixXThe following formula is shown:
in the method, in the process of the invention,τis a delay time.
According to the matrixAConstructing hamilton matrixM：
Order of principleC=M 2 Thereby constructing Xin Zhengjiao matrixQ：
In the method, in the process of the invention,Bfor the upper triangular matrix, calculating the characteristic value asλ 1 ，λ 2 ，……，λ d Matrix thenAThe characteristic values of (2) are:
in the method, in the process of the invention,σ i according to a matrixAIs characterized by the octyl geometric spectrum distribution,Q(i=1,2,……,d)from the following componentsσ i The reconstruction of each component matrix is as follows:
in the method, in the process of the invention,i=1,2,……,d. Then an initial single component trajectory matrixZThe following formula is shown:
in the method, in the process of the invention,Z∈R mⅹd . Definition of the definitionZ i Middle elementz ij ，1≦i≦d，1≦j≦mAnd (2) andd * =min(m,d)，m * =max(m,d)，n=m+(d-1)τthen:
the diagonal average transition matrix is then obtained by:
matrix can be averaged by diagonalZConversion intod×nMatrix of dimensionsYThereby inputting variable signalsxIs decomposed intodXin Jihe modal components with different center frequencies:
in the method, in the process of the invention,Yis a primitive sequence;Y i （i=1,2,……,d) To decompose the subsequence.
The step (5) specifically comprises the following steps:
(5.1) in order to better mine environmental factors, periodicity of photovoltaic power generation and existing space-time characteristics, the invention adopts CNN-BiLSTM consisting of a convolutional neural network and a two-way long-short-term memory neural network layer as a characteristic extraction and power prediction model, and takes photovoltaic power generation power, meteorological characteristics and GCN space aggregation characteristics as inputs, so that the model can deeply mine correlation among input characteristics and variation trend of characteristic time sequences.
(5.1.1) the convolutional neural network model consists of three layers, an input layer, a hidden layer and an output layer. The hidden layer is a feature extraction layer and consists of a convolution layer and a pooling layer. To improve the feature extraction capability of the model, a plurality of convolution layers are designed. The mathematical expression is as follows:
in the method, in the process of the invention,α i is the firstiA plurality of convolutional layer input features;W i andb i and respectively represent the firstiA weight matrix and a paranoid vector of the convolution kernels of the convolution layers;f i is the firstiOutputting characteristics by the convolution layers;σrepresenting the ReLU activation function.
(5.1.2) photovoltaic power generation power as periodic data exhibits a great timing dependence. Therefore, the feature vector extracted by the CNN module is flattened by the flat layer and then is input into the BiLSTM neural network, so that the time sequence characteristic of the input feature is extracted, the forward propagation characteristic and the backward propagation characteristic are calculated, the prediction precision is further improved, and the mathematical expression is as follows:
in the method, in the process of the invention,n t is an input vector;for the forward propagation layer output value: />An output value that is a backward propagation layer;θ t is the output of the output layer;δactivating a function for Tanh; />、/>、/>、/>、/>And->Is a weight matrix; />、/>And->Is a bias vector.
A computer storage medium having stored thereon a computer program which when executed by a processor implements a short-term photovoltaic output prediction method based on spatio-temporal correlation as described above.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing a short-term photovoltaic output prediction method based on spatio-temporal correlation as described above when executing the computer program.
The beneficial effects are that: compared with the prior art, the invention has the following advantages:
1. according to the invention, a short-term photovoltaic output prediction mixed model based on space-time correlation is provided, the effects of time sequence characteristics, regional topological characteristics and external influence factors are comprehensively considered, a connection diagram of a plurality of photovoltaic power stations in the same region is constructed through a GCN network, the topological structure of non-Euclidean data is captured, node characteristics and connection edge characteristics of peripheral photovoltaic power stations are utilized to deduce information of the power stations to be detected, and space input characteristics are formed, so that the space characteristics of the power stations to be detected are mined, weather mutation can be captured more quickly and accurately, and fluctuation of photovoltaic power generation power is predicted;
2. according to the invention, an SGMD modal decomposition method is introduced simultaneously, influence factors, spatial characteristics and photovoltaic power generation power are decomposed respectively, the time sequence change characteristics of historical time sequence data are extracted better, the influence of volatility, instability and randomness of an original data sequence on a prediction model learning photovoltaic output curve is reduced, and the prediction precision is improved;
3. the invention utilizes the CNN-BiLSTM combination prediction model to dynamically pay attention to the space characteristic and the time sequence characteristic, digs the potential law and the time sequence dependence of the input characteristic, comprehensively considers the factors in multiple aspects influencing the photovoltaic output, and has higher prediction precision and feasibility.
Drawings
FIG. 1 is a flow chart of the steps of the method of the present invention;
FIG. 2 is a schematic diagram of a time delay input feature;
fig. 3 is a GCN model diagram.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings.
