Non-stationary reservoir hydrological drought identification method and device, medium and product

Abstract

The invention discloses a non-stationary reservoir hydrological drought identification method and device, a medium and a product, and relates to the technical field of hydrology and drought identification. The method comprises the following steps: carrying out detrending processing on historical original storage capacity data of a target reservoir within a period of historical time; calculating an average value of detrending day-by-day storage amounts corresponding to the Lth day and S days before the Lth day in the day-by-day storage amount sequence; dividing the average detrending day-by-day storage capacity of the same date into a group, and calculating a plurality of drought quantile preliminary thresholds of each group; calculating an average value of each drought quantile initial threshold value on the Dth day of the day ordinal number in one year and M days before and after the Dth day as a drought threshold value of the Dth day, and obtaining a drought threshold value group of each day; and determining a drought event of the target reservoir in the target year according to a comparison value of the de-trending day-by-day storage capacity of each day of the target year and each drought threshold value of the corresponding same date. According to the invention, the precision of identifying the hydrological drought of the non-stationary reservoir is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrology and drought identification, and in particular to a method, device, medium and product for identifying hydrological drought in a non-stationary reservoir. Background Art

[0002] Hydrological droughts significantly impact water resource management, agricultural production, and the ecological environment. Hydrological droughts can be categorized into reservoir droughts, riverine droughts, and groundwater droughts, depending on the water storage status. Reservoirs are a crucial source of regional water supply, are deeply influenced by regional water resource allocation, and exhibit significant nonstationary characteristics.

[0003] Traditional drought identification methods are mostly based on fixed thresholds or static statistical analysis, which cannot effectively deal with non-stationary and highly seasonal hydrological data, resulting in insufficient identification accuracy. In addition, the quantitative evaluation of drought characteristics often lacks precision and dynamism. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, device, medium and product for identifying hydrological drought in non-stationary reservoirs, which can improve the accuracy of identifying hydrological drought in non-stationary reservoirs.

[0005] To achieve the above object, the present invention provides the following solutions: The first aspect of the present invention provides a method for identifying hydrological drought in a non-stationary reservoir, the method comprising: performing detrending processing on the historical original storage data of the target reservoir within a historical period to obtain a detrended daily storage sequence; calculating the average of the detrended daily storage corresponding to the Lth day and the S days before the Lth day in the detrended daily storage sequence as the average detrended daily storage of the Lth day to obtain a full daily storage sequence; wherein the full daily storage sequence includes the average detrended daily storage of each day in the detrended daily storage sequence; dividing the average detrended daily storage of the same day in the full daily storage sequence into a group, and calculating A plurality of preliminary drought quantile threshold values ​​in each group are used to obtain a preliminary drought quantile threshold value group for each day in the above-mentioned full sequence of daily storage; the above-mentioned preliminary drought quantile threshold value group includes a plurality of preliminary drought quantile threshold values; the average value of each preliminary drought quantile threshold value in the above-mentioned preliminary drought quantile threshold value group on the D-th day of the year and the M days before and after the D-th day is calculated as the drought threshold value for the D-th day, to obtain a drought threshold value group for each day; the above-mentioned drought threshold value group includes a plurality of drought threshold values; the drought event of the above-mentioned target reservoir in the target year is determined based on the comparison value of the detrended daily storage of each day in the target year and each drought threshold value on the corresponding same day.

[0006] According to an embodiment of the present invention, the above-mentioned detrending processing of the historical original storage data of the target reservoir within a historical period of time to obtain the detrended daily storage sequence includes: applying a Gaussian filtering method to detrend the historical original storage data of the target reservoir within a historical period of time to obtain the detrended daily storage sequence.

[0007] According to an embodiment of the present invention, the above-mentioned calculation of the average of the detrended daily accumulations corresponding to the Lth day and the S days before the Lth day in the detrended daily accumulation sequence as the average detrended daily accumulation of the Lth day to obtain the full sequence of daily accumulations includes: when the number of days between the Lth day and the starting date in the detrended daily accumulation sequence is greater than or equal to S, calculating the average of the detrended daily accumulations corresponding to the Lth day and the S days before the Lth day in the detrended daily accumulation sequence to obtain the daily accumulation of the Lth day; when the number of days between the Lth day and the starting date in the detrended daily accumulation sequence is less than S, calculating the average of the detrended daily accumulations corresponding to the Lth day and the number of days from the Lth day to the starting date in the detrended daily accumulation sequence to obtain the daily accumulation of the Lth day; and obtaining the full sequence of daily accumulations based on the daily accumulation of the Lth day.

