A multi-channel circular access method and system for meteorological observation data

Abstract

An embodiment of the present invention discloses a multi-channel circular access method and system for meteorological observation data, belonging to the technical field of meteorological data processing. The embodiment of the present invention uses a circular buffer structure to manage meteorological observation data. Compared with the traditional linear stack, the circular buffer has obvious advantages. The linear stack has low efficiency when the data volume is large, while the circular buffer supports multiple independent data streams to share the storage space, realizes continuous writing and circular overwrite, such as different sensor data streams can be stored in an orderly manner. The circular movement of its read and write pointers ensures continuous access to data, and the parallel access ability improves the speed and efficiency, meeting the multi-task concurrency requirements of meteorological services. The overflow and underflow detection mechanism and capacity management strategy ensure data integrity and system stability, preventing problems such as data loss or system crashes. Attribute knowledge mining and related mapping help to accurately manage data, manage according to data characteristics, improve decision-making accuracy and efficiency, and optimize data management and operation of the meteorological business system.

Description

Technical Field

[0001] The embodiments of the present invention belong to the technical field of meteorological data processing, and particularly relate to a multi-channel circular access method and system for meteorological observation data. Background Art

[0002] In the development process of meteorological business systems, equipment operation control and data management have always been core issues. The traditional linear stack access method could meet the basic needs of meteorological business systems in the early stage when the data volume was small and specific requirements for sequence existed. For example, in the storage and sequential execution of simple equipment control instructions, the last-in, first-out principle of the linear stack could ensure that instructions were executed by the equipment in a specific order, and when the data volume was limited, the efficiency of managing data was acceptable.

[0003] However, with the continuous development of meteorological observation technologies, meteorological business systems are facing new challenges. On the one hand, the increase in meteorological observation equipment and the improvement of observation accuracy have led to a sharp increase in data volume. In this case, the limitations of the linear stack gradually become apparent. Since the insertion and deletion of data elements in the linear stack are both performed at the top of the stack, as the data volume increases, the storage and reading efficiency of data will decrease significantly. For example, when a large amount of meteorological observation data (such as long-term sequence data from multiple meteorological stations) needs to be stored and processed, the operations of the linear stack will become very slow, resulting in system response delays and affecting the normal operation of meteorological services.

[0004] On the other hand, in scenarios of unstable networks and complex multi-threading in meteorological business systems, the lack of flexibility of the linear stack becomes a serious problem. When the network is unstable, data transmission may be interrupted or delayed, and the linear stack is difficult to effectively handle this situation, which may lead to data loss or incorrect command execution order. In a multi-threading scenario, such as when meteorological data collection, equipment control, data query, etc. are carried out simultaneously, the linear stack cannot meet the concurrent access requirements of multiple threads for data, which limits the overall performance of the system.

[0005] In addition, the linear stack lacks effective capacity management and data integrity protection mechanisms in data management. As the data volume increases, the linear stack will occupy more bandwidth, which not only affects data transmission efficiency but may also cause problems in the operation of the business system. Moreover, due to the lack of dedicated overflow and underflow detection mechanisms, when the data volume exceeds the stack capacity or data reading is abnormal, it is easy to cause data loss or system errors, and the integrity of data and the stability of the system in meteorological business systems cannot be guaranteed. Summary of the Invention

[0006] The embodiments of the present invention provide a multi-channel circular access method and system for meteorological observation data, which can solve or partially solve the technical problems involved in the above background art.

[0007] An embodiment of the present invention provides a multi-channel circular access method for meteorological observation data, which is applied to a meteorological observation data access system. The method includes: respectively taking multiple meteorological observation data streams in a target database as initial meteorological observation data streams, performing a timing marking operation based on a circular buffer structure on the initial meteorological observation data streams to obtain multiple timing marked observation information, and each piece of the timing marked observation information includes multiple target meteorological observation data; respectively performing attribute knowledge mining on the multiple target meteorological observation data to obtain meteorological observation attribute knowledge codes of the multiple target meteorological observation data; respectively taking the multiple timing marked observation information as data access marking information, performing a read-write state mapping on the meteorological observation attribute knowledge codes of the multiple target meteorological observation data in the data access marking information to obtain a read-write state description vector of the data access marking information; using the read-write state description vectors of the multiple timing marked observation information to determine an access management recommendation vector of the initial meteorological observation data stream; performing a read-write state mapping on the access management recommendation vectors of the multiple meteorological observation data streams in the target database to obtain a global read-write state label of the target database, and the global read-write state label of the target database is used to determine a data access strategy of the target database.

[0008] An embodiment of the present invention provides a meteorological observation data access system, which includes at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the above method.

[0009] An embodiment of the present invention provides a readable storage medium, on which a program or instructions are stored, and when the program or instructions are executed by a processor, the steps of the above method are implemented.

[0010] The embodiment of the present invention manages meteorological observation data by using a circular buffer structure. Compared with the traditional linear stack, the circular buffer has obvious advantages. The linear stack has low efficiency when the data volume is large, while the circular buffer supports multiple independent data streams to share the storage space, realizes continuous writing and circular overwrite. For example, data streams from different sensors can be stored in an orderly manner. The circular movement of its read and write pointers ensures continuous access to data, and the parallel access ability improves the speed and efficiency, meeting the concurrent needs of multiple tasks in meteorological services. The overflow and underflow detection mechanism and capacity management strategy ensure data integrity and system stability, preventing problems such as data loss or system crashes. Attribute knowledge mining and related mapping help to manage data accurately, manage according to data characteristics, improve the accuracy and efficiency of decision-making, and optimize the data management and operation of the meteorological business system. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1It is a flowchart of a multi-channel circular access method for meteorological observation data provided by an embodiment of the present invention.

[0012] Figure 2 It is a schematic structural diagram of a meteorological observation data access system provided by an embodiment of the present invention. Detailed implementation manners

[0013] Figure 1 Disclosed is a multi-channel circular access method for meteorological observation data, which is applied to a meteorological observation data access system. The method includes the following steps 110 - step 150.

[0014] Step 110: Respectively take multiple meteorological observation data streams in a target database as initial meteorological observation data streams, and perform a time series marking operation based on a circular buffer structure on the initial meteorological observation data streams to obtain multiple time series marked observation information, and each piece of the time series marked observation information includes multiple target meteorological observation data.

[0015] In an optional embodiment, the determining step of multiple meteorological observation data streams includes: performing data extraction on an original meteorological observation report to obtain the multiple meteorological observation data streams.

[0016] Step 120: Respectively perform attribute knowledge mining on the multiple target meteorological observation data to obtain meteorological observation attribute knowledge encodings of the multiple target meteorological observation data.

[0017] Step 130: Respectively take the multiple time series marked observation information as data access mark information, and perform a read-write state mapping on the meteorological observation attribute knowledge encodings of the multiple target meteorological observation data in the data access mark information to obtain a read-write state description vector of the data access mark information.

[0018] Step 140: Use the read-write state description vectors of the multiple time series marked observation information to determine an access management suggestion vector of the initial meteorological observation data stream.

[0019] Step 150: Perform a read-write state mapping on the access management suggestion vectors of the multiple meteorological observation data streams in the target database to obtain a global read-write state label of the target database, and the global read-write state label of the target database is used to determine a data access strategy of the target database.

[0020] Meteorological observation data is of great significance in many fields such as meteorological research and meteorological forecasting, and the effective access to its data is crucial. The meteorological observation data access system involved in the embodiment of the present invention realizes efficient data access through steps 110 - step 150.

[0021] First, in step 110, the meteorological observation data access system operates on multiple meteorological observation data streams in the target database. These meteorological observation data streams are the basic data sources processed by the entire system. The meteorological observation data streams in the embodiments of the present invention may come from various sources. In an alternative embodiment, they are obtained by extracting data from the original meteorological observation reports. For example, the original meteorological observation reports may contain observation records of various meteorological elements such as temperature, humidity, air pressure, wind speed, and wind direction at different time points. After data extraction, the observation data of different types or different observation stations are separated into multiple meteorological observation data streams. Then, these meteorological observation data streams are used as the initial meteorological observation data streams, and a circular buffer structure is used to perform operations based on time sequence tags on them, thereby obtaining multiple time sequence tag observation information. This circular buffer structure is a core structure in the embodiments of the present invention and has unique advantages. For example, each time sequence tag observation information covers 100 target meteorological observation data, and these target meteorological observation data are the meteorological observation results in a specific time sequence. Through the time sequence tag operation, the position of each data in the entire meteorological observation time sequence can be determined, which provides a reference basis in the time dimension for subsequent data processing.

[0022] Next, entering step 120, the system respectively performs attribute knowledge mining on the multiple target meteorological observation data obtained previously. Meteorological observation data has rich attribute information. For example, temperature data may have seasonal attributes, regional attributes, etc., and humidity data may have associated attributes with precipitation probability, etc. Through attribute knowledge mining, meteorological observation attribute knowledge codes are generated for each target meteorological observation data. This coding process is similar to attaching tags with specific meanings to each data, making the connotation of the data richer and easier to understand. For example, for the high-temperature data in a specific region in summer, its meteorological observation attribute knowledge code may include information such as the climate characteristics of that region in summer and the association between high temperature and other meteorological elements.

[0023] In the subsequent step 130, the system uses multiple time-series marker observation information as data access marker information to perform a read-write state mapping for the meteorological observation attribute knowledge encoding of multiple target meteorological observation data. The read-write state mapping in the embodiments of the present invention is an operation that associates the attribute knowledge of data with read-write operations. Through this mapping, a read-write state description vector of the data access marker information is obtained. For example, if the meteorological observation attribute knowledge encoding of a target meteorological observation data indicates that the data is frequently updated recently and is of great significance to meteorological forecasting, then in the read-write state mapping, it may be marked as a high-frequency read-write state, and this state will have a corresponding representation in the read-write state description vector. This read-write state description vector can be understood as a comprehensive description of each time-series marker observation information in terms of read-write operations, and it covers the information of all target meteorological observation data in terms of read-write states.

[0024] In step 140, the system uses the read-write state description vectors of multiple time-series marker observation information to determine an access management recommendation vector for the initial meteorological observation data stream. Since each time-series marker observation information has its read-write state description vector, these vectors combined can reflect the overall read-write demand characteristics of the initial meteorological observation data stream. For example, if the read-write state description vectors of multiple time-series marker observation information all show that a certain type of meteorological observation data (such as wind speed data in a specific area) has a very high read-write frequency within a certain period, then in the access management recommendation vector of the initial meteorological observation data stream, it will be recommended to perform special access management on this type of data, such as allocating more cache space or using a more efficient read-write algorithm, etc.

