CN115689055A - Short-term solar irradiance prediction method and device - Google Patents
Short-term solar irradiance prediction method and device Download PDFInfo
- Publication number
- CN115689055A CN115689055A CN202211463998.XA CN202211463998A CN115689055A CN 115689055 A CN115689055 A CN 115689055A CN 202211463998 A CN202211463998 A CN 202211463998A CN 115689055 A CN115689055 A CN 115689055A
- Authority
- CN
- China
- Prior art keywords
- sequence
- meteorological
- dimensional
- irradiance
- matrix
- 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
Images
Landscapes
- Complex Calculations (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a short-term solar irradiance prediction method and a short-term solar irradiance prediction device, wherein the prediction method comprises the following steps: collecting data; decomposing an original irradiance sequence into multi-scale modal components by using an ICEEMDAN algorithm, combining the multi-scale modal components, and constructing a multi-dimensional radiation characteristic sequence capable of reflecting irradiance change characteristics; constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to time steps based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence; a residual error attention mechanism is introduced to reconstruct the two-dimensional meteorological feature matrix to obtain a novel meteorological feature matrix; respectively extracting the time sequence characteristics of the two-dimensional radiation characteristic matrix and the novel meteorological characteristic matrix, and fusing; and fusing the obtained time sequence characteristics as the input of the multilayer perceptron to predict the short-term solar irradiance. The method can improve the prediction precision of the short-term solar irradiance.
Description
Technical Field
The invention belongs to the technical field of photovoltaic power generation, and relates to a short-term solar irradiance prediction method and device.
Background
The output power of the photovoltaic power generation system has volatility and intermittence, and safe and stable operation of the power system during photovoltaic grid connection is not facilitated. The solar irradiance is a main factor influencing the photovoltaic power generation power, and the solar irradiance is accurately predicted, so that the accurate prediction of the photovoltaic power generation power is facilitated.
In recent years, artificial intelligence methods based on machine learning methods such as artificial neural networks and random forests and deep learning methods such as convolutional neural networks and long-short term memory networks have been widely used in the field of solar radiation prediction.
However, the existing short-term solar irradiance prediction method is difficult to capture the fluctuation and mutation of an irradiance sequence, so that the prediction accuracy is low. Researches show that the solar irradiance is closely related to meteorological characteristics, and the meteorological characteristics are reasonably utilized to improve the prediction precision. Most of the existing prediction methods artificially select meteorological features according to correlation coefficients or experiences, are relatively complex, neglect the influence of different meteorological features on different degrees of prediction tasks, lack rationality and further influence prediction precision.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide a short-term solar irradiance prediction method and device, which can improve the short-term solar irradiance prediction precision.
The technical scheme is as follows: the invention provides a short-term solar irradiance prediction method, which comprises the following steps:
data acquisition, including radiation data and meteorological data, wherein the radiation data is the total irradiance of a horizontal plane and comprises an irradiance sequence, and the meteorological data comprises a multi-dimensional meteorological characteristic sequence;
decomposing an original irradiance sequence into multi-scale modal components by using an ICEEMDAN algorithm, combining the multi-scale modal components, and constructing a multi-dimensional radiation characteristic sequence capable of reflecting irradiance change characteristics;
constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to time steps based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence;
a residual error attention mechanism is introduced to reconstruct the two-dimensional meteorological feature matrix to obtain a novel meteorological feature matrix;
respectively extracting time sequence characteristics of the two-dimensional radiation characteristic matrix and the novel meteorological characteristic matrix, and fusing;
and fusing the obtained time sequence characteristics as the input of the multilayer perceptron to predict the short-term solar irradiance.
Further, the collected meteorological data includes solar zenith angle, temperature, type of cloud, dew point temperature, wind direction, wind speed, relative humidity and water reducible amount.
