Data processing method for river basin pollution reduction and carbon reduction collaborative management facing river and lake recovery

Abstract

The application discloses a kind of river basin pollution reduction and carbon reduction collaborative management data processing methods for river and lake recovery, comprising: obtaining basic data to establish river basin operation benchmark data package;Collect multi-source observation data to carry out attribution consistency review, determine control credible scene type;Identify environmental risk type and generate candidate control action set;Predict the water quality improvement benefit of candidate action, quantitatively evaluate the degree of sediment interface activation, and then calculate the endogenous pollution rebound risk and greenhouse gas rebound risk;Calculate the collaborative feasibility index;Based on the index, configure control instructions for scheduling engineering, point source or area source, generate collaborative control instruction package;Rolling simulation and evaluation verification are carried out on the instruction package, and the executable collaborative control scheme is output.The present application is helpful to reduce the secondary environmental risk caused by blind hydraulic scheduling, improve the robustness of decision-making under complex conditions, and is conducive to promoting the scientific collaboration of river basin water quality improvement and carbon reduction.

Description

Technical Field

[0001] This invention relates to the field of watershed water environment management and data processing technology, and in particular to a data processing method for coordinated management and control of watershed pollution reduction and carbon reduction for river and lake recovery. Background Technology

[0002] Improving water quality and controlling greenhouse gas emissions in drinking water source basins are key research areas in modern water environment engineering. Watersheds typically contain a complex interplay of heterogeneous hydrological and pollution inputs, including reservoirs, dams, wastewater treatment plant outlets, and agricultural non-point sources. Under complex hydrodynamic conditions, engineering interventions within the basin not only directly alter the physical transport and diffusion pathways of pollutants but also profoundly impact the physicochemical dynamics of the bottom sediment-water interface. Therefore, implementing scientific hydraulic interventions and source term regulation in watersheds, along with precise processing of multi-source heterogeneous monitoring data, is technically significant for maintaining the long-term stability of the overall water environment capacity of the basin.

[0003] Currently, conventional watershed water quality control systems mainly rely on single-target flow scheduling or end-of-pipe pollution interception commands. Existing data processing methods typically involve directly collecting online water quality monitoring data from control sections. When a specific pollutant concentration approaches or exceeds a set threshold, the system directly triggers an increased discharge command from reservoirs or dams, aiming to reduce pollutant concentrations by increasing water flow and dilution. This control logic, driven solely by short-term water quality compliance, often leads to drastic changes in river velocity and water level during actual implementation. This results in strong instantaneous hydrodynamic disturbances to bottom sediments, thereby altering the thermodynamics and dissolved oxygen distribution of the bottom water.

[0004] Existing methods struggle to balance short-term concentration reduction with long-term aquatic ecosystem stability when dealing with multi-source, heterogeneous monitoring data and water physicochemical evolution processes. This single-dimensional approach to management decisions can easily trigger difficult-to-control secondary environmental risks under complex hydrological conditions, threatening the overall environmental governance effect under multi-objective constraints. Therefore, it is necessary to investigate a watershed data processing method that can comprehensively assess the deep-seated impacts of regulatory disturbances and improve the robustness of decision-making under complex data conditions. Summary of the Invention

[0005] Purpose of the invention: To provide a data processing method for coordinated management and control of pollution reduction and carbon reduction in watersheds, in order to solve the above-mentioned problems existing in the prior art.

[0006] Technical solution: A data processing method for coordinated management and control of watershed pollution reduction and carbon reduction, comprising:

[0007] Acquire basic watershed data and establish a watershed operation benchmark data package that includes spatial correlations, pollution source accounting parameters, and scheduling engineering constraints;

[0008] Collect multi-source observation data for the current regulation cycle, combine it with the basin operation benchmark data package to perform multi-source attribution consistency verification, obtain attribution credibility and determine the current management and control credibility scenario type;

[0009] Based on multi-source observation data, the environmental risk types of control sections are identified, and candidate control action sets are generated by combining the reliable control scenario types with the scheduling engineering constraints in the basin operation benchmark data package.

[0010] Predict the water quality improvement benefits after the execution of each action in the candidate control action set, assess the degree of sediment interface activation caused by each action, and then calculate the risk of endogenous pollution rebound and greenhouse gas rebound induced by interface disturbance.

[0011] By combining the benefits of water quality improvement, the risk of rebound of endogenous pollution, the risk of rebound of greenhouse gases, and the credibility of attribution, the synergistic feasibility index of each candidate regulatory action is calculated.

[0012] Based on collaborative feasibility indicators, control instructions for scheduling projects, point sources or area sources are selected and configured, and integrated to obtain a collaborative management and control instruction package.

[0013] Perform rolling simulations and evaluations of the collaborative management instruction package, and output an executable collaborative management solution.

[0014] Beneficial effects: Compared with the prior art, the present invention identifies and intercepts monitoring data conflicts by introducing a multi-source attribution consistency verification mechanism, thereby reducing the risk of falsely triggering high-intensity scheduling actions due to sensor failure or data anomalies.

[0015] By establishing a sediment interface activation assessment model and incorporating the risk of endogenous pollution rebound and greenhouse gas rebound as negative penalty terms into decision indicators, it is helpful to predict the secondary environmental risks caused by sediment disturbance before hydraulic scheduling.

[0016] By constructing collaborative feasibility indicators and marginal collaborative efficiency evaluations, emission reduction tasks can be transferred and allocated to point source or area source entities with lower carbon emission costs when scheduling projects are constrained. This is conducive to improving the decision-making robustness of coordinated management and control of watershed pollution reduction and carbon reduction under complex operating conditions. Attached Figure Description

[0017] Figure 1 This is a flowchart of the data processing method for watershed pollution reduction and carbon reduction collaborative management of river and lake restoration, based on the present invention.

[0018] Figure 2 This invention is a flowchart for conducting multi-source attribution consistency verification, solving attribution credibility, and determining the trustworthy management and control scenario by combining the basin operation benchmark data package.

[0019] Figure 3This is a flowchart illustrating the water quality improvement benefits predicted by the present invention after the execution of each action within the candidate control action set.

[0020] Figure 4 This is a flowchart of the present invention for evaluating the degree of activation of the sediment interface caused by each action.

