CN-122020430-A - Detection analysis method for river and lake water quality anomaly process identification and pollution contribution rate inversion

CN122020430ACN 122020430 ACN122020430 ACN 122020430ACN-122020430-A

Abstract

The invention relates to the technical field of river and lake water quality detection and analysis, solves the technical problem that the type identification and contribution rate interval quantification of exogenous input and endogenous release/substrate resuspension cannot be realized in the prior art, and particularly relates to a detection and analysis method for river and lake water quality abnormal process identification and pollution contribution rate inversion, which comprises the steps of calculating a disturbance index trigger event time window by a water level sequence, synchronously acquiring a water quality parameter sequence and normalizing a base line; the method comprises the steps of constructing a phase locking fingerprint matrix by aligning the phase of a main period of trigger time or water level fluctuation, extracting characteristics such as peak lag, slope, water level-parameter hysteresis loop and the like, establishing an exogenous input and endogenous release/resuspension fingerprint template library to carry out matching judgment and output confidence, adopting non-negative hybrid Bayesian inversion under the prior constraint of water level disturbance, and outputting an exogenous/endogenous contribution rate trusted interval, thereby realizing type identification of exogenous input and endogenous release/substrate resuspension and detection analysis of contribution rate interval quantification.

Inventors

WANG CUNSHI
ZHANG XIAOKE
ZHANG TINGYU
WANG HUILI
ZHU JIANZHONG

Assignees

安庆师范大学

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (10)

