CN117272002B - Solar radiation amount estimation method and device, electronic equipment and storage medium - Google Patents
Solar radiation amount estimation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN117272002B CN117272002B CN202311568129.8A CN202311568129A CN117272002B CN 117272002 B CN117272002 B CN 117272002B CN 202311568129 A CN202311568129 A CN 202311568129A CN 117272002 B CN117272002 B CN 117272002B
- Authority
- CN
- China
- Prior art keywords
- solar radiation
- data
- vectors
- radiation amount
- vector
- 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
Links
- 230000005855 radiation Effects 0.000 title claims abstract description 212
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000012545 processing Methods 0.000 claims abstract description 15
- 239000013598 vector Substances 0.000 claims description 258
- 230000004927 fusion Effects 0.000 claims description 39
- 230000009977 dual effect Effects 0.000 claims description 14
- 230000009466 transformation Effects 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 description 16
- 238000004364 calculation method Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 13
- 238000012549 training Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 230000007246 mechanism Effects 0.000 description 8
- 238000009826 distribution Methods 0.000 description 6
- 238000011176 pooling Methods 0.000 description 5
- 238000010248 power generation Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 239000013307 optical fiber Substances 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 101150064138 MAP1 gene Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000004821 distillation Methods 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The disclosure provides a solar radiation amount estimation method, a device, electronic equipment and a storage medium, and relates to the field of data prediction. The method comprises the following steps: and acquiring meteorological time sequence data corresponding to the target region, constructing multidimensional time sequence data based on the meteorological time sequence data, and inputting the multidimensional time sequence data into a pre-trained solar radiation amount estimation model to obtain solar radiation amount estimation data corresponding to the target region, wherein the solar radiation amount estimation model comprises a model of a double-encoder structure based on probability sparse self-attention. According to the solar radiation quantity estimation method and device, the two encoders are used for respectively processing the theoretical meteorological factor sequence information and the real meteorological factor time sequence information of the region to be measured, so that the problem that the solar radiation quantity estimation model is slower in processing speed due to too high time complexity when processing long sequence data is avoided, and the solar radiation quantity estimation efficiency and accuracy are improved.
Description
Technical Field
The present disclosure relates to the field of data prediction, and in particular, to a method and apparatus for estimating solar radiation amount, an electronic device, and a storage medium.
Background
This section is intended to provide a background or context to the embodiments of the disclosure recited in the claims. The description herein is not admitted to be prior art by inclusion in this section.
For the economical and feasible clean energy power generation technology of photo-thermal power generation, the discontinuity, volatility and randomness of sunlight lead to the power generation process to be very dependent on sunlight conditions, and the generation of energy can be reduced or even stopped in overcast and rainy days or at night, so that an additional thermal energy storage system is needed for time and time, the complexity and cost of the system are increased, and therefore, the solar radiation amount is one of the most critical parameters for estimating the amount of energy generated by photo-thermal power generation.
At present, in the estimation method of solar radiation quantity, more students select a converter model (transducer) to perform time series estimation, but due to the existence of a self-attention secondary time complexity problem, the estimation efficiency of the solar radiation quantity is reduced, meanwhile, the influence of priori knowledge of a clear sky solar irradiance model is not fully considered in the estimation process in the existing research, and the accuracy of solar radiation quantity estimation is reduced.
Disclosure of Invention
In order to overcome the problems in the related art, the present disclosure provides a method, an apparatus, an electronic device, and a storage medium for estimating solar radiation amount, so as to solve the problems of low efficiency and poor accuracy of the estimation result of the solar radiation amount in the related scheme at least to a certain extent.
According to a first aspect of embodiments of the present disclosure, there is provided a method of estimating solar radiation amount, the method comprising:
acquiring meteorological time sequence data corresponding to a target area;
constructing multidimensional time series data based on the weather time series data;
inputting the multidimensional time series data into a pre-trained solar radiation amount estimation model to obtain solar radiation amount estimation data corresponding to a target area;
wherein the solar radiation level estimation model comprises a model of a probabilistic sparse self-attention based dual encoder structure.
According to a second aspect of embodiments of the present disclosure, there is provided a solar radiation amount estimation apparatus comprising:
the data acquisition module is used for acquiring weather time sequence data corresponding to the target area;
the data construction module is used for constructing multi-dimensional time sequence data based on the weather time sequence data;
the solar radiation amount estimation module is used for inputting the multidimensional time series data into a pre-trained solar radiation amount estimation model to obtain solar radiation amount estimation data corresponding to a target area;
wherein the solar radiation level estimation model comprises a model of a probabilistic sparse self-attention based dual encoder structure.
According to a third aspect of embodiments of the present disclosure, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements the method of estimating solar radiation in the first aspect of embodiments of the present disclosure.
According to a fourth aspect of embodiments of the present disclosure, there is provided an electronic device comprising:
a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program when executed by the processor implementing the method of estimating solar radiation in the first aspect of the embodiments of the present disclosure.
The technical scheme provided by the embodiment of the disclosure can comprise the following beneficial effects:
according to the solar radiation amount estimation method in the disclosed example embodiment, the solar radiation amount estimation data corresponding to the target region can be obtained by acquiring the weather time series data corresponding to the target region, constructing the multi-dimensional time series data based on the weather time series data, and inputting the multi-dimensional time series data into the pre-trained solar radiation amount estimation model. On the one hand, by acquiring the weather time sequence data corresponding to the target area, weather influence factors of the weather time sequence data can be better considered in the estimation process of the solar radiation quantity, so that the accuracy of a solar radiation quantity estimation result is improved, meanwhile, the weather time sequence data corresponding to the target area is introduced, the solar radiation quantity of the target area can be individually estimated according to the regional difference, and the solar radiation quantity estimation capability of the model in the target area can be improved; on the other hand, by adopting the double-encoder structure model based on the probability sparse self-attention mechanism, different semantic recognition strategies can be used for capturing context semantic information, the accuracy of the model on solar radiation quantity estimation is improved, meanwhile, the data coding rate can be accelerated by the double-encoder structure model, and the time complexity of processing long-sequence time sequence data is reduced, so that the estimation efficiency of the model is improved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure.
Fig. 1 is a system example diagram of a solar radiation amount estimation method that may be applied to embodiments of the present disclosure.
Fig. 2 is a flow chart of a method of estimating solar radiation amount according to an exemplary embodiment of the present disclosure.
FIG. 3 is a flow chart illustrating the input of meteorological time series data and theoretical solar irradiance data into a probability sparse self-attention based dual encoder, resulting in a context fusion feature vector, according to an exemplary embodiment of the present disclosure.