[0021] Figure 5 This is a flowchart of the present invention for calculating the feasibility index of the coordination of each candidate regulatory action. Detailed Implementation

[0022] Example 1: This example provides a data processing method for coordinated management and control of watershed pollution reduction and carbon reduction. The method includes the following steps:

[0023] like Figure 1 As shown, in step 101, basic watershed data is obtained, and a watershed operation baseline data package containing spatial correlation, pollution source accounting parameters, and scheduling engineering constraints is established.

[0024] Specifically, the basic data of a watershed is manifested as static geographic and engineering attribute information within the target watershed.

[0025] Read basic spatial data such as river network topology data, sub-basin boundary data, and reservoir and dam location data. Divide the basin into control units according to the river connectivity. Establish spatial relationships between upstream water inflow, lateral inflow, and scheduling projects for each control unit to form a data set that reflects the spatial network topology.

[0026] Furthermore, the system reads pollution source and carbon emission parameters such as sewage treatment plant discharge data, agricultural non-point source parameter data, and power emission factor data, and associates the pollutant emissions of various pollution sources with the corresponding carbon emission accounting parameters to form a source term data set containing pollution source accounting parameters.

[0027] In addition, it will read reservoir operation rules data, sediment phosphorus release sensitive parameters, bottom hypoxia sensitive threshold and other scheduling engineering and sediment sensitive parameters to establish specific scheduling engineering constraints.

[0028] Based on this, according to a unified time step and control unit number, a data index relationship is established between the aforementioned spatial correlations, pollution source accounting parameters, and scheduling engineering constraints, ultimately integrating them into a unified-dimensional watershed operation benchmark data package. This watershed operation benchmark data package is represented as a relational database table or network graph node at the system's underlying level, providing a unified static data foundation for subsequent scheduling rebound identification and coordinated control.

[0029] Step 102: Collect multi-source observation data for the current control cycle, combine it with the basin operation benchmark data package to perform multi-source attribution consistency verification, obtain attribution credibility and determine the current control credibility scenario type.

[0030] In this embodiment, the multi-source observation data is acquired in real time through a multi-source monitoring network deployed in the target watershed, and is represented as tensor data containing timestamps, spatial three-dimensional coordinates, and physical property quantities.

[0031] Specifically, water quality monitoring data at control sections and online monitoring data at sewage outlets are collected in real time by online water quality sensors deployed at key nodes of the river network; reservoir and dam operation status data and hydrological monitoring data are collected by hydrological stations and engineering automatic monitoring and reporting systems; and watershed land use update data are obtained by analyzing satellite remote sensing or UAV aerial images.

[0032] After completing data acquisition and timeline alignment, the system cross-validates the dynamic emission contribution characteristics corresponding to multi-source observation data with the preset load simulation contribution benchmark in the basin operation benchmark data package, calculates the degree of dispersion or deviation of various environmental evidence sources pointing to the same pollution source, and then quantifies the attribution credibility in the form of a single scalar.

[0033] By utilizing this attribution credibility, it is possible to identify whether there are abnormal conflicts in the multi-sensor data array, classify the current control status into specific control credibility scenario types, and avoid triggering high-risk scheduling interventions due to abnormal data from a single physical sensor.

[0034] Step 103: Identify the environmental risk type of the control section based on multi-source observation data. Based on the environmental risk type, and combined with the management and control credible scenario type and the scheduling engineering constraints in the basin operation benchmark data package, generate a set of candidate control actions.

[0035] Furthermore, the hydrological status and water quality physical parameters of the control section are extracted and quantitatively compared with the set water environment safety target limits to identify the types of environmental risks currently faced by the control section, such as excessive pollution or rainfall pulses. Based on the aforementioned identified credible control scenarios, it is assessed whether the current objective environmental conditions meet the safety threshold for initiating proactive intervention.

[0036] In business scenarios where proactive intervention is permitted, the system invokes scheduling engineering constraints such as adjustable flow ranges, adjustable water level ranges, and dam operation rules defined in the basin operation baseline data package. Under the premise of meeting the structural safety of water conservancy projects and the ecological base flow baseline, multiple basic control strategies are generated, including different combinations of discharge flow amplitudes, discharge durations, or low-carbon process operation modes, forming a set of candidate control actions. This set of candidate control actions primarily constitutes the basic search space for subsequent optimization and compensation decisions by the system.

[0037] Step 104: Predict the water quality improvement benefits after each action in the candidate control action set is executed, and assess the degree of sediment interface activation caused by each action, and then calculate the risk of endogenous pollution rebound and greenhouse gas rebound induced by interface disturbance.

[0038] For each generated candidate control action, a water quality concentration response relationship matrix is ​​trained and calibrated using historical monitoring data to predict the positive reduction contribution of the target action to the concentration of specific pollutants at the control section. The water quality improvement benefits are quantified through numerical normalization.

[0039] Based on this core assessment, hydraulic scheduling behavior itself is considered as a potential source of environmental dynamic disturbance. The incremental characteristics of hydrodynamic changes and estimated bottom-layer thermodynamic environmental parameters are extracted from candidate actions during their predetermined execution periods. Through comprehensive calculation of multi-dimensional environmental state parameter deviations, the degree of activation of the sediment interface, resulting in changes in the physicochemical state of the bottom-layer soil-water interface, is quantitatively measured.

[0040] By combining the activation status of the interface and the physical coverage area directly affected by specific control actions on the water surface, the absolute load of secondary phosphorus release that may be induced by strong disturbance processes can be further mapped and calculated, that is, the risk of endogenous pollution rebound can be quantified.

[0041] Simultaneously, by applying a similar environmental mapping relationship, the incremental carbon dioxide emissions corresponding to potential greenhouse gases such as methane or nitrous oxide can be calculated, thus quantifying the risk of greenhouse gas rebound. This allows for the prediction of potential problems arising from scheduling actions during the pre-control decision-making stage, including short-term water quality dilution illusions and subsequent long-term rebound issues.

[0042] Step 105: Calculate the synergistic feasibility index of each candidate regulatory action by combining the benefits of water quality improvement, the risk of rebound in endogenous pollution, the risk of rebound in greenhouse gases, and the credibility of attribution.

[0043] Construct a comprehensive hedging evaluation model that balances positive environmental quality improvement with negative ecological and environmental rebound.

