X-band weather radar reflectivity data quality control method based on deep learning

Abstract

The invention discloses an X-band weather radar reflectivity data quality control method based on deep learning, and belongs to the technical field of meteorological data processing. The deep learning technology is introduced, the convolutional neural network is used for extracting the space and time characteristics of the original reflectivity data, high-precision data quality evaluation and classification are achieved, noise, low-quality and high-quality data can be efficiently recognized under the complex meteorological condition, and through real-time monitoring and a periodic optimization mechanism, the real-time performance of the system is improved. Model parameters are dynamically adjusted to adapt to data input changing in real time, the accuracy and reliability of data processing are guaranteed, through modular integration of a deep learning model, the deep learning model can be embedded into an existing meteorological system, automatic and intelligent application of data quality control is achieved, the data processing efficiency and quality control precision are remarkably improved, and the data processing efficiency is improved. And technical support is provided for the fields of weather prediction, climate research, disaster early warning and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological data processing, and in particular to a method for quality control of X-band weather radar reflectivity data based on deep learning. Background Art

[0002] In modern meteorological observation and forecasting, weather radar is an important detection method and is widely used in precipitation monitoring, wind field analysis and disaster weather warning. Among them, X-band weather radar is particularly suitable for monitoring local severe convective weather due to its high resolution and sensitivity.

[0003] However, the quality of raw reflectivity data is often affected by a variety of factors, such as noise interference, equipment failure, and echoes from non-meteorological targets (such as flocks of birds and insects), which can lead to anomalies or errors in radar data, thus affecting the accuracy of meteorological analysis and prediction. Traditional radar data quality control methods usually rely on empirical thresholds and rule-based algorithms, which are difficult to effectively cope with complex and changing meteorological conditions and lack the ability to deeply mine the abnormal features of data. Summary of the invention

[0004] The purpose of the present invention is to provide an X-band weather radar reflectivity data quality control method based on deep learning to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for quality control of X-band weather radar reflectivity data based on deep learning, comprising the following steps: Data acquisition and processing: obtaining raw reflectivity data through the X-band weather radar system and preprocessing the raw reflectivity data; Build a deep learning model and select a convolutional neural network to build a deep learning model; Construct a training data set, obtain raw reflectance data under different weather conditions, classify the reflectance data into different quality levels based on expert annotations, obtain reflectance data samples and integrate them into a training data set; Deep learning model training optimization: Use the constructed training data set to train the deep learning model, use the cross entropy loss function and gradient descent algorithm to optimize the deep learning model, and use the back propagation algorithm to update the network weights of the deep learning model and adjust the hyperparameters; Noise and abnormal data detection: input the original reflectivity data to be detected into the deep learning model, and the deep learning model outputs the quality assessment result of the reflectivity data. Based on the quality assessment result, the original reflectivity data is processed and a three-dimensional spatiotemporal reflectivity dataset is generated; Data quality assessment feedback, real-time monitoring of raw reflectivity data processed by the deep learning model, and periodic optimization of the deep learning model; Application integration and actual deployment: Integrate the optimized deep learning model into the meteorological forecast and climate research system to automatically control the quality of reflectivity data acquired in real time.

[0006] It is characterized in that the data collection and processing further includes the following steps: Collecting raw reflectivity data through an X-band weather radar system, the raw reflectivity data including radar echo intensity values ​​and corresponding spatial coordinates and time coordinates, to form an original three-dimensional spatiotemporal reflectivity data set, wherein the radar echo intensity value is measured by detecting the echo characteristics of a target object by radar, the spatial coordinates are used to identify the radar echo position, and the time coordinates are used to record the time point of radar echo collection; According to the working characteristics of the radar equipment and the set threshold, the echo data below the noise threshold is eliminated, and the echo data of repeated acquisition and abnormal time points are deduplicated; According to the input requirements of the deep learning model, the spatial coordinates and time coordinates in the original three-dimensional spatiotemporal reflectivity dataset are mapped to a unified grid space, and the radar echo intensity values ​​are normalized; The raw reflectance data is segmented according to the time series, and the raw reflectance data in continuous time periods are organized into input segments.

[0007] Furthermore, the operating quality of the X-band weather radar system is monitored in real time, specifically comprising the following steps: Real-time monitoring of transmitter operating parameters of an X-band weather radar system, wherein the transmitter operating parameters include pulse peak power, pulse width, and pulse repetition frequency; Obtaining a transmitter operating coefficient using the pulse peak power, pulse width and pulse repetition frequency; The transmitter operation coefficient is obtained by the following formula: ; Where F is the transmitter operating factor; Q is the pulse peak power; t is the pulse width; f is the pulse repetition frequency; Q c Indicates the preset pulse peak power reference value; t c Indicates the preset pulse width reference value; comparing the transmitter operating coefficient with a preset transmitter operating coefficient threshold; When the transmitter operation coefficient exceeds a preset transmitter operation coefficient threshold, it is determined that the transmitter operation is abnormal, and an abnormal transmitter operation warning is issued.

[0008] Furthermore, the operating quality of the X-band weather radar system is monitored in real time, which specifically includes the following steps: Real-time monitoring of receiver operating parameters of an X-band weather radar system, wherein the receiver operating parameters include actual minimum measurable power and noise figure; Obtaining a receiver operating coefficient using the actual minimum measurable power and the noise figure; The receiver operating coefficient is obtained by the following formula: ; Where S represents the receiver operating coefficient; P m represents the actual minimum measurable power; Z represents the noise factor; Z y Indicates the preset noise coefficient reference value; P mb represents the rate of change of the actual minimum measurable power; x and y represent the first adjustment coefficient and the second adjustment coefficient respectively, and the value ranges of the first adjustment coefficient and the second adjustment coefficient are 0.31-0.42 and 0.12-3.14; comparing the receiver operating coefficient with a preset receiver operating coefficient threshold; When the receiver operation coefficient exceeds a preset receiver operation coefficient threshold, it is determined that the receiver operation is abnormal, and an abnormal receiver operation warning is issued.

