CN116227554A - Analog data correction method and device for meteorological data and electronic equipment - Google Patents
Analog data correction method and device for meteorological data and electronic equipment Download PDFInfo
- Publication number
- CN116227554A CN116227554A CN202310232619.4A CN202310232619A CN116227554A CN 116227554 A CN116227554 A CN 116227554A CN 202310232619 A CN202310232619 A CN 202310232619A CN 116227554 A CN116227554 A CN 116227554A
- Authority
- CN
- China
- Prior art keywords
- data
- layer
- simulation
- convolution
- 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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/02—Instruments for indicating weather conditions by measuring two or more variables, e.g. humidity, pressure, temperature, cloud cover or wind speed
-
- 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
- G06N3/08—Learning methods
-
- 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)
- Physics & Mathematics (AREA)
- Environmental & Geological Engineering (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Ecology (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Biodiversity & Conservation Biology (AREA)
- Data Mining & Analysis (AREA)
- Environmental Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Atmospheric Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
Abstract
The application discloses a simulation data correction method and device of meteorological data and electronic equipment. The method comprises the following steps: acquiring actual measurement data of meteorological data at a target moment; processing the measured data according to a weather forecast simulation mode to obtain simulation data; the simulation data is input into a trained correction model, the correction model outputs target data corrected by the simulation data, wherein the correction model is a convolution long-short-term memory network model and is trained by a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data. The method solves the problem that the accuracy of the corrected simulation data is still low only by a statistical method when the simulation data of the meteorological data is corrected in the related art.
Description
Technical Field
The application relates to the field of new energy power generation, in particular to a simulation data correction method and device of meteorological data and electronic equipment.
Background
Existing weather forecast data is typically simulated based on a WRF (weather forecast, weather Research and Forecasting) model, but comparing WRF simulated data with actual wind data can find that there is typically some deviation between the two. To ensure accuracy of the data, the WRF analog data needs to be modified so that the forecast data matches the actual data as much as possible.
Existing WRF mode corrections are typically based on traditional data statistics methods or machine learning algorithms, and are mostly corrected for wind speed data. For example, patent CN 106934191A provides a WRF mode wind speed correction method based on self-similarity, and establishes a real-time processing module based on a correction algorithm to implement real-time correction of WRF mode forecast wind speed; patent CN 114358398A proposes a numerical weather forecast wind speed correction method based on a deep neural network, which compensates the problem of insufficient data feature extraction caused by small data volume by establishing a dynamic correction strategy, and reduces the numerical weather forecast wind speed prediction error. The correction of wind direction data generally adopts the same technical route as wind speed correction, and no clear distinction is made.
In modeling the wind direction correction, a mean square error function is still generally adopted for selecting a model cost function, and a more accurate cost function is not designed for the unique characteristics of the wind direction (for example, 360 degrees and 0 degrees are actually the same wind direction, the error calculated value is supposed to be 0, but the error value calculated by the mean square error function is 360).
Aiming at the problem that the accuracy of the corrected analog data is still low only by a statistical method when the analog data of the meteorological data is corrected in the related art, no effective solution is proposed at present.
Disclosure of Invention
The main purpose of the application is to provide a method and a device for correcting analog data of meteorological data and electronic equipment, so as to solve the problem that the accuracy of the corrected analog data is still low only by a statistical method when the analog data of the meteorological data is corrected in the related technology.
In order to achieve the above object, according to one aspect of the present application, there is provided a simulation data correction method of meteorological data, including: acquiring actual measurement data of meteorological data at a target moment; processing the measured data according to a weather forecast simulation mode to obtain simulation data; and inputting the simulation data into a trained correction model, and outputting target data corrected by the simulation data by the correction model, wherein the correction model is a convolution long-short-term memory network model and is formed by training a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data.
Optionally, inputting the simulation data into a trained correction model, and outputting the target data corrected by the simulation data by the correction model includes: after the analog data is processed through an input layer, the analog data is input into a convolution layer and a pooling layer for processing, wherein the pooling layer is arranged behind each convolution layer; carrying out dimension reduction tiling on the output data of the pooling layer through a data recombination layer; inputting the data subjected to dimension reduction tiling into a long-period and short-period memory network layer; and determining the target data according to the output data of the long-term and short-term memory network layer.
Optionally, determining the target data according to the output data of the long-term memory network layer includes: performing dimension reduction integration treatment on the output data of the long-period memory network layer through a full-connection layer; and carrying out linear transformation on the data subjected to the dimension reduction integration through an output layer to obtain the target data.
Optionally, after the analog data is processed by the input layer, the processing of the input convolution layer and the pooling layer includes: after the analog data is processed by an input layer, the analog data is input into a first convolution layer; outputting first convolution data by the first layer convolution layer, and inputting the first convolution data into a first pooling layer arranged after the first convolution layer; outputting first pooled data by the first pooled layer, inputting the first pooled data into a second convolutional layer; outputting second convolution data by the second convolution layer, and inputting the second convolution data into a second pooling layer arranged after the second convolution layer; until the last pooling layer outputs the processing result.
