Cloud cover parameter prediction method and device based on multi-mode data

Through the cloud parameter prediction method of multi-modal data, a hierarchical scoring and fusion prediction model is constructed, which solves the problem of insufficient cloud forecast accuracy in numerical weather forecasts, and improves the accuracy and adaptability of forecasts.

CN119087548APending Publication Date: 2024-12-06CHINESE PEOPLES LIBERATION ARMY UNIT 61540
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411210966.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-06

Smart Images

  • Figure CN119087548A_ABST
    Figure CN119087548A_ABST
Patent Text Reader

Abstract

The invention discloses a cloud cover parameter prediction method and device based on multi-mode data. The method comprises the following steps: acquiring a cloud cover data set; the cloud cover data set comprises a multi-mode cloud cover forecast data set and a cloud cover observation data set; the multi-mode cloud cover forecasting data set comprises a plurality of cloud cover mode forecasting data sets; each piece of cloud amount prediction data has corresponding cloud amount observation data of the same place in the cloud amount observation data; preprocessing the cloud cover data set to obtain a preprocessed data set; the preprocessed data set comprises preprocessed multi-mode cloud cover forecast data and cloud cover observation data; and performing prediction processing on the preprocessed data set to obtain a cloud cover parameter prediction value. According to the invention, aiming at the characteristic that non-contact variables such as the cloud cover cannot be directly integrated, the clustering integration prediction is carried out according to different magnitudes of the variables, and the accuracy of cloud cover prediction is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the application field of numerical weather forecasting, and to a cloud amount parameter prediction method and device based on multi-mode data. Background Art

[0002] With the rapid development of the economy and society and the continuous improvement of people's living standards, the government and the public have higher and higher requirements for weather forecasts. With the increasing maturity of numerical weather forecasting technology, the forecasting ability of continuous variables of the situation field such as temperature, air pressure, humidity, wind, and potential height is relatively high. However, non-continuous variables such as precipitation and cloud cover cannot fully reflect the cloud-rainfall process due to their complex physical processes. The parameterization of cloud microphysical processes as a sub-grid parameterization process cannot fully reflect the cloud-rainfall process, and the forecasting level still needs to be improved. Weather processes such as clouds and rain have a great impact on human production and life, so it is of great significance to improve the forecasting level of non-continuous variables.

[0003] With the rapid development of numerical weather forecasting technology, different meteorological forecasting agencies have established their own numerical weather forecast simulation systems. Numerical models of different scales provide new opportunities for improving forecasting capabilities. Numerical forecast data of different models are increasing in variety, but the "singleness" and "uncertainty" of a single model in the forecasting process affect the accuracy of the model forecast. Therefore, each model has its own advantages and disadvantages in forecasting under different geographical conditions. How to use multi-model forecast data to improve the forecast accuracy of discontinuous variables such as cloud cover is an effective way to improve numerical weather forecast products and an important means to promote the use of numerical forecast products. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a cloud parameter prediction method based on multi-mode data. The method is suitable for integrated forecasting of discontinuous variables such as cloud amount. In view of the fact that non-connected variables such as cloud amount cannot be directly integrated, the method realizes clustered integrated forecasting according to different magnitudes of variables, thereby improving the accuracy of the forecast.

[0005] In order to solve the above technical problems, the first aspect of the embodiment of the present invention discloses a cloud amount parameter prediction method based on multi-mode data, including:

[0006] S1, obtaining a cloud amount data set; the cloud amount data set includes a multi-mode cloud amount forecast data set and a cloud amount observation data set; the multi-mode cloud amount forecast data set includes a plurality of cloud amount model forecast data sets; the cloud amount observation data set includes cloud amount observation data of a plurality of locations; the cloud amount model forecast data set includes cloud amount forecast data of a plurality of locations obtained by using a specific forecast model; each cloud amount forecast data has corresponding cloud amount observation data of the same location in the cloud amount observation data;

[0007] S2, preprocessing the cloud amount data set to obtain a preprocessed data set; the preprocessed data set includes a preprocessed multi-mode cloud amount forecast data set and a cloud amount observation data set;

[0008] S3, performing prediction processing on the preprocessed data set to obtain a predicted value of a cloud amount parameter.

[0009] The preprocessing of the cloud amount data set to obtain a preprocessed data set includes:

[0010] S21, performing data cleaning processing on the cloud amount data set to obtain a first data set;

[0011] S22, performing a spatiotemporal information alignment process on the first data set to obtain a cleaned data set;

[0012] S23, performing boundary detection processing on the cleaned data set to obtain a preprocessed data set.