Further, decomposing the original irradiance sequence by using an ICEEMDAN algorithm comprises the following steps:
defining an original irradiance sequence as s;
constructing a new sequence based on the sequence s:
s i ＝s+α 0 E 1 (w i )
wherein s is i For a new sequence constructed after adding i sets of white noise, w i To add i groups of white noise to the sequence s, E k (. H) represents the k-order modal components decomposed by an empirical mode decomposition algorithm;
calculating to obtain a first group of residual errors R 1 ：
R 1 ＝<M(s i )>
Wherein < · > represents averaging over the whole; m (-) is a local mean of the sequence generated based on the empirical mode decomposition algorithm;
calculating to obtain a first modal component IMF 1 ：
IMF 1 ＝s-R 1
At the obtained IMF of the first modal component 1 On the basis of the first group of residual errors, white noise is continuously added, and a second group of residual errors R are calculated by utilizing local mean decomposition 2 And a second modal component IMF 2 ：
R 2 ＝<M(R 1 +α 1 E 2 (w i ))>
IMF 2 ＝R 1 -R 2
By analogy, the k group residual error R k And the k-th modal component IMF k Comprises the following steps:
R k ＝<M(R k-1 +α k-1 E k (w i ))>
IMF k ＝R k-1 -R k
repeating the calculation process of the residual error and the modal component until the calculation is finished to obtain all the modal components and the final residual error;
alpha in the above formula k Expressed as:
wherein epsilon 0 The inverse of the signal-to-noise ratio between the Gaussian white noise sequence with the average value of 0 and the analyzed original irradiance sequence which are added for the first time; std represents a standard deviation;
and combining and decomposing the obtained modal components and residual errors of different modes to obtain a multi-dimensional radiation characteristic sequence capable of reflecting the variation characteristic of irradiance.
Further, decomposing the original irradiance sequence by using an empirical mode decomposition algorithm comprises: continuously (1) solving the mean value of the upper envelope line and the lower envelope line of the sequence; (2) subtracting the mean envelope curve of the original sequence; (3) Repeating iteration until the obtained sequence meets two constraint conditions of the inherent modal function; an IMF component is obtained at this time, and the local mean value refers to a part obtained by subtracting the IMF from the original sequence; based on the obtained i groups of local mean values, the integral mean value is calculated to obtain a first group of residual errors R 1 。
Further, reconstructing the two-dimensional meteorological feature matrix by using a residual attention mechanism, wherein the method comprises the following steps:
acquiring an attention weight matrix based on the two-dimensional meteorological feature matrix:
A＝σ(W 2 (δ(W 1 X+b 1 ))+b 2 )
wherein A represents an attention weight matrix obtained based on two-dimensional meteorological features; sigma is a sigmoid activation function; w 1 、W 2 Representing an updatable weight matrix; δ is the ReLU activation function;
the method comprises the steps of obtaining a two-dimensional meteorological characteristic matrix, wherein T is the time step number, F is the input characteristic number, and the two-dimensional meteorological characteristic matrix represents the measured values of F meteorological characteristics in the past T time steps; b 1 、b 2 The bias corresponding to the updatable weight matrix;
adding attention weight to the two-dimensional meteorological feature matrix:
X att ＝A⊙X
wherein, X att The meteorological characteristic matrix after attention weight is introduced; an indication of a hadamard product;
introducing residual connection:
X'＝X+X att
and X' is the novel meteorological feature matrix reconstructed by residual attention.
Further, the time sequence characteristics of the two-dimensional radiation characteristic matrix and the time sequence characteristics of the novel meteorological characteristic matrix are respectively extracted by utilizing the stacked long-term and short-term memory network.
Further, the time sequence characteristics of the two-dimensional radiation characteristic matrix and the time sequence characteristics of the novel meteorological characteristic matrix are fused by using a concatenate operation.
A second aspect of the present invention provides a short-term solar irradiance prediction apparatus, comprising:
the system comprises a data acquisition module, a data acquisition module and a data processing module, wherein the data acquisition module is used for acquiring radiation data and meteorological data, the radiation data is the total irradiance of a horizontal plane and comprises an irradiance sequence, and the meteorological data comprises a multi-dimensional meteorological characteristic sequence;
the data processing module is used for decomposing the original irradiance sequence into multi-scale modal components through an ICEEMDAN algorithm, combining the multi-scale modal components and constructing a multi-dimensional radiation characteristic sequence capable of reflecting irradiance change characteristics; constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to time steps based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence; reconstructing the two-dimensional meteorological feature matrix by utilizing a residual error attention mechanism to obtain a novel meteorological feature matrix;
the time sequence characteristic extraction module is used for respectively extracting time sequence characteristics of the two-dimensional radiation characteristic matrix and the novel meteorological characteristic matrix and fusing the time sequence characteristics;
and the multilayer perceptron is used for taking the fused time sequence characteristics as input to predict the short-term solar irradiance.