[0025] Finally, in step 150, the system performs a read-write state mapping on the access management recommendation vectors of multiple meteorological observation data streams in the target database to obtain a global read-write state label for the target database. This global read-write state label is a comprehensive identifier for the entire target database in terms of data access. It can reflect the read-write demand situations of different meteorological observation data streams in the entire target database. For example, if there are multiple meteorological observation data streams in the target database from different meteorological observation stations, and the data of some stations is updated frequently and is crucial for meteorological analysis, while the data of other stations is relatively stable and has a lower read-write frequency, then this difference will be reflected in the global read-write state label. The importance of this global read-write state label lies in that it can be used to determine the data access strategy of the target database. Based on this label, an access strategy suitable for the entire target database can be formulated. For example, for data with a high read-write frequency, a faster storage medium or a more optimized read algorithm can be used, and for data with a low read-write frequency, a more economical storage method can be adopted, etc.

[0026] Throughout the process, the circular buffer structure plays a crucial role. It supports multiple independent meteorological observation data streams to share storage space in the same buffer, enabling continuous data writing and circular overwriting. For example, if the total capacity of the circular buffer is 1000 data units, when new data is continuously generated by the meteorological observation data stream, the new data can be written into the buffer sequentially according to the time sequence. When the buffer is full, the new data will circularly overwrite the earliest data, just like a circular storage structure. This structure ensures continuous data access and storage through the circular movement of read and write pointers and supports parallel access. For example, when performing the read-write state mapping of data access marker information, multiple time-sequence marker observation information can be operated on simultaneously, greatly improving the data access speed and efficiency.

[0027] Meanwhile, the embodiment of the present invention also has an overflow and underflow detection mechanism and a capacity management strategy for multi-channel data streams. During the process of accessing and storing meteorological observation data, since meteorological observation is continuous, the data volume may continuously increase. If not controlled, overflow may occur. For example, when the circular buffer receives a large amount of meteorological observation data in a short period, exceeding the buffer capacity, the overflow detection mechanism will be triggered, and the system will take corresponding measures, such as temporarily stopping the writing of some unimportant data streams or compressing and storing the data. Underflow may occur when the data reading speed is too fast and the writing speed cannot keep up. The underflow detection mechanism will also detect and take measures, such as prompting the data source to accelerate data generation speed or adjusting the reading strategy. The capacity management strategy for multi-channel data streams reasonably allocates the capacity of the circular buffer according to factors such as the importance and read-write frequency of different meteorological observation data streams. For example, for those meteorological observation data streams that are of great significance to weather forecasting and have a high read-write frequency, a larger buffer capacity can be allocated to ensure data integrity and system stability.

[0028] The meteorological observation data access system involved in the embodiment of the present invention is applicable to various scenarios, such as real-time systems, streaming media processing, and sensor networks. In a real-time system, it can timely process continuously updated meteorological observation data, providing fast and accurate data support for weather forecasting; in the streaming media processing scenario, it can efficiently process the continuous stream of meteorological observation data to ensure the orderly access of data; in the sensor network scenario, it can flexibly handle meteorological observation data streams generated by different sensors and can flexibly expand the number of channels according to the characteristics and requirements of the sensors, thereby enhancing the system processing ability. In summary, through a series of closely related operation steps and effective mechanisms, the meteorological observation data access system realizes the efficient, stable, and intelligent access management of meteorological observation data.

[0029] In some examples, the illustration of the knowledge encoding of meteorological observation attributes, the description vector of the read / write state, the access management suggestion vector, and the global read / write state label is as follows.

[0030] I. Knowledge Encoding of Meteorological Observation Attributes

[0031] Meteorological observation data contains various meteorological elements, and different elements have different attribute characteristics, which can be encoded through numerical feature vectors.

[0032] For example, for temperature observation data, a feature vector of a knowledge encoding of meteorological observation attributes is [1, 0.8, 0.2, 0.5, 0.3].

[0033] The first element "1" represents the observation type to which the temperature data belongs. Here, for example, "1" represents ground temperature observation (if there are multiple observation types, different numerical values can be used for differentiation. For example, 0 represents upper-air temperature observation, etc.).

[0034] The second element "0.8" represents the degree of correlation between the temperature data and the season. The higher the numerical value, the stronger the correlation. Here, 0.8 indicates that the temperature data is within a relatively typical temperature range in the current season, possibly close to the average temperature of the season or within the common fluctuation range.

[0035] The third element "0.2" represents the association between the temperature data and altitude. A lower numerical value indicates that the temperature is relatively less affected by altitude (if it is temperature observation in mountainous areas, this numerical value may be higher).

[0036] The fourth element "0.5" represents the relationship between the temperature data and the surrounding geographical environment (such as the ocean, forest, etc.). Here, 0.5 indicates a certain association. For example, the observation point is located on land near the ocean, and the ocean has a certain regulatory effect on temperature.

[0037] The fifth element "0.3" represents the fluctuation of the temperature data in historical data of the same period. 0.3 indicates relatively small fluctuations, which may imply that the current meteorological system is relatively stable or the climate in this area has a certain stability.

[0038] For humidity observation data, the feature vector of its knowledge encoding of meteorological observation attributes may be [0, 0.6, 0.3, 0.7, 0.4]. Among them, "0" represents the humidity observation type (distinguished from the temperature observation type), "0.6" represents the correlation between humidity and the season, "0.3" represents the association with altitude, "0.7" represents the relationship with the surrounding geographical environment (for example, the humidity may be higher in forest areas, and 0.7 in the embodiments of the present invention represents a strong association), and "0.4" represents the fluctuation in historical data of the same period.

[0039] II. Read / Write Status Description Vector

[0040] Taking the time-series marked observation information containing three meteorological elements of temperature, humidity, and air pressure as an example, for instance, its read / write status description vector is [0.6, 0.4, 0.8].

[0041] The first element "0.6" corresponds to the read / write status of temperature data. 0.6 indicates that the read / write frequency of temperature data is relatively high, probably because temperature data is a key element in meteorological analysis and forecasting, and is often read for analyzing weather trends, and new temperature observation data also needs to be written in a timely manner.

[0042] The second element "0.4" corresponds to the read / write status of humidity data. Compared with temperature data, the read / write frequency of humidity data is slightly lower, perhaps because in the current meteorological analysis process, the influence weight of humidity data on the results is slightly lower than that of temperature data, or the update frequency of humidity data itself is relatively low.

[0043] The third element "0.8" corresponds to the read / write status of air pressure data. 0.8 in the embodiments of the present invention indicates that the read / write frequency of air pressure data is high, because air pressure changes are very important for the analysis of the meteorological system. Whether it is short-term weather changes or long-term climate research, air pressure data is often read and written.

[0044] III. Access Management Suggestion Vector

[0045] For example, for a certain initial meteorological observation data stream (containing multiple meteorological elements), its access management suggestion vector is [0.3, 0.5, 0.2, 0.4].

[0046] The first element "0.3" is related to the storage priority of data. 0.3 indicates a relatively low storage priority, which may mean that some data in this data stream (such as some auxiliary meteorological observation data with little impact on the overall meteorological analysis) can be stored with appropriate delay or using a lower-performance storage medium when storage resources are scarce.

[0047] The second element "0.5" is related to the optimization strategy of data reading. 0.5 indicates that a certain degree of reading optimization is required. For example, a caching mechanism can be adopted to improve the reading speed, because there may be certain delays or efficiency issues when reading some data in this data stream, and optimization is needed to meet the real-time requirements of meteorological analysis.

[0048] The third element "0.2" represents the redundant storage strategy of data. 0.2 indicates a low redundant storage requirement, that is, there is no need to perform a large amount of redundant storage on the data in this data stream, because these data may be relatively easy to obtain or regenerate, or their importance is not sufficient to support high redundant storage.

[0049] The fourth element "0.4" represents the secure storage level of the data. 0.4 indicates that certain secure storage measures are required, such as data verification and encryption means, to ensure the integrity and security of the data, but the highest level of secure storage measures (such as advanced encryption and multiple backups, etc.) are not required.

[0050] IV. Global Read / Write Status Tags

[0051] For multiple meteorological observation data streams in the target database, the eigenvector of the global read / write status tag may be [0.5, 0.4, 0.6, 0.3].

[0052] The first element "0.5" represents the overall read / write frequency of the meteorological observation data in the entire database. 0.5 indicates that the read / write frequency is at a medium level, which may be because the database contains meteorological observation data streams with different read / write frequencies, and when combined, such an overall medium-level read / write frequency is formed.

[0053] The second element "0.4" represents the timeliness requirement of the data. 0.4 indicates that the data in the database as a whole has a certain timeliness requirement, but not a very strict real-time requirement. For example, some meteorological observation data streams may be updated hourly instead of minute-by-minute or second-by-second, so the overall timeliness requirement is at a medium level.

[0054] The third element "0.6" represents the integrity requirement of the data. 0.6 indicates a relatively high data integrity requirement because the accuracy of meteorological observation data is crucial for meteorological analysis and forecasting. Even though the read / write frequency of some data is not very high, the loss or damage of any data may affect the final meteorological analysis results, so there is a relatively high requirement for data integrity as a whole.

[0055] The fourth element "0.3" represents the concurrent access requirement of the data. 0.3 indicates a relatively low concurrent access requirement, which means that in most cases, the number of users or application programs accessing the database simultaneously is relatively small, and a highly concurrent read / write operation design is not required, but certain concurrent processing capabilities still need to be considered to handle occasional concurrent access situations.

[0056] In an exemplary embodiment, the step of using the multiple time-series marker observation information as data access marker information respectively, and performing a read-write state mapping on the meteorological observation attribute knowledge encoding of multiple target meteorological observation data in the data access marker information to obtain a read-write state description vector of the data access marker information includes: using the multiple time-series marker observation information as data access marker information respectively, and performing a read-write state mapping on the meteorological observation attribute knowledge encoding of multiple target meteorological observation data in the data access marker information to obtain a first read-write state mapping feature, where the first read-write state mapping feature includes multiple first read-write state mapping clusters; determining the read-write state description vector of the data access marker information based on the read-write state description vectors of the multiple first read-write state mapping clusters.

[0057] Based on this exemplary embodiment, the step of performing a read-write state mapping on the access management recommendation vectors of multiple meteorological observation data streams in the target database to obtain the global read-write state label of the target database includes: performing a read-write state mapping on the access management recommendation vectors of multiple meteorological observation data streams in the target database to obtain a second read-write state mapping feature, where the second read-write state mapping feature includes multiple second read-write state mapping clusters; determining the global read-write state label of the target database based on the read-write state description vectors of the multiple second read-write state mapping clusters.

[0058] In the meteorological observation data access system, the core inventive point involves the specific operation methods in steps 130 and 150.

[0059] For step 130, first, take multiple time-series marker observation information as data access marker information, and perform a read-write state mapping on the meteorological observation attribute knowledge encoding of multiple target meteorological observation data among them to obtain a first read-write state mapping feature. This first read-write state mapping feature contains multiple first read-write state mapping clusters. For example, there are three time-series marker observation information, and each time-series marker observation information contains five target meteorological observation data, and their meteorological observation attribute knowledge encoding is represented in the form of a feature vector. For the first target meteorological observation data in the first time-series marker observation information, its meteorological observation attribute knowledge encoding feature vector is [0.2, 0.5, 0.3, 0.1, 0.4]. After the read-write state mapping, it may be divided into a certain first read-write state mapping cluster, and this cluster may have specific read-write attributes, such as high-frequency writing, low-frequency reading, etc. For example, other data in this cluster also have similar read-write attribute associations, and there may be commonalities in aspects such as the source of the data and the relationship between the data and other meteorological elements, resulting in their similarity in read-write states. For different target meteorological observation data in each time-series marker observation information, the read-write state mapping is performed in this way to form multiple first read-write state mapping clusters.

