CN116090635A - Meteorological-driven new energy generation power prediction method - Google Patents
Meteorological-driven new energy generation power prediction method Download PDFInfo
- Publication number
- CN116090635A CN116090635A CN202310050073.0A CN202310050073A CN116090635A CN 116090635 A CN116090635 A CN 116090635A CN 202310050073 A CN202310050073 A CN 202310050073A CN 116090635 A CN116090635 A CN 116090635A
- Authority
- CN
- China
- Prior art keywords
- new energy
- meteorological
- power
- data
- power generation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 75
- 238000010248 power generation Methods 0.000 claims abstract description 75
- 238000005457 optimization Methods 0.000 claims abstract description 10
- 238000010219 correlation analysis Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 43
- 238000013528 artificial neural network Methods 0.000 claims description 22
- 230000008859 change Effects 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 11
- 238000011176 pooling Methods 0.000 claims description 8
- 238000010206 sensitivity analysis Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 230000002779 inactivation Effects 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 230000008878 coupling Effects 0.000 claims description 4
- 238000010168 coupling process Methods 0.000 claims description 4
- 238000005859 coupling reaction Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000005315 distribution function Methods 0.000 claims description 3
- BTCSSZJGUNDROE-UHFFFAOYSA-N gamma-aminobutyric acid Chemical compound NCCCC(O)=O BTCSSZJGUNDROE-UHFFFAOYSA-N 0.000 claims description 3
- 230000007787 long-term memory Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000004445 quantitative analysis Methods 0.000 claims description 3
- 230000006403 short-term memory Effects 0.000 claims description 3
- 230000003213 activating effect Effects 0.000 claims description 2
- 125000004122 cyclic group Chemical group 0.000 claims description 2
- 230000010354 integration Effects 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 5
- 238000012098 association analyses Methods 0.000 abstract description 4
- 230000009467 reduction Effects 0.000 abstract description 4
- 230000004913 activation Effects 0.000 description 9
- 238000012549 training Methods 0.000 description 7
- 238000013527 convolutional neural network Methods 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 5
- 229910052799 carbon Inorganic materials 0.000 description 4
- 210000004027 cell Anatomy 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 230000015654 memory Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000012097 association analysis method Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008033 biological extinction Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000006880 cross-coupling reaction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Images
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- General Engineering & Computer Science (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Operations Research (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Entrepreneurship & Innovation (AREA)
- Medical Informatics (AREA)
- Game Theory and Decision Science (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Public Health (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a new energy power generation power prediction method driven by weather, which comprises a new energy power generation power-weather factor correlation analysis method, a new energy power generation power core weather factor optimization identification method and a new energy power generation power prediction method driven by weather. The prediction accuracy of the generated power prediction model is high, and the method considers the generated characteristic difference generated by geographic and meteorological factors, and has more advantages compared with the traditional prediction method; the prediction efficiency is high, key meteorological factors can be optimally selected as input variables based on meteorological association analysis by taking dimension reduction measures, the complexity of a prediction model is reduced, and the prediction efficiency is improved; the method can be applied to the low-data new energy stations, and aims at solving the problem that part of new energy stations lack historical power generation data, comprehensively considers the spatial correlation and the data source of the supplementary prediction model of the meteorological correlation analysis, and improves the prediction accuracy of the power generation of the low-data new energy stations. Therefore, the method is suitable for popularization and application.
Description
Technical Field
The invention belongs to the technical field of new energy generated power prediction, and particularly relates to a new energy generated power prediction method driven by weather.
Background
Under the strategic background of the strong implementation of the 'double carbon policy', the traditional fossil energy taking coal, natural gas and petroleum as main bodies cannot meet the requirement of low-carbon energy development. The development of clean low-carbon energy is becoming a trend, and the development and utilization of energy green low-carbon transformation and clean energy are developed to achieve a harsher energy efficiency target. Different from the traditional thermal power generating unit, the power generation of the new energy unit is related to multidimensional meteorological factors such as temperature, wind speed, humidity and the like, and has volatility and intermittence. The large-scale access of the high-proportion clean energy can compress the operation space of the conventional unit, reduce the operation flexibility and inertia of the system, and pose great challenges to the stable balance, safe operation and economic benefit of the power system. Under the background, the new energy generated power needs to be predicted, and the high-precision power prediction can be used as a basis for multi-energy coordinated development while providing a reliable basis for the scheduling decision-making behavior of the power grid. However, the new energy generated power prediction model has the characteristics of multi-dimensionality, multi-scale, cross coupling, nonlinearity and the like, and the traditional prediction method does not consider the characteristics and has the disadvantages of low precision, low efficiency and the like.