A third aspect of the present invention provides a short-term solar irradiance prediction apparatus, comprising:
one or more processors; and one or more reservoirs;
wherein one or more programs are stored in the one or more memories, which when executed by the one or more processors implement the prediction method of the first aspect described above.
A fourth aspect of the present invention provides a computer storage medium having stored therein computer instructions which, when executed, implement the prediction method of the first aspect described above.
A fifth aspect of the present invention provides a computer program product for causing a computer to perform the prediction method of the first aspect when the computer program product is run on a computer.
Has the advantages that: compared with the prior art, the invention has the following remarkable advantages:
(1) The ICEEMDAN algorithm is used for decomposing the original irradiance sequence, so that more multi-scale modal components representing irradiance change characteristics can be obtained, the influence of fluctuation and mutation of the original irradiance sequence on solar radiation prediction is reduced, and the accuracy and reliability of a prediction result are improved.
(2) A residual error attention mechanism is introduced to reconstruct the two-dimensional meteorological feature matrix, the importance degree of different meteorological features in prediction can be fully considered, loss of original meteorological feature information is avoided, solar radiation prediction precision is improved, and high robustness and feasibility are achieved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments of the present invention are briefly described below, and it is obvious that the drawings described below are only embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a block diagram of a prediction method of the present invention;
FIG. 2 is a block diagram of the prediction apparatus of the present invention;
FIG. 3 is a decomposition result graph of an original irradiance sequence;
FIG. 4 is a schematic illustration of residual attention being applied to an original meteorological feature;
FIG. 5 is a graph comparing RMSE and MAE error indicators for different models;
FIG. 6 is a graph showing model curve fitting under different degrees of irradiance fluctuation.
Detailed Description
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 not all embodiments. 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 scope of protection of the present invention.
Fig. 1 is a block diagram of a short-term solar irradiance prediction method provided in an embodiment of the present application, where the prediction method specifically includes the following steps:
(1) Data acquisition, including radiation data and meteorological data;
the radiation data is an irradiance sequence comprising horizontal plane total irradiance (GHI), and the meteorological data is a multidimensional meteorological feature sequence comprising solar zenith angle, temperature, type of cloud, dew point temperature, wind direction, wind speed, relative humidity and water reducible quantity.
(2) Decomposing an original irradiance sequence into multi-scale modal components by using an ICEEMDAN (improved adaptive noise complete set empirical mode decomposition) algorithm, combining the multi-scale modal components, and constructing a multi-dimensional radiation characteristic sequence capable of reflecting irradiance change characteristics;
the step (2) specifically comprises the following steps:
defining an original irradiance sequence as s;
constructing a new sequence based on the sequence s:
s i ＝s+α 0 E 1 (w i )
wherein s is i For a new sequence constructed after adding i sets of white noise, w i To add i groups of white noise to the sequence s, E k (. H) represents the k-order modal components decomposed by an empirical mode decomposition algorithm;
calculating to obtain a first group of residual errors R 1 ：
R 1 ＝<M(s i )>
Wherein the content of the first and second substances,<·>means averaging over the whole; m (-) is a local mean value of a sequence generated based on an empirical mode decomposition algorithm, and specifically, an original irradiance sequence is decomposed by using the empirical mode decomposition algorithm: continuously(1) Calculating the mean values of the upper envelope line and the lower envelope line of the sequence; (2) subtracting the mean envelope curve of the original sequence; (3) Repeating iteration until the obtained sequence meets two constraint conditions of the inherent modal function; an IMF component is obtained at this time, and the local mean value refers to a part obtained by subtracting the IMF from the original sequence; based on the obtained i groups of local mean values, the i groups of local mean values are averaged in a whole way to obtain a first group of residual errors R 1 。
Calculating to obtain a first modal component IMF 1 ：
IMF 1 ＝s-R 1
At the obtained IMF of the first modal component 1 On the basis of the first group of residual errors, white noise is continuously added, and a second group of residual errors R are calculated by utilizing local mean decomposition 2 And a second modal component IMF 2 ：
R 2 ＝<M(R 1 +α 1 E 2 (w i ))>
IMF 2 ＝R 1 -R 2
By analogy, the k group residual error R k And the k-th modal component IMF k Comprises the following steps:
R k ＝<M(R k-1 +α k-1 E k (w i ))>
IMF k ＝R k-1 -R k
repeating the calculation process of the residual error and the modal component until the calculation is finished to obtain all the modal components and the final residual error;
alpha in the above formula k Expressed as:
wherein epsilon 0 The inverse of the signal-to-noise ratio between the Gaussian white noise sequence with the average value of 0 and the analyzed original irradiance sequence which are added for the first time; std represents a standard deviation;
and combining and decomposing the obtained modal components and residual errors of different modes to obtain a multi-dimensional radiation characteristic sequence capable of reflecting the variation characteristic of irradiance.