[0013] The performing boundary detection processing on the cleaned data set to obtain a preprocessed data set includes:

[0014] S231, for each data attribute of the cleaned data set, taking the data collection information of the data as an independent variable and the data value of the data as a dependent variable, a cluster analysis process is performed to obtain clustering result information of the data attribute of the class; the clustering result information includes the cluster category to which all data of the data attribute of the class belongs;

[0015] S232, determining the number of data included in each cluster category; setting a data volume threshold; deleting the data included in the cluster category whose data volume is less than the data volume threshold from the cleaned data set;

[0016] S233, executing S231 and S232 for all data of the cleaned data set to obtain a preprocessed data set.

[0017] The step of performing prediction processing on the preprocessed data set to obtain a predicted value of a cloud amount parameter includes:

[0018] S31, construct a cloud amount forecasting graded scoring model; the cloud amount forecasting graded scoring model includes a first threshold range, a second threshold range and a third threshold range; the first cloud amount forecast data value range corresponding to the first threshold range is [0, k1], and the first threshold range is [x1, x2]; the second cloud amount forecast data value range corresponding to the second threshold range is (k1, k2], and the second threshold range is [x3, x4]; the third cloud amount forecast data value range corresponding to the third threshold range is (k2, k3], and the third threshold range is [x5, x6];

[0019] S32, using the cloud amount prediction hierarchical scoring model, evaluating the preprocessed data set to obtain a data scoring set; the data scoring set includes a scoring value of each cloud amount model prediction data set under each cloud amount forecast data value range;

[0020] S33, performing score optimization processing on the data score set to obtain a weight vector corresponding to the value range of each cloud amount forecast data;

[0021] S34, constructing a fusion prediction model based on the weight vector corresponding to the value range of all cloud amount forecast data;

[0022] S35, using the fusion prediction model, processing the cloud cover prediction data of the same location in all cloud cover pattern prediction data sets to obtain the final cloud cover parameter prediction value of the location.

[0023] The cloud cover prediction hierarchical scoring model is used to evaluate the preprocessed data set to obtain a data scoring set, including:

[0024] S3201, obtaining a first threshold range, a second threshold range, and a third threshold range;

[0025] S3202, for each cloud amount prediction data of the preprocessed data set, according to the cloud amount forecast data value range to which the value belongs, determining a threshold range corresponding to the cloud amount forecast data value range as a threshold judgment range;

[0026] S3203, determining a threshold discrimination interval of the cloud amount prediction data according to the cloud amount observation data corresponding to the cloud amount prediction data and the threshold discrimination range;

[0027] S3204, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the first cloud amount forecast data, to obtain a first data set;

[0028] S3205, performing scoring processing on the first data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount pattern prediction data set within the value range of the first cloud amount forecast data;

[0029] S3206, executing S3204 to S3205 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the first cloud amount forecast data value range;

[0030] S3207, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the second cloud amount forecast data, to obtain a second data set;

[0031] S3208, performing scoring processing on the second data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount model prediction data set within the value range of the second cloud amount forecast data;

[0032] S3209, executing S3207 to S3208 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the second cloud amount forecast data value range;

[0033] S3210, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the third cloud amount forecast data, to obtain a third data set;

[0034] S3211, performing scoring processing on the third data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount pattern prediction data set within the value range of the third cloud amount forecast data;

[0035] S3212, executing S3210 to S3211 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the third cloud amount forecast data value range;

[0036] S3213, fusing the score values ​​of each cloud amount model prediction data set within each cloud amount forecast data value range to obtain a data score set.

[0037] The expression of the scoring process is:

[0038]

[0039] Among them, TS k represents the score value under the value range of the k-th cloud amount forecast data, A k is the total number of cloud forecast data in a cloud model prediction data set whose data values ​​are within the corresponding threshold judgment range, B k C is the total number of cloud forecast data in a cloud model prediction data set whose data value is less than the lower limit of the corresponding threshold judgment range. k The total number of cloud cover forecast data in a cloud cover model prediction data set whose data value is less than the upper limit of the corresponding threshold judgment range.