According to the existing new energy power generation prediction method considering meteorological factors, the original measured multiple meteorological factors are directly input into a prediction model, the model structure is complex, so that the prediction model has weak generalization capability, the training efficiency is low, and key features cannot be extracted. In addition, because redundant information is input into the model without screening meteorological influence factors, the prediction model is influenced on the learning of important association characteristics, and the prediction accuracy is reduced. The influence condition and the action degree of the meteorological factors on the new energy generated power are different, and the meteorological factors have a coupling relation and have complex and various influence paths, so that a quantitative result is obtained on the degree of influence of the meteorological factors on the photovoltaic output, and further, the core meteorological influence factors are accurately identified and optimally selected, so that the basic analysis of the photovoltaic generated power prediction is necessary to be completed. However, the mainstream association analysis algorithm at the present stage is only suitable for processing linear association relation between bivariate, is difficult to process association characteristics between high-dimensional nonlinear data, and is not suitable for high-dimensional complex association identification between new energy power generation and meteorological factors such as wind speed, irradiance, temperature and the like.
The existing new energy power generation power prediction method only utilizes the meteorological data of historical monitoring and combines the related power generation scenes according to the establishment, however, the method depends on the accuracy of the established scenes, and the new energy power generation characteristic difference generated by geographic and climate factors is not considered. In recent years, new energy power generation in China is rapidly developed, a plurality of new energy stations are newly opened up, the power generation characteristics of the new energy stations cannot be incorporated into the traditional scene, and a certain deviation exists in the prediction result. Along with the improvement of the permeability of new energy, the deviation may cause the estimated deviation of the future power generation condition, so as to cause the mismatch of the power grid planning scheme and the access and delivery planning scheme.
Disclosure of Invention
The invention aims to provide a new energy power generation power prediction method driven by weather, which realizes the excavation of the correlation characteristic between the new energy power generation power and the weather factors and the construction of a new energy power generation power prediction model with the weather factors considered precisely, and improves the prediction precision of the new energy power generation power.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
a new energy generated power prediction method driven by weather comprises the following steps:
s1, analyzing the correlation characteristics of the new energy power generation and the meteorological factors;
s11, data integration is carried out on the input new energy power generation data and meteorological data, and a new energy power generation-meteorological factor data set is constructed;
s12, acquiring nonlinear correlations between each meteorological element and new energy power generation power in different periods by adopting a gray correlation analysis method;
s13, calculating the consistency of a new energy power generation curve and a meteorological curve in the dynamic change process, and obtaining the correlation characteristic of the correlation degree between the meteorological factors and the photovoltaic power generation power;
s2, optimizing and identifying core meteorological factors of the new energy generation power;
s21, taking the correlation characteristic of the new energy power generation power and the meteorological factors as the basis of the prediction model input variable identification optimization, and identifying the core meteorological influence factors of the power generation power of each new energy station;
s22, determining the dimension of an input variable according to the time period characteristic of the new energy generated power, and constructing an input variable set of a new energy generated power prediction model driven by weather based on the dimension of the input variable;
s3, new energy generated power prediction of meteorological driving is achieved;
s31, constructing a CNN-LSTM hybrid neural network;
s32, constructing a new energy power generation power prediction model driven by weather based on a CNN-LSTM algorithm according to the optimized identification of core weather factors;
s33, comprehensively considering the spatial correlation and the meteorological correlation characteristics to form a reference power station combination, and using the reference power station combination as a supplementary data source to realize the prediction of the power generation power of the new energy station with less data by using a new energy power generation prediction model driven by meteorological.
