Cloud picture prediction method and device
By collecting and merging cloud maps and meteorological data, hybrid characteristics are formed, and deep learning models are used to predict, which solves the problem of insufficient accuracy of panoramic image prediction in the existing technology, and achieves higher prediction accuracy.
Patent Information
- Application Number
- CN202310932605.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-26
- Publication Date
- 2025-01-28
AI Technical Summary
The existing methods of sky panoramic image prediction fail to make full use of meteorological data, resulting in insufficient accuracy of the predicted sky panoramic image.
By collecting multi -time data data and meteorological data at multiple time points, combined to form hybrid characteristics, and use deep learning models to predict based on these hybrid characteristics data to improve the predictive accuracy of the panoramic image of the sky.
By combining the characteristics of cloud maps and meteorological data, the accuracy of predicting cloud maps is significantly improved, and the accurate predictive needs of control systems such as photovoltaic systems to control systems in the sky are met.
Smart Images
Figure CN119377872A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the fields of data processing and deep learning, and in particular to a cloud image prediction method and device. Background Art
[0002] Photovoltaic systems need to be controlled according to changes in the sky in order to maximize power generation efficiency. Therefore, predicting future sky panoramic images is of great significance for the prediction and control tasks of photovoltaic control systems. At present, other meteorological data such as temperature, humidity, wind speed and radiation are also collected while collecting sky panoramic images. Meteorological data such as temperature, humidity, wind speed and radiation are closely related to changes in clouds and obstruction of sunlight. However, in the current sky panoramic image prediction method, only the collected sky panoramic images are used as feature data, and the corresponding meteorological data are not considered, resulting in insufficient accuracy of the predicted sky panoramic images. Summary of the invention
[0003] The present application provides a cloud map prediction method and device to solve the above-mentioned technical problems.
[0004] To this end, an embodiment of the present application provides a cloud map prediction method, the method comprising:
[0005] Collecting cloud images to be measured and meteorological data to be measured at multiple first time points, wherein the multiple first time points are different time points at intervals of preset time;
[0006] Determine the mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point;
[0007] Prediction is performed based on the mixed feature data of the multiple first time points to obtain a prediction cloud map of a preset number of second time points, wherein the multiple second time points are different time points separated by the preset time, and the earliest second time point is later than the latest first time point and separated by the preset time.
[0008] Wherein, determining the mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point includes:
[0009] Determine image feature data based on the cloud image to be measured at the first time point, wherein the image feature data is a matrix of three channels;
[0010] Determine a matrix of at least one channel based on the matrix sizes of the meteorological data to be measured and the image feature data at the first time point;
[0011] The matrix of the at least one channel is combined with the image feature data to obtain mixed feature data at the first time point.
[0012] The predicting based on the mixed feature matrix of the multiple time points includes:
[0013] Traversing the multiple first time points in chronological order;
[0014] Determine the first time feature data and the second time feature data of the current first time point based on the mixed feature data of the current first time point, the first time feature data and the second time feature data of the previous first time point, the first time feature data used for calculation at the first first time point is the preset first time feature data, and the second time feature data is the preset second time feature data;
[0015] After the traversal is completed, the first time feature data and the second time feature data of the last first time point are obtained.
[0016] Wherein, after the first time characteristic data and the second time characteristic data at the first second time point, it also includes:
[0017] Traversing the preset predicted number of second time points in chronological order;
[0018] Determine the first time feature data and the second time feature data of the current second time point based on the mixed feature data of the current second time point, the first time feature data and the second time feature data of the previous second time point, the first time feature data used for calculation at the first first time point is the first time feature data of the last first time point, and the second time feature data is the second time feature data of the last first time point;
[0019] Perform prediction based on the first time feature data and the second time feature data at the current second time point to obtain a prediction cloud map at the current second time point;
[0020] After the traversal is completed, a prediction cloud graph of a preset prediction number of second time points is obtained.
