A machine learning-based flood peak flow inversion method

Abstract

The application discloses a flood peak flow inversion method based on machine learning and relates to the technical field of flood peak flow data processing. The flood peak flow inversion method based on machine learning comprises the following steps: hydrological signal interference monitoring, hydrological signal transmission monitoring and hydrological signal data quality monitoring. The application analyzes hydrological signal interference characteristics to determine whether a dynamic anti-interference strategy is adopted, then performs flood peak hydrological signal transmission integrity checking to determine whether transmission link optimization is adopted, and finally performs hydrological data quality evaluation to determine whether flood peak inversion is performed, so that the accuracy of flood peak flow inversion results is improved, and the problem of low accuracy of flood peak flow inversion results in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of flood peak flow inversion data processing technology, and in particular to a flood peak flow inversion method based on machine learning. Background Technology

[0002] Because many river basins currently lack long-term, continuous direct observation stations for peak flood discharge, and historical flood events often lack complete discharge records, it is impossible to directly obtain peak flood discharge data to support flood risk assessment and prevention. Therefore, there is an urgent need to achieve peak flood discharge inversion. Peak flood discharge inversion typically refers to extrapolating past peak flood discharges using observed flood discharge data (such as hydrological data) or pre-set models, or estimating the peak flood discharge for a specific area based on a pre-set machine learning model. The specific process is as follows: First, collect basic data related to peak flood discharge within the river basin, including historical hydrological data (such as water level and discharge observation records), meteorological data (such as rainfall and duration), and topographic data (such as basin area and slope). Then, perform preprocessing operations on these basic data (such as filling missing values, removing outliers, and data normalization) to construct a standardized peak flood discharge dataset. Next, input the standardized peak flood discharge dataset into a pre-set machine learning model, such as LSTM (Long Short-Term Flow Mechanism). Using methods such as Long Short-Term Memory (LSTM) networks, random forests, and support vector machines, peak flow inversion is performed within a pre-defined machine learning model to obtain the corresponding peak flow inversion results, ultimately forming a stable and reliable peak flow inversion method.

[0003] In the process of monitoring river flood flow, firstly, the flood levels of the river basin are classified by combining historical rainfall data and flood conditions, and flood process lines of different levels are drawn, and the 24-hour design rainfall for each level of flood is determined. Next, a one-dimensional hydrodynamic model of the river is established using cross-sectional survey data, and the water level-water surface width relationship of the cross-section to be measured is constructed. By solving the model, the water level-discharge relationship curves of each level of flood during the rising and receding stages are obtained. Then, the recurrence level of the current flood is identified based on real-time remote sensing images, and the current water surface width of the cross-section to be measured is calculated, thereby inverting the corresponding water level. Finally, according to the flood level and flood flow process stage (such as rising or receding water) of the river, the corresponding water level-discharge relationship curve is selected and substituted to obtain the flood flow of the cross-section for the current period. Thus, the accuracy of satellite remote sensing inversion of peak flow is improved through multi-dimensional data fusion.

[0004] For example, Chinese invention patent application CN117743862A discloses a method and system for calculating peak flow in urban small watersheds, which includes: first, acquiring flow data of the monitored watershed (such as a small river watershed), and simultaneously collecting data on the catchment area, area, and topographic distribution of the small watershed; next, selecting a peak flow calculation formula based on the collected catchment area and topographic conditions; then, substituting the known data into the peak flow calculation formula to obtain the initial peak flow; finally, internally verifying the initial peak flow and comparing it with data from similar watersheds in the surrounding area, and determining the final peak flow after this verification and optimization process.

[0005] The above-mentioned technology has at least the following technical problems:

[0006] In some torrential flood events, lightning and flood peak phases may partially overlap, leading to strong electromagnetic disturbances (such as lightning discharges, induced currents, and power / communication system noise) during hydrological data acquisition for flood peak flow inversion. These disturbances disrupt the original signal transmission of hydrological sensors (such as water level gauges, rain gauges, and current meters). Current mainstream engineering applications primarily rely on fixed hardware parameters and lack real-time adaptive algorithms, potentially causing signal attenuation and distortion in the collected hydrological data (such as water level and rainfall), and signal interruption during critical flood peak periods. When training machine learning models for flood peak flow inversion based on such interfered signal data, the presence of numerous spurious interference signals may prevent the models from capturing the true correlation between flood peak flow, rainfall, and topography, ultimately resulting in low accuracy of the flood peak flow inversion results. Summary of the Invention

[0007] To address the technical problem of low accuracy in peak flow inversion results in existing technologies, this invention provides a machine learning-based peak flow inversion method. This method includes: during peak flow inversion, performing hydrological signal interference feature analysis to assess the degree of strong magnetic interference affecting the hydrological signal, and determining whether dynamic anti-interference strategies are needed to ensure the authenticity and effectiveness of the hydrological signal; after the hydrological signal interference feature analysis is deemed satisfactory, performing a peak hydrological signal transmission integrity check to verify the integrity, temporal continuity, and core feature fidelity of data transmission during the core peak period, and determining whether transmission link optimization is needed to reduce the transmission distortion rate of the peak core data features during hydrological data transmission; after the peak hydrological signal transmission integrity check is deemed satisfactory, performing a hydrological data quality assessment to quantify the accuracy of matching hydrological data with the peak core features, and determining whether peak flow inversion is necessary. Peak flow inversion is used to quantify key peak characteristic parameters to improve the accuracy of peak prediction.

[0008] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0009] 1. By analyzing the interference characteristics of hydrological signals and determining whether dynamic anti-interference strategies are needed, it is helpful to specifically eliminate strong impulse noise, high-frequency oscillation noise, and baseline drift noise, ensuring the authenticity and integrity of core features of hydrological signals. This provides reliable raw data support for subsequent flood peak hydrological signal transmission integrity verification and hydrological data quality assessment. After the hydrological signal interference characteristic analysis is qualified, flood peak hydrological signal transmission integrity verification is performed, and it is determined whether transmission link optimization is needed. This helps to reduce problems such as sampling loss, residual errors, and feature distortion during hydrological data transmission, ensuring that key time-series data during the core period of the flood peak is complete and unbiased. After the flood peak hydrological signal transmission integrity verification is qualified, hydrological data quality assessment is performed, and it is determined whether flood peak flow inversion is needed. This helps to quantitatively screen out high-quality hydrological data with accuracy and consistency, accurately determine the feasibility of flood peak flow inversion, reduce flood peak flow misjudgment caused by low-quality data, and ultimately improve the accuracy of flood peak flow inversion results.

[0010] 2. By acquiring noise mode characteristic parameters covering the proportion of hydrological signal interference pulse duration, hydrological signal oscillation frequency, and hydrological signal baseline drift ratio, three core dimensions of hydrological signal interference—time proportion, frequency characteristics, and trend shift—can be captured simultaneously. This avoids the limitations of single-parameter evaluation in existing technologies. Each parameter is quantified in a targeted manner to accurately quantify the degree of composite interference in hydrological signals, effectively solving the problem of the lack of multi-dimensional interference discrimination in existing technologies. Based on the noise mode characteristic parameters, strong impulse noise intervention conditions are identified and filtering is implemented according to different situations, forming an automated closed loop from interference feature extraction to targeted processing. This fills the gap in existing technologies that lack multi-dimensional self-inspection, classification, and precise intervention mechanisms for hydrological signal interference, providing a stable and reliable preliminary foundation for subsequent flood peak hydrological signal transmission integrity verification and high-quality hydrological data acquisition, and improving the accuracy and reliability of hydrological data processing and flood peak flow inversion.

