Severe convective weather classification forecasting method, system, equipment and medium

Abstract

The invention relates to the technical field of weather classification forecasting, and discloses a severe convection weather classification forecasting method, system and device and a medium, and the method comprises the steps: obtaining historical data of a target meteorological station, carrying out the first division operation of the historical data, and providing a basis for the subsequent data processing. The second division operation further distinguishes the training set and the test set, and provides a reliable data source for the training of the quantum machine learning classifier. Through establishment of the quantum machine learning classifier, especially a plurality of different classifiers included in the quantum machine learning classifier, the diversity and accuracy of classification are improved. A training subset is randomly formed as input, so that the generalization ability of the model is enhanced. The application of the preset ensemble learning strategy not only optimizes the weight of the quantum machine learning classifier, but also improves the stability and reliability of the final prediction result through a voting mechanism. Compared with the prior art, the method has the advantage that the forecasting efficiency and accuracy are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of weather classification and prediction, and particularly relates to a severe convective weather classification and prediction method, system, device and medium. BACKGROUND

[0002] In meteorology, a convective phenomenon describes vertical air currents within the atmosphere that result from buoyancy effects. These currents can transport heat upward from the ground. When convective activity is particularly strong, it is referred to as severe convective weather. This type of weather is often triggered by a phenomenon known as deep moist convection (DMC), which can lead to a range of extreme weather events such as hail, strong winds, tornadoes, and heavy rain. These weather events share some common characteristics: they tend to occur suddenly, last for a relatively short period of time, and have a relatively localized impact. However, once they occur, they can cause significant damage and are characterized by their suddenness, short life span, strong local nature, and susceptibility to disaster.

[0003] To prevent and reduce the damage caused by severe convective weather, the national meteorological department has conducted in-depth research on the patterns of severe convective weather disasters and, in combination with breakthrough research on severe convective weather by the academic community, has established a series of precise research and short-term forecast and warning mechanisms. This enables timely dissemination of warning information, allowing the public to be aware of potentially dangerous weather and take appropriate preventive measures. However, current methods for classifying and predicting severe convective weather largely rely on traditional statistical methods and numerical weather prediction models. These methods have limitations in terms of accuracy and warning time when dealing with complex and variable severe convective weather.

[0004] In view of the above-mentioned existing problems, the present application is proposed.

[0005] Therefore, the present application provides a severe convective weather classification and prediction method, system, device and medium, which can solve the problem of low prediction accuracy and insufficient warning time in current methods for classifying and predicting severe convective weather when dealing with complex and variable weather conditions.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a severe convective weather classification and prediction method, comprising: obtaining historical data of a target meteorological observation station and performing a first division operation on the historical data; the first division operation divides the historical data into a plurality of different characteristic factor classes according to random numbers; the plurality of different characteristic factor classes include a water vapor factor, a dynamic factor, a height layer factor, a thermal factor, and a comprehensive index; The historical data after the first division operation is subjected to a second division operation, and the second division operation is used to distinguish a training set and a test set; A quantum machine learning classifier is established, the quantum machine learning classifier includes a plurality of different classifiers, and an input of the quantum machine learning classifier is a randomly composed training subset; The preset ensemble learning strategy is used to optimize the weight of the quantum machine learning classifier, and a voting is performed on the quantum machine learning classifier after optimization, so that a strong convective weather classification prediction result is obtained.

[0007] As a preferred scheme of the strong convective weather classification prediction method, the first division operation includes: The historical data of the target meteorological observation station is collected; A preset random number generation rule is used to allocate a random number to each sample, and the sample is a sample data set obtained through historical data; An interval range corresponding to each feature factor category is defined; The sample after the random number division is divided into the corresponding feature factor category according to the interval range.

[0008] As a preferred scheme of the strong convective weather classification prediction method, the second division operation includes: The data set proportion of each feature factor category obtained by the first division operation is divided; The training set and the test set after the division are detected; The detection includes detecting whether the training set and the test set overlap and whether the data distribution of the training set and the test set is consistent.

[0009] As a preferred scheme of the strong convective weather classification prediction method, the establishment of the quantum machine learning classifier includes: The training set is divided into a plurality of groups of randomly composed training subsets; A quantum machine learning classifier is established, and the quantum machine learning classifier includes a plurality of different classification models; Each classification model is trained using a randomly composed training subset; The plurality of different classification models include a plurality of quantum support vector machine models and a plurality of models established by a quantum nearest neighbor algorithm.

[0010] The preferred scheme improves the generalization ability of the model through group training, and avoids the occurrence of overfitting phenomenon. Meanwhile, the model established by combining the quantum support vector machine model and the quantum nearest neighbor algorithm can fully utilize the parallelism and high efficiency of quantum computing, improve the accuracy and speed of classification and prediction. This combination not only enhances the robustness of the model, but also makes the prediction result more reliable.

[0011] As a preferred scheme of the strong convective weather classification and prediction method, the voting on the optimized quantum machine learning classifier to obtain the strong convective weather classification and prediction result includes four classification targets, and the four classification targets include heavy rain, thunderstorm gale, hail and no strong convection.

[0012] As a preferred scheme of the strong convective weather classification and prediction method, the voting on the optimized quantum machine learning classifier to obtain the strong convective weather classification and prediction result further includes: If the target meteorological observation station has precipitation greater than the first precipitation threshold at a certain time, it is determined as the heavy rain category; If the target meteorological observation station has wind level reaching the second preset wind level or above at a certain time, it is determined as the thunderstorm gale category; If the target meteorological observation station has hail at a certain time, it is determined as the hail category; The remaining cases are all no strong convection categories.

