CN-121980173-A - Flood event automatic identification method, system and medium based on frequency domain characteristics and multiple constraint conditions

CN121980173ACN 121980173 ACN121980173 ACN 121980173ACN-121980173-A

Abstract

The application relates to a flood event automatic identification method, a system and a medium based on frequency domain features and multiple constraint conditions, wherein the method comprises the following steps of data preprocessing, namely acquiring original daily scale hydrologic time sequence data, carrying out missing value detection and interpolation to generate a flow process line overview chart, separating and smoothing a base flow, identifying candidate peaks and calculating threshold values, carrying out frequency domain analysis and period determination, determining optimal separation time length, separating and outputting flood events, merging and screening the candidate flood peaks by applying the optimal separation time length, the valley depth threshold values and the high-salient absolute threshold values, determining final flood peak positions and event start-stop times, retaining events with flood peak flow being larger than average flood in a flood period, and outputting a flood event list and feature statistical results.

Inventors

LI XIN
LIU YA
JIN ZHONGWU
LIU BAONAN
Lv Binghan
MAO BING
CHU DONGDONG
LONG RUI

Assignees

长江水利委员会长江科学院

Dates

Publication Date: 20260505
Application Date: 20260106

Claims (10)

1. The flood event automatic identification method based on the frequency domain characteristics and the multiple constraint conditions is characterized by comprising the following steps: Step 1, data preprocessing, namely acquiring original daily scale hydrologic time series data, detecting and interpolating missing values, and generating a flow process line overview chart; Step 2, separating and smoothing the base flow, namely separating the total runoff into the base flow and the direct runoff by adopting a Lyne-Hollick filtering method, and carrying out smoothing treatment on the direct runoff to generate a smoothed direct runoff sequence; Step 3, identifying candidate peaks and calculating threshold values, namely identifying all candidate flood peaks on the smoothed direct runoff sequence, calculating the saliency of each candidate flood peak and the relative valley depth of adjacent candidate flood peak pairs, and determining a saliency absolute threshold value and a valley depth threshold value based on statistical quantiles; step 4, frequency domain analysis and period determination, namely performing fast Fourier transform FFT analysis on an original flow sequence, calculating a power spectrum, identifying a dominant period, and setting a scanning range of the minimum separation duration of a flood event based on the dominant period; Step 5, determining optimal separation duration, namely traversing each separation duration candidate value in the scanning range, calculating the corresponding flood event number by combining triple constraint conditions consisting of the separation duration, the valley depth threshold and the saliency absolute threshold, and determining the optimal separation duration by adopting an initial platform average method; and 6, flood event separation and output, namely merging and screening candidate flood peaks by applying the optimal separation duration, the valley depth threshold and the highlight absolute threshold, determining final flood peak positions and event start-stop time, reserving events with flood peak flow being greater than average flow in flood season, and outputting a flood event list and characteristic statistical results.
2. The method for automatically identifying the flood event based on the frequency domain characteristics and the multiple constraint conditions according to claim 1, wherein in the step 2, the value of a filtering parameter alpha of the Lyne-Hollick filtering method is 0.925, and a forward and backward three-time recursive filtering mode is adopted for base stream separation.
3. The method for automatically identifying flood event based on frequency domain features and multiple constraint conditions according to claim 1 is characterized in that in the step 2, a Savitzky-Golay filter is adopted for smoothing the direct runoffs, the window length of the Savitzky-Golay filter is 7 days, the polynomial order is 2, when the filter cannot be called, a rolling median method with a window of 7 days is adopted for smoothing, peak position deviation detection is needed after smoothing, and the proportion of time deviation of candidate peaks before and after smoothing exceeding 1 day is ensured not to exceed 5%.
4. The automatic flood event identification method based on the frequency domain characteristics and the multi-constraint condition according to claim 1, wherein in the step 3, the absolute threshold of the salience is determined based on the statistical quantile, 60% quantiles of all candidate flood peak salience values are taken as the absolute threshold, the valley depth threshold is determined based on the statistical quantile, and the median or 25% quantiles of all adjacent candidate flood peak pairs relative valley depth values are taken as the valley depth threshold.
5. The method for automatically identifying flood events based on frequency domain features and multiple constraint conditions according to claim 1, wherein in the step 4, a dominant period is identified, annual period interference of 340 to 390 days is eliminated from a power spectrum, and a period corresponding to a maximum power value is selected as the dominant period.
6. The method for automatically identifying flood event based on frequency domain features and multiple constraint conditions according to claim 1, wherein in the step 5, a platform averaging method is initiated, specifically, a platform segment with a first length not less than 2 in a curve of "flood event number-separation duration" is positioned from a minimum separation duration candidate value, an arithmetic average value of all separation duration values in the platform segment is calculated and rounded up and rounded down to determine the optimal separation duration.
7. The method for automatically identifying a flood event based on the frequency domain characteristics and the multiple constraints of claim 1, wherein in the step6, the triple constraints comprise: (1) The time interval between the two peaks must be not less than the optimal separation duration; (2) If the interval between two peaks is smaller than the optimal separation time length, checking whether the relative valley depth is larger than or equal to the valley depth threshold value or not, and if so, dividing into two events; (3) The salience of both peaks competing with each other must be higher than the absolute threshold of salience.
8. The method for automatically identifying flood events based on frequency domain features and multiple constraint conditions according to claim 1, wherein in the step 6, the start-stop time of the event is determined by adopting a peak-to-valley method, taking the finally determined flood peak as a core, and taking the lowest point between adjacent flood peaks as a start-stop boundary of the event in a direct runoff sequence.
9. An automatic flood event recognition system based on frequency domain characteristics and multiple constraint conditions is characterized by comprising, The data preprocessing module is used for acquiring original daily scale hydrologic time series data, detecting and interpolating missing values and generating a flow process line overview chart; the base flow separation and smoothing module separates the total runoff into a base flow and a direct runoff by adopting a Lyne-Hollick filtering method, and carries out smoothing treatment on the direct runoff to generate a smoothed direct runoff sequence; The candidate peak identification and threshold calculation module is used for identifying all candidate flood peaks on the smoothed direct runoff sequence, calculating the saliency of each candidate flood peak and the relative valley depth of adjacent candidate flood peak pairs, and determining a saliency absolute threshold and a valley depth threshold based on the statistical quantile; the frequency domain analysis and period determination module is used for carrying out fast Fourier transform FFT analysis on the original flow sequence, calculating a power spectrum, identifying a dominant period and setting a scanning range of the minimum separation duration of the flood event based on the dominant period; The optimal separation duration module is used for traversing each separation duration candidate value in the scanning range, calculating the corresponding flood event number by combining triple constraint conditions consisting of the separation duration, the valley depth threshold and the highlight absolute threshold, and determining the optimal separation duration by adopting an initial platform average method; And the flood event separation and output module is used for merging and screening candidate flood peaks by applying the optimal separation duration, the valley depth threshold and the highlight absolute threshold, determining the final flood peak position and event start-stop time, reserving the event with the flood peak flow greater than the average flood period flow, and outputting a flood event list and a characteristic statistical result.
10. A computer readable storage medium, characterized in that the computer readable storage medium stores a program code which, when executed by a processor, implements the steps of the flood event automatic identification method based on frequency domain features and multiple constraints according to any one of claims 1-8.