[0009] Furthermore, the operating quality of the X-band weather radar system is monitored in real time, which specifically includes the following steps: Real-time monitoring of scanning operation parameters of the X-band weather radar system, wherein the scanning operation parameters include scanning speed, scanning angle deviation rate and scanning acceleration; comparing the scanning speed with a preset scanning speed threshold; When the scanning speed is lower than a preset scanning speed threshold, the first scanning operation parameter is acquired using the scanning speed and the scanning acceleration; Wherein, the first scanning operation parameter is obtained by the following formula: ; Among them, Y 01 represents the first scanning operation parameter; v represents the scanning speed; a represents the scanning acceleration; P s Indicates the scanning speed change response delay rate; P a Indicates the scanning acceleration change efficiency; comparing the first scanning operation parameter with a preset first scanning parameter threshold; When the first scanning operation parameter exceeds a preset first scanning parameter threshold, it is determined that the scanning operation is abnormal, and a scanning operation abnormality alarm is issued; When the scanning speed is not lower than a preset scanning speed threshold, the second scanning operation parameter is acquired using the scanning speed, the scanning angle deviation rate and the scanning acceleration; The second scanning operation parameter is obtained by the following formula: ; Among them, Y 02 represents the second scanning operation parameter; v represents the scanning speed; a represents the scanning acceleration; P s Indicates the scanning speed change response delay rate; P a represents the scanning acceleration change efficiency; k represents the scanning angle deviation rate; comparing the second scanning operation parameter with a preset second scanning parameter threshold; When the second scanning operation parameter exceeds a preset second scanning parameter threshold, it is determined that the scanning operation is abnormal, and a scanning operation abnormality alarm is issued.

[0010] Furthermore, the data collection and processing further includes the following steps: Collecting raw reflectivity data through an X-band weather radar system, the raw reflectivity data including radar echo intensity values ​​and corresponding spatial coordinates and time coordinates, to form an original three-dimensional spatiotemporal reflectivity data set, wherein the radar echo intensity value is measured by detecting the echo characteristics of a target object by radar, the spatial coordinates are used to identify the radar echo position, and the time coordinates are used to record the time point of radar echo collection; According to the working characteristics of the radar equipment and the set threshold, the echo data below the noise threshold is eliminated, and the echo data of repeated acquisition and abnormal time points are deduplicated; According to the input requirements of the deep learning model, the spatial coordinates and time coordinates in the original three-dimensional spatiotemporal reflectivity dataset are mapped to a unified grid space, and the radar echo intensity values ​​are normalized; The raw reflectance data is segmented according to the time series, and the raw reflectance data in continuous time periods are organized into input segments.

[0011] Furthermore, the construction of the deep learning model specifically includes the following steps: Constructing a basic architecture of a convolutional neural network, wherein the convolutional neural network includes a convolutional layer, a pooling layer, and a fully connected layer; In the convolution layer, the convolution kernel is designed to extract the local spatial features and time series features of the original reflectivity data; in the pooling layer, the average pooling strategy is introduced to compress the feature map; in the fully connected layer, the extracted local spatial features and time series features are analyzed and classified through the activation function; Initialize the deep learning model weights and set hyperparameters, including learning rate, batch size, and number of training rounds, determine the number of convolution kernels, convolution kernel size, and pooling window size; Configure the input layer and output layer for the deep learning model, visually verify the deep learning model and perform preliminary tests, and verify the deep learning model structure through preliminary tests.

[0012] Furthermore, the construction of the training data set further includes the following steps: Collect reflectivity data covering multiple weather modes, including sunny, rainy, foggy and extreme weather conditions, through a weather radar system, and the time span of the collected reflectivity data covers the all-weather operation cycle; The collected reflectivity data are preliminarily classified according to the meteorological mode. By combining the radar working records and the real-time data of the meteorological station, the weather category corresponding to the reflectivity data is clarified and the reflectivity data is sorted. Each set of reflectivity data contains the reflectivity value and the corresponding time and space coordinate information; Experts label the reflectivity data with quality level labels based on the temporal and spatial distribution characteristics of the reflectivity data and the meteorological observation results. The quality level labels include high-quality samples, low-quality samples, and noise samples. Among them, normal data is marked as high-quality samples; abnormal data caused by interference sources and equipment failures are marked as low-quality samples; and data without physical meaning and invalid noise are marked as noise samples; The normal data specifically includes the following data types: Data whose values ​​conform to physical laws, fluctuate within a reasonable range, and are consistent with surrounding data; Data that conforms to specific spatiotemporal distribution patterns and is highly correlated with meteorological observations; Data that is consistent with meteorological observations and does not have abnormal fluctuations that contradict meteorological conditions; Data with fluctuations within a reasonable range and without obvious abnormal fluctuations or jumps; The abnormal data specifically includes the following data types: Data outside the normal range; Data that may be inconsistent with surrounding data and show prominent fluctuations or jumps; Data inconsistent with meteorological observations; The invalid noise data specifically includes the following data types: Data that has no physical meaning and whose fluctuations do not conform to any physical or meteorological model; Data that appear to be random noise and have no obvious spatial or temporal patterns; After the labeling is completed, the reflectivity data and the corresponding quality level labels are integrated into reflectivity data samples, and the reflectivity data samples are integrated to generate a training data set.

[0013] Furthermore, the deep learning model training optimization further includes the following steps: The training data set is randomly divided into a training set, a validation set, and a test set, and the reflectance values ​​in the training data set are normalized; Train the deep learning model using the training dataset; Define the cross entropy loss function, calculate the loss value by comparing the probability distribution of the deep learning model output with the one-hot encoded vector of the true label, optimize the model parameters through the gradient descent algorithm, adjust the network weights using mini-batch stochastic gradient descent, calculate the gradient value of the loss function to the deep learning model parameters through the back propagation algorithm, update the network weights and biases according to the gradient value, and dynamically adjust the hyperparameters; Monitor the training status of the deep learning model in real time, adjust hyperparameters based on the performance feedback of the validation set, and optimize the model structure and training process.

[0014] Furthermore, the noise and abnormal data detection specifically includes the following steps: According to the input format of the deep learning model, the original reflectivity data to be detected is processed according to the time series and spatial region segments, the abnormal boundary values ​​in the original reflectivity data to be detected are detected, and invalid data segments are eliminated; The pre-processed raw reflectivity data segments are sequentially input into the deep learning model, and the deep learning model discriminates the reflectivity data according to a preset classification threshold, and outputs a quality assessment result for each raw reflectivity data segment, wherein the quality assessment result includes three classifications: high quality, low quality, and noise; Eliminate noise data and low-quality data in quality assessment results; Among them, the original reflectivity data segments that the deep learning model determines to be noise are directly removed from the data set; for low-quality data, correction is performed by combining adjacent time and space segments and real-time data from the meteorological station. If the correction fails to meet the standards, it is marked as invalid data and removed; After eliminating invalid data, the remaining high-quality data are reintegrated into a continuous three-dimensional spatiotemporal reflectivity dataset.