[0060] Then, determine the read-write state description vector of the data access marker information based on the read-write state description vectors of these multiple first read-write state mapping clusters. The read-write state description vectors of these clusters can be regarded as a comprehensive description of the read-write states of the data within the clusters. For example, a first read-write state mapping cluster contains three target meteorological observation data, and their meteorological observation attribute knowledge encoding forms the read-write state description vector of this cluster after the read-write state mapping. For example, the read-write state description vector of this cluster is [0.6, 0.3, 0.4], where the first value 0.6 represents the overall write frequency weight of the data within the cluster, 0.3 represents the read frequency weight, and 0.4 represents the importance weight of the data in the entire meteorological observation data stream (this weight is related to the role of the data in meteorological analysis. For example, data that has a greater impact on the accuracy of meteorological forecasts has a higher weight). By comprehensively processing the read-write state description vectors of each cluster, such as weighted averaging or combining according to specific rules, the read-write state description vector of the data access marker information is finally determined.

[0061] For step 150, first, perform read-write state mapping on the access management suggestion vectors of multiple meteorological observation data streams in the target database to obtain a second read-write state mapping feature, which includes multiple second read-write state mapping clusters. For example, there are four meteorological observation data streams in the target database, and each meteorological observation data stream obtains its own access management suggestion vector through the previous operations. For example, the access management suggestion vector feature vector of the first meteorological observation data stream is [0.3, 0.5, 0.2, 0.4]. After read-write state mapping, it may be assigned to a certain second read-write state mapping cluster. The access management suggestion vectors of other meteorological observation data streams in this cluster may have similar read-write state mapping features, such as having commonalities in data storage strategies, read optimization requirements, etc. Perform such read-write state mapping on the access management suggestion vectors of each meteorological observation data stream to form multiple second read-write state mapping clusters.

[0062] Then, determine the global read-write state label of the target database based on the read-write state description vectors of multiple second read-write state mapping clusters. Each second read-write state mapping cluster has its own read-write state description vector, and these vectors reflect the overall read-write state characteristics of the meteorological observation data streams within the cluster. For example, the read-write state description vector of a second read-write state mapping cluster is [0.5, 0.4, 0.6, 0.3], where 0.5 represents the overall read-write frequency of the meteorological observation data streams within the cluster, 0.4 represents the requirement for data timeliness, 0.6 represents the requirement for data integrity, and 0.3 represents the concurrent access requirement for data. By comprehensively processing the read-write state description vectors of each second read-write state mapping cluster, such as performing operations like summing according to certain weights or taking the maximum value, minimum value, etc., finally determine the global read-write state label of the target database. The feature vector of this global read-write state label can accurately reflect the overall characteristics of the entire target database in terms of data access, such as it may be [0.4, 0.3, 0.5, 0.2], representing the overall read-write frequency, timeliness requirement, integrity requirement, and concurrent access requirement of the entire database respectively.

[0063] With such a design, first, by mapping the read-write status of meteorological observation attribute knowledge encoding, it is possible to more accurately determine the read-write status according to the characteristics of the data itself, avoiding the problem of unified processing that ignores the internal differences of the data. For example, in the process of determining the read-write status description vector, the read-write characteristics of data clustering are considered, enabling different types of data to be processed appropriately. Second, in determining the global read-write status label, comprehensive processing based on the read-write status description vectors of multiple clusters can fully reflect the read-write status requirements of the entire database. This helps to formulate more reasonable data access strategies. For example, more system resources can be allocated for optimization processing of data with high read-write frequencies, and stronger data protection measures can be taken for data with high integrity requirements. Third, this cluster-based processing method improves the scalability of the system. When new meteorological observation data streams are added or the data characteristics change, it is relatively easy to adjust the cluster division and mapping rules without the need for large-scale reconstruction of the entire system. In summary, this technical solution improves the efficiency, accuracy, and adaptability of the meteorological observation data access system, providing reliable technical support for the effective management and utilization of meteorological data.

[0064] Under the above further design ideas for steps 130 and 150, determining the read-write status description vector of the data access tag information based on the read-write status description vectors of the multiple first read-write status mapping clusters includes: fusing the vector elements of the key status mapping members of the multiple first read-write status mapping clusters to determine the read-write status description vector of the data access tag information. Determining the global read-write status label of the target database based on the read-write status description vectors of the multiple second read-write status mapping clusters includes: fusing the vector elements of the key status mapping members of the multiple second read-write status mapping clusters to determine the global read-write status label of the target database.

[0065] Among them, each second read-write status mapping feature includes local mapping features corresponding to multiple different second feature dimensions, and each local mapping feature includes multiple second read-write status mapping clusters; the second feature dimension is used to indicate the number of second read-write status mapping clusters in the corresponding second read-write status mapping feature. And the read-write status description vector of each first feature dimension of the target database includes multiple local read-write status vectors of the first feature dimension of the target database corresponding to multiple second feature dimensions respectively.

[0066] In the meteorological observation data access system, the above technical solution further elaborates on how to determine the read-write status description vector and the global read-write status label based on the vector elements of the key status mapping members of the clusters.

[0067] For the process of determining the read-write state description vector of data access tag information based on the read-write state mapping sub-clusters in step 130, the key lies in fusing the vector elements of the key state mapping members of multiple first read-write state mapping sub-clusters. Taking the actual situation of a certain meteorological observation data access as an example, for instance, there are four first read-write state mapping sub-clusters, and the vector elements of the key state mapping members of each sub-cluster have different numerical representations. For example, the vector elements of the key state mapping member of the first sub-cluster are [0.3, 0.5, 0.2], where 0.3 may represent the set weight related to data writing, 0.5 represents the read-related weight, and 0.2 represents the set importance weight of the data in the current sub-cluster; the vector elements of the second sub-cluster are [0.4, 0.3, 0.3]; the third sub-cluster is [0.2, 0.6, 0.2]; the fourth sub-cluster is [0.5, 0.2, 0.3]. When fusing these vector elements, a specific fusion rule may be adopted, such as weighted averaging according to the proportion of the data volume in each sub-cluster. If the data volume contained in the first sub-cluster accounts for 20% of the total data volume, the second sub-cluster accounts for 30%, the third sub-cluster accounts for 25%, and the fourth sub-cluster accounts for 25%, then the first element of the fused read-write state description vector is calculated as: (0.3×0.2 + 0.4×0.3 + 0.2×0.25 + 0.5×0.25), and the second and third elements are calculated in the same way. The read-write state description vector of the data access tag information determined in this way can comprehensively reflect the key read-write state characteristics of each sub-cluster, thus more accurately describing the read-write state of the entire data access tag information.

[0068] For the process of determining the global read-write status label of the target database based on the read-write status description vectors of multiple second read-write status mapping clusters in step 150, it is also to fuse the vector elements of the key status mapping members of multiple second read-write status mapping clusters. For example, there are three second read-write status mapping features in the target database, and each second read-write status mapping feature has a different second feature dimension, which means there are different numbers of second read-write status mapping clusters. For example, the first second read-write status mapping feature has two second read-write status mapping clusters, and the vector elements of its key status mapping members are [0.4, 0.3, 0.3] and [0.5, 0.2, 0.3] respectively; the second second read-write status mapping feature has three clusters, and the vector elements are [0.3, 0.4, 0.3], [0.4, 0.3, 0.3] and [0.3, 0.3, 0.4] respectively; the third second read-write status mapping feature has two clusters, and the vector elements are [0.2, 0.5, 0.3] and [0.3, 0.4, 0.3]. For each second read-write status mapping feature, first fuse the vector elements of the key status mapping members of its clusters according to the set weights of the clusters (such as the importance weights of the meteorological observation data streams within the clusters, etc.) to obtain the intermediate result vector of each second read-write status mapping feature. Then, according to the overall weights of these three second read-write status mapping features in the target database (such as determined according to the importance of different meteorological observation types), fuse these three intermediate result vectors again to finally determine the global read-write status label of the target database.

[0069] In this way, first of all, by fusing the vector elements of the key status mapping members of the clusters, whether it is in determining the read-write status description vector of the data access mark information or in determining the global read-write status label of the target database, the key information within the clusters can be fully utilized to avoid information loss and misjudgment. For example, the vector elements of different clusters can accurately reflect the read-write status characteristics of the data under different hierarchical structures after fusion. Secondly, when processing the clusters related to multiple feature dimensions, fusing according to certain rules can make the determination of the global read-write status label and the read-write status description vector more scientific and reasonable. This helps the system formulate more practical data access strategies based on these accurate status labels and vectors, improving the accuracy, efficiency and adaptability of the meteorological observation data access system, and ensuring the effective management and utilization of meteorological data.

[0070] Based on the above exemplary embodiments, the step of using the multiple time-series marker observation information as data access marker information respectively and performing a read-write state mapping on the meteorological observation attribute knowledge encoding of multiple target meteorological observation data in the data access marker information to obtain a first read-write state mapping feature includes: using the multiple time-series marker observation information as data access marker information respectively, and based on multiple different first feature dimensions, performing a read-write state mapping on the meteorological observation attribute knowledge encoding of multiple target meteorological observation data in the data access marker information to obtain first read-write state mapping features respectively corresponding to the multiple first feature dimensions, where the first feature dimension is used to indicate the number of first read-write state mapping clusters in the corresponding first read-write state mapping feature; the read-write state description vector of the data access marker information includes read-write state description vectors respectively corresponding to the multiple first feature dimensions of the data access marker information.

[0071] In step 140, the step of determining the access management recommendation vector of the initial meteorological observation data stream by using the read-write state description vectors of the multiple time-series marker observation information includes: fusing the read-write state description vectors corresponding to the same first feature dimension in the read-write state description vectors of the multiple time-series marker observation information to obtain access management recommendation vectors respectively corresponding to multiple first feature dimensions of the initial meteorological observation data stream.

[0072] In a further refined embodiment of step 150, the step of performing a read-write state mapping on the access management recommendation vectors of multiple meteorological observation data streams in the target database to obtain a second read-write state mapping feature includes: performing a read-write state mapping on the access management recommendation vectors corresponding to the same first feature dimension among the access management recommendation vectors of multiple meteorological observation data streams in the target database to obtain second read-write state mapping features respectively corresponding to multiple first feature dimensions; the global read-write state label of the target database includes read-write state description vectors respectively corresponding to the multiple first feature dimensions of the target database.

