A method for correcting and predicting tidal levels by integrating measured data feedback and tidal range analysis

Abstract

This invention discloses a tidal level boundary correction and forecasting method that integrates measured data feedback and tidal range analysis, belonging to the field of tidal level boundary prediction technology. The method includes the following steps: based on the preprocessed tidal level signal, a reduced-order physical constraint semi-diurnal tide model is fitted sequentially, followed by peak-valley effectiveness screening guided by the fitting template, to extract tidal morphological features; based on the tidal morphological features, amplitude correction coefficients, basal correction coefficients, and initial phase correction values ​​are extracted, and the initial phase correction values ​​are optimized to obtain diurnal variation coefficients and cumulative coefficients for performing tidal level boundary correction and forecasting. This invention achieves unmanned, real-time, automatic tidal level boundary correction and forecasting.

Description

Technical Field

[0001] This invention relates to the field of tidal boundary prediction technology, and more specifically, to a tidal boundary correction and prediction method that integrates measured data feedback and tidal range analysis. Background Technology

[0002] Currently, harmonic analysis is commonly used for tidal boundary condition correction. This method follows the Rayleigh criterion and requires long-term tidal time-series data of more than 15 days to decompose into various tidal constituents such as M2, S2, K1, and O1. The boundary conditions are manually adjusted by comparing the amplitude and phase deviations of each tidal constituent. Due to the frequency difference between the M2 and S2 tidal constituents, this method theoretically requires at least 14.8 days of observation data to effectively separate adjacent tidal constituents. The entire process relies on local file storage of observation data, and the correction cycle is as long as half a month or more. The forecast results are manually copied and distributed. It can only rely on long-term time-series data for batch correction and cannot use short-term observation deviations of 2 to 3 days for real-time dynamic correction.

[0003] This traditional harmonic analysis scheme has several inherent flaws. First, as a frequency domain method, harmonic analysis's frequency resolution is directly tied to data length, a fundamental limitation that cannot be compensated for through algorithmic optimization. It not only fails to analyze short-term tide sequences, and forcibly fitting short sequences easily leads to severe overfitting and distorted correction coefficients, but also struggles to distinguish between meteorological tidal disturbances and true astronomical tidal signals, with tidal anomalies caused by meteorological forcing being mistakenly included in the tidal parameters. Second, the various stages of the entire operational chain are fragmented, lacking an automated path from raw observation data to correction decisions. Key nodes such as data quality inspection, tidal deviation reliability assessment, deviation source differentiation, and correction magnitude determination all rely on manual judgment by professionals, failing to form an unmanned feedback loop of "perception → analysis → decision → correction." Furthermore, the existing system lacks supporting maintenance and fault-tolerance mechanisms. There are no alternative degradation schemes when station data is missing, historical correction coefficients are not persistently stored in a database, and retrospective reuse and error accumulation tracking are not supported. There is also no automatic distribution channel for corrected forecast results, requiring manual operation throughout, resulting in poor business continuity and data flow efficiency.

[0004] In summary, existing harmonic analysis correction techniques suffer from two core bottlenecks that are difficult to address through local optimization: First, the underlying algorithms are constrained by fundamental principles, making it impossible to accurately separate astronomical tidal components from short-term observation data, thus hindering real-time correction of tidal boundaries. This limitation can only be overcome by abandoning the approach of tidal decomposition. Second, the overall business architecture lacks an autonomous intelligent decision-making framework and supporting capabilities such as automatic verification, fault degradation, historical parameter backtracking, and automatic distribution of results. Numerous manual intervention nodes hinder the implementation of unmanned and automated tidal correction systems. The industry urgently needs a new correction technology solution that can adapt to short-term tidal sequences, possesses a fully automated closed-loop system, and supports fault tolerance and historical parameter management. Summary of the Invention

[0005] To address the aforementioned issues, the present invention aims to provide a tidal level boundary correction and forecasting method that integrates measured data feedback and tidal range analysis. This method establishes a complete automated closed loop for tidal level data acquisition, correction, forecasting, distribution, and feedback. It replaces traditional harmonic analysis with tidal range analysis combined with measured feedback, eliminating the need for manual intervention throughout the entire process. It automatically completes data quality inspection, astronomical tide extraction, correction coefficient calculation and verification, fault tolerance degradation, and boundary correction, thus eliminating the need for manual experience-based judgment and achieving unmanned, real-time, automatic correction and forecasting of tidal level boundaries.

[0006] To achieve the above technical objectives, this application provides a tidal level boundary correction and prediction method that integrates measured data feedback and tidal range analysis, including the following steps: Based on the preprocessed tidal signal, the tidal morphology features are extracted after sequentially performing the reduced-order physical constraint semi-diurnal tide model fitting and the peak-valley effectiveness screening guided by the fitting template. Based on tidal morphological characteristics, amplitude correction coefficient, datum correction coefficient, and initial phase correction value are extracted. After optimizing the initial phase correction value, the diurnal variation coefficient and cumulative coefficient are obtained to perform tidal boundary correction forecast.

[0007] Preferably, when preprocessing the tide level signal, integer multiple tide period completion and low-pass filtering are used to perform the tide level signal preprocessing.