[0021] The determining of the second time feature data at the current time point based on the mixed feature data at the current time point, the first time feature data at the previous time point, and the second time feature data includes:
[0022] Determine first weighted feature data and second weighted feature data at the current time point based on the mixed feature data at the current time point, the first time feature data and the second time feature data at the previous time point;
[0023] The second time feature data at the current time point is determined based on the mixed feature data at the current time point, the first weighted feature data, the second weighted feature data and the first time feature data at the previous time point.
[0024] The step of determining the first time feature data at the next time point based on the mixed feature data at the current time point, the first time feature data at the previous time point, and the second time feature data includes:
[0025] Determine third weighted feature data at the current time point based on the mixed feature data at the current time point, the second time feature data, and the first time feature data at the previous time point;
[0026] The first time feature data at the current time point is determined based on the third weighted feature data and the second time feature data at the current time point.
[0027] Another aspect of the present application provides a cloud image prediction device, the device comprising:
[0028] A collection module, used for collecting cloud images to be measured and meteorological data to be measured at multiple first time points, wherein the multiple first time points are different time points with preset time intervals;
[0029] A calculation module, used to determine the mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point;
[0030] A deep learning module is used to make predictions based on the mixed feature data of the multiple first time points to obtain a prediction cloud map of a preset number of second time points, wherein the multiple second time points are different time points separated by the preset time, and the earliest second time point is later than the latest first time point and separated by the preset time.
[0031] Among them, it also includes:
[0032] The calculation module is further used to determine image feature data based on the cloud image to be measured at the first time point, wherein the image feature data is a matrix of three channels;
[0033] The calculation module is further used to determine the matrix of at least one channel based on the matrix size of the meteorological data to be measured and the image feature data at the first time point;
[0034] A processing module is used to merge the matrix of the at least one channel with the image feature data to obtain mixed feature data at the first time point.
[0035] Wherein, the calculation module is further used to traverse the multiple first time points in chronological order;
[0036] The calculation module is further used to determine the first time feature data and the second time feature data at the current first time point based on the mixed feature data at the current first time point, the first time feature data and the second time feature data at the previous first time point, the first time feature data used for calculation at the first first time point is the preset first time feature data, and the second time feature data is the preset second time feature data;
[0037] The calculation module is also used to obtain the first time feature data and the second time feature data of the last first time point after the traversal is completed.
[0038] Wherein, the calculation module is further used to traverse the preset predicted number of second time points in chronological order;
[0039] The calculation module is further used to determine the first time feature data and the second time feature data of the current second time point based on the mixed feature data of the current second time point, the first time feature data and the second time feature data of the previous second time point, the first time feature data used for calculation at the first first time point is the first time feature data of the last first time point, and the second time feature data is the second time feature data of the last first time point;
[0040] The deep learning module is further used to make a prediction based on the first time feature data and the second time feature data at the current second time point to obtain a prediction cloud map at the current second time point;
[0041] The calculation module is also used to obtain a prediction cloud map of a preset number of second time points after the traversal is completed. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0043] Figure 1 A flow chart of a cloud map prediction method according to an embodiment of the present application is shown;
[0044] Figure 2 A flowchart of a cloud map prediction method according to another embodiment of the present application is shown;
[0045] Figure 3 A flowchart of a cloud map prediction method according to another embodiment of the present application is shown;
[0046] Figure 4A flowchart of a cloud map prediction method according to another embodiment of the present application is shown;
[0047] Figure 5 A flowchart of a cloud map prediction method according to another embodiment of the present application is shown;
[0048] Figure 6 A flowchart of a cloud map prediction method according to another embodiment of the present application is shown;
[0049] Figure 7 A schematic structural diagram of a cloud map prediction device according to an embodiment of the present application is shown. DETAILED DESCRIPTION
[0050] In order to make the purpose, features, and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.