[0015] Furthermore, the data quality assessment feedback specifically includes the following steps: Monitor the deviation of the original reflectivity data processed by the deep learning model in real time, compare the quality assessment results of the original reflectivity data output by the deep learning model with the actual meteorological observation results, and calculate the reflectivity value, spatial distribution characteristics and consistency deviation of the time series; Set the deviation tolerance range threshold based on historical data and expert experience. When the consistency deviation exceeds the deviation tolerance range threshold, mark it as abnormal data that needs further optimization. Based on the deviation monitoring results, the deep learning model is periodically optimized, the output data of the deep learning model in real-time operation is regularly extracted, the consistency deviation is automatically compared and the optimization process is triggered.

[0016] Furthermore, the application integration and actual deployment specifically include the following steps: The optimized deep learning model is modularly integrated with the existing meteorological system, and the deep learning model is embedded in the real-time data processing link of the meteorological system to perform quality assessment, classification, and remove noise and abnormal data for the received real-time reflectivity data; Conduct periodic performance evaluation of deployed deep learning models and evaluate the deep learning models by comparing the newly generated quality assessment results with the actual meteorological conditions.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention significantly improves the quality and processing efficiency of raw reflectivity data through automated data collection, processing and optimization methods. Through unified grid-based spatial coordinate and time coordinate mapping, the data organization is made more scientific and standardized, which is conducive to subsequent deep learning model processing. The deep learning technology is used to achieve fully automated data processing, which greatly reduces the workload and error risk of human intervention. The efficient data collection and processing capabilities enable the method to quickly respond to real-time data input from radar equipment, providing a guarantee for the meteorological system to obtain high-quality radar data in a short time, which is particularly important when dealing with extreme weather and rapidly changing meteorological conditions. It can significantly improve the efficiency of radar data quality control and provide reliable support for subsequent meteorological forecasts and research.

[0018] 2. The present invention takes convolutional neural network as the core, and realizes accurate classification of data quality according to the characteristics of original reflectivity data. By constructing a dedicated deep learning architecture, the convolution layer is used to extract the local spatial features and time series features of the data, and the feature compression strategy of the pooling layer and the classification ability of the fully connected layer are combined, the classification accuracy is significantly improved. Through the quality evaluation results of the model, noise data can be eliminated, low-quality data can be corrected, and a high-quality three-dimensional spatiotemporal reflectivity data set can be generated, which not only improves the signal-to-noise ratio and credibility of the data, but also provides strong support for complex analysis and prediction in subsequent meteorological services. At the same time, the dynamic optimization capability of the deep learning model ensures that it can still maintain excellent performance under new data conditions, and has strong adaptability and scalability.

[0019] 3. The present invention embeds the optimized model into the existing meteorological system in a modular way, so that it can seamlessly connect to the real-time radar data stream and automatically complete quality control. This integration method improves the intelligence of the system and reduces the frequency of manual intervention. At the same time, it can continuously optimize according to the system operating status to improve the operating efficiency of the entire meteorological system. In the fields of weather forecasting, climate research, and disaster warning, the present invention provides an efficient and reliable technical solution, which provides strong technical support for the deepening development of social and economic benefits and meteorological scientific research. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the X-band weather radar reflectivity data quality control method based on deep learning of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] See also Figure 1 , the present invention provides the following technical solutions: The X-band weather radar reflectivity data quality control method based on deep learning includes the following steps: Data acquisition and processing: obtaining raw reflectivity data through the X-band weather radar system and preprocessing the raw reflectivity data; Build a deep learning model and select a convolutional neural network to build a deep learning model; Construct a training data set, obtain raw reflectance data under different weather conditions, classify the reflectance data into different quality levels based on expert annotations, obtain reflectance data samples and integrate them into a training data set; Deep learning model training optimization: Use the constructed training data set to train the deep learning model, use the cross entropy loss function and gradient descent algorithm to optimize the deep learning model, and use the back propagation algorithm to update the network weights of the deep learning model and adjust the hyperparameters; Noise and abnormal data detection: input the original reflectivity data to be detected into the deep learning model, and the deep learning model outputs the quality assessment result of the reflectivity data. Based on the quality assessment result, the original reflectivity data is processed and a three-dimensional spatiotemporal reflectivity dataset is generated; Data quality assessment feedback, real-time monitoring of raw reflectivity data processed by the deep learning model, and periodic optimization of the deep learning model; Application integration and actual deployment: Integrate the optimized deep learning model into the meteorological forecast and climate research system to automatically control the quality of reflectivity data acquired in real time.

[0023] In the above embodiment, with the deep learning model as the core, by constructing a dedicated data set and optimizing the model training process, efficient and accurate evaluation and control of the quality of the original reflectivity data are achieved. Compared with the traditional rule-based quality control method, the deep learning model's efficient computing power can automatically identify noise and abnormal data, reduce manual intervention, and greatly improve the efficiency and real-time performance of quality control. The model is trained with a variety of meteorological pattern data to enhance its adaptability to complex meteorological conditions and is suitable for different climate regions and weather conditions. The high-precision classification capability of the deep learning model can effectively eliminate invalid data, and the generated high-quality three-dimensional spatiotemporal reflectivity data set provides reliable data support for subsequent meteorological forecasts and climate research. The method is designed to be compatible with a variety of X-band weather radar systems, and supports seamless integration with meteorological business systems through modular integration, so that the model can be continuously optimized and updated according to demand.

[0024] Specifically, real-time monitoring of the operating quality of the X-band weather radar system includes the following steps: Real-time monitoring of transmitter operating parameters of an X-band weather radar system, wherein the transmitter operating parameters include pulse peak power, pulse width, and pulse repetition frequency; Obtaining a transmitter operating coefficient using the pulse peak power, pulse width and pulse repetition frequency; The transmitter operation coefficient is obtained by the following formula: ; Where F is the transmitter operating factor; Q is the pulse peak power; t is the pulse width; f is the pulse repetition frequency; Q c Indicates the preset pulse peak power reference value; t c Indicates the preset pulse width reference value; comparing the transmitter operating coefficient with a preset transmitter operating coefficient threshold; When the transmitter operation coefficient exceeds a preset transmitter operation coefficient threshold, it is determined that the transmitter operation is abnormal, and an abnormal transmitter operation warning is issued.