Further, in step S13, the correlation characteristic analysis step of the degree of correlation between the meteorological factors and the photovoltaic power generation power includes:
s13.1, defining new energy generation power as a reference sequence, a multi-element meteorological factor as a comparison sequence, and the difference between the new energy generation power sequence and the meteorological factor sequence is delta i (k) The method comprises the steps of carrying out a first treatment on the surface of the Wherein,,
Δ i (k)=|x 0 (k)-x i (k)|(i=0,1…,m,k=1,2,…n)
wherein n is the number of reference sequences, m is the number of comparison sequences, and delta i (k)=(Δ i (1),Δ i (2),…,Δ i (k) Essentially a poor sequence;
s13.2, defining the minimum value in each difference sequence as the minimum range min i min k Δ i (k) Definition of eachThe maximum value in the difference sequence is the maximum difference max i max k Δ i (k) The relation coefficient gamma of the ith meteorological factor and photovoltaic power generation power 0i (k) The calculation formula is as follows:
wherein, xi is the resolution coefficient, xi is E [0,1], and the value is 0.5;
s13.3, obtaining a correlation coefficient gamma 0i (k) Then, according to the formula:
obtaining gray correlation degree of each meteorological factorWherein n is the number of reference sequences;
s13.4, carrying out normalization processing on the gray correlation degree, and calculating a factor weight coefficient gamma which is the ith meteorological factor i :
Further, in step S13, for the weather elements with strong relevance, a kernel density estimation method is adopted to calculate the probability distribution between the change amount of the kernel element and the change amount of the new energy generated power, if the density function of the random variable X is F (X) =f (X), according to the kernel density estimation method, a simple F (X) estimate is obtainedThe calculation formula is as follows:
wherein h is window width, is a non-negative constant, and F (X) is an empirical distribution function of a random variable X;
let N total sample values resulting from the same unknown probability density: x is x 1 ,x 2 ,…x n When a non-negative constant h is selected, a kernel density estimation function is obtained by setting n to →++infinity, h to 0, and nh to++infinity:
in the method, in the process of the invention,represents the probability density function of the population, h is window width, N represents the total number of samples, K h Is a kernel function;
the kernel function is selected as a Gaussian kernel function, and the calculation formula is as follows:
the overall gaussian kernel density estimation function is thus as follows:
further, in step S13, the method further includes the steps of implementing a full state space fitting of historical new energy data based on a neural network algorithm, extracting new energy power generation power changes caused by core meteorological condition fluctuation in different periods, and implementing quantitative analysis of a 'new energy power generation-meteorological factor' association relationship under the influence of multiple variables; and (3) carrying out sensitivity analysis on the core meteorological factors on the basis of the fitting result of the full-state space, and obtaining the quantitative association relation between the multiple meteorological factors and the new energy generated power through the change trend and the amplitude of the new energy generated power when the core meteorological factors change and the influence condition of other meteorological factors.
Further, the specific steps for obtaining the quantitative association relation between the multi-element meteorological factors and the new energy generated power are as follows:
a: before new energy power fitting is realized by using a neural network, historical data with different dimensions are required to be normalized;
b: selecting normalized wind speed or irradiance data as a horizontal axis, and meshing meteorological data by taking core meteorological factors as a vertical axis;
c: obtaining the optimal combination of the neural network model parameters through a search traversal method;
d: inversely normalizing the output value of the fitting result;
and e, researching the quantitative relation of the coupling relation between meteorological variables on the influence of the new energy generated power by applying sensitivity analysis.
Further, in step S21, the core weather factor optimization recognition basis includes:
1) Redundancy of input information is reduced;
2) Reasonably selecting the input variable dimension of the prediction model;
3) And (5) pertinently selecting according to the time period characteristics of the new energy generated power.
Further, the new energy generated power prediction model in step S32 is a prediction model that inputs time-series data including multiple characteristic variables and outputs a single variable prediction result, and the information and related functions of each layer of the model are as follows:
s32.1, an input and convolution layer, which is used for distributing the input of the convolution unit to the next layer;
s32.2, activating a function layer, wherein the function layer is used for introducing nonlinear factors into a prediction model, improving the expression capability of data characteristics and enabling the neural network to better solve the nonlinear problem;
s32.3, a pooling layer is used for realizing the downsampling operation of the output vector of the convolution layer, which is equivalent to secondary feature extraction;
s32.4, a random inactivation layer and a flat layer, wherein the random inactivation layer is used for solving the over fitting problem in the deep neural network, and the flat layer is used for processing data into a format required by the LSTM layer;
s32.5, a long-term and short-term memory layer is used as a variant of the cyclic neural network to enable the LSTM unit to be at the moment
S32.6, a fully connected layer for receiving the output of each LSTM cell from the previous layer, the inputs being spread and connected to the output layer.
Compared with the prior art, the invention has the following beneficial effects:
the prediction accuracy of the power generation power prediction model is high, and the method considers the power generation characteristic difference generated by geographic and meteorological factors, and has more advantages compared with the traditional prediction method; the prediction efficiency is high, key meteorological factors can be optimally selected as input variables based on meteorological association analysis by taking dimension reduction measures, the complexity of a prediction model is reduced, and the prediction efficiency is improved; the method can be applied to the low-data new energy stations, and aims at solving the problem that part of new energy stations lack historical power generation data, comprehensively considers the spatial correlation and the data source of the supplementary prediction model of the meteorological correlation analysis, and improves the prediction accuracy of the power generation of the low-data new energy stations.
Drawings
FIG. 1 is a schematic flow chart of the method of the invention.
FIG. 2 is a schematic representation of a Gaussian kernel density estimation fit in an embodiment of the invention.
FIG. 3 is a schematic diagram of a CNN-LSTM prediction model according to an embodiment of the present invention.
Detailed Description
The invention will be further illustrated by the following description and examples, which include but are not limited to the following examples.