[0051] In order to improve the accuracy of predicting sky panoramic images, an embodiment of the present application provides a cloud image prediction method, such as Figure 1 As shown, the method includes:
[0052] Step 101 , collecting cloud images to be measured and meteorological data to be measured at multiple first time points, wherein the multiple first time points are different time points separated by a preset time.
[0053] The fisheye camera is used to collect cloud images to be measured at multiple time points, and the meteorological sensor is used to collect corresponding meteorological data to be measured at each first time point. The multiple first time points are different time points with preset time intervals.
[0054] For example, set the cloud map and corresponding meteorological data to be collected once every hour, for a total of three collections. The first collection is at 5 a.m., and the cloud map and meteorological data for 5 a.m. are collected. The second collection is at 6 a.m., and the cloud map and meteorological data for 6 a.m. are collected. The third collection is at 7 a.m., and the cloud map and meteorological data for 7 a.m. are collected.
[0055] Step 102: determining mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point.
[0056] In this embodiment, there are three ways to obtain mixed feature data:
[0057] The first method is to increase the number of channels. Since the cloud image to be tested is a matrix of three channels, the number of channels of the matrix is increased, and the meteorological data to be tested is placed in the matrix of the increased channels, and then merged with the three-channel matrix of the cloud image to be tested to obtain mixed feature data.
[0058] For example, after the cloud image to be tested is converted into feature data, The matrix of three channels, the meteorological data to be tested contains data of four dimensions: temperature, humidity, wind speed and wind direction. Then the meteorological data to be tested is converted into a matrix: The matrix converted from the meteorological data to be measured is used as an additional channel to be merged with the characteristic data converted from the cloud image to be measured, and the mixed characteristic data is obtained as follows:
[0059] The second type is splicing. The cloud image to be tested is a matrix of three channels. The meteorological data to be tested is divided into three parts. The meteorological data to be tested in each part is converted into a vector, and then spliced with the matrices of the three channels respectively to obtain mixed feature data.
[0060] For example, after the cloud image to be tested is converted into feature data, The matrix of three channels, the meteorological data to be tested contains data of two dimensions: temperature and humidity. Then the temperature and humidity are converted into vectors as [5,3]. The vector converted from the meteorological data to be tested is concatenated with the rows or columns of the feature data converted from the cloud image to be tested, and the mixed feature data is obtained as or
[0061] The third type is weighted merging, which performs weighted merging of the cloud image to be measured and the meteorological data to be measured based on preset weights to obtain mixed feature data.
[0062] For example, after the cloud image to be tested is converted into feature data, The matrix of three channels contains the meteorological data of four dimensions: temperature, humidity, wind speed and wind direction. The preset weights are The meteorological data to be measured is converted into a matrix: Based on the preset weights, the matrix converted from the meteorological data to be tested and the characteristic data converted from the cloud image to be tested are weighted and merged. The calculation method of each element in the matrix is to multiply the weight by the corresponding data in the meteorological data matrix to be tested, add 1 minus the weight and then multiply it by the corresponding data in the cloud image matrix to be tested. Finally, the mixed characteristic data is obtained as
[0063] Step 103, making predictions based on the mixed feature data of the multiple first time points to obtain a prediction cloud map of a preset number of second time points, wherein the multiple second time points are different time points separated by the preset time, and the earliest second time point is later than the latest first time point and separated by the preset time.
[0064] Based on the mixed feature data of multiple first time points, prediction is performed to obtain at least one prediction cloud map at a second time point. In this embodiment, the second time point and number corresponding to the obtained prediction cloud map can be controlled by setting a preset time and a preset prediction number. The preset time can be set to any time, and the preset prediction number can be set to any positive integer.
[0065] For example, the preset time is set to 1 hour, the preset prediction number is 3, and the last first time point corresponding to the mixed feature data used for prediction is 5 o'clock in the morning. After the prediction, the prediction cloud map for 6 o'clock in the morning, the prediction cloud map for 7 o'clock in the morning and the prediction cloud map for 8 o'clock in the morning are obtained.