Fig. 4 is a model frame diagram of a solar radiation amount estimation method according to an exemplary embodiment of the present disclosure.
Fig. 5 is a schematic diagram of the operation of the probability sparse self-attention based encoder shown in accordance with an exemplary embodiment of the present disclosure.
FIG. 6 is a flow chart illustrating the input of meteorological time series data and theoretical solar irradiance data into a probability sparse self-attention based dual encoder, resulting in a context fusion feature vector, according to another example embodiment of the present disclosure.
Fig. 7 is an exemplary diagram of probability sparse self-attention computation principle, according to an exemplary embodiment of the present disclosure.
Fig. 8 is an overall flowchart of a solar radiation amount estimation method according to an exemplary embodiment of the present disclosure.
Fig. 9 is a block diagram of a solar radiation amount estimation device according to an exemplary embodiment of the present disclosure.
Fig. 10 is a schematic diagram of a structure of an electronic device according to an exemplary embodiment of the present disclosure for implementing an embodiment of the present disclosure.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the present disclosure. One skilled in the relevant art will recognize, however, that the aspects of the disclosure may be practiced without one or more of the specific details, or with other methods, components, devices, steps, etc. In other instances, well-known technical solutions have not been shown or described in detail to avoid obscuring aspects of the present disclosure.
Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor devices and/or microcontroller devices.
Fig. 1 shows an exemplary diagram of a system to which a solar radiation amount estimation method of an embodiment of the present disclosure may be applied.
As shown in fig. 1, the scenario architecture may include a server 100 and a terminal device 200. The terminal device 200 may be various electronic devices with interactive functions, which may have a display screen thereon, which may be used to present the solar radiation estimate to the user. In an actual implementation of the present disclosure, the terminal device 200 may be a mobile terminal, a desktop computer, a handheld device, or the like, which is not particularly limited in this example embodiment.
It should be understood that the number of servers 100 and terminal devices 200 in fig. 1 is merely illustrative. There may be any number of servers 100 and terminal devices 200 as needed for implementation, for example, the server 100 may be a server cluster formed by a plurality of servers, etc.
The solar radiation amount estimation method provided by the embodiments of the present disclosure may be performed by the terminal apparatus 200, and accordingly, the solar radiation amount estimation device may be provided in the terminal apparatus 200. However, it will be readily understood by those skilled in the art that the solar radiation amount estimation method provided in the embodiment of the present disclosure may be performed by the server 100, and accordingly, the solar radiation amount estimation apparatus may be provided in the server 100, which is not particularly limited in this exemplary embodiment.
In the related art, there are the following problems:
due to the problem of self-attention secondary time complexity, the traditional transducer model is difficult to be used for long-sequence estimation tasks, so that the method is difficult to be applied to solar radiation quantity estimation tasks with longer time such as week estimation and month estimation, the calculated quantity is large, the system performance requirement is high, and the completion efficiency of the solar radiation quantity estimation tasks is low; meanwhile, the matching property of the currently obtained solar radiation amount estimation result and each region is low, so that the accuracy of the solar radiation amount estimation result is low.
Based on one or more problems in the related schemes, the embodiment of the disclosure first provides a solar radiation amount estimation method, which can improve the estimation efficiency of the solar radiation amount and the accuracy of the estimation result of the solar radiation amount. The method for estimating solar radiation amount in the embodiment of the present disclosure will be described below by taking the server as an example, and referring to fig. 2, the method for estimating solar radiation amount in the embodiment of the present disclosure may include the steps of:
in step S201, weather time series data corresponding to a target region is acquired;
in step S202, constructing multi-dimensional time series data based on weather time series data;
in step S203, the multidimensional time series data is input into a pre-trained solar radiation amount estimation model, so as to obtain solar radiation amount estimation data corresponding to a target region; wherein the solar radiation level estimation model comprises a model of a probabilistic sparse self-attention based dual encoder structure.
According to the solar radiation amount estimation method in the disclosed example embodiment, on one hand, by acquiring the weather time series data corresponding to the target area, weather influence factors of the weather time series data can be better considered in the estimation process of the solar radiation amount, so that the accuracy of a solar radiation amount estimation result is improved, meanwhile, the weather time series data corresponding to the target area is introduced, the solar radiation amount of the target area can be individually estimated according to the regional difference, and the solar radiation amount estimation capability of the model in the target area can be improved; on the other hand, by adopting the double-encoder structure model based on the probability sparse self-attention mechanism, different semantic recognition strategies can be used for capturing context semantic information, the accuracy of the model on solar radiation quantity estimation is improved, meanwhile, the data coding rate can be accelerated by the double-encoder structure model, and the time complexity of processing long-sequence time sequence data is reduced, so that the estimation efficiency of the model is improved.
Next, step S201 to step S203 will be described in detail.
In step S201, weather time series data corresponding to a target region is acquired.
In an exemplary embodiment of the present disclosure, the weather time series data refers to continuous observation data obtained by observing and recording weather elements over a period of time. In the present exemplary embodiment, the acquisition interval data of the weather time series data may be several hours, several days, several months, or the like, and the present exemplary embodiment is not particularly limited thereto. Meanwhile, the weather time series data in the present exemplary embodiment may include at least one of data of real solar direct radiation intensity, real solar total radiation intensity, real solar scattering intensity, average atmospheric pressure, relative humidity, wind direction and speed, temperature, precipitation amount, and the like, which is not particularly limited in the present exemplary embodiment.
For example, the corresponding data may be obtained in real time from the monitoring device of each weather time series data, or the corresponding data stored in advance may be obtained from a database, which is not particularly limited in this example embodiment.
By acquiring the meteorological time sequence corresponding to the target area, the estimation process of the solar radiation quantity can be more attached to the actual situation of the target area, so that the accurate estimation of the solar radiation quantity is realized, meanwhile, the corresponding long-sequence meteorological time sequence data is analyzed, and the generated solar radiation quantity estimation data can be more accurate.
In step S202, multidimensional time series data is constructed based on weather time series data.
In an exemplary embodiment of the present disclosure, the multi-dimensional time series data refers to a data set composed of a plurality of variables that change in time. In this example embodiment, the multidimensional time series data may be a combination of weather time series data and theoretical solar radiation amount data, where the theoretical solar radiation amount data refers to a theoretical value of solar radiation amount data of a target area that is calculated in combination with information such as latitude and longitude or time zone of the target area.