Examples
As shown in FIG. 1, the new energy power generation power prediction method driven by the weather disclosed by the invention comprises a new energy power generation power-weather factor correlation analysis method, a new energy power generation power core weather factor optimization identification method and a new energy power generation power prediction method driven by the weather.
The new energy power generation-meteorological factor association analysis method comprises the following steps: acquiring nonlinear correlations between each meteorological element and new energy power generation power in different periods by adopting a gray correlation analysis method; and obtaining the correlation characteristic of the correlation degree between the meteorological factors and the photovoltaic power generation power by calculating the consistency of the new energy power generation curve and the meteorological curve in the dynamic change process. The method specifically comprises the following steps:
1) Defining new energy power generation power as a reference sequence, a multi-element meteorological factor as a comparison sequence, and the difference between the new energy power generation power sequence and the meteorological factor sequence as delta i (k) The method comprises the steps of carrying out a first treatment on the surface of the Wherein,,
Δ i (k)=|x 0 (k)-x i (k)|(i=0,1…,m,k=1,2,…n)
wherein n is the number of reference sequences, m is the number of comparison sequences, and delta i (k)=(Δ i (1),Δ i (2),…,Δ i (k) Essentially a poor sequence.
2) Defining the minimum value in each difference sequence as the minimum polar difference min i min k Δ i (k) Defining the maximum value in each difference sequence as the maximum difference max i max k Δ i (k) The relation coefficient gamma of the ith meteorological factor and photovoltaic power generation power 0i (k) The calculation formula is as follows:
in the formula, xi is a resolution coefficient, and xi is E [0,1], and the value is 0.5.
3) Obtaining the correlation coefficient gamma 0i (k) Then, according to the formula:
obtaining gray correlation degree of each meteorological factorWherein n is the number of reference sequences.
4) Carrying out normalization processing on gray correlation degree, and calculating a factor weight coefficient gamma which is obtained as an ith meteorological factor i :
Aiming at weather elements with strong relevance, calculating probability distribution conditions between the change quantity of the core elements and the change quantity of the power generated by the new energy by adopting a nuclear density estimation method, and obtaining F (X) simple estimation according to the nuclear density estimation method if the density function of the random variable X is F (X) =F (X)The calculation formula is as follows:
wherein h is window width, is a non-negative constant, and F (X) is an empirical distribution function of a random variable X;
as shown in fig. 2, there are N total sample values generated from the same unknown probability density: x is x 1 ,x 2 ,…x n When a non-negative constant h is selected, a kernel density estimation function is obtained by setting n to →++infinity, h to 0, and nh to++infinity:
in the method, in the process of the invention,represents the probability density function of the population, h is window width, N represents the total number of samples, K h Is a kernel function;
the kernel function is selected as a Gaussian kernel function, and the calculation formula is as follows:
the overall gaussian kernel density estimation function is thus as follows:
when different confidence intervals, such as 50% and 90% confidence intervals, are selected, the non-parameter estimation models can be estimated through Gaussian kernel density, and the specific range of up-and-down fluctuation of the new energy power variation is observed more intuitively, so that the correlation characteristics of the new energy power and other meteorological factors are analyzed. Therefore, the non-parameter estimation model of Gaussian kernel density estimation can be used for realizing more effective fitting for weather data changing according to days, so that data loss caused by fitting by a probability distribution model is avoided.
The method is based on a neural network algorithm to realize the full-state space fitting of the historical data and eliminate the influence caused by the fluctuation of the data. And then extracting new energy power generation power change caused by core meteorological condition fluctuation in different periods, and realizing quantitative analysis of the 'new energy power generation power-meteorological factor' association relation under the influence of multiple variables. According to the analysis result of the action degree of the meteorological factors, wind speed, irradiance and core meteorological influence factors are selected as a horizontal axis and a vertical axis respectively, the meteorological data are presented in a gridding mode, and the boundary of the coordinate axis is the minimum value and the maximum value of the historical meteorological data respectively. The meshed weather data may cover all weather conditions that may occur in the analyzed power plant. And taking the meshed meteorological data as the input of a model in a matrix form, and outputting the model as a new energy generated power meshed fitting result under the corresponding meteorological conditions, so as to realize the full-state space fitting of the historical data. On the basis of the fitting result, sensitivity analysis is carried out on the core meteorological factors, and the quantitative association relation between the multiple meteorological factors and the new energy generated power is obtained through the change trend and amplitude of the new energy generated power when the core meteorological factors change and the influence condition of other meteorological factors. The specific implementation steps are as follows:
1) And (5) normalization treatment. The different variables in the raw weather and process data have different units, and the magnitude of the differences are typically large. The algorithm model is sensitive to a range of values, and an excessive range of values will negatively impact the training of the neural network algorithm model. Therefore, before the new energy power fitting is realized by using the neural network, the historical data with different dimensions are required to be normalized, so that the negative influence of the data with different dimensions on the fitting is avoided. The historical data has been scaled to within the range of [0,1] in the step of data preprocessing.