[0008] Preferably, when performing the fitting of the reduced-order physical constraint semi-diurnal tide model, the reduced-order physical constraint semi-diurnal tide model represents: f(x)=a0+a1·cos(2π / 89400·x+t1)+a2·cos(2π / 44700·x+t2) Where: x represents the time from the origin of the timekeeping; a0 represents the mean sea level offset; a1 and t1 represent the amplitude and initial phase of the diurnal tide component, respectively, with a fixed period of 89400 seconds; a2 and t2 represent the amplitude and initial phase of the semi-diurnal tide component, respectively, with a fixed period of 44700 seconds.

[0009] Preferably, when performing peak and valley validity screening guided by the fitting template, peak and valley detection, duplicate peak and valley merging, intersection of two sequences, and matching verification based on the fitting template are performed sequentially to perform peak and valley validity screening guided by the fitting template.

[0010] Preferably, during the matching verification, if there are no matching items, it means that the measured tide level during the intersection period of the two sequences is severely affected by meteorological disturbances and there is no corresponding astronomical tide extreme value. This is determined to be an abnormal period, and all measured peaks and valleys in the data segment area during the intersection period of the two sequences are discarded and not included in subsequent correction calculations. If only one match is found, the measured peak and valley are consistent with the theoretical astronomical tide pattern and are considered valid astronomical tide signals, which are retained for subsequent feature extraction. If two or more extreme values ​​of the same type are matched, the occurrence of multiple extreme values ​​of the same type within the same tidal cycle is considered noise interference. Only the peak value with the highest water level is retained, and only the valley value with the lowest water level is retained. Other adjacent extreme values ​​are discarded.

[0011] Preferably, when extracting the amplitude correction coefficient, the measured tidal range and the tidal range output by the forecast model are used to obtain the amplitude correction coefficient.

[0012] Preferably, when extracting the datum correction coefficient, based on the measured average tidal crossing zero level and the predicted average crossing zero level, the predicted datum is first multiplied by the amplitude correction coefficient to complete the tidal range scaling correction, and then the measured datum is subtracted from the scaled predicted datum to obtain the datum correction coefficient.

[0013] Preferably, when extracting and optimizing the initial value of phase correction, based on the occurrence time of high tide and low tide, the measured and predicted high and low tide times are summed and the difference is calculated. Finally, the average value is taken to obtain the initial value of phase correction, and the initial value of phase correction is optimized by using a fine search based on minimizing error.

[0014] Preferably, when performing tide boundary forecasting, in response to multiple tide observation stations, a comprehensive coefficient for open boundary correction is obtained by weighted averaging of the diurnal variation coefficients of each station, so as to perform tide boundary correction forecasting.

[0015] Preferably, when performing tidal boundary forecasting, a linear gradual transition strategy for the first 6 time steps is adopted based on the cumulative coefficient to perform tidal boundary correction forecasting.

[0016] The present invention discloses the following technical effects: This invention relies on algorithms such as a simplified semi-diurnal tide physical model, short-sequence periodic completion filtering, and template matching peak-valley screening to break through the limitations of traditional harmonic analysis on long-term data of more than 15 days from the underlying principle. It can accurately remove meteorological noise and extract real astronomical tide signals based on short-term tide level sequences of 2 to 3 days, avoiding problems such as short-sequence overfitting, filtering distortion, and tide pattern confusion.

[0017] This invention establishes a full-link autonomous decision-making architecture, equipped with automated logic such as adaptive sampling and detection, two-level correction coefficient chain accumulation, multi-station weight normalization and degradation, two-stage phase optimization, and independent verification of three physical coefficients. It completely replaces the manual experience-based judgment process and realizes fully automated processing of data quality inspection, coefficient calculation, validity verification, fault degradation, and multi-station fusion.

[0018] This invention employs a smooth transition boundary loading scheme to avoid numerical model oscillations caused by abrupt changes in correction coefficients, thus ensuring stable operation of ocean models such as FVCOM.

[0019] This invention constructs a complete unmanned feedback loop covering data storage, automatic correction, numerical forecasting, encrypted distribution, and measured feedback. It is compatible with different sampling equipment, supports continuous fault tolerance and degradation of multiple stations, and distinguishes between short-term fluctuations and long-term drift to achieve iterative accumulation of correction effects. Ultimately, it completes all-weather, real-time, and fully automatic stable correction and forecasting of tidal boundary. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the method described in this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] like Figure 1As shown, this invention provides a tidal level boundary correction and forecasting method that integrates measured data feedback and tidal range analysis. This invention constructs a fully automated feedback closed loop encompassing measured data, correction analysis, numerical forecasting, encrypted distribution, and measured return flow. It employs a chain of methods including reduced-order physical constraint semi-diurnal tide model fitting, template-guided peak and valley selection, and tidal morphology feature extraction. It can solve for three types of correction coefficients with clear physical meaning—amplitude, datum, and phase—based solely on a 2-3 day short-term measured tidal level sequence. By nesting a double-layer correction coefficient system, it tracks short-term tidal fluctuations and long-term model system drift, and is equipped with a multi-stage autonomous degradation strategy. In scenarios such as missing station data or invalid daily coefficients, it sequentially performs historical valid coefficient backtracking and neutral coefficient degradation processing, requiring no manual judgment or intervention throughout the process, achieving real-time, stable, and unmanned automatic correction of tidal level boundary conditions.