[0008] According to an embodiment of the present invention, the above-mentioned dividing the average detrended daily storage on the same date in the above-mentioned daily storage full sequence into a group, and calculating multiple drought quantile preliminary thresholds for each group to obtain the drought quantile preliminary threshold group for each day in the above-mentioned daily storage full sequence, includes: dividing the average detrended daily storage on the same date of each year in the above-mentioned daily storage full sequence into a group; calculating the first percentile, the second percentile and the third percentile for the multiple average detrended daily storage in each group, and using the above-mentioned first percentile as the severe drought preliminary threshold, the second percentile as the moderate drought preliminary threshold, and the third percentile as the mild drought preliminary threshold; the drought quantile preliminary threshold group for each day in the above-mentioned daily storage full sequence includes the above-mentioned severe drought preliminary threshold, the above-mentioned moderate drought preliminary threshold and the above-mentioned mild drought preliminary threshold.

[0009] According to an embodiment of the present invention, the above-mentioned calculation of the average value of each drought quantile preliminary threshold value in the above-mentioned drought quantile preliminary threshold value group on the Dth day of the year and the M days before and after the Dth day as the drought threshold value on the Dth day, to obtain the drought threshold value group for each day, includes: calculating the average value of the above-mentioned light drought preliminary threshold value on the Dth day and the M days before and after the Dth day to obtain the light drought threshold value; calculating the average value of the above-mentioned moderate drought preliminary threshold value on the Dth day and the M days before and after the Dth day to obtain the moderate drought threshold value; calculating the average value of the above-mentioned severe drought preliminary threshold value on the Dth day and the M days before and after the Dth day to obtain the severe drought threshold value, and using the above-mentioned light drought threshold value, the above-mentioned moderate drought threshold value and the above-mentioned severe drought threshold value as the drought threshold value group for date T.

[0010] According to an embodiment of the present invention, the above-mentioned determination of the drought event of the above-mentioned target reservoir in the target year based on the comparison value of the detrended daily storage of each day in the target year and each drought threshold value on the corresponding same date includes: obtaining the drought identification result of the above-mentioned target reservoir on each day in the target year based on the comparison value of the detrended daily storage of each day in the target year and each drought threshold value on the corresponding same date; the above-mentioned drought identification results include non-drought and drought; the above-mentioned drought levels include light drought, moderate drought and severe drought; when the drought identification results of each day in a continuous time period of the target year are all drought, then a drought event occurs in the above-mentioned continuous time period; when the time interval between two adjacent drought events does not exceed a preset interval number of days, the two adjacent drought events are merged into one drought event.

[0011] According to an embodiment of the present invention, the method further comprises: extracting characteristic values ​​of drought events, and generating corresponding statistical tables based on the characteristic values ​​and the drought events.

[0012] A second aspect of the present invention provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any one of the above-mentioned methods for identifying non-stationary reservoir hydrological drought.

[0013] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for identifying non-stationary reservoir hydrological drought.

[0014] A fourth aspect of the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned methods for identifying non-stationary reservoir hydrological drought.

[0015] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: The present invention provides a method for identifying hydrological drought in non-stationary reservoirs. By using detrending and averaging processing, and on the basis of eliminating the influence of long-term trends in hydrological sequences, the daily quantile threshold is smoothed. This not only ensures the sensitivity of drought identification on a daily scale, but also effectively avoids the identification instability problem caused by excessive fluctuations in daily quantile calculations, thereby significantly improving the accuracy and stability of drought identification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 This is a flow chart of the non-stationary reservoir hydrological drought identification method provided by the present invention.

[0018] Figure 2 The present invention provides a curve chart of the original reservoir storage data, Gaussian filtered trend items and detrended storage.

[0019] Figure 3 The present invention provides a threshold curve chart for daily drought judgment after smoothing.

[0020] Figure 4 This is a comparison chart of the occurrence time of reservoir hydrological drought events identified using the non-stationary reservoir hydrological drought identification method provided by the present invention and the reservoir hydrological drought events identified directly using the original storage sequence.

[0021] Figure 5 This is a diagram of the internal structure of a computer device. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] Currently, hydrological drought identification methods are mainly divided into two categories: stationary methods and non-stationary methods. Stationary methods assume that the statistical characteristics of hydrological series (such as mean and variance) remain unchanged over time, while non-stationary methods take into account the temporal changes of hydrological series and can better reflect the impact of climate change and human activities on hydrological systems. Common non-stationary methods include: 1. Time-varying threshold method: Dynamically adjust the threshold based on the long-term trend or seasonal changes of the hydrological series to avoid the limitations of fixed thresholds. Common methods include sliding window quantiles, adaptive quantiles, and seasonal factor correction.

[0024] 2. Method based on time series decomposition: decompose the hydrological series and analyze them separately.

[0025] 3. Machine learning-based methods: Use machine learning models (such as support vector machines, random forests, etc.) to capture the nonlinear characteristics of hydrological sequences.