[0066] In the above scheme, the cloud image to be measured and the meteorological data to be measured are combined in a variety of ways to obtain mixed feature data, so that the mixed feature data simultaneously contains the image features of the cloud image and the features of the corresponding meteorological data. Prediction based on the mixed feature data of multiple first time points can enable the prediction method to fully consider the features of the image and the features of the meteorological data, thereby significantly improving the accuracy of the predicted cloud image at at least one second time point.
[0067] In an example of the present application, a cloud map prediction method is also provided, such as Figure 2 As shown, the method includes:
[0068] Step 201 : determining image feature data based on the cloud image to be measured at the first time point, wherein the image feature data is a matrix of three channels.
[0069] For example, after the cloud image to be tested is converted into image feature data, A matrix of three channels.
[0070] Step 202: determining a matrix of at least one channel based on the matrix sizes of the meteorological data to be measured and the image feature data at the first time point.
[0071] For example, each channel in the image feature data is a 16*16 matrix, and the meteorological data to be tested contains data in four dimensions: temperature, humidity, wind speed, and wind direction. Then the meteorological data to be tested is converted into a matrix: Then create a 16*16 matrix, divide the newly added 16*16 matrix into 64 4*4 matrices, and repeatedly fill the 4*4 matrix converted from the meteorological data to be measured into the 64 4*4 matrices. Finally, merge the newly added 16*16 matrix with the image feature data to obtain a 16*16 matrix with 4 channels.
[0072] For another example, each channel in the image feature data is a 16*16 matrix, and the meteorological data to be tested contains data of five dimensions: temperature, humidity, wind speed, wind direction, and precipitation. Then the meteorological data to be tested can be converted into two matrices, one for If the other matrix has only one dimension of data left, the other positions can be filled with data of that dimension repeatedly or with zero values. Then the other can be or Create two 16*16 matrices, and divide the two newly added 16*16 matrices into 64 4*4 matrices. Repeat the first 4*4 matrix converted from the meteorological data to be tested into the 64 4*4 matrices divided from the first 16*16 matrix. Repeat the second 4*4 matrix converted from the meteorological data to be tested into the 64 4*4 matrices divided from the second 16*16 matrix. Finally, merge the two newly added 16*16 matrices with the image feature data to obtain a 16*16 matrix with 5 channels.
[0073] Step 203: merge the matrix of the at least one channel with the image feature data to obtain mixed feature data at the first time point.
[0074] In the above scheme, the features of the meteorological data to be measured are combined with the features of the cloud map by adding new channels to obtain mixed feature data. In the prediction process, since the meteorological data to be measured is flattened in the matrix of the newly added channels, the convolution kernel can obtain the features of the meteorological data to be measured each time it is convolved, thereby further increasing the accuracy of the predicted cloud map obtained after the prediction.
[0075] In an example of the present application, a cloud map prediction method is also provided, such as Figure 3 As shown, the method includes:
[0076] Step 301, traverse the multiple first time points in chronological order.
[0077] For example, if cloud images and corresponding meteorological data are collected at 5 a.m., 6 a.m., and 7 a.m., the three first time points are traversed in the order of 5 a.m., 6 a.m., and 7 a.m.
[0078] Step 302, determine the first time feature data and the second time feature data of the current first time point based on the mixed feature data of the current first time point, the first time feature data and the second time feature data of the previous first time point, the first time feature data used for calculation at the first first time point is the preset first time feature data, and the second time feature data is the preset second time feature data.
[0079] The second time feature data includes all previously input feature data, while the first time feature data includes currently input feature data.
[0080] The preset first time feature data and the preset second time feature data can be filled with random fixed numbers or zero values, or can be filled with variables, and these variables are updated during training.
[0081] The first time feature data and the second time feature data at the current first time point are determined based on the mixed feature data at the current first time point, the first time feature data and the second time feature data at the previous first time point.
