CN115826092A - Multi-meteorological-element prediction method and system based on information self-attention model - Google Patents
Multi-meteorological-element prediction method and system based on information self-attention model Download PDFInfo
- Publication number
- CN115826092A CN115826092A CN202211350817.2A CN202211350817A CN115826092A CN 115826092 A CN115826092 A CN 115826092A CN 202211350817 A CN202211350817 A CN 202211350817A CN 115826092 A CN115826092 A CN 115826092A
- Authority
- CN
- China
- Prior art keywords
- meteorological
- self
- attention
- data
- training
- 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
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention discloses a multi-meteorological-element prediction method and a multi-meteorological-element prediction system based on an information self-attention model, wherein the method comprises the following steps: acquiring weather collection data of a weather station, wherein the weather collection data comprises: the numerical value of a plurality of meteorological elements and corresponding acquisition time, the meteorological elements include: temperature, dew point temperature, solar radiation down, barometric pressure; carrying out normalization processing on the meteorological acquisition data; and based on the information self-attention model, performing medium-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing to obtain a multi-meteorological-element prediction result. By predicting historical meteorological collected data based on the information self-attention model, the problems of high calculation complexity and forecast delay of traditional meteorological data forecasting are solved, and efficient near-real-time forecasting and medium-short term multi-meteorological-element forecasting are realized.
Description
Technical Field
The invention relates to the technical field of meteorological prediction, in particular to a multi-meteorological-element prediction method and system based on an information self-attention model.
Background
Since the 21 st century, with the economic development, the exhaustion of traditional fossil energy and the increase of greenhouse gas emission, renewable energy represented by photovoltaic and wind power begins to become the trend and trend of world energy development. At present, the technical capabilities of renewable energy related meteorological observation, resource evaluation and prediction are improved, and support is provided for renewable energy resource general survey, project development and power system operation. Therefore, accurate meteorological prediction can provide guarantee for accurate and effective prediction of photovoltaic and wind power, and therefore uncertainty of photovoltaic and wind power is improved, and safety and reliability of a power system are improved.
The technical development of weather forecast goes through four stages of civil skills, single station forecast, weather map forecast and numerical forecast, the development process mainly adopts more mathematical and physical methods to replace artificial experience decision in the forecasting process, the most common numerical weather forecast development at present still faces many theoretical and technical challenges, such as initial error, mode error, predictability of weather system evolution, delay of high model complexity forecast, high computation complexity needed by higher space-time resolution forecast and the like, and in addition, the existing accumulated mass numerical weather forecast information is not fully mined and expanded and applied in the traditional numerical forecast product.
Disclosure of Invention
The embodiment of the invention aims to provide a multi-meteorological-element prediction method and a multi-meteorological-element prediction system based on an information self-attention model, which are used for predicting historical meteorological collected data based on the information self-attention model, solve the problems of high computation complexity and prediction delay in the traditional meteorological data prediction, and realize efficient near-real-time prediction and medium-short term multi-meteorological-element prediction.
In order to solve the above technical problem, a first aspect of an embodiment of the present invention provides a multi-meteorological-element prediction method based on an information self-attention model, including the following steps:
acquiring weather collection data of a weather station, wherein the weather collection data comprises: the numerical value of a plurality of meteorological element and corresponding acquisition time, meteorological element includes: temperature, dew point temperature, solar radiation down, barometric pressure;
carrying out normalization processing on the meteorological acquisition data;
and based on the information self-attention model, performing medium-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing to obtain a multi-meteorological-element prediction result.
Further, the information self-attention model includes: an encoder and a decoder;
the encoder is obtained by alternately stacking a one-dimensional convolutional layer and a probability sparse self-attention layer;
the decoder is derived from cross-stacking of probabilistic self-attention layers and self-attention layers.
Further, after the normalization processing is performed on the meteorological collection data, the method further includes:
dividing historical meteorological acquisition data of a preset time length into a training data set, a verification data set and a test data set according to a preset proportion;
determining hyper-parameters and training parameters of the information self-attention model, wherein the hyper-parameters comprise: model convolutional layers, probability sparse self-attention layers and channel parameters and the number of layers of the self-attention layers, wherein the training parameters comprise: learning rate, input batch size, iteration times and model storage conditions;
generating a training data set and a verification data set;
and training the information self-attention model according to the training data set, outputting a precision report and determining an optimal model.