(3) Constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to time steps based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence;
(4) Reconstructing the two-dimensional meteorological feature matrix by using a residual error attention (RA) mechanism to obtain a novel meteorological feature matrix;
the step (4) specifically comprises the following steps:
acquiring an attention weight matrix based on the two-dimensional meteorological feature matrix:
A＝σ(W 2 (δ(W 1 X+b 1 ))+b 2 )
wherein A represents an attention weight matrix obtained based on two-dimensional meteorological features; sigma is a sigmoid activation function; w 1 、W 2 Representing an updatable weight matrix; δ is the ReLU activation function;
the method comprises the steps of obtaining a two-dimensional meteorological characteristic matrix, wherein T is the time step number, F is the input characteristic number, and the two-dimensional meteorological characteristic matrix represents the measured values of F meteorological characteristics in the past T time steps; b 1 、b 2 The bias corresponding to the updatable weight matrix;
as shown in FIG. 3, attention weights are added to the two-dimensional meteorological feature matrix:
X att ＝A⊙X
wherein, X att The meteorological characteristic matrix after attention weight is introduced; an h-Dama product;
introducing residual connection:
X'＝X+X att
and X' is the novel meteorological feature matrix reconstructed by residual attention.
(4) Respectively extracting the time sequence characteristics of the two-dimensional radiation characteristic matrix and the time sequence characteristics of the novel meteorological characteristic matrix by using a stacked long-short term memory network (LSTM);
and fusing the time sequence characteristics of the multi-two-dimensional radiation characteristic matrix and the time sequence characteristics of the novel meteorological characteristic matrix by using a configure operation.
(5) And (3) using the fused time sequence characteristics as the input of a multilayer perceptron (MLP) to predict the short-term horizontal-plane total irradiance.
The embodiment of the present application further provides a short-term solar irradiance prediction device, including:
the system comprises a data acquisition module, a data acquisition module and a data processing module, wherein the data acquisition module is used for acquiring radiation data and meteorological data, the radiation data is the total irradiance of a horizontal plane and comprises an irradiance sequence, and the meteorological data comprises a multi-dimensional meteorological characteristic sequence;
the data processing module is used for decomposing the original irradiance sequence into multi-scale modal components through an ICEEMDAN algorithm, combining the multi-scale modal components and constructing a multi-dimensional radiation characteristic sequence capable of reflecting irradiance change characteristics; constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to time steps based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence; reconstructing the two-dimensional meteorological feature matrix by utilizing a residual error attention mechanism to obtain a novel meteorological feature matrix;
the time sequence characteristic extraction module is used for respectively extracting time sequence characteristics of the two-dimensional radiation characteristic matrix and the novel meteorological characteristic matrix and fusing the time sequence characteristics;
and the multilayer perceptron is used for taking the fused time sequence characteristics as input to predict the short-term solar irradiance.
The embodiment of the present application further provides another short-term solar irradiance prediction device, including:
one or more processors; and one or more reservoirs;
wherein the one or more programs are stored in the one or more memories and, when the one or more programs are executed by the one or more processors, implement the prediction method in the above-described embodiments.