[0013] As a preferred scheme of the strong convective weather classification and prediction method, the several different characteristic factor categories further include: The physical quantities as water vapor factors include total precipitable water, specific humidity, water vapor flux, relative humidity, water vapor flux divergence, temperature dew point difference, 925 hPa dew point temperature; The physical quantities as dynamic factors include divergence, vorticity, vertical velocity, vertical wind shear; The physical quantities as height layer factors include 0℃ layer height, -10℃ layer height, -20℃ layer height; The physical quantities as thermal factors include pseudo-equivalent potential temperature, K index, Showalter index, best lifting index, best unstable energy, unstable energy, total index, 850-500 hPa pseudo-equivalent potential temperature difference, 850 hPa temperature, conditional-convection stability index, sinking available potential energy, temperature difference; The physical quantities as comprehensive indexes include strong weather threat index, Swiss thunderstorm index, modified deep convection index, storm intensity index.

[0014] In a second aspect, the present application provides a strong convective weather classification and prediction system, which comprises: The first division operation is used for obtaining historical data of a target meteorological observation station and performing a first division operation on the historical data. The first division operation divides the historical data into several different characteristic factor classes according to random numbers. The several different characteristic factor classes include a water vapor factor, a dynamic factor, a height layer factor, a thermal factor, and a comprehensive index. The second division operation is used for performing a second division operation on the historical data after the first division operation, and the second division operation is used for distinguishing a training set and a test set. The classifier establishment module is used for establishing a quantum machine learning classifier, the quantum machine learning classifier includes several different classifiers, and an input of the quantum machine learning classifier is a training subset composed of random numbers. The prediction module is used for presetting an integrated learning strategy to optimize weights of the quantum machine learning classifier, and performing voting on the quantum machine learning classifier after optimization to obtain a strong convective weather classification prediction result.

[0015] In a third aspect, the present application provides an electronic device including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method described above when executing the computer program.

[0016] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the method described above.

[0017] Compared with the prior art, the present application has the following beneficial effects: the present application provides a strong convective weather classification prediction method, by obtaining historical data of a target meteorological observation station, the authenticity and representativeness of the data are ensured. Then, the first division operation is performed on the historical data, which provides a basis for subsequent data processing. The second division operation further distinguishes the training set and the test set, which provides a reliable data source for the training of the quantum machine learning classifier. The establishment of the quantum machine learning classifier, especially the several different classifiers contained therein, improves the diversity and accuracy of classification. By randomly composing a training subset as an input, the generalization ability of the model is enhanced. The application of the preset integrated learning strategy not only optimizes the weights of the quantum machine learning classifier, but also improves the stability and reliability of the final prediction result through the voting mechanism. Therefore, the strong convective weather classification prediction system of the present application has significantly improved the prediction efficiency and accuracy compared with the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only represent some of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0019] Figure 1 A method flow chart of a strong convective weather classification and prediction method provided by an embodiment of the present application.

[0020] Figure 2 A schematic diagram of an overall model of a strong convective weather classification and prediction method provided by an embodiment of the present application.

[0021] Figure 3 An internal structure diagram of an electronic device of a strong convective weather classification and prediction method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the above objectives, characteristics and advantages of the present application more apparent, the specific embodiments of the present application will be described in detail below with reference to the drawings in the specification. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort should fall within the scope of protection of the present application.

[0023] Embodiment 1, with reference to Figure 1 For the first embodiment of the present application, the embodiment provides a strong convective weather classification and prediction method, comprising: In the prior art, there are some problems, such as low prediction accuracy, insufficient early warning time, etc., which limit the accuracy and timeliness of the strong convective weather classification and prediction.

[0024] The present application provides a method that can effectively solve the above-mentioned problems. Next, how to implement the strong convective weather classification and prediction method will be described in detail in combination with multiple embodiments. Figure 1 A method flow chart of a strong convective weather classification and prediction method is shown, comprising: S101, obtaining historical data of a target meteorological observation station, and performing a first division operation on the historical data; It should be noted that in order to realize the classification of severe convective weather, it is necessary to know what severe convective weather is and its main characteristics. Severe convective weather is generally an extreme weather event caused by deep moist convection, which has the characteristics of suddenness, short life history, and strong locality. In the power grid system, wind power generation system and aviation transportation system, severe convective weather may cause serious disasters. Therefore, a method that can correctly and truly identify specific severe convective weather is needed.

[0025] In some specific embodiments, in order to make specific classification judgment of severe convective weather, it is necessary to consider which types of severe convective weather need to be judged. Severe convective weather includes many specific classifications, such as heavy precipitation, thunderstorm gale, hail, heavy snow and the like. Each type of severe convective weather may have different effects on the power grid industry. For example, heavy precipitation may cause power grid line short circuit or equipment immersion, thunderstorm gale may cause equipment damage or tree collapse affecting line safety, hail mainly affects the power grid in that its strong impact force may cause equipment damage, and heavy snow may cause line or equipment collapse due to excessive snow. Therefore, when constructing a severe convective weather classification and forecasting system, corresponding judgment modules need to be designed for these different types of severe convective weather to ensure that the system can accurately identify and warn weather conditions that may have a significant impact on the power grid industry.