[0082] For example, if the first first time point traversed is 5 a.m., the first time feature data and the second time feature data of the current first time point are determined based on the mixed feature data, the preset first time feature data and the preset second time feature data of 5 a.m. Then traverse to 6 a.m., the first time feature data and the second time feature data of the current first time point are determined based on the mixed feature data of 6 a.m. and the first time feature data and the second time feature data of the previous first time point (that is, 5 a.m.).
[0083] Step 303: After the traversal is completed, the first time feature data and the second time feature data of the last first time point are obtained.
[0084] In the above example, the last traversal is at 7 a.m., and based on the mixed feature data at 7 a.m., the first time feature data and the second time feature data of the previous first time point (that is, 6 a.m.), the first time feature data and the second time feature data of 7 a.m. are determined.
[0085] In an example of the present application, a cloud map prediction method is also provided, such as Figure 4 As shown, the method includes:
[0086] Step 401, traverse the preset predicted number of second time points in chronological order.
[0087] For example, the last first time point is 7 a.m., the preset time is 1 hour, and the preset number of predictions is 3, then the preset number of predicted second time points are 8 a.m., 9 a.m., and 10 a.m. Then the three second time points are traversed in the order of 8 a.m., 9 a.m., and 10 a.m.
[0088] Step 402, determines the first time feature data and the second time feature data of the current second time point based on the mixed feature data of the current second time point, the first time feature data and the second time feature data of the previous second time point, the first time feature data used for calculation at the first first time point is the first time feature data of the last first time point, and the second time feature data is the second time feature data of the last first time point.
[0089] When traversing the first second time point, the first time feature data and the second time feature data are the first time feature data and the second time feature data of the last first time point.
[0090] For example, if the first second time point traversed is 8 a.m., the first time feature data and the second time feature data of 8 a.m. are determined based on the mixed feature data of 8 a.m., the first time feature data and the second time feature data of 7 a.m. (the last first time point). Then traverse 9 a.m., and determine the first time feature data and the second time feature data of 9 a.m. based on the mixed feature data of 9 a.m., the first time feature data and the second time feature data of 8 a.m.
[0091] Step 403, making a prediction based on the first time feature data and the second time feature data at the current second time point to obtain a prediction cloud map at the current second time point.
[0092] Based on the first time feature data and the second time feature data at the current second time point, a prediction cloud map at the current second time point is obtained. The prediction can be made through the ConvLSTM model (convolutional long short-term memory neural network), which can more accurately capture the local features and spatiotemporal dependencies in the image data.
[0093] Step 404, after the traversal is completed, a prediction cloud graph of a preset prediction number of second time points is obtained.
[0094] In the above method, by traversing the second time points, the first time feature data and the second time feature data of each second time point are determined. The second time feature data includes all previously input feature data, while the first time feature data includes the currently input feature data. The prediction cloud map of the second time point is predicted based on the first time feature data and the second time feature data of the second time point, so that the accuracy of the obtained prediction cloud map is further improved.
[0095] In an example of the present application, a cloud map prediction method is also provided, such as Figure 5 As shown, the method of determining the second time feature data at the current time point based on the mixed feature data at the current time point, the first time feature data at the previous time point, and the second time feature data includes:
[0096] Step 501: determine first weighted feature data and second weighted feature data at the current time point based on mixed feature data at the current time point, first time feature data and second time feature data at the previous time point.
[0097] The first weighted feature data i at the current time point is determined based on the following formula t :
[0098]
[0099] Among them, X t is the mixed feature data at the current time point, H t-1 is the first time feature data of the previous time point, C t-1 is the second time feature data of the previous time point, W xi is a mixed feature data weight parameter for determining the first weighted feature data, W hi is a first time feature data weight parameter for determining the first weighted feature data, W ci is a weight parameter of the second time feature data used to determine the first weighted feature data, b i is a weight adjustment parameter for determining the first weighted feature data. * is a convolution operation, ° is a Hadamard product operation, and σ is normalization. xi , W hi , W ci and b i These are parameters in the model and can be optimized as the model is trained.
