CN116011608A - Photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium - Google Patents
Photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN116011608A CN116011608A CN202211130291.7A CN202211130291A CN116011608A CN 116011608 A CN116011608 A CN 116011608A CN 202211130291 A CN202211130291 A CN 202211130291A CN 116011608 A CN116011608 A CN 116011608A
- Authority
- CN
- China
- Prior art keywords
- data
- photovoltaic power
- historical
- power generation
- power
- 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 45
- 238000003860 storage Methods 0.000 title claims abstract description 8
- 238000010248 power generation Methods 0.000 claims abstract description 78
- 238000012216 screening Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000010586 diagram Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 230000002123 temporal effect Effects 0.000 claims description 5
- 230000002596 correlated effect Effects 0.000 claims description 4
- 238000010219 correlation analysis Methods 0.000 claims description 4
- 230000000295 complement effect Effects 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000053 physical method Methods 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000000547 structure data Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- 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
- Supply And Distribution Of Alternating Current (AREA)
- Photovoltaic Devices (AREA)
Abstract
The invention discloses a photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium, comprising the following steps of S1: acquiring historical power generation power data and historical meteorological data of each photovoltaic power station in an area; s2: preprocessing the historical power generation data and the historical meteorological data; s3: feature screening is carried out on the preprocessed historical meteorological data, so that meteorological variables related to photovoltaic power generation power are obtained; s4: according to the meteorological variable and the preprocessed historical power generation data, a sample set is constructed and obtained, wherein the sample set is data in three dimensions of time, space and characteristics; s5: constructing a space-time feature map convolution network, constructing an ultra-short-term photovoltaic power prediction model based on the space-time feature map convolution network, and training the ultra-short-term photovoltaic power prediction model by using the sample set; s6: and inputting the sample set to be predicted into a trained ultra-short-term photovoltaic power prediction model, and outputting a prediction result. The invention can fully excavate the space-time correlation of each photovoltaic power station and realize accurate power prediction of multi-station cooperation.
Description
Technical Field
The present disclosure relates to the field of photovoltaic power prediction technologies, and in particular, to a photovoltaic power ultra-short term prediction method and apparatus, an electronic device, and a storage medium.
Background
Because solar energy resources in nature have intermittent, random and wave characteristics, photovoltaic power stations have difficulty in continuously and stably providing energy output as in conventional thermal power plants. Large-scale photovoltaic grid connection can impact a power grid, so that the safe and stable operation of a power system is affected, and the difficulty of optimal scheduling of the power grid is increased. The power grid is unwilling to accept large-scale photovoltaic, the light rejection phenomenon occurs, and the development space of photovoltaic power generation is compressed. Therefore, photovoltaic power prediction is an important link to construct an electrical power system containing a high proportion of renewable energy sources and to ensure its stable operation.
According to the difference of the prediction time ranges, the prediction of the photovoltaic power generation power can be divided into: ultra-short term power prediction, medium-long term power prediction, and long term power prediction. According to the specification in the national power grid 'photovoltaic power generation power prediction system function specification', the ultra-short-term power prediction can predict the output power of 15min to 4h in the future, and the time resolution is 15min. The ultra-short-term prediction can be used for optimizing and configuring the rotary spare capacity of the system, power grid dispatching, power quality evaluation, generator set control and the like, and has wide application range.
The current modeling method for photovoltaic power prediction can be divided into a physical method, a statistical method and an artificial intelligence method. With the rapid development of artificial intelligence, the emerging artificial intelligence method can better fit a nonlinear relation between input and output variables, and the prediction precision is further improved. In the field of photovoltaic power prediction, a plurality of methods are currently applied, such as a wavelet analysis method, a support vector machine, a neural network and the like.
The power generated by the photovoltaic power station has strong correlation with meteorological conditions and atmospheric movement. Since the atmospheric movement is spatially continuous, the weather conditions in the same region are often closely related. For a plurality of photovoltaic power stations in the same area, the generated power has not only continuity in time but also relevance in space. Therefore, if the space-time characteristics of the photovoltaic are fully utilized to conduct power prediction, prediction errors can be effectively reduced, and the method has important significance for combined power prediction of photovoltaic power stations in the same area. However, the conventional machine learning method is not good at processing non-European structure data containing spatial information, and it is difficult to fully learn spatial features of data while learning temporal features of the data, and prediction accuracy is difficult to be greatly improved.
Disclosure of Invention
The embodiment of the application aims to provide a photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium, so as to solve the technical problem of low prediction precision in the related technology.