[0011] 3. Targeted selection of indicators for hydrological data integrity, accuracy, and consistency helps to comprehensively capture the three core dimensions of hydrological data quality: transmission integrity, data accuracy, and time-series consistency. This fully covers the multifaceted requirements of flood peak flow inversion for data quality, and addresses the problems of existing technologies that often rely on a single quality dimension for evaluation, fail to consider the combined effects of multiple dimensions, and are prone to biased quality assessments and omissions of key quality information. By weighting and coupling the quantification parameters of hydrological data quality with corresponding hydrological data quality weighting parameters, a comprehensive quantitative hydrological data quality assessment index for hydrological data integrity, accuracy, and consistency is obtained, which helps to reflect... The study identifies the differences in the actual impact of different quality dimensions on the overall quality assessment results, reduces quality assessment bias caused by quantification of a single parameter, and determines whether the hydrological data quality assessment index exceeds the preset data quality threshold. If so, the corresponding hydrological data is marked as qualified hydrological data and flood peak flow inversion is performed; otherwise, a warning of unqualified hydrological data quality is issued. This helps to establish a closed-loop mechanism from hydrological data quality assessment to inversion decision-making, ensuring that only hydrological data that meets the quality requirements enters the subsequent flood peak flow inversion stage. This reduces problems such as distortion of inversion results and deviation in flood control scheduling decisions caused by low-quality data directly participating in flood peak flow inversion, and provides reliable data support for the accuracy of flood peak flow inversion. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0013] Figure 1 This is a flowchart of a peak flow inversion method based on machine learning provided in an embodiment of the present invention;

[0014] Figure 2 This is a general overview of a machine learning-based peak flow inversion method provided in this embodiment of the invention. Figure 1 ;

[0015] Figure 3 This is a general overview of a machine learning-based peak flow inversion method provided in this embodiment of the invention. Figure 2 ;

[0016] Figure 4 This invention provides a dynamic anti-interference strategy logic for a machine learning-based peak flow inversion method. Figure 1 ;

[0017] Figure 5This invention provides a dynamic anti-interference strategy logic for a machine learning-based peak flow inversion method. Figure 2

[0018] Figure 6 This is a comparison chart of the predicted flow rates using a machine learning model for a peak flow inversion method based on machine learning, as provided in this embodiment of the invention. Detailed Implementation

[0019] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0020] like Figure 1 As shown in the flowchart of the peak flow inversion method based on machine learning provided in this application embodiment, the flowchart of the peak flow inversion method based on machine learning includes: First, hydrological signal interference monitoring, performing hydrological signal interference characteristic analysis to assess the degree of strong magnetic interference on hydrological signals, and determining whether dynamic anti-interference strategies need to be adopted to specifically suppress different types of interference, so as to ensure the authenticity and effectiveness of hydrological signals; through hydrological signal interference monitoring, it is helpful to filter invalid interference signals from the source, provide a clean hydrological signal base for subsequent hydrological signal transmission monitoring and hydrological data quality assessment, and reduce the chain hydrological data distortion caused by the step-by-step transmission of strong magnetic disturbances.

[0021] Secondly, hydrological signal transmission monitoring, after the hydrological signal interference characteristic analysis is qualified, is used to verify the integrity, temporal continuity, and core feature fidelity of the data transmission during the core period of the flood peak. It is also used to determine whether transmission link optimization is needed to reduce the transmission distortion rate of the core data features of the flood peak during the data transmission process, and to ensure that the key time series data required for flood peak inversion are complete and unbiased. Through hydrological signal transmission monitoring, it is helpful to build a stable hydrological signal transmission link, completely preserve the original form of the core time series characteristics of the flood peak, and provide temporally continuous and feature-undistorted transmission data for hydrological data quality assessment.

[0022] Finally, hydrological signal data quality monitoring involves assessing the accuracy, consistency, and usability of the hydrological data's matching with the core characteristics of the flood peak after the integrity of the flood peak hydrological signal transmission is verified. This assessment determines whether flood peak flow inversion is necessary. Flood peak flow inversion is used to quantify key characteristic parameters of the flood peak, reconstruct the temporal evolution of the flood peak, capture the core dynamic information of the flood peak, and improve the accuracy of flood peak prediction. Through hydrological signal data quality monitoring, high-quality data that meets the requirements of flood peak inversion can be accurately selected, ensuring the reliability of the inversion process from a data perspective and providing high-quality hydrological data for the accurate extraction and pattern analysis of the core characteristics of the flood peak.

[0023] It is important to note that the flood peak flow inversion method based on machine learning provided in this application is implemented with a database that integrates key parameters and reference information. The information sources of this database include basic parameters pre-set by technicians, such as preset hydrological signal peak values, preset strong magnetic disturbance thresholds, and preset pulse duration ratio thresholds. It also includes reference datasets compiled from historical watershed flood peak cases (such as flow inversion records under different climatic conditions and feature parameter correlation data) and hydrological simulation experimental data (such as model error test results under multiple scenarios and feature importance verification data). This provides practical evidence for the scientific validity of the basic parameters. The database storage adopts an encrypted distributed architecture, combining a relational database to store structured preset parameters (such as the specific values ​​of various thresholds and the value range of feature weights) and a non-relational database to store unstructured reference information (such as historical flood peak flow time series maps and model training loss curves). Technicians can dynamically adjust the preset parameters based on newly accumulated watershed hydrological data and feedback on inversion results, so that the stored information continuously adapts to the needs of actual watershed flood peak inversion scenarios.

[0024] In this embodiment, a comprehensive flood peak flow inversion data assurance system is constructed through hydrological signal interference monitoring, hydrological signal transmission monitoring, and hydrological signal data quality monitoring. This system covers the entire process from suppressing strong magnetic disturbances at their source and ensuring the transmission of hydrological signals to verifying the quality of hydrological data. Hydrological signal interference monitoring specifically weakens the pollution of hydrological signals by various interferences such as strong magnetic interference, high-frequency oscillations, and baseline drift at the signal source, reducing subsequent inversion deviations caused by initial signal distortion. Hydrological signal transmission monitoring solves the problems of data loss, bit errors, and feature distortion during signal transmission, balancing data integrity and transmission efficiency in the hydrological data transmission link. Hydrological signal data quality monitoring promptly identifies residual quality defects or data anomalies caused by sudden interference after preceding links. The three systems work together to effectively improve the authenticity, integrity, and usability of flood peak flow inversion data.

[0025] like Figure 2 As shown, this application provides a general overview of a machine learning-based peak flow inversion method. Figure 1 ,Depend on Figure 2 It can be seen that: hydrological signal interference characteristic analysis is performed, and strong magnetic disturbance quantification index is obtained to determine whether the hydrological data interference verification is qualified. If the verification is qualified, the integrity verification of the peak hydrological signal transmission is performed; otherwise, a dynamic anti-interference strategy is adopted. After the dynamic anti-interference strategy ends, the hydrological signal interference characteristic analysis is determined to be qualified. If it is not qualified, a filtering failure warning is sent; otherwise, the integrity verification of the peak hydrological signal transmission is performed, and the hydrological signal transmission integrity verification parameters are obtained. It is determined whether the hydrological signal transmission integrity verification parameters meet the data transmission integrity discrimination conditions. If they meet the conditions, hydrological data quality assessment is performed; otherwise, transmission link optimization trigger discrimination is performed.

[0026] like Figure 3 As shown, this application provides a general overview of a machine learning-based peak flow inversion method. Figure 2 ,Depend on Figure 3 It is known that: transmission link optimization is triggered and the accumulated value of the transmission verification counter is obtained. It is determined whether the accumulated value of the transmission verification counter is greater than the preset maximum threshold of the transmission verification counter. If not, continuous judgment of the integrity of the peak hydrological signal transmission is performed. Otherwise, transmission link optimization is performed. After the transmission link optimization is completed, it is determined whether the integrity of the peak hydrological signal transmission is qualified. If not, a link optimization failure warning is sent. Otherwise, hydrological data quality assessment is performed and hydrological data quality assessment indicators are obtained. It is determined whether the hydrological data quality assessment indicators are greater than the preset data quality threshold. If so, peak flow inversion is performed. Otherwise, a hydrological data quality failure warning is sent.