Further, after generating the training data set and the verification data set, the method further includes:
generating Q, K and V through an embedding layer based on the meteorological acquisition data, and entering the probability sparse self-attention layer to extract time sequence characteristics;
concatenating all of the temporal features of the probabilistic sparse self-attention layer as input to the self-attention layer in the encoder;
inputting sequence union and corresponding time variable which comprise the meteorological acquisition data with the length of l before the time of t + i +1 and the sequence union with the length of o which is all 0 into the decoder, and obtaining the final prediction output with the length of l + o through the probability sparse self-attention layer and the self-attention layer in the decoder, wherein the output with the length of o after the time is the prediction result of the corresponding meteorological elements;
calculating loss with the real label, updating parameters according to gradient back transmission, stopping training until the established iteration times are met, the loss does not decrease any more and the performance on the verification data set does not increase any more, and storing the model.
Further, the accuracy evaluation index of the accuracy report includes: determining the coefficient R 2 And a square root error RMSE.
Accordingly, a second aspect of the embodiments of the present invention provides a multi-meteorological-element prediction system based on an information self-attention model, including:
a data acquisition module for acquiring weather collection data of a weather station, the weather collection data comprising: the numerical value of a plurality of meteorological elements and corresponding acquisition time, the meteorological elements include: temperature, dew point temperature, solar radiation down, barometric pressure;
the data processing module is used for carrying out normalization processing on the meteorological acquisition data;
and the data prediction module is used for performing medium-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing based on the information self-attention model to obtain a multi-meteorological-element prediction result.
Further, the information self-attention model includes: an encoder and a decoder;
the encoder is obtained by alternately stacking a one-dimensional convolutional layer and a probability sparse self-attention layer;
the decoder is derived from cross-stacking of probabilistic self-attention layers and self-attention layers.
Further, the multi-meteorological element prediction system based on the information self-attention model further comprises: a data training module, the data training module comprising:
the data dividing unit is used for dividing historical meteorological acquisition data of a preset time length into a training data set, a verification data set and a test data set according to a preset proportion;
a parameter selection unit for determining hyper-parameters and training parameters of the information self-attention model, the hyper-parameters including: the method comprises the following steps of model convolutional layer, probability sparse self-attention layer and channel parameters and layer number of the self-attention layer, wherein the training parameters comprise: learning rate, input batch size, iteration times and model storage conditions;
a data generation unit for generating a training data set and a validation data set;
and the model selection unit is used for training the information self-attention model according to the training data set, outputting a precision report and determining an optimal model.
Further, the data training module further comprises:
the timing sequence feature extraction unit is used for generating Q, K and V through an imbedding layer based on the meteorological acquisition data, entering the probability sparse self-attention layer and extracting timing sequence features;
a temporal feature input unit for concatenating all the temporal features of the probabilistic sparse self-attention layer as input to the self-attention layer in the encoder;
a meteorological element prediction unit for inputting a sequence combination of length l including t + i +1 time-before meteorological acquisition data of length l plus all 0 and length o and corresponding time variable to the decoder, and obtaining a final prediction output of length l + o via the probability sparse self-attention layer and self-attention layer in the decoder, wherein the output of length o is a corresponding meteorological element prediction result;
and the training control unit is used for calculating loss with the real label, reversely transmitting the updated parameters according to the gradient, stopping training until the established iteration times are met, the loss does not decrease any more and the performance on the verification data set does not increase any more, and storing the model.
Further, the accuracy evaluation index of the accuracy report includes: determining the coefficient R 2 And a square root error RMSE.
Accordingly, a third aspect of an embodiment of the present invention provides an electronic device, including: at least one processor; and a memory coupled to the at least one processor; wherein the memory stores instructions executable by the one processor to cause the at least one processor to perform the above-described multi-meteorological element prediction method based on an informational self-attention model.