[0100] The second weighted feature data f at the current time point is determined based on the following formula: t :
[0101]
[0102] Among them, W xf is a mixed feature data weight parameter for determining the second weighted feature data, W hf is the weight parameter of the first time feature data used to determine the second weighted feature data, W cf is a second time feature data weight parameter for determining the second weighted feature data, b f is a weight adjustment parameter used to determine the second weighted feature data. xf , W hf , W cf and b f These are parameters in the model and can be optimized as the model is trained.
[0103] Step 502: Determine the second time feature data at the current time point based on the mixed feature data at the current time point, the first weighted feature data, the second weighted feature data, and the first time feature data at the previous time point.
[0104] The second time characteristic data C at the current time point is determined based on the following formula t :
[0105]
[0106] Among them, f t is the second weighted feature data at the current time point, i t is the first weighted feature data at the current time point, W xc is a mixed feature data weight parameter for determining the second time feature data, W hc is a weight parameter of the first time characteristic data used to determine the second time characteristic data, b c is a weight adjustment parameter used to determine the second time feature data. xc , W hc and b c These are parameters in the model and can be optimized as the model is trained.
[0107] In an example of the present application, a cloud map prediction method is also provided, such as Figure 6 As shown, the method of determining the first time feature data at the next time point based on the mixed feature data at the current time point, the first time feature data at the previous time point, and the second time feature data includes:
[0108] Step 601: determine third weighted feature data at the current time point based on the mixed feature data at the current time point, the second time feature data, and the first time feature data at the previous time point.
[0109] The third weighted feature data o at the current time point is determined based on the following formula t :
[0110]
[0111] Among them, C t is the second time feature data at the current time point, W xo is a mixed feature data weight parameter for determining the third weighted feature data, W ho is the weight parameter of the first time feature data used to determine the third weighted feature data, W co is a weight parameter of the second time feature data used to determine the third weighted feature data, b o is a weight adjustment parameter used to determine the third weighted feature data. xo , W ho, W co and b o These are parameters in the model and can be optimized as the model is trained.
[0112] Step 602: Determine the first time feature data at the current time point based on the third weighted feature data and the second time feature data at the current time point.
[0113] Determine the first time feature data H at the current time point based on the following formula t :
[0114]
[0115] Among them, t is the third weighted feature data at the current time point.
[0116] In order to realize the above cloud map prediction method, Figure 7 As shown, the present application provides a cloud image prediction device according to an example, including:
[0117] The acquisition module 10 is used to acquire the cloud images to be measured and the meteorological data to be measured at multiple first time points, where the multiple first time points are different time points at intervals of preset time;
[0118] A calculation module 20, configured to determine the mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point;
[0119] The deep learning module 30 is used to make predictions based on the mixed feature data of the multiple first time points to obtain a prediction cloud map of a preset number of second time points, wherein the multiple second time points are different time points separated by the preset time, and the earliest second time point is later than the latest first time point and separated by the preset time.
[0120] Among them, it also includes:
[0121] The calculation module 20 is further used to determine image feature data based on the cloud image to be measured at the first time point, wherein the image feature data is a matrix of three channels;
[0122] The calculation module 20 is further used to determine the matrix of at least one channel based on the matrix size of the meteorological data to be measured and the image feature data at the first time point;
[0123] The processing module 40 is used to merge the matrix of the at least one channel with the image feature data to obtain mixed feature data at the first time point.
[0124] The calculation module 20 is further configured to traverse the plurality of first time points in chronological order;
[0125] The calculation module 20 is further used to determine the first time feature data and the second time feature data at the current first time point based on the mixed feature data at the current first time point, the first time feature data and the second time feature data at the previous first time point, the first time feature data used for calculation at the first first time point is the preset first time feature data, and the second time feature data is the preset second time feature data;
[0126] The calculation module 20 is further used to obtain the first time feature data and the second time feature data of the last first time point after the traversal is completed.