[0027] Furthermore, the specific process of hydrological signal interference characteristic analysis is as follows: Quantification parameters of strong magnetic disturbance are obtained. These parameters include: the proportion of strong magnetic disturbance based on the peak value of the hydrological signal, used to assess the degree of interference of strong magnetic disturbance on the peak flood peak characteristics; the proportion of signal distortion based on the hydrological signal, used to measure the distortion level of the hydrological signal's temporal regularity and amplitude accuracy caused by strong magnetic disturbance; and the proportion of interference signal based on the duration of invalid signals, used to quantify the continuous influence range of strong magnetic disturbance. The proportion of strong magnetic disturbance is represented by the average calculation result after quantifying the proportion of the hydrological signal amplitude to the preset hydrological signal amplitude during the hydrological signal interference characteristic analysis period. The proportion of signal distortion is represented by the mean square error of the hydrological signal amplitude monitored by the hydrological signal amplitude monitor and the preset hydrological signal amplitude during the hydrological signal interference characteristic analysis period, compared with the preset hydrological signal amplitude. The amplitude standard deviation of the signal is used to represent the proportion of the quantified result; the proportion of the interference signal is represented by the proportion of the interference signal duration to the total hydrological signal duration during the hydrological signal interference characteristic analysis period; the harmonic averaging process is performed based on the strong magnetic disturbance quantification parameters, and the specific process is as follows: the result of the harmonic averaging process of the strong magnetic disturbance quantification parameters is used as the strong magnetic disturbance quantification index to evaluate the degree of interference of the hydrological signal by strong magnetic disturbance; hydrological data interference verification is performed based on the strong magnetic disturbance quantification index. If the verification is qualified, the integrity verification of the flood peak hydrological signal transmission is performed. If the verification is unqualified, a dynamic anti-interference strategy is adopted; the specific process of hydrological data interference verification is as follows: it is determined whether the strong magnetic disturbance quantification index is greater than the preset strong magnetic disturbance threshold. If it is, the verification is deemed unqualified, otherwise, the verification is deemed qualified. The preset strong magnetic disturbance threshold is represented by the average value of the strong magnetic disturbance quantification index over a historical time period.

[0028] Specifically, the formula for the proportion of strong magnetic interference is as follows:

[0029]

[0030] Where R1 represents the proportion of strong magnetic interference, j represents the sampling point number of the hydrological signal amplitude, j=1,2,3...m, m represents the total number of hydrological signal amplitude sampling points, P j P0 represents the hydrological signal amplitude at the j-th sampling point, and P0 represents the preset hydrological signal amplitude. Specifically, the hydrological signal amplitude is obtained by the hydrological signal amplitude monitor, and the preset hydrological signal amplitude is represented by the average value of the hydrological signal amplitude over a historical time period.

[0031] Specifically, the formula for the percentage of signal distortion is as follows:

[0032]

[0033] Wherein, R2 represents the proportion of signal distortion, and σ0 represents the preset standard deviation of hydrological signal amplitude. Specifically, the preset standard deviation of hydrological signal amplitude is represented by the average value of the standard deviation of hydrological signal amplitude over a historical time period.

[0034] Specifically, the formula for the proportion of interference signals is as follows:

[0035]

[0036] Where R3 represents the proportion of interference signal, T inter T represents the duration of the interference signal. total The total hydrological signal duration is represented by the cumulative duration monitored by the hydrological signal duration monitor at each sampling point where the signal amplitude is outside the preset effective signal amplitude range. The total hydrological signal duration is represented by the duration obtained by the hydrological signal duration monitor.

[0037] Specifically, the formula for the quantitative index of strong magnetic disturbance is as follows:

[0038]

[0039] Among them, Q in This represents a quantitative indicator of strong magnetic disturbance.

[0040] In this embodiment, through hydrological signal interference feature analysis, three core disturbance features under strong magnetic interference were accurately captured: peak deviation of hydrological signals, distortion of hydrological signals, and duration of invalid hydrological signals. This formed a quantitative index and verification closed loop, which helps to trigger dynamic anti-interference strategies in a targeted manner, achieve targeted suppression of different types of strong magnetic disturbances, reduce the risk of distortion of core features of hydrological signals caused by strong magnetic interference, improve the purity and reliability of the original hydrological signals, and realize the quantitative assessment and automated processing of the degree of strong magnetic disturbance interference.

[0041] like Figure 4 As shown in the embodiment of this application, a dynamic anti-interference strategy logic for a machine learning-based peak flow inversion method is provided. Figure 1 ,Depend on Figure 4 It can be seen that: a dynamic anti-interference strategy is implemented, and noise mode characteristic parameters are obtained. It is then determined whether the noise mode characteristic parameters meet the strong impulse noise intervention conditions. If not, high-frequency oscillation noise discrimination is performed; otherwise, strong impulse noise filtering is performed. After the strong impulse noise filtering is completed, it is determined whether the hydrological data interference verification is qualified. If yes, the integrity verification of the flood peak hydrological signal transmission is performed; otherwise, high-frequency oscillation noise discrimination is performed. The noise mode characteristic parameters are then determined whether they meet the high-frequency oscillation noise intervention conditions. If not, baseline drift noise discrimination is performed; otherwise, high-frequency oscillation noise filtering is performed. After the high-frequency oscillation noise filtering is completed, it is determined whether the hydrological data interference verification is qualified. If yes, the integrity verification of the flood peak hydrological signal transmission is performed; otherwise, baseline drift noise discrimination is performed.

[0042] like Figure 5 As shown in the embodiment of this application, a dynamic anti-interference strategy logic for a machine learning-based peak flow inversion method is provided. Figure 2 ,Depend on Figure 5 It can be seen that: baseline drift noise discrimination is performed to determine whether the noise mode characteristic parameters meet the baseline drift noise intervention conditions. If not, a slight interference warning is sent; otherwise, baseline drift noise filtering is performed. After the baseline drift noise filtering is completed, the hydrological data interference verification is determined to be qualified. If not, a filtering failure warning is sent; otherwise, the integrity of the flood peak hydrological signal transmission is verified.

[0043] Furthermore, the specific process of the dynamic anti-interference strategy is as follows: Noise mode characteristic parameters are acquired, including the proportion of hydrological signal interference pulse duration used to quantify the sustained impact of strong pulse interference on hydrological signals, the hydrological signal oscillation frequency used to quantify the core frequency characteristics of high-frequency oscillation noise, and the hydrological signal baseline drift ratio used to assess the overall deviation of the hydrological signal baseline. The proportion of hydrological signal interference pulse duration is represented by the quantified result of the ratio of the cumulative duration of hydrological signal interference pulses to the total duration of the hydrological signal interference characteristic analysis period. The cumulative duration of hydrological signal interference pulses is represented by the result of the cumulative calculation of the duration of signal segments with hydrological signal amplitudes greater than the preset hydrological signal amplitude, monitored by the hydrological signal duration monitoring instrument, within the hydrological signal interference characteristic analysis period. The hydrological signal oscillation frequency is represented by the frequency of the hydrological signal oscillation during the hydrological signal interference characteristic analysis period, based on spectrum analysis (such as short-term...). The center frequency of the high-frequency interference signal extracted by Fourier transform is represented; the baseline drift ratio of the hydrological signal is represented by the result of quantifying the cumulative baseline offset and the preset standard deviation of the hydrological signal amplitude within the hydrological signal interference characteristic analysis period. The cumulative baseline offset is represented by the result of accumulating the deviation quantification between the hydrological signal baseline and the preset hydrological signal baseline. The hydrological signal baseline is represented by the trend line extracted based on the linear fitting algorithm, and the preset hydrological signal baseline is represented by the average value of the historical hydrological signal baseline. The deviation quantification is represented by the difference operation; it is determined whether the noise mode characteristic parameters meet the strong impulse noise intervention conditions. If so, strong impulse noise filtering is performed; otherwise, high-frequency oscillation noise discrimination is performed. The strong impulse noise intervention condition means that the proportion of hydrological signal interference pulse duration is greater than the preset pulse duration proportion threshold. The preset pulse duration proportion threshold is represented by the average value of the proportion of hydrological signal interference pulse duration in the historical time period.

[0044] Specifically, high-frequency oscillation noise discrimination involves determining whether the noise mode characteristic parameters meet the high-frequency oscillation noise intervention conditions. If so, the strong magnetic disturbance quantization index is reacquired and marked as the high-frequency oscillation strong magnetic disturbance quantization index. High-frequency oscillation noise filtering is then performed based on this index. Otherwise, baseline drift noise discrimination is performed. Baseline drift noise discrimination involves determining whether the noise mode characteristic parameters meet the baseline drift noise intervention conditions. If so, the strong magnetic disturbance quantization index is reacquired and marked as the baseline drift strong magnetic disturbance quantization index. Baseline drift noise filtering is then performed based on this index. If the signal is not properly handled, a minor interference warning will be issued; the high-frequency oscillation noise intervention condition indicates that the hydrological signal oscillation frequency is greater than the preset high-frequency oscillation threshold, and the duration of the high-frequency oscillation is greater than the preset oscillation duration threshold. The preset high-frequency oscillation threshold is represented by the average value of the hydrological signal oscillation frequency over a historical time period; the baseline drift noise intervention condition indicates that the baseline drift ratio of the hydrological signal is within the preset baseline drift ratio range, and the baseline offset is less than the preset drift rate threshold within the hydrological signal interference characteristic analysis time period. The preset baseline drift ratio range includes both ends of the baseline drift ratio range and is set in advance by preset personnel. The preset drift rate threshold is represented by the average value of the baseline offset over a historical time period.