[0044] Specifically, the system sets the quantified water quality improvement benefits as a positive driving factor for evaluation decisions, and uses the predicted absolute values ​​of endogenous pollution rebound risk, greenhouse gas rebound risk, and data uncertainty caused by the lack of credibility of multi-source attribution at the front end as negative penalty constraints in the control equation.

[0045] By introducing a unified dimensionless reference parameter, linear penalty offset calculations are performed on various heterogeneous hydrodynamic and biochemical environmental indicators under a unified scale, resulting in a single quantitative scalar value, namely the collaborative feasibility index. This index can be used to intuitively determine the overall technical adaptability of candidate scheduling actions in the face of multi-objective conflict environments.

[0046] Step 106: Based on the collaborative feasibility index, select and configure control instructions for scheduling projects, point sources or area sources, and integrate them to obtain a collaborative management and control instruction package.

[0047] Based on the collaborative feasibility index calculated for each action, candidate control strategies that are deemed infeasible due to excessive environmental backlash penalties are filtered out according to the set feasibility threshold rules.

[0048] Based on the collaborative feasibility index values ​​of each candidate control action, suitable control actions are uniformly selected, and corresponding control commands are configured according to the selection results.

[0049] For actions that meet the execution conditions based on the collaborative feasibility indicators, they are directly configured as scheduling engineering control commands; for actions that do not meet the conditions as a whole, point source or area source control commands are configured instead.

[0050] For highly feasible actions that pass threshold verification, the system parses and extracts action parameters such as the specific execution control node object, trigger execution timestamp, and operation intensity, and translates and configures them into direct control instructions for scheduling engineering agencies such as reservoirs and dams.

[0051] In some optional implementations, if the system assessment finds that instructions solely for scheduling projects cannot be issued due to extremely low indicators caused by penalties such as rebound of endogenous pollution, a task shift compensation mechanism will be triggered to generate supplementary constraint instructions for point source emission entities such as upstream sewage treatment plants or other execution entities such as agricultural non-point source areas.

[0052] The discrete control instructions belonging to different levels of execution terminals are encapsulated and integrated into a complete collaborative management and control instruction package after time and space conflict investigation and collaborative arrangement.

[0053] Step 107: Perform rolling simulation and evaluation verification on the collaborative management and control instruction package, and output an executable collaborative management and control scheme.

[0054] Before actually issuing control commands over the network, the execution effect of the full-time control actions contained in the collaborative management and control command package is virtually simulated and verified using the previously established basin operation benchmark data network environment.

[0055] The dynamic evolution curves of the predicted cross-sectional water quality parameters over simulation time are tracked. If the simulation evaluation confirms that the execution trajectory of the instruction package meets the preset water environment capacity rebound constraint thresholds and the physical constraints of the hydraulic engineering structure, the relevant action parameter settings are locked, the internal simulation iterative optimization process ends, and the finally confirmed executable collaborative control scheme is output externally for the downstream scheduling execution control module to directly call and execute.

[0056] This rolling simulation verification step constitutes the last automatic safety defense network before the command takes effect and triggers on-site physical operations.

[0057] Through the coordinated implementation of steps 101-107 above, the quantitative calculation of the bottom sediment disturbance risk (step 104) and the multi-objective hedging assessment (step 105) were completed before the hydraulic scheduling decision took effect, so that the final output control scheme could avoid the problems of bottom sediment resuspension and anaerobic layer exposure caused by direct large-flow discharge in traditional methods.

[0058] Specifically, when the collaborative feasibility index of a certain high-flow discharge scheme is calculated to be negative, the system automatically downgrades the scheme to a segmented correction action or transfers the reduction task to a wastewater treatment plant with higher marginal efficiency. In this way, while maintaining the water quality of the section, the additional carbon emissions caused by the scheduling are controlled within the preset proportion of the baseline emissions.

[0059] Example 2 further explains the process of verifying the consistency of multi-source evidence and determining the type of trustworthy scenario for control.

[0060] In one possible implementation, the following steps are included:

[0061] like Figure 2 As shown, in step 201, the online monitoring contribution ratio corresponding to the multi-source observation data and the load simulation contribution ratio corresponding to the basin operation benchmark data package are extracted respectively.

[0062] After obtaining multi-source observation data and basin operation benchmark data, the contribution of pollutant sources to the same control section is analyzed.

[0063] The contribution ratio of online monitoring is reflected in the pollution source input ratio parameters fed back in real time by the actual flow rate of sewage outlets, the effluent concentration sensors of sewage treatment plants, and the rainfall and runoff monitoring stations; the contribution ratio of load simulation is reflected in the theoretical contribution ratio parameters of each type of pollution source derived based on historical model accumulation and source term accounting parameters.

[0064] The two sets of proportional arrays with the same spatiotemporal dimension labels are read through the data interface and used as the input basis for subsequent attribution consistency calculation.

[0065] Step 202: Calculate the sum of the absolute deviations between the online monitoring contribution ratio and the load simulation contribution ratio of various pollution sources, and normalize the sum of absolute deviations based on the theoretical maximum value to calculate the attribution confidence.

[0066] The L1 norm algorithm is used to process the discreteness of multi-source data arrays in order to quantify the consistency between physical data from multiple sensors and model prediction data.

[0067] Furthermore, the two contribution ratios extracted above are substituted into the following formula for calculation:

[0068] G t =1-∑(|A s,t -L s,t |) / D max ;

[0069] Among them, G t Let be the attribution confidence level within the t-th regulation period, s be the pollution source category number, and A be the attribution confidence level. s,t L represents the contribution ratio of the s-th pollution source obtained from online monitoring. s,t D represents the contribution ratio of the s-th type of pollution source obtained from the load simulation. max The theoretical maximum value of the sum of absolute deviations is given when the normalized proportions of each contribution satisfy ∑(A s,t )=1 and ∑(L s,t When )=1, D max The value is 2; ∑() is the summation operation.

[0070] Assuming a region has only two main pollution sources, the extracted online monitoring contribution ratios A are 0.8 and 0.2, and the extracted load simulation contribution ratios L are 0.1 and 0.9. Substituting these into the above calculation logic, the sum of absolute deviations is |0.8-0.1|+|0.2-0.9|=1.4; after normalization using the theoretical maximum value of 2, we get 1.4 / 2=0.7; after offsetting and subtracting using 1, we obtain the attribution confidence G. t =1-0.7=0.3.