[0024] Step S100 utilizes integer multiples of the tidal period for completion and low-pass filtering to perform tidal level signal preprocessing.

[0025] In one implementation, adaptive sampling step size detection is performed.

[0026] For example, since the measured tide level data comes from different types of acquisition terminals, the sampling intervals are not uniform and there may be some missing data. At the same time, a fixed sampling step size is not assumed. Instead, the actual sampling step size is adaptively detected through statistical methods.

[0027] For example, the mode of the time difference between all adjacent time points in the tide level sequence is statistically analyzed—the frequency distribution of the time difference is calculated, and the time difference with the highest frequency is taken as the actual sampling step size dt. Specifically, this includes: inputting the complete tide level observation sequence 'tide' arranged chronologically; traversing all observation records in the sequence except the last one, calculating the time difference between adjacent data points for each record, and storing the result in a delta_t array; counting the frequency of all time differences, selecting the time interval with the highest frequency as the final sampling step size dt, and outputting it. This logic can automatically adapt to observation intervals of various acquisition devices, such as 5 minutes, 10 minutes, and 30 minutes, to achieve cross-device compatibility.

[0028] For example, when performing integer multiples of the lunar cycle for completion, the boundary effects of short-sequence filtering need to be addressed before low-pass filtering. Traditional zero-padding or mirror-padding introduces artificial amplitude attenuation or phase distortion, making it unsuitable for tidal signals.

[0029] For example, this invention utilizes the quasi-periodicity of tidal signals to complete the sequence to an integer multiple of the complete lunar day cycle (89400 seconds = 24 hours and 50 minutes). Specifically, the current sequence length is calculated and divided by an integer multiple of 89400 to obtain the target length. Data is copied from the previous cycle segment of the sequence end (i.e., the data between 89400 seconds before the end of the current sequence and 89400 seconds before the target end) and appended to the end of the sequence, ensuring that the completed sequence length is exactly an integer multiple of 89400 seconds. This includes: Calculate the end time (end_time) after the cycle is completed: First, calculate the total duration of the beginning and end of the original tide level sequence, divide it by the duration of a single lunar day (89400 seconds, 24 hours and 50 minutes), and round up to obtain the number of complete lunar days that need to accommodate the original data; then multiply this number by 89400 seconds, add it to the sequence start time, and obtain the total end time of the completed sequence, ensuring that the total duration after completion is exactly an integer multiple of a lunar day; Extract the original tidal segment data buchong for completion: Extract a segment of tidal data with a duration of 1 lunar day from the end of the original sequence. The filtering criteria are: the time of this segment is later than the end time of the original sequence minus 89400 seconds and earlier than the completion end time minus 89400 seconds. This segment has a complete tidal cycle shape and is used for subsequent filling. The time axis of the completed segment is shifted: the time of all data in the completed segment buchong is uniformly increased by 89400 seconds, and this tidal waveform is shifted to the blank interval after the end of the original sequence to conform to the quasi-periodic pattern of the tide. Splicing to generate a complete periodic sequence: The time-shifted complete segment is appended to the end of the original tide level sequence to obtain the complete sequence pt_tide_bu. The total duration of this sequence is strictly an integer multiple of 89,400 seconds of the lunar day.

[0030] For example, the physical principle is that the Lunar Day (89400 seconds) is the time interval between two consecutive moon transits and is the dominant cycle of tidal changes in the semi-diurnal tidal region. By completing the sequence to an integer multiple of this cycle, the boundary effect falls only within the completed artificial data segment. After filtering, these segments are truncated to recover the signal within the original time range.

[0031] For example, low-pass filtering and data recovery are performed. An 8th-order Butterworth low-pass filter is applied to the completed sequence, with the cutoff frequency set to the frequency value corresponding to a 2-hour period (f_cutoff=10×7200 / dt / N, where N is the length of the completed sequence, dt is the sampling step size, and a factor of 10 is the safety factor). After filtering, the manually completed segments are truncated, retaining only the filtered results within the original time range.

[0032] For example, the present invention takes as input the tide level sequence pt_tide_bu after lunar diurnal periodicity completion, and finally outputs smoothed tide level data that retains only the original observation period and filters out high-frequency noise. The complete process is as follows: The normalized cutoff frequency of the filter is calculated, with a target filtering period of 2 hours. The normalized low-frequency threshold freq_lowlimit is calculated by combining the sampling step size dt, the total length N of the completed sequence, and a safety factor of 10. This threshold is used to distinguish between short-period high-frequency interference such as wind waves and ship waves that need to be filtered out and the full-day and half-day tidal signals that need to be retained. Construct an 8th-order Butterworth low-pass filter. Based on the cutoff frequency calculated in the previous step, generate 8th-order Butterworth low-pass filter parameters b and a. Higher-order filters can achieve steep frequency boundaries and accurately separate the effective tidal signal from high-frequency noise. Zero-phase bidirectional filtering is performed by calling the zero-phase bidirectional filtering function filtfilt to filter the complete periodic completion sequence. This method does not introduce additional phase shifts into the tide time series, ensuring that no distortion occurs at the time of tide occurrence. After removing manually completed sections and restoring the original observation interval, all manually completed data with times later than the end of the original tide level sequence are removed, and only the smooth tide levels within the original observation time range are retained, resulting in the final tide level sequence pt_tide_new without high-frequency noise and boundary distortion.