[0045] Specifically, strong impulse noise filtering is used to specifically suppress continuous strong impulse interference caused by strong magnetic disturbances in hydrological signals, thereby improving the stability and reliability of hydrological signal amplitude. The specific process is as follows: Interference pulse marking is performed by comparing the hydrological signal amplitude within a preset time window with a preset hydrological signal amplitude one by one. When the corresponding hydrological signal amplitude is greater than the preset pulse amplitude, the corresponding hydrological signal sampling point is marked as a pulse interference point. The preset time window is pre-set by designated personnel. Median filtering is performed based on the pulse interference points until the median filtering of the hydrological signal within the entire preset time window is completed, obtaining the median filtered hydrological signal. After median filtering, a secondary correction process is performed on the median filtered hydrological signal based on nonlinear amplitude limiting filtering. After the secondary correction process, the strong magnetic disturbance quantification index is re-acquired, and hydrological data interference verification is performed. If the verification is successful, the integrity of the flood peak hydrological signal transmission is verified; otherwise, high-frequency oscillation noise is identified.

[0046] High-frequency oscillation noise filtering is used to specifically suppress high-frequency oscillation noise caused by strong magnetic disturbances in hydrological signals, improving the continuity of low-frequency trends and the fidelity of core features in hydrological signals. The specific process is as follows: The quantification index of high-frequency oscillation strong magnetic disturbance and the bandpass output signal frequency monitored by the signal frequency monitor are input into a preset oscillation frequency-notch filter mapping table in the database for lookup, obtaining the center frequency adjustment coefficient of the adaptive notch filter; The notch filter center frequency is then adjusted, specifically as follows: Using the amplitude corresponding to the center frequency adjustment coefficient as the adjustment step size, the center frequency of the adaptive notch filter is adjusted step by step in the direction of decreasing center frequency deviation. (After each center frequency adjustment, the residual high-frequency oscillation level is reacquired, and it is determined whether the residual level is less than the preset residual threshold. If not, the adjusted center frequency is used as the initial value for the next adjustment, and the adjustment continues step by step in the direction of decreasing center frequency deviation.) This helps to dynamically adapt to the actual frequency fluctuations of strong magnetic interference from high-frequency oscillations, reduce frequency deviation or over-suppression problems caused by one-time adjustments, and improve the filtering accuracy and stability of high-frequency oscillation noise. The center frequency deviation is represented by the absolute value of the deviation quantized between the current center frequency of the adaptive notch filter and the center frequency of the high-frequency oscillation. Continuous monitoring of high-frequency oscillation... The residual degree of high-frequency oscillation is considered. When the residual degree of high-frequency oscillation is less than the preset residual threshold, high-frequency oscillation noise filtering is performed on subsequent hydrological signals based on the center frequency of the current adaptive notch filter. After the high-frequency oscillation noise filtering is completed, hydrological data interference verification is performed. If the verification is successful, the integrity of the flood peak hydrological signal transmission is verified; otherwise, baseline drift noise is identified. The preset residual threshold is represented by the average value of the residual degree of high-frequency oscillation over a historical time period. When the residual degree of high-frequency oscillation is not less than the preset residual threshold, high-frequency oscillation noise filtering is continuously performed. When the number of high-frequency oscillation noise filtering operations exceeds the preset high-frequency oscillation threshold, the high-frequency oscillation noise filtering is further processed. If the hydrological data interference check fails during the maximum number of filtering operations, baseline drift noise discrimination will be performed. The maximum number of high-frequency oscillation filtering operations is preset by the preset personnel. The high-frequency oscillation residual degree is represented by the quantification result of the ratio of the energy value of the hydrological signal in the preset high-frequency oscillation target frequency range to the energy value in the same frequency range before filtering. This is used to quantify the actual suppression effect of high-frequency oscillation noise filtering on high-frequency interference in the target frequency range. The preset high-frequency oscillation target frequency range is preset by the preset personnel, and the energy value is represented by the signal energy quantization value obtained by spectrum analysis based on short-time Fourier transform.

[0047] Specifically, baseline drift noise filtering is used to specifically suppress baseline drift noise caused by strong magnetic disturbances in hydrological signals, improving the stability and baseline fit of the hydrological signal baseline. The specific process is as follows: The baseline drift strong magnetic disturbance quantification index and the hydrological signal baseline drift ratio are input into a preset baseline drift-filter correction parameter mapping table in the database for lookup, obtaining the moving average window adjustment coefficient of the moving average filter; the moving average window is then adjusted, specifically as follows: using the amplitude corresponding to the moving average window adjustment coefficient as the adjustment step size, the window size of the moving average filter is adjusted incrementally in the direction of increasing the moving average window size (after each adjustment of the moving average filter window size, the signal baseline fit is re-acquired). It determines whether the signal baseline fit is less than the preset baseline fit threshold. If not, the adjusted window size of the moving average filter is used as the initial value for the next adjustment, and the window size is gradually increased in the direction of the moving average window. This helps to dynamically adapt to the actual severity of hydrological signal baseline drift, reduce the distortion of core hydrological signal features (such as flood peak rise trend and peak relative to the reference position) caused by a one-time large adjustment of the window, and ensure the accuracy and stability of hydrological signal baseline correction. The signal baseline fit is continuously monitored. The signal baseline fit is calculated by taking the absolute value of the deviation between the hydrological drift signal baseline and the preset signal baseline at each sampling point, averaging all the absolute values ​​of the deviations, and then dividing the average by the preset signal amplitude standard deviation. The sampling point deviation represents the difference between the amplitude of the current hydrological drift signal baseline and the amplitude of the preset signal baseline at the same sampling time point. The hydrological drift signal baseline is represented by a trend line extracted by linear fitting after baseline drift noise filtering. The signal baseline fit reflects the suppression effect of baseline drift noise filtering on hydrological signal baseline drift, ensuring the consistency of the benchmark for subsequent hydrological data quality assessment and flood peak flow inversion. The preset signal amplitude standard deviation is represented by the average of the standard deviations of hydrological signal amplitudes over historical time periods. When the signal baseline fit is less than the preset baseline fit threshold, baseline drift noise filtering is continued on subsequent hydrological signals based on the current moving average window size. After wave processing, hydrological data interference verification is performed. If the verification is successful, the integrity of the flood peak hydrological signal transmission is verified; otherwise, a filtering failure warning is issued. The preset baseline fit threshold is represented by the average value of the signal baseline fit over a historical time period. When the signal baseline fit is not less than the preset baseline fit threshold, baseline drift noise filtering is continuously performed. When the number of times baseline drift noise filtering is performed exceeds the preset maximum number of times baseline drift noise filtering is performed, hydrological data interference verification is performed. If the verification fails, a filtering failure warning is issued; otherwise, the integrity of the flood peak hydrological signal transmission is verified. The preset maximum number of times baseline drift noise filtering is performed is preset by designated personnel.

[0048] It should be noted that in the embodiments of this application, when performing high-frequency oscillation noise filtering, baseline drift noise filtering, and transmission link optimization, the preset oscillation frequency-notch filter parameter mapping table, preset baseline drift-filter correction parameter mapping table, and preset strong magnetic disturbance-segment size mapping table are all pre-set by professional technicians and stored in the database. This provides precise support for matching corresponding adjustment parameters to the combination of high-frequency oscillation strong magnetic disturbance quantification index and bandpass output signal frequency, the combination of baseline drift strong magnetic disturbance quantification index and hydrological signal baseline drift ratio, and the combination of flood peak hydrological signal transmission integrity verification parameter and total amount of hydrological data. When the system performs the above processing and optimization, the appropriate adjustment coefficient or adjustment factor can be quickly retrieved from the corresponding mapping table to ensure that the adjustment range of noise filtering and link optimization accurately matches the actual strong magnetic disturbance scenario requirements, avoiding secondary problems such as signal characteristic distortion and transmission efficiency imbalance caused by insufficient or excessive adjustment.