[0025] The technical effect of the above technical solution is: the technical solution can timely detect abnormal conditions of the transmitter by real-time monitoring of the transmitter operating parameters (including pulse peak power, pulse width and pulse repetition frequency) of the X-band weather radar system. When the transmitter operating coefficient exceeds the preset threshold, the system can automatically issue an early warning to ensure the stable operation of the radar system. By calculating the transmitter operating coefficient, the technical solution provides a method for quantitatively evaluating transmitter performance. The transmitter operating coefficient combines three key parameters: pulse peak power, pulse width and pulse repetition frequency, and can fully reflect the working status of the transmitter. This helps operation and maintenance personnel to understand the performance of the transmitter more accurately and take corresponding maintenance measures. Since the X-band weather radar system is often used for monitoring and early warning of disastrous weather, its reliability is crucial. Through real-time monitoring and early warning, the technical solution can detect and handle transmitter failures in time before they occur, thereby avoiding the radar system from being paralyzed due to transmitter failures. This greatly improves the reliability and stability of the radar system. Through real-time monitoring and early warning, operation and maintenance personnel can perform maintenance before the transmitter fails, avoiding downtime and maintenance costs caused by failures. In addition, the method of quantitatively evaluating transmitter performance also helps operation and maintenance personnel to formulate maintenance plans more accurately, reducing operation and maintenance costs. This technical solution provides a new transmitter performance monitoring and early warning method, and provides new ideas for the technological advancement of X-band weather radar systems. By continuously optimizing and improving this technical solution, the performance and reliability of the radar system can be further improved, providing more accurate and timely data support for meteorological monitoring and early warning.

[0026] In summary, this technical solution achieves real-time monitoring and early warning of the radar system operation quality by real-time monitoring of the transmitter operating parameters of the X-band weather radar system and calculating the transmitter operating coefficient for performance evaluation. This greatly improves the reliability and stability of the radar system, reduces operation and maintenance costs, and promotes the advancement of radar system technology.

[0027] Specifically, real-time monitoring of the operating quality of the X-band weather radar system further includes the following steps: Real-time monitoring of receiver operating parameters of an X-band weather radar system, wherein the receiver operating parameters include actual minimum measurable power and noise figure; Obtaining a receiver operating coefficient using the actual minimum measurable power and the noise figure; The receiver operating coefficient is obtained by the following formula: ; Where S represents the receiver operating coefficient; P m represents the actual minimum measurable power; Z represents the noise factor; Z y Indicates the preset noise coefficient reference value; Pmb represents the rate of change of the actual minimum measurable power; x and y represent the first adjustment coefficient and the second adjustment coefficient respectively, and the value ranges of the first adjustment coefficient and the second adjustment coefficient are 0.31-0.42 and 0.12-3.14; comparing the receiver operating coefficient with a preset receiver operating coefficient threshold; When the receiver operation coefficient exceeds a preset receiver operation coefficient threshold, it is determined that the receiver operation is abnormal, and an abnormal receiver operation warning is issued.

[0028] The technical effect of the above technical solution is: by real-time monitoring of the receiver operating parameters of the X-band weather radar system, including the actual minimum measurable power and noise factor, the technical solution can fully reflect the performance status of the receiver. These parameters are important indicators for evaluating the sensitivity, anti-interference ability and signal processing ability of the receiver. By introducing the receiver operating coefficient S, the technical solution provides a method for quantitatively evaluating the performance of the receiver. The coefficient comprehensively considers the actual minimum measurable power, noise factor and their rate of change, and introduces the first adjustment coefficient x and the second adjustment coefficient y, so that the evaluation result is more accurate and comprehensive. This helps the operation and maintenance personnel to understand the performance status of the receiver more accurately, find problems in time and take corresponding measures. When the receiver operating coefficient exceeds the preset threshold, the system can automatically issue an early warning to prompt the operation and maintenance personnel that the receiver may be operating abnormally. This real-time abnormal early warning function can greatly shorten the time for fault detection and processing, and improve the reliability and stability of the radar system. The value ranges of the first adjustment coefficient x and the second adjustment coefficient y are 0.31-0.42 and 0.12-3.14, respectively, which provides greater flexibility. According to different application scenarios and receiver performance requirements, the values ​​of these coefficients can be adjusted to better adapt to the actual situation. Through real-time monitoring and early warning functions, operation and maintenance personnel can perform maintenance before the receiver fails, avoiding downtime and maintenance costs caused by failures. At the same time, the method of quantitatively evaluating receiver performance also helps operation and maintenance personnel to formulate maintenance plans more accurately and improve operation and maintenance efficiency. This technical solution provides a new method for receiver performance monitoring and early warning, and provides new ideas for the technological advancement of X-band weather radar systems. By continuously optimizing and improving this technical solution, the performance and reliability of the radar system can be further improved, providing more accurate and timely data support for meteorological monitoring and early warning.

[0029] In summary, this technical solution monitors the receiver operating parameters of the X-band weather radar system in real time and introduces the receiver operating coefficient for quantitative evaluation, thus achieving real-time monitoring and early warning functions for the receiver operating quality. This greatly improves the reliability and stability of the radar system, reduces operation and maintenance costs, and promotes the advancement of radar system technology.

[0030] Specifically, real-time monitoring of the operating quality of the X-band weather radar system further includes the following steps: Real-time monitoring of scanning operation parameters of the X-band weather radar system, wherein the scanning operation parameters include scanning speed, scanning angle deviation rate and scanning acceleration; comparing the scanning speed with a preset scanning speed threshold; When the scanning speed is lower than a preset scanning speed threshold, the first scanning operation parameter is acquired using the scanning speed and the scanning acceleration; Wherein, the first scanning operation parameter is obtained by the following formula: ; Among them, Y 01 represents the first scanning operation parameter; v represents the scanning speed; a represents the scanning acceleration; P s Indicates the scanning speed change response delay rate; P a Indicates the scanning acceleration change efficiency; comparing the first scanning operation parameter with a preset first scanning parameter threshold; When the first scanning operation parameter exceeds a preset first scanning parameter threshold, it is determined that the scanning operation is abnormal, and a scanning operation abnormality alarm is issued; When the scanning speed is not lower than a preset scanning speed threshold, the second scanning operation parameter is acquired using the scanning speed, the scanning angle deviation rate and the scanning acceleration; The second scanning operation parameter is obtained by the following formula: ; Among them, Y 02 represents the second scanning operation parameter; v represents the scanning speed; a represents the scanning acceleration; P s Indicates the scanning speed change response delay rate; P a represents the scanning acceleration change efficiency; k represents the scanning angle deviation rate; comparing the second scanning operation parameter with a preset second scanning parameter threshold; When the second scanning operation parameter exceeds a preset second scanning parameter threshold, it is determined that the scanning operation is abnormal, and a scanning operation abnormality alarm is issued.