Description

Flood event automatic identification method, system and medium based on frequency domain characteristics and multiple constraint conditions Technical Field The invention relates to the technical field of hydrologic information processing and flood analysis, in particular to a flood event automatic identification method, system and medium based on frequency domain characteristics and multiple constraint conditions, which are suitable for monitoring data analysis of river, river basin or reservoir inflow processes, flood process decomposition and characteristic statistics. Background Flood disasters are one of the most frequently occurring natural disasters, and form serious threats to national economy and life and property safety of people, and accurate identification of flood events is an important basis for hydrographic data reorganization, flood characteristic analysis, hydrographic model parameter calibration and flood control decision. The field flood is used as a complete hydrologic process from rising to peak top and then falling back, and the accurate identification of the field flood has important scientific significance and practical value for water resource management, reservoir dispatching and flood control early warning. As climate changes are exacerbated, the frequency of occurrence of extreme hydrologic events increases, placing higher demands on the accuracy, efficiency and degree of automation of flood event identification. Current flood event identification relies primarily on manual interpretation or semi-automatic methods based on fixed thresholds. The traditional manual identification method takes a base flow as a reference, combines a rainfall process to manually determine a flood rising point and a flood fading point, can ensure certain precision, has low efficiency, is difficult to cope with the batch processing requirement of long-sequence hydrologic data, and has larger influence on the result by subjective experience. In the automatic method, the threshold segmentation method is most widely applied, flood event separation is realized by setting parameters such as a flow threshold, a flood peak threshold and the like, but the method has the remarkable limitation that the threshold selection depends on manual experience, the threshold set by different operators can be greatly different, so that the identification result lacks consistency and repeatability, and meanwhile, the selected threshold is always different for different watershed hydrologic stations, but the correlation of the threshold and the hydrologic data is rarely considered in the current method. Therefore, there is a need for a flood event identification method capable of adaptively adjusting key parameters and fusing multiple constraint conditions and frequency domain features in combination with drainage basin characteristics, which not only solves the subjectivity and inefficiency problems of the traditional method, but also breaks through the adaptation bottleneck of the cross-drainage basin application, and realizes efficient, accurate and automatic identification of flood events under drainage basins with different hydrologic characteristics. Disclosure of Invention The embodiment of the application aims to provide a flood event automatic identification method, a flood event automatic identification system and a flood event automatic identification medium based on frequency domain features and multiple constraint conditions so as to improve the accuracy, robustness and interpretation of event identification. In order to achieve the above purpose, the present application provides the following technical solutions: in a first aspect, an embodiment of the present application provides a method for automatically identifying a flood event based on frequency domain features and multiple constraints, including the following steps: Step 1, data preprocessing, namely acquiring original daily scale hydrologic time series data, detecting and interpolating missing values, and generating a flow process line overview chart; Step 2, separating and smoothing the base flow, namely separating the total runoff into the base flow and the direct runoff by adopting a Lyne-Hollick filtering method, and carrying out smoothing treatment on the direct runoff to generate a smoothed direct runoff sequence; Step 3, identifying candidate peaks and calculating threshold values, namely identifying all candidate flood peaks on the smoothed direct runoff sequence, calculating the saliency of each candidate flood peak and the relative valley depth of adjacent candidate flood peak pairs, and determining a saliency absolute threshold value and a valley depth threshold value based on statistical quantiles; step 4, frequency domain analysis and period determination, namely performing fast Fourier transform FFT analysis on an original flow sequence, calculating a power spectrum, identifying a dominant period, and setting a scanning range of the minimum