[0026] It should be noted that in the process of specific severe convective weather judgment, the rules need to be found according to some historical data of meteorological observation stations, including but not limited to temperature, air pressure, humidity, wind speed, wind direction, precipitation and other key meteorological indicators. Through in-depth analysis of these historical data, the typical meteorological characteristics before the occurrence of different types of severe convective weather can be revealed, providing a scientific basis for subsequent classification and forecasting.

[0027] In the embodiment of the present application, the first division operation divides the historical data into several different characteristic factor classes according to random numbers; In the embodiment of the present application, the several different characteristic factor classes include water vapor factor, dynamic factor, height layer factor, thermal factor and comprehensive index; In some specific embodiments, water vapor is the basis for cloud and precipitation formation, and sufficient water vapor supply is one of the important conditions for the occurrence of severe convective weather. By analyzing the total column precipitable water, specific humidity, water vapor flux and other indicators, the content and transport of water vapor in the atmosphere can be evaluated. High water vapor content is usually associated with heavy precipitation, and specific water vapor flux divergence patterns may indicate the occurrence of thunderstorm gale or hail.

[0028] In some specific embodiments, the dynamic factors describe the state of motion in the atmosphere, including vertical velocity, divergence, vorticity, etc., which are crucial for understanding the dynamic processes of the atmosphere. Strong updrafts (positive vertical velocity) often accompany heavy precipitation; strong divergence and vorticity changes can indicate thunderstorm activity, especially when accompanied by strong vertical wind shear, which can be a precursor to tornadoes or severe thunderstorms.

[0029] In some specific embodiments, the height layer factors involve the temperature distribution at different height layers, such as the heights of the 0℃, -10℃, and -20℃ layers, which are very important for understanding ice phase processes. The presence or absence of these specific height layers and their height positions can help determine whether hail will occur, as hail formation requires specific temperature conditions and sufficient vertical development space.

[0030] In some specific embodiments, the thermal factors involve measurements of energy and stability, such as pseudo-equivalent potential temperature, Showalter index, etc., which provide key information about atmospheric stability. Higher unstable energy and lower stability index indicate that the atmospheric environment is conducive to convective development, especially in high-energy environments, which are prone to strong precipitation and thunderstorms. For example, the smaller the Showalter index, the more unstable the atmosphere, and the greater the likelihood of severe convective weather.

[0031] In some specific embodiments, the comprehensive index combines multiple physical quantity factors to calculate a comprehensive index, aiming to provide a more comprehensive perspective to assess the likelihood of severe convective weather. For example, the severe weather threat index, the Swiss thunderstorm index, etc., by considering multiple factors through weighted consideration, can more accurately predict the probability and intensity of short-term heavy precipitation, thunderstorm wind, hail, and other severe convective weather events.

[0032] In the embodiments of the present application, the several different characteristic factor categories also include: The physical quantities as water vapor factors include total column precipitable water, specific humidity, water vapor flux, relative humidity, water vapor flux divergence, temperature dew point difference, 925 hPa dew point temperature; The physical quantities as dynamic factors include divergence, vorticity, vertical velocity, vertical wind shear; The physical quantities as height layer factors include 0℃ layer height, -10℃ layer height, -20℃ layer height; The physical quantities as thermal factors include pseudo-equivalent potential temperature, K index, Showalter index, best lift index, best unstable energy, unstable energy, total index, 850-500 hPa pseudo-equivalent potential temperature difference, 850 hPa temperature, conditional-convection stability index, sinking available potential energy, temperature difference; The physical quantities as comprehensive indexes include severe weather threat index, Swiss thunderstorm index, modified deep convection index, storm intensity index.

[0033] Note that the total precipitable water is measured by ground and satellite remote sensing techniques, commonly using GPS meteorology or microwave radiometer.

[0034] Note that the specific humidity is measured directly by radiosonde at different heights and calculated from the water vapor content in the air.

[0035] Note that the water vapor flux is calculated based on wind speed and specific humidity data, using meteorological stations or reanalysis data.

[0036] Note that the relative humidity is calculated from the difference between temperature and dew point temperature, which can be obtained through the automatic weather station network.

[0037] Note that the water vapor flux divergence is calculated using numerical weather prediction model output data.

[0038] Note that the temperature-dew point difference is measured directly by temperature and dew point temperature, commonly found in ground meteorological stations.

[0039] Note that the 925 hPa dew point temperature is extracted from the information of a specific pressure level in the upper air sounding data.

[0040] Note that the divergence and vorticity can be based on wind field data, usually provided by numerical weather prediction models, or indirectly estimated by Doppler radar.

[0041] Note that the vertical velocity can be simulated by numerical model or directly measured by acoustic radar.

[0042] Note that the vertical wind shear can be calculated from wind speed and wind direction data at different heights, commonly derived from sounding data.

[0043] Note that the 0°C layer height, -10°C layer height, -20°C layer height: these height layer information mainly comes from sounding data, and can also be predicted by numerical model.

[0044] Note that the pseudo-equivalent temperature, K index, Showalter index, best lift index, best unstable energy, unstable energy, total index, 850-500 hPa pseudo-equivalent temperature difference, 850 hPa temperature, conditional-convection stability index, sinking effective potential energy, temperature difference: Most of the thermal parameters are calculated based on sounding data (such as temperature, humidity), and some parameters can also be estimated by numerical weather prediction model.