[0026] Although non-stationary methods have shown good adaptability in hydrological drought identification, they still have some limitations: 1. Insufficient decomposition granularity: Existing methods mostly perform time series decomposition based on monthly data. However, hydrological drought events may be triggered by short-term fluctuations. Coarse-grained decomposition can obscure key details, leading to errors in the description of the temporal characteristics of drought. 2. Rigid threshold setting: Existing non-stationary methods often rely on fixed quantile values ​​of the original series to set thresholds, without considering the coupling effect of trend terms and residual terms. The thresholds do not match the actual drought conditions well enough. 3. Risk of overfitting during trend fitting: For non-stationary series, detrending methods are often used to obtain stationary series before identification. Traditional trend extraction methods may result in large fluctuations in the fitted curve due to insufficient control of model complexity. Abnormal fluctuations in the trend line can distort the determination of the intensity and duration of drought events, leading to systematic biases in drought attribution analysis.

[0027] The purpose of the present invention is to provide a method, device, medium and product for identifying non-stationary reservoir hydrological drought. By detrending the daily reservoir storage data, a daily varying drought threshold is constructed to identify and evaluate reservoir hydrological drought events. It combines technical means such as time series decomposition, sliding average, and quantile threshold analysis. It can accurately and effectively identify non-stationary reservoir hydrological drought and calculate its characteristic value, which is of great significance for accurately identifying reservoir hydrological drought.

[0028] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] Example 1

[0030] like Figure 1 As shown, the non-stationary reservoir hydrological drought identification method in this embodiment includes: Step S1: Detrending the original historical storage data of the target reservoir within a historical period to obtain a detrended daily storage sequence.

[0031] Specifically, the Gaussian filtering method is applied to detrend the historical raw storage data of the target reservoir within a historical period to obtain the detrended daily storage series.

[0032] In practice, the raw storage data is first detrended using a Gaussian filter. The filter's standard deviation parameter (σ) is set to 731, and the boundary processing parameter (mode) is set to "reflect" to obtain a long-term trend term. The smoothed trend term is then subtracted from the raw daily storage data to generate a detrended daily storage series. In this step, the raw storage data is measured historical data and can cover any time period, but continuous data is recommended to ensure the accuracy of the smoothed trend.

[0033] Gaussian filtering is a filtering method commonly used for signal smoothing and trend extraction. In hydrological time series analysis, it can be used to extract long-term trends in storage series, thereby achieving detrending and adapting to the needs of non-stationary drought identification. With Gaussian filtering, each raw daily storage data point has a corresponding smoothed trend term. Gaussian filtering applies a filtering operation to each data point, generating a smoothed long-term trend. After deducting this trend, the detrended data is obtained. Each smoothed trend term is based on the entire time series, so its shape is the same (i.e., it is a Gaussian filtered trend), but its value varies at different time points. The trend term generated by Gaussian filtering is dynamic and not constant. Therefore, the value of the smoothed trend term varies at each time point.

[0034] The Gaussian kernel function expression is as follows: (1) in, Represents the weight at the time t from the center; is the standard deviation of the Gaussian kernel, which controls the width of the smoothing window; t is the time offset relative to the center point. Take 731.

[0035] Given original storage capacity , the smoothed trend part Calculated by the following formula: (2) Among them, 2n+1 represents the width of the sliding window. In actual application, Select an appropriate window size.

[0036] After the smoothing process is completed, the trend item can be removed from the original sequence to obtain the detrended accumulation sequence. : (3) Among them, t is the index value of the daily data, indicating the tth day in the data, surface shows the detrended data for day t, Represents the original number of day t according to, Represents the trend part of the t-th day after smoothing by Gaussian filtering.

[0037] Calculated The series does not contain a long-term trend term and can be used in the subsequent drought identification process.

[0038] Step S2: Calculate the average of the detrended daily accumulations corresponding to the Lth day and the S days before the Lth day in the detrended daily accumulation sequence as the average detrended daily accumulation of the Lth day, and obtain the full sequence of daily accumulations; wherein the full sequence of daily accumulations includes the average detrended daily accumulation of each day in the detrended daily accumulation sequence.

[0039] S2 includes: Step S21: When the number of days between the Lth day and the starting date in the detrended daily accumulation sequence is greater than or equal to S, the average of the detrended daily accumulations corresponding to the Lth day and the S days before the Lth day in the detrended daily accumulation sequence is calculated to obtain the daily accumulation of the Lth day.

[0040] Step S22: When the number of days between the Lth day and the starting date in the detrended daily accumulation sequence is less than S, calculate the average of the detrended daily accumulations corresponding to the Lth day in the detrended daily accumulation sequence and the number of days between the Lth day and the starting date in the detrended daily accumulation sequence to obtain the daily accumulation of the Lth day.

[0041] Step S23: Obtain a full sequence of daily storage amounts based on the daily storage amounts on the Lth day.