[0049] Specifically, these mapping tables are constructed based on a large amount of effective data from historical hydrological monitoring scenarios, covering parameter combinations under different intensities of strong magnetic interference and different hydrological signal characteristics (including combinations of high-frequency oscillation strong magnetic disturbance quantification index, bandpass output signal frequency and corresponding center frequency adjustment coefficient, baseline drift strong magnetic disturbance quantification index, hydrological signal baseline drift ratio and corresponding moving average window adjustment coefficient, flood peak hydrological signal transmission integrity verification parameters, total amount of hydrological data and corresponding data segmentation adjustment factor). Each parameter combination is assigned a quantified value based on its adaptability to the adjustment effect (e.g., when the high-frequency oscillation strong magnetic disturbance quantification index significantly exceeds...). When the preset interference threshold is set, a larger center frequency adjustment coefficient is matched to enhance high-frequency noise suppression. When the baseline drift ratio only slightly exceeds the preset benchmark, a smaller window adjustment coefficient is matched to avoid excessive smoothing that would cause the loss of peak trend characteristics. At the same time, the actual effective values ​​of the matching parameters (such as center frequency adjustment coefficient, window adjustment coefficient, and segmentation adjustment factor) under each historical scenario are recorded. Abnormal correlation data caused by instantaneous anomalies of monitoring equipment (such as frequency monitor fluctuations and data counter errors) and sudden environmental interference (such as short-term strong magnetic pulses) are eliminated through correlation analysis (such as Spearman rank correlation coefficient), while retaining the parameter correspondence with statistical stability.

[0050] In this embodiment, a dynamic anti-interference strategy is employed to selectively filter out three noise types: strong impulse noise, high-frequency oscillation noise, and baseline drift noise. These methods precisely address the signal distortion caused by strong impulse noise, high-frequency oscillation noise, and baseline drift noise resulting from strong magnetic disturbances. The three methods work together to form a multi-dimensional anti-interference closed loop, which helps to accurately match the interference characteristics of different types of noise, avoids the limitations of a single filtering method, reduces the risk of various noises derived from strong magnetic disturbances damaging the core characteristics of flood peaks (such as the flood peak initiation point), and improves the purity, trend stability, and core characteristic fidelity of hydrological signals. This achieves efficient and precise suppression of interference with flood peak-related hydrological signals, providing reliable signal assurance for subsequent flood peak hydrological signal transmission integrity verification, data quality assessment, and flood peak flow inversion accuracy.

[0051] Furthermore, the specific process for verifying the integrity of flood peak hydrological signal transmission is as follows: Obtaining hydrological signal transmission integrity verification parameters, including the percentage of valid sampling points reflecting the absence of lost or missing data during transmission; the data transmission error rate reflecting the accuracy and reliability of data transmission; and the water level change rate transmission fidelity used to assess the retention of core temporal characteristics of the water level change rate after transmission, ensuring the authenticity of the key trends of the flood peak. The percentage of valid sampling points is represented by quantifying the ratio of the number of valid hydrological signal sampling points monitored by the hydrological data acquisition and monitoring instrument after transmission to the total number of sampling points to be transmitted after filtering. The transmission error rate (BER) is represented by quantizing the percentage of bit errors monitored by the BER tester relative to the total number of transmitted bits. The bit error rate represents the total number of bits that fail to pass the verification (such as CRC check) after the receiver checks the transmitted data. The water level change rate transmission fidelity is represented by quantizing the percentage of mean square error (MSE) calculated between the water level change rate sequence and the filtered original signal water level change rate sequence, and then proportionally comparing this result to the original change rate amplitude. The water level change rate sequence represents the sequence composed of the ratios of the water level difference between adjacent sampling points to the sampling time interval in the transmitted hydrological signal, arranged chronologically. The original change rate amplitude represents the largest water level change rate value in the original hydrological signal's water level change rate sequence after implementing dynamic anti-interference strategies. The determination of whether the hydrological signal transmission is complete... If the integrity check parameters meet the data transmission integrity criteria, the corresponding hydrological data is marked as qualified and evaluated. Otherwise, the transmission check counter is incremented to obtain the incremented value, and transmission link optimization is triggered based on the incremented value. The data transmission integrity criteria indicate that the proportion of valid sampling points is greater than a preset sampling point proportion threshold, the data transmission error rate is less than a preset error rate threshold, and the transmission fidelity of the water level change rate is greater than a preset water level change rate transmission threshold. The preset sampling point proportion threshold is represented by the average proportion of valid sampling points over a historical period, and the preset error rate threshold is represented by the average data transmission error rate over a historical period. The water level change rate transmission threshold is represented by the average value of the water level change rate transmission fidelity over a historical time period. The transmission link optimization trigger judgment means that it is determined whether the accumulated value of the transmission verification counter is greater than the preset maximum threshold of the transmission verification counter. If so, transmission link optimization is performed; otherwise, continuous judgment of the integrity of the flood peak hydrological signal transmission is performed. The continuous judgment of the integrity of the flood peak hydrological signal transmission means that the flood peak hydrological signal transmission integrity verification is continued. When the number of times the flood peak hydrological signal transmission integrity verification is performed is greater than the preset maximum number of hydrological signal transmission verifications, if the accumulated value of the transmission verification counter is still not greater than the preset maximum threshold of the transmission verification counter, hydrological data quality assessment is performed. The preset maximum threshold of the transmission verification counter is set in advance by preset personnel.

[0052] In this embodiment, by performing integrity verification of flood peak hydrological signal transmission, it helps to specifically identify problems such as lost signal sampling points, data errors, and distortion of core features during transmission. This provides a qualified foundation of transmitted hydrological data for subsequent hydrological data quality assessment, reduces the risk of missing and deviation of key time-series data of the flood peak (such as rise rate and peak amplitude) during transmission, improves the integrity, temporal continuity, and fidelity of core features of the hydrological signal transmission, and realizes quantitative verification and closed-loop optimization of the transmission quality of flood peak hydrological signals. This ensures that the key data required for flood peak flow inversion has low deviation at the transmission level, providing a solid transmission guarantee for subsequent high-quality data quality assessment and accurate inversion.

[0053] Furthermore, the specific process of transmission link optimization is as follows: Dynamic data fragmentation is performed, specifically as follows: The newly acquired flood peak hydrological signal transmission integrity verification parameters and the total amount of hydrological data are input into a preset strong magnetic disturbance-fragment size mapping table in the database for querying. A data fragmentation adjustment factor is obtained, and the magnitude corresponding to the data fragmentation adjustment factor is used as the adjustment step size. The data fragment size is adjusted incrementally in the direction of decreasing data fragmentation. (After each data fragmentation size adjustment, the flood peak hydrological signal transmission integrity verification parameters are reacquired, and it is determined whether the flood peak hydrological signal transmission integrity verification parameters meet the data transmission integrity judgment conditions. If not, the adjusted data fragment size is used as the initial value for the next adjustment, continuing to adjust incrementally in the direction of decreasing data fragmentation.) This helps reduce the problem of a sudden drop in transmission efficiency caused by a one-time large reduction in fragmentation, ensuring that while reducing the data transmission load, the transmission integrity of the core flood peak data is gradually improved, achieving dynamic matching between the data fragment size and the transmission link status; continuous monitoring of flood peak hydrological signals... Transmission integrity verification parameters are used. If the transmission integrity verification parameters of the flood peak hydrological signal meet the data transmission integrity judgment conditions, flood peak flow inversion is performed; otherwise, dynamic continuous data fragmentation is performed. Dynamic continuous data fragmentation means continuing to execute dynamic data fragmentation. If the transmission integrity verification parameters of the flood peak hydrological signal meet the data transmission integrity judgment conditions, hydrological data quality assessment is performed; otherwise, if the number of times dynamic continuous data fragmentation is executed exceeds the preset maximum number of times dynamic data fragmentation is executed, and if the transmission integrity verification parameters of the flood peak hydrological signal still do not meet the data transmission integrity judgment conditions, the current main transmission link of hydrological data is switched to a backup anti-interference link (such as a backup channel using differential signal transmission). The preset maximum number of times dynamic data fragmentation is executed is set in advance by preset personnel. After the transmission link optimization is completed, the transmission integrity verification parameters of the flood peak hydrological signal are retrieved again. If the transmission integrity verification parameters of the flood peak hydrological signal meet the data transmission integrity judgment conditions, flood peak flow inversion is performed; otherwise, a link optimization failure warning is sent.