[0040] The step of performing score optimization processing on the data score set to obtain a weight vector corresponding to the value range of each cloud amount forecast data includes:

[0041] S331, initialize the weight vector corresponding to each cloud amount forecast data value range, and the variable of the weight vector corresponding to the k-th cloud amount forecast data value range is expressed as Wik , i=1,2,3,4, k=1,2,3;

[0042] S332, constructing an integrated equation based on the value relationship between each cloud amount prediction data and the corresponding threshold discrimination interval; the expression of the integrated equation is:

[0043]

[0044] Among them, TS is the total data score, N is the total number of cloud cover model prediction data sets, and TS 1i It represents the score value of the ith cloud amount model prediction data set in the first cloud amount forecast data value range, and so on. TS 3i represents the score value of the i-th cloud amount model prediction data set within the value range of the first cloud amount forecast data;

[0045] S333, with the goal of maximizing the integrated equation, is a constraint condition, for all W ik , i = 1, 2, 3, 4, k = 1, 2, 3, solve and obtain the weight optimization vector corresponding to each cloud amount forecast data value range. The i-th weight of the weight optimization vector corresponding to the k-th cloud amount forecast data value range is expressed as W′ ik , i=1,2,3,4, k=1,2,3.

[0046] The expression of the fusion prediction model is:

[0047]

[0048] Among them, Y′ j is the final cloud cover parameter prediction value at the jth location, Y ij is the cloud cover prediction data of the jth location in the i-th cloud cover pattern prediction data set, It is the mean value calculated for the cloud cover prediction data of the jth location in all cloud cover model prediction data sets.

[0049] In a second aspect of an embodiment of the present invention, a cloud amount parameter prediction device based on multi-mode data is proposed, the device comprising:

[0050] A memory storing executable program code;

[0051] a processor coupled to the memory;

[0052] The processor calls the executable program code stored in the memory to execute the cloud parameter prediction method based on multi-mode data.

[0053] According to a third aspect of an embodiment of the present invention, a computer storable medium is proposed, wherein the computer storable medium stores computer instructions, and when the computer instructions are called by a computer, they are used to execute the cloud parameter prediction method based on multi-mode data.

[0054] The beneficial effects of the present invention are:

[0055] The present invention proposes a data optimization integration of non-continuous variables in multiple modes. The method calculates the multi-mode integration weights by associating the value ranges of forecast elements under multiple modes, and obtains the optimization integration forecast product through weighted calculation of the multi-mode forecast products.

[0056] The present invention comprehensively considers the forecasting characteristics of non-continuous variables and the forecasting levels of various models at different magnitudes to perform optimal integration.

[0057] The present invention can timely adjust the weight index according to the magnitude change of the integrated variable, which is beneficial to further improve the integrated forecasting effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a flow chart for implementing the method of the present invention. DETAILED DESCRIPTION

[0059] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0060] Figure 1 It is a flow chart for implementing the method of the present invention.

[0061] In order to construct a more accurate variable integrated forecast product based on multi-mode forecasting of non-continuous variables, the present invention combines meteorological statistics and numerical models, clarifies the range of weights through the distribution scores of non-continuous variable elements over a period of time; obtains the optimal weights through exhaustive training, and obtains the optimal integrated forecast product at the next moment through weighted iteration of multi-mode forecast variables, thereby improving the accuracy of non-continuous variables.

[0062] In a first aspect of an embodiment of the present invention, a cloud amount parameter prediction method based on multi-mode data is disclosed, comprising:

[0063] S1, obtaining a cloud amount data set; the cloud amount data set includes a multi-mode cloud amount forecast data set and a cloud amount observation data set; the multi-mode cloud amount forecast data set includes a plurality of cloud amount model forecast data sets; the cloud amount observation data set includes cloud amount observation data of a plurality of locations; the cloud amount model forecast data set includes cloud amount forecast data of a plurality of locations obtained by using a specific forecast model; each cloud amount forecast data has corresponding cloud amount observation data of the same location in the cloud amount observation data;

[0064] S2, preprocessing the cloud amount data set to obtain a preprocessed data set; the preprocessed data set includes a preprocessed multi-mode cloud amount forecast data set and a cloud amount observation data set;

[0065] S3, performing prediction processing on the preprocessed data set to obtain a predicted value of a cloud amount parameter.

[0066] The preprocessing of the cloud amount data set to obtain a preprocessed data set includes:

[0067] S21, performing data cleaning processing on the cloud amount data set to obtain a first data set;

[0068] S22, performing a spatiotemporal information alignment process on the first data set to obtain a cleaned data set;

[0069] S23, performing boundary detection processing on the cleaned data set to obtain a preprocessed data set.

[0070] The performing boundary detection processing on the cleaned data set to obtain a preprocessed data set includes:

[0071] S231, for each data attribute of the cleaned data set, taking the data collection information of the data as an independent variable and the data value of the data as a dependent variable, a cluster analysis process is performed to obtain clustering result information of the data attribute of the class; the clustering result information includes the cluster category to which all data of the data attribute of the class belongs;

[0072] S232, determining the number of data included in each cluster category; setting a data volume threshold; deleting the data included in the cluster category whose data volume is less than the data volume threshold from the cleaned data set;

[0073] S233, executing S221 and S222 for all data of the cleaned data set to obtain a preprocessed data set.