[0042] In practical applications, the value of t is the Lth day in the daily sequence. Based on the detrended accumulation data, taking each day t as the benchmark, the accumulation of the previous 30 days (including the day) is taken to calculate the average, and the rolling average is used to obtain a new sequence. . is the daily accumulated data after detrending, and t is the index value of the daily data, indicating the tth day in the data. The specific calculation steps are as follows: when When (i.e., on the 31st day and thereafter), for each date t, the rolling average is obtained by calculating the mean of the data on that day and the data of the previous 30 days: (4) Where t represents the current day, and t-30 represents the 30 days preceding the current date (including the current date). This rolling average window includes the daily storage data for the current date and the previous 30 days.

[0043] When t < 31 (i.e., day 1 to day 30), since there are less than 30 days of data, the rolling average calculation will use all data from day 1 to the current date t: (5) When t < 31, the rolling average is the mean of all data from day 1 to the current day, t. This sliding average of the preceding 31 days (including the current day) takes into account the possibility that the initial data may be less than 31 days long, avoids missing values ​​at the beginning of the series, and ensures that the entire series contains results. Therefore, in this invention, t refers to L, and the value of S can be selected as 30.

[0044] Step S3: Divide the average detrended daily storage on the same date in the full daily storage sequence into a group, and calculate multiple drought quantile preliminary thresholds for each group to obtain a drought quantile preliminary threshold group for each day in the full daily storage sequence; the drought quantile preliminary threshold group includes multiple drought quantile preliminary thresholds.

[0045] S3 specifically includes: Step S31: Divide the average detrended daily accumulation on the same day of each year in the full sequence of daily accumulation into one group.

[0046] Step S32: For each group of multiple average detrended daily storage, calculate the first percentile, second percentile and third percentile, and use the first percentile as the preliminary threshold for severe drought, the second percentile as the preliminary threshold for moderate drought, and the third percentile as the preliminary threshold for mild drought; the drought percentile preliminary threshold group for each day in the entire daily storage sequence includes the preliminary threshold for severe drought, the preliminary threshold for moderate drought and the preliminary threshold for mild drought.

[0047] Specifically, the smoothed sequence Group by "calendar day", that is, collect the data of the same date in all years, and then calculate the 6th, 16th and 30th percentiles of the daily storage for each date.

[0048] Since leap year data are less and errors are more likely to occur, in the present invention, leap year data are not considered when grouping by "calendar day" and calculating the 6th, 16th, and 30th percentiles of each group of data after grouping. Indicates the order of calendar days (Date). Each set of data is: (6) in, Denotes the date corresponding to day t. For each calendar day D, calculate the preliminary quantile threshold of its residual sequence, which is defined as follows: Preliminary threshold for severe drought (6th percentile, corresponding to a standardized index threshold of -1.5): (7) Preliminary threshold for moderate drought (16th percentile, corresponding to the normalized index threshold -1): (8) Preliminary threshold for mild drought (30th percentile, corresponding to a standardized index threshold of -0.5): (9) Finally, a preliminary set of daily drought quantile thresholds covering 365 days a year is formed.

[0049] The 6th percentile represents the preliminary threshold for severe drought (approximately corresponding to the standardized index threshold of -1.5), the 16th percentile represents the preliminary threshold for moderate drought (approximately corresponding to the standardized index threshold of -1), and the 30th percentile serves as the preliminary threshold for mild drought (approximately corresponding to the standardized index threshold of -0.5).

[0050] Step S4: Calculate the average value of each drought quantile preliminary threshold value in the drought quantile preliminary threshold value group on the Dth day of the year and the M days before and after the Dth day as the drought threshold value on the Dth day, and obtain the drought threshold value group for each day; the drought threshold value group includes multiple drought threshold values.

[0051] S4 includes: Step S41: Calculate the average value of the preliminary threshold value of mild drought on the D day and the M days before and after the D day to obtain the threshold value of mild drought.

[0052] Step S42: Calculate the average value of the preliminary threshold value of moderate drought on the D day and the M days before and after the D day to obtain the moderate drought threshold value.

[0053] Step S43: Calculate the average value of the preliminary severe drought threshold on the D day and the M days before and after the D day to obtain the severe drought threshold, and use the mild drought threshold, moderate drought threshold and severe drought threshold as the drought threshold group for date T.

[0054] In practical applications, the 31-day sliding average algorithm is applied to the preliminary thresholds of daily drought quantiles of different levels. The storage volume of the 31 days before and after the day is averaged to form a smooth daily drought threshold from January 1 to December 31. : (10) in, Represents a specific percentile Preliminary drought thresholds for corresponding levels , D represents the day sequence within a year. When using the 31-day moving average algorithm, M can be set to 15, with the Dth day, the preceding and following M days forming 31 days.

[0055] When the index is out of range, the annual cycle filling method is used: For dates before D=1 (e.g. D=-1), fill in the number with the last digit (e.g. D=364). For dates after D=365, fill in the number with the first digit (e.g. D=1,2,...).

