Deep and far sea meteorological and hydrological data transmission method, electronic equipment and medium

Abstract

The invention provides a deep and far sea meteorological and hydrological data transmission method, electronic equipment and a medium. The method comprises the following steps: acquiring a real-time data segment with a preset length; determining data characteristics of the real-time data segment, wherein the data characteristics comprise a mean value, a variance, a time gradient, a baseline standard deviation, a power spectral density, a signal-to-noise ratio and autocorrelation of data in the real-time data segment; according to the data characteristics of the real-time data segment, determining the data type of the real-time data segment, the data type including transient data, slowly varying data and periodic data; according to the data type of the real-time data segment, data compression is carried out on the real-time data segment to obtain a compression result, and different data types adopt different data compression modes; and transmitting the compression result. The method can improve the transmission efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of data transmission, and more specifically to a method, electronic equipment and medium for transmitting deep-sea meteorological and hydrological data. Background Art

[0002] With the advancement of deep-sea survey activities, the amount of hydrological and meteorological data in deep-sea areas continues to increase. These data include air temperature, air pressure, wind speed, sea water temperature, salinity, depth and other information. The real-time collection and transmission of these data are crucial to supporting the efficient operation and intelligent development of deep-sea projects.

[0003] In the related technologies, there are many defects in the transmission of deep-sea meteorological and hydrological data, which restrict the efficient transmission of data. Specifically, in order to adapt to the increasing amount of data and improve the efficiency of data transmission, it is usually necessary to compress the data when transmitting deep-sea meteorological and hydrological data. However, the existing compression algorithms (such as LZW, DEFLATE, etc.) do not optimize the mutation and periodic characteristics of meteorological and hydrological data, and the compression effect is limited, resulting in low transmission efficiency under limited bandwidth conditions.

[0004] In view of this, the present invention is proposed. Summary of the invention

[0005] The present invention is proposed in consideration of the above problems. According to one aspect of the present invention, a method for transmitting deep-sea meteorological and hydrological data is provided, comprising: obtaining a real-time data segment of a preset length; determining data characteristics of the real-time data segment, the data characteristics comprising: mean, variance, time gradient, baseline standard deviation, power spectrum density, signal-to-noise ratio and autocorrelation of the data in the real-time data segment; determining the data type of the real-time data segment according to the data characteristics of the real-time data segment, wherein the data type comprises transient data, slowly varying data and periodic data; compressing the real-time data segment according to the data type of the real-time data segment to obtain a compression result, wherein different data types adopt different data compression methods; and transmitting the compression result.

[0006] Exemplarily, determining the data type of the real-time data segment according to the data characteristics of the real-time data segment includes: determining that the data type of the real-time data segment is slowly varying data when the real-time data segment meets the following conditions: ;in, is the variance of the data in the real-time data segment, is the mean value of the data in the real-time data segment, is the power spectrum density of the data in the real-time data segment; and / or, when the real-time data segment satisfies the following conditions, determining that the data type of the real-time data segment is transient data: ;in, is the time gradient of the data in the real-time data segment, is the baseline standard deviation of the data in the real-time data segment, is the power spectrum density of the data in the real-time data segment; and / or, when the real-time data segment satisfies the following conditions, determining that the data type of the real-time data segment is periodic data: ;in, is the power spectrum density of the data in the real-time data segment, is the signal-to-noise ratio of the data in the real-time data segment, is the autocorrelation of the data in the real-time data segment.

[0007] Exemplarily, data compression is performed on the real-time data segment according to the data type of the real-time data segment, including: when the data type of the real-time data segment is unique, if the data type of the real-time data segment is slowly varying data, the real-time data segment is compressed by residual coding based on linear prediction, wherein the compression result is a residual value greater than a residual threshold; if the data type of the real-time data segment is transient data, the real-time data segment is compressed by sparse representation based on wavelet transform, wherein the compression result is a sparse coefficient; if the data type of the real-time data segment is periodic data, the real-time data segment is compressed by Fourier parameterized compression, wherein the compression result is the amplitude, phase and frequency of some frequency components in the real-time data segment, and the sum of the energies of the some frequency components is greater than the energy threshold.

[0008] Exemplarily, the residual threshold is positively correlated with the standard deviation and range of the residual values ​​of each data point in the real-time data segment; and / or, the allowable error of the wavelet transform is positively correlated with the baseline error of the data in the real-time data segment and the L2 gradient norm of the data in the real-time data segment; and / or, the energy threshold is determined based on the total energy and the main frequency energy ratio of the real-time data segment, and the energy threshold is positively correlated with the total energy and the main frequency energy ratio of the real-time data segment.

[0009] Exemplarily, data compression is performed on the real-time data segment according to the data type of the real-time data segment, including: when the data type of the real-time data segment is not unique, selecting a segmentation strategy that minimizes the objective function to segment the real-time data segment to obtain multiple sub-data segments, wherein the objective function is a weighted sum of the KL divergence and the number of compressed bits of the sub-data segments obtained after segmentation, the KL divergence is a degree of match between the corresponding sub-data segment and an expected distribution of a compression model of the sub-data segment, and the sub-data segment compression model is a compression model corresponding to the data type of the sub-data segment; and data compression is performed on the multiple sub-data segments respectively according to the data type of each sub-data segment in the multiple sub-data segments obtained after segmentation, wherein the compression result includes results obtained by respectively compressing the multiple sub-data segments.

[0010] Exemplarily, before transmitting the compression result, the method also includes: determining the reliability weight of the compression result according to the priority and urgency of the compression result; determining the transmission mode of the compression result based on the priority and reliability weight; and / or, calculating the transmission deadline of the compression result according to the urgency of the compression result; wherein, when the transmission time of the compression result is greater than or equal to the transmission deadline, switching the transmission mode of the compression result to a degraded transmission mode; wherein, when the transmission mode of the compression result is the degraded transmission mode, only the low-frequency part of the compression result is transmitted; and / or, selecting a satellite link according to the data type of the compression result; determining the initial number of redundant packets in the transmission process according to the packet loss rate estimate and the channel reliability factor of the selected satellite link.

[0011] Exemplarily, the initial number of redundant packets in the transmission process is determined according to the estimated value of the packet loss rate of the selected satellite link and the channel reliability factor, and is performed when the estimated value of the packet loss rate of the current satellite link is greater than the packet loss rate threshold.

[0012] Exemplarily, the method further includes: monitoring the channel quality and bandwidth usage during data transmission; and adjusting the bandwidth during transmission of the compression result according to the priority, channel quality and bandwidth usage of the compression result.

[0013] According to another aspect of the present invention, there is provided an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to execute the computer program to implement the above method.

[0014] According to another aspect of the present invention, a computer-readable storage medium is provided, storing a computer program / instruction, and the computer program / instruction implements the above method when executed by a processor.

[0015] In the above technical scheme, by classifying the real-time data segments and selecting the best compression method in real time according to the data type of the real-time data segments, this adaptive compression method can achieve compression optimization for the slow-changing, sudden and periodic characteristics of deep-sea meteorological and hydrological data, and make full use of the potential compression space of the data, which helps to improve the data compression effect and compression efficiency, reduce the amount of transmitted data, and reduce the bandwidth resources required during the transmission process, thereby realizing efficient transmission of deep-sea meteorological and hydrological data under limited bandwidth conditions.

[0016] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other purposes, features and advantages of the present invention will become more apparent by describing the embodiments of the present invention in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 A schematic flow chart showing a method for transmitting deep-sea meteorological and hydrological data according to an embodiment of the present invention; Figure 2 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical scheme and advantages of the present invention more obvious, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described in the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the protection scope of the present invention.

[0020] As mentioned above, there are many defects in the transmission of deep-sea meteorological and hydrological data, which restrict the efficient transmission of data. These defects include: limited data compression effect, bandwidth waste and unnecessary redundant transmission during transmission, and poor transmission effect of the adopted transmission protocol. Specifically, first, traditional data transmission methods usually directly transmit raw data, fail to fully utilize the time correlation, periodicity and spatial redundancy of data, resulting in a large amount of bandwidth waste and unnecessary redundant transmission. Secondly, some data transmission processes compress data before compression, but the existing compression algorithms (such as LZW, DEFLATE, etc.) do not optimize the mutation and periodic characteristics of meteorological and hydrological data in a targeted manner, and the compression effect is limited, resulting in low transmission efficiency under limited bandwidth conditions. In addition, existing transmission protocols (such as TCP and UDP) are difficult to effectively apply in satellite link environments with high latency and high packet loss rate. The TCP protocol is less efficient under high latency, while UDP lacks reliability guarantees and cannot take into account both real-time and data integrity, resulting in data loss and unstable transmission quality. To solve at least one of the above problems, the present invention proposes a transmission method, electronic device and storage medium for deep-sea meteorological and hydrological data. The method can achieve targeted compression of different meteorological and hydrological data, which helps to improve the data compression effect and reduce the bandwidth resources required during the transmission process, thereby achieving efficient transmission under limited bandwidth conditions. The method, electronic device and storage medium are described in detail below.

[0021] According to one aspect of an embodiment of the present invention, a method for transmitting deep-sea meteorological and hydrological data is provided. In this article, meteorological and hydrological data include but are not limited to meteorological elements such as air temperature, air pressure, wind direction, wind speed, maximum wind speed, air pressure, 3-hour voltage change, temperature, dew point temperature, visibility, total cloud cover, typhoon warning, typhoon path, etc.; hydrological element data such as waves, tide level, seawater temperature, flow velocity, flow direction, depth, salinity, etc.

[0022] Figure 1 FIG. 2 is a schematic flow chart showing a method for transmitting deep-sea meteorological and hydrological data according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps S110, S120, S130, S140 and S150.

[0023] In step S110, a real-time data segment of a preset length is obtained.

[0024] It can be understood that the real-time data segment is a continuous data segment obtained from the original data stream of deep-sea meteorological and hydrological data. In some implementations, a real-time data segment of a preset length (to facilitate the distinction from the length of the sliding window hereinafter, it may be referred to as the first preset length) may be obtained from the original data stream by a sliding window. Of course, it may also be obtained by direct interception, which will not be described in detail.

[0025] The first preset length can be selected according to actual needs, for example, it can be selected as half an hour, 1 hour, 2 hours, etc., which will not be elaborated.

[0026] In step S120, data features of the real-time data segment are determined, the data features including: mean, variance, time gradient, baseline standard deviation, power spectrum density, signal-to-noise ratio and autocorrelation of the data in the real-time data segment.

[0027] After obtaining the real-time data segment, the data features of the real-time data segment can be statistically analyzed. Specifically, the time-frequency domain features such as the mean, variance, time gradient, baseline standard deviation, power spectrum density, signal-to-noise ratio and autocorrelation of the data in the real-time data segment can be calculated. These features can more accurately reflect the changing trend of the data in the real-time data segment. These data features can be calculated by traditional calculation methods. For example, the power spectrum density and signal-to-noise ratio of the real-time data segment can be obtained by Fourier transform, and the autocorrelation of the data in the real-time data segment can be obtained by autocorrelation function (ACF). I will not go into details.

[0028] In step S130, the data type of the real-time data segment is determined according to the data characteristics of the real-time data segment, wherein the data type includes transient data, slowly changing data and periodic data.

[0029] In this embodiment, the data structure of deep sea meteorological and hydrological data is divided into transient data, slowly changing data and periodic data. Table 1 below is a data type table of a specific implementation of this embodiment.

[0030] Table 1 Data type table

[0031] As shown in Table 1, different data types are used to represent different data change patterns. The characteristic of slowly changing data is that the data changes relatively slowly and the change amplitude is small, which is usually suitable for low-frequency acquisition and compression. The characteristic of transient data is that the data changes dramatically in a short period of time, which is usually manifested as local sharp fluctuations. Periodic data is characterized by obvious periodic fluctuations in the data, which is suitable for periodic modeling and transmission. Since different types of data have different change characteristics, the data type of the real-time data segment can be judged by the data characteristics of the real-time data segment. This data type can provide a more accurate basis for the compression and transmission of real-time data segments.

[0032] In step S140, data compression is performed on the real-time data segment according to the data type of the real-time data segment to obtain a compression result, wherein different data types use different data compression methods.

[0033] As mentioned above, different data types have different changing rules, so different data compression methods can be used to compress them in a targeted manner. For example, since the data of slowly varying data changes smoothly, the data with large deviation (i.e., large residual) can be captured as the compression result through compression. Since transient data has high-frequency local mutations, its mutation information can be concentrated on a few coefficients through compression. Since periodic data has obvious periodic fluctuations, its main spectral information rather than the complete data sequence can be captured through compression. In this way, the amount of data required to be transmitted can be effectively reduced.

[0034] In step S150, the compression result is transmitted. After the compression result is obtained, the compression result can be directly transmitted. Of course, the transmission method of the compression result can also be predetermined according to the data priority, urgency and other information of the compression result, and the transmission strategy can be dynamically adjusted during the data transmission process to achieve more efficient data transmission.

[0035] In the related technology, although some schemes compress the transmitted deep-sea meteorological and hydrological data, they usually adopt a fixed compression strategy and cannot optimize the slow-changing, sudden and periodic characteristics of deep-sea meteorological and hydrological data, resulting in insufficient data compression efficiency and failure to fully utilize the potential compression space of the data. In the above technical scheme, by classifying the real-time data segments and selecting the best compression method for compression in real time according to the data type of the real-time data segments, this adaptive compression method can achieve compression optimization for the slow-changing, sudden and periodic characteristics of deep-sea meteorological and hydrological data, and fully utilize the potential compression space of the data, which helps to improve the data compression effect and compression efficiency, reduce the amount of transmitted data, and reduce the bandwidth resources required during the transmission process, so that the efficient transmission of deep-sea meteorological and hydrological data can be achieved under limited bandwidth conditions.

[0036] Exemplarily, the data type of the real-time data segment is determined based on the data characteristics of the real-time data segment, including: judging whether the real-time data segment is slowly varying data based on the variance, mean and power spectrum density of the data in the real-time data segment; and / or judging whether the real-time data segment is transient data based on the time gradient, baseline standard deviation and power spectrum density of the data in the real-time data segment; and / or judging whether the real-time data segment is periodic data based on the power spectrum density, signal-to-noise ratio and autocorrelation of the data in the real-time data segment.

[0037] In the scheme of this example, the variance, mean and power spectrum density of the data in the real-time data segment can be used to determine whether the data change trend in the real-time data segment is gentle. If the change is gentle, it can be determined that at least part of the data in the real-time data segment meets the data characteristics of slowly varying data, and the data type of the real-time data segment can be considered to be slowly varying data. In some embodiments, judging whether the data type of the real-time data segment is slowly varying data based on the variance, mean and power spectrum density of the data in the real-time data segment includes: determining that the real-time data segment is slowly varying data when the real-time data segment meets the following conditions: ; in, is the variance of the data in the real-time data segment, is the mean value of the data in the real-time data segment, is the power spectrum density of the data in the real-time data segment. This shows that the data changes are relatively stable. This indicates that the spectrum energy in the real-time data segment is concentrated in the low-frequency range (0-0.01 Hz), indicating that the data changes slowly.

[0038] In the scheme of this example, it is possible to determine whether there is a local sharp fluctuation in the real-time data segment based on the time gradient, baseline standard deviation and power spectrum density of the data in the real-time data segment. If there is, it can be determined that at least part of the data in the real-time data segment meets the data characteristics of transient data, and the data type of the real-time data segment can be considered to be transient data. In a specific embodiment, judging whether the real-time data segment is transient data based on the time gradient, baseline standard deviation and power spectrum density of the data in the real-time data segment includes: determining that the data type of the real-time data segment is transient data when the real-time data segment meets the following conditions: ; in, is the time gradient of the data in the real-time data segment, is the baseline standard deviation of the data in the real-time data segment, is the power spectrum density of the data in the real-time data segment. This indicates that the data is changing at a high rate and significant mutations have occurred. This indicates that the data contains more high-frequency components, which means that the data has stronger instantaneous fluctuations.

[0039] In the scheme of this example, it is possible to determine whether there is obvious periodic fluctuation in the real-time data segment based on the power spectrum density, signal-to-noise ratio and autocorrelation of the data in the real-time data segment. If so, it can be determined that at least part of the data in the real-time data segment meets the data characteristics of periodic data, and the data type of the real-time data segment can be considered to be periodic data. In a specific embodiment, judging whether the real-time data segment is periodic data based on the power spectrum density, signal-to-noise ratio and autocorrelation of the data in the real-time data segment includes: determining that the data type of the real-time data segment is periodic data when the real-time data segment meets the following conditions: ; in, is the power spectrum density of the data in the real-time data segment, is the signal-to-noise ratio of the data in the real-time data segment, is the autocorrelation of the data in the real-time data segment. This indicates that there are significant periodic components in the data spectrum. It means that the data has high autocorrelation and strong periodic volatility, which is suitable for periodic modeling.

[0040] The above technical solution can more accurately determine the data type of the real-time data segment according to the data change law of the real-time data segment, which can provide a reliable basis for determining the compression and transmission strategy of the real-time data segment in subsequent steps.

[0041] Exemplarily, according to the data type of the real-time data segment, data compression is performed on the real-time data segment, including: when the data type of the real-time data segment is unique, if the data type of the real-time data segment is slowly varying data, the real-time data segment is compressed by using a residual coding method based on linear prediction, wherein the compression result is a residual value greater than a residual threshold; if the data type of the real-time data segment is transient data, the real-time data segment is compressed by using a sparse representation method based on wavelet transform, wherein the compression result is a sparse coefficient; if the data type of the real-time data segment is periodic data, the real-time data segment is compressed by using Fourier parameterized compression, wherein the compression result is the amplitude, phase and frequency of some frequency components in the real-time data segment, and the sum of the energy of some frequency components is greater than the energy threshold. This method of compressing the real-time data segment according to the data type can be called a dynamic adaptive segmentation compression algorithm (Dynamic Adaptive Segmentation Compression, DASC).

[0042] In the above, the data type in the real-time data segment is determined based on whether the judgment condition of the data type is met. In this case, the data type of the real-time data segment may have the following possibilities: the real-time data segment only meets the judgment condition of one data type, and the data type of the real-time data segment is unique; the real-time data segment meets the judgment conditions of multiple data types at the same time, and the data type of the real-time data segment is multiple, for example, the data type of the real-time data segment may be slowly varying data and periodic data; the real-time data segment does not meet the judgment conditions of each data type, and the real-time data segment cannot be classified. In this article, the situation where the real-time data segment meets the judgment conditions of multiple data types at the same time and the situation where the real-time data segment cannot be classified can be uniformly expressed as the data type of the real-time data segment is not unique.

[0043] In the scheme of this embodiment, when the real-time data segment is only slowly varying data, a compression strategy based on linear prediction residual coding is used to compress the real-time data segment. This method is adopted because the data changes of slowly varying data are slow, and linear prediction can effectively capture the data trend. The residual part reflects the actual change. When the residual is greater than a certain threshold, the residual is transmitted. Only when the data has a large deviation (i.e., the residual is large) do these deviated parts need to be transmitted, thereby greatly reducing unnecessary data transmission, especially when the data changes slowly.

[0044] When the real-time data segment is only transient data, a sparse representation compression strategy based on wavelet transform is used to compress the real-time data segment. Transient data has high-frequency local mutations, which can be converted into sparse representation through wavelet transform, so that most of the information is concentrated on a few coefficients. In this way, the amount of data that needs to be transmitted can be effectively reduced while retaining a large amount of information in the real-time data segment.

[0045] When the real-time data segment is only periodic data, the Fourier parameterized compression strategy is used to compress the real-time data segment. Periodic data has periodic characteristics. The frequency domain parameters of periodic data (i.e., the amplitude, phase, and frequency of the frequency components) can be obtained through Fourier transform. During transmission, only the amplitude, phase, and frequency of several frequency components with relatively concentrated energy are transmitted, rather than the complete data sequence. In this way, the amount of data transmission can be reduced, the transmission efficiency can be improved, and most of the energy can be captured during transmission.

[0046] In the above technical solution, targeted compression of real-time data segments can be achieved by matching corresponding compression strategies according to the data type of the real-time data segment. This compression method can make full use of the potential compression space of the data while ensuring data quality, thereby improving the compression effect and reducing the amount of data transmitted, which in turn helps to increase the transmission speed and reduce the loss of bandwidth resources during data transmission.

[0047] Exemplarily, the real-time data segment includes a plurality of data points; and the real-time data segment is compressed by using a residual coding method based on linear prediction, including: for each of the plurality of data points, using the following formula to predict a predicted value of the data point: ; in, is the predicted value, is the prediction coefficient, which can be selected according to actual needs. is the historical data value of the previous data point; It is the historical data value of the second data point before this data point; Calculate the residual value between the predicted value and the actual data value of the data point; When the residual value of the data point is greater than the residual threshold, it is determined that the compression result includes the residual value of the data point.

[0048] The residual value can be calculated using the following formula: ; in, Represents the residual value, i.e., the prediction error; Indicates the actual data value of the data point.

[0049] In this example, only when the residual value is greater than the residual threshold Transmit only when ∣> The residual value of the real-time data segment is shown in Table 2.

[0050] Table 2 Residual table

[0051] The above technical solution can significantly reduce the amount of data transmission, and this method is particularly suitable for compressing slowly varying data.

[0052] Exemplarily, the real-time data segment is compressed by using a sparse representation method based on wavelet transform, including: determining the sparsest wavelet coefficient combination to obtain a compression result under the premise of satisfying constraint conditions; wherein the constraint conditions are: , is the wavelet dictionary, is the sparse coefficient, is the vector representation of the real-time data segment, For the allowable error.

[0053] In this example scenario, is the vector representation of the entire real-time data segment. For example, if the real-time data segment contains 1000 sampling points, then is a 1000-dimensional column vector, where each element corresponds to the value of a data point. Each sparse coefficient corresponds to a wavelet basis function, not a single data point. Wavelet dictionary yes where N is the total number of data points (i.e. the dimension of x) and M is the number of wavelet basis functions. yes The dimensional column vector represents the weight coefficients of each wavelet basis, and its sparsity (few non-zero items) is the key to compression.

[0054] Determining the sparsest combination of wavelet coefficients is to find the wavelet coefficients that meet the optimization goal, which can be expressed as , which is to minimize The L1 norm (i.e., sparsity) of the data. The allowable error can be a fixed value set based on experience, or a dynamic value dynamically adjusted based on data in the real-time data segment.

[0055] The above technical solution realizes sparse coding through L1 norm optimization instead of traditional wavelet threshold compression, which can ensure the reconstruction error ( ) is controllable, find the sparsest ( This method uses L1 norm minimization to approximate the original data, making the representation of most data sparse (that is, only a small number of coefficients are significant), which greatly reduces the amount of data required for transmission.

[0056] Exemplarily, the real-time data segment is compressed by Fourier parameterized compression, including: performing Fourier transform on the real-time data segment to obtain the amplitude, phase and frequency of each of multiple frequency components used to represent the spectrum characteristics of the real-time data segment; determining some frequency components that meet the following conditions in the multiple frequency components as compression results: ; in, K is the total number of frequency components in the multiple frequency components, For the k The amplitude of a frequency component.

[0057] The energy threshold can be selected according to actual needs. For example, it can be 95% of the total energy. In a specific embodiment, if the first three frequency components (such as 12h, 24h, and 6h harmonics of the tide) have covered 95% of the energy, only these three groups are transmitted. ,in, For the k The phase of the frequency component, For the k Of course, the energy threshold value can also be a dynamic value dynamically adjusted according to actual needs, which will not be described in detail.

[0058] The above technical solution can ensure that most of the energy is captured when compressing the periodic data, so that the periodic data can be transmitted in an efficient manner.

[0059] Exemplarily, the residual threshold is positively correlated with the standard deviation and range of the residual values ​​of each data point in the real-time data segment; and / or, the allowable error of the wavelet transform is positively correlated with the baseline error of the data in the real-time data segment and the L2 gradient norm of the data in the real-time data segment; and / or, the energy threshold is determined based on the total energy and the main frequency energy ratio of the real-time data segment, and the energy threshold is positively correlated with the total energy and the main frequency energy ratio of the real-time data segment.

[0060] In some implementations of this example, the residual threshold is positively correlated with the standard deviation and range of the residual values ​​of each data point in the real-time data segment. That is, the residual threshold can be dynamically adjusted according to the standard deviation and range of the residual values ​​of each data point in the real-time data segment. In this case, the residual threshold can be adaptively scaled with the intensity of data fluctuations to avoid excessive residual transmission or information loss caused by a fixed threshold, thereby helping to ensure the efficiency and authenticity of data transmission.

[0061] In some embodiments, the residual threshold may be determined by the following formula: ; ; in, is the standard deviation of the residual values ​​of each data point in the real-time data segment; , is a weight coefficient, which can be set according to actual needs or obtained through historical data training and optimization. In some embodiments, ; is the extreme difference, which can be used as a sensitive indicator of the abnormal amplitude of the residual; is the maximum residual value among all data points in the real-time data segment; is the minimum residual value among all data points in the real-time data segment; is the residual threshold.

[0062] In a specific embodiment, the standard deviation and range of the residual value of the real-time data segment can be calculated first, and then the residual threshold is updated according to the standard deviation and range. Then, the real-time data segment is compressed using the residual threshold. For example, if , then the residual threshold can be calculated to be 0.13. In this case, when compressing the real-time data segment, the residual value greater than 0.13 is transmitted, and the residual value less than 0.13 is discarded.

[0063] In some implementations of this example, the allowable error of the wavelet transform is positively correlated with the baseline error of the data in the real-time data segment and the L2 gradient norm of the data in the real-time data segment. w The sparsity of ) and local gradient Adjust the tolerance, which helps to relax the tolerance when the signal changes suddenly (keep more high frequency details), tighten when stable , which helps improve the compression ratio.

[0064] In some embodiments, the allowable error is determined by the following formula: ; in, Indicates the allowable error; represents the baseline error, =0.1 , is the standard deviation of the original data (i.e., the real-time data segment); Represents the L2 gradient norm of the data in the real-time data segment, which is used to quantify the mutation intensity; represents the smoothing factor. In this embodiment, ).

[0065] In a specific embodiment, the gradient of the real-time data segment can be first calculated and The tolerance is then dynamically adjusted and the real-time data segment is then compressed based on the adjusted tolerance.

[0066] In a specific embodiment, =0.1, =5, α =10, in this case, δ =0.1×(1+5 / 10)=0.15.

[0067] In some implementations of this example, the energy threshold is determined based on the total energy of the real-time data segment and the main frequency energy ratio, and the energy threshold is positively correlated with the total energy of the real-time data segment and the main frequency energy ratio. That is, the size of the energy threshold can be dynamically adjusted based on the total energy of the real-time data segment and the main frequency energy ratio. Thus, the energy threshold can be increased in a strong periodic signal, thereby reducing the number of transmission coefficients; and the energy threshold can be lowered in a weak periodic signal, thereby avoiding the loss of critical frequencies.

[0068] In some embodiments, the energy threshold is determined by the following formula: ; in, is the total energy; is the main frequency energy ratio, η ∈[0,1], ; is the sensitivity coefficient, which can be set as needed, for example, 0.5. In a specific embodiment, if =10, η =0.9, then the threshold = 95% × 10 × (1 − 0.5 × 0.1) = 9.025.

[0069] In a specific embodiment, the total energy and the main frequency energy ratio of the real-time data segment can be calculated first, and then the energy threshold can be dynamically adjusted. Finally, the real-time data segment can be compressed according to the adjusted energy threshold.

[0070] The above technical solution can dynamically adjust parameters such as residual threshold, allowable error and energy threshold according to the statistical characteristics of the data, thereby avoiding overfitting or information loss caused by fixed parameters.

[0071] Exemplarily, data compression is performed on the real-time data segment according to the data type of the real-time data segment, including: when the data type of the real-time data segment is not unique, the segmentation strategy that minimizes the objective function is selected to segment the real-time data segment to obtain multiple sub-data segments, wherein the objective function is the weighted sum of the KL divergence and the number of compressed bits of the sub-data segment obtained after segmentation, the KL divergence is the matching degree between the corresponding sub-data segment and the expected distribution of the compression model of the sub-data segment, and the sub-data segment compression model is a compression model corresponding to the data type of the sub-data segment; according to the data type of each sub-data segment in the multiple sub-data segments obtained after segmentation, data compression is performed on the multiple sub-data segments respectively, wherein the compression results include the results obtained by respectively compressing the multiple sub-data segments. This scheme can be called a dynamic segmentation KL divergence selection algorithm.

[0072] During the research, the inventor found that for time series data with dynamic mixed characteristics, especially for scenarios where certain meteorological and hydrological data belong to multiple types of data (slowly changing data, transient data, and periodic data), the core feature is that the dynamic pattern of the data changes over time. At this time, the compression efficiency is low and cannot be fully utilized by only using a single compression algorithm. In view of this, the inventor considers using a dynamic KL divergence (Kullback-Leibler Divergence) selection algorithm to quantify the matching degree between each segment in the real-time data segment and the compression model corresponding to different data types, and optimizes and selects the optimal segment in combination with the compression rate weight, so as to solve the problem of information loss and low compression efficiency caused by fixed compression strategies in the compression algorithm.

[0073] In this example, the objective function can be expressed as: ; in, ; in, Indicates the measurement of the current data segment Expected distribution with compression model The smaller the value, the higher the fidelity of the compressed data. Indicates the number of compressed bits, that is, the number of bits in the compressed data segment. The smaller the value, the higher the compression rate. is a weight coefficient used to balance fidelity and compression rate (the default value =0.5, which can be adjusted dynamically according to the channel status). λ Role: λ →0: Prioritize fidelity (select the model with the smallest KL divergence); λ →∞: Prioritize compression rate (select the model with the least number of bits).

[0074] In some implementations of this example, the expected distribution of compression models corresponding to different data types is for: Slowly changing data: , assuming that the data follows a low-variance Gaussian distribution, this compression model can be called an SV model; Transient data: , assuming that the data obeys a sparse Laplace distribution in the wavelet domain, reflecting the spike characteristics of the mutation, this compression model can be called a TR model; Cycle data: , assuming that the data is composed of a few sinusoidal components, the compression model can be called a TD model.

[0075] in, It can be trained using the following methods. First, a large number of samples can be extracted from historical data and compressed using the SV / TR / PD model respectively; the distribution of the residuals after compression of each model is statistically analyzed and fitted into the SV / TR / PD model respectively. For example, for the SV model, the residual can be calculated e t The mean and variance of N(0, ); For the TR model, the decay rate of the wavelet coefficients can be counted to determine the scale parameter b of the Laplace distribution. In addition, it can also be dynamically adjusted according to real-time data Parameters (such as sliding window updates or the scale parameter b of the Laplace distribution) to achieve dynamic updating of the model.

[0076] In some implementation schemes of this example, the number of compressed bits can be calculated in the following manner: for one of the segments in the real-time data segment, compress the segment using a compression model corresponding to the data type of the segment; and count the number of bits of the compressed output (including frequency parameters and residuals).

[0077] Example, SV model: Transfer initial value x 0 (32-bit floating point number) + large residual { ti , ei} (each residual occupies 16 bits of timestamp + 16 bits of value), ; TR model: transmission wavelet basis identifier (8 bits) + non-zero coefficients { index , value} (each coefficient occupies 10 bits of index + 16 bits of value), ;PD model: transmission (16 bits per parameter, 48× K Bit), .

[0078] In a specific embodiment, a segment of the real-time data segment has a length of 1 minute and contains tidal (PD) and wave (TR) components.

[0079] KL divergence: D KL ( P seg∥ Q PD )=0.1 (high tidal matching); D KL ( P seg∥ Q TR )=0.3 (waves partially match); Number of bits: BitsPD=160 (3 frequency parameters); BitsTR=240 (8 wavelet coefficients); Select result: like λ =0.5, total cost PD=0.1+80=80.1, TR=0.3+120=120.3→ Select PD model.

[0080] Optionally, the dynamic piecewise KL divergence selection algorithm may include the following steps: First, a sliding window analysis is performed on the real-time data segment to extract multiple candidate segments from the real-time data segment, wherein the length of the sliding window may be a second preset length, which is smaller than the first preset length; for example, the second preset length T=5min; Then, the data features of each candidate segment obtained are calculated in real time, and the data type of each candidate segment is determined according to the data features. The method of determining the data type is similar to the method of determining the data type of the real-time data segment, and will not be repeated; Next, for each candidate segment, the weighted sum of its KL divergence and the number of compressed bits is calculated, and the segmentation strategy that minimizes the objective function is quickly solved through dynamic programming; Finally, the real-time data segment is segmented according to the segmentation strategy of minimizing the objective function to obtain multiple sub-data segments, and data compression is performed on the multiple sub-data segments respectively according to the data type of each sub-data segment in the multiple sub-data segments obtained after segmentation to obtain compression parameters.

[0081] The above technical solution achieves a dynamic trade-off between fidelity and compression rate through the joint optimization of KL divergence + compression bit number, which is suitable for mixed feature data. This design significantly improves the compression efficiency of complex meteorological and hydrological data while avoiding the limitations of fixed strategies.

[0082] Exemplarily, before transmitting the compression result, the method further includes: determining the reliability weight of the compression result according to the priority and urgency of the compression result; determining the transmission mode of the compression result based on the priority and reliability weight; and / or, calculating the transmission deadline of the compression result according to the urgency of the compression result; wherein, when the transmission time of the compression result is greater than or equal to the transmission deadline, switching the transmission mode of the compression result to a degraded transmission mode; wherein, when the transmission mode of the compression result is the degraded transmission mode, only the low-frequency part of the compression result is transmitted; and / or, selecting a satellite link according to the data type of the compression result; and determining the initial number of redundant packets in the transmission process according to the packet loss rate estimation value and the channel reliability factor of the selected satellite link. This scheme can be called a Hybrid Reliable Transmission Protocol (HRTP).

[0083] For ease of description, the following definitions are used for priority and urgency: priority is 1-5, with 5 being the highest; urgency is 1-3, with 3 being the most urgent.

[0084] In the solution of this embodiment, the priority of the compression result can be determined according to the data type of the real-time data segment corresponding to the compression result. For example, when the data type is slowly changing data, the priority is 1 or 2; when the data type is periodic data, the priority is 3; when the data type is transient data, the priority is 4 or 5.

[0085] When the data type of the compression result is not unique, the priority of the compression result can be calculated based on the proportion of different data features (slow-changing features, transient features, and periodic features) in the compression result. For example, the priority of the compression result can be calculated using the following formula:

[0086] in, is the feature weight, which can be an empirical value; is the proportion of data features in the compression result. For example, if the compression result contains 60% TR features + 40% PD features → DataPriority = 0.6×4 + 0.4×3 = 3.6, then the high reliability mode is triggered ( Wr >0.5).

[0087] The urgency can be preset manually or automatically generated according to the current data monitoring type. For example, when obtaining deep-sea meteorological and hydrological data by conventional monitoring, the urgency can be 1, and when obtaining deep-sea meteorological and hydrological data by disaster warning monitoring, the urgency can be 3.

[0088] Optionally, before transmitting the compression result, the method further includes: determining a reliability weight of the compression result according to the priority and urgency of the compression result; and determining a transmission mode of the compression result based on the priority and the reliability weight.

[0089] It can be understood that the reliability weight is a factor used to measure the reliability of data transmission, which is calculated based on the priority of the data and its sensitivity to transmission. In this embodiment, the reliability weight can be calculated using the following formula: ; in is the reliability weight; is the slope parameter, which is used to adjust the impact of the priority and can be set based on experience; It is the priority of the data; is the priority threshold, which is used to determine the critical value of the priority.

[0090] As mentioned above, the slope parameter k is used to control the weight growth rate. For example, when k =0.5, the priority increases by 1. Increase by about 0.2-0.3 (smooth transition); when k = 1.0, the priority difference is more sensitive, and the weight increases sharply (for example, priority 4 → =0.98).

[0091] In some embodiments, the results of the priority and reliability weights may be compared with corresponding thresholds (e.g., priority threshold and reliability threshold), and when both are greater than the corresponding thresholds, a high reliability transmission mode is triggered, otherwise, a balanced mode is triggered. In this embodiment, the parameters of each transmission mode may be predetermined, and the parameters of the transmission mode (i.e., transmission parameters) may include a redundancy multiple, a number of retransmissions, and a bandwidth preemption priority (which may be represented by a reliability weight, i.e., the higher the reliability weight, the higher the priority in occupying the channel to ensure low latency).

[0092] In other embodiments, determining the transmission mode of the compression result based on the priority and reliability weight may include the following steps: when the priority is greater than the priority threshold, determining the type of the transmission mode to be a high-reliability transmission mode; when the priority is less than or equal to the priority threshold, determining the type of the transmission mode to be a balanced mode; and determining the transmission parameters according to the reliability weight. The higher the reliability weight, the higher the redundancy multiple, the number of retransmissions, and the bandwidth preemption priority. In this embodiment, the priority is used as a logical judgment to intuitively distinguish between high and low priorities. For example, if the priority is greater than the priority threshold, it is directly determined that the compression result is transmitted in a high-reliability transmission mode. The reliability weight is used for dynamic control to quantify the intensity of the reliability requirement, so that policy mutations caused by relying solely on priority for judgment can be avoided. For example, when the priority threshold is 3, if only relying on the priority for binary judgment, the difference between priority 4 and priority 5 cannot be distinguished. Further judgment through the reliability weight can achieve progressive control of transmission parameters. For example, priority 4 → Wr =0.88 → Redundancy factor = 1.76; Priority 5 → Wr =0.95 → Redundancy factor = 2.0; Priority 3 → Wr =0.5→ Redundancy multiple=1.0 (baseline). Therefore, this scheme can effectively improve transmission efficiency and transmission effect by quickly classifying by priority and finely controlling the transmission strategy by reliability weight.

[0093] The priority threshold can be a fixed value set according to user experience, or a dynamic value that changes dynamically according to the degree of channel congestion. For example, when channel congestion is detected, the priority threshold is determined to be the first threshold; when sufficient channel bandwidth is detected, the priority threshold is determined to be the second threshold, and the first threshold is greater than the second threshold. In a specific embodiment, when the channel is congested: the priority threshold is 4, at which time only priority 5 triggers the high-reliability transmission mode, concentrating resources to protect the most critical data; when the bandwidth is sufficient, the priority threshold can be 2, at which time the high-reliability range can be expanded (priorities 3-5 all enhance transmission).

[0094] In a specific implementation, the priority of the compression result is 4, and the slope parameter k is 0.5, the priority threshold is 3, and the reliability weight is 0.88. In this embodiment, the transmission mode type of the compression result can be determined as a high reliability transmission mode according to the priority, and then the redundancy multiple is determined according to the reliability weight = 1 + 2 × (0.88-0.5) = 1.76 → 2 (rounded up); the number of retransmissions: increased from 3 times to 5 times; bandwidth preemption: priority occupation of the channel to ensure low latency.

[0095] The above scheme can accurately determine the transmission mode of the compression result by utilizing the priority and reliability weights, thereby ensuring that high-priority data is transmitted first, which helps to ensure the transmission efficiency of the compression result.

[0096] Optionally, before transmitting the compression result, the method also includes: calculating the transmission deadline of the compression result based on the urgency of the compression result; wherein, when the transmission time of the compression result is greater than or equal to the transmission deadline, switching the transmission mode of the compression result to a degraded transmission mode; wherein, when the transmission mode of the compression result is the degraded transmission mode, only the low-frequency part of the compression result is transmitted.

[0097] The transmission deadline can be used to control the delay of data transmission, ensure timeliness, and provide a common real-time constraint for compressed data. In some embodiments, the transmission deadline can be calculated using the following formula: ; in, is the transmission deadline; is the base time, usually the system initialization time; is the real-time adjustment factor, which is used to indicate the impact of urgency on the deadline; The urgency of the compression result.

[0098] In this embodiment, when the transmission mode of the compression result is the degraded transmission mode, only the low-frequency part of the compression result is transmitted, thereby ensuring the real-time performance of data transmission by reducing the amount of data transmission.

[0099] Optionally, before transmitting the compression result, the method further includes: selecting a satellite link according to the data type of the compression result; and determining the initial number of redundant packets in the transmission process according to the estimated packet loss rate and channel reliability factor of the selected satellite link.

[0100] In some embodiments, a satellite link is selected according to the data type of the compression result, including: when the data type is slowly changing data, it is transmitted through the VAST satellite; when the data type is transient data, it is transmitted through the Starlink and Beidou dual links; when the data type is periodic data, it is stably transmitted through the VAST baseline link. This solution is only an example, and the user can pre-define the satellite links corresponding to different data types according to actual conditions.

[0101] It can be understood that the satellite link selected according to the data type of the compression result can be regarded as the initially determined satellite link (ie, the default link). In some schemes of this embodiment, the satellite link can be further determined according to the priority and / or urgency of the compression result. For example, if a slowly varying data is manually marked as DataPriority=5 due to task requirements, even if its type is slowly varying data, it can still trigger the Starlink link (50Mbps-1Gbps) to be transmitted in parallel with the Beidou short message. In a specific embodiment, the link switching rules are shown in Table 3 below.

[0102] Table 3 Link switching rules

[0103] In this optional embodiment, the packet loss rate estimation value can be calculated by a signaling feedback method. For example, the receiving end can periodically send NACK packets, and the sending end can count the recent N The packet loss ratio of each packet is calculated to obtain the estimated packet loss rate. Of course, the estimated packet loss rate can also be obtained by updating the sliding window. For example, the exponential weighted moving average (EWMA) can be used to avoid instantaneous fluctuations. In this case, the estimated packet loss rate can be calculated by the following formula: .

[0104] In some embodiments, when multi-link parallel transmission is used, the estimated packet loss rate of the link with the highest packet loss rate can be selected as the basis for calculating the initial number of redundant packets. For example, when using Starlink + Beidou dual-link transmission, if the Starlink packet loss rate = 20% and the Beidou packet loss rate = 30%, the number of redundant packets is calculated as 30%.

[0105] In this optional embodiment, the channel reliability factor can be determined based on the channel type. For example, when the channel type is acoustic communication, the channel reliability factor can be 0.3-0.5, which indicates that the current channel has high packet loss and low bandwidth. When the channel type is satellite relay, the channel reliability factor can be 0.7-1, which indicates that the current channel has strong reliability. Of course, the channel reliability factor can also be determined by historical data statistics. For example, the channel reliability factor of a satellite link can be 0.6, and the channel reliability factor of an optical fiber can be 0.95.

[0106] In some embodiments, determining the initial number of redundant packets in the transmission process according to the estimated value of the packet loss rate of the selected satellite link and the channel reliability factor may include: determining the initial number of redundant packets by the following formula: ; in, is the estimated value of packet loss rate; is the channel reliability factor, ∈[0,1]; Indicates rounding up.

[0107] In a specific embodiment, the packet loss rate estimation value of the current channel can be estimated first, then the initial redundant packet number can be calculated, and finally, the number of transmission redundant packets can be adjusted according to the calculated initial redundant packet number to ensure that data can be successfully received in an environment with severe packet loss. Increasing the number of redundant packets helps to improve data reliability, especially when the channel quality is unstable.

[0108] The above scheme can calculate the number of redundant packets that need to be transmitted based on the estimated packet loss rate and the channel reliability factor to ensure that data can be fully received in an environment with high packet loss. Increasing the number of redundant packets can improve data reliability, especially in a high packet loss environment.

[0109] As mentioned above, traditional transmission protocols in related technologies, such as TCP and UDP, have their advantages in some scenarios, but cannot meet the multiple requirements of high efficiency, reliability and low latency in special environments such as the deep sea. The core idea of ​​the hybrid reliable transmission protocol (HRTP) proposed in the present invention is to dynamically adjust the data transmission strategy according to factors such as the real-time requirements, priority, channel status, etc. of different data. Through the above scheme, HRTP can ensure that high-priority and urgent data are transmitted first. At the same time, in environments with poor network conditions and high packet loss rates, the reliability of the data is enhanced by the number of redundant packets to ensure that key data is not lost. In this way, some special challenges faced in deep-sea communication environments can be addressed, including high latency, high packet loss rate, limited bandwidth, and conflicts between real-time and reliability requirements.

[0110] Optionally, the initial number of redundant packets in the transmission process is determined according to the estimated packet loss rate of the selected satellite link and the channel reliability factor, and is executed when the estimated packet loss rate of the current satellite link is greater than the packet loss rate threshold. Thus, in a low packet loss rate environment, a fixed number of redundant packets can be used, and in a high packet loss rate environment, a dynamic number of redundant packets can be used to improve data transmission reliability, ensuring that data will not be lost due to packet loss.

[0111] Exemplarily, the method further includes: monitoring the channel quality and bandwidth usage during data transmission; and adjusting the bandwidth during transmission of the compression result according to the priority, channel quality and bandwidth usage of the compression result.

[0112] In this example, the data priority reflects the importance and urgency of the data. Different data may have different priorities. Especially in deep sea environments, disaster warning information needs to be transmitted more preferentially than conventional meteorological data. The quality of the channel will directly affect the transmission rate and stability of the data. It can be measured using parameters such as signal-to-noise ratio and packet loss rate. In this embodiment, the channel quality is expressed in signal-to-noise ratio. Bandwidth usage reflects the current degree of bandwidth occupancy. The remaining amount of bandwidth directly determines whether the system can smoothly transmit additional data.

[0113] Optionally, according to the priority, channel quality and bandwidth usage of the compression result, the bandwidth in the transmission process of the compression result is adjusted, including: inputting the priority, channel quality and bandwidth usage (i.e., bandwidth utilization) of the compression result into the bandwidth allocation model to obtain the optimal bandwidth, wherein the adjusted bandwidth is the optimal bandwidth; wherein the bandwidth allocation model is constructed by a fuzzy reward function and a Q-learning algorithm, and the fuzzy reward function is used to calculate the fuzzy reward value of the current bandwidth allocation according to the priority membership corresponding to the priority of the compression result, the quality membership corresponding to the channel quality and the bandwidth usage; the Q-learning algorithm is used to update the Q value according to the fuzzy reward value, and determine the optimal bandwidth according to the updated Q value. This method can be called satellite bandwidth fuzzy Q optimization (FQO).

[0114] In some embodiments, the fuzzy reward function can be expressed as: ; in, Indicates priority membership; Indicates the quality membership; Indicates the used bandwidth, that is, the bandwidth used to transmit the compressed result; Indicates the total bandwidth currently available.

[0115] According to the above fuzzy reward function, it can be seen that if the data priority is high (P is high), the channel quality is good (Q is high), and the bandwidth utilization is good ( near ), the reward value is larger, which means that the current bandwidth allocation strategy is more effective, and the current transmission strategy can be maintained. If the channel quality is poor or the bandwidth utilization is too low, the reward value will be smaller, indicating that the current strategy may need to be adjusted, such as adjusting the bandwidth allocation, or increasing the number of redundant packets.

[0116] In this embodiment, the fuzzy reward function combines multiple factors that affect bandwidth allocation (such as priority, channel quality, and bandwidth usage) through fuzzy logic to evaluate the effect of the current bandwidth allocation strategy, so that the bandwidth allocation strategy can be dynamically adjusted according to different system states (such as channel quality fluctuations or data priority changes), ensuring that bandwidth utilization and system performance can be effectively improved under different network conditions. And this method of making nonlinear decisions through fuzzy logic can be applied to processing uncertainty and fuzzy information, achieving a more efficient and flexible decision-making process than traditional algorithms.

[0117] In this optional embodiment, the priority and channel quality are converted into corresponding memberships through fuzzy logic, and the fuzzy characteristics of the corresponding variables are quantified through the membership, thereby ensuring that the fuzzy reward function can flexibly adjust the reward value under different input values. In some embodiments, the priority membership corresponding to the priority can be determined in the following way: ; in, Indicates priority membership; P Indicates priority.

[0118] In some embodiments, the quality membership can be determined in the following manner: ; in, Indicates the quality membership; Indicates the best channel quality; represents the channel quality; σ represents the standard deviation of channel fluctuation.

[0119] In some embodiments, the update formula of the Q-learning algorithm can be expressed as: ; in, Yes Status and actions The value function of value; is the learning rate, which is used to control the update speed during the learning process; Is the current state and actions The immediate reward is the fuzzy reward value; is a discount factor, indicating the importance of future rewards; is the next state The maximum value.

[0120] In the Q-learning algorithm of this embodiment, the state Indicates the current status of the system, including data priority, channel quality, and remaining bandwidth, as well as actions , indicating the bandwidth allocation ratio, in different states according to The optimal bandwidth allocation strategy is selected based on the value. By using the fuzzy Q optimization algorithm, FQO can automatically adjust the bandwidth allocation strategy according to the real-time system status (data priority, channel quality, bandwidth resources, etc.), thereby ensuring the optimal use of satellite bandwidth resources. With the continuous updating of the Q-learning algorithm and the optimization of the reward function, FQO can optimize the reliability and real-time performance of data transmission in a dynamically changing deep-sea environment.

[0121] In some implementations of this example, when high-priority data (P≥3) and high-quality channels (SNR≥25dB) coexist, the fuzzy reward value is maximized and bandwidth is allocated to the VAST link first; if the channel quality is poor (SNR<15dB), the fuzzy reward value decreases, triggering Q-learning to switch to low-orbit satellites (Starlink) or Beidou short messages. In this implementation, by dynamically adjusting R(s,a) of Q-learning, the Q value update is affected, so that the fuzzy reward value can be used to guide Q-learning to give priority to VAST or Starlink under specific conditions.

[0122] In some embodiments, the specific method for determining the optimal bandwidth based on the updated Q value is as follows: Default mode: When the Q value points to VAST, 70%~100% of the bandwidth (regular data) is allocated, and emergency mode: When the Q value turns to Starlink due to the fuzzy reward mechanism, 40% of the bandwidth (high-priority transient data) is forcibly allocated. That is, in the default mode, VAST allocates 70%~100% of the bandwidth to transmit regular data (slowly changing data or periodic data), and Starlink is used as a backup; in emergency mode (for example, high-priority transient data is detected), 40% of the bandwidth is forcibly allocated to Starlink and 10% to Beidou short messages to achieve dual-channel redundancy for disaster data. This method can convert the actions output by Q-learning (such as "select VAST") into specific bandwidth ratios, thereby achieving dynamic adjustment of bandwidth.

[0123] In summary, the above technical solution combines the fuzzy reward function with the Q-learning algorithm, so that the bandwidth allocation decision is not only based on the current system state, but also gradually learns and optimizes the decision-making strategy. Through the combination of fuzzy logic and reinforcement learning, it can better adapt to the complex and changing transmission requirements in deep sea environments, improve the utilization efficiency of satellite bandwidth and the reliability of data transmission.

[0124] Exemplarily, the method further includes: adjusting the number of redundant packets (which may be referred to as the current number of redundant packets) during transmission of the compression result according to the channel quality.

[0125] Optionally, adjusting the current number of redundant packets according to the channel quality includes: determining whether the quality membership corresponding to the channel quality is less than a quality threshold; and increasing the current number of redundant packets when the quality membership is less than the quality threshold.

[0126] In some embodiments, the current number of redundant packets can be increased according to a preset growth value. For example, if the quality threshold is 0.5, when When <0.5, 1-2 packets can be added based on the current number of redundant packets (e.g. 3→5).

[0127] In some other embodiments, when the quality membership is less than the quality threshold, the current number of redundant packets may be increased in the following manner: .

[0128] The above scheme can dynamically adjust the number of redundant packets during the transmission of the compression result according to the channel quality, thereby further ensuring data transmission efficiency and transmission quality.

[0129] Exemplarily, the method further includes: adjusting the satellite link according to the channel quality. For example, when the quality membership is less than the quality threshold, the Starlink + Beidou dual-link parallel transmission can be triggered.

[0130] In a specific embodiment, the compression result is transmitted through the following steps: first, a default satellite link is selected according to the type of compression result, and the transmission deadline and the initial number of redundant packets are calculated according to the priority and urgency of the compression result. Then, during the transmission of the compression result, the channel quality and bandwidth usage are detected in real time, and the fuzzy reward value is calculated using the fuzzy reward function. Next, the Q-learning algorithm is used to update the Q value, and the bandwidth allocation strategy is dynamically adjusted according to the Q value, and the current number of redundant packets and the satellite link are adjusted according to the channel quality. When the transmission result is received, if the transmission is successful, the Q-learning algorithm is strengthened to maintain the current parameters; if the transmission fails, the transmission is downgraded, and inefficient actions are punished in the Q-learning algorithm (i.e., the corresponding Q value is reduced).

[0131] In a specific embodiment of the present invention, the DASC algorithm is used to compress the real-time data segment, the HRTP protocol is used to transmit the compression result, and the fuzzy reward function and the Q-learning algorithm are combined to perform dynamic adjustment during the transmission of the compression result. It has been verified that this embodiment has the following effects compared with the prior art: 1. Data compression efficiency is significantly improved: Slowly changing data (such as temperature and salinity): through linear prediction residual coding, the compression rate is stable at 92%-95% (the original 10MB / h is reduced to 0.8MB / h for a measured data set in a certain sea area); sudden change data (such as typhoon): wavelet sparse compression rate is 78%-82%, and the retention error of key wind speed mutation points is ≤3%; periodic data (such as tides): Fourier parameterization compression rate is 80%-85%, and the daily cycle waveform reconstruction error is <1%; 2. Greatly enhanced transmission reliability: High-priority disaster data (such as typhoon path): Through Starlink + Beidou dual-channel redundancy, the transmission success rate is increased from 82% of traditional TCP to 99.3% (measured in a deep sea environment with a packet loss rate of 30%); Real-time guarantee: Emergency data (UrgencyLevel=3) end-to-end delay ≤60ms, timeout rate reduced from 35% to below 5%; 3. Satellite bandwidth utilization optimization: The average daily bandwidth usage of VAST satellites has been reduced by 40% (from 100Mbps to 60Mbps), and the peak rate of Starlink burst transmission has reached 900Mbps (used to transmit typhoon high-definition radar data); the dynamic bandwidth allocation strategy has reduced the channel idle rate by 50%, avoiding resource waste; 4. Verification of adaptability to complex environments: Under severe sea conditions with fluctuating signal-to-noise ratio (10-25dB) and packet loss rate of 20%-50%, the system maintained a 98% data integrity rate through three-level disaster recovery (VAST→Starlink→Beidou), and the interruption recovery time was less than 3 seconds; a deep-sea exploration ship operated continuously for 3 months, with a cumulative data transmission volume of 12TB and a bit error rate of only 0.02%; 5. Energy efficiency and cost advantages: Compared with the traditional solution (full transmission + fixed redundancy), satellite communication costs are reduced by 55%, and the power consumption of shipborne equipment is reduced by 30%. It meets the low power consumption requirements of deep-sea exploration vessels for long-term operations.

[0132] According to yet another aspect of the embodiments of the present invention, an electronic device is provided. Figure 2 FIG. 2 shows a schematic block diagram of an electronic device according to an embodiment of the present invention. Figure 2 As shown, the electronic device 200 includes: a processor 210 and a memory 220. The memory 220 stores a computer program, and the processor 210 is used to execute the computer program to implement the above method.

[0133] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is also provided. The storage medium stores a computer program / instruction, and the computer program / instruction implements the above method when executed by a processor. The storage medium may include, for example, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0134] A person skilled in the art can easily understand the implementation structure, working principle and beneficial effects of the electronic device and the computer-readable storage medium by reading the above method, and will not be described in detail for the sake of brevity.

[0135] Although the exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the exemplary embodiments described above are merely exemplary and are not intended to limit the scope of the present invention thereto. Various changes and modifications may be made therein by those skilled in the art without departing from the scope and spirit of the present invention.

[0136] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0137] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0138] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0139] Similarly, it should be understood that in order to streamline the present invention and help understand one or more of the various inventive aspects, in the description of the exemplary embodiments of the present invention, the various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, the method of the present invention should not be interpreted as reflecting the following intention: the claimed invention requires more features than the features explicitly stated in each claim. More specifically, as reflected in the corresponding claims, the inventive point is that the corresponding technical problem can be solved with less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are hereby expressly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present invention.

[0140] Those skilled in the art will understand that, except for mutually exclusive features, all features disclosed in this specification (including the accompanying claims, abstract and drawings) and all processes or units of any method or device disclosed in this specification may be combined in any combination. Unless otherwise explicitly stated, each feature disclosed in this specification (including the accompanying claims, abstract and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0141] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims, any one of the claimed embodiments may be used in any combination.

[0142] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all functions of some modules in an electronic device according to an embodiment of the present invention. The present invention may also be implemented as a device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0143] The above description is only a specific implementation mode of the present invention or an explanation of a specific implementation mode, and the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for transmitting deep sea meteorological and hydrological data, characterized in that: include: Get a real-time data segment of preset length; Determining data features of the real-time data segment, the data features comprising: mean, variance, time gradient, baseline standard deviation, power spectral density, signal-to-noise ratio and autocorrelation of data in the real-time data segment; Determining the data type of the real-time data segment according to the data characteristics of the real-time data segment, wherein the data type includes transient data, slowly varying data and periodic data; According to the data type of the real-time data segment, the real-time data segment is compressed to obtain a compression result, wherein different data types use different data compression methods; The compression result is transmitted.

2. The method according to claim 1, characterized in that The step of determining the data type of the real-time data segment according to the data feature of the real-time data segment includes: When the real-time data segment meets the following conditions, it is determined that the data type of the real-time data segment is slowly varying data: ； in, is the variance of the data in the real-time data segment, is the mean value of the data in the real-time data segment, is the power spectrum density of the data in the real-time data segment; and / or, When the real-time data segment meets the following conditions, it is determined that the data type of the real-time data segment is transient data: ； in, is the time gradient of the data in the real-time data segment, is the baseline standard deviation of the data in the real-time data segment, is the power spectrum density of the data in the real-time data segment; and / or, When the real-time data segment meets the following conditions, it is determined that the data type of the real-time data segment is periodic data: ； in, is the power spectrum density of the data in the real-time data segment, is the signal-to-noise ratio of the data in the real-time data segment, is the autocorrelation of the data in the real-time data segment.

3. The method according to claim 1, characterized in that The step of compressing the real-time data segment according to the data type of the real-time data segment comprises: When the data type of the real-time data segment is unique, If the data type of the real-time data segment is slowly varying data, the real-time data segment is compressed by using a residual coding method based on linear prediction, wherein the compression result is a residual value greater than a residual threshold; If the data type of the real-time data segment is transient data, the real-time data segment is compressed by using a sparse representation method based on wavelet transform, wherein the compression result is a sparse coefficient; If the data type of the real-time data segment is periodic data, the real-time data segment is compressed using Fourier parametric compression, wherein the compression result is the amplitude, phase and frequency of some frequency components in the real-time data segment, and the sum of the energies of the some frequency components is greater than an energy threshold.

4. The method according to claim 3, characterized in that The residual threshold is positively correlated with the standard deviation and range of the residual values ​​of each data point in the real-time data segment; and / or, the allowable error of the wavelet transform is positively correlated with the baseline error of the data in the real-time data segment and the L2 gradient norm of the data in the real-time data segment; and / or, the energy threshold is determined based on the total energy of the real-time data segment and the proportion of the main frequency energy, and the energy threshold is positively correlated with the total energy of the real-time data segment and the proportion of the main frequency energy.

5. The method according to claim 1, characterized in that The step of compressing the real-time data segment according to the data type of the real-time data segment comprises: When the data type of the real-time data segment is not unique, Selecting a segmentation strategy that minimizes an objective function to segment the real-time data segment to obtain a plurality of sub-data segments, wherein the objective function is a weighted sum of a KL divergence and a number of compressed bits of the sub-data segments obtained after segmentation, the KL divergence is a matching degree between a corresponding sub-data segment and an expected distribution of a compression model of the sub-data segment, and the sub-data segment compression model is a compression model corresponding to a data type of the sub-data segment; According to the data type of each sub-data segment in the multiple sub-data segments obtained after segmentation, data compression is performed on the multiple sub-data segments respectively, wherein the compression results include results obtained by performing data compression on the multiple sub-data segments respectively.

6. The method according to any one of claims 1 to 5, characterized in that: Before transmitting the compression result, the method further includes: Determining a reliability weight of the compression result according to the priority and urgency of the compression result; Determining a transmission mode of the compression result based on the priority and the reliability weight; and / or, Calculating a transmission deadline of the compression result according to the urgency of the compression result; Wherein, when the transmission time of transmitting the compression result is greater than or equal to the transmission deadline, switching the transmission mode of the compression result to a degraded transmission mode; Wherein, when the transmission mode of the compression result is the degraded transmission mode, only the low-frequency part of the compression result is transmitted; and / or, Selecting a satellite link according to the data type of the compression result; The number of initial redundant packets in the transmission process is determined according to the estimated packet loss rate of the selected satellite link and the channel reliability factor.

7. The method according to claim 6, characterized in that The method of determining the initial number of redundant packets in the transmission process according to the estimated value of the packet loss rate of the selected satellite link and the channel reliability factor is performed when the estimated value of the packet loss rate of the current satellite link is greater than the packet loss rate threshold.

8. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: Monitor channel quality and bandwidth usage during data transmission; The bandwidth during transmission of the compression result is adjusted according to the priority of the compression result, channel quality and bandwidth usage.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: A computer program / instruction is stored, and when the computer program / instruction is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images