[0074] The performing the spatiotemporal acquisition information alignment processing on the first data set to obtain a cleaned data set includes:

[0075] S231, setting a unified data collection time and a unified data collection location for the observation data with the same data attribute included in the first data set;

[0076] S232, for observation data with the same data attribute, aligning the collection time and collection location of the observation data according to the unified data collection time and the unified data collection location to obtain a cleaned data set;

[0077] The aligning of the collection time and the collection location of the observation data comprises:

[0078] When the time interval for collecting the observation data is greater than the time interval for the unified data collection time, interpolation processing is performed on the adjacent observation data to obtain the observation data value at the unified data collection time, and the value is used as the cleaned data set;

[0079] When the time interval for collecting the observation data is less than the time interval for the unified data collection time, sampling the observation data to obtain observation data consistent with the unified data collection time, and use the observation data as a cleaned data set;

[0080] When the spatial interval of the observation data collection is greater than the spatial interval of the unified data collection location, interpolation processing is performed on the adjacent observation data to obtain the observation data value at the unified data collection location, which is used as the cleaned data set;

[0081] When the spatial interval of the observation data collection is smaller than the spatial interval of the unified data collection location, the observation data is sampled to obtain observation data consistent with the unified data collection location, which is used as a cleaned data set.

[0082] S233, executing S232 on the observed data of each data attribute to obtain a standard data set;

[0083] The data cleaning process includes filling missing values, smoothing noise data, and smoothing or deleting outliers; the smoothed noise data is obtained by first determining the noise data, and then smoothing the noise data according to the data before and after the noise data; the noise data is a value whose value is less than the detection sensitivity of the sensor of the observed data, or greater than the measurement upper limit of the sensor of the observed data. The outlier point can be determined by the Kalman filter method. The filling value of the missing value can be determined by averaging the measured values ​​within a certain sampling interval before and after the missing value.

[0084] The step of performing prediction processing on the preprocessed data set to obtain a predicted value of a cloud amount parameter includes:

[0085] S31, construct a cloud amount forecasting graded scoring model; the cloud amount forecasting graded scoring model includes a first threshold range, a second threshold range and a third threshold range; the first cloud amount forecast data value range corresponding to the first threshold range is [0, k1], and the first threshold range is [x1, x2]; the second cloud amount forecast data value range corresponding to the second threshold range is (k1, k2], and the second threshold range is [x3, x4]; the third cloud amount forecast data value range corresponding to the third threshold range is (k2, k3], and the third threshold range is [x5, x6];

[0086] S32, using the cloud amount prediction hierarchical scoring model, evaluating the preprocessed data set to obtain a data scoring set; the data scoring set includes a scoring value of each cloud amount model prediction data set under each cloud amount forecast data value range;

[0087] S33, performing score optimization processing on the data score set to obtain a weight vector corresponding to the value range of each cloud amount forecast data;

[0088] S34, constructing a fusion prediction model based on the weight vector corresponding to the value range of all cloud amount forecast data;

[0089] S35, using the fusion prediction model, processing the cloud cover prediction data of the same location in all cloud cover pattern prediction data sets to obtain the final cloud cover parameter prediction value of the location.

[0090] The cloud cover prediction hierarchical scoring model is used to evaluate the preprocessed data set to obtain a data scoring set, including:

[0091] S3201, obtaining a first threshold range, a second threshold range, and a third threshold range;

[0092] S3202, for each cloud amount prediction data of the preprocessed data set, according to the cloud amount forecast data value range to which the value belongs, determining a threshold range corresponding to the cloud amount forecast data value range as a threshold judgment range;

[0093] S3203, determining a threshold discrimination interval of the cloud amount prediction data according to the cloud amount observation data corresponding to the cloud amount prediction data and the threshold discrimination range;

[0094] S3204, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the first cloud amount forecast data, to obtain a first data set;

[0095] S3205, performing scoring processing on the first data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount pattern prediction data set within the value range of the first cloud amount forecast data;

[0096] S3206, executing S3204 to S3205 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the first cloud amount forecast data value range;

[0097] S3207, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the second cloud amount forecast data, to obtain a second data set;

[0098] S3208, performing scoring processing on the second data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount model prediction data set within the value range of the second cloud amount forecast data;

[0099] S3209, executing S3207 to S3208 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the second cloud amount forecast data value range;

[0100] S3210, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the third cloud amount forecast data, to obtain a third data set;

[0101] S3211, performing scoring processing on the third data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount pattern prediction data set within the value range of the third cloud amount forecast data;

[0102] S3212, executing S3210 to S3211 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the third cloud amount forecast data value range;

[0103] S3213, fusing the score values ​​of each cloud amount model prediction data set within each cloud amount forecast data value range to obtain a data score set.