[0031] The technical effect of the above technical solution is: the technical solution monitors the scanning operation parameters of the X-band weather radar system in real time, including scanning speed, scanning angle deviation rate and scanning acceleration, and these parameters can fully reflect the operating status of the radar scanning system. By monitoring these parameters, abnormal conditions in the scanning process can be discovered in time to ensure the normal operation of the radar system. In the technical solution, the scanning speed is first compared with the preset scanning speed threshold, and different calculation methods and thresholds are selected according to the comparison results for further judgment. This flexible processing method can more accurately identify abnormal conditions in the scanning operation and avoid false alarms and missed alarms. When the scanning speed is lower than the preset threshold, the first scanning operation parameter is calculated using the scanning speed and scanning acceleration, and compared with the preset first scanning parameter threshold; when the scanning speed is not lower than the preset threshold, the second scanning operation parameter is calculated using the scanning speed, scanning angle deviation rate and scanning acceleration, and compared with the preset second scanning parameter threshold. This precise calculation and comparison method can timely discover abnormalities in the scanning process and ensure the stability and reliability of the radar system. When the scanning operation parameter exceeds the preset threshold, the system can automatically issue an abnormal warning to prompt the operation and maintenance personnel to take timely measures. This real-time abnormal warning function can greatly shorten the time for fault discovery and processing, and improve the maintenance efficiency of the radar system. Through real-time monitoring and early warning functions, operation and maintenance personnel can perform maintenance before the scanning system fails, avoiding downtime and maintenance costs caused by failures. At the same time, accurate operation parameter calculation and abnormality judgment also help operation and maintenance personnel to formulate maintenance plans more accurately, improving operation and maintenance efficiency. This technical solution provides a new method for scanning operation quality monitoring and abnormality early warning, and provides new ideas for the technological advancement of X-band weather radar systems. By continuously optimizing and improving this technical solution, the performance and reliability of the radar system can be further improved, providing more accurate and timely data support for meteorological monitoring and early warning.

[0032] In summary, this technical solution achieves real-time monitoring of the scanning operation parameters of the X-band weather radar system and performs accurate calculations and comparisons, thus realizing the functions of real-time monitoring of the scanning operation quality and abnormal warning. This greatly improves the stability and reliability of the radar system, reduces the operation and maintenance costs, and promotes the advancement of radar system technology.

[0033] The data collection and processing specifically includes the following steps: Collecting raw reflectivity data through an X-band weather radar system, the raw reflectivity data including radar echo intensity values ​​and corresponding spatial coordinates and time coordinates, to form an original three-dimensional spatiotemporal reflectivity data set, wherein the radar echo intensity value is measured by detecting the echo characteristics of a target object by radar, the spatial coordinates are used to identify the radar echo position, and the time coordinates are used to record the time point of radar echo collection; According to the working characteristics of the radar equipment and the set threshold, the echo data below the noise threshold is eliminated, and the echo data of repeated acquisition and abnormal time points are deduplicated; According to the input requirements of the deep learning model, the spatial coordinates and time coordinates in the original three-dimensional spatiotemporal reflectivity dataset are mapped to a unified grid space, and the radar echo intensity values ​​are normalized; The raw reflectance data is segmented according to the time series, and the raw reflectance data in continuous time periods are organized into input segments.

[0034] In the above embodiment, in the data acquisition and processing link, an X-band weather radar system is used to collect the original reflectivity data with high precision, and the data is preprocessed in combination with the input requirements of the deep learning model, thereby laying a good foundation for subsequent model training and quality assessment. By setting reasonable noise threshold values ​​and abnormal time point detection rules, noise data and redundant data generated by environmental interference or equipment failure can be effectively removed, and the purity and reliability of the data can be significantly improved. The dimensional difference of the radar echo intensity value is eliminated through normalization processing, and the space and time coordinates are unified through grid processing. A consistent data input format is provided for the deep learning model to ensure the stability and efficiency of the training and reasoning process. The original reflectivity data is organized as input segments according to the time series, which retains the dynamic characteristics of the data, so that the deep learning model can more accurately learn the spatiotemporal characteristics of the data and improve the classification performance.

[0035] Building a deep learning model also includes the following steps: Constructing a basic architecture of a convolutional neural network, wherein the convolutional neural network includes a convolutional layer, a pooling layer, and a fully connected layer; In the convolution layer, the convolution kernel is designed to extract the local spatial features and time series features of the original reflectivity data; in the pooling layer, the average pooling strategy is introduced to compress the feature map; in the fully connected layer, the extracted local spatial features and time series features are analyzed and classified through the activation function; Initialize the deep learning model weights and set hyperparameters, including learning rate, batch size, and number of training rounds, determine the number of convolution kernels, convolution kernel size, and pooling window size; Configure the input layer and output layer for the deep learning model, visually verify the deep learning model and perform preliminary tests, and verify the deep learning model structure through preliminary tests.

[0036] In the above embodiment, in the model construction link, a convolutional neural network is selected as the basic model architecture, and targeted optimization is performed in combination with the characteristics of weather radar data. The convolutional neural network can efficiently extract local spatial features and time series features in the original reflectivity data by designing appropriate convolution kernels, laying the foundation for subsequent quality assessment. Through the organic combination of convolution layers, pooling layers and fully connected layers, the functional division of feature extraction, feature compression and classification is realized. At the same time, the redundancy of feature maps is reduced through the average pooling strategy, and the generalization ability of the model is improved. By optimizing hyperparameters such as the number of convolution kernels, the size of the convolution kernel, and the size of the pooling window, the computational complexity and performance of the model are balanced to ensure efficient operation on different hardware platforms. The feasibility and accuracy of the model structure are verified through preliminary tests, providing a reliable basis for subsequent model training optimization.