[0127] The calculation module 20 is further configured to traverse the preset predicted number of second time points in chronological order;
[0128] The calculation module 20 is further used to determine the first time feature data and the second time feature data of the current second time point based on the mixed feature data of the current second time point, the first time feature data and the second time feature data of the previous second time point, the first time feature data used for calculation at the first first time point is the first time feature data of the last first time point, and the second time feature data is the second time feature data of the last first time point;
[0129] The deep learning module 30 is further used to make a prediction based on the first time feature data and the second time feature data at the current second time point to obtain a prediction cloud map at the current second time point;
[0130] The calculation module 20 is further used to obtain a prediction cloud map of a preset prediction number of second time points after the traversal is completed.
[0131] The calculation module 20 is further used to determine the first weighted feature data and the second weighted feature data at the current time point based on the mixed feature data at the current time point, the first time feature data and the second time feature data at the previous time point;
[0132] The calculation module 20 is further used to determine the second time feature data at the current time point based on the mixed feature data at the current time point, the first weighted feature data, the second weighted feature data and the first time feature data at the previous time point.
[0133] Wherein, the calculation module 20 is further used to determine the third weighted feature data at the current time point based on the mixed feature data at the current time point, the second time feature data and the first time feature data at the previous time point;
[0134] The calculation module 20 is further used to determine the first time feature data at the current time point based on the third weighted feature data and the second time feature data at the current time point.
[0135] In the above scheme, the cloud image to be measured and the meteorological data to be measured are combined in a variety of ways to obtain mixed feature data, so that the mixed feature data simultaneously contains the image features of the cloud image and the features of the corresponding meteorological data. Prediction based on the mixed feature data of multiple first time points can enable the prediction method to fully consider the features of the image and the features of the meteorological data, thereby significantly improving the accuracy of the predicted cloud image at at least one second time point.
[0136] In one example, an embodiment of the present application further provides a mobile terminal, the mobile terminal comprising at least one memory and a processor communicatively connected to the at least one memory; wherein the memory stores instructions executable by the at least one processor, the instructions being configured to execute the above Figures 1 to 6 The cloud map prediction method according to any one of the embodiments.
[0137] In addition, the present invention also provides a computer-readable storage medium storing computer-executable instructions for executing the above-mentioned Figures 1 to 6 The cloud map prediction method process described in any one of the embodiments.
[0138] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.
[0139] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the embodiments of the present application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0140] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.
[0141] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.
[0142] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0143] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.
[0144] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (Processor) to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.
[0145] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A cloud map prediction method, the method comprising: Collecting cloud images to be measured and meteorological data to be measured at multiple first time points, wherein the multiple first time points are different time points at intervals of preset time; Determine the mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point; Prediction is performed based on the mixed feature data of the multiple first time points to obtain a prediction cloud map of a preset number of second time points, wherein the multiple second time points are different time points separated by the preset time, and the earliest second time point is later than the latest first time point and separated by the preset time.
2. The cloud image prediction method according to claim 1, wherein the step of determining the mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point comprises: Determine image feature data based on the cloud image to be measured at the first time point, wherein the image feature data is a matrix of three channels; Determine a matrix of at least one channel based on the matrix sizes of the meteorological data to be measured and the image feature data at the first time point; The matrix of the at least one channel is combined with the image feature data to obtain mixed feature data at the first time point.
3. The cloud map prediction method according to claim 1, wherein the prediction based on the mixed feature matrix of the multiple time points comprises: Traversing the multiple first time points in chronological order; Determine the first time feature data and the second time feature data of the current first time point based on the mixed feature data of the current first time point, the first time feature data and the second time feature data of the previous first time point, the first time feature data used for calculation at the first first time point is the preset first time feature data, and the second time feature data is the preset second time feature data; After the traversal is completed, the first time feature data and the second time feature data of the last first time point are obtained.