Accordingly, a fourth aspect of embodiments of the present invention provides a computer-readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the above-described multi-meteorological-element prediction method based on an information self-attention model.
The technical scheme of the embodiment of the invention has the following beneficial technical effects:
by predicting historical meteorological collected data based on the information self-attention model, the problems of high calculation complexity and forecast delay of traditional meteorological data forecasting are solved, and efficient near-real-time forecasting and medium-short term multi-meteorological-element forecasting are realized.
Drawings
FIG. 1 is a flowchart of a multi-meteorological-element prediction method based on an information self-attention model according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a multi-meteorological-element prediction method based on an information self-attention model according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an information self-attention model provided by an embodiment of the invention;
FIG. 4 is a flow chart of information self-attention model training provided by an embodiment of the present invention;
FIG. 5a is a first schematic diagram illustrating a multi-meteorological-element prediction result according to an embodiment of the present invention;
FIG. 5b is a schematic diagram illustrating a prediction result of multiple meteorological elements according to an embodiment of the present invention;
FIG. 5c is a schematic diagram of a multi-meteorological-element prediction result provided by the embodiment of the invention;
FIG. 5d is a diagram illustrating a predicted result of multiple meteorological elements according to an embodiment of the present invention;
FIG. 6 is a block diagram of a multi-meteorological element prediction system module based on an information self-attention model according to an embodiment of the present invention;
fig. 7 is a block diagram of a data training module according to an embodiment of the present invention.
Reference numerals:
1. the system comprises a data acquisition module, a data processing module, a data prediction module, a data training module, a data division unit, a parameter selection unit, a data generation unit, a model selection unit, a time sequence feature extraction unit, a data acquisition module, a data processing module, a data prediction module, a data training module, a data division unit, a parameter selection unit, a data generation unit, a model selection unit, a time sequence feature extraction unit, a time sequence feature input unit, a meteorological element prediction unit, a training control unit and a training control unit.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings in conjunction with the following detailed description. It should be understood that the description is intended to be exemplary only, and is not intended to limit the scope of the present invention. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present invention.
Referring to fig. 1, fig. 2, fig. 3 and fig. 4, a first aspect of the embodiments of the present invention provides a method for predicting multiple meteorological elements based on an information self-attention model, including the following steps:
step S200, acquiring meteorological acquisition data of a meteorological station, wherein the meteorological acquisition data comprises: the numerical value of a plurality of meteorological element and corresponding acquisition time, meteorological element includes: temperature, dew point temperature, solar radiation down, barometric pressure.
And step S400, performing maximum and minimum normalization processing on the meteorological acquisition data.
And S600, performing middle-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing based on the information self-attention model to obtain multi-meteorological-element prediction results.
Specifically, the information self-attention model includes: an encoder and a decoder; the encoder is obtained by alternately stacking one-dimensional convolutional layers and probability sparse self-attention layers; the decoder is obtained by cross-stacking the probability self-attention layer and the self-attention layer.
The information self-attention structure includes:
(1) Probabilistic sparsity self-attention:
the self-attention formula of the self-attention model is as follows: for queue Q, key value K and value V, the corresponding attention is:
where A denotes the attention of the ith dimension of the queue, q i For the ith dimension of the queue Q, T represents the matrix transposition, d is the input dimension, and k is the asymmetric index kernell is the length of the key value K, K j Is the j dimension, v, of the key value K j For the value V dimension j, i is the dimension of the queue and j is the dimension of the key value.
This attention is the binding value and the need to be based on the calculated probabilityObtaining an output whose computation amount requires a power dot product computation and whose storage space complexity is O (L) Q L K ) Wherein L is Q Is the length of the queue, L K Is the length of the key value, and the difficulty of model prediction is increased due to the increase of the calculation amount and storage when the length of the prediction sequence needs to be increased. The probability sparsity self-attention can solve the problem by firstly carrying out queue sparsity evaluationIn the above formula, the attention of the ith queue on all keys is given by p (k) j ,q i ) By definition, the output is the sum of the product of the probability and the corresponding value, and in fact, the attention is to encourage the probability distribution of the queue not to be uniform, so that the important queue can be distinguished to carry out simplified calculation, and the KL divergence is used to measure the importance of the queue, and the formula is as follows:
wherein the first item is the ith dimension q of the queue i The logarithms and expressions over all the key values, the second term is their arithmetic mean. ln is logarithm, L K To indicate the length of the key value K.