[0056] For the smoothing threshold on February 29, the interpolation between the two days before and after is used, and the calculation formula is as follows: (11) in, Corresponding level on February 29th The threshold value, and These are the thresholds corresponding to February 28 and March 1, respectively. This data is suitable for interpolation and filling on February 29 in leap years to ensure that the threshold sequence for the entire year of 366 days is continuous and complete.

[0057] The final result The quantile thresholds for severe drought, moderate drought and mild drought for each day of the year form a complete basis for judgment. The data comes from ordinary years. , when used to calculate The data comes from leap years. ; D represents the day of the year.

[0058] Step S5: Determine the drought event of the target reservoir in the target year based on the comparison value of the detrended daily storage capacity of each day in the target year and each drought threshold value on the corresponding same day.

[0059] S5 includes: Step S51: Based on the comparison value of the detrended daily storage volume on each day of the target year and each drought threshold value on the corresponding same day, the drought identification result of the target reservoir on each day of the target year is obtained; the drought identification results include non-drought and drought; the levels of drought include light drought, moderate drought and severe drought.

[0060] Step S52: When the drought identification result for each day in the continuous time period of the target year is drought, a drought event occurs in the continuous time period.

[0061] Step S53: When the time interval between two adjacent drought events does not exceed a preset number of days, the two adjacent drought events are merged into one drought event.

[0062] In practical applications, based on the daily accumulation series after detrending , and the smoothed daily quantile threshold Compare, among them, is the day sequence of the tth day in the year, and determines whether a day is a drought day. If the storage on a certain day is less than the threshold of the day, it is marked as a drought day: like , then the day is a dry day (12) in, Corresponding to severe drought, moderate drought, and mild drought, t is the index value of the daily data, indicating the tth day in the data. is the calendar day sequence corresponding to the tth day, indicating the actual position of the day in the year. In an ordinary year, , in a leap year, .

[0063] Drought event identification will be based on the severe drought, moderate drought and mild drought thresholds corresponding to the 6th, 16th and 30th percentiles, respectively, and drought events of different levels will be extracted to comprehensively reflect the spatiotemporal variation characteristics of drought levels.

[0064] If there are 20 or more consecutive dry days, this period is considered a drought event. If the interval between two drought periods is less than 5 days (inclusive), the two drought events are combined into one drought.

[0065] In addition, the non-stationary reservoir hydrological drought identification method also includes: The characteristic values ​​of drought events are extracted, and corresponding statistical tables are generated based on the characteristic values ​​and drought events.

[0066] In practical applications, for each drought event (indexed by ), calculate the following eigenvalues: Drought start time : The time of occurrence of the drought event, that is, the first day in the drought sequence that meets the drought conditions: (13) in, It is a drought event The collection of indices of all dry days in .

[0067] When the drought ends : The last time point when the drought event storage recovers above the threshold and no longer meets the drought conditions.

[0068] (14) Drought Center Time : The midpoint of the drought event duration, calculated by adding half the drought duration to the start time.

[0069] (15) Duration of drought : The number of days the drought lasts, the length of time from the beginning to the end of the drought.

[0070] (16) Drought severity : The maximum deficit of storage relative to the threshold during a drought event. The maximum deficit refers to the maximum difference between daily storage and the corresponding drought threshold (e.g., moderate drought threshold) during a drought event.

[0071] (17) Drought intensity : The cumulative deficit in water storage relative to the threshold during a drought event. Drought severity is the deficit on the most severe day of the drought, emphasizing the "maximum deficit" and measuring the severity of the drought at its peak. Drought intensity is the sum of the daily deficits during the drought, also understood as the "total water deficit," emphasizing "overall loss." Compared to severity, intensity is more influenced by duration.

[0072] (18) Drought frequency : The relative duration of a drought event in the year to which it belongs, calculated by dividing the drought duration by the number of days in the year, and the corresponding year is the year of the center of the drought event. is the year of the drought center, and the number of days in that year is (usually 365 or 366): (19) The annual attribution method used for drought frequency is to use the year of the central time of the drought event as its statistical attribution year. For example, suppose there is a drought event: Duration: December 20, 2022, to January 10, 2023, a total of 22 days. The drought's central day falls on December 30, 2022. Therefore, this drought event falls into the year 2022. The frequency of this event is calculated as 22 / 365.

[0073] Finally, a detailed information table of drought events is output, including: drought event annual serial number (assigned to the corresponding year according to the start time), start time, end time, drought duration, drought severity, drought intensity and frequency, providing a scientific basis for subsequent water resources management and drought early warning.

[0074] The present invention provides a non-stationary reservoir hydrological drought identification method based on a detrended variable drought threshold, which comprises the following steps: (1) Detrending the original data: Calculate the annual mean of the daily storage data to generate a yearly sequence. Smooth the data using a Gaussian filter to extract the trend term. Then deduct the trend term from the original data to generate a detrended storage sequence. In this example, the daily storage data of a reservoir in a certain area from 2001 to 2023 are selected for research. The specific steps are as follows: Gaussian filter smoothing: the original storage data First, the Gaussian filtering method is used for detrending, and the standard deviation parameter (σ) of the filter is set to 731; the boundary processing parameter (mode) is set to "reflect mode" to obtain the long-term trend term T(t).