According to a first aspect of embodiments of the present application, there is provided a photovoltaic power ultra-short term prediction method, including:
acquiring historical power generation power data and historical meteorological data of each photovoltaic power station in an area;
preprocessing the historical power generation data and the historical meteorological data;
feature screening is carried out on the preprocessed historical meteorological data, so that meteorological variables related to photovoltaic power generation power are obtained;
according to the meteorological variable and the preprocessed historical power generation data, a sample set is constructed and obtained, wherein the sample set is data in three dimensions of time, space and characteristics;
constructing a space-time feature map convolution network, constructing an ultra-short-term photovoltaic power prediction model based on the space-time feature map convolution network, and training the ultra-short-term photovoltaic power prediction model by using the sample set;
and inputting the sample set to be predicted into a trained ultra-short-term photovoltaic power prediction model, and outputting a prediction result.
Optionally, preprocessing the historical generated power data and the historical meteorological data includes:
the average value of the historical power generation data and the data of the time before and after the historical meteorological data is utilized to complement the missing value in the data;
and normalizing the completed historical power generation data and the historical meteorological data.
Optionally, normalizing the completed historical power data and the historical meteorological data, including:
aiming at the completed historical power generation data, taking the maximum value of the power generation power of a certain photovoltaic power station as the maximum value, taking the minimum value of the power generation power of a certain photovoltaic power station as the minimum value, and carrying out normalization processing on the completed historical power generation data by utilizing the maximum value and the minimum value;
and aiming at the completed historical meteorological data, taking the maximum value of all the meteorological variables of the photovoltaic power station to be tested as the maximum value, taking the minimum value of all the meteorological variables of the photovoltaic power station to be tested as the minimum value, and carrying out normalization processing on the completed historical meteorological data by utilizing the maximum value and the minimum value.
Optionally, feature screening is performed on the preprocessed historical meteorological data to obtain meteorological variables associated with photovoltaic power generation power, including:
and adopting a gray correlation analysis method to perform characteristic screening on the preprocessed historical meteorological data to obtain meteorological variables correlated with the photovoltaic power generation power.
Optionally, according to the preprocessed meteorological variable and the preprocessed historical power data, a sample set is constructed, including:
defining the sample at time k as a time slice X k I.e. X k ∈R n×v N is the number of photovoltaic power stations, v is the characteristic number contained in each photovoltaic power station, namely the sum of the historical power and the types of meteorological variables;
x at successive c moments k Stacked chronologically into a three-dimensional tensor:
U k ＝[X k-c+1 ,X k-c+2 ,...,X k ]∈R n×v×c
the three dimensions are the spatial, characteristic and temporal dimensions, respectively, and the three-dimensional tensor can be used as a sample set of the model.
Optionally, constructing a spatio-temporal feature graph convolution network includes:
constructing a space-time feature graph convolution network, wherein the space-time feature graph convolution network consists of space-time feature graph convolution network layers, and the transfer function of a first layer of the network is as follows:
wherein H is (0) ＝U k Namely, the input of the whole network;
and->
Respectively, a time parameter, a space parameter, a characteristic parameter and a bias parameter, c, n and v are respectively a time dimension number, a space dimension number and a characteristic dimension number, +.>
The input matrix and the output matrix are represented,
representing tensor mixture multiplier,>
the sub-matrices representing tensors X and Y decomposed in 2-dimensional correspond to matrix multiplication,
sub-matrices representing tensors X and Y decomposed in 1, 3-dimension are respectively and correspondingly multiplied by matrix, and sigma (& gt) representsActivating a function;
preferably, n is taken l ＝c l ＝1，
Become->
/>
Optionally, constructing an ultrashort-term photovoltaic power prediction model based on the spatio-temporal feature graph convolution network includes:
and according to the space-time characteristic diagram convolution network, a multi-step prediction model of the photovoltaic power generation power is established by adopting a multi-input multi-output strategy, and the model can output the power generation power predicted values of n photovoltaic power stations at a plurality of moments in the future.
According to a second aspect of embodiments of the present application, there is provided a photovoltaic power ultra-short term prediction apparatus, including:
the acquisition module is used for acquiring historical power generation power data and historical meteorological data of each photovoltaic power station in the area;
the preprocessing module is used for preprocessing the historical power generation data and the historical meteorological data;
the screening module is used for carrying out characteristic screening on the preprocessed historical meteorological data to obtain meteorological variables related to the photovoltaic power generation power;
the sample set construction module is used for constructing and obtaining a sample set according to the meteorological variable and the preprocessed historical power generation data, wherein the sample set is data of three dimensions of time, space and characteristics;
the model construction and training module is used for constructing a space-time characteristic graph convolution network layer, constructing an ultra-short-term photovoltaic power prediction model based on the space-time characteristic graph convolution network layer, and training the ultra-short-term photovoltaic power prediction model by using the sample set;
and the prediction module is used for inputting the sample set to be predicted into the trained ultra-short-term photovoltaic power prediction model and outputting a prediction result.
According to a third aspect of embodiments of the present application, there is provided an electronic device, including:
one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of the first aspect.
According to a third aspect of embodiments of the present application, there is provided a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the steps of the method according to the first aspect.