[0037] Constructing a training data set also includes the following steps: Collect reflectivity data covering multiple weather modes, including sunny, rainy, foggy and extreme weather conditions, through a weather radar system, and the time span of the collected reflectivity data covers the all-weather operation cycle; The collected reflectivity data are preliminarily classified according to the meteorological mode. By combining the radar working records and the real-time data of the meteorological station, the weather category corresponding to the reflectivity data is clarified and the reflectivity data is sorted. Each set of reflectivity data contains the reflectivity value and the corresponding time and space coordinate information; Experts label the reflectivity data with quality level labels based on the temporal and spatial distribution characteristics of the reflectivity data and the meteorological observation results. The quality level labels include high-quality samples, low-quality samples, and noise samples. Among them, normal data is marked as high-quality samples; abnormal data caused by interference sources and equipment failures are marked as low-quality samples; and data without physical meaning and invalid noise are marked as noise samples; The normal data specifically includes the following data types: Data whose values ​​conform to physical laws, fluctuate within a reasonable range, and are consistent with surrounding data; Data that conforms to specific spatiotemporal distribution patterns and is highly correlated with meteorological observations; Data that is consistent with meteorological observations and does not have abnormal fluctuations that contradict meteorological conditions; Data with fluctuations within a reasonable range and without obvious abnormal fluctuations or jumps; The abnormal data specifically includes the following data types: Data outside the normal range; Data that may be inconsistent with surrounding data and show prominent fluctuations or jumps; Data inconsistent with meteorological observations; The invalid noise data specifically includes the following data types: Data that has no physical meaning and whose fluctuations do not conform to any physical or meteorological model; Data that appear to be random noise and have no obvious spatial or temporal patterns; After the labeling is completed, the reflectivity data and the corresponding quality level labels are integrated into reflectivity data samples, and the reflectivity data samples are integrated to generate a training data set.

[0038] In the above embodiment, in the data set construction stage, the quality and comprehensiveness of the training data set are ensured by introducing diversified meteorological patterns and expert-annotated data, thereby improving the classification ability and adaptability of the deep learning model. By collecting original reflectivity data under sunny, rainy, foggy and extreme weather conditions, the model's adaptability to complex weather conditions is enhanced, and it can be widely used in different climatic regions. Combined with the real-time data of the meteorological station and the radar working records, experts annotate the data one by one to ensure the accuracy of the quality grade labels, provide high-quality labeled samples for the supervised learning of the model, and divide the data into three levels: high quality, low quality and noise, so that the model can accurately distinguish different types of data, significantly improving the accuracy and efficiency of data classification.

[0039] Deep learning model training optimization also includes the following steps: The training data set is randomly divided into a training set, a validation set, and a test set, and the reflectance values ​​in the training data set are normalized; Train the deep learning model using the training dataset; Define the cross entropy loss function, calculate the loss value by comparing the probability distribution of the deep learning model output with the one-hot encoded vector of the true label, optimize the model parameters through the gradient descent algorithm, adjust the network weights using mini-batch stochastic gradient descent, calculate the gradient value of the loss function to the deep learning model parameters through the back propagation algorithm, update the network weights and biases according to the gradient value, and dynamically adjust the hyperparameters; Monitor the training status of the deep learning model in real time, adjust hyperparameters based on the performance feedback of the validation set, and optimize the model structure and training process.

[0040] In the above embodiment, in the training optimization link, by constructing a reasonable training process and optimization algorithm, efficient training and performance improvement of the model are ensured. By randomly dividing the training set, validation set and test set, it is ensured that the model can be fully evaluated at different stages to avoid the occurrence of overfitting. The cross entropy loss function and small batch stochastic gradient descent algorithm are used to dynamically adjust the model weights and biases through back propagation to improve the classification accuracy and convergence speed of the model. According to the performance feedback of the validation set, the hyperparameters such as learning rate, batch size and number of training rounds are adjusted in real time to ensure that the model can maintain the best state at different training stages.

[0041] Noise and abnormal data detection specifically includes the following steps: According to the input format of the deep learning model, the original reflectivity data to be detected is processed according to the time series and spatial region segments, the abnormal boundary values ​​in the original reflectivity data to be detected are detected, and invalid data segments are eliminated; The pre-processed raw reflectivity data segments are sequentially input into the deep learning model, and the deep learning model discriminates the reflectivity data according to a preset classification threshold, and outputs a quality assessment result for each raw reflectivity data segment, wherein the quality assessment result includes three classifications: high quality, low quality, and noise; Eliminate noise data and low-quality data in quality assessment results; Among them, the original reflectivity data segments that the deep learning model determines to be noise are directly removed from the data set; for low-quality data, correction is performed by combining adjacent time and space segments and real-time data from the meteorological station. If the correction fails to meet the standards, it is marked as invalid data and removed; After eliminating invalid data, the remaining high-quality data are reintegrated into a continuous three-dimensional spatiotemporal reflectivity dataset.

[0042] In the above embodiment, in the noise and abnormal data detection link, through the efficient classification ability of the deep learning model, invalid data can be accurately eliminated and the quality of the data set can be improved. By detecting and eliminating abnormal boundary values, invalid data caused by radar failure or environmental interference can be reduced, ensuring the high credibility of the data. The low-quality data is corrected by combining the adjacent space-time fragments and the real-time data of the meteorological station, thereby improving the overall validity of the data. After eliminating the invalid data, the remaining high-quality data is reintegrated into a continuous three-dimensional space-time reflectivity data set to provide reliable data support for subsequent applications.

[0043] Data quality assessment feedback also includes the following steps: Monitor the deviation of the original reflectivity data processed by the deep learning model in real time, compare the quality assessment results of the original reflectivity data output by the deep learning model with the actual meteorological observation results, and calculate the reflectivity value, spatial distribution characteristics and consistency deviation of the time series; Set the deviation tolerance range threshold based on historical data and expert experience. When the consistency deviation exceeds the deviation tolerance range threshold, mark it as abnormal data that needs further optimization. Based on the deviation monitoring results, the deep learning model is periodically optimized, the output data of the deep learning model in real-time operation is regularly extracted, the consistency deviation is automatically compared and the optimization process is triggered.