2) And (5) meshing meteorological data. And (3) selecting the normalized wind speed or irradiance data as a horizontal axis, taking a core meteorological factor as a vertical axis, and setting grid precision to be 0.05 to obtain a meteorological condition matrix of 20 rows by 20 columns as the input of the model.
3) And setting model parameters. And obtaining the optimal combination of the neural network model parameters through a search traversal method. The optimization algorithm for model training was selected from Adam algorithm, the training batch size was 128, and the total number of training of the model was set to 200 here.
4) And (5) inversely normalizing the fitting result. In order to obtain the fitting result of the normal dimension, the output value in the range of [0,1] needs to be inversely normalized, and the formula is an inversely normalized calculation formula. Fitting results are displayed in a thermodynamic diagram mode, and the new energy power generation condition under the meteorological conditions can be intuitively seen from the color shade of each grid.
5) Sensitivity analysis. And (3) researching the quantitative relation of the coupling relation between meteorological variables on the influence of the new energy generated power by applying sensitivity analysis. And selecting a meteorological factor to be analyzed, keeping the fixed irradiance value unchanged, and calculating the difference value of the new energy power generation power when the meteorological factor unit changes. The sensitivity analysis result is displayed in a thermodynamic diagram mode, and the grid depth represents the change trend and the change amplitude of the new energy power generation power when the meteorological factor unit changes under the meteorological condition.
The new energy power generation core meteorological factor optimizing and identifying method comprises the following steps: and taking the association characteristic of the new energy power generation power and the meteorological factors as the basis of the prediction model input variable identification optimization, identifying the core meteorological influence factors of the power generation power of each new energy station, determining the dimension of the input variable according to the time period characteristic of the new energy power generation power, and constructing the input variable set of the new energy power generation power prediction model driven by the meteorological based on the dimension. The core meteorological factor optimization and identification is based on the following:
1) Redundancy of input information is reduced. The unreasonable meteorological factors may be selected to input excessive redundant information for the model, so that the complexity of model calculation is improved.
2) And reasonably selecting the input variable dimension of the prediction model. If the meteorological factors input into the prediction model are too many, the structure of the prediction model becomes complex, and the sparse distribution of the limited input data in the high-dimensional space can lead to the failure of the prediction model to be effectively trained.
3) And (5) pertinently selecting according to the time period characteristics of the new energy generated power. The new energy power generation power and the associated characteristics of meteorological factors in different time periods have great difference, and when the input variable is selected, the high water period and the dead water period need to be considered separately.
The new energy generated power prediction method driven by weather comprises the following steps: the advantages of the Convolutional Neural Network (CNN) feature extraction and the long-short-term memory neural network (LSTM) can be combined to effectively process time sequence data, so that a CNN-LSTM hybrid neural network is formed. According to the optimization and identification of the core meteorological factors, the dimension reduction measure of the input variables of the prediction model is realized, and then the new energy power generation prediction model driven by the meteorological and based on the CNN-LSTM algorithm is constructed. For a new energy station with less data, comprehensively considering spatial correlation and meteorological correlation characteristics to form a reference power station combination, and using the reference power station combination as a supplementary data source to realize the power generation prediction of the new energy station with less data by using a new energy power generation prediction model driven by meteorological. The CNN-LSTM prediction model constructed by the method is characterized in that time series data containing multiple characteristic variables is input, a single variable prediction result is output, and the information and related functions of each layer of the model are as follows:
(1) Input and convolution layers. The purpose of the input layer is to distribute the input of the convolution unit to the next layer, accepting one data sample at a time, each data sample having n features including historical new energy generation power, irradiance, wind speed, temperature, etc. CNNs are adapted to process grid-like data, including multivariate time series data sets for feature extraction, whereby the data sets represent a one-dimensional grid of data samples acquired at equidistant time steps. Because CNN is convolutional neural network, layer unit adopts convolution operation to replace traditional multiplication operation.
Where, represents a convolution operation, I is an input with dimension M, and K is a convolution kernel with dimension M. In the predictive model constructed herein, M is equal to the number of input meteorological factor variables, characteristic k 1 To k M And respectively corresponding to the weather factors selected and input.
The layer contains a number of convolution kernels of size 1, step size 1 and no padding. The kernel weight initialization is to use a gloort unified initializer to prevent the oversaturation of the activation function:
where U [ -i, i ] is a uniform distribution within the interval [ -i, i ], and S (=M) is the size of the previous layer. The bias value is initialized to 0.