[0075] Generate a detrended accumulation sequence: Deduct the trend term from the original daily accumulation data to generate a detrended accumulation sequence .

[0076] (2) Calculate the daily accumulation of different quantiles of the sequence: In this embodiment, based on the generated detrended accumulation, a new sequence is formed after smoothing. , group and count the data for multiple years, and calculate the different percentile values ​​of each day's storage by date. In this step, the 6th, 16th, and 30th percentiles are selected to reflect different levels of storage. The specific steps are as follows: Smoothed detrended storage: Based on the detrended storage data, the storage of the previous 30 days (including the current day) is averaged daily, and the rolling average is used to obtain a new sequence. .

[0077] Group by year and date: Group the data by year and date, and get 365 groups of data excluding February 29 .

[0078] Calculate quantiles: For each group, calculate multiple statistical values ​​for that day, including: 6th percentile : The 6th percentile of the daily storage capacity corresponds approximately to the standardized index threshold of -1.5), which is used to identify severe drought events.

[0079] 16th percentile : The 16th percentile of the daily storage (approximately corresponding to the normalized index threshold of -1), used to identify moderate drought events.

[0080] 30th percentile : The 30th percentile of the daily storage (approximately corresponding to the standardized index threshold of -0.5), used to identify mild drought events.

[0081] Selection of quantiles: The accumulation values ​​of these quantiles can be used to identify the threshold of drought events. In this process, the values ​​on February 29 are not considered because they are less common and prone to errors.

[0082] (3) Perform a 31-day sliding average of the storage capacity at different quantiles to form a daily drought threshold: Based on the different quantile sequences calculated in the second step, further smoothing is performed using the 31-day moving average method. The specific steps are as follows: Sliding average calculation: Apply the 31-day sliding average algorithm to average the storage volume of the 31 days before and after the current day, smooth out the excessive fluctuation of the daily drought threshold, and form a smooth daily drought threshold from January 1 to December 31. , taking into account the leap year, the threshold value of February 29 is interpolated using different quantile values ​​of February 28 and March 1.

[0083] (4) Based on the daily quantile threshold of the sliding average, reservoir hydrological drought is identified and the reservoir hydrological drought characteristic value is calculated: Reservoir hydrological drought is usually identified by comparing it with the drought threshold. The smoothed storage sequence generated in the second step is used for identification based on run theory, and the daily quantile threshold obtained in the third step is used as the basis for drought determination. The specific steps are as follows: Identify drought days: If the storage on a certain day is less than the threshold value for that day, the day is considered a drought day.

[0084] The beginning and end of the drought period: from the storage capacity Below the threshold for the first time The drought event begins on the day of Recover to threshold If the drought period exceeds a certain number of days (20 days), the drought is considered to be over.

[0085] Drought merging: If the interval between two drought periods does not exceed 5 days (inclusive), the two drought events will be merged into one drought.

[0086] Calculation of drought characteristic values: For each drought event, characteristic values ​​are calculated to quantify the severity and duration of the drought. The specific calculation method is as follows: Drought start time : The occurrence time of the drought event, that is, the first day in the drought sequence that meets the drought conditions.

[0087] When the drought ends : The last time point when the drought event storage recovers above the threshold and no longer meets the drought conditions.

[0088] Drought Center Time : The midpoint of the drought event duration, calculated by adding half the drought duration to the start time.

[0089] Duration of drought : The number of days the drought lasts, the length of time from the beginning to the end of the drought.

[0090] Drought severity : The maximum deficit of storage relative to the threshold during a drought event.

[0091] Drought intensity : The cumulative deficit of storage relative to the threshold during a drought event.

[0092] Drought frequency : The relative duration of a drought event in the year to which it belongs, calculated by dividing the drought duration by the number of days in the year, and the corresponding year is the year of the center of the drought event. is the year of the drought center, and the number of days in that year is (Usually 365 or 366).

[0093] Result output: After the identification and characteristic calculation of drought events are completed, all drought events and their characteristic values ​​will be summarized and output as a table of mild drought, a table of moderate drought, and a table of severe drought.

[0094] Comparing the hydrological drought results of a certain reservoir before detrending the storage capacity, it is almost impossible to identify drought events after 2015 in the original data, while drought events can still be identified in the detrended data, as shown in Tables 1, 2, and 3. This proves that this method is applicable to non-stationary data.