[0033] For example, the reason for choosing 2 hours as the cutoff period is that the semi-diurnal tide cycle is about 12.4 hours and the diurnal tide cycle is about 24.8 hours. Fluctuations less than 2 hours are high-frequency noise (wind waves, ship waves, etc.) and need to be filtered out; while tidal signals longer than 2 hours are completely preserved.

[0034] Step S200 performs a reduced-order physical constraint semi-diurnal tide model fitting.

[0035] In one implementation, a simplified 5-parameter physical constraint model replaces the dozens of tidal parameters in traditional harmonic analysis. The model is defined as follows: f(x)=a0+a1·cos(2π / 89400·x+t1)+a2·cos(2π / 44700·x+t2) Where: x represents the time (seconds) from the origin of the timekeeping; a0 represents the mean sea level offset (meters); represents the reference surface; a1 and t1 represent the amplitude (meters) and initial phase (radians) of the diurnal tide component, respectively, with a period fixed at 89400 seconds (lunar day); a2 and t2 represent the amplitude (meters) and initial phase (radians) of the semi-diurnal tide component, respectively, with a period fixed at 44700 seconds (lunar semi-diurnal).

[0036] For example, in most sea areas dominated by semi-diurnal tides, tidal energy is mainly concentrated in two frequency bands: diurnal (24.8h cycle) and semi-diurnal (12.4h cycle). Traditional harmonic analysis requires dozens of tidal constituents to accurately distinguish tidal constituent pairs with extremely close frequencies within the same frequency band (e.g., the frequency difference between M2 and S2 is only 0.0000014Hz), which requires long sequences that satisfy the Rayleigh criterion. However, the purpose of this invention is not to accurately decompose tidal constituents, but to extract the tidal range and phase characteristics required for correction, i.e., a simplified physical model containing only two dominant frequencies. Because the frequency is "locked" to known values ​​as a physical constraint, the data only needs to fit the amplitude and phase parameters, significantly reducing the requirement for data length.

[0037] For example, nonlinear least squares fitting is performed using scipy's curve_fit, with initial parameter values ​​set to mean tidal level and zero phase, and parameter boundaries set within a reasonable physical range (non-negative amplitude and phase within [0, 2π]).

[0038] For example, the fitted curve represents the signal of an "ideal astronomical tide" generated solely by celestial tidal forces, a pure tidal reference that removes meteorological forcing (wind stress, pressure variations) and local topographic effects. All subsequent analyses are performed on this reference.

[0039] Step S300 performs peak and valley validity screening guided by the fitted template.

[0040] In one implementation, in addition to the peaks and troughs of astronomical tides, the actual tide sequence also contains a large number of "pseudo-peaks and troughs" caused by meteorological forcing (storm surge, sudden pressure changes) and local effects. If these abnormal peaks and troughs are included in the calculation of correction coefficients, they will seriously contaminate the results.

[0041] For example, the present invention uses the fitted curve as an "astronomical tide template" to screen effective astronomical tide peaks and valleys, specifically including the following steps. Step 301 Peak and Valley Detection (set_flag): Iterate through each sampling point of the tide level sequence and identify local maxima (peak value, flag=1) and local minima (valley value, flag=-1) using a three-point comparison method. A special case of consecutive equal values ​​(tide level plateaus) is handled by backtracking to the previous point of unequal value for comparison.

[0042] Step 302 repeats peak-valley merging (check_flag2). When the time difference between two adjacent extreme points of the same type (two peaks or two valleys) is less than 8 hours, only the more extreme one is retained: the higher of the two peaks is retained, and the lower of the two valleys is retained. This step eliminates false extremes caused by high-frequency fluctuations remaining due to incomplete filtering.

[0043] Step 303: Intersection of the two sequences (the first two parts are steps S301 and S302). The time range intersection of the peaks and valleys of the measured sequence and the predicted sequence is calculated respectively, and the peak and valley values ​​of the two sequences with no overlap at the beginning and end are removed.

[0044] Step 304 is based on the matching verification of the fitting template (intersection, the third part, i.e., step S303). It iterates through each standard extreme value (peak / valley) fit_extreme extracted from the fitting curve and searches for matching items in all the peak and valley data identified in the actual sequence. The matching must meet two hard conditions at the same time: Condition 1: The type of the actual extreme value label is consistent with the template extreme value, both being peak values ​​or both being valley values; Condition 2: The time difference between the time of occurrence of the actual extreme value and the time of occurrence of the template extreme value is less than 3 hours, falling within a reasonable tidal time window; According to the number of matching items found, they are processed into three categories: (1) No matching items found Matching item: indicates that the measured tide level during this period (intersection of two sequences) is severely affected by meteorological disturbances and there is no corresponding astronomical tide extreme value. It is judged as an abnormal period. All measured peaks and valleys in this area (data segment area during the intersection of two sequences) are invalidated and will not participate in subsequent correction calculations; (2) Only one match is found: the measured peak and valley are consistent with the theoretical astronomical tide pattern and belong to the effective astronomical tide signal. They are retained for subsequent feature extraction; (3) Two or more extreme values ​​of the same type are matched: multiple extreme values ​​of the same type appearing in the same tidal cycle are noise interference. Only the point with the highest water level is retained for the peak value and only the point with the lowest water level is retained for the valley value. Other adjacent extreme values ​​are removed.