(2) The function is activated. The function of the activation function is to introduce nonlinear factors into the prediction model, so that the expression capacity of the data characteristics is improved, and the neural network can better solve the nonlinear problem. Common activation functions include sigmoid functions, tanh functions, reLU functions, and the like. The ReLU function is selected as the activation function, because the calculation speed and the convergence speed are faster than those of the sigmoid function and the tanh function, and the specific expression is as follows:
a l(i,j) =f(y l(i,j) )=max{a l(i,j) }
wherein l is the characteristic matrix of the convolution network, y l(i,j ) For the output value of the convolution layer, a l(i,j) Is an activation value.
(3) And (5) pooling the layers. The pooling layer realizes the downsampling operation on the output vector of the convolution layer and can be equivalently used for secondary feature extraction. The pooling layer selects maximum value pooling, takes the maximum value in the neuron as output so as to effectively reduce network parameters, and the calculation formula is as follows:
wherein a is l(i,j) For the activation value, p l(i,j) For pooling results, W is the width of the pooling window.
(4) Random inactivation (Dropout) layer and flame layer. The Dropout layer is used for solving the common overfitting problem in the deep neural network, and the principle is that the units are temporarily discarded from the network according to a certain probability, so that the activation value is discarded according to a certain proportion in the process of network training, and the generalization performance of the neural network is improved. The flat layer is used to handle the format required by the LSTM layer.
(5) A long-term and short-term memory layer.
LSTM as a variant of a recurrent neural network, whose units remain memorized at time tAnd outputs the hidden state of the loop to solve the problem of gradient extinction.
In the method, in the process of the invention,the LSTM network output obtained by the dynamic control of the memory unit is obtained by the following formula:
wherein sigma is an s-shaped function, x t 、W o 、U o And V o The input vector, weight matrix and diagonal matrix of the LSTM cell, respectively. The memory unit passes through the state information of the current momentAnd the memory information of the previous moment is calculated by the formula>
In the method, in the process of the invention,and->Is the activation vector of the forget gate and the input gate, respectively.
(6) And (5) a full connection layer. The only neurons in this layer receive the output from each LSTM cell of the previous layer, with the inputs unwrapped and connected by way to the output layer:
o=f(Wx+b)
where f, W, x and b are the activation function, weight, input and bias vectors, respectively.
As shown in fig. 3, according to the CNN-LSTM prediction model structure, the input layer time step is set to 12, i.e., data 1 hour after prediction by data of the previous 12 hours. In general, for simple mappings between input variables and output variables, a neural network using a single hidden layer is sufficient for description. Along with the increase of the number of hidden layers and the number of neurons in each hidden layer, the fitting capacity of the model to the nonlinear function is enhanced, a more complex mapping relation can be extracted, but the complexity of the model is increased, and the training efficiency is reduced. The embodiment is provided with a one-dimensional convolution layer, and the layer has 128 convolution units; an LSTM layer, which has 100 cells.
Through the design, the prediction accuracy of the power generation power prediction model is high, and the method considers the power generation characteristic difference generated by geographic and meteorological factors, and has more advantages compared with the traditional prediction method; the prediction efficiency is high, key meteorological factors can be optimally selected as input variables based on meteorological association analysis by taking dimension reduction measures, the complexity of a prediction model is reduced, and the prediction efficiency is improved; the method can be applied to the low-data new energy stations, and aims at solving the problem that part of new energy stations lack historical power generation data, comprehensively considers the spatial correlation and the data source of the supplementary prediction model of the meteorological correlation analysis, and improves the prediction accuracy of the power generation of the low-data new energy stations.
The above embodiment is only one of the preferred embodiments of the present invention, and should not be used to limit the scope of the present invention, but all the insubstantial modifications or color changes made in the main design concept and spirit of the present invention are still consistent with the present invention, and all the technical problems to be solved are included in the scope of the present invention.