[0095] Table 1 Mild drought events in a certain reservoir in a certain year

[0096]

[0097] Table 2 Moderate drought events in a certain reservoir in a certain year

[0098] Table 3 Severe drought events in a certain reservoir in a certain year

[0099] The present invention effectively solves the interference of trend interference and seasonal fluctuation on drought identification in traditional methods through Gaussian filtering detrending processing and dynamic threshold construction. Specifically, it includes: firstly, Gaussian filtering detrending processing is performed on the original daily storage data of reservoirs, and the stable storage sequence is obtained by deducting the long-term trend term. Secondly, based on the rolling 30-day (including the current day) storage average sequence A preliminary threshold system of daily quantiles was constructed (the 6th, 16th, and 30th percentiles correspond to severe / moderate / mild droughts), and the threshold curve was smoothed by sliding 31-day average after filling with annual cycle data. Finally, the thresholds generated were , using the run theory to identify drought events and calculate their duration intensity strength frequency Compared with the traditional fixed threshold method, this method innovatively couples trend stripping with dynamic thresholds, eliminates non-stationary trend interference through Gaussian filtering, and combines sliding average to maintain the continuity of the threshold within the year, achieving dual correction of historical trends and seasonal fluctuations. This significantly improves the spatiotemporal accuracy of drought identification and provides a more scientific quantitative basis for reservoir drought risk assessment.

[0100] Figure 2 The present invention provides a curve chart of the original reservoir storage data, Gaussian filtered trend items and detrended storage.

[0101] like Figure 2 As shown in the figure, the original storage data and trend item graphs show the original storage data (solid gray line) and the trend item after Gaussian filtering (dashed black line). The detrended storage graph shows the storage after deducting the trend item (solid black line). The dashed trend item line shows that the storage of a certain reservoir showed an upward trend from 2001 to 2023. Comparing the storage curves before and after detrending, it can be seen that the detrending method using Gaussian filtering is effective and feasible.

[0102] Figure 3 The present invention provides a threshold curve chart for daily drought judgment after smoothing.

[0103] like Figure 3 As shown, the square line indicates the mild drought threshold, based on a 31-day moving average; the triangle line indicates the moderate drought threshold, based on a 31-day moving average; and the cross-shaped line indicates the severe drought threshold. Dynamic thresholds generated using quantiles and moving averages can adapt to seasonal changes in storage capacity and provide a scientific basis for determining drought severity.

[0104] Figure 4This is a comparison chart of the occurrence time of reservoir hydrological drought events identified using the non-stationary reservoir hydrological drought identification method provided by the present invention and the reservoir hydrological drought events identified directly using the original storage sequence.

[0105] like Figure 4 As shown in the figure, the gray histogram is the reservoir hydrological drought events identified by the original storage data, and the black histogram is the reservoir hydrological drought events identified by the detrended storage data. The comparison results show that the original storage data can hardly identify drought events after 2015, while the drought events identified by the detrended data are more uniform in time distribution, which proves that the present invention has a good recognition effect when processing data with obvious trends and has good applicability for non-stationary data.

[0106] The present invention performs Gaussian filtering detrending on the storage data, uses Gaussian smoothing to extract long-term trends, and removes trend items from the original data. After detrending, the detrended data is combined with sliding average and quantile threshold analysis to effectively identify reservoir hydrological drought events and quantify the characteristic values ​​of drought (such as drought duration, drought severity, and drought intensity). Compared with traditional methods, this method uses Gaussian filtering to detrend, avoiding the impact of long-term trends on quantile calculations; by calculating quantiles on a daily basis, it captures seasonal changes within the year, making the threshold more accurate and avoiding systematic deviations in the drought identification process caused by trend interference. In addition, by quantifying characteristic values ​​such as drought duration, severity, and intensity, the present invention provides reliable data support for the quantitative analysis of drought events, thereby helping to optimize water resource management strategies.

[0107] Example 2

[0108] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the non-stationary reservoir hydrological drought identification method in Example 1.

[0109] Example 3

[0110] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the non-stationary reservoir hydrological drought identification method in embodiment 1.

[0111] Example 4

[0112] A computer program product includes a computer program, which implements the non-stationary reservoir hydrological drought identification method in embodiment 1 when executed by a processor.

[0113] Example 5 A computer device, which may be a database, may have an internal structure as shown in FIG. Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store pending transactions. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the non-stationary reservoir hydrological drought identification method in Example 1 is implemented.

[0114] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the present invention are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions.

[0115] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, database, or other media used in the embodiments provided herein may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, and the like.