[0045] For example, verified peaks and valleys retain their flag values; those that fail verification are remarked as flag=0 (non-extreme values) and do not participate in subsequent tidal range and phase calculations.

[0046] For example, the fitted curve represents the "location where the astronomical tide should occur". If an actual peak and valley deviates from the fitted peak and valley by more than 3 hours in time, or if the fitted curve does not have a corresponding peak and valley at the location where the actual peak and valley occur, then this actual peak and valley is very likely a non-astronomical tide signal caused by wind stress or sudden changes in air pressure.

[0047] Step S400: Extraction of tidal morphological features.

[0048] In one embodiment, among the effective peaks and valleys after screening, the last 26 hours and 50 minutes (89400+9000 seconds) of the intersection of the two sequences is taken as the feature extraction window. This window covers the most recent 1.1 lunar days, ensuring that it contains at least one complete tidal process. The feature values ​​are defined as shown in Table 1.

[0049] Table 1 For example, linear interpolation is used to calculate the time across zero points during rapid rises and falls. In the tidal level sequence between high tide and low tide, the two adjacent sampling points whose tidal level values ​​are closest to the MTL are found, and the time across zero points is accurately calculated by linear interpolation.

[0050] Based on the strength of the core, the feature quantities corresponding to the measured sequence in the forecast sequence are obtained through time nearest neighbor matching. Using the time corresponding to the HW and LW of the measured sequence as anchor points, the corresponding peak and valley with the closest time is searched in the effective peak and valley of the forecast sequence.

[0051] Step S500: Calculation of three correction coefficients and two-stage phase optimization.

[0052] In one implementation, a first-stage physical coarse estimation based on morphological features is performed.

[0053] The three correction coefficients are each directly derived from tidal morphological characteristics and have clear physical meanings: Amplitude correction coefficient scale: TR_obs is the measured tidal range, and TR_sim is the predicted tidal range output by the model. The scale is obtained by dividing the two. This coefficient represents the overall scaling deviation of the predicted tidal range: if scale is greater than 1, it means that the model prediction continuously underestimates the tidal range; if it is less than 1, it means that the model systematically overestimates the tidal range. Subsequent corrections will use this coefficient to uniformly scale the fluctuation amplitude of all predicted tidal levels.

[0054] The datum correction factor shift: MTL_obs is the measured tidal mean sea level across zero, and MTL_sim is the predicted tidal mean sea level across zero. The calculation first multiplies the predicted datum by the amplitude factor scale to correct for tidal scaling. Then, the measured datum is subtracted from the scaled predicted datum to obtain the datum offset shift. This factor corrects for the systematic shift in mean sea level, ensuring that the overall water level reference remains consistent with the measured value after tidal scaling.

[0055] The initial phase correction value phase_0: T_HW and T_LW represent the times of high tide and low tide, respectively. First, the measured and predicted high and low tide times are summed, then the difference is calculated, and finally the average is taken to obtain the initial phase difference. By jointly calculating the high and low tide times, the interference caused by the error of a single extreme value detection can be reduced, obtaining the initial value of the overall time offset of the predicted tidal wave that is lagging or leading the measured time. This serves as the physically reasonable initial value for subsequent fine phase optimization.

[0056] In one implementation, a second phase of fine-grained search based on error minimization is performed.

[0057] For example, the phase_0 obtained in the first stage is based on the average value of a finite number of peak and valley moments, and its accuracy is limited by the accuracy of peak and valley detection. Therefore, the second stage performs fine optimization based on it, which specifically includes the following process: Define the optimization interval and traversal step size. Centered on the physical coarse estimate phase_0, set a reasonable time search range of ±3 hours (±10800 seconds). Every 5 minutes (300 seconds), take a candidate phase offset Δphase for traversal calculation.

[0058] Tide level correction is performed on each candidate phase. For each candidate offset Δphase, three corrections to the forecast tide level are performed simultaneously: the total phase offset phase_0+Δphase is superimposed on the forecast time series to achieve time series translation, and the tidal range is scaled using the amplitude coefficient scale and the water level reference is corrected by superimposing the datum coefficient shift, thereby generating the corrected forecast tide level sequence corrected_sim.

[0059] The error calculation measures the matching effect. A tidal window of about 50 minutes in the past 24 hours is selected, and the root mean square error (RMS) between each set of corrected forecast sequences and the measured tidal level sequences is calculated. The smaller the error, the higher the degree of fit between the forecast and the actual measurement under that phase.

[0060] Determine the optimal phase and output the final phase value. Select the candidate offset that minimizes the RMS error as the optimal additional phase difference. Add it to the initial rough estimate phase_0 to obtain the final accurate phase correction coefficient phase.

[0061] For example, the above process belongs to a two-stage strategy of "physical coarse estimation + numerical fine optimization": the first stage calculates phase_0 based on tidal physical characteristics, locking the phase within a reasonable range that conforms to tidal laws, thus avoiding the numerical search from getting trapped in physically meaningless local optima; the second stage performs a small-step grid search within the reasonable range, slightly improving the matching accuracy. Choosing 5 minutes as the traversal step size avoids losing the optimal solution and reducing correction accuracy due to an excessively large step size, while also avoiding massive repetitive calculations due to an excessively small step size, thus balancing computational speed and correction accuracy.