Based on the importance index, probability sparse attention is obtained as follows:
whereinThe sparse matrix is the same as q in size, only comprises a front top-k queue according to the importance, and k is a self-defined parameter.
Further, after the normalization processing is performed on the weather collection data in step S400, the method further includes:
step S510, dividing historical meteorological acquisition data of a preset time length into a training data set, a verification data set and a test data set according to a preset proportion.
Step S520, determining hyper-parameters and training parameters of the information self-attention model, wherein the hyper-parameters comprise: the channel parameters and the number of layers of the model convolution layer, the probability sparse self-attention layer and the self-attention layer, and the training parameters comprise: learning rate, input batch size, iteration times and model storage conditions.
Model input and true output tag generation: for the t-th time, the meteorological data and the time variable from t to t + i and the time variable from t + i +1 to t + i + o +1 are input, the meteorological data with the real label from t + i +1 to t + i + o +1 is output, and the meteorological data is generated through a sliding window according to the length s. Wherein i is the time length of the input meteorological collection data sequence, and o is the time length of the output meteorological collection prediction sequence.
In step S530, a training data set and a verification data set are generated.
And S540, training the information self-attention model according to the training data set, outputting a precision report and determining an optimal model.
Further, after generating the training data set and the verification data set in step S530, the method further includes:
and S531, generating Q, K and V through an embedding layer based on meteorological acquisition data, and entering a probability sparse self-attention layer to extract time sequence characteristics.
Step S532, all the timing characteristics of the probability sparse self-attention layer are connected as input to the self-attention layer in the encoder.
Step S533, inputting sequence union including the meteorological acquisition data with length l before time t + i +1 plus all 0 sequences with length o and corresponding time variables into a decoder, and obtaining a final prediction output with length l + o through the probability sparse self-attention layer and the self-attention layer in the decoder, wherein the output with the length o is the corresponding meteorological element prediction result.
And step S534, calculating loss with the real label, reversely transmitting the updated parameters according to the gradient, stopping training until the established iteration times are met, the loss does not decrease any more and the performance on the verification data set does not increase any more, and storing the model.
Referring to fig. 5a, fig. 5b, fig. 5c, and fig. 5d, it can be seen that: the accuracy statistics of the temperature, the dew point temperature, the solar radiation downward direction and the atmospheric pressure within the future 48 hours and hour by hour are respectively carried out on the test set, and from the view point of the figure, the accuracy of the information self-attention model is slightly reduced along with the time, but the overall accuracy is better, and the fact that the information self-attention model can realize accurate multi-meteorological-element prediction is verified.
The block diagram of the model building part in fig. 2 is a schematic diagram, and the specific building process is shown in fig. 3.
In addition, the accuracy evaluation indexes of the accuracy report include: determining the coefficient R 2 And a square root error RMSE.
Accordingly, referring to fig. 6, a second aspect of the embodiments of the present invention provides a multi-meteorological-element prediction system based on an information self-attention model, including:
the data acquisition module 1 is used for acquiring meteorological acquisition data of a meteorological station, wherein the meteorological acquisition data comprises: the numerical value of a plurality of meteorological element and corresponding acquisition time, meteorological element includes: temperature, dew point temperature, solar radiation down, barometric pressure;
the data processing module 2 is used for carrying out normalization processing on the meteorological acquisition data;
and the data prediction module 3 is used for performing medium-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing based on the information self-attention model to obtain a multi-meteorological-element prediction result.
Further, the information self-attention model includes: an encoder and a decoder; the encoder is obtained by alternately stacking one-dimensional convolutional layers and probability sparse self-attention layers; the decoder is obtained by cross-stacking the probability self-attention layer and the self-attention layer.