As shown in fig. 1, a short-term photovoltaic output prediction method based on space-time correlation includes the following steps:
(1) And acquiring historical power of adjacent photovoltaic power stations in a certain area and historical data of power station influence factors to be tested, establishing a characteristic database, removing abnormal values and filling missing values of original data, and normalizing all data by using a maximum-minimum method due to different dimensionalities of different characteristics.
The historical data of meteorological factors of a plurality of photovoltaic power stations in a certain area are arranged; and carrying out outlier removal and data complementation on the sample set.
And (1.1) giving a confidence probability by using a Laida criterion method, determining a confidence limit, considering that errors exceeding the confidence limit do not belong to a random error range, and eliminating abnormal values to obtain historical power of a plurality of photovoltaic power stations and weather factor historical data of the power station to be tested.
And (1.2) filling missing data in the historical output data and the meteorological factor historical data respectively by adopting a bilinear interpolation method on the basis of eliminating abnormal values.
And (1.3) carrying out normalization processing on the cleaned historical output data and the meteorological factor data of the power station to be tested, wherein the normalization formula is as follows:
in the method, in the process of the invention,Xis the data before the normalization and is used for the data,X´is the corresponding normalized data of the data set,X min andX max is the minimum and maximum in the sequence.
(2) Selecting a strongly-correlated meteorological factor by pearson correlation analysis; and introducing historical generated power and constructing a delay input characteristic.
(2.1) analyzing the correlation between the photovoltaic output and the meteorological factors by using the Pearson correlation coefficient so as to determine a strong correlation meteorological factor, wherein the calculation formula is as follows:
in the method, in the process of the invention,P XY for the Pearson correlation coefficient,X、Yis a correlation factor;、/>respectively the average of the two factors.
(2.2) As shown in FIG. 2, considering that the photovoltaic output has strong time sequence continuity in the period, the moment to be measuredtHas strong correlation with the historical power, and needs to introduce the historical output as the predicted delay inputFeatures; the delay step length can directly influence the prediction effect of the model, if the delay step length is too long, the prediction model is over-fitted, and if the delay step length is too short, the model cannot completely learn the time dependence of the input characteristics; and according to the selection result of the delay step length of the photovoltaic power, carrying out delay input characteristic expansion on the target photovoltaic power station, and introducing historical photovoltaic power generation power as input characteristics.
(3) Establishing a GCN model based on a topological structure of the regional photovoltaic power station, transversely tracking the photovoltaic spatial evolution modes of the multi-photovoltaic power station, aggregating the spatial features of adjacent photovoltaic electric fields, and outputting a spatial feature set containing the spatial evolution modes.
And (3.1) the output forces of the photovoltaic power stations in the area have spatial similarity, when strong weather mutation occurs, the influence on the photovoltaic power stations in the same area also has time sequence, when the output force of one adjacent power station first fluctuates greatly, the GCN network forms the spatial characteristic of the power station to be tested by aggregating the output forces of the adjacent photovoltaic power stations, the output fluctuation caused by weather mutation is quickly captured, the photovoltaic output connectivity in space is established, and the prediction precision is improved.
(3.1.1) As shown in FIG. 3, a regional photovoltaic power plant graph structure is defined, and it is necessary to define a graph structure data in consideration of spatial similarity and characteristic divergence of output changes of a plurality of photovoltaic power plants in the same regionGDescribing their spatial dependencies. The undirected graph of a plurality of photovoltaic power stations in a certain area is expressed as:
in the formula, each photovoltaic power station is regarded as one node in the undirected graphv i ，VA group of photovoltaic power plants representing a region,Nthe number of the regional photovoltaic power stations is represented,Erepresenting a set of relationships between each node (i.e., each photovoltaic power plant).
The connection feature between the nodes uses oneN×NMatrix of dimensionsAIs expressed, called adjacency matrix, which can reflect the whole graphIs a structural feature of (a). The invention adopts the absolute value of the Pearson correlation coefficient between nodes to represent the spatial correlation of adjacent photovoltaic power stations, and uses the correlation matrix between multiple photovoltaic power stations to represent the adjacent matrix. The historical power generation data of each photovoltaic power station is regarded as node attribute characteristics, so that one is constructedN×FFeature matrix of dimensionX，FRepresenting the historical time series length.
Taking multiple photovoltaic power plants as an example, the elapsed time is collectedt-hTo the time to be measuredtIs used for constructing an adjacency matrix according to historical power data of the power generationAFeature matrixXIs input into the GCN to obtain spatial characteristics of a plurality of adjacent photovoltaic power plants.