[0062] Step S600 constructs a nested double-layer correction coefficient system.

[0063] In one implementation, the coefficients at the two levels being maintained correspond to different time scales and physical meanings, respectively.

[0064] For example, the Layer 1 diurnal variation coefficient (the difference between adjacent days) represents the tide level deviation of the current forecast day relative to the previous day's forecast. It is obtained by comparing the observed sequence with the previous day's forecast sequence. The validation function isgoodlst() uses a strict threshold, as shown in Table 2.

[0065] Table 2 For example, the Layer2 cumulative coefficient (cumulative deviation relative to astronomical tides) is the cumulative deviation of the forecast day relative to the original astronomical tide (output of the DTU10 global tide model). It is obtained by chaining the diurnal variation coefficient and the historical cumulative coefficient. The validation function isgood() uses a lenient threshold, as shown in Table 3.

[0066] Table 3 For example, the chain compounding formula: scale_cum(today)=scale_day×scale_cum(yesterday); shift_cum(today)=shift_day+shift_cum(yesterday)×scale_day; phase_cum(today)=phase_day+phase_cum(yesterday).

[0067] For example, the step shift_cum(yesterday)×scale_day in the composite formula of shift reflects the "scaling effect" of scale correction on datum offset, that is, the datum offset accumulated on the previous day needs to be adjusted accordingly on the scale after the tidal range correction on the current day.

[0068] For example, phase and scale are simple addition and multiplication accumulations.

[0069] For example, a strict threshold is used for the daytime coefficient layer because the tidal driving conditions (changes in celestial positions) between physically adjacent days are very slow, and the tidal characteristics should not have drastic abrupt changes. That is, if the calculated daytime variation exceeds the strict threshold, it is almost certainly due to data quality issues or algorithmic anomalies, and should be rejected and reverted to historical coefficients.

[0070] For example, a lenient threshold is used for the cumulative coefficient layer because long-term accumulated deviations may reflect real systematic errors (such as systematic deviations of the DTU10 model in a specific sea area, parameter deviations of the FVCOM model, etc.) and should be tracked and corrected.

[0071] For example, the fallback strategy is as follows: when the daytime coefficient is invalid, the most recent valid daytime coefficient is searched from the database (corrC table) day by day and used as the coefficient for the current day. If no historical valid coefficients are found in the database and the system is not in restart mode (restart=false), then the neutral coefficients (scale=0, shift=0, phase=0) are used, which is equivalent to not making any corrections to the astronomical tides and relying only on the original DTU10 tide level driving model.

[0072] Step S700 performs multi-station weighted synthesis.

[0073] In one implementation, when deployed at multiple tide level observation stations, each station independently runs a complete correction analysis process to obtain its own diurnal variation coefficient. The final comprehensive coefficient used for open boundary correction is obtained by weighted averaging, specifically including the following steps: Identify and mark failed stations. Use [999.0, 999.0, 999.0] as placeholders for invalid coefficients in advance. Judge the three sets of correction coefficients output by each station one by one. If all the coefficients of a station are equal to the placeholder values, it means that the data of the station is abnormal and no valid correction result can be calculated. Mark the station as an invalid station.

[0074] The weights of valid sites are renormalized, all stations marked as invalid are removed, and only the preset weights of stations with normal output valid coefficients are retained. The weights of the remaining valid sites are divided by the sum of the weights to complete the weight renormalization, ensuring that the sum of the weights of all valid sites is 1, thus eliminating the interference of invalid sites on the overall results.

[0075] The weighted fusion yields a comprehensive correction coefficient for the entire sea area. The magnitude coefficient (scale), datum coefficient (shift), and phase coefficient (phase) of all valid stations are then summed by dot product with the normalized station weights to finally output a set of comprehensive correction coefficients adapted to the entire sea area.

[0076] For example, when individual stations have missing observations or abnormal data failures, the failed stations are automatically removed, and only normal stations participate in the calculation, without the need for manual removal of faulty stations; as long as there is at least one valid station, the correction parameters can be output normally, realizing continuous fault tolerance and degradation of multiple stations.

[0077] For example, the weight of each station is pre-configured by the user based on the station's representativeness and data quality, reflecting the station's ability to represent the tidal characteristics of the overall sea area.

[0078] Step S800: Soft start and gradual transition.

[0079] In one implementation, when applying the cumulative coefficient to the open-boundary tidal sequence of the model, it is not possible to directly jump from the coefficient of the previous period to the new coefficient in the first time step—this would trigger a transient shockwave in the numerical model, which would not only cause instability in the model calculation but also produce a physically unrealistic spurious flow field.

[0080] For example, a linear gradual transition strategy is adopted for the first 6 time steps, specifically including the following process: The first 6 time steps involve a linear gradual transition (t ranges from 0 to 5): The amplitude coefficients scale and shift coefficients are gradually changed. Starting with the old correction coefficients scale_old and shift_old, and ending with the new target coefficients scale_new and shift_new, the numerical transition is completed linearly and uniformly within 6 time steps. The gradual transition weight for each step is (t+1) / 6, and the proportion of the new coefficients is gradually increased as the time step increases. The original open boundary tide level is then corrected using the current gradually changed coefficients to obtain the smoothed tide level value obc_corrected[t].