[0045] It should be noted that the severe weather threat index, the Swiss thunderstorm index, the modified deep convection index, and the storm intensity index: these indices are based on a variety of basic meteorological elements (such as temperature, humidity, wind, etc.) calculated by a specific algorithm, mainly used to assess the risk level of severe convective weather, usually relying on the results of numerical weather prediction models.

[0046] It should also be noted that after obtaining the historical data of the target meteorological observation station, a first division operation needs to be performed on the historical data. The first division operation here is used to divide the historical data into different characteristic factor categories.

[0047] In some specific embodiments, the first division operation can be implemented through a pre-set division logic. For example, a machine learning algorithm can be used to perform cluster analysis on the historical data, grouping data with similar characteristics into a category, thereby forming different characteristic factor categories. This process helps to identify key meteorological characteristics related to severe convective weather, providing a more accurate basis for subsequent classification forecasts. In addition, the first division operation can also consider factors such as data timeliness and regional characteristics to ensure the representativeness and accuracy of the selected characteristic factors. Through such a first division operation, the present application can more effectively utilize historical data, improving the accuracy and efficiency of severe convective weather classification forecasts.

[0048] In the embodiments of the present application, the first division operation includes: Collecting historical data of the target meteorological observation station; Defining different characteristic factor categories corresponding to the interval range; Assigning a random number to each sample based on a pre-set random number generation rule, where the sample is a sample data set obtained from the historical data; Dividing the samples after dividing the random numbers into corresponding characteristic factor categories according to the interval range.

[0049] It should be noted that the random number generation rule refers to the method or algorithm for generating random numbers. Commonly used are linear congruential generators, Mason rotation algorithms, etc. The role here is to assign a random number value to each sample to facilitate subsequent data processing.

[0050] It should be noted that in statistics and machine learning, a sample refers to a data unit extracted from a population for analysis. In this context, it refers to specific observation records extracted from historical meteorological data.

[0051] It should be noted that the characteristic factor category refers to the key meteorological factors selected for the analysis of severe convective weather. These factors can be divided into different categories, such as water vapor factors, dynamic factors, etc., each category containing several specific meteorological parameters.

[0052] It is important to note that for each characteristic factor category, the interval range defines one or more numerical ranges of its possible values. The purpose of this is to be able to assign samples to the correct category according to their characteristic values.

[0053] Specifically, the specific steps of dividing historical data into several different characteristic factor categories according to a fixed random number can be as follows: Step 1, determine the programming environment used (such as Python), and set a fixed random seed value to ensure consistency and repeatability of the division process. For example, in Python, you can achieve this by random.seed(fixed value) or numpy.random.seed(fixed value).

[0054] Step 2, collect the historical data of the target meteorological observation station, ensuring that these data contain all necessary characteristic factors, including but not limited to water vapor factors, dynamic factors, height layer factors, thermal factors, and comprehensive indexes, etc.

[0055] Step 3, for each record in the historical data set, generate a corresponding random number according to the set random seed. This step is based on the previously set fixed random seed, ensuring consistent results each time it is executed.

[0056] Step 4, define the random number range for each category according to the number of characteristic factor categories required for division. Assuming that you need to divide into 5 different characteristic factor categories, you can set the first category to correspond to [0, 0.2), the second category to correspond to [0.2, 0.4), and so on.

[0057] Step 5, iterate through the entire historical data set, and according to the random number value associated with each record and its corresponding preset category range, assign it to the corresponding characteristic factor category. This process may need to be automated by writing scripts to ensure efficient and accurate processing of large amounts of data.

[0058] Step 6, check the sample distribution in each characteristic factor category to ensure there is no obvious deviation or imbalance. If you find that the number of samples in some categories is too large or too small, you can adjust the random number generation logic or category boundary conditions as appropriate, and repeat the above steps until you are satisfied.

[0059] Step 7, store the divided data by category for subsequent use in building quantum machine learning classifiers as training sets and test sets.

[0060] It should be noted that obtaining the historical data of the target meteorological observation station and performing the first division operation on the historical data can significantly improve the efficiency and accuracy of data processing. By performing the first division operation on the historical data, i.e., preliminarily classifying the data according to the characteristic factor categories, the subsequent data processing and analysis work can be more targeted. In this way, not only unnecessary data redundancy can be reduced, but also the data under each characteristic factor category can have similar characteristics, thereby providing more accurate training set and test set for constructing the quantum machine learning classifier.

[0061] S102, performing a second division operation on the historical data after the first division operation, the second division operation being used to distinguish between a training set and a test set; It should be noted that after the first division operation, the law needs to be obtained according to the divided data, at which time some models can be established to assist in obtaining the law, therefore, a second division operation needs to be designed to facilitate the establishment of the final model.

[0062] In some specific embodiments, the second division operation can be performed by a preset division ratio, for example, 70% of the data is divided into a training set and the remaining 30% is divided into a test set. The training set is used to construct and train the quantum machine learning classifier, and the test set is used to evaluate the performance and accuracy of the classifier. Such division ratio can be adjusted according to actual needs to ensure that the model can be fully trained and verified.

[0063] When performing the second division operation, a random sampling method can be used to ensure that the data in the training set and the test set is representative. Specifically, a part of the historical data can be randomly selected as the training set, and the remaining part as the test set. In order to maintain the consistency of the data, the historical data can be shuffled before the second division operation is performed to avoid the influence of data order on the division result.