Claims (7)
1. A meteorological-driven new energy generation power prediction method is characterized by comprising the following steps:
s1, analyzing the correlation characteristics of the new energy power generation and the meteorological factors;
s11, data integration is carried out on the input new energy power generation data and meteorological data, and a new energy power generation-meteorological factor data set is constructed;
s12, acquiring nonlinear correlations between each meteorological element and new energy power generation power in different periods by adopting a gray correlation analysis method;
s13, calculating the consistency of a new energy power generation curve and a meteorological curve in the dynamic change process, and obtaining the correlation characteristic of the correlation degree between the meteorological factors and the photovoltaic power generation power;
s2, optimizing and identifying core meteorological factors of the new energy generation power;
s21, taking the correlation characteristic of the new energy power generation power and the meteorological factors as the basis of the prediction model input variable identification optimization, and identifying the core meteorological influence factors of the power generation power of each new energy station;
s22, determining the dimension of an input variable according to the time period characteristic of the new energy generated power, and constructing an input variable set of a new energy generated power prediction model driven by weather based on the dimension of the input variable;
s3, new energy generated power prediction of meteorological driving is achieved;
s31, constructing a CNN-LSTM hybrid neural network;
s32, constructing a new energy power generation power prediction model driven by weather based on a CNN-LSTM algorithm according to the optimized identification of core weather factors;
s33, comprehensively considering the spatial correlation and the meteorological correlation characteristics to form a reference power station combination, and using the reference power station combination as a supplementary data source to realize the prediction of the power generation power of the new energy station with less data by using a new energy power generation prediction model driven by meteorological.
2. The method for predicting the generated power of a new energy source driven by a meteorological according to claim 1, wherein in step S13, the correlation characteristic analysis step of the degree of correlation between the meteorological factors and the generated power of the photovoltaic comprises:
s13.1, defining new energy generation power as a reference sequence, a multi-element meteorological factor as a comparison sequence, and the difference between the new energy generation power sequence and the meteorological factor sequence is delta i (k) The method comprises the steps of carrying out a first treatment on the surface of the Wherein,,
Δ i (k)=|x 0 (k)-x i (k)|(i=0,1…,m,k=1,2,…n)
wherein n is the number of reference sequences, m is the number of comparison sequences, and delta i (k)=(Δ i (1),Δ i (2),…,Δ i (k) A sequence that is essentially poor;
s13.2, defining the minimum value in each difference sequence as the minimum range min i min k Δ i (k) Defining the maximum value in each difference sequence as the maximum difference max i max k Δ i (k) The relation coefficient gamma of the ith meteorological factor and photovoltaic power generation power 0i (k) The calculation formula is as follows:
wherein, xi is the resolution coefficient, xi is E [0,1], and the value is 0.5;
s13.3, obtaining a correlation coefficient gamma 0i (k) Then, according to the formula:
obtaining gray correlation degree of each meteorological factorWherein n is the number of reference sequences;
s13.4, carrying out normalization processing on the gray correlation degree, and calculating a factor weight coefficient gamma which is the ith meteorological factor i :
3. The method for predicting the power generated by new energy driven by meteorological according to claim 2, wherein in step S13, for meteorological elements with strong relevance, a kernel density estimation method is adopted to calculate probability distribution conditions between the change amount of the kernel element and the change amount of the power generated by new energy, if the density function of the random variable X is F (X) =f (X), F (X) is obtained according to the kernel density estimation method, and F (X) is simply estimatedThe calculation formula is as follows:
wherein h is window width, is a non-negative constant, and F (X) is an empirical distribution function of a random variable X;
let N total sample values resulting from the same unknown probability density: x is x 1 ,x 2 ,…x n When a non-negative constant h is selected, a kernel density estimation function is obtained by setting n to →++infinity, h to 0, and nh to++infinity:
in the method, in the process of the invention,represents the probability density function of the population, h is window width, N represents the total number of samples, K h Is a kernel function;
the kernel function is selected as a Gaussian kernel function, and the calculation formula is as follows:
the overall gaussian kernel density estimation function is thus as follows:
4. the method for predicting the new energy generation power driven by meteorological according to claim 1, wherein in step S13, the method further comprises the steps of realizing full-state space fitting of historical new energy data based on a neural network algorithm, extracting new energy generation power change caused by core meteorological condition fluctuation in different periods, and realizing quantitative analysis of 'new energy generation power-meteorological factor' association under the influence of multiple variables; and (3) carrying out sensitivity analysis on the core meteorological factors on the basis of the fitting result of the full-state space, and obtaining the quantitative association relation between the multiple meteorological factors and the new energy generated power through the change trend and the amplitude of the new energy generated power when the core meteorological factors change and the influence condition of other meteorological factors.
5. The meteorological-driven new energy generation power prediction method according to claim 4, wherein the specific step of obtaining the quantitative association relation between the multiple meteorological factors and the new energy generation power is as follows:
a: before new energy power fitting is realized by using a neural network, historical data with different dimensions are required to be normalized;
b: selecting normalized wind speed or irradiance data as a horizontal axis, and meshing meteorological data by taking core meteorological factors as a vertical axis;
c: obtaining the optimal combination of the neural network model parameters through a search traversal method;
d: inversely normalizing the output value of the fitting result;
and e, researching the quantitative relation of the coupling relation between meteorological variables on the influence of the new energy generated power by applying sensitivity analysis.