The embodiment of the present application further provides a computer storage medium, in which computer instructions are stored, and when the computer instructions are executed, the prediction method in the foregoing embodiment is implemented.
The computer storage medium may be, for example, a U disk, a removable hard disk, a ROM memory, a RAM memory, a magnetic or optical disk, or any other medium that can be used to store computer instructions.
The embodiment of the present application further provides a computer program product, which when running on a computer, causes the computer to execute the above related steps, so as to implement the prediction method in the above embodiment.
The prediction device, the computer storage medium and the computer program product provided by the embodiment of the application are all used for executing the prediction method provided by the above, so that the prediction device, the computer storage medium and the computer program product have the same beneficial effects as the prediction method.
The implementation process of the prediction method for short-term solar radiation prediction is described in detail below with reference to specific examples.
The radiation data and meteorological data of Nanjing city of Jiangsu province are selected for testing, specifically data of 1 month and 1 day in 2016 to 12 months and 31 days in 2020, and the time interval of data acquisition is 1 hour, and 43800 samples are counted. The data from 1/2016 to 31/12/2019 are used as training samples to train the model, and the data from 1/2020 to 31/12/2020 are used as test samples to evaluate the model performance.
The collected data comprises radiation data and meteorological data, the radiation data is total irradiance of a horizontal plane, and the meteorological data comprises solar zenith angle, temperature, cloud type, dew point temperature, wind direction, wind speed, relative humidity and water reducible amount.
The original irradiance sequence is decomposed into multi-scale modal components using the iceelmdan algorithm.
As shown in fig. 2, the decomposition result includes 15 Intrinsic Mode Functions (IMF) and residuals (Res), wherein the high frequency component corresponds to the component with larger volatility and catastrophe in the original irradiance sequence, the low frequency component corresponds to the component with stronger regularity in the original irradiance sequence, and the 16-dimensional feature sequence capable of reflecting the irradiance variation characteristic can be obtained by combining the 15 intrinsic mode functions and residuals.
And constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to the time step based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence. In actual testing, the number of time steps was set at 48, i.e., the multi-dimensional radiation and meteorological features over the past 48 hours.
And reconstructing the two-dimensional meteorological feature matrix by using a residual error attention mechanism to obtain a novel meteorological feature matrix.
Respectively extracting the time sequence characteristics of the two-dimensional radiation characteristic matrix and the time sequence characteristics of the novel meteorological characteristic matrix by using a stacked long-short term memory network (LSTM); and fusing the time sequence characteristics of the two-dimensional radiation characteristic matrix and the time sequence characteristics of the novel meteorological characteristic matrix by using the coordinate operation.
And (3) taking the fused time sequence characteristics as the input of a multilayer perceptron (MLP), predicting the total irradiance of a short-term horizontal plane, and outputting a final prediction result.
In order to verify the performance of the prediction method provided by the embodiment of the application, the prediction effect of the model established according to the prediction method is evaluated based on the test sample, the evaluation indexes of the selected model are Root Mean Square Error (RMSE), absolute mean error (MAE) and correlation coefficient (R), and the calculation formulas are respectively as follows:
wherein n represents the total number of test samples,
and y i Respectively representing the predicted value and the actual value of the ith sample,
and y a The predicted mean and the actual mean are indicated separately.
In order to further evaluate the prediction performance of the prediction model, a total of six comparison models are set, namely MLP, LSTM-ANN, bi-LSTM, CNN-Bi-LSTM and ICEEMDAN-LSTM, wherein the LSTM-ANN is added with an ANN part on the basis of the LSTM, and the aim of increasing the network depth is to improve the nonlinear fitting capability of the model; the proposed model iceelmdan-LSTM does not enforce residual attention on the original meteorological features, but rather uses it directly as an input to the model.
Table 1 shows the prediction error of each model when the horizontal total irradiance is predicted 1 hour ahead, and fig. 5 shows the comparison between the model constructed according to the present invention and the errors of the other models RMSE and MAE.