4. The cloud map prediction method according to claim 3, after the first time feature data and the second time feature data at the first second time point, further comprising: Traversing the preset predicted number of second time points in chronological order; Determine the first time feature data and the second time feature data of the current second time point based on the mixed feature data of the current second time point, the first time feature data and the second time feature data of the previous second time point, the first time feature data used for calculation at the first first time point is the first time feature data of the last first time point, and the second time feature data is the second time feature data of the last first time point; Perform prediction based on the first time feature data and the second time feature data at the current second time point to obtain a prediction cloud map at the current second time point; After the traversal is completed, a prediction cloud graph of a preset prediction number of second time points is obtained.
5. The cloud map prediction method according to claim 3 or 4, wherein the determining the second time feature data at the current time point based on the mixed feature data at the current time point, the first time feature data at the previous time point, and the second time feature data comprises: Determine first weighted feature data and second weighted feature data at the current time point based on the mixed feature data at the current time point, the first time feature data and the second time feature data at the previous time point; The second time feature data at the current time point is determined based on the mixed feature data at the current time point, the first weighted feature data, the second weighted feature data and the first time feature data at the previous time point.
6. The cloud map prediction method according to claim 5, wherein the determining the first time feature data at the next time point based on the mixed feature data at the current time point, the first time feature data at the previous time point, and the second time feature data comprises: Determine third weighted feature data at the current time point based on the mixed feature data at the current time point, the second time feature data, and the first time feature data at the previous time point; The first time feature data at the current time point is determined based on the third weighted feature data and the second time feature data at the current time point.
7. A cloud map prediction device, comprising: A collection module, used for collecting cloud images to be measured and meteorological data to be measured at multiple first time points, wherein the multiple first time points are different time points with preset time intervals; A calculation module, used to determine the mixed characteristic data at the first time point based on the cloud image to be measured and the meteorological data to be measured at the first time point; A deep learning module is used to make predictions based on the mixed feature data of the multiple first time points to obtain a prediction cloud map of a preset number of second time points, wherein the multiple second time points are different time points separated by the preset time, and the earliest second time point is later than the latest first time point and separated by the preset time.
8. The cloud image prediction device according to claim 7, further comprising: The calculation module is further used to determine image feature data based on the cloud image to be measured at the first time point, wherein the image feature data is a matrix of three channels; The calculation module is further used to determine the matrix of at least one channel based on the matrix size of the meteorological data to be measured and the image feature data at the first time point; A processing module is used to merge the matrix of the at least one channel with the image feature data to obtain mixed feature data at the first time point.
9. The cloud image prediction device according to claim 7, comprising: The calculation module is further used to traverse the multiple first time points in chronological order; The calculation module is further used to determine the first time feature data and the second time feature data at the current first time point based on the mixed feature data at the current first time point, the first time feature data and the second time feature data at the previous first time point, the first time feature data used for calculation at the first first time point is the preset first time feature data, and the second time feature data is the preset second time feature data; The calculation module is also used to obtain the first time feature data and the second time feature data of the last first time point after the traversal is completed.
10. The cloud image prediction device according to claim 9, comprising: The calculation module is further used to traverse the preset predicted number of second time points in chronological order; The calculation module is further used to determine the first time feature data and the second time feature data of the current second time point based on the mixed feature data of the current second time point, the first time feature data and the second time feature data of the previous second time point, the first time feature data used for calculation at the first first time point is the first time feature data of the last first time point, and the second time feature data is the second time feature data of the last first time point; The deep learning module is further used to make a prediction based on the first time feature data and the second time feature data at the current second time point to obtain a prediction cloud map at the current second time point; The calculation module is also used to obtain a prediction cloud map of a preset number of second time points after the traversal is completed.