(3.1.2) region photovoltaic power station information aggregation, wherein the graph convolution neural network derives information of nodes to be detected by utilizing information of other nodes, the core idea is to aggregate by utilizing edge characteristics and node characteristics to generate new node representations, and the essential purpose is to extract spatial characteristics of a graph, namely, spatial characteristics of the power stations to be detected are obtained by aggregating spatial characteristic data of other photovoltaic power stations in the same region; the graph book integration is divided into two major categories, namely a spectrum-based method and a space-based method; the present invention models spatial features using a space-based GCN.
Adjacency matrix for multi-photovoltaic power stationAFeature matrixXInput into GCN model, GCN is implemented by pairingAAndXcoding is carried out, so that the spatial correlation between the photovoltaic power station to be tested and other peripheral power station units can be obtained; the operation formula between the graph convolution layers is expressed as:
in the method, in the process of the invention,His a feature of each layer, and for the input layer,Hnamely the feature matrixX；σA nonlinear activation function;is->A degree matrix of (2);Wis a learning weight.
(4) The advantage of nonlinear unsteady sequence processing by SGMD is utilized, irradiance, temperature, humidity, rainfall, regional photovoltaic output aggregation characteristics and photovoltaic output of a power station to be tested are decomposed into a group of relatively stable subsequences, error components with larger partial volatility are removed, a multi-layer modal subsequence characteristic matrix which can reflect time sequence change characteristics of historical data is constructed, and subsequence input matrixes are formed by corresponding subsequences of different input characteristics and photovoltaic output.
(4.1) irradiance and photovoltaic power signals are highly oscillatory and unstable signals whose volatility can make it difficult to predict such signals. The invention adopts a mode decomposition method to decompose the time sequence signal into a plurality of components, so as to generate a time sequence with more stability and smaller fluctuation. These temporal sub-sequence signals may improve the prediction accuracy.
Time series data of meteorological factors and photovoltaic power generation power are set asx=[x 1 ,x 2 ,……，x n ]WhereinnFor time series signal length, builddDimensional track matrixXThe following formula is shown:
in the method, in the process of the invention,τis a delay time.
According to the matrixAConstructing hamilton matrixM：
Order of principleC=M 2 Thereby constructing Xin Zhengjiao matrixQ：
In the method, in the process of the invention,Bto get up toTriangle matrix, calculating its characteristic value asλ 1 ,λ 2 ,……,λ d Matrix thenAThe characteristic values of (2) are:
in the method, in the process of the invention,σ i according to a matrixAIs characterized by the octyl geometric spectrum distribution,Q(i=1,2,……,d)from the following componentsσ i The reconstruction of each component matrix is as follows:
in the method, in the process of the invention,i=1,2,……,d. Then an initial single component trajectory matrixZThe following formula is shown:
in the method, in the process of the invention,Z∈R mⅹd . Definition of the definitionZ i Middle elementz ij 1 +.i +.d, 1 +.j +.m, andd * =min(m,d)，m * =max(m,d)，n=m+(d-1)τthen:
the diagonal average transition matrix is then obtained by:
matrix can be averaged by diagonalZConversion intod×nMatrix of dimensionsYThereby inputting variable signalsxIs decomposed intodXin Jihe modal components with different center frequencies:
in the method, in the process of the invention,Yis a primitive sequence;Y i （i=1,2,……,d) To decompose the subsequence.
(5) And performing feature extraction and photovoltaic power generation power prediction by adopting a CNN-BiLSTM neural network, obtaining a final prediction result through inverse normalization and subsequence reconstruction, and performing error evaluation on the final prediction result.
(5.1) in order to better mine environmental factors, periodicity of photovoltaic power generation and existing space-time characteristics, the invention adopts CNN-BiLSTM consisting of a convolutional neural network and a two-way long-short-term memory neural network layer as a characteristic extraction and power prediction model, and takes photovoltaic power generation power, meteorological characteristics and GCN space aggregation characteristics as inputs, so that the model can deeply mine correlation among input characteristics and variation trend of characteristic time sequences.
(5.1.1) the convolutional neural network model consists of three layers, an input layer, a hidden layer and an output layer. The hidden layer is a feature extraction layer and consists of a convolution layer and a pooling layer. To improve the feature extraction capability of the model, a plurality of convolution layers are designed. The mathematical expression is as follows:
in the method, in the process of the invention,α i is the firstiA plurality of convolutional layer input features;W i andb i and respectively represent the firstiA weight matrix and a paranoid vector of the convolution kernels of the convolution layers;f i is the firstiOutputting characteristics by the convolution layers; σ represents the ReLU activation function.
(5.1.2) photovoltaic power generation power as periodic data exhibits a great timing dependence. Therefore, the feature vector extracted by the CNN module is flattened by the flat layer and then is input into the BiLSTM neural network, so that the time sequence characteristic of the input feature is extracted, the forward propagation characteristic and the backward propagation characteristic are calculated, the prediction precision is further improved, and the mathematical expression is as follows:
in the method, in the process of the invention,n t is an input vector;for the forward propagation layer output value: />An output value that is a backward propagation layer;θ t is the output of the output layer;δactivating a function for Tanh; />、/>、/>、/>、/>And->Is a weight matrix; />、/>And->Is a bias vector.