[0073] For the process of using multiple time-series tagged observation information as data access tag information and performing a read-write state mapping on the meteorological observation attribute knowledge encoding of the target meteorological observation data therein to obtain the first read-write state mapping feature, it is carried out based on multiple different first feature dimensions. For example, there are three different first feature dimensions, which are 3, 4, and 5 respectively. For the case where the first feature dimension is 3, when performing a read-write state mapping on the meteorological observation attribute knowledge encoding of the target meteorological observation data in the data access tag information, a first read-write state mapping feature corresponding to this first feature dimension will be obtained. The number of first read-write state mapping clusters included in this feature is 3. For example, in this mapping process, for a certain target meteorological observation data, its meteorological observation attribute knowledge encoding is represented by the feature vector [0.3, 0.5, 0.2], and after mapping, it is divided into a certain first read-write state mapping cluster, and the overall read-write state of this cluster is associated with the encoding feature of this data. Similarly, for the cases where the first feature dimensions are 4 and 5, the mapping will also be carried out according to their respective rules to obtain the corresponding first read-write state mapping features, and the number of first read-write state mapping clusters in each feature is 4 and 5 respectively. Moreover, the read-write state description vector of the data access tag information contains read-write state description vectors corresponding to multiple first feature dimensions respectively, which means that the read-write state descriptions under different first feature dimensions are independent of each other and jointly constitute the overall read-write state description.

[0074] Next, in step 140, when determining the access management recommendation vector of the initial meteorological observation data stream using the read-write state description vectors of multiple time-series tagged observation information, the read-write state description vectors corresponding to the same first feature dimension in the read-write state description vectors of multiple time-series tagged observation information are fused. For example, for the three first feature dimensions mentioned above, the read-write state description vectors under each first feature dimension are fused. For example, for the case where the first feature dimension is 3, there are three time-series tagged observation information, and their read-write state description vectors under this first feature dimension are [0.4, 0.3, 0.3], [0.3, 0.4, 0.3], and [0.3, 0.3, 0.4] respectively. In the fusion process, the fusion weights may be determined according to factors such as the importance of each element of the vector or the data volume. If the simple addition and then averaging method with equal weights is used, the first element of the fused access management recommendation vector is (0.4 + 0.3 + 0.3) / 3, and the second and third elements are calculated in the same way. In this way, the access management recommendation vector of the initial meteorological observation data stream under this first feature dimension is obtained. The same operation is carried out for other first feature dimensions, and finally, the access management recommendation vectors of the initial meteorological observation data stream corresponding to multiple first feature dimensions are obtained.

[0075] In a further refined embodiment of step 150, the process of obtaining the second read-write state mapping feature by performing read-write state mapping on the access management recommendation vectors of multiple meteorological observation data streams in the target database is as follows. For the access management recommendation vectors of multiple meteorological observation data streams in the target database, first perform read-write state mapping on the access management recommendation vectors corresponding to the same first feature dimension. For example, there are four meteorological observation data streams. For the case where the first feature dimension is 3, the access management recommendation vectors of these four meteorological observation data streams under this first feature dimension are [0.3, 0.5, 0.2], [0.4, 0.3, 0.3], [0.2, 0.6, 0.2], and [0.5, 0.2, 0.3] respectively. Through specific read-write state mapping rules, obtain the second read-write state mapping feature corresponding to this first feature dimension. Perform mapping operations on other first feature dimensions in the same way, so as to obtain the second read-write state mapping features corresponding to multiple first feature dimensions respectively. Moreover, the global read-write state label of the target database contains the read-write state description vectors corresponding to multiple first feature dimensions of the target database, which enables the global read-write state label to comprehensively reflect the read-write state of the target database under different first feature dimensions.

[0076] It can be seen that by performing operations such as read-write state mapping based on different first feature dimensions, the processing of meteorological observation data becomes more meticulous and accurate. For example, when determining the first read-write state mapping feature, different first feature dimensions can correspond to different meteorological observation data features or data grouping methods, so as to better mine the read-write state characteristics of the data. In terms of the fusion operation, whether it is fusing the read-write state description vectors of time-series marked observation information to obtain the access management recommendation vector, or performing read-write state mapping on the access management recommendation vectors of meteorological observation data streams, it can integrate the information of multiple related vectors and avoid one-sidedness. This helps to accurately determine the data access management strategy, improve the accuracy and reliability of the meteorological observation data access system, and can better adapt to the complexity and diversity of meteorological observation data, providing strong technical support for the effective management and utilization of meteorological data.

[0077] Under an optional design idea, based on steps 110 - 150, the method further includes: (1) performing feature matching on the global read-write state label of the target database based on the policy vector coordinate system to obtain the read-write execution policy vector of the target database; (2) processing the read-write execution policy vector of the target database to obtain the data access policy of the target database.

[0078] In a meteorological observation data access system, based on the previous step 110150, there are further optional design ideas. First, perform feature matching on the global read / write status label of the target database based on the policy vector coordinate system to obtain the read / write execution policy vector of the target database. The global read / write status label of the target database exists in the form of a feature vector, which contains comprehensive information about the entire target database in terms of data access. For example, the feature vector of the global read / write status label is [0.4, 0.3, 0.5, 0.2], where 0.4 may represent the overall read / write frequency of the entire database, 0.3 represents the requirement for data timeliness, 0.5 represents the requirement for data integrity, and 0.2 represents the concurrent access requirement for data.

[0079] The policy vector coordinate system is a reference system constructed to determine the read / write execution policy vector. In this coordinate system, different coordinate axes represent different factors related to data access policies, such as read / write frequency, timeliness, integrity, concurrent access, etc. For the feature vector [0.4, 0.3, 0.5, 0.2] of the global read / write status label, when performing feature matching in the policy vector coordinate system, the read / write execution policy vector will be determined according to the meaning represented by each coordinate axis and the preset matching rules. For example, in the policy vector coordinate system, for the read / write frequency coordinate axis, if the value is between 0 and 0.3, it represents a low-frequency read / write policy; between 0.3 and 0.7, it represents a medium-frequency read / write policy; between 0.7 and 1, it represents a high-frequency read / write policy. Then, the read / write frequency of 0.4 corresponds to the medium-frequency read / write policy. For aspects such as timeliness, integrity, and concurrent access requirements, they are also matched according to their respective matching rules. For example, for the timeliness aspect, 0.3 corresponds to a general timeliness policy; for integrity, 0.5 corresponds to a high-integrity policy; for the concurrent access requirement, 0.2 corresponds to a low-concurrent access policy. Combining these matching results, the read / write execution policy vector of the target database is obtained. This vector reflects the specific execution policy determined in the policy vector coordinate system based on the global read / write status label. For example, the read / write execution policy vector may be expressed as [medium-frequency read / write policy, general timeliness policy, high-integrity policy, low-concurrent access policy].

[0080] Next, the read-write execution policy vector of the target database is processed to obtain the data access policy of the target database. For the obtained read-write execution policy vector [intermediate-frequency read-write policy, general timeliness policy, high integrity policy, low concurrent access policy], corresponding processing is performed according to different policies to determine the data access policy. Under the intermediate-frequency read-write policy, a moderately optimized read-write algorithm may be adopted, which is neither as highly speed-optimized as the high-frequency read-write policy nor as lenient as the low-frequency read-write policy. For the general timeliness policy, the update and storage cycle of data can be set according to regular time intervals, without particularly frequent or lenient time arrangements. The high integrity policy requires strict data verification, backup and other measures during data access to ensure the accuracy and integrity of data. For example, data verification is performed after each data write, and data is regularly backed up to multiple storage locations. The low concurrent access policy means that there is no need to make excessive preparations for a large number of concurrent accesses in the system resource allocation, such as not requiring the configuration of too many concurrent access channels or complex concurrent control mechanisms. By processing each policy in the read-write execution policy vector, the data access policy of the target database is finally determined. This data access policy is a comprehensive and integrated solution that covers specific operation methods in multiple aspects such as data read-write frequency, timeliness, integrity, and concurrent access.

[0081] Thus, by obtaining the read-write execution policy vector through feature matching based on the policy vector coordinate system, the complex information contained in the global read-write state label can be transformed into specific execution policies in a targeted and regularized manner. For example, numerical features are transformed into actual policy descriptions through clear matching rules, avoiding ambiguity and uncertainty. Then, by processing the read-write execution policy vector to obtain the data access policy, all aspects of data access can be comprehensively considered, making the data access policy more scientific and reasonable. This step-by-step derivation process from the global read-write state label to the read-write execution policy vector and then to the data access policy improves the accuracy, reliability, and adaptability of the meteorological observation data access system, ensuring that meteorological observation data can be effectively managed and operated according to its own characteristics and requirements during the access process, and guaranteeing the quality and usability of meteorological data.

[0082] In a preferred technical solution, the processing of the read-write execution policy vector of the target database to obtain the data access policy of the target database includes: based on the data access requirements indicated by the channel capacity regulation vector, processing the read-write execution policy vector of the target database through a feedforward neural network based on conditional probability to determine the data access policy of the target database.

[0083] In the technical solution of the meteorological observation data access system, for the process of processing the read / write execution policy vector of the target database into the data access policy of the target database, a method based on the data access requirements indicated by the channel capacity regulation vector and with the help of a feedforward neural network based on conditional probability is adopted.

[0084] The read / write execution policy vector of the target database contains various policy information about database read / write operations. For example, the read / write execution policy vector may be expressed as [medium-frequency read / write policy, general timeliness policy, high integrity policy, low concurrent access policy]. Each policy element in the embodiments of the present invention reflects specific requirements or attributes of the target database in terms of data access.

[0085] The channel capacity regulation vector, on the other hand, indicates the data access requirements from another perspective. It may contain various information related to the data channel capacity, such as the data flow limit of each channel, the data storage capacity limit, etc. For example, the channel capacity regulation vector is [0.6, 0.4, 0.3], where 0.6 may represent the capacity utilization limit of a certain main data channel, 0.4 represents the capacity limit of another related channel, and 0.3 represents the additional capacity limit under specific conditions or the capacity limit of the standby channel, etc. This vector provides constraint and requirement information about the channel capacity for the determination of the data access policy.

[0086] The feedforward neural network based on conditional probability plays a key role in this process. The feedforward neural network is a neural network structure with an input layer, a hidden layer, and an output layer. In this context, the input layer receives the read / write execution policy vector and the channel capacity regulation vector of the target database. For example, the read / write execution policy vector [medium-frequency read / write policy, general timeliness policy, high integrity policy, low concurrent access policy] and the channel capacity regulation vector [0.6, 0.4, 0.3] are used as inputs. The neurons in the network are connected by weights, and these weights are determined during the training process and are related to the conditional probability.

[0087] The conditional probability is reflected here as the probability of adopting a set data access policy under different read / write execution policies and channel capacity conditions. For example, under the medium-frequency read / write policy, general timeliness policy, high integrity policy, low concurrent access policy, and the capacity limits represented by the channel capacity regulation vector, there is a certain probability for a set data storage method (such as using a set storage medium, storage format, etc.). This probability is obtained through the analysis and learning of a large number of meteorological observation data access situations.

[0088] In the hidden layer of the network, neurons process the input information. They transform and combine the input vector information according to pre-determined weights and activation functions. During this process, different neurons may focus on the relationships between different strategies and capacity information, thereby uncovering the features hidden behind this information. For example, some neurons may focus on the relationship between high-integrity strategies and channel capacity limitations to determine how to ensure data integrity while meeting channel capacity requirements in such a scenario.

[0089] Finally, the data access strategy for the target database is obtained at the output layer. This data access strategy is a result obtained by comprehensively considering the read / write execution strategy vector and the channel capacity regulation vector and processing through a feedforward neural network based on conditional probability. For example, the generated data access strategy may include using a specific storage medium (such as a high-speed hard disk to meet certain speed requirements under medium-frequency read / write strategies), setting a specific data update cycle (determined according to general timeliness strategies), adopting a strict data verification and backup mechanism (determined by high-integrity strategies), and configuring an appropriate concurrent access processing capacity (based on low-concurrency access strategies), while also meeting the channel capacity limitations indicated by the channel capacity regulation vector, such as reasonably allocating the storage amount of data on different channels to avoid exceeding the channel capacity.

[0090] In this way, by introducing the channel capacity regulation vector, it is possible to fully consider the channel capacity limitations during the data access process, avoiding problems such as data storage failures or performance degradation caused by ignoring capacity issues. The use of a feedforward neural network based on conditional probability makes the determination of the data access strategy more scientific and accurate. It can utilize a large amount of historical data and empirical knowledge to reflect the complex relationships between different factors through conditional probability, thereby providing a more reliable basis for the data access strategy. This method of comprehensively considering multiple factors and leveraging advanced neural network technology improves the adaptability, reliability, and efficiency of the meteorological observation data access system, ensuring that appropriate data access strategies can be formulated under different read / write execution strategies and channel capacity requirements, and guaranteeing the effective management and utilization of meteorological observation data.

[0091] In an alternative preferred technical concept, based on the data access requirements indicated by the channel capacity regulation vector, the read-write execution policy vector of the target database is processed by a feedforward neural network based on conditional probability to determine the data access policy of the target database, including: combining the read-write execution policy vector of the target database and the overflow detection mechanism execution vector of the target database through a feedforward neural network based on conditional probability to obtain the multi-channel access interaction vector of the target database; based on the data access requirements indicated by the channel capacity regulation vector, processing the multi-channel access interaction vector of the target database to determine the data access policy of the target database.

[0092] In the technical solution of the meteorological observation data access system, for the key link of determining the data access policy of the target database, there is a more detailed operation process under an alternative preferred technical concept.

[0093] First, it involves combining the read-write execution policy vector of the target database and the overflow detection mechanism execution vector of the target database through a feedforward neural network based on conditional probability to obtain the multi-channel access interaction vector of the target database.

[0094] The read-write execution policy vector of the target database contains important information related to database read-write operations. For example, the read-write execution policy vector may be [medium-frequency read-write policy, relatively loose timeliness policy, high integrity policy, low concurrent access policy]. The "medium-frequency read-write policy" in the embodiments of the present invention indicates that the read-write frequency of the data is at a medium level, meaning that the frequency of read-write operations is neither very high nor very low within a certain time range; the "relatively loose timeliness policy" means that the requirement for timeliness of the data is not particularly strict, and it may allow updates or reads at relatively long time intervals; the "high integrity policy" reflects that the integrity of the data is crucial, and various measures need to be taken during data access to ensure that the data is not damaged or lost; the "low concurrent access policy" reflects that the system does not need to handle a large number of concurrent read-write requests, and there is no need to overly consider high-concurrency situations in resource allocation and architecture design.

[0095] The overflow detection mechanism execution vector of the target database is also an important component. For example, the overflow detection mechanism execution vector is [0.3, 0.6, 0.1]. Here, 0.3 may represent the overflow detection sensitivity in a specific channel or data block. The lower the value, the lower the sensitivity, meaning that for a small amount of data overflow, the strict processing mechanism may not be immediately triggered. 0.6 represents a relatively high overflow detection sensitivity in another situation (such as for a specific type of data or a specific storage area), indicating that once the data volume approaches or may exceed the set capacity limit, the overflow detection mechanism will be triggered relatively quickly and corresponding measures will be taken. 0.1 represents a very low overflow detection sensitivity in other special situations (such as a backup channel or a temporary storage area), perhaps because these areas have a certain elasticity in accommodating data volume or the temporariness of data in these areas means that a small amount of overflow will not cause serious consequences.

[0096] The feedforward neural network based on conditional probability plays a core role in this combination process. The feedforward neural network consists of an input layer, a hidden layer, and an output layer. In this context, the input layer receives the read-write execution strategy vector and the overflow detection mechanism execution vector. The neurons in the neural network are connected by weights, and these weights are related to the conditional probability. The conditional probability reflects the probabilities of various combination situations under different read-write execution strategies and overflow detection mechanism execution states. For example, under the combination of the medium-frequency read-write strategy and an overflow detection sensitivity of 0.3, there is a specific probability indicating the likelihood of this combination occurring in the actual data access process. This probability is obtained through the analysis and learning of the access situations of a large amount of meteorological observation data.

[0097] In the hidden layer of the neural network, the neurons process the input vector information. The neurons in the hidden layer will transform and combine the read-write execution strategy vector and the overflow detection mechanism execution vector according to the pre-determined weights and activation functions. For example, some neurons may focus on the relationship between the high-integrity strategy and a relatively high overflow detection sensitivity (such as 0.6), because the high-integrity strategy may require a more strict overflow detection mechanism to ensure that data will not be lost or damaged due to overflow under any circumstances. Different neurons will discover the potential relationships between different strategy elements and overflow detection mechanism elements, thus integrating this information.

[0098] Through the processing of this feedforward neural network, a multi-channel access interaction vector of the target database is obtained at the output layer. This multi-channel access interaction vector synthesizes the information of the read / write execution policy vector and the overflow detection mechanism execution vector, and it reflects the interaction relationships and requirements of data access among channels considering the read / write policy and the overflow detection mechanism. For example, the multi-channel access interaction vector may be expressed as [channel 1 interaction relationship, channel 2 interaction relationship, channel 3 interaction relationship], where the interaction relationship of each channel may include comprehensive information in multiple aspects such as read / write frequency, integrity requirements, and overflow detection sensitivity.

[0099] Next, based on the data access requirements indicated by the channel capacity regulation vector, the multi-channel access interaction vector of the target database is processed to determine the data access strategy of the target database.

[0100] The channel capacity regulation vector contains key information regarding the data channel capacity. For example, the channel capacity regulation vector is [0.5, 0.4, 0.3], where 0.5 may represent the upper limit of the capacity utilization rate of the main data channel, that is, this channel can use at most 50% of its capacity during data access to ensure the stable operation of the system and the efficient transmission of data; 0.4 represents the capacity limit of another auxiliary channel, which may be mainly used to store some less frequently accessed data or as a data cache area; 0.3 represents the capacity limit of a backup channel or a specific function channel (such as a channel dedicated to data verification or temporary data storage).

[0101] When processing the multi-channel access interaction vector, the constraints brought by the channel capacity regulation vector need to be fully considered. For example, if the interaction relationship of a certain channel in the multi-channel access interaction vector indicates that this channel requires a high read / write frequency (such as the high-frequency read / write requirement in the interaction relationship of channel 1), but the upper limit of the capacity utilization rate of this channel in the channel capacity regulation vector is 0.5, then when determining the data access strategy, it is necessary to adjust the read / write operation mode of this channel. It may adopt data compression technology to reduce the amount of data read / written each time, or adjust the read / write time interval to avoid exceeding the channel capacity. For a channel with high integrity requirements (such as the high integrity requirement reflected in the interaction relationship of channel 2 in the multi-channel access interaction vector), if the capacity limit of this channel in the channel capacity regulation vector is 0.4, it may be necessary to preferentially allocate more system resources for data verification and backup on the premise of meeting the capacity limit to ensure data integrity. For a channel sensitive to the overflow detection mechanism (such as the high overflow detection sensitivity shown in the interaction relationship of channel 3 in the multi-channel access interaction vector), if the capacity limit of this channel in the channel capacity regulation vector is 0.3, then when approaching this capacity limit, it is necessary to more strictly execute the overflow detection and processing mechanism, and may perform data migration or cleaning operations in advance.

[0102] Through such a processing procedure, the data access strategy of the target database is finally determined. This data access strategy is a comprehensive solution that covers various aspects such as the read-write frequencies of each channel, integrity requirements, overflow detection mechanisms, and channel capacity limitations. For example, the data access strategy may include: for Channel 1, medium-frequency read-write operations are adopted, data is updated once an hour, moderate data compression technology is used, and a capacity warning is issued when the capacity utilization rate reaches 40%; for Channel 2, it is set to low-frequency read-write operations, data verification is performed once a day, a dual-backup mechanism is adopted to ensure data integrity when the channel capacity utilization rate does not exceed 30%; for Channel 3, due to its high overflow detection sensitivity and low capacity limitation, it is used as a temporary data storage channel, only storing temporary data generated in the most recent hour, and automatically clearing the earliest data when the capacity utilization rate reaches 20%.

[0103] Based on the above preferred technical ideas, first, by combining the read-write execution policy vector and the overflow detection mechanism execution vector into a multi-channel access interaction vector, various factors and their interrelationships in the data access process, such as read-write operations, integrity requirements, and overflow detection, can be comprehensively considered. This comprehensive consideration avoids data access problems caused by only focusing on a single factor. For example, the coordinated consideration between the high-integrity policy and the high overflow detection sensitivity can ensure that in cases where data integrity is crucial, data corruption or loss caused by overflow can be detected and avoided in a timely manner. Second, by processing the multi-channel access interaction vector based on the channel capacity regulation vector to determine the data access strategy, the limitations of the channel capacity can be fully considered. This makes the data access strategy more in line with the actual system architecture and hardware conditions, avoiding problems such as performance degradation and data loss caused by the mismatch between data access requirements and channel capacity. For example, adjusting the read-write frequency and data processing method according to the capacity limitations of different channels can maximize the utilization of channel capacity on the premise of ensuring normal data access, improving the overall efficiency of the system. Finally, the entire technical solution performs vector combination and processing through a feedforward neural network based on conditional probability, leveraging the neural network's ability to learn and process complex relationships. The conditional probability obtained through learning a large amount of data can more accurately reflect the internal connections between different factors, thereby providing a more scientific and reasonable basis for determining the data access strategy, improving the accuracy, reliability, and adaptability of the meteorological observation data access system, and ensuring the effective management and utilization of meteorological observation data.

[0104] Based on the above preferred technical idea, the feature matching of the global read-write status tag of the target database based on the policy vector coordinate system to obtain the read-write execution policy vector of the target database includes: performing feature matching of the global read-write status tag of the target database based on the policy vector coordinate system through a deep learning network to obtain the read-write execution policy vector of the target database.