[0081] After the 6th time step, the new coefficients are fully utilized (t≥6): After completing the 6-step smooth transition, the original tide level is fixedly corrected by the brand-new correction coefficients scale_new and shift_new throughout the entire process, without any further gradual changes.

[0082] For example, if the coefficients are switched abruptly and instantaneously, the tidal boundary will change abruptly, generating false shock waves and abnormal flow fields in the numerical model. A gradual, six-step linear transition to switch between old and new coefficients eliminates the non-physical numerical disturbances caused by abrupt boundary changes, ensuring the computational stability of ocean numerical models such as FVCOM. Note: Phase correction is a uniform time shift of the entire time series and is not involved in this step-by-step transition logic.

[0083] For example, a transition length of 6 steps (approximately corresponding to a time window of 6 times the external model time step in the FVCOM model) is selected, which ensures the smoothness of the transition without excessively delaying the effectiveness of the new coefficients; the phase correction is applied to the systematic translation of the entire time axis (obc.time=dtu10.time+phase_lag_cum) and does not participate in the gradual change. That is, because abrupt changes in the time axis will also introduce non-physical numerical effects, the phase correction is applied once at the sequence level, rather than gradually changing step by step.

[0084] In summary, this invention constructs a fully autonomous decision-making architecture from perception to correction decision-making. It establishes a complete autonomous decision-making chain from raw measured data input to final correction boundary output, with each decision node automatically completed by the algorithm: (a) autonomous judgment of data availability (adaptive sampling step size detection, sequence integrity check); (b) autonomous screening of peak and valley signals (automatic exclusion of abnormal peaks and valleys in meteorological tides through template matching verification, eliminating the need for manual identification); (c) autonomous calculation and verification of correction coefficients (each of the three coefficients is calculated and verified independently; automatic downgrading is triggered if verification fails); (d) autonomous selection of correction strategies (valid daytime coefficients → normal correction; invalid → backtracking to the most recent valid historical coefficients; no valid historical coefficients and not in restart mode → downgrading to a neutral value, waiting for the next round of data recovery). This architecture, for the first time, transforms multiple key decision-making steps in traditional tide level correction that require expert judgment based on experience into automatically executable algorithmic logic, providing a fundamental guarantee for achieving unmanned operation "from perception to decision-making".

[0085] This invention designs a reduced-order physical constraint semi-diurnal tide model fitting method. Using the full diurnal tide cycle (89400 seconds) and the semi-diurnal tide cycle (44700 seconds) as hard constraint frequencies, a simplified physical model with only 5 parameters is constructed: f(x) = a0 + a1·cos(2π / 89400·x+t1) + a2·cos(2π / 44700·x+t2). This completely abandons the dozens of tidal decomposition paradigms of traditional harmonic analysis and solves the problem of the inability to effectively separate tidal constituents in short-term (2-3 day) sequences from the principle level. This is the core theoretical basis of this method.

[0086] This invention designs a short-sequence filtering preprocessing method with integer multiple tidal period padding. Before low-pass filtering, the length of the tidal sequence is padded to an integer multiple of the lunar day (89400 seconds)—data from the previous period is appended to the end of the copied sequence—and the padded portion is truncated after filtering. This method utilizes the quasi-periodicity of tidal signals to avoid artificial signal distortion caused by traditional zero-padding / mirror-padding, and solves the boundary effect problem of short-sequence digital filtering.

[0087] This invention designs a template-guided method for screening the effectiveness of astronomical tide peaks and valleys. Using the peaks and valleys of the semi-diurnal tide fitting curve as the "astronomical tide template," it performs type consistency checks and time window (±3 hours) matching verification on the peaks and valleys in actual data, automatically eliminating abnormal peaks and valleys caused by non-astronomical factors such as meteorological forcing. This method achieves effective separation of astronomical tide signals from meteorological tide noise in short-term sequences, ensuring that the correction coefficients only reflect the systematic bias of the astronomical tides.

[0088] This invention designs a two-stage phase difference optimization method. The first stage calculates a physical coarse estimate of the phase difference based on the average difference between the measured and predicted fluctuation rest times. The second stage, centered on the coarse estimate, performs a grid search within a ±3-hour range with a step size of 300 seconds, using the minimization of the RMS error between the corrected predicted and measured sequences as the objective function to obtain the optimal phase difference. This method balances physical rationality and numerical accuracy.

[0089] This invention designs a nested two-layer correction coefficient system and a chain-compound method, comprising two layers: a daily coefficient (strict threshold verification) and a cumulative coefficient (relaxed threshold verification). The daily coefficient captures short-term fluctuations between adjacent days, while the cumulative coefficient tracks long-term drift through a chain-compound formula. The two layers are linked by the mathematical relationships: scale_cum = scale_day × scale_prev, shift_cum = shift_day + shift_prev × scale_day, and phase_cum = phase_day + phase_prev.

[0090] This invention designs an adaptive sampling step size detection method based on the mode statistics of time intervals. By statistically analyzing the frequency distribution of the time difference between adjacent tide level observations, the actual sampling step size is automatically identified, enabling the correction system to be independent of the sampling frequency of a specific acquisition device and possessing cross-platform versatility.