Claims (10)

1. The short-term photovoltaic output prediction method based on space-time correlation is characterized by comprising the following steps of:
(1) Acquiring historical power of a plurality of photovoltaic power stations in a certain area and historical data of power station influence factors to be tested, establishing a characteristic database, removing abnormal values and filling missing values of original data, and normalizing all data;
(2) Selecting a weather factor with strong correlation by utilizing pearson correlation analysis, introducing historical power generation, and constructing a delay input characteristic;
(3) Establishing a GCN model based on a topological structure of the regional photovoltaic power station, transversely tracking the photovoltaic spatial evolution modes of the multi-photovoltaic power station, aggregating the spatial features of adjacent photovoltaic electric fields, and outputting a spatial feature set containing the spatial evolution modes;
(4) Utilizing SGMD to process the advantage of a nonlinear unstable sequence, decomposing irradiance, temperature, humidity, rainfall, regional photovoltaic output aggregation characteristics and photovoltaic output of a power station to be tested into a group of relatively stable subsequences, eliminating error components with larger partial volatility, constructing a multi-layer modal subsequence characteristic matrix which can reflect the time sequence change characteristics of historical data, and forming subsequence input matrixes by corresponding subsequences of different input characteristics and photovoltaic output;
(5) And performing feature extraction and photovoltaic power generation power prediction by adopting a CNN-BiLSTM neural network, obtaining a final prediction result through inverse normalization and subsequence reconstruction, and performing error evaluation on the final prediction result.
2. The method for predicting short-term photovoltaic output based on spatio-temporal correlation according to claim 1, wherein said step (1) specifically comprises:
the method comprises the steps of sorting historical power data of a plurality of photovoltaic power stations in a certain area and historical data of meteorological factors of the power stations to be tested; removing abnormal values and completing data of the sample set;
the method comprises the steps of (1.1) giving a confidence probability by using a Laida criterion method, determining a confidence limit, and regarding errors exceeding the confidence limit as outlier rejection to obtain historical power of a plurality of photovoltaic power stations and historical data of meteorological factors of the power stations to be tested, wherein the errors exceeding the confidence limit do not belong to a random error range;
(1.2) on the basis of eliminating abnormal values, filling missing data in historical output data and meteorological factor historical data respectively by adopting a bilinear interpolation method;
and (1.3) carrying out normalization processing on the cleaned historical output data and the meteorological factor data of the power station to be tested.
3. The method of claim 2, wherein the normalization in step (1.3) uses the following formula:
；
in the method, in the process of the invention,Xis the data before the normalization and is used for the data,X´is the corresponding normalized data of the data set,X min andX max is the minimum and maximum in the sequence.
4. The method for predicting short-term photovoltaic output based on spatio-temporal correlation according to claim 1, wherein said step (2) specifically comprises:
(2.1) analyzing the correlation between the photovoltaic output and the meteorological factors by using the Pearson correlation coefficient to determine a strong correlation meteorological factor, wherein the calculation formula is as follows:
；
in the method, in the process of the invention,P XY for the Pearson correlation coefficient,X、Yis a correlation factor;、/>respectively the average value of the two factors;
and (2.2) according to the selection result of the delay step length of the photovoltaic power, carrying out delay input characteristic expansion on the target photovoltaic power station, and introducing historical photovoltaic power generation power as an input characteristic.
5. The method for predicting short-term photovoltaic output based on spatio-temporal correlation according to claim 1, wherein said step (3) specifically comprises:
(3.1) when the output of a certain adjacent power station first fluctuates greatly, the GCN network forms the spatial characteristics of the power station to be tested by aggregating the output of the adjacent photovoltaic power station, rapidly captures the output fluctuation caused by meteorological mutation, and establishes the photovoltaic output connectivity in space;
(3.1.1) defining a graph structure dataGTo describe their spatial dependencies; the undirected graph of a plurality of photovoltaic power stations in a certain area is expressed as:
；
in the formula, each photovoltaic power station is regarded as one node in the undirected graphv i ，VA group of photovoltaic power plants representing a region,Nthe number of the regional photovoltaic power stations is represented,Erepresenting a set of relationships between each node;
the connection feature between the nodes uses oneN×NMatrix of dimensionsAIs referred to as an adjacency matrix; the absolute value of the Pearson correlation coefficient among the nodes is adopted to represent the spatial correlation of adjacent photovoltaic power stations, and the correlation matrix among the multiple photovoltaic power stations is adopted to represent the adjacent matrix; the historical power generation data of each photovoltaic power station is regarded as node attribute characteristics, so that one is constructedN×FFeature matrix of dimensionX，FRepresenting a historical time series length;