[0064] It should be noted that the second division operation is one of the important steps for constructing the quantum machine learning classifier. By reasonably dividing the training set and the test set, the classifier can learn the internal law of the data during the training process, and accurately evaluate its performance during the testing process. This is of great significance to improve the accuracy and reliability of the severe convective weather classification and prediction.

[0065] In the embodiment of the present application, the second division operation includes: performing data set ratio division on each characteristic factor category obtained by the first division operation; detecting the training set and the test set after the division is completed; The detection includes detecting whether the training set and the test set overlap, and whether the data distribution of the training set and the test set is consistent.

[0066] Specifically, the detailed operation steps of the second division operation can be as follows: First, a division ratio needs to be set for the training set and the test set, such as 80% of the data for the training set and 20% of the data for the test set.

[0067] Further, according to the selected ratio, the data of each feature factor category is respectively assigned to the training set and the test set using a random allocation method. In order to ensure the repeatability and consistency of the process, a fixed random seed value can be set.

[0068] Further, it is necessary to check whether there is any overlapping data sample between the training set and the test set. This step is to ensure that the model does not "see" the data in the test set during the training process, thereby avoiding overestimation of the model performance. This can be achieved by comparing the unique identifiers (such as timestamps, observation station IDs, etc.) of the samples in the two data sets.

[0069] Further, in addition to confirming that there is no overlap, it is also necessary to ensure that the data distribution of the training set and the test set is similar. This step is important because if there is a significant difference in the distribution of the two, the model trained on the training set may not generalize well to the test set. The basic statistics (mean, variance, etc.) of key features (such as temperature, humidity, etc.) in each data set can be calculated, and then these statistics can be compared to ensure their similarity.

[0070] Further, if significant differences in data distribution or overlap are found between the training set and the test set, the division method needs to be adjusted (for example, change the random seed value or reconsider the ratio), and the above steps need to be repeated until the requirements are met.

[0071] It should be noted that the second division operation on the historical data after the first division operation helps to improve the generalization ability of the model. By dividing the historical data into a training set and a test set first, it can be ensured that the model only contacts the training set data during the training process, while the test set data is used to evaluate the performance of the model. This approach can avoid the situation where the model overfits the training set data and performs poorly in real-world scenarios. At the same time, by carefully dividing the training set and the test set to ensure their similar data distribution, the prediction accuracy of the model on unknown data can be further improved. This double division strategy is one of the key steps in building a robust and reliable strong convective weather classification and forecasting system.

[0072] S103, a quantum machine learning classifier is established, the quantum machine learning classifier includes a plurality of different classifiers, and an input of the quantum machine learning classifier is a randomly composed training subset; In some specific embodiments, the established classification model can be based on neural networks or machine learning. When using neural networks and machine learning to establish, the steps can be as follows: First, collect and clean the historical data of the target weather observation station. This includes removing outliers, filling in missing values, and other operations.

[0073] Further, according to the needs, select the characteristic factors related to severe convective weather, such as water vapor factors, dynamic factors, and height layer factors, and perform necessary conversion or combination to improve the performance of the model.

[0074] Further, divide the processed data into training set and test set (usually the ratio of 80% training set and 20% test set), ensure that the distribution of the two is as consistent as possible and there is no overlap.

[0075] Further, select a suitable classifier from existing machine learning algorithms, such as logistic regression, decision tree, random forest, support vector machine (SVM), K nearest neighbor (KNN), etc., or select a neural network model, such as convolutional neural network (CNN), recurrent neural network (RNN), etc.

[0076] Further, use the training set to train the selected model. In this process, it may be necessary to adjust the hyperparameters to optimize the performance of the model.

[0077] Further, use the test set to evaluate the performance of the model. Common evaluation indicators include accuracy, precision, recall, F1 score, etc. According to the evaluation results, further adjust the model or reselect the model.

[0078] Further, once the model reaches a satisfactory performance level, it can be deployed to the production environment for actual prediction.

[0079] It should be noted that although traditional machine learning methods and neural networks have achieved remarkable success in many fields, they still face some challenges when dealing with complex and variable severe convective weather forecasting: Severe convective weather involves a large number of meteorological parameters, and there may be complex nonlinear relationships between these parameters. Quantum machine learning has the potential to handle high-dimensional data and can more effectively capture these complex relationships.

[0080] For certain types of computing tasks, quantum computers can theoretically provide exponential speedup. This means that quantum machine learning classifiers may be faster than traditional methods when dealing with large-scale meteorological data.

[0081] By adopting the superposition principle of quantum states and entanglement characteristics, quantum machine learning models can enhance the generalization ability of the model to some extent and reduce the occurrence of overfitting phenomenon.

[0082] The quantum machine learning classifier in the present application adopts a preset ensemble learning strategy for weight optimization and voting mechanism, which not only improves the stability of the model, but also enhances the reliability of the final prediction result.

[0083] In the embodiments of the present application, establishing a quantum machine learning classifier comprises: dividing the training set into several randomly composed training subsets; establishing a quantum machine learning classifier, the quantum machine learning classifier comprising several different classification models; each classification model is trained using a randomly composed training subset; the several different classification models include several quantum support vector machine models and several models established by quantum nearest neighbor algorithm.

[0084] It should be noted that the training subset is a part of data randomly extracted from the entire training set. In machine learning, a large training set is usually divided into multiple smaller subsets for cross-validation or ensemble learning. By training models using different training subsets, the generalization ability of the model can be improved and the risk of overfitting can be reduced.