The technical scheme provided by the embodiment of the application can comprise the following beneficial effects:
according to the embodiment, the photovoltaic power ultra-short-term prediction method can simultaneously predict the photovoltaic power for a plurality of photovoltaic power stations in the area, wherein the non-European data structure can be effectively processed by utilizing the strong characteristic extraction and fusion capability of the graph rolling network, so that the space-time relevance of each photovoltaic power station is fully excavated, and the multi-station collaborative accurate power prediction is realized. The method provided by the invention adopts multi-step prediction of a multi-input multi-output strategy, can output power prediction data at a plurality of moments in the future at one time, meets the requirement of ultra-short-term power prediction, and has higher practicability compared with single-step prediction.
Aiming at the photovoltaic ultra-short-term power prediction task, the method provided by the invention further improves the time characteristic graph convolution network model, changes the structure of the characteristic parameter matrix, greatly reduces the number of model parameters, reduces the possibility of the occurrence of the phenomenon of excessive fitting, and enables the model to be lighter. The invention can provide new ideas and references for photovoltaic power prediction and application of the graph neural network in the power system.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.
FIG. 1 is a flow chart illustrating a method of photovoltaic power ultra-short term prediction according to an exemplary embodiment.
FIG. 2 is a schematic diagram of a spatio-temporal signature convolution network hybrid multiplier operation, according to an exemplary embodiment.
FIG. 3 is a schematic diagram illustrating an employed multiple-input multiple-output strategy according to an exemplary embodiment.
FIG. 4 (a) is a schematic diagram of a multi-step predictive model first-layer network based on a spatio-temporal feature map convolution network, according to an example embodiment.
FIG. 4 (b) is a schematic diagram of a multi-step predictive model second-level network based on a spatio-temporal feature map convolution network, according to an example embodiment.
Fig. 5 (a) is a timing chart showing the ultra-short term generated power prediction result of the photovoltaic power plant No. 1 according to an exemplary embodiment.
Fig. 5 (b) is a timing chart showing the ultra-short term generated power prediction result of the No. 2 photovoltaic power plant according to an exemplary embodiment.
Fig. 5 (c) is a timing chart showing the ultra-short term generated power prediction result of the No. 3 photovoltaic power plant according to an exemplary embodiment.
Fig. 5 (d) is a timing chart showing the ultra-short term generated power prediction result of the photovoltaic power plant No. 4 according to an exemplary embodiment.
Fig. 6 is a block diagram illustrating a photovoltaic power ultra-short term prediction apparatus according to an example embodiment.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application as detailed in the accompanying claims.
The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. As used in this application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any or all possible combinations of one or more of the associated listed items.
It should be understood that although the terms first, second, third, etc. may be used herein to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first message may also be referred to as a second message, and similarly, a second message may also be referred to as a first message, without departing from the scope of the present application. The word "if" as used herein may be interpreted as "at … …" or "at … …" or "responsive to a determination", depending on the context.
FIG. 1 is a flow chart illustrating a method of photovoltaic power ultra-short term prediction, according to an exemplary embodiment, as shown in FIG. 1, the method may include the steps of:
s1: acquiring historical power generation power data and historical meteorological data of each photovoltaic power station in an area;
s2: preprocessing the historical power generation data and the historical meteorological data;
s3: feature screening is carried out on the preprocessed historical meteorological data, so that meteorological variables related to photovoltaic power generation power are obtained;
s4: according to the meteorological variable and the preprocessed historical power generation data, a sample set is constructed and obtained, wherein the sample set is data in three dimensions of time, space and characteristics;
s5: constructing a space-time feature map convolution network, constructing an ultra-short-term photovoltaic power prediction model based on the space-time feature map convolution network, and training the ultra-short-term photovoltaic power prediction model by using the sample set;
s6: and inputting the sample set to be predicted into a trained ultra-short-term photovoltaic power prediction model, and outputting a prediction result.
According to the embodiment, the photovoltaic power ultra-short-term prediction method can simultaneously predict the photovoltaic power for a plurality of photovoltaic power stations in the area, wherein the space-time relevance of each photovoltaic power station can be fully excavated by utilizing the strong characteristic extraction and fusion capability of the graph rolling network, and the accurate power prediction of multi-station cooperation is realized.
The method provided by the application adopts multi-step prediction of a multi-input multi-output strategy, can output power prediction data at a plurality of moments in the future at one time, meets the requirement of ultra-short-term power prediction, and has higher practicability compared with single-step prediction.
Aiming at the photovoltaic ultra-short-term power prediction task, the method provided by the application further improves the time space characteristic graph convolution network model, changes the structure of the characteristic parameter matrix, greatly reduces the number of model parameters, reduces the possibility of the occurrence of the phenomenon of overfitting, and enables the model to be lighter. The method can provide new ideas and references for photovoltaic power prediction and application of the graph neural network in the power system.
The following describes in detail, with reference to specific examples, a specific implementation of ultra-short-term prediction of photovoltaic power using the method of the present invention. Taking 4 photovoltaic power stations in the same province in south China as an example, live power generation data and environment monitor data of each power station, which are three months from 10 months in 2019 to 12 months in 2019, are selected for modeling and analysis, and the data resolution is 15min.
In the implementation of S1: acquiring historical power generation power data and historical meteorological data of each photovoltaic power station in an area;
specifically, historical power generation data and historical meteorological data of each photovoltaic power station to be predicted in the area are obtained, wherein the meteorological data can comprise historical time sequence data such as total irradiance, air temperature, air pressure, relative humidity and the like measured from an environment monitor of the photovoltaic power station. Because photovoltaic power generation is largely affected by changes in meteorological conditions, the addition of historical meteorological data may result in more accurate prediction results than training and prediction using historical power generation data alone.
In the implementation of S2: preprocessing the historical power generation data and the historical meteorological data; the following sub-steps may be included:
s21: the average value of the historical power generation data and the data of the time before and after the historical meteorological data is utilized to complement the missing value in the data;
in particular, photovoltaic power generation power data and meteorological data collected from the field often have a certain amount of missing values due to transmission problems and failure of the data collection equipment. Moreover, since rated capacities of stations are different, units of different meteorological factor variables are different, and input characteristics of the network have a problem of non-uniform dimensions. Therefore, before various data input models are predicted, the data needs to be preprocessed, and the average value of the data at the front and rear moments is used for complementing the missing value in the time sequence data. The data quality can be improved by complementing the missing value, and the prediction accuracy is increased.
S22: normalizing the completed historical power generation data and the historical meteorological data;
specifically, the normalization formula is:
wherein X is nom For the normalized data, X is the original data, X min Is the minimum value of the original data, X max Is the maximum value of the original data.
This step can be divided into two aspects:
(1) Aiming at the completed historical power generation data, taking the maximum value of the power generation power of a certain photovoltaic power station as the maximum value, taking the minimum value of the power generation power of a certain photovoltaic power station as the minimum value, and carrying out normalization processing on the completed historical power generation data by utilizing the maximum value and the minimum value;
in particular, due to the present inventionThe standing model can simultaneously predict the power generation power of a plurality of different photovoltaic power stations in the same region, and the rated capacities of different stations can be greatly different, but the meteorological conditions are basically similar. Thus, for photovoltaic power generation power data, X max X is the maximum value of the generated power of a certain photovoltaic power station min Is the minimum value of the generated power of a certain photovoltaic power station.
(2) And aiming at the completed historical meteorological data, taking the maximum value of all the meteorological variables of the photovoltaic power station to be tested as the maximum value, taking the minimum value of all the meteorological variables of the photovoltaic power station to be tested as the minimum value, and carrying out normalization processing on the completed historical meteorological data by utilizing the maximum value and the minimum value.
Specifically, for meteorological data such as total irradiance, air temperature, air pressure, relative humidity, X max For the maximum value of the meteorological variable in all photovoltaic power stations to be tested, X min The minimum value of the meteorological variable in all photovoltaic power stations to be tested is the meteorological variable.
If the original data with different dimensions and different orders of magnitude are directly input into the model, the training and calculation of the network can be greatly interfered due to overlarge numerical value difference. It is therefore necessary to transform dimensional data into dimensionless data that is substantially in the range of 0, 1. The power generation power of several different photovoltaic power stations in the same region is predicted, and the rated capacities of different stations can be greatly different, so that the power generation power is greatly different, but the meteorological conditions are basically similar. Therefore, the historical power generation data and the historical meteorological data are normalized in different modes, the data can be better mapped into the range of 0 and 1, the dimension is eliminated, the convergence speed of the model is improved, and the prediction accuracy is improved.
In the implementation of S3: feature screening is carried out on the preprocessed historical meteorological data, so that meteorological variables related to photovoltaic power generation power are obtained;
specifically, a gray correlation analysis method is adopted to perform feature screening on the preprocessed historical meteorological data, so as to obtain meteorological variables correlated with photovoltaic power generation power.
Since there are many meteorological variables, if all the meteorological variables are used as input samples for photovoltaic power prediction, the model parameters are excessively large, and data redundancy is easy to occur. Therefore, it is necessary to select a meteorological variable having a strong correlation with the photovoltaic power generation power as an input of the model. However, the photovoltaic power generation power has strong correlation with weather variables and weak correlation with the weather variables, and cannot be determined empirically only, and quantitative analysis is required.
Therefore, the gray correlation analysis method is adopted to perform feature screening on the preprocessed historical meteorological data to obtain meteorological variables correlated with the photovoltaic power generation power, so that the prediction accuracy is improved.
In the implementation of S4: according to the meteorological variable and the preprocessed historical power generation data, a sample set is constructed and obtained, wherein the sample set is data in three dimensions of time, space and characteristics;
specifically, a sample at time k is defined as a time slice X k I.e. X k ∈R n×v N is the number of photovoltaic power stations, v is the characteristic number contained in each photovoltaic power station, namely the sum of the historical power and the types of meteorological variables;
x at successive c moments k Stacked chronologically into a three-dimensional tensor:
U k ＝[X k-c+1 ,X k-c+2 ,...,X k ]∈R n×v×c
the three dimensions are the spatial, characteristic and temporal dimensions, respectively, and the three-dimensional tensor can be used as a sample set of the model.
By firstly sorting the data into tensors with three dimensions of time, space and characteristics, a sample set is constructed, so that the sorted samples can be conveniently input into a model for subsequent prediction.
In actual use, the sample set may be divided into a training set, a validation set and a test set, and the proportions may be 70%, 10% and 20%, respectively.
In the implementation of S5: constructing a space-time feature map convolution network, constructing an ultra-short-term photovoltaic power prediction model based on the space-time feature map convolution network, and training the ultra-short-term photovoltaic power prediction model by using the sample set;
specifically, constructing a spatio-temporal feature graph convolution network, including:
constructing a space-time feature graph convolution network, wherein the space-time feature graph convolution network consists of space-time feature graph convolution network layers, and the transfer function of a first layer of the network is as follows:
wherein H is (0) ＝U k Namely, the input of the whole network;
and->
Respectively, a time parameter, a space parameter, a characteristic parameter and a bias parameter, c, n and v are respectively a time dimension number, a space dimension number and a characteristic dimension number, +.>
The input matrix and the output matrix are represented,
representing tensor mixture multiplier,>
the sub-matrices representing tensors X and Y decomposed in 2-dimensional correspond to matrix multiplication,
sub-matrices representing tensors X and Y decomposed in 1, 3-th dimensions are respectively corresponding to matrix multiplicationIn the method, σ (·) represents the activation function, which is preferably a ReLU function. Fig. 2 shows the above operation. After inputting the sample into the model. The model firstly extracts the information of the space dimension and the time dimension respectively, then fixes the space dimension and the time dimension, and performs information fusion in the characteristic dimension, so that a power prediction result can be finally obtained.
Preferably, for ultra-short-term prediction in the prediction of the generated power, the distribution of the aggregation relation between the features in the time dimension and the space dimension is approximately considered to be constant in the range of the prediction time step due to the short prediction time step. Therefore, when the features are fused, the space dimension and the time dimension can be unchanged, and the feature parameters W F The variability in the temporal and spatial dimensions need not be considered. I.e. n in the characteristic parameters of the master model l ＝c l ＝1，
Becomes as follows
After improvement, characteristic parameter W F The parameter amount of the model is changed to be very small, which is equivalent to degradation into a two-dimensional matrix, so that the number of model parameters is greatly reduced, and the possibility of over fitting phenomenon is reduced.
The construction of the ultra-short-term photovoltaic power prediction model based on the space-time characteristic graph convolution network can comprise the following steps:
and according to the space-time characteristic diagram convolution network, a multi-step prediction model of the photovoltaic power generation power is established by adopting a multi-input multi-output strategy, and the model can output the power generation power predicted values of n photovoltaic power stations at a plurality of moments in the future.
Specifically, according to the space-time characteristic diagram convolution network layer, a multi-step prediction model of the photovoltaic power generation power is established by adopting a multi-input multi-output strategy, and the model can output the power generation power prediction values of n photovoltaic power stations at a plurality of moments in the future at one time, and a schematic diagram is shown in fig. 3. The multi-input multi-output strategy keeps the relevance and the dependence among a plurality of predicted points, avoids the error caused by the direct strategy for independently modeling each predicted point, and is successfully applied to time sequence prediction tasks in a plurality of fields at present.
To predict the next 16 point in time power, the input-output relationship of the model can be expressed as:
[Y k+16 ，Y k+15 ，…Y k+1 ]＝model[X k ，X k-1 ，…，X k-c+1 ]
in the embodiment, the model can output the predicted value of the generated power of 4 photovoltaic power stations at 16 future moments at one time. Because the data time interval is 15min, the model can predict the power of 0-4 time in the future each time, and can complete the task requirement of ultra-short-term prediction.
The model includes 2 spatiotemporal feature map convolution network layers. The schematic diagrams of the model are shown in fig. 4 (a) and 4 (b). The network parameter settings are shown in table 1.
Table 1 prediction model parameter settings based on spatiotemporal feature map convolutional network:
the input training set sample is used for training the model, the loss function adopted during training is an MAE function, and the input time slice length c=20 is set. The network trains 200 epochs, and the accuracy of the current model is tested after each iteration using a validation set.
In the implementation of S6: and inputting the sample set to be predicted into a trained ultra-short-term photovoltaic power prediction model, and outputting a prediction result.
Specifically, a test sample is input into a trained space-time characteristic diagram convolution network model, an ultra-short-term photovoltaic power prediction result of each photovoltaic power station is output, and the accuracy of the model is verified, specifically:
and inputting the test set sample into a trained space-time characteristic graph convolution network model, setting the input time slice length c=20, and outputting the ultra-short-term photovoltaic power prediction result of each photovoltaic power station.
The accuracy of the model can be verified by normalized root mean square error (NRMSE, root Mean Square Error), normalized mean absolute error (NMAE, mean Absolute Error), the mathematical expression of which is as follows:
wherein y is N (i) And y' N (i) The true value and the predicted value after normalization, respectively.
The prediction results of the 4 stations for 16 times in the future 0 to 4 are shown in fig. 5 (a) to 5 (d), wherein the broken line is a predicted value and the solid line is an actual value. Table 2 shows the prediction errors for each station for an input time slice length of 20 for a multi-step prediction model employing a multiple-input multiple-output strategy. At c=20, the average NRMSE of each station is only 4.43%, and NMAE is only 1.90%, so that a good prediction effect is obtained.
Table 2 prediction error for photovoltaic power multi-step prediction model:
corresponding to the embodiment of the photovoltaic power ultra-short-term prediction method, the application also provides an embodiment of the photovoltaic power ultra-short-term prediction device.
Fig. 6 is a block diagram of a photovoltaic power ultra-short term prediction device, according to an example embodiment. Referring to fig. 6, the apparatus includes:
an acquisition module 21, configured to acquire historical power generation data and historical meteorological data of each photovoltaic power station in the area;
a preprocessing module 22 for preprocessing the historical generated power data and the historical meteorological data;
the screening module 23 is used for performing characteristic screening on the preprocessed historical meteorological data to obtain meteorological variables related to photovoltaic power generation power;
the sample set construction module 24 is configured to construct a sample set according to the meteorological variable and the preprocessed historical power data, where the sample set is data of three dimensions including time, space and features;
the model construction and training module 25 is configured to construct a spatiotemporal feature graph convolution network layer, construct an ultrashort-term photovoltaic power prediction model based on the spatiotemporal feature graph convolution network layer, and train the ultrashort-term photovoltaic power prediction model by using the sample set;
the prediction module 26 is configured to input the sample set to be predicted into a trained ultra-short term photovoltaic power prediction model, and output a prediction result.
The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
For the device embodiments, reference is made to the description of the method embodiments for the relevant points, since they essentially correspond to the method embodiments. The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purposes of the present application. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
Correspondingly, the application also provides electronic equipment, which comprises: one or more processors; a memory for storing one or more programs; the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the photovoltaic power ultra-short term prediction method as described above.
Accordingly, the present application also provides a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement a photovoltaic power ultra-short term prediction method as described above.
Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.
It is to be understood that the present application is not limited to the precise arrangements and instrumentalities shown in the drawings, which have been described above, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the application is limited only by the appended claims.
Claims (10)
1. A photovoltaic power ultra-short term prediction method, comprising:
acquiring historical power generation power data and historical meteorological data of each photovoltaic power station in an area;
preprocessing the historical power generation data and the historical meteorological data;
feature screening is carried out on the preprocessed historical meteorological data, so that meteorological variables related to photovoltaic power generation power are obtained;
according to the meteorological variable and the preprocessed historical power generation data, a sample set is constructed and obtained, wherein the sample set is data in three dimensions of time, space and characteristics;
constructing a space-time feature map convolution network, constructing an ultra-short-term photovoltaic power prediction model based on the space-time feature map convolution network, and training the ultra-short-term photovoltaic power prediction model by using the sample set;
and inputting the sample set to be predicted into a trained ultra-short-term photovoltaic power prediction model, and outputting a prediction result.
2. The method of claim 1, wherein preprocessing the historical generated power data and historical meteorological data comprises:
the average value of the historical power generation data and the data of the time before and after the historical meteorological data is utilized to complement the missing value in the data;
and normalizing the completed historical power generation data and the historical meteorological data.
3. The method of claim 1, wherein normalizing the completed historical generated power data and the historical meteorological data comprises:
aiming at the completed historical power generation data, taking the maximum value of the power generation power of a certain photovoltaic power station as the maximum value, taking the minimum value of the power generation power of a certain photovoltaic power station as the minimum value, and carrying out normalization processing on the completed historical power generation data by utilizing the maximum value and the minimum value;
and aiming at the completed historical meteorological data, taking the maximum value of all the meteorological variables of the photovoltaic power station to be tested as the maximum value, taking the minimum value of all the meteorological variables of the photovoltaic power station to be tested as the minimum value, and carrying out normalization processing on the completed historical meteorological data by utilizing the maximum value and the minimum value.
4. The method of claim 1, wherein the feature screening of the preprocessed historical meteorological data to obtain meteorological variables associated with photovoltaic power generation comprises:
and adopting a gray correlation analysis method to perform characteristic screening on the preprocessed historical meteorological data to obtain meteorological variables correlated with the photovoltaic power generation power.
5. The method of claim 1, wherein constructing a sample set from the preprocessed meteorological variable and preprocessed historical power generation data comprises:
defining the sample at time k as a timeFilm X k I.e. X k ∈R n×v N is the number of photovoltaic power stations, v is the characteristic number contained in each photovoltaic power station, namely the sum of the historical power and the types of meteorological variables;
x at successive c moments k Stacked chronologically into a three-dimensional tensor:
U k ＝[X k-c+1 ,X k-c+2 ,...,X k ]∈R n×v×c
the three dimensions are the spatial, characteristic and temporal dimensions, respectively, and the three-dimensional tensor can be used as a sample set of the model.
6. The method of claim 1, wherein constructing a spatio-temporal feature graph convolution network comprises:
constructing a space-time feature graph convolution network, wherein the space-time feature graph convolution network consists of space-time feature graph convolution network layers, and the transfer function of a first layer of the network is as follows:
wherein H is (0) ＝U k Namely, the input of the whole network;
and->
Respectively, a time parameter, a space parameter, a characteristic parameter and a bias parameter, c, n and v are respectively a time dimension number, a space dimension number and a characteristic dimension number, +.>
Representing an input matrix and an output matrix,>
representing tensor mixture multiplier,>
the sub-matrices representing tensors X and Y decomposed in 2-dimensional correspond to matrix multiplication,
matrix multiplication is respectively and correspondingly performed on submatrices representing tensors X and Y decomposed according to 1 st dimension and 3 rd dimension, and sigma (°) represents an activation function;
preferably, n is taken l ＝c l ＝1，
Become->
7. The method of claim 6, wherein constructing an ultrashort-term photovoltaic power prediction model based on the spatiotemporal feature graph convolution network comprises:
and according to the space-time characteristic diagram convolution network, a multi-step prediction model of the photovoltaic power generation power is established by adopting a multi-input multi-output strategy, and the model can output the power generation power predicted values of n photovoltaic power stations at a plurality of moments in the future.
8. A photovoltaic power ultra-short term prediction device, comprising:
the acquisition module is used for acquiring historical power generation power data and historical meteorological data of each photovoltaic power station in the area;
the preprocessing module is used for preprocessing the historical power generation data and the historical meteorological data;
the screening module is used for carrying out characteristic screening on the preprocessed historical meteorological data to obtain meteorological variables related to the photovoltaic power generation power;
the sample set construction module is used for constructing and obtaining a sample set according to the meteorological variable and the preprocessed historical power generation data, wherein the sample set is data of three dimensions of time, space and characteristics;
the model construction and training module is used for constructing a space-time characteristic graph convolution network layer, constructing an ultra-short-term photovoltaic power prediction model based on the space-time characteristic graph convolution network layer, and training the ultra-short-term photovoltaic power prediction model by using the sample set;
and the prediction module is used for inputting the sample set to be predicted into the trained ultra-short-term photovoltaic power prediction model and outputting a prediction result.
9. An electronic device, comprising:
one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-7.
10. A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the steps of the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211130291.7A CN116011608A (en) | 2022-09-16 | 2022-09-16 | Photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211130291.7A CN116011608A (en) | 2022-09-16 | 2022-09-16 | Photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116011608A true CN116011608A (en) | 2023-04-25 |
Family
ID=86030536
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211130291.7A Pending CN116011608A (en) | 2022-09-16 | 2022-09-16 | Photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116011608A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116384593A (en) * | 2023-06-01 | 2023-07-04 | 深圳市国电科技通信有限公司 | Distributed photovoltaic output prediction method and device, electronic equipment and medium |
| CN116432874A (en) * | 2023-06-14 | 2023-07-14 | 青岛鼎信通讯科技有限公司 | Distributed photovoltaic power prediction method based on characteristic power |
| CN116523145A (en) * | 2023-06-29 | 2023-08-01 | 长江三峡集团实业发展(北京)有限公司 | Photovoltaic power ultra-short-term prediction method and device, computer equipment and storage medium |
| 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 (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113052389A (en) * | 2021-04-01 | 2021-06-29 | 中国电力科学研究院有限公司 | Distributed photovoltaic power station ultra-short-term power prediction method and system based on multiple tasks |
| CN113822467A (en) * | 2021-08-24 | 2021-12-21 | 华南理工大学 | Graph neural network prediction method for electric power area load |
| CN114004139A (en) * | 2021-09-24 | 2022-02-01 | 沈阳工程学院 | Photovoltaic power generation power prediction method |
| CN114819264A (en) * | 2022-03-11 | 2022-07-29 | 中国长江三峡集团有限公司 | Photovoltaic power station irradiance ultra-short term prediction method based on space-time dependence and storage medium |
-
2022
- 2022-09-16 CN CN202211130291.7A patent/CN116011608A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113052389A (en) * | 2021-04-01 | 2021-06-29 | 中国电力科学研究院有限公司 | Distributed photovoltaic power station ultra-short-term power prediction method and system based on multiple tasks |
| CN113822467A (en) * | 2021-08-24 | 2021-12-21 | 华南理工大学 | Graph neural network prediction method for electric power area load |
| CN114004139A (en) * | 2021-09-24 | 2022-02-01 | 沈阳工程学院 | Photovoltaic power generation power prediction method |
| CN114819264A (en) * | 2022-03-11 | 2022-07-29 | 中国长江三峡集团有限公司 | Photovoltaic power station irradiance ultra-short term prediction method based on space-time dependence and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 阚博文 等: ""基于图机器学习的分布式光伏发电预测"", 《供用电》, vol. 36, no. 11, 30 November 2019 (2019-11-30), pages 20 - 27 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116384593A (en) * | 2023-06-01 | 2023-07-04 | 深圳市国电科技通信有限公司 | Distributed photovoltaic output prediction method and device, electronic equipment and medium |
| CN116384593B (en) * | 2023-06-01 | 2023-08-18 | 深圳市国电科技通信有限公司 | Distributed photovoltaic output prediction method and device, electronic equipment and medium |
| CN116432874A (en) * | 2023-06-14 | 2023-07-14 | 青岛鼎信通讯科技有限公司 | Distributed photovoltaic power prediction method based on characteristic power |
| CN116523145A (en) * | 2023-06-29 | 2023-08-01 | 长江三峡集团实业发展(北京)有限公司 | Photovoltaic power ultra-short-term prediction method and device, computer equipment and storage medium |
| CN116523145B (en) * | 2023-06-29 | 2023-10-20 | 长江三峡集团实业发展(北京)有限公司 | Photovoltaic power ultra-short-term prediction method and device, computer equipment and storage medium |
| 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 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070226B (en) | Photovoltaic power prediction method and system based on convolutional neural network and meta-learning | |
| CN116011608A (en) | Photovoltaic power ultra-short-term prediction method and device, electronic equipment and storage medium | |
| Huang et al. | Multiple-input deep convolutional neural network model for short-term photovoltaic power forecasting | |
| CN109842373B (en) | Photovoltaic array fault diagnosis method and device based on space-time distribution characteristics | |
| Liu et al. | Fault diagnosis approach for photovoltaic array based on the stacked auto-encoder and clustering with IV curves | |
| Shams et al. | Artificial intelligence-based prediction and analysis of the oversupply of wind and solar energy in power systems | |
| CN112329339B (en) | Short-term wind speed prediction method for wind power plant | |
| CN104102951A (en) | Short-term wind power prediction method based on EMD (Empirical Mode Decomposition) historical data preprocessing | |
| CN114462718A (en) | CNN-GRU wind power prediction method based on time sliding window | |
| Nelega et al. | Prediction of power generation of a photovoltaic power plant based on neural networks | |
| Wang et al. | An approach for day-ahead interval forecasting of photovoltaic power: A novel DCGAN and LSTM based quantile regression modeling method | |
| Gong et al. | Fast fault detection method for photovoltaic arrays with adaptive deep multiscale feature enhancement | |
| CN116706907B (en) | Photovoltaic power generation prediction method based on fuzzy reasoning and related equipment | |
| CN113837434A (en) | Solar photovoltaic power generation prediction method and device, electronic equipment and storage medium | |
| CN111091141B (en) | Photovoltaic backboard fault diagnosis method based on layered Softmax | |
| CN113240217A (en) | Photovoltaic power generation prediction method and device based on integrated prediction model | |
| CN113131515A (en) | Intelligent microgrid AI fault detection and network optimization reconstruction system and method | |
| CN116317937A (en) | Distributed photovoltaic power station operation fault diagnosis method | |
| Zhang et al. | Graph neural network based short-term solar irradiance forcasting model considering surrounding meteorological factors | |
| Rosato et al. | A Combined Deep Learning Approach for Time Series Prediction in Energy Environments | |
| CN114861993A (en) | Regional photovoltaic power generation prediction method based on federal learning and deep neural network | |
| CN114913036A (en) | Combined fault diagnosis system and fault diagnosis method for photovoltaic power station | |
| CN111027816A (en) | Photovoltaic power generation efficiency calculation method based on data envelope analysis | |
| CN112560898B (en) | Load space-time prediction method based on deep learning | |
| Guanoluisa-Pineda et al. | Short-Term forecasting of photovoltaic power in an isolated area of Ecuador using deep learning techniques |
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 |