[0105] Among them, the method further includes: debugging the deep learning network based on debugging examples including a meteorological observation debugging vector, an overflow detection mechanism debugging vector, a channel capacity regulation training vector, and a data access annotation policy.

[0106] It can be understood that the deep learning network is used to perform feature matching of the meteorological observation debugging vector based on the policy vector coordinate system to obtain a read-write debugging policy vector. The feedforward neural network based on conditional probability is used to combine the read-write debugging policy vector and the overflow detection mechanism debugging vector to obtain a multi-channel access debugging vector; based on the data access requirements indicated by the channel capacity regulation training vector, the multi-channel access debugging vector is processed to determine a data access debugging policy, and the network weights of the deep learning network are optimized according to the data access debugging policy and the data access annotation policy.

[0107] Under the technical solution framework of the meteorological observation data access system, a deep learning network is used as a tool for the process of performing feature matching of the global read-write status tag of the target database based on the policy vector coordinate system to obtain the read-write execution policy vector.

[0108] The global read-write status label of the target database exists in the form of a specific feature vector, which contains comprehensive information about the overall read-write status of the target database. For example, the feature vector of the global read-write status label is [0.4, 0.3, 0.5, 0.2], where 0.4 may represent the read-write frequency weight of the overall database, 0.3 represents the data timeliness weight, 0.5 represents the data integrity weight, and 0.2 represents the data concurrent access weight. The policy vector coordinate system is a reference system constructed for feature matching. Different coordinate axes represent various factors related to data access strategies, such as read-write frequency, timeliness, integrity, concurrent access, etc. In this process, the deep learning network matches the features of the global read-write status label in the policy vector coordinate system according to its complex structure and algorithm. Through the calculation and processing of multiple hidden layers and neurons in the deep learning network, each element in the global read-write status label is associated and matched with each factor in the policy vector coordinate system, and finally the read-write execution policy vector of the target database is obtained. This read-write execution policy vector reflects the specific execution policy determined according to the matching result of the global read-write status label in the policy vector coordinate system. For example, the read-write execution policy vector may be expressed as [medium-frequency read-write policy, general timeliness policy, high-integrity policy, low-concurrent access policy].

[0109] On this basis, the method also includes a process of debugging the deep learning network. This debugging process is based on a variety of debugging examples, including meteorological observation debugging vectors, over-inspection mechanism debugging vectors, channel capacity regulation training vectors, and data access annotation strategies.

[0110] The meteorological observation debugging vector is a vector related to meteorological observation data, which contains relevant information about the meteorological observation data used to debug the deep learning network. For example, the meteorological observation debugging vector is [0.3, 0.5, 0.2], where 0.3 may represent the weight related to the set meteorological observation data type, 0.5 represents the weight of the set features of the data (such as data stability or volatility, etc.), and 0.2 represents the weight related to the data source or acquisition device. The deep learning network first performs feature matching on the meteorological observation debugging vector based on the policy vector coordinate system to obtain the read-write debugging policy vector. This process is similar to the feature matching of the global read-write status label of the target database. The deep learning network matches the information in the meteorological observation debugging vector with the factors in the policy vector coordinate system to obtain the read-write debugging policy vector. For example, the read-write debugging policy vector may be expressed as [low-frequency read-write policy, loose timeliness policy, medium integrity policy, low-concurrent access policy].

[0111] The overflow detection mechanism debugging vector is also an important part of the debugging process. For example, the overflow detection mechanism debugging vector is [0.2, 0.6, 0.2], where 0.2 represents the initial value of the overflow detection sensitivity in a specific channel or data block, 0.6 represents a relatively high value of the overflow detection sensitivity in another case (such as a specific type of data or a specific storage area), and 0.2 represents the value of the overflow detection sensitivity in other special cases (such as spare channels or temporary storage areas). The feedforward neural network based on conditional probability plays a role in this process. It combines the previously obtained read-write debugging strategy vector and the overflow detection mechanism debugging vector to obtain a multi-channel access debugging vector. The neurons in the feedforward neural network are connected by weights, which are related to the conditional probabilities. The conditional probability reflects the probability of various combinations under different read-write debugging strategies and overflow detection mechanism debugging states. For example, under the combination of low-frequency read-write strategy and overflow detection sensitivity of 0.6, there is a specific probability that represents the possibility of this combination in the actual data access process. The input vector is processed through the hidden layer of the feedforward neural network to explore the potential relationship between the read-write debugging strategy vector and the overflow detection mechanism debugging vector, and finally a multi-channel access debugging vector is obtained at the output layer. For example, the multi-channel access debugging vector may be expressed as [channel 1 debugging relationship, channel 2 debugging relationship, channel 3 debugging relationship], where the debugging relationship of each channel contains comprehensive information such as read-write frequency, integrity requirements, overflow detection sensitivity, etc.

[0112] The channel capacity regulation training vector contains key information for training regarding the data channel capacity. For example, the channel capacity regulation training vector is [0.4, 0.3, 0.3], where 0.4 may represent the training value of the upper limit of the capacity utilization rate of the main data channel, 0.3 represents the training value of the capacity limit of another auxiliary channel, and 0.3 represents the training value of the capacity limit of a spare channel or a specific function channel (such as a channel dedicated to data verification or temporary data storage). Based on the data access requirements indicated by the channel capacity regulation training vector, the multi-channel access debugging vector is processed to determine the data access debugging strategy. During the processing, the constraints brought by the channel capacity regulation training vector need to be fully considered. For example, if a channel debugging relationship in the multi-channel access debugging vector indicates that the channel requires a high read / write frequency (such as the high-frequency read / write requirement in the channel 1 debugging relationship), but the upper limit of the capacity utilization rate of this channel in the channel capacity regulation training vector is 0.4, then when determining the data access debugging strategy, the read / write operation mode of this channel needs to be adjusted. It may adopt data compression technology to reduce the amount of data read / written each time, or adjust the read / write time interval to avoid exceeding the channel capacity. For channels with high integrity requirements (such as the high integrity requirement reflected in the channel 2 debugging relationship in the multi-channel access debugging vector), if the capacity limit of this channel in the channel capacity regulation training vector is 0.3, more system resources may need to be preferentially allocated for data verification and backup under the premise of meeting the capacity limit to ensure data integrity. For channels sensitive to the overflow detection mechanism (such as the high overflow detection sensitivity shown in the channel 3 debugging relationship in the multi-channel access debugging vector), if the capacity limit of this channel in the channel capacity regulation training vector is 0.3, then when approaching this capacity limit, the overflow detection and handling mechanism needs to be executed more strictly, and data migration or cleaning operations may be carried out in advance.

[0113] The data access annotation strategy is a pre-set standard used to measure the correctness of the data access debugging strategy. The network weights of the deep learning network are optimized based on the data access debugging strategy and the data access annotation strategy. During this process, the weights in the deep learning network are adjusted by comparing the differences between the data access debugging strategy and the data access annotation strategy. If there is a large difference between the data access debugging strategy and the data access annotation strategy, it indicates that the current weights of the deep learning network may not be appropriate and need to be adjusted. For example, if the read-write frequency in the data access debugging strategy differs significantly from the read-write frequency in the data access annotation strategy, then the deep learning network will adjust the neuron connection weights related to the read-write frequency so that more results closer to the data access annotation strategy can be obtained in subsequent processing. By continuously debugging with debugging examples, the deep learning network can continuously optimize its network weights, thereby improving the accuracy of feature matching of the global read-write state label of the target database based on the policy vector coordinate system, and then obtaining a more reasonable read-write execution policy vector.

[0114] It can be seen that, first, the read-write execution policy vector is obtained through feature matching based on the policy vector coordinate system by the deep learning network, making use of the powerful feature learning and pattern matching capabilities of the deep learning network. It can accurately extract the relationships with various factors in the policy vector coordinate system from the complex information of the global read-write state label, thus obtaining a read-write execution policy vector that better meets the actual needs. For example, the deep learning network can accurately match the appropriate read-write execution policy according to the integrity weight and timeliness weight in the global read-write state label, avoiding the limitations that may exist in manually setting rules. Second, debugging the deep learning network based on debugging examples including meteorological observation debugging vectors, overflow detection mechanism debugging vectors, channel capacity regulation training vectors, and data access annotation strategies can enable the deep learning network to continuously optimize its network weights. This helps to improve the accuracy and reliability of the entire data access system. For example, by comparing the data access debugging strategy and the data access annotation strategy to adjust the network weights, it can ensure that the deep learning network can obtain more accurate results under different data conditions and requirements. Moreover, the interaction operations between the various vectors involved in the debugging process, such as combining the read-write debugging policy vector and the overflow detection mechanism debugging vector by a feedforward neural network based on conditional probability, and processing the multi-channel access debugging vector based on the channel capacity regulation training vector, fully consider various factors in the data access process, such as the overflow detection mechanism, channel capacity, etc. This way of comprehensively considering various factors can make the data access strategy more perfect and reasonable, improve the adaptability and effectiveness of the meteorological observation data access system, and ensure the efficient management and accurate utilization of meteorological observation data.

[0115] In a technical concept, the operation of taking multiple meteorological observation data streams in the target database as initial meteorological observation data streams respectively and performing a time series marking operation based on a circular buffer structure on the initial meteorological observation data streams to obtain multiple time series marked observation information in step 110 includes: taking multiple meteorological observation data streams in the target database as initial meteorological observation data streams respectively, and performing grouped marking based on time series nodes on the initial meteorological observation data streams to obtain meteorological observation marked groups; dividing the meteorological observation marked groups into multiple time series marked observation information according to the observation priority.

[0116] In the technical solution of the meteorological observation data access system, the operation in step 110 is the starting part of the entire data processing flow, which includes a time series marking operation based on a circular buffer structure on the initial meteorological observation data streams.

[0117] First, take multiple meteorological observation data streams in the target database as initial meteorological observation data streams respectively, and perform grouped marking based on time series nodes on them to obtain meteorological observation marked groups. The meteorological observation data streams contain rich meteorological observation data, and these data have certain laws and characteristics in the time series. For example, a certain meteorological observation data stream may contain the observation data of meteorological elements such as temperature, humidity, and air pressure in a certain area for consecutive years. The time series nodes can be set at a certain time interval, such as in hours, and the observation data for each hour is used as a time series node. By performing grouped marking on these initial meteorological observation data streams according to the time series nodes, meteorological observation marked groups can be obtained. For example, a meteorological observation marked group contains the meteorological observation data with hours as time series nodes within a certain day, and these data cover the meteorological element information at different times within that day.