(3.1.2) obtaining the spatial characteristics of the power station to be tested by aggregating the spatial characteristic data of other photovoltaic power stations in the same area; the graph book integration is divided into two major categories, namely a spectrum-based method and a space-based method; modeling spatial features using a space-based approach, specifically:
adjacency matrix for multi-photovoltaic power stationAFeature matrixXInput into GCN model, GCN is implemented by pairingAAndXcoding to obtain the spatial correlation between the photovoltaic power station to be tested and other peripheral power station units; the operation formula between the graph convolution layers is expressed as:
；
in the method, in the process of the invention,His a feature of each layer, and for the input layer,Hnamely the feature matrixX；σA nonlinear activation function;is->A degree matrix of (2);Wis a learning weight.
6. The method for predicting short-term photovoltaic output based on spatio-temporal correlation according to claim 1, wherein said step (4) specifically comprises:
(4.1) decomposing the time series signal into a plurality of components by adopting a mode decomposition method to generate a more stable time series with smaller fluctuation;
time series data of meteorological factors and photovoltaic power generation power are set asx=[x 1 ,x 2 ,……，x n ]WhereinnFor time series signal length, builddDimensional track matrixXThe following formula is shown:
；
in the method, in the process of the invention,τis a delay time;
according to the matrixAConstructing hamilton matrixM：
；
Order of principleC=M 2 Thereby constructing Xin Zhengjiao matrixQ：
；
In the method, in the process of the invention,Bfor the upper triangular matrix, calculating the characteristic value asλ 1 ,λ 2 ,……,λ d Matrix thenAThe characteristic values of (2) are:
；
in the method, in the process of the invention,σ i according to a matrixAIs characterized by the octyl geometric spectrum distribution,Q(i=1,2,……,d)from the following componentsσ i The reconstruction of each component matrix is as follows:
；
in the method, in the process of the invention,i=1,2,……,dthe method comprises the steps of carrying out a first treatment on the surface of the Then an initial single component trajectory matrixZThe following formula is shown:
；
in the method, in the process of the invention,Z∈R mⅹd the method comprises the steps of carrying out a first treatment on the surface of the Definition of the definitionZ i Middle elementz ij 1 +.i +.d, 1 +.j +.m, andd * =min(m,d)，m * =max(m,d)，n=m+ (d-1)τthen:
；
the diagonal average conversion matrix is obtained by:
；
matrix can be averaged by diagonalZConversion intod×nMatrix of dimensionsYThereby inputting variable signalsxIs decomposed intodXin Jihe modal components with different center frequencies:
；
in the method, in the process of the invention,Yis a primitive sequence;Y i （i=1,2,……,d) To decompose the subsequence.
7. The method for predicting short-term photovoltaic output based on spatio-temporal correlation according to claim 1, wherein said step (5) specifically comprises:
(5.1) adopting CNN-BiLSTM consisting of a convolutional neural network and a two-way long-short-term memory neural network layer as a feature extraction and power prediction model, and taking photovoltaic power generation power, meteorological features and GCN space aggregation features as inputs;
(5.1.1) the convolutional neural network model consists of three layers, an input layer, a hidden layer and an output layer; the hidden layer is a characteristic extraction layer and consists of a convolution layer and a pooling layer;
(5.1.2) flattening the feature vector extracted by the CNN module by the flat layer, inputting the feature vector into the BiLSTM neural network, extracting the time sequence characteristic of the input feature, and calculating the forward propagation characteristic and the backward propagation characteristic, wherein the mathematical expression is as follows:
；
in the method, in the process of the invention,n t is an input vector;for the forward propagation layer output value: />An output value that is a backward propagation layer;θ t is the output of the output layer;δactivating a function for Tanh; />、/>、/>、/>、/>And->Is a weight matrix; />、/>And->Is a bias vector.
8. The method of claim 7, wherein in the step (5.1.1), the convolution layer is designed as a multi-layer convolution layer, and the mathematical expression is as follows:
；
in the method, in the process of the invention,α i is the firstiA plurality of convolutional layer input features;W i andb i and respectively represent the firstiA weight matrix and a paranoid vector of the convolution kernels of the convolution layers;f i is the firstiOutputting characteristics by the convolution layers;σrepresenting the ReLU activation function.
9. A computer storage medium having stored thereon a computer program which, when executed by a processor, implements a short-term photovoltaic output prediction method based on spatio-temporal correlation according to any of claims 1-8.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements a short-term photovoltaic output prediction method based on spatio-temporal correlation according to any of claims 1-8 when executing the computer program.
CN202310837787.6A 2023-07-10 2023-07-10 Short-term photovoltaic output prediction method based on space-time correlation Active CN116565863B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310837787.6A CN116565863B (en) 2023-07-10 2023-07-10 Short-term photovoltaic output prediction method based on space-time correlation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310837787.6A CN116565863B (en) 2023-07-10 2023-07-10 Short-term photovoltaic output prediction method based on space-time correlation