[0071] Step 203: Compare the attribution credibility with the pre-configured credibility threshold, and classify the current control cycle's control credibility scenario type into a credible control scenario, a control scenario pending review, or an untrustworthy control scenario.

[0072] The pre-configured confidence threshold is reflected as the boundary of a set numerical range. In practice, the calculated attribution confidence scalar is compared with two confidence thresholds (for example, setting the high threshold to 0.7 and the low threshold to 0.4).

[0073] When the attribution credibility is greater than or equal to the high threshold, the system determines that the multi-source evidence points to the same conclusion and outputs the credible control scenario type as a credible control scenario.

[0074] When the attribution confidence level is between the high and low thresholds, output the management scenario to be reviewed;

[0075] When the attribution confidence level is below a low threshold, it indicates a large-scale conflict between the sensor's physical data acquisition and the theoretical model, resulting in an unreliable control scenario.

[0076] Step 204: When the trusted control scenario type is classified as an untrusted control scenario, the generation of candidate control action set is paused. Based on the current operating status data of each project and pollution source collected from the multi-source observation data, the minimum intervention plan to maintain the current operating status of each project and pollution source is executed, and the minimum intervention plan is output as an executable collaborative control plan.

[0077] When the scenario type is an untrusted control scenario, the foolproof control logic is triggered. At this time, the system cuts off the instruction path to the generation of high-intensity scheduling actions, and no longer generates reservoir release or sewage treatment plant strong control instructions with the risk of destructive disturbance based on conflicting data.

[0078] The current physical operating parameters of each scheduling project and sewage discharge unit are extracted to construct a minimum intervention plan to maintain the original state. This plan is then directly assigned to the execution layer as the final output of the current control cycle. This step provides a safety net for system operation in cases where data sources are contaminated or sensors fail.

[0079] Step 205: Simultaneously, increase the frequency of cross-sectional water quality monitoring and online monitoring of sewage outlets within the current control cycle, and write the increased monitoring data into the multi-source observation data for the next control cycle.

[0080] In the parallel process of outputting the minimum intervention plan, control commands are issued to the multi-source monitoring network to increase the sampling frequency. For example, the trigger interval of the water quality sensor, which originally collected samples every 4 hours, is shortened to once every 1 hour.

[0081] The additional physical sampling data packets are received and added to the basic database for use by the calculation module in the next cycle. This data reconstruction feedback action provides more timing verification evidence to fix multi-source consistency conflicts in subsequent runs.

[0082] By adopting a control mechanism based on absolute deviation consistency verification and bottom-line error prevention, the control system can significantly reduce blind high-intensity hydraulic intervention when facing complex operating conditions such as partial water quality sensor failure or inverted multi-source data. This improves the decision robustness of the basin scheduling system from passively accepting data distortion to actively tolerating and intercepting errors.

[0083] Example 3 further explains the process of identifying the environmental risk types of control sections and conducting preliminary safety screening of candidate control action sets.

[0084] In one possible implementation, the following steps are included:

[0085] Step 301: Extract the current flow rate of the control section from the multi-source observation data, and calculate the flow quantile by combining it with the historical seasonal flow rate data stored in the basin operation benchmark data package; extract the hourly dissolved oxygen data, water temperature data, effluent variation coefficient of upstream sewage treatment plant, and the number of consecutive drought days in the basin.

[0086] We analyze the time series characteristics in multi-source observation data and calculate and extract various environmental dynamics and statistical parameters.

[0087] Traffic quantiles are obtained by sorting historical traffic data sequences for the same season and calculating the current traffic position.

[0088] Daily hourly dissolved oxygen and water temperature data are directly fed back through the underlying water quality sensor array;

[0089] The effluent coefficient of variation is calculated by extracting the ratio of the standard deviation to the average value of the wastewater treatment plant's discharge concentration within a previously set time window;

[0090] The number of consecutive drought days in the watershed is calculated based on the continuous absence of effective precipitation records at meteorological rainfall monitoring nodes. These parameters constitute the basis for the criterion of scene quantitative identification.

[0091] Step 302: Based on the comparison results between the extracted parameters and the preset quantitative judgment conditions, the environmental risk type is specifically identified as at least one of the following: scheduling sensitive scenario, point source marginal control scenario, or rainfall surface source pulse scenario.

[0092] The preset quantitative criteria are manifested as a series of multi-dimensional logical judgment rules. The extracted physical parameters are then matched with these criteria rules.

[0093] In one execution judgment rule, when the extracted current flow rate is lower than the 25th percentile of the historical flow rate in the same season, and the hourly dissolved oxygen is lower than the environmental state reference value and the water temperature is higher than the corresponding reference value, the system will identify the environmental risk type as a scheduling sensitive scenario.

[0094] In another execution judgment rule, when the effluent variation coefficient of the upstream sewage treatment plant is greater than 0.3 and the plant's discharge load accounts for more than 20% of the total cross-sectional load, the system identifies it as a point source marginal control scenario.

[0095] In another set of enforcement rules, when the rainfall in the basin exceeds 20 mm in the past 48 hours and the previous consecutive drought days in the basin are greater than 7 days, the system identifies it as a rainfall pulse scenario. This identification result allows subsequent control decisions to focus on specific pollution sources and mechanisms.

[0096] Step 303: Based on the water supply safety boundary, dam operation safety boundary and minimum ecological flow constraint in the basin operation benchmark data package, conduct preliminary safety screening of the actions in the candidate control action set.

[0097] After generating a set of candidate control actions containing multiple scheduling options, a preliminary physical condition filtering operation is performed.

[0098] The water supply safety boundary is defined as the minimum guaranteed water level that the reservoir should maintain.

[0099] The safety boundary for dam operation is defined by the maximum allowable discharge flow and the upper limit of the hoist operating rate.

[0100] The minimum ecological flow constraint is reflected in the lower limit of the baseflow that maintains continuous flow in the river channel.

[0101] The estimated water level change, instantaneous flow rate, and operation command rate for each candidate action are extracted and compared one by one with the rigid safety constraint values ​​stored in the database.

[0102] Step 304: Eliminate candidate control actions that trigger any of the above safety boundary constraints, and retain actions that satisfy the engineering safety boundary to enter the subsequent rebound risk calculation process.