[0118] Then, group the meteorological observation marks and divide them into multiple time-series mark observation information according to the observation priority. The observation priority is determined according to the importance of meteorological elements in applications such as meteorological analysis and forecasting. For example, in meteorological forecasting, barometric data may have a relatively high observation priority because the change in barometric pressure has a crucial indicative effect on the evolution of weather systems; while some auxiliary meteorological elements, such as the local wind speed change at certain specific locations (when the impact on the overall weather conditions is relatively small), may have a relatively low observation priority. For a group of meteorological observation marks containing multiple meteorological elements, it is divided according to the observation priority. For example, the data feature vector in a group of meteorological observation marks is [0.6, 0.3, 0.1], where 0.6 represents the weight related to barometric data (reflecting its relative importance in the group), 0.3 represents the weight related to temperature data, and 0.1 represents the weight related to local wind speed data. According to such a weight relationship, this group of meteorological observation marks can be divided into multiple time-series mark observation information. For example, the part closely related to barometric data is taken as a time-series mark observation information with a relatively high priority. Because the weight of barometric data is relatively high, this part of information may be first concerned and processed in subsequent meteorological analysis; while the part related to local wind speed data is taken as a time-series mark observation information with a relatively low priority.

[0119] This way of dividing the meteorological observation mark group according to the observation priority to obtain the time-series mark observation information is of great significance in the entire meteorological observation data access system. It enables subsequent data processing to perform targeted operations according to the importance of different data. For example, in terms of data storage, more high-quality storage resources, such as high-speed storage media or larger storage spaces, can be allocated for the time-series mark observation information with high priority to ensure the fast storage and reliable preservation of these key data; while for the time-series mark observation information with low priority, a relatively economical storage method can be adopted. During the data reading and analysis process, the time-series mark observation information with high priority can also be first concerned and processed to improve the efficiency of meteorological analysis and forecasting. At the same time, this operation method based on the circular buffer structure lays a foundation for the subsequent continuous writing, circular overwriting of data, and efficient management of multi-channel data. For example, the circular buffer structure can enable multiple independent meteorological observation data streams to share the storage space in the same buffer. When processing multiple time-series mark observation information, through reasonable movement of the read and write pointers, continuous access to data can be ensured, and parallel access is supported, thus greatly improving the data access speed and efficiency.

[0120] In addition, the data partitioning method in this technical solution helps with the classified management and orderly processing of data. Different management strategies can be adopted for the observation information of time-series tags with different priorities. For example, the observation information of time-series tags with high priority may require more frequent integrity checks and backups to prevent serious impacts on meteorological analysis caused by data loss or damage; while the observation information of time-series tags with low priority can appropriately reduce the inspection frequency to reduce the consumption of system resources. In the data processing flow, different observation information of time-series tags can also be processed in the order of priority, making the entire data processing process more orderly and efficient.

[0121] Based on the above technical solution, firstly, through the grouping tags based on time-series nodes and the partitioning according to the observation priority, the inherent value differences of meteorological observation data can be better reflected. For example, by dividing the priority according to the importance of meteorological elements in meteorological analysis, key data can be focused on and processed, improving the pertinence and effectiveness of meteorological observation data processing. Secondly, in terms of data storage and management, allocating resources according to priority can improve resource utilization efficiency. For example, providing high-quality storage resources for high-priority data can not only ensure the security and availability of important data but also optimize the overall allocation of storage resources. Moreover, this partitioning method is conducive to improving the efficiency and accuracy of data processing. In the process of meteorological analysis, processing data in the order of priority can first process key data, reducing unnecessary calculations and processing processes, and thus obtaining accurate meteorological analysis results more quickly. At the same time, it provides an orderly and efficient data management mode for the entire meteorological observation data access system, ensuring the effective utilization and long-term management of meteorological observation data.

[0122] In an independent embodiment, the obtaining of the meteorological observation attribute knowledge encoding of the multiple target meteorological observation data by respectively performing attribute knowledge mining on the multiple target meteorological observation data described in step 120 includes:

[0123] (1) Input each target meteorological observation data into the spatio-temporal feature mining layer in the MetNet model to obtain the first spatio-temporal attribute vector and the second spatio-temporal attribute vector of each target meteorological observation data generated by the spatio-temporal feature mining layer, where the spatio-temporal feature mining layer includes multiple interconnected residual kernels, the first spatio-temporal attribute vector is the spatio-temporal attribute vector generated by the non-terminal residual kernels among the interconnected multiple residual kernels, and the second spatio-temporal attribute vector is the spatio-temporal attribute vector generated by the terminal residual kernel among the interconnected multiple residual kernels;

[0124] (2) Input the second spatio-temporal attribute vector into the recursive layer in the MetNet model to obtain the target observation attribute linear feature generated by the recursive layer;

[0125] (3) Input the first spatio-temporal attribute vector, the second spatio-temporal attribute vector, the third spatio-temporal attribute vector, and the linear features of the target observation attributes into the fully connected layer in the MetNet model to obtain the meteorological observation attribute knowledge encoding of each target meteorological observation data generated by the fully connected layer, where the third spatio-temporal attribute vector is a spatio-temporal attribute vector generated by the residual kernel in the recurrent layer according to the augmented attribute knowledge encoding, and the augmented attribute knowledge encoding is obtained by feature augmentation of the second spatio-temporal attribute vector.

[0126] In the meteorological observation data access system, the attribute knowledge mining operation in step 120 is of crucial significance for understanding and processing meteorological observation data.

[0127] First, input each target meteorological observation data into the spatio-temporal feature mining layer in the MetNet model. The MetNet model is a model specifically designed for processing meteorological data, and its spatio-temporal feature mining layer contains multiple interconnected residual kernels. These residual kernels play an important role in mining the spatio-temporal features of target meteorological observation data. For example, when a target meteorological observation data is input, in the spatio-temporal feature mining layer, multiple residual kernels process the data step by step. Among them, the non-terminal residual kernels generate the first spatio-temporal attribute vector. For example, the first spatio-temporal attribute vector obtained after processing a certain target meteorological observation data by a non-terminal residual kernel is [0.3, 0.5, 0.2]. The numerical values in the embodiments of the present invention can represent different spatio-temporal feature weights. For example, 0.3 may represent the feature weight in a specific spatial region, 0.5 represents the feature weight in a certain time period, and 0.2 represents the spatio-temporal correlation weight with other relevant meteorological elements. The terminal residual kernel generates the second spatio-temporal attribute vector. For example, the second spatio-temporal attribute vector corresponding to this target meteorological observation data is [0.4, 0.4, 0.2], and each numerical value also reflects specific spatio-temporal features.

[0128] Then, input the second spatio-temporal attribute vector into the recurrent layer in the MetNet model. The recurrent layer processes the input second spatio-temporal attribute vector based on its own structure and algorithm to generate the linear features of the target observation attributes. For example, the linear features of the target observation attributes may be represented as [0.6, 0.3, 0.1], where 0.6 may represent the linear feature weight related to the long-term change trend of meteorological elements, 0.3 represents the linear feature weight related to short-term fluctuations, and 0.1 represents the linear feature weight related to other interference factors.

[0129] Then, input the first spatio-temporal attribute vector, the second spatio-temporal attribute vector, the third spatio-temporal attribute vector, and the linear features of the target observation attributes into the fully connected layer in the MetNet model. The third spatio-temporal attribute vector in the embodiment of the present invention is a spatio-temporal attribute vector encoded by the residual kernel in the recurrent layer according to the augmented attribute knowledge, and the augmented attribute knowledge encoding is obtained by feature augmentation of the second spatio-temporal attribute vector. For example, the augmented third spatio-temporal attribute vector is [0.5, 0.3, 0.2]. The fully connected layer synthesizes the vector information of these inputs and generates the meteorological observation attribute knowledge encoding for each target meteorological observation data through its complex neuron connections and calculations. This encoding comprehensively reflects various characteristic information of the target meteorological observation data and is obtained on the basis of comprehensively considering the outputs of different-level residual kernels in the spatio-temporal feature mining layer, the linear features in the recurrent layer, and the augmented spatio-temporal attribute vector.

[0130] The coherence of this technical solution lies in that starting from the spatio-temporal feature mining layer, it gradually and deeply mines the features of the target meteorological observation data. First, it distinguishes spatio-temporal attributes at different levels through multiple residual kernels, then uses the recurrent layer to extract linear features, and finally synthesizes all relevant information in the fully connected layer to obtain the complete meteorological observation attribute knowledge encoding. This way makes full use of the functional characteristics of each layer of the MetNet model, enabling the meteorological observation attribute knowledge encoding to accurately reflect the essential characteristics of the target meteorological observation data.

[0131] Designed in this way, first, by using the spatio-temporal feature mining layer of the MetNet model and mining spatio-temporal attribute vectors through multiple residual kernels, it can deeply analyze the spatio-temporal characteristics of the target meteorological observation data. The output of different-level residual kernels are different spatio-temporal attribute vectors, and this hierarchical processing helps to comprehensively capture the features of the data at different spatio-temporal scales, avoiding the limitations of single-scale analysis. Secondly, the linear features of the target observation attributes generated by the recurrent layer add information about the change trend of meteorological elements to the meteorological observation attribute knowledge encoding. This helps to better understand the dynamic change laws of meteorological elements in subsequent data processing and analysis. Moreover, the fully connected layer synthesizes various vector information to generate the meteorological observation attribute knowledge encoding, making this encoding contain rich and comprehensive information. This helps to improve the accuracy and effectiveness of meteorological observation data in subsequent processing. For example, in applications such as data classification, association analysis, and meteorological prediction, a more comprehensive and accurate attribute knowledge encoding can provide a more reliable basis for these operations, thereby enhancing the performance of the entire meteorological observation data access system.

[0132] In summary, the embodiment of the present invention adopts a circular buffer structure to manage meteorological observation data, which has great advantages compared with the traditional linear stack method. When accessing commands and data in the traditional linear stack, due to its last-in-first-out principle and the fact that data operations are concentrated at the top of the stack, in the face of a large amount of meteorological observation data streams, problems of low efficiency will occur. The circular buffer structure of the embodiment of the present invention can support multiple independent data streams to share storage space in the same buffer, realizing continuous writing and circular overwriting of data. For example, in meteorological observations, data streams from different sensors (such as temperature sensors, humidity sensors, wind speed sensors, etc.) can be stored orderly in the circular buffer at the same time, avoiding the rapid performance degradation caused by the increase in data volume like the linear stack.

[0133] The continuous access and storage of data are ensured by the circular movement of the read and write pointers, and this feature is crucial in the meteorological business system. Meteorological observation is a continuous process, and data is constantly generated. The read and write pointer mechanism of the circular buffer ensures that data can be stored and read in a timely and continuous manner, without the blocking or delay phenomena that may occur in the linear stack when the data volume is large. Moreover, the circular buffer supports parallel access, which greatly improves the data access speed and efficiency. In the meteorological business system, there may be multiple tasks (such as real-time meteorological data monitoring, historical data query, device control instruction transmission, etc.) that need to access data at the same time. The parallel access ability of the circular buffer enables these tasks to be executed efficiently in parallel, greatly improving the operating efficiency of the entire system.