[0054] In this embodiment, by optimizing the transmission link with dynamic data fragmentation adjustment as the core, combined with real-time adaptation of transmission parameters and strong magnetic disturbance status, it helps to accurately match the transmission link carrying capacity under strong magnetic disturbance with the flood peak data transmission requirements. This reduces the risk of sampling point loss, bit error rate, and transmission distortion of core flood peak features during hydrological data transmission. At the same time, it avoids the problem of transmission efficiency imbalance caused by one-time optimization, improves the stability, integrity, and dynamic balance of transmission efficiency of flood peak hydrological signal transmission, and achieves accurate adaptation of transmission link status with the impact of strong magnetic disturbance and data fragmentation scale, ensuring the efficient transmission of high-quality hydrological data to subsequent stages.

[0055] Furthermore, the specific process of hydrological data quality assessment is as follows: Hydrological data quality quantification parameters are obtained, including hydrological data integrity indicators to assess the integrity of data transmission during transmission, hydrological data accuracy indicators to measure the impact of residual noise and accuracy, and hydrological data consistency indicators to quantify the continuity and rationality of hydrological data temporal changes. The results of weighted coupling processing of the hydrological data quality quantification parameters and their corresponding hydrological data quality weights are used as hydrological data quality assessment indicators to comprehensively quantify the integrity, accuracy, and consistency of hydrological data. Here, weighted coupling processing refers to performing a product operation; hydrological data quality weights... The influencing parameters include the hydrological data integrity influence coefficient, used to assess the impact of hydrological data integrity indicators on hydrological data quality assessment indicators; the hydrological data accuracy influence coefficient, used to assess the impact of hydrological data accuracy indicators on hydrological data quality assessment indicators; and the hydrological data consistency influence coefficient, used to assess the impact of hydrological data consistency indicators on hydrological data quality assessment indicators. The system determines whether the hydrological data quality assessment indicators exceed a preset data quality threshold. If so, the corresponding hydrological data is marked as qualified hydrological data, and peak flow inversion is performed. Otherwise, a hydrological data quality non-compliance warning is issued. The preset data quality threshold is represented by the average value of hydrological data quality assessment indicators over a historical time period.

[0056] Specifically, the formula for the hydrological data integrity index is as follows:

[0057]

[0058] Among them, H com The parameters represent the integrity index of hydrological data, where N0 represents the percentage of valid sampling points, N1 represents the data transmission error rate, and E represents the transmission fidelity of the water level change rate.

[0059] Specifically, the formula for the accuracy index of hydrological data is as follows:

[0060]

[0061]

[0062] Among them, H acc H represents an indicator of the accuracy of hydrological data. acc，K This represents the accuracy index of the Kth type of hydrological data, where n represents the total number of sampling points within the hydrological data quality assessment period, i represents the sampling point number within the hydrological data quality assessment period (i=1, 2, 3...n), K represents the numerical code for the hydrological data type (e.g., water level data is coded as 1, flow velocity data as 2, flow rate data as 3, etc.), K=1,2,3...Z, and Z represents the total number of hydrological data types. K y represents the total number of sampling points for Class K hydrological data during the hydrological data quality assessment period. i Represents the amplitude sequence of hydrological data, ŷ K α represents the preset hydrological data amplitude for the Kth type of hydrological data. K N represents the noise residual correction factor for the Kth type of hydrological data. total,K N represents the total number of parameter data packets for the Kth type of hydrological data. nois,K This represents the number of noise sampling points for the Kth type of hydrological data. Specifically, the hydrological data amplitude sequence is represented by the sequence of hydrological data amplitudes transmitted within the hydrological data quality assessment period. The numerical codes for hydrological data types are assigned values ​​one by one according to preset classification rules, which are set in advance by preset personnel. The preset hydrological data amplitude for the Kth type of hydrological data is represented by the average value of the hydrological data amplitude for the Kth type of hydrological data over a historical period. The number of noise sampling points is represented by the total number of sampling points identified as strong impulse interference, high-frequency oscillation noise, and baseline drift noise within the hydrological data quality assessment period. Hydrological data includes water level height, water flow velocity, etc.

[0063] Specifically, the formula for the hydrological data consistency index is as follows:

[0064]

[0065] Among them, H con Δy represents the consistency index of hydrological data. i Δy represents the change between adjacent sampling points. i =y i+1 -y i Δy0 represents the preset change between adjacent sampling points, Δy max I(Δy) represents the preset maximum change. i >β×Δy max) represents the indicator function (1 when the condition in parentheses is true, 0 otherwise), β represents the consistency threshold coefficient. Specifically, the change in adjacent sampling points represents the change in adjacent sampling points of the hydrological data amplitude sequence. The preset change in adjacent sampling points is represented by the average value of the change in adjacent sampling points of hydrological data over a historical period. The preset maximum change is set in advance by the preset personnel.

[0066] Specifically, the formulas for the hydrological data quality assessment indicators are as follows:

[0067]

[0068] Among them, H qua The values ​​represent the hydrological data quality assessment indicators, where w1 represents the hydrological data integrity impact coefficient, w2 represents the hydrological data accuracy impact coefficient, and w3 represents the hydrological data consistency impact coefficient.

[0069] Specifically, the process of peak flow inversion is as follows: Based on the time series partitioning method, historical hydrological data is divided into a hydrological data training set and a hydrological data test set; based on the hydrological data training set, the parameters of the preset machine learning model (such as LSTM, random forest, support vector machine, etc.) are iteratively optimized, and the model accuracy is verified through the hydrological data test set to complete the training of the preset machine learning model; qualified hydrological data is input into the preset machine learning model, and the real-time peak flow inversion value is output.

[0070] It is important to note that hydrological data quality assessment involves a set of weighted parameter systems used to quantify the correlation between hydrological data integrity indicators, hydrological data accuracy indicators, hydrological data consistency indicators and hydrological data quality assessment indicators, as well as the matching relationship between noise residual correction coefficient, consistency threshold coefficient and corresponding correction effect, and threshold judgment criteria. This system is pre-set by professional technicians and stored in a database, providing the core basis for the weight matching of hydrological data quality quantification parameters and various influence coefficients, and the appropriate selection of correction coefficients and threshold coefficients.

[0071] Specifically, a large amount of historical hydrological monitoring data is first extracted, covering hydrological data integrity indicators, hydrological data accuracy indicators, and hydrological data consistency indicators under different watershed scenarios and different flood peak periods, along with their corresponding preset benchmark values ​​(historical statistical averages), hydrological data integrity impact coefficients, hydrological data accuracy impact coefficients, and hydrological data consistency impact coefficients. Simultaneously, a complete parameter combination is included, comprising noise residual correction coefficients, consistency threshold coefficients and corresponding noise correction effect values, and threshold judgment adaptation values. Each parameter combination is assigned a weighted quantitative value based on its influence on the hydrological data quality assessment indicators. For example, when the hydrological data integrity indicator is below the historical normal range, leading to an increased risk of missing core data, a higher hydrological data integrity impact coefficient is matched to strengthen its proportion in the assessment indicator calculation. Simultaneously, this is recorded... The system records the actual effective values ​​of each influence coefficient, correction coefficient, and threshold coefficient under various historical scenarios. Then, through correlation analysis (such as Spearman rank correlation coefficient), it removes abnormal correlated data caused by temporary deviations of monitoring equipment (such as hydrological amplitude sensor calibration drift and instantaneous data transmission errors) and sudden environmental interference (such as short-term strong magnetic interference and sampling equipment failure). It retains the correspondence between influence coefficients and correction parameters with statistical stability. Finally, it integrates all effective data to form a weighted parameter system. Each coefficient uses a numerical range of 0-1 to represent the influence ratio, ensuring the accuracy and quantifiability of parameter matching. When the system conducts hydrological data quality assessment, it can quickly retrieve coefficient values ​​that match the current data from this system, providing reliable parameter support for the comprehensive quantification of hydrological data quality and subsequent flood peak flow inversion.