Further, referring to fig. 7, the multi-meteorological-element prediction system based on the information self-attention model further includes: data training module 4, data training module 4 includes:
a data dividing unit 41, configured to divide historical meteorological collection data of a preset time length into a training data set, a verification data set, and a test data set according to a preset ratio;
a parameter selecting unit 42, configured to determine hyper-parameters and training parameters of the information self-attention model, where the hyper-parameters include: the channel parameters and the number of layers of the model convolution layer, the probability sparse self-attention layer and the self-attention layer, and the training parameters comprise: learning rate, input batch size, iteration times and model storage conditions;
a data generation unit 43 for generating a training data set and a validation data set;
and the model selection unit 44 is used for training the information self-attention model according to the training data set, outputting a precision report and determining an optimal model.
Further, referring to fig. 7, the data training module 4 further includes:
a time sequence feature extraction unit 45, configured to generate Q, K, V through an embedding layer based on weather collection data, enter a probability sparse self-attention layer, and extract time sequence features;
a timing feature input unit 46 for connecting all timing features of the probability sparse self-attention layer as input of the self-attention layer in the encoder;
a meteorological element prediction unit 47, configured to input, to a decoder, a sequence union including meteorological acquisition data of length l before time t + i +1 and sequence unions of length o of all 0 and corresponding time variables, and obtain a final prediction output of length l + o via probability sparse self-attention layers and self-attention layers in the decoder, where the output of length o is a corresponding meteorological element prediction result;
a training control unit 48 for calculating losses with the real tags, passing back the updated parameters according to the gradient, stopping training until the specified number of iterations is reached, the losses do not decrease and the performance on the validation dataset does not increase, and saving the model.
In addition, the accuracy evaluation indexes of the accuracy report include: determining the coefficient R 2 And a square root error RMSE.
Accordingly, a third aspect of an embodiment of the present invention provides an electronic device, including: at least one processor; and a memory coupled to the at least one processor; wherein the memory stores instructions executable by the processor to cause the processor to perform the multi-meteorological element prediction method based on the information self-attention model.
Accordingly, a fourth aspect of embodiments of the present invention provides a computer-readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the above-described multi-meteorological-element prediction method based on an information self-attention model.
The embodiment of the invention aims to protect a multi-meteorological-element prediction method and a multi-meteorological-element prediction system based on an information self-attention model, wherein the method comprises the following steps: acquiring weather collection data of a weather station, wherein the weather collection data comprises: the numerical value of a plurality of meteorological elements and corresponding acquisition time, the meteorological elements include: temperature, dew point temperature, solar radiation down, barometric pressure; carrying out normalization processing on the meteorological acquisition data; and on the basis of the information self-attention model, performing medium-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing to obtain a multi-meteorological-element prediction result. The technical scheme has the following effects:
by predicting historical meteorological collected data based on the information self-attention model, the problems of high calculation complexity and forecast delay of traditional meteorological data forecasting are solved, and efficient near-real-time forecasting and medium-short term multi-meteorological-element forecasting are realized.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (12)
1. A multi-meteorological-element prediction method based on an information self-attention model is characterized by comprising the following steps:
acquiring weather collection data of a weather station, wherein the weather collection data comprises: the numerical value of a plurality of meteorological elements and corresponding acquisition time, the meteorological elements include: temperature, dew point temperature, solar radiation down, barometric pressure;
carrying out normalization processing on the meteorological acquisition data;
and on the basis of the information self-attention model, performing medium-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing to obtain a multi-meteorological-element prediction result.
2. The method of multi-meteorological-element prediction based on an information self-attention model according to claim 1,
the information self-attention model comprises: an encoder and a decoder;
the encoder is obtained by alternately stacking a one-dimensional convolution layer and a probability sparse self-attention layer;
the decoder is derived from cross-stacking of probabilistic self-attention layers and self-attention layers.