Publications (2)

Publication Number Publication Date
CN116565863A true CN116565863A (en) 2023-08-08
CN116565863B CN116565863B (en) 2023-09-26

Family

ID=87495151

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310837787.6A Active CN116565863B (en) 2023-07-10 2023-07-10 Short-term photovoltaic output prediction method based on space-time correlation

Country Status (1)

Country Link
CN (1) CN116565863B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117032385A (en) * 2023-10-09 2023-11-10 威胜能源技术股份有限公司 High-efficiency MPPT control method applied to BUCK topology
CN117134334A (en) * 2023-08-28 2023-11-28 淮阴工学院 Mechanism-data driving hybrid integration-based photovoltaic power generation power prediction method
CN117154704A (en) * 2023-08-29 2023-12-01 淮阴工学院 Photovoltaic power prediction method based on multiscale space-time diagram attention convolution network

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111260126A (en) * 2020-01-13 2020-06-09 燕山大学 Short-term photovoltaic power generation prediction method considering correlation degree of weather and meteorological factors
CN112598180A (en) * 2020-12-21 2021-04-02 北京华能新锐控制技术有限公司 Distributed regional wind power prediction method
CN114254556A (en) * 2021-11-17 2022-03-29 中国华能集团清洁能源技术研究院有限公司 Photovoltaic power generation power prediction method and device, electronic equipment and storage medium
CN114282646A (en) * 2021-11-29 2022-04-05 淮阴工学院 Light power prediction method and system based on two-stage feature extraction and improved BilSTM
CN115115125A (en) * 2022-07-13 2022-09-27 西安理工大学 Photovoltaic power interval probability prediction method based on deep learning fusion model