[0134] The overflow and underflow detection mechanism and the capacity management strategy for multi-channel data streams provided by the embodiment of the present invention provide a strong guarantee for the data integrity and system stability of the meteorological business system. During the continuous generation of meteorological observation data, if there is no effective overflow detection mechanism, data loss or system crashes may occur when the data volume exceeds the storage capacity, while the overflow detection mechanism of the circular buffer can detect and handle this situation in a timely manner. Similarly, the underflow detection mechanism can prevent abnormal situations that occur when the data reading speed is too fast or the data generation is insufficient. The capacity management strategy for multi-channel data streams can reasonably allocate storage space according to the needs of different data streams to ensure that the data of each channel can be properly processed, which is particularly important in the multi-source data fusion environment of the meteorological business system, avoiding the impact on the operation of the entire system due to the excessive or insufficient data volume of a certain channel.

[0135] Mining the attribute knowledge of meteorological observation data and generating meteorological observation attribute knowledge codes, and then performing a series of read-write state mappings based on this to obtain relevant vectors and labels, which helps to manage meteorological observation data more accurately. Meteorological observation data has rich connotations, and different data has different importance and functions in aspects such as meteorological analysis, forecasting, and equipment control. Through attribute knowledge mining, the characteristics of each target meteorological observation data can be deeply understood, and then targeted management can be carried out according to these characteristics during the data access process. For example, for data that plays a key role in meteorological forecasting, a higher priority can be given in the read-write state mapping, so as to be reflected in the data access management advice vector and the global read-write state label, in order to formulate a more optimized data access strategy and improve the decision-making accuracy and efficiency of the meteorological business system.

[0136] Furthermore, Figure 2 FIG. 5 is a schematic structural diagram of a meteorological observation data access system 200 provided by an embodiment of the present invention. As Figure 2 shown, the meteorological observation data access system 200 includes a processor 210. The processor 210 can call and run a computer program from a memory to implement the method in the embodiment of the present invention.

[0137] Optionally, as Figure 2 shown, the meteorological observation data access system 200 may further include a memory 230. Among them, the processor 210 can call and run a computer program from the memory 230 to implement the method in the embodiment of the present invention. Among them, the memory 230 can be a separate device independent of the processor 210, or can be integrated in the processor 210. Optionally, as Figure 2 shown, the meteorological observation data access system 200 may further include a transceiver 220. The processor 210 can control the transceiver 220 to interact with other devices. Specifically, it can send information or data to other devices, or receive information or data sent by other devices. Optionally, the meteorological observation data access system 200 can implement the corresponding processes of the storage engine or components (such as processing modules) in the storage engine or the device deployed with the storage engine in each method of the embodiment of the present invention. For the sake of brevity, it will not be described in detail here. It should be understood that the processor in the embodiment of the present invention may be an integrated circuit chip with signal processing capabilities. It can be understood that the memory in the embodiment of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. It should be noted that the memory of the system and method described in this article is intended to include but not limited to suitable types of memories.

[0138] On the above basis, a readable storage medium is provided. A program or instruction is stored on the readable storage medium, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0139] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the embodiments of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the embodiments of the present invention and the scope protected by the embodiments of the present invention, and all of them fall within the protection scope of the embodiments of the present invention.

Claims

1. A multi-channel ring access method for meteorological observation data, characterized in that: The method is applied to a meteorological observation data access system, and the method comprises: Taking multiple meteorological observation data streams in the target database as initial meteorological observation data streams respectively, performing a timing marking operation based on a ring buffer structure on the initial meteorological observation data streams to obtain multiple timing marked observation information, each of which includes multiple target meteorological observation data; Performing attribute knowledge mining on the multiple target meteorological observation data to obtain meteorological observation attribute knowledge codes of the multiple target meteorological observation data; The plurality of time series mark observation information are respectively used as data access mark information, and the meteorological observation attribute knowledge encoding of the plurality of target meteorological observation data in the data access mark information is mapped to a read-write state to obtain a read-write state description vector of the data access mark information; Determine the access management suggestion vector of the initial meteorological observation data stream by using the read and write state description vectors of the plurality of time-series-tagged observation information; Performing read-write status mapping on access management suggestion vectors of multiple meteorological observation data streams in the target database to obtain a global read-write status label of the target database, wherein the global read-write status label of the target database is used to determine a data access strategy of the target database; The process of determining the data access strategy of a target database according to a read-write status tag to achieve data access includes: performing feature matching on the global read-write status tag of the target database based on a strategy vector coordinate system to obtain the read-write execution strategy vector of the target database; processing the read-write execution strategy vector of the target database to obtain the data access strategy of the target database, wherein the strategy vector coordinate system is a reference system constructed to determine the read-write execution strategy vector, in which different coordinate axes represent different factors related to the data access strategy, and when performing feature matching in the strategy vector coordinate system, the read-write execution strategy vector is determined according to the meaning represented by each coordinate axis and the pre-set matching rules.

2. The method according to claim 1, characterized in that The method of using the multiple time series mark observation information as data access mark information, performing read-write state mapping on the meteorological observation attribute knowledge encoding of multiple target meteorological observation data in the data access mark information, and obtaining the read-write state description vector of the data access mark information includes: using the multiple time series mark observation information as data access mark information, performing read-write state mapping on the meteorological observation attribute knowledge encoding of multiple target meteorological observation data in the data access mark information, and obtaining a first read-write state mapping feature, wherein the first read-write state mapping feature includes a plurality of first read-write state mapping subgroups; determining the read-write state description vector of the data access mark information based on the read-write state description vectors of the plurality of first read-write state mapping subgroups; The method of performing read-write status mapping on the access management suggestion vectors of multiple meteorological observation data streams in the target database to obtain a global read-write status label of the target database includes: performing read-write status mapping on the access management suggestion vectors of multiple meteorological observation data streams in the target database to obtain a second read-write status mapping feature, wherein the second read-write status mapping feature includes multiple second read-write status mapping clusters; and determining the global read-write status label of the target database based on the read-write status description vectors of the multiple second read-write status mapping clusters.

3. The method according to claim 2, characterized in that The determining the read / write state description vector of the data access tag information based on the read / write state description vectors of the plurality of first read / write state mapping subgroups comprises: fusing vector elements of key state mapping members of the plurality of first read / write state mapping subgroups to determine the read / write state description vector of the data access tag information; Determining the global read-write status label of the target database based on the read-write status description vectors of the multiple second read-write status mapping subgroups includes: fusing vector elements of key status mapping members of the multiple second read-write status mapping subgroups to determine the global read-write status label of the target database.

4. The method according to claim 2, characterized in that The method of using the plurality of time series mark observation information as data access mark information respectively, and performing read-write state mapping on the meteorological observation attribute knowledge encoding of the plurality of target meteorological observation data in the data access mark information to obtain a first read-write state mapping feature includes: The multiple time-series mark observation information are respectively used as data access mark information, and based on multiple different first feature dimensions, the meteorological observation attribute knowledge encodings of multiple target meteorological observation data in the data access mark information are mapped to read and write states, so as to obtain first read and write state mapping features respectively corresponding to the multiple first feature dimensions, wherein the first feature dimension is used to indicate the number of first read and write state mapping subgroups in the corresponding first read and write state mapping features; the read and write state description vector of the data access mark information includes the read and write state description vectors respectively corresponding to the multiple first feature dimensions of the data access mark information; The method of using the read / write state description vectors of the plurality of timing mark observation information to determine the access management suggestion vector of the initial meteorological observation data stream comprises: fusing the read / write state description vectors corresponding to the same first characteristic dimension in the read / write state description vectors of the plurality of timing mark observation information to obtain the access management suggestion vectors of the initial meteorological observation data stream corresponding to the plurality of first characteristic dimensions respectively; The access management suggestion vectors of the multiple meteorological observation data streams in the target database are mapped to read and write states to obtain a second read and write state mapping feature, including: among the access management suggestion vectors of the multiple meteorological observation data streams in the target database, access management suggestion vectors corresponding to the same first feature dimension are mapped to read and write states to obtain second read and write state mapping features corresponding to multiple first feature dimensions respectively; the global read and write state label of the target database includes the read and write state description vectors corresponding to the multiple first feature dimensions of the target database respectively.

5. The method according to claim 3, characterized in that Each second read-write state mapping feature includes local mapping features corresponding to multiple different second feature dimensions, and each local mapping feature includes multiple second read-write state mapping clusters; the second feature dimension is used to indicate the number of second read-write state mapping clusters in the corresponding second read-write state mapping feature; the read-write state description vector of each first feature dimension of the target database includes multiple local read-write state vectors of the first feature dimension of the target database corresponding to multiple second feature dimensions.

6. The method according to claim 1, characterized in that The processing of the read / write execution strategy vector of the target database to obtain the data access strategy of the target database includes: based on the data access demand indicated by the channel capacity control vector, processing the read / write execution strategy vector of the target database through a feedforward neural network based on conditional probability to determine the data access strategy of the target database; The data access demand indicated by the channel capacity control vector is based on the data access demand indicated by the channel capacity control vector, and the read / write execution strategy vector of the target database is processed by a feed-forward neural network based on conditional probability to determine the data access strategy of the target database, including: combining the read / write execution strategy vector of the target database and the overflow detection mechanism execution vector of the target database by a feed-forward neural network based on conditional probability to obtain a multi-channel access interaction vector of the target database; based on the data access demand indicated by the channel capacity control vector, the multi-channel access interaction vector of the target database is processed to determine the data access strategy of the target database; The performing feature matching based on the strategy vector coordinate system on the global read-write status tag of the target database to obtain the read-write execution strategy vector of the target database includes: performing feature matching based on the strategy vector coordinate system on the global read-write status tag of the target database through a deep learning network to obtain the read-write execution strategy vector of the target database; The method further includes: debugging the deep learning network based on debugging examples including meteorological observation debugging vectors, overflow detection mechanism debugging vectors, channel capacity regulation training vectors, and data access annotation strategies; Among them, the deep learning network is used to perform feature matching on the meteorological observation debugging vector based on the strategy vector coordinate system to obtain the read-write debugging strategy vector; the feedforward neural network based on conditional probability is used to combine the read-write debugging strategy vector and the overflow detection mechanism debugging vector to obtain a multi-channel access debugging vector; based on the data access demand indicated by the channel capacity regulation training vector, the multi-channel access debugging vector is processed to determine the data access debugging strategy, and the network weights of the deep learning network are optimized according to the data access debugging strategy and the data access labeling strategy.

7. The method according to any one of claims 1 to 5, characterized in that: The method of using the multiple meteorological observation data streams in the target database as initial meteorological observation data streams respectively and performing a timing marking operation based on a ring buffer structure on the initial meteorological observation data streams to obtain multiple timing marked observation information includes: Taking multiple meteorological observation data streams in the target database as initial meteorological observation data streams respectively, and performing group marking based on time series nodes on the initial meteorological observation data streams to obtain meteorological observation marking groups; The meteorological observation mark group is divided into a plurality of time-series mark observation information according to the observation priority.

8. The method according to any one of claims 1 to 5, characterized in that: The method further includes: extracting data from the original meteorological observation report to obtain the multiple meteorological observation data streams.

9. A meteorological observation data access system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 8.

Images