[0072] In this embodiment, by conducting hydrological data quality assessment and flood peak flow inversion, based on hydrological data that has undergone dynamic anti-interference measurement and data link transmission optimization, high-quality hydrological data matching the core characteristics of the flood peak are quantitatively screened. Combined with machine learning models, the key parameters of the flood peak are accurately restored, which helps to connect the results of previous signal processing and transmission guarantee, provides objective and reliable quantitative support for flood peak situation analysis, reduces the risk of misjudgment of flood peak characteristics and deviation of inversion results caused by low-quality data, improves the accuracy of flood peak flow inversion, the controllability of data quality, and the extraction accuracy of core dynamic information of the flood peak, and realizes a closed loop of the whole process from hydrological data quality screening to accurate quantification of flood peak flow.

[0073] like Figure 6 As shown in the figure, this application provides a comparison chart of the predicted flow rates using a machine learning model for a peak flow inversion method based on machine learning. Figure 6It can be seen that the predicted flow of the preset machine learning models (such as SVM, random forest, LSTM) can fit the temporal variation trend of the measured peak flow well. Among them, the curves of the predicted flow and the measured flow of each model have a high degree of fit, which fully verifies the effectiveness of the peak flow inversion method based on machine learning in this application in capturing the dynamic characteristics of peak flow and realizing accurate peak flow inversion.

[0074] In summary, by analyzing the interference characteristics of hydrological signals and determining whether dynamic anti-interference strategies are needed, it is helpful to specifically eliminate strong impulse noise, high-frequency oscillation noise, and baseline drift noise, ensuring the authenticity and integrity of core features of hydrological signals. This provides reliable raw data support for subsequent verification of the integrity of flood peak hydrological signal transmission and assessment of hydrological data quality. After the hydrological signal interference characteristic analysis is qualified, verifying the integrity of flood peak hydrological signal transmission and determining whether transmission link optimization is needed helps to reduce problems such as sampling loss, residual errors, and feature distortion during hydrological data transmission, ensuring the integrity and unbiasedness of key time-series data during the core period of the flood peak. After the flood peak hydrological signal transmission integrity verification is qualified, hydrological data quality assessment and determining whether flood peak flow inversion is needed help to quantitatively screen out high-quality hydrological data with accuracy and consistency, accurately determine the feasibility of flood peak flow inversion, reduce misjudgments of flood peak flow caused by low-quality data, and ultimately improve the accuracy of flood peak flow inversion results.

[0075] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for machine learning based flood peak flow inversion, characterized in that, The method comprises: In the flood peak flow inversion process, hydrological signal interference feature analysis for evaluating the degree of hydrological signal interference is performed, and it is judged whether a dynamic anti-interference strategy needs to be taken to guarantee the authenticity and effectiveness of the hydrological signal; After the hydrological signal interference feature analysis is qualified, flood peak hydrological signal transmission integrity checking for verifying the integrity, time sequence continuity and core feature fidelity of the flood peak core period data transmission is performed, and it is judged whether transmission link optimization needs to be taken to reduce the flood peak core data feature transmission distortion rate in the hydrological data transmission process; After the flood peak hydrological signal transmission integrity checking is qualified, hydrological data quality evaluation for quantifying the accuracy of the matching of the hydrological data and the flood peak core feature is performed, and it is judged whether flood peak flow inversion needs to be performed; The core feature comprises a flood peak rising trend, a peak value relative reference position and a flood peak rising point; The specific process of the dynamic anti-interference strategy is as follows: Obtaining noise mode feature parameters, the noise mode feature parameters comprising a hydrological signal interference pulse duration proportion for quantifying the degree of the continuous influence of strong impulse interference on the hydrological signal, a hydrological signal oscillation frequency for quantifying the core frequency feature of high-frequency oscillation noise and a hydrological signal baseline drift ratio for evaluating the overall offset degree of the hydrological signal baseline; Judging whether the noise mode feature parameters meet a strong impulse noise intervention condition, if yes, performing strong impulse noise filtering processing, otherwise, performing high-frequency oscillation noise discrimination; The strong impulse noise intervention condition means that the hydrological signal interference pulse duration proportion is greater than a preset pulse duration proportion threshold value; The high-frequency oscillation noise discrimination means judging whether the noise mode feature parameters meet a high-frequency oscillation noise intervention condition, if yes, performing high-frequency oscillation noise filtering processing, otherwise, performing baseline drift noise discrimination; The baseline drift noise discrimination means judging whether the noise mode feature parameters meet a baseline drift noise intervention condition, if yes, performing baseline drift noise filtering processing, otherwise, sending a slight interference warning; The high-frequency oscillation noise intervention condition means that the hydrological signal oscillation frequency is greater than a preset high-frequency oscillation threshold value, and the high-frequency oscillation duration is greater than a preset oscillation duration threshold value; The baseline drift noise intervention condition means that the hydrological signal baseline drift ratio is within a preset baseline drift ratio range, and the baseline offset amount in a hydrological signal interference feature analysis time period is less than a preset drift rate threshold value; The specific process of the flood peak flow inversion is as follows: Dividing historical hydrological data into a hydrological data training set and a hydrological data test set based on a time sequence division method; Performing parameter iterative optimization on a preset machine learning model based on the hydrological data training set, verifying the model accuracy through the hydrological data test set to complete the preset machine learning model training, inputting qualified hydrological data into the preset machine learning model, and outputting real-time flood peak flow inversion values.

2. The machine learning based flood peak inversion method of claim 1, wherein, The specific process of the hydrological signal interference feature analysis is as follows: Obtaining a strong magnetic disturbance quantification parameter, the strong magnetic disturbance quantification parameter including a strong magnetic disturbance proportion for evaluating a degree of interference of the strong magnetic disturbance on a flood peak value feature, a signal distortion degree proportion for measuring a distortion level of a hydrological signal timing rule and amplitude accuracy caused by the strong magnetic disturbance, and a disturbance signal proportion for quantifying a continuous influence range of the strong magnetic disturbance; Performing a harmonic mean processing based on the strong magnetic disturbance quantification parameter, and a result of the harmonic mean processing is taken as a strong magnetic disturbance quantification index for evaluating a degree of interference of a hydrological signal by the strong magnetic disturbance; Performing a hydrological data interference verification based on the strong magnetic disturbance quantification index, if the verification is qualified, performing a flood peak hydrological signal transmission integrity verification, if the verification is unqualified, taking a dynamic anti-interference strategy; A specific process of the hydrological data interference verification is as follows: judging whether the strong magnetic disturbance quantification index is greater than a preset strong magnetic disturbance threshold value, if yes, determining that the verification is unqualified, otherwise, determining that the verification is qualified.

3. The machine learning based flood peak inversion method of claim 2, wherein, If the noise mode feature parameter meets the high-frequency oscillation noise intervention condition, the strong magnetic disturbance quantification index is reacquired, and the corresponding strong magnetic disturbance quantification index is marked as a high-frequency oscillation strong magnetic disturbance quantification index, and a high-frequency oscillation noise filtering processing is performed based on the high-frequency oscillation strong magnetic disturbance quantification index. If the noise mode feature parameter meets the baseline drift noise intervention condition, the strong magnetic disturbance quantification index is reacquired, and the corresponding strong magnetic disturbance quantification index is marked as a baseline drift strong magnetic disturbance quantification index, and a baseline drift noise filtering processing is performed based on the baseline drift strong magnetic disturbance quantification index.

4. The machine learning based flood peak inversion method of claim 3, wherein, The strong pulse noise filtering processing is used for targetedly suppressing continuous strong pulse interference in the hydrological signal caused by the strong magnetic disturbance, and a specific process is as follows: Performing interference pulse marking, and a specific process is as follows: comparing the hydrological signal amplitude in a preset time window with a preset hydrological signal amplitude one by one, when the corresponding hydrological signal amplitude is greater than a preset pulse amplitude, marking the corresponding hydrological signal sampling point as a pulse interference point; Performing median filtering based on the pulse interference point, until the median filtering processing of the hydrological signal in the entire preset time window is completed, and a median filtered hydrological signal is acquired; After the median filtering is completed, performing secondary correction processing on the median filtered hydrological signal based on nonlinear amplitude limiting filtering; After the secondary correction processing is completed, the strong magnetic disturbance quantification index is reacquired, and the hydrological data interference verification is performed, if the verification is qualified, the flood peak hydrological signal transmission integrity verification is performed, otherwise, the high-frequency oscillation noise discrimination is performed.