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111260126A (en) * 2020-01-13 2020-06-09 燕山大学 Short-term photovoltaic power generation prediction method considering correlation degree of weather and meteorological factors
CN112598180A (en) * 2020-12-21 2021-04-02 北京华能新锐控制技术有限公司 Distributed regional wind power prediction method
CN114254556A (en) * 2021-11-17 2022-03-29 中国华能集团清洁能源技术研究院有限公司 Photovoltaic power generation power prediction method and device, electronic equipment and storage medium
CN114282646A (en) * 2021-11-29 2022-04-05 淮阴工学院 Light power prediction method and system based on two-stage feature extraction and improved BilSTM
CN115115125A (en) * 2022-07-13 2022-09-27 西安理工大学 Photovoltaic power interval probability prediction method based on deep learning fusion model

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117134334A (en) * 2023-08-28 2023-11-28 淮阴工学院 Mechanism-data driving hybrid integration-based photovoltaic power generation power prediction method
CN117154704A (en) * 2023-08-29 2023-12-01 淮阴工学院 Photovoltaic power prediction method based on multiscale space-time diagram attention convolution network
CN117154704B (en) * 2023-08-29 2024-06-07 淮阴工学院 Photovoltaic power prediction method based on multiscale space-time diagram attention convolution network
CN117032385A (en) * 2023-10-09 2023-11-10 威胜能源技术股份有限公司 High-efficiency MPPT control method applied to BUCK topology
CN117032385B (en) * 2023-10-09 2023-12-22 威胜能源技术股份有限公司 High-efficiency MPPT control method applied to BUCK topology

Also Published As

Publication number Publication date
CN116565863B (en) 2023-09-26

Similar Documents

Publication Publication Date Title
Huang et al. Multiple-input deep convolutional neural network model for short-term photovoltaic power forecasting
CN116565863B (en) Short-term photovoltaic output prediction method based on space-time correlation
Wu et al. A novel hybrid system based on multi-objective optimization for wind speed forecasting
Tian et al. Multi-step short-term wind speed prediction based on integrated multi-model fusion
He et al. A hybrid system for short-term wind speed forecasting
CN111091233B (en) Short-term wind power prediction modeling method for wind power plant
Mehrmolaei et al. Time series forecasting using improved ARIMA
Bo et al. Wind speed forecasting system based on the variational mode decomposition strategy and immune selection multi-objective dragonfly optimization algorithm
Yang et al. Hybrid prediction method for wind speed combining ensemble empirical mode decomposition and Bayesian ridge regression
Li et al. An integrated missing-data tolerant model for probabilistic PV power generation forecasting
CN108717579B (en) Short-term wind power interval prediction method
Liang et al. A combined model based on CEEMDAN, permutation entropy, gated recurrent unit network, and an improved bat algorithm for wind speed forecasting
CN109412161B (en) Power system probability load flow calculation method and system
Yang et al. A novel deep learning approach for short and medium-term electrical load forecasting based on pooling LSTM-CNN model
Wang et al. Swarm Intelligence‐Based Hybrid Models for Short‐Term Power Load Prediction
Thanh et al. Short-term three-phase load prediction with advanced metering infrastructure data in smart solar microgrid based convolution neural network bidirectional gated recurrent unit
Kosana et al. Hybrid wind speed prediction framework using data pre-processing strategy based autoencoder network
Sun et al. Fast transient stability prediction using grid-informed temporal and topological embedding deep neural network
CN112418504B (en) Wind speed prediction method based on mixed variable selection optimization deep belief network
Zhou et al. A short-term power prediction method based on temporal convolutional network in virtual power plant photovoltaic system
Chen et al. Graph neural network-based wind farm cluster speed prediction
Wang et al. An optimized deep nonlinear integrated framework for wind speed forecasting and uncertainty analysis
CN116167465A (en) Solar irradiance prediction method based on multivariate time series ensemble learning
Wu et al. Combined IXGBoost-KELM short-term photovoltaic power prediction model based on multidimensional similar day clustering and dual decomposition
CN116384574A (en) W-LSTM-based non-parameter load interval prediction method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant