CN106815659B - Ultra-short-term solar radiation prediction method and device based on hybrid model - Google Patents

Ultra-short-term solar radiation prediction method and device based on hybrid model Download PDF

Info

Publication number
CN106815659B
CN106815659B CN201710042017.7A CN201710042017A CN106815659B CN 106815659 B CN106815659 B CN 106815659B CN 201710042017 A CN201710042017 A CN 201710042017A CN 106815659 B CN106815659 B CN 106815659B
Authority
CN
China
Prior art keywords
sequence
clear sky
solar radiation
wavelet
model
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.)
Active
Application number
CN201710042017.7A
Other languages
Chinese (zh)
Other versions
CN106815659A (en
Inventor
张雪松
朱想
赵波
周海
崔方
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
State Grid Corp of China SGCC
State Grid Zhejiang Electric Power Co Ltd
China Electric Power Research Institute Co Ltd CEPRI
Electric Power Research Institute of State Grid Zhejiang Electric Power Co Ltd
Original Assignee
State Grid Corp of China SGCC
State Grid Zhejiang Electric Power Co Ltd
China Electric Power Research Institute Co Ltd CEPRI
Electric Power Research Institute of State Grid Zhejiang Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by State Grid Corp of China SGCC, State Grid Zhejiang Electric Power Co Ltd, China Electric Power Research Institute Co Ltd CEPRI, Electric Power Research Institute of State Grid Zhejiang Electric Power Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN201710042017.7A priority Critical patent/CN106815659B/en
Publication of CN106815659A publication Critical patent/CN106815659A/en
Application granted granted Critical
Publication of CN106815659B publication Critical patent/CN106815659B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • 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

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Marketing (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a hybrid model-based ultra-short-term solar radiation prediction method and a hybrid model-based ultra-short-term solar radiation prediction device, which comprise the steps of periodically obtaining solar radiation observation samples collected by an observation station; obtaining a clear sky index time sequence according to a solar radiation observation sample and a solar radiation-clear sky index relational expression; performing wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence; respectively inputting clear sky radiation sequences into a preset linear prediction model, a first support vector machine model and a neural network model to obtain three input sequences; the three output sequences are input into a second support vector machine model after being overlapped; and taking the output result of the second support vector machine model as a predicted clear sky index time sequence to be brought into a solar radiation-clear sky index relational expression to obtain a predicted solar radiation value. According to the invention, the influence of the prediction error of a single prediction model on the overall preset precision is weakened in a mode of mixing and predicting a plurality of prediction models.

Description

Ultra-short-term solar radiation prediction method and device based on hybrid model
Technical Field
The invention relates to the technical field of photovoltaic power generation, in particular to an ultrashort-term solar radiation prediction method and device based on a hybrid model.
Background
Under the influence of policy excitation of the countries in the aspect of photovoltaic power generation in recent years, large-scale photovoltaic power stations are connected into a power grid in a successive way, and because the output power of photovoltaic power generation has randomness and volatility, the safety, stability and economic operation of the power grid are influenced. Therefore, the output power of the photovoltaic power station needs to be accurately predicted, so that the coordination and coordination of the conventional power supply and the photovoltaic power generation are arranged overall, and the photovoltaic resources are fully utilized.
The output power of the photovoltaic power generation system is largely determined by the amount of solar radiation which can be received by the photovoltaic panel, the amount of solar radiation is easily affected by weather factors, and the photovoltaic power generation system has the defects of intermittence, volatility and randomness, so that the output power is unstable and is difficult to predict.
At present, in the aspect of solar radiation prediction, two models, namely a support vector machine and a neural network, in a linear model and a nonlinear prediction model are widely used. For the linear model, when the prediction time is longer, the correlation between data is weakened, and the linear model is difficult to realize accurate prediction; the kernel function selection of the support vector machine has great influence on the operation result, the neural network has the problem that the optimal network structure is difficult to determine, and the two may have an overfitting phenomenon during model training.
Therefore, how to provide a hybrid model-based ultra-short-term solar radiation prediction method and a hybrid model-based ultra-short-term solar radiation prediction device, which can reduce the influence of the prediction error of a single model on the overall prediction accuracy, is a problem that needs to be solved by those skilled in the art at present.
Disclosure of Invention
The invention aims to provide a hybrid model-based ultra-short-term solar radiation prediction method and a hybrid model-based ultra-short-term solar radiation prediction device, which can weaken the influence of the prediction error of a single prediction model on the overall preset precision in a hybrid prediction mode of multiple prediction models.
In order to solve the technical problem, the invention provides a hybrid model-based ultra-short-term solar radiation prediction method, which comprises the following steps:
periodically acquiring solar radiation observation samples collected by an observation station;
obtaining a clear sky index time sequence according to the solar radiation observation sample and a solar radiation-clear sky index relational expression;
performing wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence;
respectively inputting the clear sky radiation sequence into a preset linear prediction model, a first support vector machine model and a neural network model to obtain three input sequences;
inputting the three output sequences into a second support vector machine model after superposition;
and taking the output result of the second support vector machine model as a predicted clear sky index time sequence to be brought into the solar radiation-clear sky index relational expression to obtain a predicted solar radiation value.
Preferably, the process of performing wavelet transform processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence specifically includes:
respectively performing one-dimensional wavelet decomposition of multiple scales on the clear sky index time sequence according to wavedec functions to obtain a multilayer wavelet sequence;
reconstructing the wavelet sequences of each layer according to preset wavelet low-frequency coefficients of each layer and a wavelet reconstruction function respectively to obtain a reconstruction sequence obtained after reconstructing each wavelet sequence of each layer;
wherein the wavelet reconstruction function is:
Figure BDA0001215138630000021
wherein f (n) is the wavelet sequence; wf(j, k) is a reconstruction sequence corresponding to the wavelet sequence; psij,kThe wavelet low-frequency coefficient corresponding to the wavelet sequence is obtained;
respectively carrying out column direction mean value removing operation on each reconstruction sequence and the clear sky index time sequence to obtain each mean value sequence
Figure BDA0001215138630000022
Wherein each sequence is ∈ Rm×n；
Performing partial least squares regression on each mean sequence;
calculating the standard deviation s of each row of each mean value sequence according to the standard deviation relationjWherein the standard deviation relation is as follows:
Figure BDA0001215138630000023
according to the standard deviation and the coefficient relation, obtaining partial least square projection coefficients r corresponding to the reconstruction sequences and the clear sky index time sequencesj；
rj＝|sj×fj|，j＝1,2…,n
Wherein f isjIs a partial least squares regression factor;
to rjCarrying out normalization processing to obtain partial least square projection coefficient vector Rj；
Each partial least squares projection coefficient vector RjAnd as a weighting coefficient, performing weighted summation on each reconstruction sequence and the clear sky index time sequence to obtain the clear sky radiation sequence.
Preferably, the process of periodically acquiring the solar radiation observation samples collected by the observation station specifically includes:
collecting solar radiation observation data at an observation point once every preset collection period;
calculating the average value of a plurality of solar radiation observation data collected in the period every preset statistical period to obtain the solar radiation observation sample; and the preset statistical period is an integral multiple of the preset acquisition period.
Preferably, before obtaining the clear sky index time sequence according to the solar radiation observation sample and the solar radiation-clear sky index relational expression, the method further includes:
deleting erroneous data in the solar radiation observation sample;
and screening the data in the solar radiation observation sample after the error data is deleted, and taking a data set in a preset angle of the solar zenith angle in the solar radiation observation sample as an effective solar radiation observation sample.
In order to solve the above technical problem, the present invention further provides an ultra-short term solar radiation prediction apparatus based on a hybrid model, including:
the system comprises a sample acquisition module, a data acquisition module and a data processing module, wherein the sample acquisition module is used for periodically acquiring solar radiation observation samples acquired by an observation station;
the clear sky conversion module is used for obtaining a clear sky index time sequence according to the solar radiation observation sample and a solar radiation-clear sky index relational expression;
the wavelet transformation module is used for carrying out wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence;
the hybrid model processing module is used for respectively inputting the clear sky radiation sequence into a preset linear prediction model, a first support vector machine model and a neural network model to obtain three input sequences;
the weighted model processing module is used for inputting the three output sequences into a second support vector machine model after superposition; sending an output result of the second support vector machine model to a clear sky inverse conversion module;
and the clear sky inverse conversion module is used for taking an output result of the second support vector machine model as a predicted clear sky index time sequence to be brought into the solar radiation-clear sky index relational expression to obtain a predicted solar radiation value.
Preferably, the wavelet transform module specifically includes:
the wavelet decomposition unit is used for performing one-dimensional wavelet decomposition of multiple scales on the clear sky index time sequence according to wavedec functions respectively to obtain a multilayer wavelet sequence;
the reconstruction unit is used for reconstructing the wavelet sequences of each layer according to the preset wavelet low-frequency coefficients of each layer and the wavelet reconstruction function respectively to obtain a reconstruction sequence obtained after reconstructing each layer of the wavelet sequences; wherein the wavelet reconstruction function is:
Figure BDA0001215138630000041
wherein f (n) is the wavelet sequence; wf(j, k) is a reconstruction sequence corresponding to the wavelet sequence; psij,kThe wavelet low-frequency coefficient corresponding to the wavelet sequence is obtained;
a partial least square projection coefficient processing unit, configured to perform column direction mean value removing operation on each reconstruction sequence and the clear sky index time sequence to obtain each mean value sequence
Figure BDA0001215138630000042
Wherein each sequence is ∈ Rm×n(ii) a Performing partial least squares regression on each mean sequence; calculating the standard deviation s of each row of each mean value sequence according to the standard deviation relationjWherein the standard deviation relation is as follows:
Figure BDA0001215138630000043
according to the standard deviation and the coefficient relational expression, obtaining partial least squares corresponding to each reconstruction sequence and the clear sky index time sequenceProjection coefficient rj；
rj＝|sj×fj|，j＝1,2…,n
Wherein f isjIs a partial least squares regression factor; to rjCarrying out normalization processing to obtain partial least square projection coefficient vector Rj(ii) a Each partial least squares projection coefficient vector RjAnd as a weighting coefficient, performing weighted summation on each reconstruction sequence and the clear sky index time sequence to obtain the clear sky radiation sequence.
The invention provides an ultra-short-term solar radiation prediction method and device based on a hybrid model, which are characterized in that an obtained solar radiation observation sample is converted into a clear sky index time sequence and subjected to wavelet transformation, then the obtained solar radiation observation sample is respectively input into a linear prediction model, a first support vector machine model and a neural network model for prediction, output results of the three models are input into a second support vector machine model for weighting treatment, and then the output of the second support vector machine model is converted into a predicted solar radiation value through a solar radiation-clear sky index relational expression. Therefore, the method adopts a mode of hybrid processing of the linear prediction model, the first support vector machine model and the neural network model, comprehensively considers the prediction effect of each model, weakens the prediction error brought by single prediction of each model, and ensures the prediction precision.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed in the prior art and the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of the process of a hybrid model-based ultra-short-term solar radiation prediction method provided by the present invention;
fig. 2 is a schematic structural diagram of an ultra-short-term solar radiation prediction apparatus based on a hybrid model according to the present invention.
Detailed Description
The core of the invention is to provide a hybrid model-based ultra-short-term solar radiation prediction method and a hybrid model-based ultra-short-term solar radiation prediction device, and the influence of the prediction error of a single prediction model on the overall preset precision is weakened in a hybrid prediction mode of multiple prediction models.
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a hybrid model-based ultra-short-term solar radiation prediction method, which is shown in fig. 1, wherein fig. 1 is a flow chart of a process of the hybrid model-based ultra-short-term solar radiation prediction method provided by the invention; the method comprises the following steps:
step s 101: periodically acquiring solar radiation observation samples collected by an observation station;
step s 102: obtaining a clear sky index time sequence according to a solar radiation observation sample and a solar radiation-clear sky index relational expression;
the solar radiation-clear sky index relational expression is as follows:
Figure BDA0001215138630000061
wherein k is a clear sky index value; GHIclrThe value of the solar radiation is clear sky; GHImAs observed solar radiation. The solar radiation observation sample comprises a solar radiation time sequence formed by solar radiation observation data and a clear sky solar radiation time sequence formed by clear sky solar radiation data. The conversion can convert the solar radiation time sequence in the solar radiation observation sample into the clear sky index time sequence, thereby eliminating the sun position change pairInfluence of the solar radiation value.
Step s 103: performing wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence;
step s 104: respectively inputting clear sky radiation sequences into a preset linear prediction model, a first support vector machine model and a neural network model to obtain three input sequences;
step s 105: the three output sequences are input into a second support vector machine model after being overlapped;
step s 106: and taking the output result of the second support vector machine model as a predicted clear sky index time sequence to be brought into a solar radiation-clear sky index relational expression to obtain a predicted solar radiation value.
Specifically, the process of step s103 is specifically:
step s 201: respectively performing one-dimensional wavelet decomposition of multiple scales on the clear sky index time sequence according to wavedec functions to obtain a multilayer wavelet sequence;
step s 202: reconstructing each layer of wavelet sequence according to a preset wavelet low-frequency coefficient and a preset wavelet reconstruction function of each layer to obtain a reconstructed sequence obtained after reconstructing each layer of wavelet sequence;
wherein the wavelet reconstruction function wrcoef is:
Figure BDA0001215138630000062
wherein f (n) is a wavelet series; wf(j, k) is a reconstruction sequence corresponding to the wavelet sequence; psij,kThe wavelet low-frequency coefficient corresponding to the wavelet sequence;
it will be appreciated that wavelet transform can decompose a non-stationary sequence to obtain a relatively stationary plurality of time subsequences of different scale (or number of layers) and then perform a single branch reconstruction. For example, the clear sky index time sequence is decomposed into a first-layer wavelet low-frequency sequence, a second-layer wavelet low-frequency sequence and a third-layer wavelet low-frequency sequence, and then the reconstruction is performed according to a preset first-layer wavelet low-frequency coefficient, a preset second-layer wavelet low-frequency coefficient, a preset third-layer wavelet low-frequency coefficient and the wavelet reconstruction function, and then a first-layer reconstruction sequence, a second-layer reconstruction sequence and a third-layer reconstruction sequence are obtained respectively.
Specifically, when the wrcoef function is reconstructed, a zero padding expansion mode is adopted, a one-dimensional signal is loaded, and single-branch reconstruction is performed on the one-dimensional wavelet sequence.
Step s 203: respectively carrying out mean value removing operation in the row direction on each reconstruction sequence and each clear sky index time sequence to obtain each mean value sequence
Figure BDA0001215138630000071
Wherein each sequence is ∈ Rm×n；
Step s 204: performing Partial Least Squares regression (PLS) on each mean sequence;
step s 205: calculating the standard deviation s of each column of each mean value sequence according to the standard deviation relational expressionjWherein, the standard deviation relation is as follows:
Figure BDA0001215138630000072
step s 206: according to the standard deviation and the coefficient relation, obtaining partial least square projection coefficients r corresponding to each reconstruction sequence and clear sky index time sequencej；
rj＝|sj×fj|，j＝1,2…,n
Wherein f isjIs a partial least squares regression factor;
step s 207: to rjCarrying out normalization processing to obtain partial least square projection coefficient vector Rj；
Step s 207: projecting each partial least square projection coefficient vector RjAnd as weighting coefficients, carrying out weighted summation on each reconstruction sequence and the clear sky index time sequence to obtain a clear sky radiation sequence.
It can be understood that the partial least square correlation method is an effective feature extraction method, which can effectively reduce features and improve an analysis model, and the purpose of the invention is to calculate the weighting coefficient when each reconstruction sequence is fitted with the clear sky index time sequence.
Preferably, the process of step s101 is specifically:
collecting solar radiation observation data at an observation point once every preset collection period;
calculating the average value of a plurality of solar radiation observation data collected in the period every preset statistical period to obtain a solar radiation observation sample; and the preset statistical period is an integral multiple of the preset acquisition period.
The preset acquisition period may be 5 minutes, the preset statistical period may be 15 minutes, and of course, both the above two periods may be set according to specific situations, which is not limited in the present invention.
Preferably, after step s101, step s102 further includes:
deleting error data in the solar radiation observation sample;
and screening the data in the solar radiation observation sample after the error data is deleted, and collecting the data in the solar zenith angle preset angle in the solar radiation observation sample as an effective solar radiation observation sample.
It is understood that the error data is an error occurring in the transmission and recording process of the sample data. In addition, the screening operation is to define the solar radiation observation sample as an observation data set within a preset angle of a solar zenith angle after comprehensively considering the influence of the topography of an observation point and the cosine effect of a radiation instrument and the effective action of solar radiation on a photovoltaic system. The preset angle is specifically 80 degrees, and of course, other angle values may be set.
In addition, the hybrid model is adopted, model order fixing is needed in the initial stage, the model order fixing is used for obtaining proper model input, certain influence is caused on the prediction performance, the selection is too small, and the prediction precision is reduced; if the selection is too large, not only the speed is reduced, but also the prediction accuracy is reduced. The criterion function method is to determine the fitting degree (generally according to residual values) of the original data by using a model, and a BIC criterion function method is adopted to determine the order of the model. The BIC criteria function is specifically:
BIC＝-2ln(L)+ln(n)*k
where k is the number of model parameters, L is the likelihood function, and n is the number of observations. The order of the model can be determined to be 7 orders according to the result of the BIC criterion.
I.e. the linear prediction model is defined to be 7 th order, and the specific parameters thereof can be obtained by effective sample training. And determining the weight and the threshold of the three-layer neural network structure and the parameters by adopting a BP learning algorithm.
In addition, when the outputs of the three models are weighted, the support vector machine model is adopted, the output results of the models are directly used as input, supervised learning is carried out by taking the final prediction result as a target, and the optimal parameters of the second support vector machine model are obtained by a cross validation method, so that the final prediction result of the mixed model is obtained, the problem of weight selection is avoided, and the prediction precision and performance of the whole mixed model are effectively improved.
The invention provides an ultra-short-term solar radiation prediction method based on a hybrid model, which comprises the steps of converting an obtained solar radiation observation sample into a clear sky index time sequence, performing wavelet transformation, respectively inputting the clear sky index time sequence into a linear prediction model, a first support vector machine model and a neural network model for prediction, inputting output results of the three models into a second support vector machine model for weighting treatment, and converting the output of the second support vector machine model into a predicted solar radiation value through a solar radiation-clear sky index relational expression. Therefore, the method adopts a mode of hybrid processing of the linear prediction model, the first support vector machine model and the neural network model, comprehensively considers the prediction effect of each model, weakens the prediction error brought by single prediction of each model, and ensures the prediction precision.
The invention also provides a hybrid model-based ultra-short-term solar radiation prediction device, which is shown in fig. 2, and fig. 2 is a schematic structural diagram of the hybrid model-based ultra-short-term solar radiation prediction device provided by the invention. The device includes:
the system comprises a sample acquisition module 1, a data acquisition module and a data processing module, wherein the sample acquisition module is used for periodically acquiring solar radiation observation samples acquired by an observation station;
the clear sky conversion module 2 is used for obtaining a clear sky index time sequence according to the solar radiation observation sample and the solar radiation-clear sky index relational expression;
the wavelet transformation module 3 is used for performing wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence;
the hybrid model processing module 4 is used for respectively inputting the clear sky radiation sequence into a preset linear prediction model, a first support vector machine model and a neural network model to obtain three input sequences;
the weighted model processing module 5 is used for inputting the superposed three output sequences into a second support vector machine model; sending the output result of the second support vector machine model to a clear sky inverse conversion module 6;
and the clear sky inverse conversion module 6 is used for taking the output result of the second support vector machine model as a predicted clear sky index time sequence to be brought into a solar radiation-clear sky index relational expression to obtain a predicted solar radiation value.
The wavelet transform module 3 specifically includes:
the wavelet decomposition unit is used for performing one-dimensional wavelet decomposition of multiple scales on the clear sky index time sequence according to wavedec functions respectively to obtain a multilayer wavelet sequence;
the reconstruction unit is used for reconstructing wavelet sequences of each layer according to preset wavelet low-frequency coefficients of each layer and wavelet reconstruction functions respectively to obtain a reconstruction sequence obtained after each wavelet sequence of each layer is reconstructed; wherein the wavelet reconstruction function is:
Figure BDA0001215138630000091
wherein f (n) is a wavelet series; wf(j, k) is a reconstruction sequence corresponding to the wavelet sequence; psij,kThe wavelet low-frequency coefficient corresponding to the wavelet sequence;
partial least squares projection coefficient processing unit forRespectively carrying out mean value removing operation in the row direction on each reconstruction sequence and each clear sky index time sequence to obtain each mean value sequenceWherein each sequence is ∈ Rm×n(ii) a Performing partial least squares regression on each mean sequence; calculating the standard deviation s of each column of each mean value sequence according to the standard deviation relational expressionjWherein, the standard deviation relation is as follows:
Figure BDA0001215138630000101
according to the standard deviation and the coefficient relation, obtaining partial least square projection coefficients r corresponding to each reconstruction sequence and clear sky index time sequencej；
rj＝|sj×fj|，j＝1,2…,n
Wherein f isjIs a partial least squares regression factor; to rjCarrying out normalization processing to obtain partial least square projection coefficient vector Rj(ii) a Projecting each partial least square projection coefficient vector RjAnd as weighting coefficients, carrying out weighted summation on each reconstruction sequence and the clear sky index time sequence to obtain a clear sky radiation sequence.
The invention provides an ultra-short-term solar radiation prediction device based on a hybrid model, which is characterized in that an obtained solar radiation observation sample is converted into a clear sky index time sequence and subjected to wavelet transformation, then the obtained solar radiation observation sample is respectively input into a linear prediction model, a first support vector machine model and a neural network model for prediction, output results of the three models are input into a second support vector machine model for weighting treatment, and then the output of the second support vector machine model is converted into a predicted solar radiation value through a solar radiation-clear sky index relational expression. Therefore, the method adopts a mode of hybrid processing of the linear prediction model, the first support vector machine model and the neural network model, comprehensively considers the prediction effect of each model, weakens the prediction error brought by single prediction of each model, and ensures the prediction precision.
It is to be noted that, in the present specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (4)

1. A hybrid model-based ultra-short-term solar radiation prediction method is characterized by comprising the following steps:
periodically acquiring solar radiation observation samples collected by an observation station;
obtaining a clear sky index time sequence according to the solar radiation observation sample and a solar radiation-clear sky index relational expression;
performing wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence;
respectively inputting the clear sky radiation sequence into a preset linear prediction model, a first support vector machine model and a neural network model to obtain three output sequences;
inputting the three output sequences into a second support vector machine model after superposition;
taking the output result of the second support vector machine model as a predicted clear sky index time sequence to be brought into the solar radiation-clear sky index relational expression to obtain a predicted solar radiation value;
the process of performing wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence specifically comprises the following steps:
respectively performing one-dimensional wavelet decomposition of multiple scales on the clear sky index time sequence according to wavedec functions to obtain a multilayer wavelet sequence;
reconstructing the wavelet sequences of each layer according to preset wavelet low-frequency coefficients of each layer and a wavelet reconstruction function respectively to obtain a reconstruction sequence obtained after reconstructing each wavelet sequence of each layer;
wherein the wavelet reconstruction function is:
Figure FDA0002526296440000011
wherein f (n) is the wavelet sequence; wf(j, k) is a reconstruction sequence corresponding to the wavelet sequence; psij,kThe wavelet low-frequency coefficient corresponding to the wavelet sequence is obtained;
respectively carrying out column direction mean value removing operation on each reconstruction sequence and the clear sky index time sequence to obtain each mean value sequence
Figure FDA0002526296440000012
Wherein each sequence is ∈ Rm×n；
Performing partial least squares regression on each mean sequence;
calculating the standard deviation s of each row of each mean value sequence according to the standard deviation relationjWherein the standard deviation relation is as follows:
Figure FDA0002526296440000013
according to the standard deviation and the coefficient relation, obtaining partial least square projection coefficients r corresponding to the reconstruction sequences and the clear sky index time sequencesj；
rj＝|sj×fj|，j＝1,2…,n
Wherein f isjIs a partial least squares regression factor;
to rjCarrying out normalization processing to obtain partial least square projection coefficient vector Rj；
Each partial least squares projection coefficient vector RjAnd as a weighting coefficient, performing weighted summation on each reconstruction sequence and the clear sky index time sequence to obtain the clear sky radiation sequence.
2. The method according to claim 1, wherein the periodically acquiring of the collected solar radiation observation samples of the observation site is specifically:
collecting solar radiation observation data at an observation point once every preset collection period;
calculating the average value of a plurality of solar radiation observation data collected in the period every preset statistical period to obtain the solar radiation observation sample; and the preset statistical period is an integral multiple of the preset acquisition period.
3. The method according to claim 2, wherein obtaining the clear sky index time series according to the solar radiation observation sample and the solar radiation-clear sky index relation further comprises:
deleting erroneous data in the solar radiation observation sample;
and screening the data in the solar radiation observation sample after the error data is deleted, and taking a data set in a preset angle of the solar zenith angle in the solar radiation observation sample as an effective solar radiation observation sample.
4. An ultra-short term solar radiation prediction device based on a hybrid model, comprising:
the system comprises a sample acquisition module, a data acquisition module and a data processing module, wherein the sample acquisition module is used for periodically acquiring solar radiation observation samples acquired by an observation station;
the clear sky conversion module is used for obtaining a clear sky index time sequence according to the solar radiation observation sample and a solar radiation-clear sky index relational expression;
the wavelet transformation module is used for carrying out wavelet transformation processing on the clear sky index time sequence to obtain a transformed clear sky radiation sequence;
the hybrid model processing module is used for respectively inputting the clear sky radiation sequence into a preset linear prediction model, a first support vector machine model and a neural network model to obtain three output sequences;
the weighted model processing module is used for inputting the three output sequences into a second support vector machine model after superposition; sending an output result of the second support vector machine model to a clear sky inverse conversion module;
the clear sky inverse conversion module is used for taking an output result of the second support vector machine model as a predicted clear sky index time sequence to be brought into the solar radiation-clear sky index relational expression to obtain a predicted solar radiation value;
the wavelet transform module specifically comprises:
the wavelet decomposition unit is used for performing one-dimensional wavelet decomposition of multiple scales on the clear sky index time sequence according to wavedec functions respectively to obtain a multilayer wavelet sequence;
the reconstruction unit is used for reconstructing the wavelet sequences of each layer according to the preset wavelet low-frequency coefficients of each layer and the wavelet reconstruction function respectively to obtain a reconstruction sequence obtained after reconstructing each layer of the wavelet sequences; wherein the wavelet reconstruction function is:
Figure FDA0002526296440000031
wherein f (n) is the wavelet sequence; wf(j, k) is a reconstruction sequence corresponding to the wavelet sequence; psij,kThe wavelet low-frequency coefficient corresponding to the wavelet sequence is obtained;
a partial least square projection coefficient processing unit, configured to perform column direction mean value removing operation on each reconstruction sequence and the clear sky index time sequence to obtain each mean value sequence
Figure FDA0002526296440000032
Wherein each sequence is ∈ Rm×n(ii) a Performing partial least squares regression on each mean sequence; calculating the standard deviation s of each row of each mean value sequence according to the standard deviation relationjWherein the standard deviation relation is as follows:
Figure FDA0002526296440000033
according to the standard deviation and the coefficient relation, obtaining partial least square projection coefficients r corresponding to the reconstruction sequences and the clear sky index time sequencesj；
rj＝|sj×fj|，j＝1,2…,n
Wherein f isjIs a partial least squares regression factor; to rjCarrying out normalization processing to obtain partial least square projection coefficient vector Rj(ii) a Each partial least squares projection coefficient vector RjAnd as a weighting coefficient, performing weighted summation on each reconstruction sequence and the clear sky index time sequence to obtain the clear sky radiation sequence.
CN201710042017.7A 2017-01-20 2017-01-20 Ultra-short-term solar radiation prediction method and device based on hybrid model Active CN106815659B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710042017.7A CN106815659B (en) 2017-01-20 2017-01-20 Ultra-short-term solar radiation prediction method and device based on hybrid model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710042017.7A CN106815659B (en) 2017-01-20 2017-01-20 Ultra-short-term solar radiation prediction method and device based on hybrid model

Publications (2)

Publication Number Publication Date
CN106815659A CN106815659A (en) 2017-06-09
CN106815659B true CN106815659B (en) 2020-09-18

Family

ID=59112871

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710042017.7A Active CN106815659B (en) 2017-01-20 2017-01-20 Ultra-short-term solar radiation prediction method and device based on hybrid model

Country Status (1)

Country Link
CN (1) CN106815659B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110175375A (en) * 2019-05-13 2019-08-27 中国科学院遥感与数字地球研究所 A kind of earth's surface Calculation method for solar radiation based on deep learning
CN111242359B (en) * 2020-01-06 2020-09-29 南京林业大学 Solar radiation online dynamic prediction method based on data drift
CN113779861B (en) * 2021-07-23 2023-08-22 国网河北省电力有限公司电力科学研究院 Photovoltaic Power Prediction Method and Terminal Equipment

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105139264A (en) * 2015-06-06 2015-12-09 安徽工程大学 Photovoltaic generation capacity prediction method based on particle swarm algorithm wavelet neural network
CN105426989A (en) * 2015-11-03 2016-03-23 河海大学 EEMD and combined kernel RVM-based photovoltaic power short-term prediction method
CN105678397A (en) * 2015-11-18 2016-06-15 江苏省电力公司泰州市姜堰区供电公司 Short-term photovoltaic power prediction method based on improved EMD algorithm and Elman algorithm
CN105760961A (en) * 2016-03-08 2016-07-13 广州供电局有限公司 Photovoltaic output power prediction method and system
JP2016194478A (en) * 2015-04-01 2016-11-17 株式会社Ｎｔｔファシリティーズ Meteorological data estimation method and power generation amount estimation method
JP2016201921A (en) * 2015-04-10 2016-12-01 株式会社東芝 Method, device and program for detecting reduction in power generation amount of photovoltaic power generation facility
JP2016224566A (en) * 2015-05-27 2016-12-28 一般財団法人電力中央研究所 Prediction device, prediction method, and prediction program

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2016194478A (en) * 2015-04-01 2016-11-17 株式会社Ｎｔｔファシリティーズ Meteorological data estimation method and power generation amount estimation method
JP2016201921A (en) * 2015-04-10 2016-12-01 株式会社東芝 Method, device and program for detecting reduction in power generation amount of photovoltaic power generation facility
JP2016224566A (en) * 2015-05-27 2016-12-28 一般財団法人電力中央研究所 Prediction device, prediction method, and prediction program
CN105139264A (en) * 2015-06-06 2015-12-09 安徽工程大学 Photovoltaic generation capacity prediction method based on particle swarm algorithm wavelet neural network
CN105426989A (en) * 2015-11-03 2016-03-23 河海大学 EEMD and combined kernel RVM-based photovoltaic power short-term prediction method
CN105678397A (en) * 2015-11-18 2016-06-15 江苏省电力公司泰州市姜堰区供电公司 Short-term photovoltaic power prediction method based on improved EMD algorithm and Elman algorithm
CN105760961A (en) * 2016-03-08 2016-07-13 广州供电局有限公司 Photovoltaic output power prediction method and system

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
光伏功率预测技术;龚莺飞;《电力系统自动化》;20160225;第40卷(第4期);全文 *
太阳能光伏发电的中长期随机特性分析;张曦;《电力系统自动化》;20140325;第38卷(第6期);全文 *
并网光伏的短期运行备用评估;汪海瑛;《电力系统自动化》;20130310;第37卷(第5期);全文 *
组合数值天气预报与地基云图的光伏超短期功率预测模型;朱想;《电力系统自动化》;20150325;第39卷(第6期);全文 *

Also Published As

Publication number Publication date
CN106815659A (en) 2017-06-09

Similar Documents

Publication Publication Date Title
CN109214575B (en) Ultrashort-term wind power prediction method based on small-wavelength short-term memory network
CN111582551B (en) Wind power plant short-term wind speed prediction method and system and electronic equipment
Rodríguez et al. Forecasting intra-hour solar photovoltaic energy by assembling wavelet based time-frequency analysis with deep learning neural networks
CN106815659B (en) Ultra-short-term solar radiation prediction method and device based on hybrid model
CN108717579B (en) Short-term wind power interval prediction method
CN111242377A (en) Short-term wind speed prediction method integrating deep learning and data denoising
CN113159102B (en) Multi-time-scale photovoltaic power prediction method and system
CN106502850B (en) A kind of the multiple faults discrimination method and device of sparse tight frame dictionary learning model
CN116307291B (en) Distributed photovoltaic power generation prediction method and prediction terminal based on wavelet decomposition
CN117633446A (en) Wind speed prediction method and system based on CEEMD-SE-EWT double decomposition and ARIMA combined prediction
CN114266416A (en) Photovoltaic power generation power short-term prediction method and device based on similar days and storage medium
CN116632820A (en) Wind power prediction method based on sequence extension VMD and wind speed fluctuation characteristics
CN115796351A (en) Rainfall shorthand prediction method and device based on variational modal decomposition and microwave attenuation
CN113361737A (en) Abnormity early warning method and system for photovoltaic module
Sun et al. Short-term power load prediction based on VMD-SG-LSTM
CN112381279B (en) Wind power prediction method based on VMD and BLS combined model
CN117200208B (en) User-level short-term load prediction method and system based on multi-scale component feature learning
CN104239689B (en) Short-term wind speed forecasting method based on optimization Method Using Relevance Vector Machine
CN113837434A (en) Solar photovoltaic power generation prediction method and device, electronic equipment and storage medium
CN113361782A (en) Photovoltaic power generation power short-term rolling prediction method based on improved MKPLS
CN117035155A (en) Water quality prediction method
CN116340724A (en) ICCEEMDAN-LRTC-based power load completion method and system
CN113343590B (en) Wind speed prediction method and system based on combined model
Şen Probabilistic innovative solar irradiation estimation
CN111932007B (en) Power prediction method and device for photovoltaic power station and storage medium

Legal Events

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