For example, if the weather time series data is expressed asTheoretical solar irradiance data is expressed asThe meteorological time series data and the theoretical solar radiation amount data can be combined in a splicing mode to construct multidimensional time series data, namely, the theoretical solar radiation amount data can be spliced after the meteorological time series data to construct multidimensional time series data such as ++>Of course, the theoretical solar radiation data can also be spliced before the meteorological time series data to constructThe present exemplary embodiment is not particularly limited thereto.
Alternatively, the multi-dimensional time series data may include weather time series data And a theoretical solar radiation data +.>Meteorological time series data may also be included +.>And a plurality of theoretical solar radiation data +.>The present exemplary embodiment is not particularly limited thereto.
In step S203, the multidimensional time series data is input into a pre-trained solar radiation amount estimation model, so as to obtain solar radiation amount estimation data corresponding to a target region; wherein the solar radiation level estimation model comprises a model of a probabilistic sparse self-attention based dual encoder structure.
In an example embodiment of the present disclosure, the solar radiation amount estimation model refers to a model used to estimate the solar radiation amount from multi-dimensional time series data. In the present exemplary embodiment, the solar radiation amount estimation model may be a model of a dual-encoder structure based on multi-head probability sparse self-attention, or may be an integrated regression model that processes multidimensional time series data, or may be a gated cyclic neural network model, or may be a model that processes multidimensional time series data and performs solar radiation amount estimation, which is not particularly limited in the present exemplary embodiment.
The multi-dimensional time series data are input into the pre-trained solar radiation amount estimation model, so that solar radiation amount estimation data corresponding to a target area are obtained, the situation that a complex nonlinear relation between solar radiation amount and various factors cannot be captured well by using a traditional statistical method is avoided, and the estimation accuracy of the solar radiation amount is improved.
The technical solutions involved in step S201 to step S203 are explained in detail below.
In an example embodiment of the present disclosure, building multi-dimensional time series data based on weather time series data may be accomplished by:
obtaining geographic position information of a target region, inputting the geographic position information into a pre-trained clear sky model, and determining theoretical solar radiation amount data of the target regionFurthermore, the meteorological time series data and the theoretical solar radiation data can be spliced to obtain multi-dimensional time series data, wherein the meteorological time series data are +.>Comprises real solar direct radiation intensity, real solar total radiation intensity, real solar scattering intensity, average atmospheric pressure, relative humidity, wind direction and wind speed, theoretical solar radiation amount data ∈>Including theoretical direct solar radiation intensity, theoretical total solar radiation intensity, and theoretical solar scattering intensity.
The geographic position information of the target region refers to information describing specific position and space coordinates of the target region on the earth. In the present exemplary embodiment, the geographical position information may be any one of longitude, latitude, time zone, altitude, and the like, and the present exemplary embodiment is not particularly limited thereto.
Alternatively, the geographic location information of the target area may be obtained through a global positioning system, or the geographic location information of the target area may be obtained through a geographic location system, or the geographic location information of the target area may be checked by using an online map service, which is not particularly limited in this example embodiment.
Illustratively, the weather time series data includes real solar direct radiation intensity, real solar total radiation intensity, real solar scattering intensity, average atmospheric pressure, relative humidity, wind direction and wind speed, and the acquired data is subjected to data division when the window length is 8 and the step length is 1, so that each sample contains 7 weather time series data with the length of 8. In mathematical terms, data can be considered as a three-dimensional matrix with a total amount of data in dimension 1, 7 in dimension 2, and 8 in dimension 3. Assume the firstThe individual real sample data are denoted +.>Wherein each element represents a time series of seven weather factors of length 8 in the same time period, e.g +.>A time series of solar radiation DNI features is represented.
A clear sky model is a mathematical model for modeling and predicting solar radiation transmission in the earth's atmosphere. For example, the geographic position information is input into a pre-trained clear sky model, and three theoretical solar radiation amounts of the obtained target region in the time period of the sample can be expressed as Will include the real solar direct radiation intensity, the real solar total radiation intensity, the real solar scattering intensity, the average atmospheric pressure, the relativeSeven meteorological time series data of humidity, wind direction and wind speed are expressed as +.>Furthermore, the meteorological time series data and the theoretical solar radiation data can be spliced to obtain multidimensional time series data expressed as。
Specifically, after the original data set is built, the subsequent output can be estimated according to the weather time sequence data of the previous months as the training data set of the model, meanwhile, the training data set is divided into a training set and a test set, for example, 70% of the training data set can be divided into the training set, the rest 30% is divided into the test set, the training set is used for training the initial solar radiation amount estimation model, the test set is used for checking and evaluating the model after the training of the initial solar radiation amount estimation model is finished, and when the model evaluation of the initial solar radiation amount estimation model is unqualified, the model parameters are adjusted to carry out model training again. When the model evaluation of the initial solar radiation amount estimation model is qualified, the model evaluation is used as a trained model to carry out subsequent solar radiation amount estimation.
The meteorological time series data and the theoretical solar radiation amount data are spliced to obtain the multidimensional time series data, so that the correlation between the meteorological time series data and the theoretical solar radiation amount data can be considered in the estimation process of the solar radiation amount, and the meteorological phenomenon can be more comprehensively analyzed and predicted.
In an example embodiment of the present disclosure, the solar radiation amount estimation model may include a probability sparse self-attention based dual encoder and a probability sparse self-attention based decoder, and the following steps may be implemented:
the meteorological time sequence data and the theoretical solar radiation amount data can be input into a double encoder based on probability sparse self-attention to obtain a context fusion feature vector, and then the context fusion feature vector can be input into a decoder to be decoded to obtain solar radiation amount estimation data corresponding to a target region.
The context fusion feature vector refers to a method for representing and fusing context information in a text by means of feature vectors in natural language processing. In the present exemplary embodiment, the context feature vectors may be fused by means of weighted summation to obtain a context fusion feature vector, or may be obtained by means of stitching, which is not particularly limited in the present exemplary embodiment.
Alternatively, the context fusion feature vector is input to the decoder, which may predict the output of the next step in a single-step decoding manner, i.e., by generating a probability distribution through the output layer of the decoder for predicting the output of the next step.