[0044] In the above embodiment, dynamic evaluation of data quality and continuous optimization of the model are achieved through real-time monitoring and periodic optimization. Through real-time monitoring of consistency deviations, abnormal data can be quickly discovered and targeted processing measures can be taken. The output results of the model are regularly compared and analyzed to trigger the optimization process, thereby continuously improving the adaptability and processing capabilities of the model. The automated optimization process reduces manual intervention and significantly improves the response speed and optimization efficiency of the system.

[0045] Application integration and actual deployment also include the following steps: The optimized deep learning model is modularly integrated with the existing meteorological system, and the deep learning model is embedded in the real-time data processing link of the meteorological system to perform quality assessment, classification, and remove noise and abnormal data for the received real-time reflectivity data; Perform periodic performance evaluation on the deployed deep learning models and evaluate the deep learning models by comparing the newly generated quality assessment results with the actual meteorological conditions.

[0046] In the above embodiments, in the actual deployment stage, modular integration and automated optimization are used to ensure the efficient operation of the model in the meteorological system. The modular design of the deep learning model is seamlessly connected with the meteorological system, realizing the automation and intelligence of real-time data processing. Through performance evaluation and continuous deployment framework, it is ensured that the model can quickly adapt to changing needs in actual applications and achieve long-term stable operation. The generated high-quality original reflectivity dataset provides important support for meteorological forecasting and climate research, and promotes technological progress in related fields.

[0047] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for quality control of X-band weather radar reflectivity data based on deep learning, characterized in that: The following steps are involved: Data acquisition and processing: obtaining raw reflectivity data through the X-band weather radar system and preprocessing the raw reflectivity data; Build a deep learning model and select a convolutional neural network to build a deep learning model; Construct a training data set, obtain raw reflectance data, classify the reflectance data into different quality levels based on expert annotation, obtain reflectance data samples and integrate them into a training data set; Deep learning model training optimization: Use the constructed training data set to train the deep learning model, use the cross entropy loss function and gradient descent algorithm to optimize the deep learning model, and use the back propagation algorithm to update the network weights of the deep learning model and adjust the hyperparameters; Noise and abnormal data detection: input the original reflectivity data to be detected into the deep learning model, and the deep learning model outputs the quality assessment result of the reflectivity data. Based on the quality assessment result, the original reflectivity data is processed and a three-dimensional spatiotemporal reflectivity dataset is generated; Data quality assessment feedback, real-time monitoring of raw reflectivity data processed by the deep learning model, and periodic optimization of the deep learning model; Application integration and actual deployment: Integrate the optimized deep learning model into the meteorological forecast and climate research system to automatically control the quality of reflectivity data acquired in real time.

2. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 1, characterized in that: The data collection and processing further includes the following steps: Collecting raw reflectivity data through an X-band weather radar system, the raw reflectivity data including radar echo intensity values ​​and corresponding spatial coordinates and time coordinates, to form an original three-dimensional spatiotemporal reflectivity data set, wherein the radar echo intensity value is measured by detecting the echo characteristics of a target object by radar, the spatial coordinates are used to identify the radar echo position, and the time coordinates are used to record the time point of radar echo collection; According to the working characteristics of the radar equipment and the set threshold, the echo data below the noise threshold is eliminated, and the echo data of repeated acquisition and abnormal time points are deduplicated; According to the input requirements of the deep learning model, the spatial coordinates and time coordinates in the original three-dimensional spatiotemporal reflectivity dataset are mapped to a unified grid space, and the radar echo intensity values ​​are normalized; The raw reflectance data is segmented according to the time series, and the raw reflectance data in continuous time periods are organized into input segments.

3. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 2, characterized in that: Real-time monitoring of the operational quality of the X-band weather radar system includes the following steps: Real-time monitoring of transmitter operating parameters of an X-band weather radar system, wherein the transmitter operating parameters include pulse peak power, pulse width, and pulse repetition frequency; Obtaining a transmitter operating coefficient using the pulse peak power, pulse width and pulse repetition frequency; The transmitter operation coefficient is obtained by the following formula: ； Where F is the transmitter operating factor; Q is the pulse peak power; t is the pulse width; f is the pulse repetition frequency; Q c Indicates the preset pulse peak power reference value; t c Indicates the preset pulse width reference value; comparing the transmitter operating coefficient with a preset transmitter operating coefficient threshold; When the transmitter operation coefficient exceeds a preset transmitter operation coefficient threshold, it is determined that the transmitter operation is abnormal, and an abnormal transmitter operation warning is issued.

4. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 3, characterized in that: The operation quality of the X-band weather radar system is monitored in real time, and specifically includes the following steps: Real-time monitoring of receiver operating parameters of an X-band weather radar system, wherein the receiver operating parameters include actual minimum measurable power and noise figure; Obtaining a receiver operating coefficient using the actual minimum measurable power and the noise figure; The receiver operating coefficient is obtained by the following formula: ； Where S represents the receiver operating coefficient; P m represents the actual minimum measurable power; Z represents the noise factor; Z y Indicates the preset noise coefficient reference value; P mb represents the rate of change of the actual minimum measurable power; x and y represent the first adjustment coefficient and the second adjustment coefficient respectively, and the value ranges of the first adjustment coefficient and the second adjustment coefficient are 0.31-0.42 and 0.12-3.14; comparing the receiver operating coefficient with a preset receiver operating coefficient threshold; When the receiver operation coefficient exceeds a preset receiver operation coefficient threshold, the receiver operation is determined to be abnormal, and an abnormal receiver operation warning is issued.

5. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 4, characterized in that: The operation quality of the X-band weather radar system is monitored in real time, and specifically includes the following steps: Real-time monitoring of scanning operation parameters of the X-band weather radar system, wherein the scanning operation parameters include scanning speed, scanning angle deviation rate and scanning acceleration; comparing the scanning speed with a preset scanning speed threshold; When the scanning speed is lower than a preset scanning speed threshold, the first scanning operation parameter is acquired using the scanning speed and the scanning acceleration; Wherein, the first scanning operation parameter is obtained by the following formula: ； Among them, Y 01 represents the first scanning operation parameter; v represents the scanning speed; a represents the scanning acceleration; P s Indicates the scanning speed change response delay rate; P a Indicates the scanning acceleration change efficiency; comparing the first scanning operation parameter with a preset first scanning parameter threshold; When the first scanning operation parameter exceeds a preset first scanning parameter threshold, it is determined that the scanning operation is abnormal, and a scanning operation abnormality alarm is issued; When the scanning speed is not lower than a preset scanning speed threshold, the second scanning operation parameter is obtained using the scanning speed, the scanning angle deviation rate and the scanning acceleration; The second scanning operation parameter is obtained by the following formula: ； Among them, Y 02 represents the second scanning operation parameter; v represents the scanning speed; a represents the scanning acceleration; P s Indicates the scanning speed change response delay rate; P a represents the scanning acceleration change efficiency; k represents the scanning angle deviation rate; comparing the second scanning operation parameter with a preset second scanning parameter threshold; When the second scanning operation parameter exceeds a preset second scanning parameter threshold, it is determined that the scanning operation is abnormal, and a scanning operation abnormality alarm is issued.

6. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 1, characterized in that: The construction of the deep learning model specifically further includes the following steps: Constructing a basic architecture of a convolutional neural network, wherein the convolutional neural network includes a convolutional layer, a pooling layer, and a fully connected layer; In the convolution layer, the convolution kernel is designed to extract the local spatial features and time series features of the original reflectivity data; in the pooling layer, the average pooling strategy is introduced to compress the feature map; in the fully connected layer, the extracted local spatial features and time series features are analyzed and classified through the activation function; Initialize the deep learning model weights and set hyperparameters, including learning rate, batch size, and number of training rounds, determine the number of convolution kernels, convolution kernel size, and pooling window size; Configure the input layer and output layer for the deep learning model, visually verify the deep learning model and perform preliminary tests, and verify the deep learning model structure through preliminary tests.

7. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 1, characterized in that: The construction of the training data set specifically further includes the following steps: Collect reflectivity data covering multiple weather modes, including sunny, rainy, foggy and extreme weather conditions, through a weather radar system, and the time span of the collected reflectivity data covers the all-weather operation cycle; The collected reflectivity data are preliminarily classified according to the meteorological mode. By combining the radar working records and the real-time data of the meteorological station, the weather category corresponding to the reflectivity data is clarified and the reflectivity data is sorted. Each set of reflectivity data contains the reflectivity value and the corresponding time and space coordinate information; Experts label the reflectivity data with quality level labels based on the temporal and spatial distribution characteristics of the reflectivity data and the meteorological observation results. The quality level labels include high-quality samples, low-quality samples, and noise samples. Among them, normal data is marked as high-quality samples; abnormal data caused by interference sources and equipment failures are marked as low-quality samples; invalid noise data is marked as noise samples; The normal data specifically includes the following data types: Data whose values ​​conform to physical laws, fluctuate within a reasonable range, and are consistent with surrounding data; Data that conforms to specific spatiotemporal distribution patterns and is highly correlated with meteorological observations; Data that is consistent with meteorological observations and does not have abnormal fluctuations that contradict meteorological conditions; Data with fluctuations within a reasonable range and without obvious abnormal fluctuations or jumps; The abnormal data specifically includes the following data types: Data outside the normal range; Data that may be inconsistent with surrounding data and show prominent fluctuations or jumps; Data inconsistent with meteorological observations; The invalid noise data specifically includes the following data types: Data that has no physical meaning and whose fluctuations do not conform to any physical or meteorological model; Data that appear to be random noise and have no obvious spatial or temporal patterns; After the labeling is completed, the reflectivity data and the corresponding quality level labels are integrated into reflectivity data samples, and the reflectivity data samples are integrated to generate a training data set.

8. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 1, characterized in that: The deep learning model training optimization specifically further includes the following steps: The training data set is randomly divided into a training set, a validation set, and a test set, and the reflectance values ​​in the training data set are normalized; Train the deep learning model using the training dataset; Define the cross entropy loss function, calculate the loss value by comparing the probability distribution of the deep learning model output with the one-hot encoded vector of the true label, optimize the model parameters through the gradient descent algorithm, adjust the network weights using mini-batch stochastic gradient descent, calculate the gradient value of the loss function to the deep learning model parameters through the back propagation algorithm, update the network weights and biases according to the gradient value, and dynamically adjust the hyperparameters; Monitor the training status of the deep learning model in real time, adjust hyperparameters based on the performance feedback of the validation set, and optimize the model structure and training process.

9. The X-band weather radar reflectivity data quality control method based on deep learning as claimed in claim 1, characterized in that: The noise and abnormal data detection specifically further includes the following steps: According to the input format of the deep learning model, the original reflectivity data to be detected is processed according to the time series and spatial region segments, the abnormal boundary values ​​in the original reflectivity data to be detected are detected, and invalid data segments are eliminated; The pre-processed raw reflectivity data segments are sequentially input into the deep learning model, and the deep learning model discriminates the reflectivity data according to a preset classification threshold, and outputs a quality assessment result for each raw reflectivity data segment, wherein the quality assessment result includes three classifications: high quality, low quality, and noise; Eliminate noise data and low-quality data in quality assessment results; Among them, the original reflectivity data segments that the deep learning model determines to be noise are directly removed from the data set; for low-quality data, correction is performed by combining adjacent time and space segments and real-time data from the meteorological station. If the correction fails to meet the standards, it is marked as invalid data and removed; After eliminating invalid data, the remaining high-quality data are reintegrated into a continuous three-dimensional spatiotemporal reflectivity dataset.

10. The X-band weather radar reflectivity data quality control method based on deep learning according to claim 1, characterized in that: The data quality assessment feedback specifically includes the following steps: Monitor the deviation of the original reflectivity data processed by the deep learning model in real time, compare the quality assessment results of the original reflectivity data output by the deep learning model with the actual meteorological observation results, and calculate the reflectivity value, spatial distribution characteristics and consistency deviation of the time series; Set the deviation tolerance range threshold based on historical data and expert experience. When the consistency deviation exceeds the deviation tolerance range threshold, mark it as abnormal data that needs further optimization. Based on the deviation monitoring results, the deep learning model is periodically optimized, the output data of the deep learning model in real-time operation is regularly extracted, the consistency deviation is automatically compared and the optimization process is triggered; The application integration and actual deployment specifically further include the following steps: The optimized deep learning model is modularly integrated with the existing meteorological system, and the deep learning model is embedded in the real-time data processing link of the meteorological system to perform quality assessment, classification, and remove noise and abnormal data for the received real-time reflectivity data; Perform periodic performance evaluation on the deployed deep learning models and evaluate the deep learning models by comparing the newly generated quality assessment results with the actual meteorological conditions.

Images