TABLE 1 comparison of the predicted results of different models
As can be seen from table 1, the prediction method of the present invention has the highest prediction accuracy no matter which evaluation index is based on, and compared with a method of decomposing an original irradiation sequence without using icemdan, the method of decomposing an original irradiation sequence using icemdan has an obviously improved prediction effect, which can be clearly seen in fig. 5. The prediction method implements residual attention on the original meteorological characteristics on the basis of decomposing the original irradiance sequence, and obtains the minimum prediction error.
To further evaluate the predicted performance of the prediction models, fig. 6 shows the fitting of the prediction curves of each model to the actual irradiance curve under different weather conditions. It can be seen that for the case of small irradiance fluctuation, most of the predicted value curves of the model can be well fitted with the actual value curves, and particularly, the predicted value curves based on ICEEMDAN-RA-LSTM in the invention have the minimum deviation from the actual value curves in the stages of irradiance value rising and falling and have the best curve fitting effect. For the situation that irradiance fluctuates greatly, the predicted value curves and the actual value curves of most models have large deviation, but the whole trend of the predicted value curves corresponding to the invention is very close to the change trend of the actual irradiance values, because the decomposition result based on ICEEMDAN can obtain irradiance components with remarkable fluctuation, the models are suitable for prediction under different weather conditions and different fluctuation degrees of the irradiance values.
In conclusion, the invention utilizes ICEEMDAN to construct a multi-dimensional characteristic sequence capable of reflecting the variation characteristics of the original irradiance sequence, thereby capturing the fluctuation and the mutation of the original irradiance sequence, being applicable to short-term solar irradiance prediction under different weather conditions and different irradiance fluctuation degrees and obtaining better prediction performance. The method can fully consider the importance degree of different meteorological features in prediction, avoids the loss of original meteorological feature information, improves the solar radiation prediction precision, and has higher robustness and feasibility. The prediction result of the method can be used for predicting the photovoltaic power, so that the safe and stable operation of the power system during large-scale photovoltaic grid connection is guaranteed.
The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method of short-term solar irradiance prediction, comprising:
data acquisition, including radiation data and meteorological data, wherein the radiation data is the total irradiance of a horizontal plane and comprises an irradiance sequence, and the meteorological data comprises a multi-dimensional meteorological characteristic sequence;
decomposing an original irradiance sequence into multi-scale modal components by using an ICEEMDAN algorithm, combining the multi-scale modal components, and constructing a multi-dimensional radiation characteristic sequence capable of reflecting irradiance change characteristics;
constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to time steps based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence;
a residual attention mechanism is introduced to reconstruct the two-dimensional meteorological feature matrix to obtain a novel meteorological feature matrix;
respectively extracting the time sequence characteristics of the two-dimensional radiation characteristic matrix and the novel meteorological characteristic matrix, and fusing;
and fusing the obtained time sequence characteristics as the input of the multilayer perceptron to predict the short-term solar irradiance.
2. The prediction method of claim 1, wherein the collected meteorological data comprises a solar zenith angle, a temperature, a type of cloud, a dew point temperature, a wind direction, a wind speed, a relative humidity, and a degradable water content.
3. The prediction method of claim 1, wherein decomposing the original irradiance sequence using the icemdan algorithm comprises:
defining an original irradiance sequence as s;
constructing a new sequence based on the sequence s:
s i ＝s+α 0 E 1 (w i )
wherein s is i For a new sequence constructed after adding i sets of white noise, w i To add i groups of white noise to the sequence s, E k (. H) represents the k-order modal components decomposed by an empirical mode decomposition algorithm;
calculating to obtain a first group of residual errors R 1 ：
R 1 ＝<M(s i )>
Wherein < · > represents averaging over the whole; m (-) is a local mean of the sequence generated based on the empirical mode decomposition algorithm;
calculating to obtain a first modal component IMF 1 ：
IMF 1 ＝s-R 1
At the obtained IMF of the first modal component 1 On the basis of the first group of residual errors, white noise is continuously added, and a second group of residual errors R are calculated by utilizing local mean decomposition 2 And a second modal component IMF 2 ：
R 2 ＝<M(R 1 +α 1 E 2 (w i ))>
IMF 2 ＝R 1 -R 2
By analogy, the k group residual error R k And the k-th modal component IMF k Comprises the following steps:
R k ＝<M(R k-1 +α k-1 E k (w i ))>
IMF k ＝R k-1 -R k
repeating the calculation process of the residual error and the modal component until the calculation is finished to obtain all the modal components and the final residual error;
alpha in the above formula k Expressed as:
wherein epsilon 0 The inverse of the signal-to-noise ratio between the Gaussian white noise sequence with the average value of 0 and the analyzed original irradiance sequence which are added for the first time; std represents a standard deviation;
and combining and decomposing the obtained modal components and residual errors of different modes to obtain a multi-dimensional radiation characteristic sequence capable of reflecting the variation characteristic of irradiance.