[0104] The expression of the scoring process is:

[0105]

[0106] Among them, TS k represents the score value under the value range of the k-th cloud amount forecast data, A k is the total number of cloud forecast data in a cloud model prediction data set whose data values ​​are within the corresponding threshold judgment range, B kC is the total number of cloud forecast data in a cloud model prediction data set whose data value is less than the lower limit of the corresponding threshold judgment range. k The total number of cloud cover forecast data in a cloud cover model prediction data set whose data value is less than the upper limit of the corresponding threshold judgment range.

[0107] The step of performing score optimization processing on the data score set to obtain a weight vector corresponding to the value range of each cloud amount forecast data includes:

[0108] S331, initialize the weight vector corresponding to each cloud amount forecast data value range, and the variable of the weight vector corresponding to the k-th cloud amount forecast data value range is expressed as W ik , i=1,2,3,4, k=1,2,3;

[0109] S332, constructing an integrated equation based on the value relationship between each cloud amount prediction data and the corresponding threshold discrimination interval; the expression of the integrated equation is:

[0110]

[0111] Among them, TS is the total data score, N is the total number of cloud pattern prediction data sets, Y is the cloud amount prediction data in the cloud amount pattern prediction data set, and TS 1i It represents the score value of the ith cloud amount model prediction data set in the first cloud amount forecast data value range, and so on. TS 3i represents the score value of the i-th cloud amount model prediction data set within the value range of the first cloud amount forecast data;

[0112] S333, with the goal of maximizing the integrated equation, is a constraint condition, for all W ik , i = 1, 2, 3, 4, k = 1, 2, 3, solve and obtain the weight optimization vector corresponding to each cloud amount forecast data value range. The i-th weight of the weight optimization vector corresponding to the k-th cloud amount forecast data value range is expressed as W i ' k , i=1,2,3,4, k=1,2,3.

[0113] The expression of the fusion prediction model is:

[0114]

[0115] Among them, Y j ′ is the final predicted value of cloud cover parameters at the jth location, Y ij is the cloud cover prediction data of the jth location in the i-th cloud cover pattern prediction data set, It is the mean value calculated for the cloud cover prediction data of the jth location in all cloud cover model prediction data sets.

[0116] The multi-model cloud forecast data set includes GRAPE model cloud forecast data, ECMWF model cloud forecast data, GFS model cloud forecast data, and WRF model cloud forecast data.

[0117] Each data of the cloud amount pattern prediction data set corresponds to a specific location; the cloud amount observation data of the cloud amount observation data set corresponds to a specific location;

[0118] The data collection information includes time information and space information corresponding to the data collection; when performing data fitting, any of the above information can be used as an independent variable, and the corresponding data value can be used as a dependent variable.

[0119] For each cloud amount prediction data of the preprocessed data set, according to the cloud amount forecast data value range to which its value belongs, the threshold range corresponding to the cloud amount forecast data value range is determined as the threshold judgment range, which can be: if the value of a cloud amount prediction data satisfies (k1, k2], it belongs to the second cloud amount forecast data value range, and the corresponding threshold range is the second threshold range [x3, x4], then the threshold judgment range is [x3, x4].

[0120] The step of determining the threshold discrimination interval according to the cloud amount observation data corresponding to the cloud amount prediction data and the threshold discrimination range includes:

[0121] For the cloud amount observation data corresponding to the cloud amount prediction data, the value is p, the corresponding threshold discrimination range is [xa, xb], and the determined threshold discrimination interval is [p-xa, p+xb]. Furthermore, if the threshold discrimination range is [x3, x4], the determined threshold discrimination interval is [p-x3, p+x4].

[0122] When evaluating the first, second or third data set of a certain type of cloud forecast data set, the threshold judgment interval of each data of the first, second or third data set is used to judge whether the data belongs to the threshold judgment interval or is less than the upper limit value or less than the upper limit value, and the total number of data that meets each type is counted. The threshold judgment interval of each data is different and needs to be judged one by one. In this way, the accuracy of data scoring is ensured.