3. The method of predicting multiple meteorological elements based on an informational self-attention model according to claim 2, wherein after normalizing the meteorological collection data, further comprising:
dividing historical meteorological acquisition data of a preset time length into a training data set, a verification data set and a test data set according to a preset proportion;
determining hyper-parameters and training parameters of the information self-attention model, wherein the hyper-parameters comprise: model convolutional layers, probability sparse self-attention layers and channel parameters and the number of layers of the self-attention layers, wherein the training parameters comprise: learning rate, input batch size, iteration times and model storage conditions;
generating a training data set and a verification data set;
and training the information self-attention model according to the training data set, outputting a precision report and determining an optimal model.
4. The method of claim 3, wherein the generating the training dataset and the validation dataset further comprises:
generating Q, K and V through an embedding layer based on the meteorological acquisition data, and entering the probability sparse self-attention layer to extract time sequence characteristics;
concatenating all of the temporal features of the probabilistic sparse self-attention layer as input to the self-attention layer in the encoder;
inputting sequence union which comprises the meteorological acquisition data with the length of l before the time of t + i +1 and the sequence union with the length of o which is all 0 and corresponding time variable into the decoder, and obtaining the final prediction output with the length of l + o through the probability sparse self-attention layer and the self-attention layer in the decoder, wherein the output with the length of o is the prediction result of the corresponding meteorological element;
calculating loss with the real label, updating parameters according to gradient back transmission, stopping training until the established iteration times are met, the loss does not decrease any more and the performance on the verification data set does not increase any more, and storing the model.
5. The method of multi-meteorological-element prediction based on an information self-attention model according to claim 3,
the accuracy evaluation indexes of the accuracy report comprise: determining the coefficient R 2 And a square root error RMSE.
6. A multi-meteorological element prediction system based on an information self-attention model is characterized by comprising:
a data acquisition module for acquiring weather collection data of a weather station, the weather collection data comprising: the numerical value of a plurality of meteorological elements and corresponding acquisition time, the meteorological elements include: temperature, dew point temperature, solar radiation down, barometric pressure;
the data processing module is used for carrying out normalization processing on the meteorological acquisition data;
and the data prediction module is used for performing medium-short term multi-meteorological-element prediction according to the meteorological acquisition data after normalization processing based on the information self-attention model to obtain a multi-meteorological-element prediction result.
7. The multi-meteorological-element prediction system based on an information self-attention model according to claim 6,
the information self-attention model comprises: an encoder and a decoder;
the encoder is obtained by alternately stacking a one-dimensional convolutional layer and a probability sparse self-attention layer;
the decoder is derived from cross-stacking of probabilistic self-attention layers and self-attention layers.
8. The information-based self-attention model multi-meteorological element prediction system of claim 7, further comprising: a data training module, the data training module comprising:
the data dividing unit is used for dividing historical meteorological acquisition data of a preset time length into a training data set, a verification data set and a test data set according to a preset proportion;
a parameter selection unit for determining hyper-parameters and training parameters of the information self-attention model, the hyper-parameters including: model convolutional layers, probability sparse self-attention layers and channel parameters and the number of layers of the self-attention layers, wherein the training parameters comprise: learning rate, input batch size, iteration times and model storage conditions;
a data generation unit for generating a training data set and a validation data set;
and the model selection unit is used for training the information self-attention model according to the training data set, outputting a precision report and determining an optimal model.
9. The information-based self-attention model multi-meteorological element prediction system of claim 8, wherein the data training module further comprises:
the timing characteristic extraction unit is used for generating Q, K and V through an embedding layer based on the meteorological acquisition data, entering the probability sparse self-attention layer and extracting timing characteristics;
a temporal feature input unit for concatenating all the temporal features of the probabilistic sparse self-attention layer as input to the self-attention layer in the encoder;
a meteorological element prediction unit, configured to input, to the decoder, a sequence union including the meteorological acquisition data with a length l before a time t + i +1 and a length o of all 0, and a corresponding time variable, and obtain a final prediction output with a length l + o via the probability sparse self-attention layer and the self-attention layer in the decoder, where an output with a length o after the time is a corresponding meteorological element prediction result;
and the training control unit is used for calculating loss with the real label, reversely transmitting the updated parameters according to the gradient, stopping training until the established iteration times are met, the loss does not decrease any more and the performance on the verification data set does not increase any more, and storing the model.