1. A detection analysis method for river and lake water quality anomaly process identification and pollution contribution rate inversion is characterized by comprising the following steps: S1, acquiring a water level time sequence of a river and lake water body to be monitored And calculate based on the trigger time Constructing a water level disturbance index for an event time window ; S2, synchronously acquiring a multi-parameter water quality in-situ monitoring sequence in the event time window, and carrying out baseline correction and normalization processing on the multi-parameter water quality in-situ monitoring sequence to obtain a normalization response sequence; S3, using the trigger time Or the main period of the water level fluctuation The phase is an alignment reference, and the normalized response sequence and the water level time sequence are processed Performing phase lock alignment to generate a phase lock fingerprint matrix Or (b) ; S4, locking the fingerprint matrix from the phase Or (b) Extracting double-domain features for realizing stable characterization of an abnormal process mechanism, wherein the double-domain features comprise time sequence domain features and hysteresis domain features; S5, respectively constructing a template library comprising exogenous input fingerprints based on historical event data Fingerprint template library re-suspended with endogenous release/substrate Matching the two-domain features with the fingerprint template library, and outputting the type of the abnormal process and the confidence level of the characterization posterior probability; S6, the phase locking fingerprint matrix Or (b) Represented as a non-negative mix of exogenous input template response and endogenous release/substrate resuspension template response, and at a water level perturbation index Weighting exogenesis under a priori constraints With endogenous weights Performing posterior deduction and outputting an exogenous contribution rate interval And endogenous contribution rate interval 。
2. The detection and analysis method according to claim 1, wherein in step S1, the specific process includes: S11, collecting a water level time sequence of the river and lake water body to be monitored at a monitoring section or a monitoring point ; S12, according to the water level time sequence Calculating the rate of change of water level Acceleration with water level The calculation formula is as follows: ; Wherein, the Representing the water level; to represent time; Representing a differential operation; S13, based on the water level change rate And water level acceleration Constructing a water level disturbance index The expression is: ; Wherein, the Is a non-negative weight coefficient; For time series by flow rate Calculated flow rate variation when the flow rate time series is not acquired Time can be made ; S14, based on water level disturbance index Setting up to determine the trigger moment The trigger conditions of (2) include at least one of: when the water level disturbance index And the duration is not less than the duration determination time threshold Time trigger, and , In order to provide a sampling interval, A preset threshold value; When the absolute value of the water level change rate Occurrence of the threshold value Transition to above threshold Triggered on rising or falling edge of (c), and ; After any triggering condition is met, taking the water level disturbance index in the time range of the triggering event The maximum time of (2) is used as the trigger time ; S15, based on trigger time Constructing an event time window by adopting an adaptive event time window strategy , For triggering time The previously preset duration of the front window is set, For triggering time The subsequent back window time period.
3. The assay of claim 1, wherein the multiparameter aqueous quality in situ monitoring sequence comprises at least turbidity Particulate phosphorus Oxidation-reduction potential And/or dissolved oxygen Temperature (temperature) Conductivity of Water quality parameters of (a); the particulate phosphorus Obtained directly from an in-line particulate phosphorus sensor, or from turbidity By pre-established local landmark model Conversion to obtain and output the uncertainty of the granular phosphorus 。
4. The detection and analysis method according to claim 3, comprising, in step S3: Setting the relative time Re-expressing each water quality parameter as normalized response sequence And form it into phase-locked fingerprint matrix The method comprises the following steps: ; Wherein, the Is the turbidity after normalization; is normalized particle phosphorus; Is normalized oxidation-reduction potential or dissolved oxygen; Is the normalized temperature; for the normalized conductivity, or, When the water level fluctuation is periodically fluctuated, determining the main period of the water level main fluctuation through autocorrelation analysis or frequency domain spectrum analysis And time within the event time window Mapping to phase ; Equiphase resampling of multiparameter water quality in-situ monitoring sequences in the phase domain to generate a phase-locked fingerprint matrix To eliminate time scale differences caused by different cycle lengths.
5. The method of claim 1, wherein the timing domain features comprise at least one or a combination of the following: Peak amplitude Slope of rising Slope of descent Peak time difference or phase Hysteresis of relative water level A ratio feature for reflecting changes in phosphorus carried by the particles; The hysteresis domain features include the water level And constructing a hysteresis loop with one or more parameters in the multi-parameter water quality in-situ monitoring sequence, extracting one or more of the hysteresis loop direction and the normalized loop area, wherein the hysteresis domain features are used for describing the hysteresis form of water level fluctuation-water quality response.
6. The method according to claim 5, wherein the hysteresis loop direction is determined by the sign of the directed area of the closed loop, normalizing the loop area Calculated as follows: ; Wherein, the Is the directed area of the closed loop; Respectively normalized water levels Maximum and minimum of (2); Respectively normalized water quality parameters Maximum and minimum of (2).
7. The method of claim 1, wherein the exogenous input fingerprint template library Fingerprint template library re-suspended with endogenous release/substrate According to the index of disturbance of water level The method comprises the steps of establishing a hierarchical model sub-library into at least two layers, wherein each model sub-library comprises a mean sequence and covariance parameters of a phase locking fingerprint matrix and/or distribution parameters of double-domain features, and the model sub-library is based on satisfying posterior probability Or (b) Performing recursive updating on the corresponding template sub-library according to the high confidence event of the template sub-library; the expression of the posterior probability is: ; ; Wherein, the Fingerprint template library for representing phase locking fingerprint matrix and exogenous input Confidence of (2); fingerprint template library representing phase-locked fingerprint matrix and endogenous release/substrate resuspension Confidence of (a) a priori probability Is constant or is related to the index of disturbance of the water level Hierarchical association; Respectively phase-locked fingerprint matrix Exogenous input fingerprint template library Fingerprint template library re-suspended with endogenous release/substrate The following likelihood.
8. The method of claim 1, wherein the fingerprint template library comprises at least two types of information: The first is the statistical characterization of the phase locking fingerprint matrix, including the mean sequence, covariance structure or principal component space; and the second is probability distribution parameters of the double-domain features, including a mean vector, a covariance matrix or distribution obtained by nuclear density estimation.
9. The detection and analysis method according to claim 1, characterized in that in step S6, it comprises: Set the response of the exogenous input template as Endogenous release/substrate resuspension template response was Then observe the fingerprint The method meets the following conditions: ; Wherein the exogenous weight Endogenous weight And (2) and ; To observe noise terms.
10. The detection and analysis method according to claim 1, further comprising, in step S6: For endogenous weight Setting and water level disturbance index Related a priori distributions, namely: ; Wherein, the Index of disturbance along with water level Monotonically increasing and/or Index of disturbance along with water level Monotonically decreasing to express a priori preference that strong perturbations are more prone to trigger resuspension/endogenous contribution enhancement; endogenous weights are calculated by Markov chain Monte Carlo or particle filtering And at a preset confidence level Lower output endogenous contribution rate interval At the same time by Obtaining an exogenous contribution rate interval 。