By way of example only, and in an illustrative,representing +.>For a time step, use->Representing the generated context fusion feature vector using +.>Representing the input sequence->Representing the length of the input sequence, the decoder output can be expressed as +.>，/>Representing the predicted sequence length, t representing the time step of the current process,/->Representing the input sequence +.>Hidden state processed by encoder, < +.>Indicate->The output of the individual predicted time steps,representing the target sequence of the previously decoded prediction output. According to the time sequence of the input->And the target sequence of the previously decoded prediction output +.>Calculating the predicted next time (th +.>Time step) target value ∈ ->And then the objective function in each prediction process is multiplied, so that the joint conditional probability shown in the formula (1) can be obtained. In theory, when the objective function value is the largest, the target value of the predicted output is the optimal, and the objective function of the whole model can be written as the following joint conditional probability formula (1):
Formula (1)
Wherein,representing model parameters, which may include mean, variance, coefficients, etc., for the present example embodiment +.>And are not limited.
The solar radiation amount is estimated through the context fusion feature vector, compared with the single use of meteorological time series data or theoretical solar radiation amount data, the method comprises comprehensive information and relativity of multi-source data, and the method can better reflect real conditions, so that the accuracy of solar radiation amount estimation can be further improved, meanwhile, probability sparsity is introduced into a traditional self-attention mechanism to reduce interaction between positions, namely, for each position, only a part of relatively important positions are selected for calculation, and therefore calculation complexity can be remarkably reduced, and model performance is maintained to a certain extent.
In an example embodiment of the present disclosure, a dual encoder based on probability sparse self-attention includes a first encoder and a second encoder, and inputting meteorological time series data and theoretical solar radiation amount data into the dual encoder based on probability sparse self-attention can be achieved by the following steps, with reference to fig. 3, to obtain a context fusion feature vector:
In step S301, weather time series data is input into a first encoder to obtain a weather feature vector;
in step S302, inputting the theoretical solar radiation amount data into the second encoder to obtain a theoretical solar radiation amount feature vector;
in step S303, the meteorological feature vector and the theoretical solar radiation feature vector are adaptively weighted and fused, so as to obtain a context fusion feature vector.
The structures of the first encoder and the second encoder may refer to fig. 4 and fig. 5, as shown in fig. 4, where the first encoder is used for encoding real values, i.e. weather time series data, and the second encoder is used for encoding theoretical values, i.e. theoretical solar radiation, and meanwhile, multi-head probability sparse attention modules are applied to both the first encoder and the second encoder. Fig. 5 shows a process of encoding input multidimensional time series data by a first encoder and a second encoder to generate corresponding feature vectors, and the following description is given in the figure:
the input of the model consists of time series data and time stamp information.
Timestamp information: namely, the time stamp information corresponding to each element in the input time sequence data is used for representing the time sequence in the sequence data;
Time stamp information is embedded: converting the time stamp information into a fixed length vector representation, typically using an embedding layer to map the time stamps into a continuous vector space;
one-dimensional convolution layer: extracting features of an input sequence through a one-dimensional convolution layer;
multi-headed probability sparse attention modules 1, 2, 3: the one-dimensional convolution layer and the two parts processed by the one-dimensional convolution layer are spliced and then input into a probability sparse attention mechanism, the representation of important elements is gradually enhanced by calculating the correlation between the elements in an input sequence, and the calculation complexity is reduced by probability sparsification of the correlation;
convolution pooling layer: the cascade decreasing is obtained by carrying out self-attention distillation on each layer of input sequence data through twice convolution pooling, the length of the input sequence is gradually shortened, namely, a main stack of the whole input sequence is received, then a second stack obtains half of the input sequence after passing through a first convolution pooling layer, and a third stack obtains one fourth of the original input sequence after passing through a second convolution pooling layer.
Feature map: the feature maps of all stacks are connected as output of the encoder.
The meteorological time sequence data and the theoretical solar radiation data are input into different encoders to obtain corresponding feature vectors, and then the corresponding feature vectors are subjected to self-adaptive weighted fusion to obtain context fusion feature vectors. On one hand, different encoders are adopted to encode weather time sequence data and theoretical solar radiation data respectively, and different semantic recognition strategies can be adopted for the weather time sequence data and the theoretical solar radiation data so as to better capture context semantic information, thereby improving the accuracy of input data feature extraction and the accuracy of model estimation. On the other hand, the multi-layer multi-head probability sparse attention module and the convolution pooling layer are applied in the encoder, and when long sequence input data is faced, the estimation efficiency of the solar radiation amount can be improved on the premise of ensuring the accuracy of the solar radiation amount estimation value.
In an example embodiment of the present disclosure, inputting meteorological time series data and theoretical solar irradiance data into a probability sparse self-attention-based dual encoder may be accomplished by:
in step S601, linear transformation is performed on the weather time sequence data and the theoretical solar radiation amount data by using a probability sparse self-attention-based double encoder, so as to obtain query vectors, key vectors and value vectors corresponding to the weather time sequence data and the theoretical solar radiation amount data;
in step S602, a key query vector is determined from the query vectors, and an attention weight is calculated from the key query vector;
in step S603, an output context fusion feature vector is calculated by the attention weight, the query vector, the key vector, and the value vector.
By way of example, each query vector of the meteorological time series data may be randomly sampled in all key vectors of the theoretical solar radiation amount data according to a preset sampling number, and then a sparsity score of each query vector may be calculated according to the sampled key vectors, and then a plurality of query vectors with high sparsity scores may be selected as key query vectors. And meanwhile, carrying out average value calculation on query vectors of all unselected meteorological time series data to obtain a feature vector representing the average value, and further combining the feature vector representing the average value with a context fusion feature vector to obtain a target context fusion feature vector, so as to estimate the solar radiation amount according to the target context fusion feature vector.
Alternatively, the sparsity scores of all the query vectors may be selected to be ranked, and the query vector ranked at the top is selected as the key query vector according to the preset number, or of course, the sparsity score threshold may be preset, and the query vector with the sparsity score of each query vector being greater than or equal to the sparsity score threshold may be used as the key query vector, which is not particularly limited in this exemplary embodiment.
Alternatively, the sparsity score calculation of each query vector may be performed by defining the attention of the query vector to all key vectors as a conditional probability distribution, and simultaneously setting up a reference probability distribution as a sparsity score reference, by calculating the degree of difference between the two probability distributions, or by performing the sparsity score calculation of each query vector by the L1 norm, or of course by performing the sparsity score calculation of each query vector by the L0 norm.
For example, the attention weight can be obtained by calculating the key query vector and all key vectors, for example, the dot product result of the key query vector of the meteorological time series data on the key vector of the theoretical solar radiation amount data can be determined, then all dot product results are accumulated to obtain a total dot product result, and then each dot product result can be divided by the total dot product result to obtain the attention weight between the query vector of each meteorological time series data and the key vector of the theoretical solar radiation amount data.