5. The machine learning based flood peak inversion method of claim 3, wherein, A specific process of the high-frequency oscillation noise filtering processing is as follows: Inputting the high-frequency oscillation strong magnetic disturbance quantification index and a band-pass output signal frequency into a preset oscillation frequency-notch parameter mapping table for querying, and acquiring a center frequency adjustment coefficient of an adaptive notch filter; Performing notch center frequency adjustment, and a specific process is as follows: taking an amplitude corresponding to the center frequency adjustment coefficient as an adjustment step, and adjusting the center frequency of the adaptive notch filter in a direction of decreasing a center frequency deviation value step by step; The center frequency deviation value is represented by taking an absolute value of a deviation quantification result of a current center frequency of the adaptive notch filter and a high-frequency oscillation center frequency. Continuously monitor the high-frequency oscillation residual degree, when the high-frequency oscillation residual degree is less than the preset residual threshold, then continue to perform high-frequency oscillation noise filtering processing on the subsequent hydrological signals based on the center frequency of the current adaptive notch filter, after the high-frequency oscillation noise filtering processing is completed, perform hydrological data interference verification, if the verification is qualified, then perform flood peak hydrological signal transmission integrity verification, otherwise, perform baseline drift noise discrimination; When the high-frequency oscillation residual degree is not less than the preset residual threshold, continuously perform high-frequency oscillation noise filtering processing, when the number of times of performing high-frequency oscillation noise filtering processing is greater than the preset maximum number of times of performing high-frequency oscillation filtering, if the hydrological data interference verification is not qualified, then perform baseline drift noise discrimination; The high-frequency oscillation residual degree is used to quantify the actual suppression effect of the high-frequency oscillation noise filtering processing on the high-frequency interference of the target frequency range.

6. The machine learning based flood peak inversion method of claim 3, wherein, The specific process of the baseline drift noise filtering processing is as follows: Input the baseline drift strong magnetic disturbance quantitative index and the hydrological signal baseline drift ratio into the preset baseline drift-filter correction parameter mapping table for query, and obtain the sliding average window adjustment coefficient of the sliding average filter; Perform sliding average window adjustment, the specific process is as follows: take the amplitude corresponding to the sliding average window adjustment coefficient as the adjustment step, and adjust the window size of the sliding average filter step by step in the direction of increasing the sliding average window; Continuously monitor the signal baseline fit degree, which is used to reflect the suppression effect of the baseline drift noise filtering processing on the baseline drift of the hydrological signal, and ensure the consistency of the benchmark for subsequent hydrological data quality evaluation and flood peak flow inversion; When the signal baseline fit degree is less than the preset baseline fit degree threshold, continue to perform baseline drift noise filtering processing on the subsequent hydrological signals based on the current sliding average window size, after the baseline drift noise filtering processing is completed, perform hydrological data interference verification, if the verification is qualified, then perform flood peak hydrological signal transmission integrity verification, otherwise, send a filtering processing failure warning; When the signal baseline fit degree is not less than the preset baseline fit degree threshold, continuously perform baseline drift noise filtering processing, when the number of times of performing baseline drift noise filtering processing is greater than the preset maximum number of times of performing baseline drift noise filtering processing, perform hydrological data interference verification, if the verification is not qualified, then send a filtering processing failure warning, otherwise, perform flood peak hydrological signal transmission integrity verification.

7. The machine learning based flood peak inversion method of claim 6, wherein, The specific process of the flood peak hydrological signal transmission integrity verification is as follows: Obtain the hydrological signal transmission integrity verification parameters, including the retention degree of effective data in the transmission process, the effective sampling point proportion reflecting the degree of no loss and no missing of sampling points, the data transmission error probability quantifying the error probability of transmission data, the data transmission bit error rate reflecting the accuracy and reliability of data transmission, and the water level change rate transmission fidelity ensuring the authenticity of the key trend of the flood peak. The method comprises the following steps: judging whether the hydrological signal transmission integrity check parameter meets the data transmission integrity judgment condition; if yes, marking the corresponding hydrological data as qualified hydrological data and performing hydrological data evaluation; otherwise, performing accumulation operation on the transmission check counter to obtain a transmission check counter accumulation value, and performing transmission link optimization trigger judgment based on the transmission check counter accumulation value; The data transmission integrity judgment condition indicates that the effective sampling point proportion is greater than a preset sampling point proportion threshold, the data transmission error code rate is less than a preset error code rate threshold, and the water level change rate transmission fidelity is greater than a preset water level change rate transmission threshold; The transmission link optimization trigger judgment indicates that whether the transmission check counter accumulation value is greater than a preset transmission check counter maximum threshold is judged; if yes, transmission link optimization is performed; otherwise, flood peak hydrological signal transmission integrity continuous judgment is performed; The flood peak hydrological signal transmission integrity continuous judgment indicates that the flood peak hydrological signal transmission integrity check is continuously performed; when the number of times of performing the flood peak hydrological signal transmission integrity check is greater than a preset maximum number of times of performing the hydrological signal transmission check, if the transmission check counter accumulation value is still not greater than the preset transmission check counter maximum threshold, hydrological data quality evaluation is performed.

8. The machine learning based flood peak inversion method of claim 7, wherein, The specific process of the transmission link optimization is as follows: Dynamic data fragmentation is performed, and the specific process is as follows: the reacquired flood peak hydrological signal transmission integrity check parameter and the total number of hydrological data are input into a preset strong magnetic disturbance-fragmentation size mapping table for query to obtain a data fragmentation adjustment factor; the amplitude corresponding to the data fragmentation adjustment factor is taken as an adjustment step, and the data fragmentation size is adjusted in the direction of being reduced step by step; The flood peak hydrological signal transmission integrity check parameter is continuously monitored; when the flood peak hydrological signal transmission integrity check parameter meets the data transmission integrity judgment condition, flood peak flow inversion is performed; otherwise, dynamic continuous data fragmentation is performed; The dynamic continuous data fragmentation indicates that the dynamic data fragmentation is continuously performed; when the flood peak hydrological signal transmission integrity check parameter meets the data transmission integrity judgment condition, hydrological data quality evaluation is performed; otherwise, when the number of times of performing the dynamic continuous data fragmentation is greater than a preset maximum number of times of performing the dynamic data fragmentation, if the flood peak hydrological signal transmission integrity check parameter still does not meet the data transmission integrity judgment condition, the current transmission main link of the hydrological data is switched to a backup anti-interference link; After the transmission link optimization is completed, the flood peak hydrological signal transmission integrity check parameter is reacquired; if the flood peak hydrological signal transmission integrity check parameter meets the data transmission integrity judgment condition, flood peak flow inversion is performed; otherwise, a link optimization failure early warning is sent.

9. The machine learning based flood peak inversion method of claim 8, wherein, The specific process of the hydrological data quality evaluation is as follows: A hydrological data quality quantization parameter is obtained, the hydrological data quality quantization parameter comprising a hydrological data integrity index for evaluating the data transmission integrity degree in the hydrological data transmission process, a hydrological data accuracy index for measuring the noise residual influence and accuracy degree of the hydrological data, and a hydrological data consistency index for quantifying the time sequence change continuity and rationality of the hydrological data; The result of the weighting and coupling processing of the hydrological data quality quantity parameter and the corresponding hydrological data quality weight influence parameter is used as a hydrological data quality evaluation index for comprehensively quantifying the integrity, accuracy and consistency of the hydrological data; The hydrological data quality weight influence parameter includes a hydrological data integrity influence coefficient for evaluating the influence degree of the hydrological data integrity index on the hydrological data quality evaluation index, a hydrological data accuracy influence coefficient for evaluating the influence degree of the hydrological data accuracy index on the hydrological data quality evaluation index, and a hydrological data consistency influence coefficient for evaluating the influence degree of the hydrological data consistency index on the hydrological data quality evaluation index; It is judged whether the hydrological data quality evaluation index is greater than a preset data quality threshold, if yes, the corresponding hydrological data is marked as qualified hydrological data, and the flood peak flow is inverted, otherwise, a hydrological data quality unqualified warning is sent.

Images