4. The prediction method of claim 3, wherein decomposing the original irradiance sequence using an empirical mode decomposition algorithm comprises: continuously (1) solving the mean value of the upper envelope line and the lower envelope line of the sequence; (2) subtracting the mean envelope curve of the original sequence; (3) Repeating iteration until the obtained sequence meets two constraint conditions of the inherent modal function; an IMF component is obtained at this time, and the local mean value refers to a part obtained by subtracting the IMF from the original sequence; based on the obtained i groups of local mean values, the i groups of local mean values are averaged in a whole way to obtain a first group of residual errors R 1 。
5. The prediction method of claim 1, wherein reconstructing the two-dimensional meteorological signature matrix using a residual attention mechanism comprises:
acquiring an attention weight matrix based on the two-dimensional meteorological feature matrix:
A＝σ(W 2 (δ(W 1 X+b 1 ))+b 2 )
wherein A represents an attention weight matrix obtained based on two-dimensional meteorological features; sigma is a sigmoid activation function; w 1 、W 2 Representing an updatable weight matrix; δ is the ReLU activation function;
the second is a two-dimensional meteorological characteristic matrix, T is a time step number, F is an input characteristic number, and the two-dimensional meteorological characteristic matrix represents measured values of F meteorological characteristics in the past T time steps; b is a mixture of 1 、b 2 The bias corresponding to the updatable weight matrix;
adding attention weight to the two-dimensional meteorological feature matrix:
X att ＝A⊙X
wherein, X att The meteorological characteristic matrix after attention weight is introduced; an indication of a hadamard product;
introducing residual connection:
X'＝X+X att
and X' is the novel meteorological feature matrix reconstructed by residual attention.
6. The prediction method according to claim 1, wherein the time sequence features of the two-dimensional radiation feature matrix and the time sequence features of the novel meteorological feature matrix are extracted respectively by using a stacked long-short term memory network.
7. The prediction method according to claim 1 or 6, characterized in that the time sequence characteristics of the two-dimensional radiation characteristic matrix and the time sequence characteristics of the novel meteorological characteristic matrix are fused by using a concatenate operation.
8. A short-term solar irradiance prediction apparatus, comprising:
the system comprises a data acquisition module, a data acquisition module and a data processing module, wherein the data acquisition module is used for acquiring radiation data and meteorological data, the radiation data is the total irradiance of a horizontal plane and comprises an irradiance sequence, and the meteorological data comprises a multi-dimensional meteorological characteristic sequence;
the data processing module is used for decomposing the original irradiance sequence into multi-scale modal components through an ICEEMDAN algorithm, combining the multi-scale modal components and constructing a multi-dimensional radiation characteristic sequence capable of reflecting irradiance change characteristics; constructing a two-dimensional radiation characteristic matrix and a two-dimensional meteorological characteristic matrix according to time steps based on the multi-dimensional radiation characteristic sequence and the multi-dimensional meteorological characteristic sequence; reconstructing the two-dimensional meteorological feature matrix by using a residual attention mechanism to obtain a novel meteorological feature matrix;
the time sequence characteristic extraction module is used for respectively extracting time sequence characteristics of the two-dimensional radiation characteristic matrix and the novel meteorological characteristic matrix and fusing the time sequence characteristics;
and the multilayer perceptron is used for taking the fused time sequence characteristics as input to predict the short-term solar irradiance.