Optionally, before inputting the simulation data into the trained correction model and outputting the target data corrected by the simulation data by the correction model, the method further includes: obtaining simulation data and actual measurement data at a target point position, wherein the simulation data and the actual measurement data carry time stamps, the simulation data and the actual measurement data at the same time are a group, and the simulation data are obtained in a weather forecast simulation mode according to the existing actual measurement data and the weather data at a certain moment before the simulation data are the moment; dividing the simulation data and the actual measurement data into a training set and a testing set according to time, wherein the time span of the testing set is not smaller than the preset proportion of the total time span of the testing set and the training set; training an initial model through a training set until the requirement of a loss function is met, wherein the initial model is created according to a convolutional neural network and a long-term and short-term memory network; testing an initial model meeting the requirement of the loss function by using a test set, determining that the training of the initial model is completed under the condition that the test result meets the preset accurate requirement, and taking the initial model after the training is completed as the correction model; and initializing the initial model under the condition that the test result does not meet the preset accurate requirement, and retraining the initialized initial model by utilizing the training set.
Optionally, the initial model is trained by the training set until the requirement of the loss function is met, and the method further comprises: determining simulation data input by an initial model and output measured data, wherein the simulation data and the measured data comprise wind directions; converting the angular coordinates of the first wind direction and the second wind direction into rectangular coordinates, wherein the input simulation data comprise the first wind direction, and the output measured data comprise the second wind direction; determining an error formula according to the first wind direction and the second wind direction, wherein the error formula defines the error as the length of a chord determined by coordinates of the first wind direction and the second wind direction on a unit circle; and determining the loss function according to a preset proportion and the error formula.
Optionally, the loss function is calculated by:
wherein error is a loss value, eta is a preset proportion, and->
For the length of the string, θ 1 For the angle of the first wind direction, θ 2 Is the angle of the second wind direction.
In order to achieve the above object, according to another aspect of the present application, there is provided an analog data correction device of meteorological data, including: the acquisition module is used for acquiring the actual measurement data of the meteorological data at the target moment; the simulation module is used for processing the actual measurement data according to a weather forecast simulation mode to obtain simulation data; the correction module is used for inputting the simulation data into a trained correction model, and outputting target data corrected by the simulation data by the correction model, wherein the correction model is a convolution long-short-term memory network model and is formed by training a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data.
According to another aspect of the present application, there is also provided a computer-readable storage medium storing a program, wherein the program performs the analog data correction method of meteorological data described in any one of the above.
According to another aspect of the present application, there is also provided an electronic device including one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method for correcting analog data of meteorological data as described in any one of the above.
Through the application, the following steps are adopted: acquiring actual measurement data of meteorological data at a target moment; processing the measured data according to a weather forecast simulation mode to obtain simulation data; the simulation data is input into a trained correction model, the correction model outputs target data corrected by the simulation data, wherein the correction model is a convolution long-short-term memory network model and is trained by a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data. The correction model corrects the simulation data obtained by the weather forecast simulation mode, so that the technical effects of the correction efficiency and accuracy of the simulation data are improved, and the problem that the accuracy of the corrected simulation data is still low due to the fact that the statistics method is only adopted when the simulation data of the weather data are corrected in the related technology is solved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application, illustrate and explain the application and are not to be construed as limiting the application. In the drawings:
FIG. 1 is a flow chart of a method for modeling data modification of meteorological data according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a modified model structure provided in accordance with an embodiment of the present application;
FIG. 3-1 is a schematic diagram of wind direction provided according to an embodiment of the present application;
FIG. 3-2 is a schematic diagram of wind direction coordinate conversion provided according to an embodiment of the present application;
FIG. 4-1 is a schematic diagram of test results for a first point location provided in accordance with an embodiment of the present application;
FIG. 4-2 is a schematic diagram of test results for a first point location provided in accordance with an embodiment of the present application;
FIGS. 4-3 are schematic diagrams of test results for a first site provided in accordance with embodiments of the present application;
FIG. 5 is a schematic diagram of an analog data correction device for meteorological data according to an embodiment of the present application;
fig. 6 is a schematic diagram of an electronic device according to an embodiment of the present application.
Detailed Description
It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate in order to describe the embodiments of the present application described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention will be described with reference to preferred embodiments, and FIG. 1 is a flowchart of a method for modifying simulated data of meteorological data according to an embodiment of the present application, as shown in FIG. 1, comprising the steps of:
step S101, obtaining actual measurement data of meteorological data at a target moment;
step S102, processing the measured data according to a weather forecast simulation mode to obtain simulation data;
step S103, inputting the simulation data into a trained correction model, and outputting target data corrected by the simulation data by the correction model, wherein the correction model is a convolution long-short-term memory network model and is trained by a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data.
Through the steps, the actual measurement data of the meteorological data at the target moment is obtained; processing the measured data according to a weather forecast simulation mode to obtain simulation data; the simulation data is input into a trained correction model, the correction model outputs target data corrected by the simulation data, wherein the correction model is a convolution long-short-term memory network model and is trained by a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data. The correction model corrects the simulation data obtained by the weather forecast simulation mode, so that the technical effects of the correction efficiency and accuracy of the simulation data are improved, and the problem that the accuracy of the corrected simulation data is still low due to the fact that the statistics method is only adopted when the simulation data of the weather data are corrected in the related technology is solved.