[0123] The objective is to maximize the integrated equation. is a constraint condition, for all W ik , i = 1, 2, 3, 4, k = 1, 2, 3, and solve to obtain the weight vector corresponding to each cloud cover forecast data value range, which can be solved by linear programming or genetic algorithm.

[0124] Specifically, in a scoring process, A k B k C k All are obtained for the same cloud cover pattern prediction data set.

[0125] In a second aspect of an embodiment of the present invention, a cloud amount parameter prediction method based on multi-mode data is proposed, comprising:

[0126] S1, obtain multi-model forecast data set and measured data set;

[0127] S2, processing the multi-model forecast data set and the measured data set to obtain a first distribution score set;

[0128] S3, processing the distribution score sets of different magnitudes of the modes to obtain an integrated prediction value.

[0129] The acquisition and collection of forecast data and true value data in multiple modes are as follows:

[0130] S11, obtain forecast data from various models, such as GRAPE forecast data, ECMWF forecast data, GFS forecast data, and WRF forecast data;

[0131] S12, collects true value data, which is site observation data.

[0132] The preprocessing of data to prepare a training data set includes:

[0133] S21, performing data cleaning processing on the GRAPE forecast data, ECMWF forecast data, GFS forecast data, WRF forecast data and observation data, including distinguishing the time range, spatial range and data attributes to obtain a cleaned data set;

[0134] S22, performing bilinear interpolation on the cleaned data set through each mode non-continuous variable, taking cloud cover as an example, to obtain a standard data set consistent with the time and space of the observation site;

[0135] S23, performing boundary check and category consistency check processing on the standard data set to obtain a consistent data set;

[0136] S24, based on each data attribute, the cloud amount observation data and the cloud amount forecast data of each model with the same time attribute in the consistency data set are combined to obtain a corresponding training data set.

[0137] The magnitude threshold of non-continuous variables is set, and the prediction accuracy of each model training data is evaluated using distribution scores to obtain the distribution scores of each model at different magnitudes, including:

[0138] S31, according to the characteristics of different cloud amounts, different magnitude thresholds are designed, 3, 7, and 10;

[0139] S32, performing distribution scoring on each model forecast cloud amount data and cloud amount observation data of the training data set according to the distribution scoring formula, and obtaining TS values ​​of different magnitudes of different model forecast results within the training data time range; obtaining distribution scores of each model at different magnitudes.

[0140]

[0141] Among them, k is the level of the non-continuous variable, A k The number of stations where the variable forecast result and the true value have the same threshold, B k The number of stations whose variable forecast results do not reach the lower limit of the true value threshold, C k The number of stations where the variable forecast results do not exceed the upper limit of the true value threshold.

[0142] The three levels of scoring, 0≤cloud amount<3, 3≤cloud amount<7, and 7≤cloud amount≤100, were obtained respectively, and the distribution score results of GRAPE, ECMWF, GFS, and WRF models at each level were obtained.

[0143] According to the distribution score, the weights of different magnitudes are obtained through exhaustive training method to make the distribution score of the magnitude reach the highest, including:

[0144] S41, judging the range of weights according to the TS values ​​of different modes at the same level k in claim 3; if the TS value of one mode is greater than that of another mode, then the weight value of the mode at level k is greater than that of the other mode; for example, at the level [0,3], the distribution score of the ECMWF mode is greater than that of the GRAPE mode, then at this level the ECMWF mode is higher than the GRAPE mode,

[0145] S42, by exhaustively enumerating the changes in weight values, according to the following weight integration formula, training is performed to obtain the highest cloud amount distribution score at level k after integration, and the weights of each mode at this time are the optimal weights at level k;

[0146] Based on the acquired weights, integrated equations of discontinuous variables of different magnitudes are established, including:

[0147] S51, establishing an integrated equation according to the weight values ​​obtained at different levels in claim 4, the formula is as follows:

[0148]

[0149] Among them, Y j is the cloud cover forecast value of the jth integration, Y j-1 is the cloud cover forecast value of the j-1th integration; Xi is the variable forecast value of the ith model, which can be GRAPE, ECMWF, GFS, WRF and other 4 models; the cloud cover level is divided into 3 levels, P1 is the threshold of the first variable level 3, P2 is the threshold of the first variable level 7, P3 is the threshold of the third variable level 10; N is the number of models participating in the integrated forecast, which is 4; W ik is the optimal weight of the i-th mode at level k, and

[0150] The initial variable values ​​of the integrated equation adopt the model values ​​with higher distribution scores, and carry out the integrated forecast of non-continuous variables predicted by each model in the future, including:

[0151] S61, based on the model forecast value with the highest comprehensive distribution score in claim 3 as Y0, carry out integrated forecasting to obtain Y1. If the sum of all magnitude distribution scores of the ECMWF model in claim 3 is the largest, then Y0 takes the forecast result of the ECMWF model and brings it into the integrated formula for calculation;

[0152] S62, according to the region where Y1 is located, the region is judged and brought into the integrated equation for the second iteration forecast; if Y0 is in the [0,3] level, and the integrated Y1 is in the [4,7] level, then the [4,7] level weight is selected and brought into the integrated formula to calculate Y2;

[0153] S63, discriminate according to the region where Y2 is located, bring it into the integrated equation for the third iterative forecast, and obtain Y3; generally three iterations are appropriate, which is used as the integrated forecast value of the variable Y3.

[0154] In a third aspect of an embodiment of the present invention, a cloud amount parameter prediction device based on multi-mode data is proposed, the device comprising:

[0155] A memory storing executable program code;

[0156] a processor coupled to the memory;

[0157] The processor calls the executable program code stored in the memory to execute the cloud parameter prediction method based on multi-mode data.

[0158] In a fourth aspect of an embodiment of the present invention, a computer storable medium is proposed, wherein the computer storable medium stores computer instructions, and when the computer instructions are called by a computer, they are used to execute the cloud parameter prediction method based on multi-mode data.

[0159] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A cloud parameter prediction method based on multi-mode data, characterized in that: include: S1, obtaining a cloud amount data set; the cloud amount data set includes a multi-mode cloud amount forecast data set and a cloud amount observation data set; the multi-mode cloud amount forecast data set includes a plurality of cloud amount model forecast data sets; the cloud amount observation data set includes cloud amount observation data of a plurality of locations; the cloud amount model forecast data set includes cloud amount forecast data of a plurality of locations obtained by using a specific forecast model; each cloud amount forecast data has corresponding cloud amount observation data of the same location in the cloud amount observation data; S2, preprocessing the cloud cover data set to obtain a preprocessed data set; The preprocessed data set includes a preprocessed multi-mode cloud cover forecast data set and a cloud cover observation data set; S3, performing prediction processing on the preprocessed data set to obtain a predicted value of a cloud amount parameter.

2. The cloud cover parameter prediction method based on multi-mode data according to claim 1, characterized in that: The preprocessing of the cloud amount data set to obtain a preprocessed data set includes: S21, performing data cleaning processing on the cloud amount data set to obtain a first data set; S22, performing a spatiotemporal information alignment process on the first data set to obtain a cleaned data set; S23, performing boundary detection processing on the cleaned data set to obtain a preprocessed data set.

3. The cloud cover parameter prediction method based on multi-mode data according to claim 2, characterized in that: The performing boundary detection processing on the cleaned data set to obtain a preprocessed data set includes: S231, for each data attribute of the cleaned data set, taking the data collection information of the data as the independent variable and the data value of the data as the dependent variable, performing cluster analysis processing to obtain clustering result information of the data attribute of the class; the clustering result information includes the clustering category to which all data of the data attribute of the class belongs; S232, determining the number of data included in each cluster category; setting a data volume threshold; deleting the data included in the cluster category whose data number is less than the data volume threshold from the cleaned data set; S233, executing S231 and S232 for all data of the cleaned data set to obtain a preprocessed data set.

4. The cloud cover parameter prediction method based on multi-mode data according to claim 1, characterized in that: The step of performing prediction processing on the preprocessed data set to obtain a predicted value of a cloud amount parameter includes: S31, construct a cloud amount forecasting graded scoring model; the cloud amount forecasting graded scoring model includes a first threshold range, a second threshold range and a third threshold range; the first cloud amount forecast data value range corresponding to the first threshold range is [0, k1], and the first threshold range is [x1, x2]; the second cloud amount forecast data value range corresponding to the second threshold range is (k1, k2], and the second threshold range is [x3, x4]; the third cloud amount forecast data value range corresponding to the third threshold range is (k2, k3], and the third threshold range is [x5, x6]; S32, using the cloud amount prediction hierarchical scoring model, evaluating the preprocessed data set to obtain a data scoring set; the data scoring set includes a scoring value of each cloud amount model prediction data set under each cloud amount forecast data value range; S33, performing score optimization processing on the data score set to obtain a weight vector corresponding to the value range of each cloud amount forecast data; S34, constructing a fusion prediction model based on the weight vector corresponding to the value range of all cloud amount forecast data; S35, using the fusion prediction model, processing the cloud cover prediction data of the same location in all cloud cover pattern prediction data sets to obtain the final cloud cover parameter prediction value of the location.