9. A short-term solar irradiance prediction apparatus, comprising:
one or more processors; and one or more reservoirs;
wherein one or more programs are stored in the one or more memories, the one or more programs when executed by the one or more processors implement the prediction method of any of claims 1-7.
10. A computer storage medium having computer instructions stored thereon which, when executed, implement the prediction method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211463998.XA CN115689055A (en) | 2022-11-22 | 2022-11-22 | Short-term solar irradiance prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211463998.XA CN115689055A (en) | 2022-11-22 | 2022-11-22 | Short-term solar irradiance prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115689055A true CN115689055A (en) | 2023-02-03 |
Family
ID=85053394
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211463998.XA Pending CN115689055A (en) | 2022-11-22 | 2022-11-22 | Short-term solar irradiance prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115689055A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116167465A (en) * | 2023-04-23 | 2023-05-26 | 杭州经纬信息技术股份有限公司 | Solar irradiance prediction method based on multivariate time series ensemble learning |
| CN117272002A (en) * | 2023-11-23 | 2023-12-22 | 中国电建集团西北勘测设计研究院有限公司 | Solar radiation amount estimation method and device, electronic equipment and storage medium |
-
2022
- 2022-11-22 CN CN202211463998.XA patent/CN115689055A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116167465A (en) * | 2023-04-23 | 2023-05-26 | 杭州经纬信息技术股份有限公司 | Solar irradiance prediction method based on multivariate time series ensemble learning |
| CN117272002A (en) * | 2023-11-23 | 2023-12-22 | 中国电建集团西北勘测设计研究院有限公司 | Solar radiation amount estimation method and device, electronic equipment and storage medium |
| CN117272002B (en) * | 2023-11-23 | 2024-02-20 | 中国电建集团西北勘测设计研究院有限公司 | Solar radiation amount estimation method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109802430B (en) | Wind power grid control method based on LSTM-Attention network | |
| CN109272156B (en) | Ultra-short-term wind power probability prediction method | |
| CN108304623B (en) | Probability load flow online calculation method based on stack noise reduction automatic encoder | |
| CN115689055A (en) | Short-term solar irradiance prediction method and device | |
| CN108336739B (en) | RBF neural network-based probability load flow online calculation method | |
| CN104573876A (en) | Wind power plant short-period wind speed prediction method based on time sequence long memory model | |
| CN111144644B (en) | Short-term wind speed prediction method based on variation variance Gaussian process regression | |
| CN112036595B (en) | Short-term wind power prediction method and system based on multi-position numerical weather forecast | |
| CN115186923A (en) | Photovoltaic power generation power prediction method and device and electronic equipment | |
| CN114498619A (en) | Wind power prediction method and device | |
| CN117172589A (en) | Industry typical user power demand response potential evaluation method | |
| CN114897204A (en) | Method and device for predicting short-term wind speed of offshore wind farm | |
| CN111967652B (en) | Double-layer collaborative real-time correction photovoltaic prediction method | |
| CN113642255A (en) | Photovoltaic power generation power prediction method based on multi-scale convolution cyclic neural network | |
| CN116739172B (en) | Method and device for ultra-short-term prediction of offshore wind power based on climbing identification | |
| CN113111592A (en) | Short-term wind power prediction method based on EMD-LSTM | |
| Su et al. | A LSTM based wind power forecasting method considering wind frequency components and the wind turbine states | |
| CN115640916B (en) | Distributed power supply output and multi-energy load situation sensing method and system | |
| CN115907131B (en) | Method and system for constructing electric heating load prediction model in northern area | |
| CN116090216A (en) | Power generation power prediction method and system based on typical correlation analysis | |
| CN112070320A (en) | Ultra-short-term wind power prediction method and system based on dynamic harmonic regression | |
| CN111861259A (en) | Load modeling method, system and storage medium considering time sequence | |
| CN113722970B (en) | Photovoltaic power ultra-short-term online prediction method | |
| Huang et al. | Probabilistic prediction intervals of wind speed based on explainable neural network | |
| Hanif et al. | Towards Optimal Solar Energy Integration: A Deep Dive into AI-Enhanced Solar Irradiance Forecasting Models |
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 |