In the present exemplary embodiment, for a process of calculating a sparsity score of each query vector from a query vector, a key vector, and a value vector of input data, there are the following examples:
equation (2) as follows is a conventional self-attention calculation equation:
formula (2)
Wherein,output result representing the attentiveness mechanism, +.>Representing a query vector->Watch (watch)
The vector of the key is shown as such,representing a value vector +_>Representing an activation function normalizing the attention weights such that the sum of the attention weights is 1,/or->Representing transpose operations on key vectors, +.>Representing the input dimension.
The traditional self-attention calculation formula (2) is improved by introducing a probability sparse self-attention mechanism, and an improved formula (3) is obtained:
formula (3)
Wherein,output result representing the attentiveness mechanism, +.>Indicate->Individual query vectors->Representing key vectors +_>Representing a value vector +_>Representing the number of query vectors, +.>Representing the number of random key vectors,/-, and>representing the number of all key vectors, +.>Indicate->Random key vector, ">Representing the +.>Personal key vector->Representation and->A value vector associated with a random key vector, +.>For a function representing the correlation between a key query vector and a random key vector >Representing the sum of the calculated correlations of the key query vectors over all key vectors,/for>，，/>Representing the number of query vectors, +.>Representing the number of random key vectors,/-, and>representing a transpose operation->Indicate->Individual query vectors->Indicate->Random key vector, ">Indicate->The random key vectors are transposed and,representing the input dimension, the formula representing the dot product of the query vector and the key vector divided by +.>Scaling is performed to achieve indexing.
The implementation step of the formula (2) may refer to fig. 7, where the input weather time sequence data and the theoretical solar radiation data are subjected to linear transformation to obtain a corresponding query vector, a key vector and a value vector, transpose the key vector matrix, calculate the attention weight result W of the key vector and the query vector, and finally perform weighted dot product summation with the value vector through a series of transformations to obtain a final dot product result Z.
In order to measure the sparsity of the input query vector, the invention will be as followsThe attention of a query vector to a random key vector is defined as a conditional probability +.>When the probability distribution is close to uniform distribution, attention is degraded into the summation of value vectors, and in order to further measure the sparsity of query vectors, the embodiment of the disclosure proposes the following formula (4):
Formula (4)
Wherein,expressed as probability distribution +.>For reference, relative to uniform distribution->Degree of difference of->Indicates the sequence length,/->Representing natural constants, approximately 2.71828, +.>Representing input dimension +.>Representing a transpose operation->Indicate->Individual query vectors->Representing the +.>Personal key vector->Indicate->Random key vector, ">Representing the number of query vectors, +.>Representing the number of random key vectors,/-, and>representing the number of all key vectors, +.>Representing +.>The individual key vectors are transposed,/->Indicate->The random key vectors are transposed.
The sparsity score for each query vector may be calculated by the following relationship (5) and the query vector and random key vector:
formula (5)
Wherein,representing the number of query vectors, +.>Representing the number of all key vectors, +.>The number of random key vectors is represented,representing key vectors +_>Indicate->Sparsity score for individual query vectors, +.>Indicate->Individual query vectors->Representing the +.>Personal key vector->Indicate->Random key vector, ">Representing +.>The individual key vectors are transposed,/- >Indicate->Transpose the random key vectors, ">Indicates the sequence length,/->Representing input dimension +.>Representing natureConstant, approximately 2.71828, < ->Is->Log sum of individual query vectors over all key vectors,/->Is->Arithmetic mean of the individual query vectors over the random key vectors.
The approximate result formula (6) of the sparsity score can be obtained according to the upper and lower bounds of formula (5):
formula (6)
Wherein,indicate->Approximate result of sparsity score for each query vector, +.>Representing key vectors +_>Indicate->Random key vector, ">Representing the number of random key vectors,/-, and>indicates the sequence length,/->Indicate->Individual query vectors->Representing input dimension +.>Indicate->The random key vectors are transposed.
By calculating the sparsity score of each query vector, and further selecting the query vector with higher score as the key query vector, the data volume in the model data processing process can be reduced, the calculation resources and the time cost are saved, the model estimation efficiency is improved, meanwhile, the vector with higher sparsity score often represents important characteristics or key information, the key information in the data can be better captured by selecting the query vectors as the key query vectors, the accuracy and generalization capability of the model are improved, and the key vectors are selected by the preset number or sparsity score threshold value, so that the sparsity of the model can be artificially controlled, and the accuracy controllability of model estimation is improved.
In an example embodiment of the present disclosure, referring to a model frame diagram of a solar radiation amount estimation method shown in fig. 4, as shown in the drawing, weather time series data, that is, a real value and a theoretical value calculated based on a clear sky model are respectively input into two encoders, and are processed by adopting a multi-head probability sparse self-attention mechanism to obtain a corresponding theoretical feature diagram, and then the two corresponding real feature diagrams are fused and input into a decoder, and the fused feature diagram is decoded to generate a solar radiation amount estimation value.
For example, referring to formulas (7) and (8), a layer of fully-connected network can be used to adaptively calculate the weights of the two encoder outputs, and finally the characteristics of the encoder fusion output are obtainedAnd performing estimation of the solar radiation amount:
formula (7)
Formula (8)
Wherein,characteristic map 1 showing the output of the meteorological time series data processed by the first encoder, i.e. the hidden state of the first encoder output +.>And a characteristic diagram 2 which is output after the theoretical solar radiation amount data is processed by the second encoder, namely the hidden state of the output of the second encoder. />And->Representing the weight of the fully connected network that is trainable, Representing a trainable bias term +.>Representing a sigmoid activation function,/->Features representing the encoder fusion output, +.>The weight coefficient calculated by a layer of fully-connected network is represented, the value range is (0, 1) and is used for adding the characteristic graphs of the two encodersWeight.