5. The cloud cover parameter prediction method based on multi-mode data according to claim 4, characterized in that: The cloud cover prediction hierarchical scoring model is used to evaluate the preprocessed data set to obtain a data scoring set, including: S3201, obtaining a first threshold range, a second threshold range, and a third threshold range; S3202, for each cloud amount prediction data of the preprocessed data set, according to the cloud amount forecast data value range to which the value belongs, determining a threshold range corresponding to the cloud amount forecast data value range as a threshold judgment range; S3203, determining a threshold discrimination interval of the cloud amount prediction data according to the cloud amount observation data corresponding to the cloud amount prediction data and the threshold discrimination range; S3204, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the first cloud amount forecast data, to obtain a first data set; S3205, performing scoring processing on the first data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount pattern prediction data set within the value range of the first cloud amount forecast data; S3206, executing S3204 to S3205 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the first cloud amount forecast data value range; S3207, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the second cloud amount forecast data, to obtain a second data set; S3208, performing scoring processing on the second data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount model prediction data set within the value range of the second cloud amount forecast data; S3209, executing S3207 to S3208 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the second cloud amount forecast data value range; S3210, for each cloud amount model prediction data set, obtaining all cloud amount prediction data whose values ​​belong to the value range of the third cloud amount forecast data, to obtain a third data set; S3211, performing scoring processing on the third data set according to the threshold discrimination interval of each data thereof, to obtain a score value of the cloud amount pattern prediction data set within the value range of the third cloud amount forecast data; S3212, executing S3210 to S3211 for each cloud amount pattern prediction data set, to obtain the score values ​​of all cloud amount pattern prediction data sets within the third cloud amount forecast data value range; S3213, fusing the score values ​​of each cloud amount model prediction data set within each cloud amount forecast data value range to obtain a data score set.

6. The cloud cover parameter prediction method based on multi-mode data according to claim 5, characterized in that: The expression of the scoring process is: Among them, TS k represents the score value under the value range of the k-th cloud amount forecast data, A k is the total number of cloud forecast data in a cloud model prediction data set whose data values ​​are within the corresponding threshold judgment range, B k C is the total number of cloud forecast data in a cloud model prediction data set whose data value is less than the lower limit of the corresponding threshold judgment range. k The total number of cloud cover forecast data in a cloud cover model prediction data set whose data value is less than the upper limit of the corresponding threshold judgment range.

7. The cloud cover parameter prediction method based on multi-mode data according to claim 5, characterized in that: The step of performing score optimization processing on the data score set to obtain a weight vector corresponding to the value range of each cloud amount forecast data includes: S331, initialize the weight vector corresponding to each cloud amount forecast data value range, and the variable of the weight vector corresponding to the k-th cloud amount forecast data value range is expressed as W ik , i=1,2,3,4, k=1,2,3; S332, constructing an integrated equation based on the value relationship between each cloud amount prediction data and the corresponding threshold discrimination interval; the expression of the integrated equation is: Among them, TS is the total data score, N is the total number of cloud cover model prediction data sets, and TS 1i It represents the score value of the ith cloud amount model prediction data set in the first cloud amount forecast data value range, and so on. TS 3i represents the score value of the i-th cloud amount model prediction data set within the value range of the first cloud amount forecast data; S333, with the goal of maximizing the integrated equation, is a constraint condition, for all W ik , i = 1, 2, 3, 4, k = 1, 2, 3, solve and obtain the weight optimization vector corresponding to each cloud amount forecast data value range. The i-th weight of the weight optimization vector corresponding to the k-th cloud amount forecast data value range is expressed as W i ' k , i=1,2,3,4, k=1,2,3.

8. The cloud parameter prediction method based on multi-mode data according to claim 7, characterized in that: The expression of the fusion prediction model is: Among them, Y j ′ is the final predicted value of cloud cover parameters at the jth location, Y ij is the cloud cover prediction data of the jth location in the i-th cloud cover pattern prediction data set, It is the mean value calculated for the cloud cover prediction data of the jth location in all cloud cover model prediction data sets.

9. A cloud parameter prediction device based on multi-mode data, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the cloud parameter prediction method based on multi-mode data as described in any one of claims 1 to 8.

10. A computer storable medium, characterized in that: The computer storable medium stores computer instructions, and when the computer instructions are called by a computer, they are used to execute the cloud parameter prediction method based on multi-mode data as described in any one of claims 1 to 8.

Citation Information

Cited By

  • Intelligent prediction method and device for cloud and rain meteorological parameters

    CN120875146A