[0103] In the first execution path, when it is detected that the estimated discharge flow of a candidate control action causes the reservoir water level to drop below the water supply safety boundary elevation, or the estimated discharge flow is lower than the minimum ecological flow constraint, the system marks the action as illegal and removes it from the candidate control action set. This operation helps reduce the inclusion of invalid actions with engineering safety hazards in the subsequent high-computing-power rebound risk quantification calculation model.

[0104] In the second execution path, when none of the engineering prediction parameters for a candidate control action trigger the alarm thresholds for the water supply safety boundary, dam operation safety boundary, and minimum ecological flow constraint, the system retains the action in the candidate control action set. The retained action, as a valid search node, is passed to the next-level calculation interface for further quantitative assessment of sediment interface activation and multidimensional environmental rebound risk.

[0105] Example 4 further explains the process of predicting the water quality improvement benefits after the execution of each action within the candidate control action set. In one possible implementation, the following steps are included:

[0106] like Figure 3 As shown, in step 401, for each candidate regulatory action, the baseline predicted concentration when the regulatory action is not performed, and the action predicted concentration after the regulatory action is performed are predicted respectively.

[0107] Historical monitoring data and corresponding scheduling records are read in advance, and multiple linear regression is used to establish the response coefficient matrix of the control section to each disturbance source.

[0108] Specifically, historical scheduling event records from the past two years are extracted, with each record containing the actual operational changes ΔX for each scheduling project. j And the measured changes in the concentration of each pollutant at the control section after scheduling, ΔCr; and the ΔX of all scheduling events. j Using ΔCr as the independent variable matrix and ΔCr as the dependent variable vector, the response coefficient α is obtained by least squares fitting. r,j .

[0109] When collinearity exists among the independent variables, ridge regression can be used for regularization. The calibrated response coefficient matrix is ​​stored in the watershed operation benchmark data package for use in the current regulation cycle.

[0110] After entering the current regulation cycle, the current measured concentration of the target section is extracted and combined with trend extrapolation to calculate the baseline state data without taking any action, which is then used as the basic predicted concentration.

[0111] Extract the discharge flow increment parameters contained in the candidate control actions and input them into the response coefficient matrix for mapping calculation. The calculation process is represented by obtaining the action prediction concentration using the following linear text formula:

[0112] ;

[0113] in, To predict the concentration of the r-th pollutant index when executing the a-th candidate action, α represents the baseline predicted concentration of the r-th pollutant indicator when no candidate action is performed. r,j Let ΔX be the element in the r-th row and j-th column of the response coefficient matrix. j,a,t Let ∑ represent the change in the j-th scheduling project or pollution source among the a-th candidate actions, and let ∑ represent the summation operation over all scheduling projects and pollution source indices j.

[0114] Through the above calculations, the system quantifies the numerical change in concentration at the downstream section caused by a single scheduling operation.

[0115] Step 402: Using the target limit values ​​corresponding to each water quality indicator as a dimensionless benchmark, normalize the concentration difference between the basic predicted concentration and the action predicted concentration.

[0116] Considering the order-of-magnitude differences among different water quality indicators, directly using the absolute concentration difference for calculation would result in the larger value of the indicator covering the smaller value of the indicator.

[0117] Extract the water environment standard concentrations corresponding to each pollutant indicator and use them as target limits.

[0118] The difference between the baseline predicted concentration and the action predicted concentration is calculated, and this difference is divided by the corresponding target limit to convert the absolute concentration difference with different physical units into a dimensionless proportionality coefficient with a unified dimension.

[0119] Step 403: Based on the preset weights of each water quality indicator, the normalized concentration difference is weighted and calculated to obtain the water quality improvement benefits corresponding to each candidate control action.

[0120] The system extracts the pre-configured weights of various water quality indicators and sums them together with the dimensionless proportional coefficients obtained in previous steps. The calculation logic is implemented through the following formula:

[0121] ;

[0122] Among them, Q a,t For the water quality improvement benefits corresponding to the candidate actions, w r The water quality improvement weight for the r-th pollutant indicator is... Based on the predicted concentration, To predict concentration for action, Let r be the target limit for the r-th pollutant indicator, max() is the maximum value operation, and ∑() is the summation operation for all water quality indicators r.

[0123] In one scenario, assuming the target limit for ammonia nitrogen is 1.0, the baseline predicted concentration is 1.2, and the action predicted concentration is 0.9, with a corresponding weight of 0.6; the target limit for permanganate index is 5.0, the baseline predicted concentration is 6.0, and the action predicted concentration is 5.5, with a corresponding weight of 0.4. Substituting these values ​​into the above calculation logic, the normalized concentration difference for ammonia nitrogen is 0.3 / 1.0 = 0.3, and the normalized concentration difference for permanganate index is 0.5 / 5.0 = 0.1. Multiplying each value by its weight and summing the results, we obtain the water quality improvement benefit. This allows pollutants with different physical properties to exert a regulatory effect on the amount of improvement on the same mathematical scale.

[0124] Example 5 further explains the steps for quantifying the activation level of sediment interfaces and calculating the risk of endogenous pollution rebound and greenhouse gas rebound.

[0125] In one possible implementation, the following steps are included:

[0126] like Figure 4 As shown, in step 501, for each candidate control action, the estimated dissolved oxygen, estimated oxidation-reduction potential and estimated water temperature for the corresponding time period are obtained.

[0127] Using the specific scheduled execution time of the candidate control action as the index condition, access the hourly state distribution database within the basin operation benchmark data package.

[0128] By interpolating data within the target time period, the system obtains the estimated dissolved oxygen background value of the control section during that execution period.

[0129] Furthermore, the system corrects the above-mentioned background dissolved oxygen estimate based on the physical increase in oxygenation generated by the large-volume discharge of candidate actions in the drop area, thus obtaining a comprehensive dissolved oxygen estimate.

[0130] Using the same timing operation logic, the system synchronously extracts the underlying redox potential and water temperature prediction values ​​corresponding to the execution period. The introduction of dynamic parameters for the execution period reflects the thermodynamic differences caused by operations within different time windows.

[0131] Step 502: Combining the predicted flow velocity amplitude and water level amplitude caused by each candidate control action, calculate the state deviation of the predicted flow velocity amplitude, water level amplitude, dissolved oxygen estimate, redox potential estimate, and water temperature estimate relative to their respective environmental state reference values.