In an example embodiment of the present disclosure, referring to fig. 8, a flowchart of a solar radiation amount estimation method is shown, including the following steps S801 to S808:
s801, acquiring meteorological time sequence data;
s802, constructing multidimensional time series data: obtaining geographic position information of a target region, inputting the geographic position information into a pre-trained clear sky model to determine theoretical solar radiation amount data of the target region, and splicing meteorological time series data with the theoretical solar radiation amount data to obtain multidimensional time series data;
s803, obtaining a query vector, a key vector and a value vector: performing linear transformation on the weather time sequence data and the theoretical solar radiation amount data through a probability sparse self-attention-based double encoder to obtain query vectors, key vectors and value vectors corresponding to the weather time sequence data and the theoretical solar radiation amount data;
S804, calculating a sparsity score: obtaining a preset number of random key vectors from all key vectors, and calculating sparsity scores of the query vectors according to a formula (5) according to the query vectors of the meteorological time sequence data and the random key vectors of the theoretical solar radiation data;
step S805, judging whether the sparsity score threshold is greater than or equal to: if the sparsity score of the query vector of the current weather time sequence data is greater than or equal to the sparsity score threshold, the query vector is used as a key query vector, the step S806 is waited for judging to finish execution, and if the sparsity score is smaller than the sparsity score threshold, the current operation is ended and the sparsity score comparison of the next query vector is executed;
step S806, calculating attention weight: calculating to obtain the attention weight through all key vectors and all key vectors with the sparsity score larger than or equal to the sparsity score threshold in S805;
s807, obtaining a context fusion feature vector;
and S808, estimating the solar radiation amount.
In an exemplary embodiment of the present disclosure, as shown in fig. 9, there is provided a solar radiation amount estimation device, including a data acquisition module 901, a data construction module 902, and a solar radiation amount estimation module 903, specifically as follows:
The data acquisition module 901 may be configured to acquire weather time series data corresponding to a target area;
the data construction module 902 may be configured to construct multi-dimensional time series data based on weather time series data;
the solar radiation amount estimation module 903 may be configured to input the multidimensional time series data into a pre-trained solar radiation amount estimation model, to obtain solar radiation amount estimation data corresponding to the target area;
wherein the solar radiation level estimation model comprises a model of a probabilistic sparse self-attention based dual encoder structure.
In an example embodiment of the present disclosure, the data construction module 902 is determined as:
the theoretical value determining module is used for acquiring the geographic position information of the target area, inputting the geographic position information into the pre-trained clear sky model and determining theoretical solar radiation amount data of the target area;
the data splicing module is used for splicing the meteorological time sequence data with the theoretical solar radiation amount data to obtain multidimensional time sequence data;
the meteorological time sequence data comprise real solar direct radiation intensity, real solar total radiation intensity, real solar scattering intensity, average atmospheric pressure, relative humidity, wind direction and wind speed, and the theoretical solar radiation amount data comprise theoretical solar direct radiation intensity, theoretical solar total radiation intensity and theoretical solar scattering intensity.
In an example embodiment of the present disclosure, the solar radiation amount estimation model includes a probabilistic sparse self-attention based dual encoder and a probabilistic sparse self-attention based decoder, and the solar radiation amount estimation module 903 is determined to:
the feature vector acquisition module is used for inputting the meteorological time sequence data and the theoretical solar radiation data into a double encoder based on probability sparse self-attention to obtain a context fusion feature vector;
and the feature vector decoding module is used for inputting the context fusion feature vector into a decoder for decoding to obtain solar radiation quantity estimation data corresponding to the target region.
In an example embodiment of the present disclosure, the probability sparse self-attention based dual encoder includes a first encoder and a second encoder, the solar radiation amount estimation module 903 is determined to:
the weather feature vector generation module is used for inputting weather time sequence data into the first encoder to obtain weather feature vectors;
the theoretical solar radiation characteristic vector generation module is used for inputting theoretical solar radiation data into the second encoder to obtain a theoretical solar radiation characteristic vector;
and the feature vector fusion module is used for carrying out self-adaptive weighted fusion on the meteorological feature vector and the theoretical solar radiation quantity feature vector to obtain a context fusion feature vector.
In an example embodiment of the present disclosure, the solar radiation amount estimation module 903 is determined as:
the linear transformation module is used for carrying out linear transformation on the weather time sequence data and the theoretical solar radiation amount data through a double encoder based on probability sparse self-attention to obtain query vectors, key vectors and value vectors corresponding to the weather time sequence data and the theoretical solar radiation amount data;
the attention weight calculation module is used for determining a key query vector from the query vectors and calculating the attention weight through the key query vector;
and the context fusion feature vector calculation module is used for calculating and outputting the context fusion feature vector through the attention weight, the query vector, the key vector and the value vector.
In an example embodiment of the present disclosure, the solar radiation amount estimation module 903 further includes:
the random sampling module is used for randomly sampling from the key vectors to obtain random key vectors;
the sparsity score calculation module is used for calculating the sparsity score of each query vector according to the query vector and the random key vector;
the key query vector determining module is used for taking a query vector with the sparsity score being greater than or equal to a preset sparsity score threshold value as a key query vector;
And the attention weight generating module is used for calculating the attention weight through the key query vector and all key vectors.
In an example embodiment of the present disclosure, the solar radiation amount estimation module 903 is determined as:
the sparsity score for each query vector is calculated by the following equation (5) and the query vector and random key vector:
；
wherein,representing the number of query vectors, +.>Representing the number of all key vectors, +.>The number of random key vectors is represented,representing key vectors +_>Indicate->Sparsity score for individual query vectors, +.>Indicate->Individual query vectors->Representing the +.>Personal key vector->Indicate->Random key vector, ">Representing +.>The individual key vectors are transposed,/->Indicate->Transpose the random key vectors, ">Indicates the sequence length,/->Representing input dimension +.>Representing natural constants, approximately 2.71828, +.>Is->Log sum of individual query vectors over all key vectors,/->Is->Arithmetic mean of the individual query vectors over the random key vectors.
The specific details of each module in the above-mentioned solar radiation amount estimation device have been described in detail in the corresponding solar radiation amount estimation method, and thus are not described here again.
Exemplary embodiments of the present disclosure also provide a computer-readable storage medium having stored thereon a program product capable of implementing the method described above in the present specification. In some possible implementations, aspects of the present disclosure may also be implemented in the form of a program product comprising program code for causing an electronic device to carry out the steps according to the various exemplary embodiments of the disclosure as described in the "exemplary methods" section of this specification, when the program product is run on an electronic device. The program product may employ a portable compact disc read-only memory (CD-ROM) and comprise program code and may be run on an electronic device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
The computer readable signal medium may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, radio frequency technology (RF), or any suitable combination of the foregoing.
Program code for carrying out operations of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C#, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).
The exemplary embodiment of the disclosure also provides an electronic device capable of implementing the method. An electronic device 1000 according to such an exemplary embodiment of the present disclosure is described below with reference to fig. 10. The electronic device 1000 shown in fig. 10 is merely an example and should not be construed as limiting the functionality and scope of use of the disclosed embodiments.