[0085] It should be noted that the quantum machine learning classifier is a machine learning algorithm based on quantum computing technology, used to solve classification problems. It uses quantum bits (qubits) and quantum gate operations for data processing and pattern recognition. Compared with traditional machine learning methods, quantum machine learning may have higher efficiency and stronger expression ability when dealing with high-dimensional data.

[0086] It should be noted that support vector machine (SVM) is a classic supervised learning method used for classification and regression analysis. Quantum support vector machine (QSVM) is a variant of SVM that uses the advantages of quantum computing to speed up the solution process. When dealing with large-scale, complex feature space data, QSVM can more efficiently find the optimal hyperplane, thereby improving classification accuracy.

[0087] It should be noted that the nearest neighbor algorithm (KNN) is a simple non-parametric method used for classification and regression tasks. Quantum nearest neighbor algorithm (QKNN) is its quantum version, aiming to speed up the search process through quantum state superposition and entanglement characteristics. QKNN can significantly improve the computing speed while maintaining the advantages of the original algorithm, especially when dealing with high-dimensional data.

[0088] It should be noted that the randomly composed training subset refers to a new data set composed of a number of samples extracted from the original training set by a certain random way. Each extraction is independent and uniformly distributed. Ensuring that each training subset can represent the feature distribution of the whole data set, while increasing the diversity of model training, helps to improve the robustness and generalization ability of the final model.

[0089] For example, the detailed steps of building a quantum machine learning classifier can be as follows: 1. Initialize the quantum computing environment and select a suitable quantum computing platform, such as IBM's Qiskit, Google's Cirq, or Microsoft's Quantum Development Kit. If using a simulator, initialize the quantum computer simulator; if using an actual quantum computer, connect to the corresponding quantum hardware.

[0090] 2. Define feature mapping and encoding. In order to convert classical data into quantum states, feature mapping needs to be defined. Common mapping methods include ZFeatureMap, PauliFeatureMap, etc. Encode the data of the training subset into quantum states through feature mapping.

[0091] 3. Train the quantum support vector machine (QSVM). Initialize QSVM using the selected feature mapping and quantum instance. Load the randomly generated training subset into QSVM for training. Call the fit function to train the model and adjust the parameters to optimize performance.

[0092] 4. Train the quantum nearest neighbor algorithm (QKNN). Define the QKNN model, set the number of neighbors k and the quantum feature mapping. Load another randomly generated training subset into QKNN for training. Since QKNN is a non-parametric method, the main work is how to efficiently perform nearest neighbor search on quantum states.

[0093] 5. Integrate multiple QSVM and QKNN models. Repeat 3 and 4, for each training subset, train a set of QSVM and QKNN models respectively. Assuming there are 10 training subsets in this invention, there will be 10 QSVM models and 10 QKNN models.

[0094] It is worth noting that the establishment of quantum machine learning classifier can significantly improve the accuracy and efficiency of severe convective weather classification prediction. First, by using quantum machine learning classifier, high-dimensional and complex meteorological data can be processed more effectively, capturing nonlinear relationships that traditional methods cannot discover. Second, quantum machine learning classifier takes advantage of quantum computing, which can provide faster computing speed when dealing with large-scale data sets, which is crucial for real-time weather forecasting. In addition, by integrating multiple quantum support vector machines (QSVM) and quantum nearest neighbor algorithm (QKNN) models and training with randomly composed training subsets, the generalization ability of the model can be further enhanced, reducing the risk of overfitting. This ensemble learning strategy not only improves the stability of individual models, but also enhances the reliability of the final prediction results through the voting mechanism. Therefore, the establishment of quantum machine learning classifier is one of the key steps to achieve high-precision severe convective weather classification prediction.

[0095] In S104, the preset ensemble learning strategy is used to optimize the weights of the quantum machine learning classifier, and the optimized quantum machine learning classifier is voted to obtain the severe convective weather classification prediction result.

[0096] It is worth noting that after the establishment of the quantum machine learning classifier, the quantum machine learning classifier needs to be optimized in combination with the test set. At this time, methods such as ensemble learning strategy optimization, model pruning, and regularization can be selected for model optimization. Among them, ensemble learning strategy is an effective method that combines the prediction results of multiple models to improve the accuracy of overall prediction. In this invention, we use a preset ensemble learning strategy to optimize the weights of the quantum machine learning classifier.

[0097] Specifically, when selecting the ensemble learning strategy optimization, the following steps are included: First, collect the prediction results of each quantum model (such as QSVM and QKNN).

[0098] Then, according to the preset weight distribution scheme, the prediction results of each classifier are weighted.

[0099] These weights can be optimized through cross-validation, grid search, etc. to ensure that they accurately reflect the prediction ability of each classifier.

[0100] Finally, the weighted prediction results are summarized, and the final severe convective weather classification prediction result is obtained through the voting mechanism.

[0101] It is worth noting that the design of the redistribution and voting mechanism based on the final optimized quantum machine learning classifier is the key to ensuring the accuracy and reliability of the prediction results.

[0102] In the embodiment of the present application, the optimized quantum machine learning classifier is voted to obtain the strong convective weather classification prediction result, which includes four classification targets, and the four classification targets include strong precipitation, thunderstorm gale, hail and no strong convection.