The main body of execution of the above steps may be a processor, a calculator, a server, or the like, a device having data calculation and data analysis processing capabilities, or may be a device having the above device having data calculation, analysis and processing capabilities, for example, a computer having a processor, a smart phone, a wearable device, or the like, a data system having a server, an arithmetic system, or the like.
The weather data may be weather parameters that may be simulated by the weather forecast simulation WRF, such as wind force, wind direction, wind speed, air temperature, pressure, etc. The weather forecast simulation WRF can simulate the possible later development conditions according to the existing weather data in a certain mode to obtain simulation data. The above simulation data can also be considered as values that may develop after the meteorological parameters.
However, due to the limitation of the weather forecast simulation mode, certain errors exist in the weather forecast simulation mode. In order to improve the accuracy of the simulation data and reduce the error of weather forecast simulation, the simulation data can be corrected based on a traditional data statistics method or be predicted directly by using a machine learning algorithm. The traditional data statistics method corrects the analog data, has low accuracy and needs to rely on a large amount of data. The way machine learns models to predict directly, its accuracy is still low due to technical limitations.
Therefore, the simulation data is corrected by using the correction model of the machine learning algorithm through the simulation data, so that the advantages of the weather simulation forecasting algorithm and the machine learning algorithm are combined, and the accuracy of the simulation data is improved.
The correction model is a convolution long-term and short-term memory network model, so that the correction model is more related to the time characteristics of data, and the accuracy of the simulation data prediction is improved.
Optionally, inputting the simulation data into a trained correction model, and outputting the target data corrected by the simulation data by the correction model includes: after the analog data is processed through an input layer, the analog data is input into a convolution layer and a pooling layer for processing, wherein the pooling layer is arranged behind each convolution layer; carrying out dimension reduction tiling on the output data of the pooling layer through the data reorganization layer; inputting the data subjected to dimension reduction tiling into a long-period and short-period memory network layer; and determining target data according to the output data of the long-term and short-term memory network layer.
As shown in FIG. 2, the correction model built based on the convolution long-short-term memory network can comprise an input layer, a convolution layer, a pooling layer, a data reorganization layer, a long-short-term memory layer, a full connection layer and an output layer. At least one layer of convolution layers is provided with a corresponding pooling layer after each convolution layer, and the output of the pooling layer corresponding to the previous convolution layer is used as the input of the next convolution layer. The number of the long-period memory layers is at least one, and the at least one long-period memory layer is connected end to end.
After the analog data is processed through an input layer, the analog data is input into a convolution layer and a pooling layer for processing, wherein the pooling layer is arranged behind each convolution layer; carrying out dimension reduction tiling on the output data of the pooling layer through the data reorganization layer; inputting the data subjected to dimension reduction tiling into a long-period and short-period memory network layer; and determining target data according to the output data of the long-term and short-term memory network layer, thereby obtaining corrected analog data. The analog data may be processed by the input layer, and may be processed by data amplification or the like.
Optionally, determining the target data according to the output data of the long-term memory network layer includes: output data of the long-period memory network layer is subjected to dimension reduction integration treatment through the full-connection layer; and performing linear transformation on the data subjected to the dimension reduction integration through an output layer to obtain target data.
The long-term and short-term memory network layer processes data by using a preset updating rule. Specifically, the LSTM cell update rule in the kth (k=1, 2, …, S) (S is the number of LSTM layers) LSTM layer is as follows:
δ(x)＝1/(1+e -x )
wherein X is t ,h t ,S t The input at the time t, the hidden layer state and the LSTM unit state; f (f) t ,i t ,o t A forget gate, an input gate and an output gate at the time t; w (W) fx ,W fh ,b f Corresponding weights and biases for the forgetting gate; w (W) ix ,W ih ,b i The input gate corresponds to the weight and the bias; w (W) ox ,W oh ,b o Corresponding weight and bias of the output gate; w (W) sx ,W sh ,b s Corresponding weights and biases for the LSTM units.
The full-connection layer performs dimension reduction integration on the output of the S-th LSTM layer, and the calculation mode is as follows:
in which W is q ,b q The weight and the bias are corresponding to the full connection layer.
The final output of the C-LSTM network can be obtained by linearly transforming the output q of the full connection layer:
in which W is yh ,b y Is the output layer weight and bias.
Representing the corrected model output, i.e., corrected simulation data.
Optionally, after the analog data is processed by the input layer, the processing of the input convolution layer and the pooling layer includes: after the analog data is processed by the input layer, inputting the analog data into the first convolution layer; outputting first convolution data by the first convolution layer, and inputting the first convolution data into a first pooling layer arranged after the first convolution layer; outputting first pooled data by the first pooled layer and inputting the first pooled data into the second convolved layer; outputting second convolution data by the second convolution layer, and inputting the second convolution data into a second pooling layer arranged after the second convolution layer; until the last pooling layer outputs the processing result.