As shown in fig. 10, the electronic device 1000 may be embodied in the form of a general purpose computing device. Components of electronic device 1000 may include, but are not limited to: at least one processing unit 1010, at least one memory unit 1020, a bus 1030 connecting the various system components (including the memory unit 1020 and the processing unit 1010), and a display unit 1040.
The memory unit 1020 stores program code that can be executed by the processing unit 1010, such that the processing unit 1010 performs steps according to various exemplary embodiments of the present disclosure described in the above "exemplary methods" section of the present specification. For example, the processing unit 1010 may perform the method steps in fig. 2.
The memory unit 1020 may include readable media in the form of volatile memory units such as Random Access Memory (RAM) 1021 and/or cache memory unit 1022, and may further include Read Only Memory (ROM) 1023.
Storage unit 1020 may also include a program/utility 1024 having a set (at least one) of program modules 1025, such program modules 1025 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.
Bus 1030 may be representing one or more of several types of bus structures including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.
The electronic device 1000 can also communicate with one or more external devices 1070 (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a user to interact with the electronic device 1000, and/or with any device (e.g., router, modem) that enables the electronic device 1000 to communicate with one or more other computing devices. Such communication may occur through an input/output (I/O) interface 1050. Also, electronic device 1000 can communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet, through network adapter 1060. As shown, the network adapter 1060 communicates with other modules of the electronic device 1000 over the bus 1030. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with the electronic device 1000, including, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.
From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or may be implemented in software in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, including several instructions to cause a computing device (may be a personal computer, a server, a terminal device, or a network device, etc.) to perform the method according to the exemplary embodiments of the present disclosure.
Furthermore, the above-described figures are only illustrative of the inclusion of a method according to exemplary embodiments of the present disclosure, and are not intended to be limiting. It will be readily appreciated that the processes shown in the above figures do not indicate or limit the temporal order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, among a plurality of modules.
It should be noted that although in the above detailed description several modules or units of a device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit in accordance with embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into a plurality of modules or units to be embodied.
This disclosure is intended to cover any adaptations, uses, or adaptations of the disclosure following the general principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope of the disclosure being indicated by the following claims.
It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (8)
1. A method of estimating solar radiation, comprising:
acquiring meteorological time sequence data corresponding to a target area;
obtaining geographic position information of the target region, inputting the geographic position information into a pre-trained clear sky model, determining theoretical solar radiation amount data of the target region, and splicing the meteorological time series data with the theoretical solar radiation amount data to obtain multidimensional time series data;
inputting the multidimensional time series data into a pre-trained solar radiation amount estimation model to obtain solar radiation amount estimation data corresponding to the target region, wherein the solar radiation amount estimation model comprises a probability sparse self-attention-based double encoder and a probability sparse self-attention-based decoder, and the probability sparse self-attention-based double encoder comprises a first encoder and a second encoder;
The step of inputting the multidimensional time series data into a pre-trained solar radiation amount estimation model to obtain solar radiation amount estimation data corresponding to the target region comprises the following steps:
inputting the meteorological time sequence data and the theoretical solar radiation data into the double encoders based on probability sparse self-attention to obtain context fusion feature vectors;
inputting the context fusion feature vector into the decoder for decoding to obtain solar radiation amount estimation data corresponding to the target area;
the inputting the meteorological time sequence data and the theoretical solar radiation data into the double encoder based on probability sparse self-attention to obtain a context fusion feature vector comprises the following steps:
inputting the weather time sequence data into the first encoder to obtain weather feature vectors;
inputting the theoretical solar radiation amount data into the second encoder to obtain a theoretical solar radiation amount characteristic vector;
and carrying out self-adaptive weighted fusion on the meteorological feature vector and the theoretical solar radiation amount feature vector to obtain a context fusion feature vector.
2. The method of claim 1, wherein the weather time series data includes true solar direct radiation intensity, true solar total radiation intensity, true solar scattering intensity, average atmospheric pressure, relative humidity, wind direction, and wind speed, and the theoretical solar radiation data includes theoretical solar direct radiation intensity, theoretical solar total radiation intensity, theoretical solar scattering intensity.
3. The method of claim 1, wherein said inputting the meteorological time series data and the theoretical solar radiation data into the probabilistic sparse self-attention based dual encoders results in a context fusion feature vector, comprising:
performing linear transformation on the meteorological time sequence data and the theoretical solar radiation amount data through the probability sparse self-attention-based double encoder to obtain query vectors, key vectors and value vectors corresponding to the meteorological time sequence data and the theoretical solar radiation amount data;
determining a key query vector from the query vectors, and calculating attention weights through the key query vector;
And calculating an output context fusion feature vector through the attention weight, the query vector, the key vector and the value vector.
4. A method of estimating solar radiation amount according to claim 3, wherein said determining a key query vector from said query vectors and calculating an attention weight from said key query vector comprises:
randomly sampling from the key vectors to obtain random key vectors;
calculating sparsity scores of the query vectors according to the query vectors and the random key vectors;
taking the query vector with the sparsity score being greater than or equal to a preset sparsity score threshold as a key query vector;
and calculating the attention weight through the key query vector and all the key vectors.
5. The method of claim 4, wherein said calculating a sparsity score for each of said query vectors from said query vector and said random key vector comprises:
calculating a sparsity score for each of the query vectors by the following relationship and the query vector and the random key vector:
；
wherein,representing the number of query vectors, +. >Representing the number of all key vectors, +.>Representing the number of random key vectors,/-, and>representing key vectors +_>Indicate->Sparsity score for individual query vectors, +.>Indicate->Individual query vectors->Representing the +.>Personal key vector->Indicate->Random key vector, ">Representing +.>The individual key vectors are transposed,/->Indicate->Transpose the random key vectors, ">Indicates the sequence length,/->Representing input dimension +.>Representing natural constant->Is->Log sum of individual query vectors over all key vectors,/->Is the firstArithmetic mean of the individual query vectors over the random key vectors.