[0091] This invention designs a multi-station joint degradation correction method based on weight renormalization. Each station independently calculates the correction coefficient, automatically removes stations marked as invalid (999.0 placeholder value), and renormalizes the weights of the remaining valid stations before performing a weighted average, thereby achieving continuous degradation from all stations being normal to only a single station being valid.

[0092] This invention designs a soft-start, gradual transition boundary condition smoothing application method. In the first 6 time steps of the open boundary tidal level sequence of the model, the correction coefficient is smoothly transitioned from the previous period value to the new value through linear interpolation, eliminating the numerical shock wave caused by the sudden change of coefficients and ensuring the computational stability of numerical models such as FVCOM.

[0093] This invention designs a morphological calculation method for three physically interpretable correction coefficients: the amplitude coefficient (scale) is determined by the ratio of the measured to the predicted tidal range, the datum coefficient (shift) is determined by the linear relationship between the high tide level (crossing zero point) (considering the scale effect), and the phase coefficient (phase_lag) is determined by the average difference between the high tide and low tide times. Each coefficient has a clear physical meaning, an independent calculation method, and an independent validity judgment criterion.

[0094] This invention designs an automatic feedback closed-loop architecture: field station data → database management → calibration analysis engine → boundary condition correction → FVCOM numerical forecast → SFTP encrypted distribution → downstream users → new field acquisition → recalibration, forming a continuous operation closed loop that requires no manual intervention.

[0095] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0096] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0097] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A tidal level boundary correction and prediction method integrating measured data feedback and tidal range analysis, characterized in that, Includes the following steps: Based on the preprocessed tidal signal, the tidal morphology features are extracted after sequentially performing the reduced-order physical constraint semi-diurnal tide model fitting and the peak-valley effectiveness screening guided by the fitting template. Based on the tidal morphological characteristics, amplitude correction coefficient, datum correction coefficient, and initial phase correction value are extracted. After optimizing the initial phase correction value, diurnal variation coefficient and cumulative coefficient are obtained to perform tidal boundary correction forecast.

2. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis as described in claim 1, characterized in that: When preprocessing the tide level signal, integer multiple tide period completion and low-pass filtering are used to perform tide level signal preprocessing.

3. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis as described in claim 1, characterized in that: When performing a reduced-order physical constraint semi-diurnal tide model fitting, the reduced-order physical constraint semi-diurnal tide model represents: f(x)=a0+a1·cos(2π / 89400·x+t1)+a2·cos(2π / 44700·x+t2) Where: x represents the time from the origin of the timekeeping; a0 represents the mean sea level offset; a1 and t1 represent the amplitude and initial phase of the diurnal tide component, respectively, with a fixed period of 89400 seconds; a2 and t2 represent the amplitude and initial phase of the semi-diurnal tide component, respectively, with a fixed period of 44700 seconds.

4. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis as described in claim 1, characterized in that: When performing peak and valley validity screening guided by the fitting template, peak and valley detection, duplicate peak and valley merging, intersection of two sequences, and matching verification based on the fitting template are performed in sequence to perform peak and valley validity screening guided by the fitting template.

5. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis as described in claim 4, characterized in that: When performing matching verification, if there are no matching items, it means that the measured tide level during the intersection period of the two sequences is severely affected by meteorological disturbances and there is no corresponding astronomical tide extreme value. This is judged as an abnormal period, and all measured peaks and valleys in the data segment area of ​​the intersection period of the two sequences are invalidated and will not be included in subsequent correction calculations. If only one match is found, the measured peak and valley are consistent with the theoretical astronomical tide law and are considered valid astronomical tide signals, which are retained for subsequent feature extraction. If two or more extreme values ​​of the same type are matched, the occurrence of multiple extreme values ​​of the same type within the same tidal cycle is considered noise interference. Only the highest water level is retained for peak values ​​and only the lowest water level is retained for valley values. Other adjacent extreme values ​​are discarded.

6. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis as described in claim 1, characterized in that: When extracting the amplitude correction coefficient, the measured tidal range and the tidal range output by the forecast model are used to obtain the amplitude correction coefficient.

7. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis as described in claim 1, characterized in that: When extracting the datum correction coefficient, based on the measured average tidal level across zero and the predicted average tidal level across zero, the predicted datum is first multiplied by the amplitude correction coefficient to complete the tidal range scaling correction, and then the measured datum is subtracted from the scaled predicted datum to obtain the datum correction coefficient.

8. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis according to claim 1, characterized in that: When extracting and optimizing the initial phase correction value, based on the occurrence times of high tide and low tide, the measured and predicted high and low tide times are summed and then the difference is calculated. Finally, the average value is taken to obtain the initial phase correction value. The initial phase correction value is then optimized using a fine search based on minimizing error.

9. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis as described in claim 1, characterized in that: When performing tide boundary forecasting, in response to multiple tide observation stations, the weighted average of the diurnal variation coefficients of each station is used to obtain a comprehensive coefficient for open boundary correction, so as to perform tide boundary correction forecasting.

10. The tidal level boundary correction and prediction method based on the fusion of measured data feedback and tidal range analysis according to claim 1, characterized in that: When performing tidal boundary forecasting, a linear gradual transition strategy for the first 6 time steps is adopted based on the cumulative coefficient to perform tidal boundary correction forecasting.

Images