[0116] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for identifying non-stationary reservoir hydrological drought, characterized in that: The method comprises: Detrending the original storage data of the target reservoir over a period of time to obtain the detrended daily storage series; Calculating the average of the detrended daily accumulations corresponding to the Lth day and the S days before the Lth day in the detrended daily accumulation sequence as the average detrended daily accumulation of the Lth day, and obtaining a full sequence of daily accumulations; wherein the full sequence of daily accumulations includes the average detrended daily accumulation of each day in the detrended daily accumulation sequence; The average detrended daily storage values ​​on the same date in the full daily storage sequence are divided into a group, and a plurality of drought quantile preliminary threshold values ​​are calculated for each group to obtain a drought quantile preliminary threshold value group for each day in the full daily storage sequence; the drought quantile preliminary threshold value group includes a plurality of drought quantile preliminary threshold values; Calculating the average of each drought quantile preliminary threshold value in the drought quantile preliminary threshold value group on the Dth day of the year and the M days before and after the Dth day as the drought threshold value for the Dth day, thereby obtaining a drought threshold value group for each day; the drought threshold value group includes a plurality of drought threshold values; The drought event of the target reservoir in the target year is determined based on the comparison value of the detrended daily storage on each day of the target year and each drought threshold value on the corresponding same day.

2. The non-stationary reservoir hydrological drought identification method according to claim 1, characterized in that: The detrending process is performed on the original historical storage data of the target reservoir within a historical period to obtain a detrended daily storage sequence, including: The Gaussian filtering method is applied to detrend the historical raw storage data of the target reservoir over a period of time, and the detrended daily storage series is obtained.

3. The non-stationary reservoir hydrological drought identification method according to claim 1, characterized in that: The step of calculating the average of the detrended daily accumulations corresponding to the Lth day and the S days before the Lth day in the detrended daily accumulation sequence as the average detrended daily accumulation on the Lth day to obtain the full sequence of daily accumulations includes: When the number of days between the Lth day and the starting date in the detrended daily accumulation sequence is greater than or equal to S, calculating the average of the detrended daily accumulations corresponding to the Lth day and the S days before the Lth day in the detrended daily accumulation sequence to obtain the daily accumulation of the Lth day; When the number of days between the Lth day and the starting date in the detrended daily accumulation sequence is less than S, calculating the average of the detrended daily accumulations corresponding to the Lth day in the detrended daily accumulation sequence and the number of days between the Lth day and the starting date in the detrended daily accumulation sequence to obtain the daily accumulation on the Lth day; According to the daily storage on the Lth day, a full sequence of daily storage is obtained.

4. The non-stationary reservoir hydrological drought identification method according to claim 1, characterized in that: The method divides the average detrended daily storage values ​​on the same date in the full daily storage sequence into a group, and calculates a plurality of preliminary drought quantile thresholds for each group, to obtain a preliminary drought quantile threshold group for each day in the full daily storage sequence, including: Divide the average detrended daily storage on the same day of each year in the full series of daily storage into one group; For each group of multiple average detrended daily storage, the first percentile, second percentile and third percentile are calculated, and the first percentile is used as the preliminary threshold for severe drought, the second percentile is used as the preliminary threshold for moderate drought, and the third percentile is used as the preliminary threshold for mild drought; the drought percentile preliminary threshold group for each day in the full sequence of daily storage includes the preliminary threshold for severe drought, the preliminary threshold for moderate drought and the preliminary threshold for mild drought.

5. The non-stationary reservoir hydrological drought identification method according to claim 1, characterized in that: The calculating of the average of each drought quantile preliminary threshold value in the drought quantile preliminary threshold value group on day D and M days before and after day D as the drought threshold value on day D, and obtaining the drought threshold value group for each day, includes: Calculate the preliminary threshold value of mild drought by taking the average value of the D-day and the M days before and after the D-day to obtain the threshold value of mild drought; Calculate the preliminary moderate drought threshold value by taking the average value of the D day and the M days before and after the D day to obtain the moderate drought threshold value; The average value of the preliminary severe drought threshold on the D day and the M days before and after the D day is calculated to obtain the severe drought threshold, and the mild drought threshold, the moderate drought threshold and the severe drought threshold are used as the drought threshold group for date T.

6. The non-stationary reservoir hydrological drought identification method according to claim 5, characterized in that: Determining the drought event of the target reservoir in the target year based on the comparison of the detrended daily storage capacity of each day in the target year and each drought threshold value on the corresponding same day includes: Obtaining drought identification results for the target reservoir on each day of the target year based on a comparison of the detrended daily storage volume on each day of the target year and each drought threshold value on the corresponding same day; the drought identification results include non-drought and drought; and the drought levels include mild drought, moderate drought, and severe drought; When the drought identification result for each day in a continuous time period of the target year is drought, a drought event occurs in the continuous time period; When the time interval between two adjacent drought events does not exceed the preset interval number of days, the two adjacent drought events are merged into one drought event.

7. The method for identifying non-stationary reservoir hydrological drought according to claim 6, characterized in that: The method further comprises: Extracting characteristic values ​​of drought events, and generating corresponding statistical tables based on the characteristic values ​​and the drought events.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the non-stationary reservoir hydrological drought identification method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the non-stationary reservoir hydrological drought identification method according to any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the non-stationary reservoir hydrological drought identification method according to any one of claims 1 to 7 is implemented.

Images