6. A solar radiation amount estimation device, characterized by comprising:
the data acquisition module is used for acquiring weather time sequence data corresponding to the target area;
the data construction module is used for acquiring the geographic position information of the target area, inputting the geographic position information into a pre-trained clear sky model, determining theoretical solar radiation amount data of the target area, and splicing the meteorological time series data with the theoretical solar radiation amount data to obtain multidimensional time series data;
The solar radiation quantity estimation module is used for inputting the multidimensional time series data into a pre-trained solar radiation quantity estimation model to obtain solar radiation quantity estimation data corresponding to the target area, the solar radiation quantity estimation model comprises a probability sparse self-attention-based double encoder and a probability sparse self-attention-based decoder, and the probability sparse self-attention-based double encoder comprises a first encoder and a second encoder;
wherein the solar radiation amount estimation module comprises:
the feature vector generation module is used for inputting the meteorological time sequence data and the theoretical solar radiation quantity data into the double encoder based on probability sparse self-attention to obtain a context fusion feature vector;
the feature vector decoding module is used for inputting the context fusion feature vector into the decoder for decoding to obtain solar radiation amount estimation data corresponding to the target area;
wherein, the feature vector generation module includes:
the first encoder data processing module is used for inputting the weather time sequence data into the first encoder to obtain weather feature vectors;
The second encoder data processing module is used for inputting the theoretical solar radiation quantity data into the second encoder to obtain a theoretical solar radiation quantity characteristic vector;
and the feature vector fusion module is used for carrying out self-adaptive weighted fusion on the meteorological feature vector and the theoretical solar radiation quantity feature vector to obtain a context fusion feature vector.
7. A storage medium having stored thereon a computer program, which when executed by a processor implements the solar radiation amount estimation method of any one of claims 1 to 5.
8. An electronic device, comprising:
memory, a processor and a computer program stored on the memory and executable on the processor, which when executed by the processor implements the solar radiation amount estimation method of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311568129.8A CN117272002B (en) | 2023-11-23 | 2023-11-23 | Solar radiation amount estimation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311568129.8A CN117272002B (en) | 2023-11-23 | 2023-11-23 | Solar radiation amount estimation method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117272002A CN117272002A (en) | 2023-12-22 |
| CN117272002B true CN117272002B (en) | 2024-02-20 |
Family
ID=89206753
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311568129.8A Active CN117272002B (en) | 2023-11-23 | 2023-11-23 | Solar radiation amount estimation method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117272002B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114089834A (en) * | 2021-12-27 | 2022-02-25 | 杭州电子科技大学 | Electroencephalogram identification method based on time-channel cascade Transformer network |
| CN114781744A (en) * | 2022-05-07 | 2022-07-22 | 东南大学 | Deep learning multi-step long radiance prediction method based on codec |
| CN115511220A (en) * | 2022-11-02 | 2022-12-23 | 河海大学 | Ultra-short-term solar radiation prediction method and system based on cross-mode attention mechanism |
| CN115689055A (en) * | 2022-11-22 | 2023-02-03 | 河海大学 | Short-term solar irradiance prediction method and device |
| CN116612393A (en) * | 2023-05-05 | 2023-08-18 | 北京思源知行科技发展有限公司 | Solar radiation prediction method, system, electronic equipment and storage medium |
| CN116757321A (en) * | 2023-06-16 | 2023-09-15 | 西安交通大学 | Solar direct radiation quantity prediction method, system, equipment and storage medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11009536B2 (en) * | 2016-10-05 | 2021-05-18 | Telecom Italia S.P.A. | Method and system for estimating energy generation based on solar irradiance forecasting |
-
2023
- 2023-11-23 CN CN202311568129.8A patent/CN117272002B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114089834A (en) * | 2021-12-27 | 2022-02-25 | 杭州电子科技大学 | Electroencephalogram identification method based on time-channel cascade Transformer network |
| CN114781744A (en) * | 2022-05-07 | 2022-07-22 | 东南大学 | Deep learning multi-step long radiance prediction method based on codec |
| CN115511220A (en) * | 2022-11-02 | 2022-12-23 | 河海大学 | Ultra-short-term solar radiation prediction method and system based on cross-mode attention mechanism |
| CN115689055A (en) * | 2022-11-22 | 2023-02-03 | 河海大学 | Short-term solar irradiance prediction method and device |
| CN116612393A (en) * | 2023-05-05 | 2023-08-18 | 北京思源知行科技发展有限公司 | Solar radiation prediction method, system, electronic equipment and storage medium |
| CN116757321A (en) * | 2023-06-16 | 2023-09-15 | 西安交通大学 | Solar direct radiation quantity prediction method, system, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117272002A (en) | 2023-12-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10762305B2 (en) | Method for generating chatting data based on artificial intelligence, computer device and computer-readable storage medium | |
| Ayala Solares et al. | Modeling and prediction of global magnetic disturbance in near‐Earth space: A case study for Kp index using NARX models | |
| CN109685246B (en) | Environment data prediction method and device, storage medium and server | |
| CN111784041B (en) | Wind power prediction method and system based on graph convolution neural network | |
| CN114493050B (en) | Multi-dimensional fusion new energy power parallel prediction method and device | |
| Hou et al. | D2CL: A dense dilated convolutional LSTM model for sea surface temperature prediction | |
| CN110991284B (en) | Optical remote sensing image statement description generation method based on scene pre-classification | |
| Gu et al. | System identification and data‐driven forecasting of AE index and prediction uncertainty analysis using a new cloud‐NARX model | |
| CN112508265A (en) | Time and activity multi-task prediction method and system for business process management | |
| CN115882454B (en) | Source load power prediction method and system based on multi-element information fusion | |
| CN112668238A (en) | Rainfall processing method, device, equipment and storage medium | |
| CN114091361B (en) | Weather event-based transform model construction method | |
| Said et al. | AI-based solar energy forecasting for smart grid integration | |
| CN114004152B (en) | Multi-wind-field wind speed space-time prediction method based on graph convolution and recurrent neural network | |
| Wang et al. | Customized deep learning for precipitation bias correction and downscaling | |
| CN117272002B (en) | Solar radiation amount estimation method and device, electronic equipment and storage medium | |
| CN112990530A (en) | Regional population number prediction method and device, electronic equipment and storage medium | |
| CN117197632A (en) | Transformer-based electron microscope pollen image target detection method | |
| CN115994317A (en) | Incomplete multi-view multi-label classification method and system based on depth contrast learning | |
| US20220327335A1 (en) | Controlling asynchronous fusion of spatio-temporal multimodal data | |
| Xu et al. | Minutely multi-step irradiance forecasting based on all-sky images using LSTM-InformerStack hybrid model with dual feature enhancement | |
| CN113011674A (en) | Photovoltaic power generation prediction method and device, electronic equipment and storage medium | |
| CN117493583B (en) | Method and system for generating flow operation sequence by combining event log and knowledge graph | |
| Tan et al. | OSP-FEAN: Optimizing Significant Wave Height Prediction with Feature Engineering and Attention Network | |
| CN117830828A (en) | Satellite signal quality prediction model construction method, device and prediction method |
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 |