Description

Detection analysis method for river and lake water quality anomaly process identification and pollution contribution rate inversion Technical Field The invention relates to the technical field of river and lake water quality detection and analysis, in particular to a detection and analysis method for river and lake water quality anomaly process identification and pollution contribution rate inversion. In particular to a method for fingerprint identification of water quality abnormality process based on a multi-parameter in-situ monitoring sequence by taking water level fluctuation as a trigger and alignment reference, and the detection analysis method for inverting the exogenous input and endogenous release/substrate resuspension pollution contribution rate interval is applied to detection analysis of river and lake water quality monitoring and pollution process identification. Background Under the influence of multiple factors such as river basin development, gate dam regulation and control, extreme weather and the like, water quality abnormality processes often occur in river and lake water bodies, for example, total Phosphorus (TP) is suddenly increased in a short time and algae burst or water bloom risks are induced. For shallow lakes, river-lake communication areas and gate-controlled river networks, water quality anomalies may come from not only exogenous inputs (surface source runoff, tributary water, instantaneous discharge from a drain, etc.), but also endogenous release/substrate resuspension (substrate sludge disturbance results in particulate matter carrying phosphorus into the water body, enhanced sediment interfacial phosphorus release, etc.). However, the two types of processes often show similar characterization of 'TP rise/turbidity rise' in conventional monitoring, so that abnormal causes are difficult to rapidly judge, and treatment strategy deviation is easy to cause, namely, if an endogenous process is misjudged as exogenous input, pollution interception control sources can be excessively emphasized, and if the exogenous input is misjudged as endogenous release, sediment treatment and hydrodynamic regulation decision can be misled. From the mechanism, the endogenous process of the shallow water body has obvious interface oxidation-reduction sensitivity and disturbance triggering. Studies have shown that sediment-water interfaces, when dissolved oxygen is significantly reduced, amplify sediment phosphorus release (internal loading) and thereby increase the phosphorus concentration in the body of water over a short period of time and further exacerbate the risk of eutrophication. At the same time, the resuspended particulate matter of shallow lakes can contribute significant phosphorus loading to the water column, and fine particles are more likely to remain suspended in the water body and extend the abnormal duration. In engineering application, optical indexes such as turbidity (Tb) and the like are often used as agents of TP/granular phosphorus (PP) due to high-frequency and online advantages, but the agents are influenced by the grain size composition, flow state and water chemistry conditions, and calibration and uncertainty management are needed. Existing river and lake water quality anomaly identification and attribution techniques can be generally divided into the following categories: (1) Statistical discrimination and empirical thresholding based on cross-section/manual sampling. The method relies on low-frequency sampling and post-hoc analysis, and is difficult to capture abnormal processes of an hour level or a tide/gate control cycle level, and particularly peak values are easy to miss and load estimation deviation is caused in high-flow or strong disturbance events. (2) An anomaly detection and early warning method based on online multi-parameter monitoring. Such methods are by turbidityConductivity ofThe abnormality points/abnormality sections are identified by multiple indexes such as dissolved oxygen DO, oxidation-reduction potential ORP and the like, so that the real-time performance can be improved, but most methods focus on 'whether abnormality' exists, a verifiable process distinguishing structure is lacking for 'abnormality belongs to exogenous input or endogenous release/resuspension', response lag caused by hydrodynamic disturbance (especially rapid fluctuation of water level) and hysteresis form change are not considered enough, and false alarm, omission or mechanism misjudgment are caused. (3) A source analysis method based on hydrodynamic-water quality model or conservation of mass. The method can realize load decomposition in theory, but often depends on complete boundary conditions, parameter calibration and exogenous input lists, and is difficult to work stably in an online scene when facing complex scenes such as gating water level, communication scheduling, short-duration high-strength disturbance and the like, and is generally difficult to give contributio