[0132] The monitoring data sequence of the same season in the preset year for the controlled river section is read, the median value is calculated and set as the environmental state reference value. Median statistics exclude the interference of extreme outlier data on the reference baseline. The predicted amplitude of river flow velocity and water level caused by candidate regulation actions is obtained. Together with the previously obtained estimated physical parameters, the absolute or relative differences between them and the corresponding reference values ​​of flow velocity, water level, dissolved oxygen, redox potential and water temperature are calculated and recorded as state deviations, and an array is constructed.

[0133] Step 503: Based on the preset sensitive weights of each state deviation, perform weighted calculations on each state deviation to quantify the degree of sediment interface activation corresponding to each candidate control action.

[0134] Environmental indicator data prior to historical sediment release events were extracted in advance. A logistic regression model was used to fit the proportion of each environmental variable deviating from the reference value, and the sensitive weight parameters of each state deviation were calculated.

[0135] The study uses the presence or absence of historical sediment release events as a binary classification label, and the deviation values ​​of five environmental states prior to each event as input features. The contribution coefficients of each feature to the probability of the release event are fitted using a logistic regression model, and the fitted feature coefficients are normalized and used as the corresponding weight parameters.

[0136] Specifically, sediment release monitoring records were collected within 72 hours after historical scheduling events in the target watershed, and each scheduling event was labeled as "triggered release event" or "not triggered release event" as a binary label;

[0137] A logistic regression classification model was established using the flow velocity deviation ratio, water level deviation ratio, dissolved oxygen deviation ratio, redox potential deviation ratio, and water temperature deviation ratio at the time of the corresponding scheduling event as independent variables.

[0138] After training, the absolute values ​​of the regression coefficients corresponding to the independent variables in the logistic regression model are extracted and normalized so that their sum is 1, thus obtaining the sensitive weight parameters corresponding to the state deviations of each variable.

[0139] The activated degree of the sediment interface is calculated using the extracted weight parameters according to the following formula:

[0140] ;

[0141] Among them, I a,t λ represents the degree of activation at the sediment interface. v , λ h , λ o , λ e , λ θ These are the weighted parameters for flow velocity, water level, dissolved oxygen, redox potential, and water temperature, respectively, Δv a,t With Δh a,t These represent the predicted variations in flow velocity and water level, respectively. ref h ref O ref E ref θ ref These are the environmental state reference values ​​corresponding to each physical quantity. , , These are the corresponding estimated values, where ε is a very small positive number, and max() is the operation to retrieve the maximum value.

[0142] The above calculation logic converts the physical disturbance index caused by hydraulic scheduling into a numerical value of the activation intensity of the reaction chemical interface.

[0143] Step 504: Determine the sediment-sensitive area affected by each candidate control action based on the hydraulic influence range of each candidate control action and the spatial distribution information of the riverbed sediment in the basin operation benchmark data package; determine the duration of the control disturbance based on the command parameters of each candidate control action.

[0144] Based on the attenuation propagation distance of the predicted velocity variation in the river channel topology model, the longitudinal range of the river segment covered by the hydraulic disturbance is determined.

[0145] Search the spatial distribution layer of sediment types within the river section, extract the water body projection area of ​​the area identified as having high release sensitivity, and use it as the sediment sensitivity area.

[0146] The gate opening time set by the read command parameters is used as the duration of the control disturbance.

[0147] Step 505: Multiply the sediment sensitive area, the duration of the regulation disturbance, and the degree of sediment interface activation, and combine them with the pre-stored sediment phosphorus release sensitivity coefficient to calculate the absolute release mass of the rebound as the risk of endogenous pollution rebound.

[0148] By reading the phosphorus release rate matrix data under different low oxygen and temperature gradients pre-tested in the laboratory, it is used as a phosphorus release sensitivity coefficient for sediments and pre-stored.

[0149] Extract the coefficient and perform the multiplication calculation according to the following formula:

[0150] ;

[0151] in, To mitigate the risk of a rebound in endogenous pollution, I a,t K represents the degree of activation at the sediment interface. p,t A is the sensitivity coefficient for phosphorus release from pre-stored sediments. sed,a,t For the sediment-sensitive area, T a,t To regulate the duration of the disturbance.

[0152] Step 506 involves combining the activation level of the sediment interface, the sensitive area of ​​the sediment, and the duration of the regulation disturbance with the pre-stored potential greenhouse gas release flux and carbon dioxide equivalent conversion parameters to estimate the corresponding increase in carbon dioxide equivalent as the risk of greenhouse gas rebound.

[0153] In one specific implementation, the potential greenhouse gas release flux is pre-stored based on the gas release rate obtained from a physical collection device deployed at the water-air interface.

[0154] Extract flux data and equivalent parameters, and calculate the increment using the following formula:

[0155] ;

[0156] in, To mitigate the risk of a rebound in greenhouse gas emissions, I a,t η represents the degree of activation at the sediment interface. m With η n The carbon dioxide equivalent conversion parameters for methane and nitrous oxide are respectively, F. m,t With F n,t A represents the potential release fluxes of pre-stored methane and nitrous oxide, respectively.sed,a,t For the sediment-sensitive area, T a,t To regulate the duration of the disturbance.

[0157] Furthermore, this embodiment provides an alternative extrapolation scheme for potential greenhouse gas release fluxes.

[0158] When there is no on-site measured data storage, the system uses a pre-configured empirical formula to estimate the potential greenhouse gas emission flux. The specific calculation operation is as follows:

[0159] ;

[0160] Among them, F m,t To estimate the potential methane release flux, k m The basic methane release constant, whose dimensions are derived from F m,t Objective Dimensions and TOC sed The dimensions of TOC are jointly determined and can be determined through nonlinear regression fitting of historical measured flux release data with corresponding sediment organic carbon content and temperature data. sed For pre-stored sediment organic carbon content parameters, β m Let θ be the temperature sensitivity coefficient. t The value is the measured or estimated bottom water temperature, and e is a natural constant. This alternative calculation logic maintains the data connectivity of the rebound risk accounting process.