10. The multi-meteorological-element prediction system based on an information self-attention model according to claim 8,
the accuracy evaluation indexes of the accuracy report comprise: determining the coefficient R 2 And a square root error RMSE.
11. An electronic device, comprising: at least one processor; and a memory coupled to the at least one processor; wherein the memory stores instructions executable by the one processor to cause the at least one processor to perform the method of multi-meteorological element prediction based on an informational self-attention model according to any one of claims 1-5.
12. A computer-readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the method for multi-meteorological element prediction based on an informational self-attention model according to any one of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211350817.2A CN115826092A (en) | 2022-10-31 | 2022-10-31 | Multi-meteorological-element prediction method and system based on information self-attention model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211350817.2A CN115826092A (en) | 2022-10-31 | 2022-10-31 | Multi-meteorological-element prediction method and system based on information self-attention model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115826092A true CN115826092A (en) | 2023-03-21 |
Family
ID=85525971
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211350817.2A Pending CN115826092A (en) | 2022-10-31 | 2022-10-31 | Multi-meteorological-element prediction method and system based on information self-attention model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115826092A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116384593A (en) * | 2023-06-01 | 2023-07-04 | 深圳市国电科技通信有限公司 | Distributed photovoltaic output prediction method and device, electronic equipment and medium |
-
2022
- 2022-10-31 CN CN202211350817.2A patent/CN115826092A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116384593A (en) * | 2023-06-01 | 2023-07-04 | 深圳市国电科技通信有限公司 | Distributed photovoltaic output prediction method and device, electronic equipment and medium |
CN116384593B (en) * | 2023-06-01 | 2023-08-18 | 深圳市国电科技通信有限公司 | Distributed photovoltaic output prediction method and device, electronic equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10290066B2 (en) | Method and device for modeling a long-time-scale photovoltaic output time sequence | |
CN107563565B (en) | A kind of short-term photovoltaic decomposition prediction technique considering Meteorology Factor Change | |
CN108711847B (en) | A kind of short-term wind power forecast method based on coding and decoding shot and long term memory network | |
CN110909919A (en) | Photovoltaic power prediction method of depth neural network model with attention mechanism fused | |
CN109142171A (en) | The city PM10 concentration prediction method of fused neural network based on feature expansion | |
KR102296309B1 (en) | Apparatus and method for predicting solar power generation | |
CN112100911B (en) | Solar radiation prediction method based on depth BILSTM | |
CN111242355A (en) | Photovoltaic probability prediction method and system based on Bayesian neural network | |
Li et al. | Deep spatio-temporal wind power forecasting | |
CN115826092A (en) | Multi-meteorological-element prediction method and system based on information self-attention model | |
CN113822418A (en) | Wind power plant power prediction method, system, device and storage medium | |
CN112862274A (en) | WRF-LES and Deepar combined wind power prediction method | |
CN114330100A (en) | Short-term photovoltaic power probability interval prediction method | |
CN111915079A (en) | Hybrid KNN wind power prediction method and system | |
KR20230076695A (en) | Solar power generation forcasting device and method reflecting solar periodic time variable | |
CN103473461B (en) | Based on the wind power prediction error estimation that data characteristics is extracted | |
CN113159437B (en) | Method for predicting short-term photovoltaic power generation output power | |
KR100928238B1 (en) | Seasonal prediction method of typhoon frequency over the east china sea | |
Castillo et al. | Wind Speed Forecasting using ARMA and Boosted Regression Tree Methods: A Case Study | |
KR20230066927A (en) | Method and apparatus for predicting power generation | |
CN112906757A (en) | Model training method, optical power prediction method, device, equipment and storage medium | |
CN114139783A (en) | Wind power short-term power prediction method and device based on nonlinear weighted combination | |
CN112580899A (en) | Medium-and-long-term wind power generation prediction method and system fused with machine learning model | |
KR102388292B1 (en) | Radiation parameterization system for correcting uncertainty of radiation parameterization emulator and operating method thereof | |
CN116739172B (en) | Method and device for ultra-short-term prediction of offshore wind power based on climbing identification |
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 |