The convolution layer processes the analog data to give a first convolution layer input X (l-1) (l=1, 2, …, L), when l=1, X (0) Namely, the input of the convolution layer; when l>1，X (l-1) The output of the first-1 pooling layer is obtained. The specific calculation mode of the convolution layer is as follows:
f(x)＝max(0,x)
wherein M and N are dimensions of input data; w (w) ij Is a convolution kernel of size (J, K); y is Y l Output for the first convolutional layer.
After the first convolution layer completes the operation, the layer outputs Y l To the first pooling layer. The pooling layer gathers and aggregates the features extracted by the convolution layer to gradually realize the feature expression from high level to low level. The first pooling layer is calculated as follows:
wherein pool () is a pooling function, and can select maximum pooling or average pooling; s is the pooling layer window size.
Optionally, before the simulation data is input into the trained correction model and the target data corrected by the simulation data is output by the correction model, the method further includes: obtaining simulation data and actual measurement data at a target point position, wherein the simulation data and the actual measurement data carry time stamps, the simulation data and the actual measurement data at the same time are a group, and the simulation data are obtained in a weather forecast simulation mode according to the existing actual measurement data and weather data at a certain moment before the simulation data are at a certain moment; dividing the simulation data and the actual measurement data into a training set and a testing set according to time, wherein the time span of the testing set is not smaller than the preset proportion of the total time span of the testing set and the training set; training an initial model through a training set until the requirement of a loss function is met, wherein the initial model is created according to a convolutional neural network and a long-term and short-term memory network; testing an initial model meeting the requirement of the loss function by using a test set, determining that the training of the initial model is completed under the condition that the test result meets the preset accurate requirement, and taking the initial model after the training is completed as a correction model; under the condition that the test result does not meet the preset accurate requirement, initializing an initial model, and retraining the initialized initial model by utilizing a training set.
The weather data is wind factor data including wind speed and wind direction. The target point is an actual meteorological data acquisition point, and may be one or more. Multiple target sites are also analyzed separately. The simulation data of the target point location is obtained by simulation of weather forecast simulation WRF. The simulation data and the actual measurement data are provided with time stamps, and the simulation data and the actual measurement data at the same time can be used as a group of data to train the correction model.
The training set is used for training the correction model, and the test data is used for testing the correction model. The above-mentioned preset proportion may be 10%. For example, the time span of the simulation data and the measured data is 1 month to 10 months in a certain year, so that the training set can select 1 month to 8 months in the same year, and the test set can select 9 months to 10 months in the same year.
After the training set trains the initial model to meet the requirement of the loss function, the testing set is utilized to test, and under the condition that the testing result of the testing set shows that the accuracy of the initial model reaches the preset threshold value, the initial model can be used as a correction model. Under the condition that the test result does not meet the preset accuracy requirement, initializing the initial model, and retraining the initialized initial model by utilizing the training set until the test result of the test set indicates that the accuracy of the initial model reaches a preset threshold value.
Optionally, training the initial model through the training set until the requirement of the loss function is met, the method further includes: determining simulation data input by an initial model and actual measurement data output by the initial model, wherein the simulation data and the actual measurement data comprise wind directions; converting the angle coordinates of the first wind direction and the second wind direction into rectangular coordinates, wherein the input simulation data comprise the first wind direction, and the output measured data comprise the second wind direction; determining an error formula according to the first wind direction and the second wind direction, wherein the error formula defines the error as the length of a chord determined by coordinates of the first wind direction and the second wind direction on a unit circle; and determining a loss function according to a preset proportion and an error formula.
The loss functions may also include wind speed loss functions, as well as wind direction loss functions. The wind speed loss function may be an average absolute error and/or a root mean square error. The wind direction loss function may be error as described above. The simulation data is corrected through the loss function of the wind speed and the wind direction, so that the correction effect is better, and the error of the corrected simulation data is smaller.
Alternatively, the loss function is calculated by:
wherein error is a loss value, eta is a preset proportion, and- >
Length of chord, θ 1 Angle of the first wind direction, θ 2 Is the angle of the second wind direction.
As shown in fig. 3-1, the corresponding angles of the first wind direction 1 and the second wind direction 2 are respectively theta 1 θ 2 The wind direction is represented on a unit circle, and the angular coordinates are converted into rectangular coordinates, as shown in fig. 3-2:
θ 1 →(x 1 ,y 1 )＝(sinθ 1 ,cosθ 1 )
θ 2 →(x 2 ,y 2 )＝(sinθ 2 ,cosθ 2 )
defining the error value between the first wind direction 1 and the second wind direction 2 as the length of the chord determined by two coordinates on the unit circle, namely d, as shown in fig. 3-2, the calculation process is as follows:
because the error 'value range is [0,2], in order to omit the data normalization process of wind direction variable in the network training process, the error' is scaled in equal proportion as follows:
and the model is used as a loss function in the training process of the convolution long-term and short-term memory network model.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in a different order than that illustrated herein.
It should be noted that this application also provides an alternative embodiment, and the following detailed description of this embodiment is provided.