[0103] In the embodiment of the present application, the optimized quantum machine learning classifier is voted to obtain the strong convective weather classification prediction result, which includes four classification targets, and the four classification targets include strong precipitation, thunderstorm gale, hail and no strong convection. If the target meteorological observation station has precipitation greater than the first precipitation threshold at a certain time, it is determined as the strong precipitation category; If the target meteorological observation station has wind level reaching the second preset wind level or above at a certain time, it is determined as the thunderstorm gale category; If the target meteorological observation station has hail at a certain time, it is determined as the hail category; The remaining cases are all no strong convection classification.

[0104] In some specific embodiments, the prediction results after weighting processing are summarized, and a voting mechanism is used to determine the final classification result. For example, a simple majority voting method can be used, that is, the category with the most votes is selected as the final prediction result; or a weighted voting method is used, and the opinions of each classifier are integrated according to the weights determined in advance.

[0105] In some cases, more complex fusion strategies such as Bayesian model averaging, stacking and other advanced integration techniques can also be applied to further improve prediction accuracy.

[0106] In summary, the present application proposes a strong convective weather classification prediction method, which obtains the historical data of the target meteorological observation station to ensure the authenticity and representativeness of the data. Then, the first division operation is performed on the historical data to provide a basis for subsequent data processing. The second division operation further distinguishes the training set and the test set to provide a reliable data source for the training of the quantum machine learning classifier. The establishment of the quantum machine learning classifier, especially the several different classifiers it contains, improves the diversity and accuracy of classification. By randomly forming a training subset as input, the generalization ability of the model is enhanced. The application of the preset ensemble learning strategy not only optimizes the weights of the quantum machine learning classifier, but also improves the stability and reliability of the final prediction result through the voting mechanism. Therefore, the strong convective weather classification prediction system of the present application has significantly improved the prediction efficiency and accuracy compared with the prior art.

[0107] Embodiment 2, refer to Figure 2 is an optional overall model architecture of the present application, in which: Firstly, the starting point of this overall architecture lies in identifying and selecting the independent variable factors closely related to severe convective weather. These factors encompass key parameters in meteorology, such as thermal factors, dynamic factors, water vapor factors, comprehensive indices, and special height layers, as input features for subsequent model construction. These factors are obtained through monitoring data and classified according to their physical significance.

[0108] Secondly, based on the selected independent variable factors, training sets and test sets are generated. This process involves randomly dividing historical data according to certain rules to form different subsets. Each subset contains various meteorological parameters as input features and the corresponding four types of severe convective weather (short-term heavy precipitation, thunderstorm gale, hail, and no severe convection) as target variables. In this way, sufficient training data can be provided for quantum machine learning classifiers to improve the generalization ability and accuracy of the model.

[0109] Thirdly, quantum support vector machines (QSVM) and quantum nearest neighbor (QKNN) algorithms are used to construct corresponding quantum machine learning classifiers for each randomly generated training subset. These classifiers focus on analyzing specific physical quantity factors and efficiently handle complex nonlinear problems using the advantages of quantum computing, achieving rapid feature space partitioning and pattern recognition. On this basis, an ensemble learning strategy is adopted to fuse the prediction results of multiple quantum classifiers through a voting principle to determine the final prediction result. This strategy not only reduces the bias of individual models but also enhances the stability and reliability of overall prediction.

[0110] Finally, to verify the effectiveness of the model, a test set independent of the training set is used for performance evaluation. By comparing the actual results of the test set with the model prediction results, error evaluation is completed, and the model is further optimized accordingly. This step is crucial for ensuring the accuracy and robustness of the model in practical applications and is an indispensable part of the entire process. Through such systematic steps, a complete chain from data preparation to model construction and performance evaluation is achieved, demonstrating the application potential of severe convective weather classification and prediction technology based on quantum machine learning classification models and ensemble learning.

[0111] In Example 3, refer to Figure 3 In this embodiment, a severe convective weather classification and prediction system is also provided, comprising: The first division operation is used to obtain the historical data of the target meteorological observation station and perform the first division operation on the historical data; The first division operation divides the historical data into several different characteristic factor classes according to random numbers; The several different characteristic factor classes include water vapor factors, dynamic factors, height layer factors, thermal factors, and comprehensive indices; A second division operation is used to perform a second division operation on the historical data after the first division operation, and the second division operation is used to distinguish the training set and the test set. The classifier establishment module is configured to establish a quantum machine learning classifier, the quantum machine learning classifier includes a plurality of different classifiers, and an input of the quantum machine learning classifier is a randomly composed training subset. The prediction module is configured to preset an ensemble learning strategy to optimize the weight of the quantum machine learning classifier, and perform voting on the quantum machine learning classifier after optimization to obtain a strong convective weather classification prediction result.

[0112] The above-mentioned various unit modules can be embedded in or independent of the processor in the electronic device in hardware form, or can be stored in the memory in the electronic device in software form, so as to call and execute the operations corresponding to the above-mentioned various modules by the processor.

[0113] The embodiment also provides an electronic device, which can be a terminal, and an internal structure diagram of the electronic device can be as shown in the figure. Figure 3 The electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the electronic device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a strong convective weather classification prediction method. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the electronic device. In addition, an external keyboard, touchpad or mouse can also be used.