[0161] By introducing environmental background prediction parameters for specific execution periods, a spatiotemporally coupled mathematical model for assessing interface activation and rebound was constructed. This model maps the previously difficult-to-quantify underlying hydrodynamic physical disturbances across scales into absolute mass scales for the rebound of endogenous phosphorus and greenhouse gases, which is beneficial for promoting pre-emptive quantitative mitigation of secondary pollution releases.

[0162] Example 6 further explains the steps of calculating the collaborative feasibility index of each candidate control action and configuring the scheduling engineering control command.

[0163] In one possible implementation, the following steps are included:

[0164] like Figure 5 As shown, in step 601, the benefits of water quality improvement are taken as a positive gain term, and the risk of rebound of endogenous pollution, the risk of rebound of greenhouse gases, and the attribution deficiency degree obtained by subtracting the attribution confidence from the value of 1 are taken as negative penalty terms.

[0165] Establish a decision-making framework for balancing the conflicting objectives of environmental governance:

[0166] The water quality improvement benefit obtained after normalization is extracted and defined as the positive gain term driving the execution of scheduling actions, representing the physical concentration reduction benefit brought about by the scheduling actions.

[0167] Extract the risk of endogenous pollution rebound and greenhouse gas rebound obtained from the prediction and assessment, and subtract the attribution confidence from the value 1 to obtain the attribution missing degree. This missing degree represents the decision uncertainty caused by the inconsistency of evidence from multiple sources of data, and is uniformly mapped as a negative penalty term that restricts the execution of scheduling actions.

[0168] Incorporating multi-dimensional environmental risks and data uncertainties into the penalty constraints provides a fundamental condition for comprehensively evaluating the destructive impact of scheduling actions.

[0169] Step 602: Introduce the allowable reference values ​​corresponding to the positive gain term and each negative penalty term, and perform denominator normalization on each index.

[0170] Since the positive gain term and each negative penalty term have different physical dimensions, direct combination will lead to numerical deviation.

[0171] Pre-configure permissible reference values ​​for each indicator at the underlying level. Extract reference values ​​corresponding to water quality improvement benefits, capacity reference values ​​corresponding to endogenous pollution rebound, and baseline carbon emission reference values ​​corresponding to greenhouse gases, and use them as the denominator for calculating the corresponding indicator items.

[0172] By dividing by their respective set allowable reference values, the system transforms variables with different physical properties and numerical magnitudes into dimensionless ratios that are uniformly on the same quantitative scale, thus eliminating the abnormal dominance of a single extreme value on the overall evaluation.

[0173] Step 603: Perform linear hedging summation on the normalized indicator items to obtain the synergistic feasibility index corresponding to each candidate regulatory action.

[0174] Extract the preset weight parameters for each indicator, and then perform a signed linear weighted summation on the normalized dimensionless ratios. The calculation process is implemented using the following formula:

[0175] ;

[0176] Among them, J a,t For collaborative feasibility indicators, As a weighting factor for water quality improvement benefits, Q a,t For the benefits of water quality improvement, Q tol This refers to the permissible reference value corresponding to the benefits of water quality improvement. Weighting for penalties to prevent a rebound in endogenous pollution. To mitigate the risk of a rebound in endogenous pollution, P tol This is the allowable value for rebound of endogenous pollution. Weighting for greenhouse gas rebound penalties To mitigate the risk of a rebound in greenhouse gas emissions, B tol This is the allowable value for greenhouse gas rebound. As a weight for attribution credibility penalty, G t For attribution credibility.

[0177] In a certain scenario, assuming that after normalization, the ratio of the water quality positive gain term for a certain candidate action is 0.9, the ratio of the endogenous phosphorus rebound penalty term is 0.3, the ratio of the greenhouse gas rebound penalty term is 0.1, and the attribution confidence of multi-source data is 0.9.

[0178] Assuming each item has a weight of 1.0, the collaborative feasibility index J is calculated by substituting these values ​​into the formula. a,t =0.9-0.3-0.1-0.1=0.4. This indicator, as a single scalar, directly reflects the degree of coordinated feasibility of actions. Weight parameters for each indicator. , , , The determination method is as follows: based on the relative priorities of the four objectives in the watershed management goals—water quality improvement, endogenous pollution control, carbon emission control, and data reliability—the watershed management decision-makers determine the relative weights using the analytic hierarchy process (AHP), and then normalize the weights so that the sum of all weights is 1. In this embodiment, Set it to 0.4. Set it to 0.25. Set it to 0.2. Set it to 0.15.

[0179] Step 604: Set the collaborative feasibility threshold and the prohibition threshold.

[0180] The decision engine pre-defines boundary conditions for classifying action levels. The collaborative feasibility threshold is set as the mathematical equilibrium point after offsetting positive returns and negative risks, for example, its value is set to 0;

[0181] The prohibition threshold is set to the limit of rebound risk that the system can tolerate, for example, setting its value to -0.5. The system loads the above threshold into the logic judge as a hard judgment basis for subsequent action status marking.

[0182] Step 605: When the collaborative feasibility index is greater than the collaborative feasibility threshold, the corresponding candidate control action is marked as an action that can be directly executed.

[0183] Extract the collaborative feasibility indicators corresponding to each candidate control action and compare them with the collaborative feasibility threshold.

[0184] If the calculated index scalar is greater than the collaborative feasibility threshold, it indicates that the positive water quality improvement effect of the scheduling action outweighs the combined negative impact of endogenous rebound, carbon emissions and data unreliability. The action is then labeled with a data tag that can be directly executed and retained in the master control queue.

[0185] Step 606: When the collaborative feasibility index is between the collaborative feasibility threshold and the prohibition threshold, the corresponding candidate control action is marked as an action that needs to be corrected in stages.

[0186] If the extracted synergistic feasibility index is less than or equal to the synergistic feasibility threshold, but greater than the prohibition threshold, it indicates that although the action carries a greater than expected risk of physical disturbance rebound, it still has a controllable contribution value in terms of water quality improvement.

[0187] The action is labeled with data tags indicating that it needs to be corrected in segments, and then it is included in the sequence of instructions to be modified. The subsequent perturbation reduction algorithm module then deconstructs and reassembles the action.

[0188] Step 607: When the collaborative feasibility index is lower than the prohibition threshold, the corresponding candidate control action is marked as an action not recommended for execution.