The embodiment provides a WRF mode wind direction correction method based on a convolution long-short term memory network, which comprises the following specific implementation steps:
1. obtaining simulated WRF (weather forecast simulation) mode wind element data at specific points, wherein the simulated WRF mode wind element data comprises variables such as wind speed, wind direction and other meteorological elements (precipitation, radiation, temperature and humidity and the like) of each height layer, and actual wind measuring wind direction data (which can be from a wind measuring tower, laser radar wind measuring equipment and the like) at the same point;
2. the actual wind-measuring wind direction data of the point location and the corresponding WRF mode element measurement data are selected to be combined into a model training and test data set, the time span of the test set is usually not less than 10% of the total data set according to the time sequence of the division of the training set and the test set.
3. Taking each related variable (mode simulation wind direction, wind speed, temperature and humidity and the like) as input of a correction model, taking measured wind direction data as output of the correction model, and adopting a convolution long-term and short-term memory network for the correction model;
4. the convolution long-short-term memory network consists of seven parts, namely an input layer, a convolution layer, a pooling layer, a data recombination layer, a long-short-term memory layer, a full connection layer and an output layer, wherein the specific structure is shown in fig. 2, and fig. 2 is a schematic diagram of a correction model structure provided according to an embodiment of the application;
5. Given the first convolutional layer input X (l-1) (l=1, 2, …, L), when l=1, X (0) Namely, the input of the network; when l>1，X (l-1) The output of the first-1 pooling layer is obtained. The specific calculation mode of the convolution layer is as follows:
f(x)＝max(0,x)
wherein M and N are dimensions of input data; w (w) ij Is a convolution kernel of size (J, K); y is Y l Output for the first convolutional layer.
6. After the first convolution layer completes the operation, the layer outputs Y l To the first pooling layer. The pooling layer gathers and aggregates the features extracted by the convolution layer to gradually realize the feature expression from high level to low level. The first pooling layer is calculated as follows:
wherein pool () is a pooling function, and can select maximum pooling or average pooling; s is the pooling layer window size.
7. After the convolution layer and the pooling layer finish calculation, the data reorganization layer outputs the data X of the last pooling layer L The multidimensional features are tiled in a dimension reducing way and used for inputting an LSTM layer.
The LSTM cell update rule in the kth (k=1, 2, …, S) (S is the number of LSTM layers) LSTM layer is as follows:
δ(x)＝1/(1+e -x )
wherein X is t ,h t ,S t The input at the time t, the hidden layer state and the LSTM unit state; f (f) t ,i t ,o t A forget gate, an input gate and an output gate at the time t; w (W) fx ,W fh ,b f Corresponding weights and biases for the forgetting gate; w (W) ix ,W ih ,b i The input gate corresponds to the weight and the bias; w (W) ox ,W oh ,b o Corresponding weight and bias of the output gate; w (W) sx ,W sh ,b s Corresponding weights and biases for the LSTM units.
9. The full-connection layer performs dimension reduction integration on the output of the S-th LSTM layer, and the calculation mode is as follows:
in which W is q ,b q Corresponding weight and bias for the full connection layer;
10. the final output of the C-LSTM network can be obtained by linearly transforming the output q of the full connection layer:
in which W is yh ,b y Is the output layer weight and bias.
11. The derivation process of the convolution long-short-term memory network model cost function is as follows:
FIG. 3-1 is a schematic diagram of wind direction according to an embodiment of the present application, as shown in FIG. 3-1, the corresponding angles of the first wind direction 1 and the second wind direction 2 are respectively θ 1 θ 2 The wind direction is represented on a unit circle, and the angular coordinate is converted into the rectangular coordinate, as shown in fig. 3-2, and fig. 3-2 is a schematic diagram of wind direction coordinate conversion provided according to an embodiment of the present application:
θ 1 →(x 1 ,y 1 )＝(sinθ 1 ,cosθ 1 )
θ 2 →(x 2 ,y 2 )＝(sinθ 2 ,cosθ 2 )
defining the error value between the first wind direction 1 and the second wind direction 2 as the length of the chord determined by two coordinates on the unit circle, namely d, as shown in fig. 3-2, the calculation process is as follows:
because the error 'value range is [0,2], in order to omit the data normalization process of wind direction variable in the network training process, the error' is scaled in equal proportion as follows:
And taking the model as a cost function in the training process of the convolution long-term and short-term memory network model;
12. the convolutional long-short-term memory network model training adopts a self-adaptive momentum random optimization algorithm, a test set is adopted to carry out model test after the training is finished, the model training is completed after the precision setting requirement is met, otherwise, the network is reinitialized and the steps 6 to 12 are repeated;
13. the trained convolution long-term and short-term memory network model is used as a WRF mode wind direction correction model at the point location, and the wind direction data correction process can be realized by inputting required relevant data.
The embodiment provides an error calculation function which is applicable to wind direction data characteristics and accords with trigonometric function rules; and establishing an applicable error correction model aiming at the WRF mode wind direction data, adopting a designed error function as a cost function for model training, and verifying the validity and applicability of the provided model and the error function through comparison and verification.
The embodiment designs an error calculation formula which is more in line with the actual rule of the wind direction error, and the formula adopts a trigonometric function form, so that the calculation of the wind direction error is more reasonable; the design of the wind direction error calculation formula adopts equal proportion scaling, the error value range is adjusted to be [0,1], and the data normalization process in the subsequent wind direction correction model training process is omitted; a correction model specially aiming at wind direction variable in WRF mode simulation data is designed, and model feasibility and applicability verification is carried out by adopting related data.