[0114] The embodiment also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the following steps: Obtaining historical data of a target meteorological observation station, and performing a first division operation on the historical data; The first division operation divides the historical data into a plurality of different characteristic factor classes according to random numbers; The plurality of different characteristic factor classes include a water vapor factor, a dynamic factor, a height layer factor, a thermal factor and a comprehensive index; Performing a second division operation on the historical data after the first division operation, and the second division operation is used to distinguish the training set and the test set; A quantum machine learning classifier is established, the quantum machine learning classifier including a plurality of different classifiers, and an input of the quantum machine learning classifier being a randomly composed training subset; The preset ensemble learning strategy optimizes the weight of the quantum machine learning classifier, and votes the quantum machine learning classifier after optimization to obtain a strong convection weather classification prediction result.

[0115] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all modifications and replacements should be covered in the scope of the claims of the present application.

[0116] Although the preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to the embodiments once they understand the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all modifications and replacements falling within the scope of the present application.

[0117] Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and changes of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and changes.

Claims

1. A method for classifying and forecasting severe convective weather, characterized in that, include: Acquire historical data from the target meteorological observation station and perform a first segmentation operation on the historical data; The first partitioning operation divides the historical data into several different feature factor classes according to random numbers; The various characteristic factor categories include water vapor factor, dynamic factor, altitude layer factor, thermal factor, and comprehensive index; The historical data after the first partitioning operation is then subjected to a second partitioning operation, which is used to distinguish between the training set and the test set. A quantum machine learning classifier is established, which includes several different classifiers, and the input of the quantum machine learning classifier is a randomly composed training subset; An ensemble learning strategy is preset to optimize the weights of the quantum machine learning classifier, and the optimized quantum machine learning classifier is voted on to obtain the classification and forecast results of severe convective weather.

2. The method for classifying and forecasting severe convective weather as described in claim 1, characterized in that, The first partitioning operation includes: Collect historical data from the target meteorological observation station; A preset random number generation rule is set, and random numbers are assigned to each sample, wherein the sample is a sample dataset obtained through historical data; Define the interval range corresponding to different feature factor categories; The samples after being randomly numbered are divided into corresponding feature factor categories according to the interval range.

3. The method for classifying and forecasting severe convective weather as described in claim 2, characterized in that, The second partitioning operation includes: Divide the dataset proportionally for each feature factor category obtained from the first partitioning operation; The training and test sets are then tested after the partitioning is completed; The detection includes detecting whether the training set and the test set overlap, and whether the data distribution of the training set and the test set is consistent.

4. The method for classifying and forecasting severe convective weather as described in claim 3, characterized in that, The establishment of the quantum machine learning classifier includes: The training set is divided into several randomly formed training subsets; A quantum machine learning classifier is established, which includes several different classification models. Each classification model is trained using a randomly composed subset of training data; The various classification models include several sets of quantum support vector machine models and several sets of models established by quantum nearest neighbor algorithms.

5. The method for classifying and forecasting severe convective weather as described in claim 4, characterized in that, The optimized quantum machine learning classifier is voted on to obtain severe convective weather classification forecast results, which include four classification targets: heavy precipitation, thunderstorms and strong winds, hail, and no severe convection.

6. The method for classifying and forecasting severe convective weather as described in claim 5, characterized in that, The process of voting on the optimized quantum machine learning classifier to obtain the severe convective weather classification and forecast results also includes: If a target meteorological observation station experiences precipitation exceeding the first precipitation threshold at a certain time, it is classified as heavy precipitation. If the wind level at a target meteorological observation station reaches or exceeds the second preset wind level at a certain moment, it will be classified as a thunderstorm gale. If a hail is recorded at a certain time at the target meteorological observation station, it is classified as a hail. All other cases are classified as having no strong convection.

7. The method for classifying and forecasting severe convective weather as described in claim 6, characterized in that, The aforementioned different feature factor classes also include: Physical quantities that are water vapor factors include total precipitable water, specific humidity, water vapor flux, relative humidity, water vapor flux divergence, temperature-dew point difference, and 925 hPa dew point temperature. Physical quantities that serve as dynamic factors include divergence, vorticity, vertical velocity, and vertical wind shear. The physical quantities that serve as altitude factors include 0°C altitude, -10°C altitude, and -20°C altitude. Physical quantities that are thermodynamic factors include pseudo-equivalent potential temperature, K index, Sablius index, optimal uplift index, optimal unstable energy, unstable energy, total index, pseudo-equivalent potential temperature difference between 850 and 500 hPa, 850 hPa temperature, conditional-convective stability index, sinking effective potential energy, and temperature difference. The physical quantities used as a composite index include the Severe Weather Threat Index, the Swiss Thunderstorm Index, the Modified Deep Convection Index, and the Storm Intensity Index.

8. A severe convective weather classification forecasting system, using the method described in any one of claims 1 to 7, characterized in that, include: The first partitioning operation is used to acquire historical data from the target meteorological observation station and perform a first partitioning operation on the historical data; The first partitioning operation divides the historical data into several different feature factor classes according to random numbers; The various characteristic factor categories include water vapor factor, dynamic factor, altitude layer factor, thermal factor, and comprehensive index; The second partitioning operation is used to perform a second partitioning operation on the historical data after the first partitioning operation. The second partitioning operation is used to distinguish between the training set and the test set. A classifier building module is used to build a quantum machine learning classifier, which includes several different classifiers, and the input of the quantum machine learning classifier is a randomly composed training subset; The prediction module is used to optimize the weights of the quantum machine learning classifier by pre-setting an ensemble learning strategy, and to vote on the optimized quantum machine learning classifier to obtain the classification and forecast results of severe convective weather.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the severe convective weather classification and forecasting method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the severe convective weather classification and forecasting method according to any one of claims 1 to 7.

Images