6. The method for predicting the generated power of a new energy source driven by meteorological according to claim 1, wherein in step S21, the core meteorological factor optimization and identification basis comprises:
1) Redundancy of input information is reduced;
2) Reasonably selecting the input variable dimension of the prediction model;
3) And (5) pertinently selecting according to the time period characteristics of the new energy generated power.
7. The method for predicting the power of new energy generation driven by meteorological according to claim 6, wherein the model for predicting the power of new energy generation in step S32 is a model for predicting the power of new energy generation by input of time-series data including multiple characteristic variables and output of single variable prediction results, and the information and related functions of each layer of model are as follows:
s32.1, an input and convolution layer, which is used for distributing the input of the convolution unit to the next layer;
s32.2, activating a function layer, wherein the function layer is used for introducing nonlinear factors into a prediction model, improving the expression capability of data characteristics and enabling the neural network to better solve the nonlinear problem;
s32.3, a pooling layer is used for realizing the downsampling operation of the output vector of the convolution layer, which is equivalent to secondary feature extraction;
s32.4, a random inactivation layer and a flat layer, wherein the random inactivation layer is used for solving the over fitting problem in the deep neural network, and the flat layer is used for processing data into a format required by the LSTM layer;
s32.5, a long-term and short-term memory layer is used as a variant of the cyclic neural network to enable the LSTM unit to be at the moment
S32.6, a fully connected layer for receiving the output of each LSTM cell from the previous layer, the inputs being spread and connected to the output layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310050073.0A CN116090635A (en) | 2023-02-01 | 2023-02-01 | Meteorological-driven new energy generation power prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310050073.0A CN116090635A (en) | 2023-02-01 | 2023-02-01 | Meteorological-driven new energy generation power prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116090635A true CN116090635A (en) | 2023-05-09 |
Family
ID=86213593
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310050073.0A Pending CN116090635A (en) | 2023-02-01 | 2023-02-01 | Meteorological-driven new energy generation power prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116090635A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116885703A (en) * | 2023-06-27 | 2023-10-13 | 大连理工大学 | Short-term wind-solar power prediction method for high-dimensional multi-element meteorological data fusion |
-
2023
- 2023-02-01 CN CN202310050073.0A patent/CN116090635A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116885703A (en) * | 2023-06-27 | 2023-10-13 | 大连理工大学 | Short-term wind-solar power prediction method for high-dimensional multi-element meteorological data fusion |
CN116885703B (en) * | 2023-06-27 | 2024-02-06 | 大连理工大学 | Short-term wind-solar power prediction method for high-dimensional multi-element meteorological data fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
Yin et al. | Multi-temporal-spatial-scale temporal convolution network for short-term load forecasting of power systems | |
CN110705743B (en) | New energy consumption electric quantity prediction method based on long-term and short-term memory neural network | |
CN112116144B (en) | Regional power distribution network short-term load prediction method | |
CN109948845A (en) | A kind of distribution network load shot and long term Memory Neural Networks prediction technique | |
Liu et al. | Heating load forecasting for combined heat and power plants via strand-based LSTM | |
CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
CN112381673B (en) | Park electricity utilization information analysis method and device based on digital twin | |
CN114462718A (en) | CNN-GRU wind power prediction method based on time sliding window | |
CN115759415A (en) | Power consumption demand prediction method based on LSTM-SVR | |
CN114862032B (en) | XGBoost-LSTM-based power grid load prediction method and device | |
CN116345555A (en) | CNN-ISCA-LSTM model-based short-term photovoltaic power generation power prediction method | |
CN115409369A (en) | Comprehensive energy system reliability evaluation method based on mechanism and data hybrid driving | |
CN115640901A (en) | Small sample load prediction method based on hybrid neural network and generation countermeasure | |
CN116090635A (en) | Meteorological-driven new energy generation power prediction method | |
CN116384583A (en) | Photovoltaic power prediction method based on multiple neural networks | |
CN115238854A (en) | Short-term load prediction method based on TCN-LSTM-AM | |
CN115049115A (en) | RDPG wind speed correction method considering NWP wind speed transverse and longitudinal errors | |
CN114091776A (en) | K-means-based multi-branch AGCNN short-term power load prediction method | |
CN117132132A (en) | Photovoltaic power generation power prediction method based on meteorological data | |
CN117151770A (en) | Attention mechanism-based LSTM carbon price prediction method and system | |
CN116885699A (en) | Power load prediction method based on dual-attention mechanism | |
CN110659775A (en) | LSTM-based improved electric power short-time load prediction algorithm | |
CN116611702A (en) | Integrated learning photovoltaic power generation prediction method for building integrated energy management | |
CN115481788A (en) | Load prediction method and system for phase change energy storage system |
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 |