The present embodiment is described in further detail below in conjunction with the associated drawings and detailed description.
And acquiring WRF mode simulation data of a certain area and actual anemometry data (acquired by a anemometer tower) at 3 point positions in the area, wherein the WRF mode simulation data are respectively a first point position 1, a second point position 2 and a third point position 3. The WRF mode simulation data comprise variables such as wind direction, wind speed, air temperature, pressure, u wind component, v wind component and the like at different height layers (100 m, 90m, … and 10 m); the actual anemometry data comprises 100 meters of wind direction data in a corresponding time period, and the time resolution is 10 minutes. The WRF mode simulation data are used for inputting the correction model, and the actual wind direction data are used for outputting the correction model.
The input source of the correction model is constructed by adopting the 54 variable data after selection, the input source is a two-dimensional data characteristic diagram, the row information represents the time sequence length of a sample, the column information represents the number of characteristic variables, the characteristic diagram in the embodiment is 72 x 54, and the output source of the correction model is actually measured wind direction data at a selected point.
Inputting the constructed input data into a correction model, wherein the model adopts 1 convolution layer and 1 pooling layer, the input data is subjected to convolution calculation in the convolution layer, the convolution kernel size is set to be 3 multiplied by 3, the number is 32, and the step length is 2; the size of a pooling layer window is set to be 3 multiplied by 3, the dimension reduction operation is carried out on the convolved feature mapping by adopting a maximum pooling method, and the step sizes are respectively 1; the data reorganization layer further performs dimension reduction compression on the pooled three-dimensional feature vectors to form two-dimensional data; the two-dimensional data is sent to the LSTM part for calculation, wherein the part uses 1 LSTM layer, and the number of neurons is set to be 32; the number of neurons in the full connection layer is set to 16; the output layer neuron is set to 1, and the final output, i.e., the corrected pattern wind direction data, is generated.
The convolution long-term and short-term memory network training adopts a self-adaptive momentum random optimization algorithm. The data at the first point 1 is used for model training and testing, the training set is selected from 1 month data to 8 months data in the same year, and the testing set is selected from 9 months data to 10 months data in the same year.
The following error indexes are selected to evaluate the model effect:
wherein, the liquid crystal display device comprises a liquid crystal display device,
representing the corrected model output (i.e. corrected WRF mode wind speed data), y i Representing actual anemometry data, n is the number of samples. MAE is the mean absolute error, mean Absolute Error, RMSE is the root mean square error, root Mean Squared Error.
The effect of the model test at the first point 1 is shown in fig. 4-1, and fig. 4-1 is a schematic diagram of the test result of the first point according to the embodiment of the present application. In order to verify the applicability of the constructed wind direction correction model, the data at the second point location 2 and the third point location 3 are adopted to train and test the convolution long-short-period memory network respectively, the test effects are shown in fig. 4-2 and fig. 4-3 respectively, fig. 4-2 is a schematic diagram of the test result of the first point location provided according to the embodiment of the application, and fig. 4-3 is a schematic diagram of the test result of the first point location provided according to the embodiment of the application.
In order to further verify the applicability of the proposed wind direction error calculation function, a conventional mean square error function is adopted to conduct remodelling on a convolution long-short-period memory network, statistics of indexes of comparison results of the two are shown in a table 1, and the table 1 is a correction model test effect data table.
Table 1 correction model test effect data table
As can be seen from the error comparison results in fig. 4-1 to fig. 4-3 and table 1, the correction model of the convolution long-short-period memory network constructed in this embodiment can be effectively applied to WRF mode wind direction correction, and the applicability of the error calculation function conforming to the trigonometric function rule is also verified.
The embodiment of the application also provides a simulation data correction device for the meteorological data, and it is to be noted that the simulation data correction device for the meteorological data in the embodiment of the application can be used for executing the simulation data correction method for the meteorological data provided in the embodiment of the application. The following describes an analog data correction device for meteorological data provided in the embodiment of the present application.
FIG. 5 is a schematic diagram of an apparatus for correcting analog data of meteorological data according to an embodiment of the present application, as shown in FIG. 5, the apparatus includes: the acquisition module 51, the simulation module 52, the correction module 53, the device is described in detail below.
An acquisition module 51, configured to acquire measured data of meteorological data at a target time; the simulation module 52 is connected with the acquisition module 51, and is used for processing the actual measurement data according to a weather forecast simulation mode to obtain simulation data; the correction module 53 is connected to the simulation module 52, and is configured to input the simulation data into a trained correction model, and output the target data corrected by the simulation data from the correction model, where the correction model is a convolutional long-short-term memory network model and is trained by multiple sets of training data, and each set of training data includes the input simulation data and the output actual measurement data.
The simulation data correction device for the meteorological data acquires actual measurement data of the meteorological data at a target moment; processing the measured data according to a weather forecast simulation mode to obtain simulation data; the simulation data is input into a trained correction model, the correction model outputs target data corrected by the simulation data, wherein the correction model is a convolution long-short-term memory network model and is trained by a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data. The correction model corrects the simulation data obtained by the weather forecast simulation mode, so that the technical effects of the correction efficiency and accuracy of the simulation data are improved, and the problem that the accuracy of the corrected simulation data is still low due to the fact that the statistics method is only adopted when the simulation data of the weather data are corrected in the related technology is solved.
The analog data correction device of meteorological data comprises a processor and a memory, wherein the acquisition module 51, the analog module 52, the correction module 53 and the like are all stored in the memory as program units, and the processor executes the program units stored in the memory to realize corresponding functions.
The processor includes a kernel, and the kernel fetches the corresponding program unit from the memory. The kernel can be provided with one or more than one kernel, and the problem that the accuracy of the corrected simulation data is still low due to a statistical method only when the simulation data of the meteorological data is corrected in the related technology is solved by adjusting the kernel parameters.
The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip.
An embodiment of the present invention provides a computer-readable storage medium having stored thereon a program that, when executed by a processor, implements a method for modifying analog data of meteorological data.
The embodiment of the invention provides a processor which is used for running a program, wherein the program runs to execute a simulation data correction method of meteorological data.
Fig. 6 is a schematic diagram of an electronic device according to an embodiment of the present application, and as shown in fig. 6, an embodiment of the present application provides an electronic device 60, where the device includes a processor, a memory, and a program stored on the memory and executable on the processor, and the processor implements steps of any of the methods described above when executing the program.
The device herein may be a server, PC, PAD, cell phone, etc.
The present application also provides a computer program product adapted to perform a program initialized with any of the above method steps when executed on an analog data modification device of meteorological data.
It will be appreciated by those skilled in the art that 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 weather data simulation data correction device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable weather data simulation data correction device, 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 correction device 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 modifying 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.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.
It will be appreciated by those skilled in the art that 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 foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the present application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. which are within the spirit and principles of the present application are intended to be included within the scope of the claims of the present application.
Claims (10)
1. A method for modifying analog data of meteorological data, comprising:
acquiring actual measurement data of meteorological data at a target moment;
processing the measured data according to a weather forecast simulation mode to obtain simulation data;
and inputting the simulation data into a trained correction model, and outputting target data corrected by the simulation data by the correction model, wherein the correction model is a convolution long-short-term memory network model and is formed by training a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data.
2. The method of claim 1, wherein inputting the simulated data into a trained correction model, and outputting the simulated data corrected target data from the correction model comprises:
after the analog data is processed through an input layer, the analog data is input into a convolution layer and a pooling layer for processing, wherein the pooling layer is arranged behind each convolution layer;
carrying out dimension reduction tiling on the output data of the pooling layer through a data recombination layer;
inputting the data subjected to dimension reduction tiling into a long-period and short-period memory network layer;
and determining the target data according to the output data of the long-term and short-term memory network layer.
3. The method of claim 2, wherein determining the target data based on the output data of the long-short term memory network layer comprises:
performing dimension reduction integration treatment on the output data of the long-period memory network layer through a full-connection layer;
and carrying out linear transformation on the data subjected to the dimension reduction integration through an output layer to obtain the target data.
4. The method of claim 2, wherein processing the analog data through the input layer, the input convolutional layer and the pooling layer comprises:
After the analog data is processed by an input layer, the analog data is input into a first convolution layer;
outputting first convolution data by the first layer convolution layer, and inputting the first convolution data into a first pooling layer arranged after the first convolution layer;
outputting first pooled data by the first pooled layer, inputting the first pooled data into a second convolutional layer;
outputting second convolution data by the second convolution layer, and inputting the second convolution data into a second pooling layer arranged after the second convolution layer;
until the last pooling layer outputs the processing result.
5. The method of claim 4, wherein the simulated data is input into a trained correction model, and wherein the method further comprises, prior to outputting the simulated data corrected target data from the correction model:
obtaining simulation data and actual measurement data at a target point position, wherein the simulation data and the actual measurement data carry time stamps, the simulation data and the actual measurement data at the same time are a group, and the simulation data are obtained in a weather forecast simulation mode according to the existing actual measurement data and the weather data at a certain moment before the simulation data are the moment;
dividing the simulation data and the actual measurement data into a training set and a testing set according to time, wherein the time span of the testing set is not smaller than the preset proportion of the total time span of the testing set and the training set;
Training an initial model through a training set until the requirement of a loss function is met, wherein the initial model is created according to a convolutional neural network and a long-term and short-term memory network;
testing an initial model meeting the requirement of the loss function by using a test set, determining that the training of the initial model is completed under the condition that the test result meets the preset accurate requirement, and taking the initial model after the training is completed as the correction model;
and initializing the initial model under the condition that the test result does not meet the preset accurate requirement, and retraining the initialized initial model by utilizing the training set.
6. The method of claim 5, wherein the initial model is trained by a training set until the requirement of the loss function is met, the method further comprising:
determining simulation data input by an initial model and output measured data, wherein the simulation data and the measured data comprise wind directions;
converting the angular coordinates of the first wind direction and the second wind direction into rectangular coordinates, wherein the input simulation data comprise the first wind direction, and the output measured data comprise the second wind direction;
Determining an error formula according to the first wind direction and the second wind direction, wherein the error formula defines the error as the length of a chord determined by coordinates of the first wind direction and the second wind direction on a unit circle;
and determining the loss function according to a preset proportion and the error formula.
7. The method of claim 6, wherein the loss function is calculated by:
wherein error is a loss value, eta is a preset proportion,
for the length of the string, θ 1 For the angle of the first wind direction, θ 2 Is the angle of the second wind direction.
8. An analog data correction device for meteorological data, comprising:
the acquisition module is used for acquiring the actual measurement data of the meteorological data at the target moment;
the simulation module is used for processing the actual measurement data according to a weather forecast simulation mode to obtain simulation data;
the correction module is used for inputting the simulation data into a trained correction model, and outputting target data corrected by the simulation data by the correction model, wherein the correction model is a convolution long-short-term memory network model and is formed by training a plurality of groups of training data, and each group of training data comprises the input simulation data and the output actual measurement data.
9. A computer-readable storage medium storing a program, wherein the program performs the analog data correction method of meteorological data according to any one of claims 1 to 7.
10. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of analog data modification of meteorological data of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310232619.4A CN116227554A (en) | 2023-03-06 | 2023-03-06 | Analog data correction method and device for meteorological data and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310232619.4A CN116227554A (en) | 2023-03-06 | 2023-03-06 | Analog data correction method and device for meteorological data and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116227554A true CN116227554A (en) | 2023-06-06 |
Family
ID=86585559
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310232619.4A Pending CN116227554A (en) | 2023-03-06 | 2023-03-06 | Analog data correction method and device for meteorological data and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116227554A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116992222A (en) * | 2023-09-27 | 2023-11-03 | 长江三峡集团实业发展(北京)有限公司 | Method, device, equipment and medium for migration learning of wind element correction model |
| CN117009716A (en) * | 2023-09-27 | 2023-11-07 | 长江三峡集团实业发展(北京)有限公司 | Weather forecast data error calculation model construction and weather forecast data correction method |
-
2023
- 2023-03-06 CN CN202310232619.4A patent/CN116227554A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116992222A (en) * | 2023-09-27 | 2023-11-03 | 长江三峡集团实业发展(北京)有限公司 | Method, device, equipment and medium for migration learning of wind element correction model |
| CN117009716A (en) * | 2023-09-27 | 2023-11-07 | 长江三峡集团实业发展(北京)有限公司 | Weather forecast data error calculation model construction and weather forecast data correction method |
| CN116992222B (en) * | 2023-09-27 | 2024-01-26 | 长江三峡集团实业发展(北京)有限公司 | Method, device, equipment and medium for migration learning of wind element correction model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gao et al. | Hourly forecasting of solar irradiance based on CEEMDAN and multi-strategy CNN-LSTM neural networks | |
| CN116227554A (en) | Analog data correction method and device for meteorological data and electronic equipment | |
| CN111310889B (en) | Evaporation waveguide profile estimation method based on deep neural network | |
| CN111815038B (en) | Photovoltaic ultra-short term prediction method and system | |
| CN112348292B (en) | Short-term wind power prediction method and system based on deep learning network | |
| CN113063737B (en) | Ocean heat content remote sensing inversion method combining remote sensing and buoy data | |
| CN109635245A (en) | A kind of robust width learning system | |
| CN111815027A (en) | Photovoltaic station generated power prediction method and system | |
| CN115859816A (en) | Wind power minute-level prediction method and system based on CNN-LSTM algorithm | |
| CN113505938A (en) | Ultra-short-term wind power combined prediction method and system | |
| CN116402203A (en) | Method, system and medium for predicting short-time photovoltaic power generation capacity considering weather conditions | |
| CN115689055A (en) | Short-term solar irradiance prediction method and device | |
| CN115186923A (en) | Photovoltaic power generation power prediction method and device and electronic equipment | |
| CN110852415B (en) | Vegetation index prediction method, system and equipment based on neural network algorithm | |
| CN110852475B (en) | Extreme gradient lifting algorithm-based vegetation index prediction method, system and equipment | |
| CN112560633A (en) | Plant key phenological period time point prediction method and system based on deep learning | |
| CN116757321A (en) | Solar direct radiation quantity prediction method, system, equipment and storage medium | |
| CN116316600A (en) | Wind power day-ahead prediction self-adaptive filtering method based on seasonal typing | |
| Chang et al. | Neural network with multi-trend simulating transfer function for forecasting typhoon wave | |
| CN115912334A (en) | Method for establishing prediction model of output guarantee rate of wind power plant and prediction method | |
| CN115907194A (en) | Photovoltaic power generation power prediction method, system and product | |
| CN112070320B (en) | Ultra-short-term wind power prediction method and system based on dynamic harmonic regression | |
| CN109460893A (en) | A kind of photovoltaic plant weather pattern correlation metric calculation method and system | |
| CN113095547B (en) | Short-term wind power prediction method based on GRA-LSTM-ICE model | |
| CN116992222B (en) | Method, device, equipment and medium for migration learning of wind element correction model |
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 |