[0189] If the extracted synergistic feasibility index is less than or equal to the prohibition threshold, it indicates that the cost of secondary pollution of bottom sediments and greenhouse gas carbon emissions that may be induced by this scheduling action far outweighs its water quality dilution benefits.

[0190] The action is labeled with a data tag indicating that it is not recommended to be performed, and it is then forcibly removed from the available scheduling queue during subsequent configuration to avoid issuing scheduling instructions with a high risk of environmental damage.

[0191] Step 608: When configuring control instructions for scheduling engineering, if the target control action is marked as an action that needs to be segmented and corrected, then perform the following configuration: split the original continuous discharge action into multiple low-disturbance sub-actions.

[0192] For the control parameter package that requires segmented modification of action labels, initiate the instruction structure reorganization operation.

[0193] The system reads the duration of the continuous high-flow discharge and the target total discharge volume set in the original instruction, and then divides the continuous process proportionally or non-proportionally along the time axis to generate several independent, low-disturbance sub-actions with reduced parameter scale. This splitting helps reduce the continuous high-intensity scouring of the river channel and reduces the intensity of single activation of the sediment interface by the bottom flow field.

[0194] Step 609: Add control conditions based on the preset limit of water level change per unit time in the basin operation reference data packet to the low-disturbance sub-actions. Based on the intraday water temperature and dissolved oxygen change trends predicted by the multi-source observation data, determine the estimated high-temperature and low-oxygen periods, set the execution sequence to avoid the estimated high-temperature and low-oxygen periods, and extend the interval reoxygenation time between each low-disturbance sub-action to generate the corrected scheduling control instructions.

[0195] Additional physical boundary constraints are injected into each generated low-disturbance sub-action. A water level drop rate limiting constant is added to the system to forcibly reduce the amplitude of instantaneous water level changes caused by engineering actions.

[0196] At the same time, the system retrieves the preset environmental prediction model, locates the time period during the day when high water temperature and low dissolved oxygen are expected, and sets the start timestamps of each low-disturbance sub-action to avoid the above-mentioned sensitive period through time-series translation operation.

[0197] The system also forces the insertion of time stagnation intervals between two adjacent low-disturbance sub-actions to provide physical time for natural reoxygenation of the water interface. After the above multi-dimensional parameter coverage and reorganization, the system completes the smooth degradation of high-rebound-risk actions and outputs corrected scheduling and control commands.

[0198] Example 7: Unlike Example 6 which focuses on the operation branch of the scheduling project, this example provides an alternative configuration scheme when scheduling actions are limited. It explains the collaborative compensation link for configuring control commands for point sources and configuring segmented control commands for area sources.

[0199] In one possible implementation, the following steps are included:

[0200] Step 701: When all candidate control actions for the scheduling project are marked as not recommended actions, determine the upstream sewage treatment plants of the control section based on the spatial correlation in the basin operation baseline data package, and calculate the reduction in section risk and the increase in carbon emissions brought about by each upstream sewage treatment plant performing enhanced control operations.

[0201] When all candidate hydraulic actions of the upstream scheduling project are identified as not recommended for execution due to low coordination feasibility indicators, the system activates the point source pollution transfer compensation logic.

[0202] By traversing the wastewater treatment plants upstream of the control section and establishing topological connections, and using the water quality response coefficient matrix, the weighted reduction value of the exceedance of each water quality index at the corresponding control section is calculated if the aeration intensity of nitrogen and phosphorus removal or the amount of carbon source added is increased. This value is then used as the risk reduction amount of the section.

[0203] The calculation method for the reduction in cross-sectional risk is the same as the calculation method for the benefits of water quality improvement, namely, a dimensionless weighted value based on the target limit.

[0204] Simultaneously, the system combines pre-stored electricity emission factors and reagent emission factors to calculate the additional greenhouse gas equivalent generated by implementing the above-mentioned enhanced control process, and uses it as the increased carbon emissions.

[0205] Step 702: Introduce the baseline carbon emissions corresponding to each wastewater treatment plant pre-stored in the basin operation baseline data package, and perform dimensionless processing on the relative increase of carbon emissions.

[0206] To reduce the unfair evaluation caused by the difference in absolute carbon emission scale between ultra-large and small wastewater treatment plants, the system extracts the historical average carbon emission equivalent data of each wastewater treatment plant under normal operating conditions and uses it as the benchmark carbon emission. The system calculates the ratio between the increased carbon emissions and the benchmark carbon emissions, generating a dimensionless value reflecting the relative increase in carbon emissions.

[0207] Step 703: Calculate the ratio between the reduction in cross-sectional risk and the increase in carbon emissions after dimensionless treatment to obtain the marginal synergistic efficiency of each wastewater treatment plant.

[0208] The carbon emission compensation efficiency of each plant was uniformly assessed using a formula.

[0209] Specifically, ;

[0210] Among them, M u,t Let ΔR be the marginal synergistic efficiency of the u-th wastewater treatment plant within the t-th control cycle. u,t ΔE represents the reduction in cross-sectional risk. u,t To increase carbon emissions, E u.base ε is the baseline carbon emissions, and ε is a preset minimum positive number to prevent the denominator from being zero.

[0211] This ratio represents the environmental risk reduction return that can be obtained by increasing the carbon emissions of the execution unit by a certain percentage, and provides a mathematical basis for the system to find pollution reduction subjects with low carbon costs.

[0212] Step 704: Select target wastewater treatment plants based on marginal synergistic efficiency and back-calculate control intensity to configure and generate point source low-carbon operation setpoints.

[0213] The calculated marginal synergy efficiencies of each wastewater treatment plant are ranked in descending order.

[0214] Prioritize selecting the wastewater treatment plant node with the highest marginal synergy efficiency as the target entity to undertake the alternative emission reduction task, and use the environmental response matrix to calculate the minimum effluent target and corresponding process operation parameters required to meet the cross-section pollution control, solidify them as point source low-carbon operation setpoints and issue them for execution.

[0215] Step 705: Divide the future timeline after the rainfall event is triggered into several equally spaced periods.

[0216] When a rainfall event is identified as a surface source pulse scenario, the surface source collaborative control logic is activated. After a rainfall event is detected, the predicted rainfall start point is used as the origin, and the forecast time length for the next few days is divided into integer hours according to the set duration, constructing a discrete time step sequence with the same time span, i.e., equally spaced time periods.

[0217] Step 706: Based on the rainfall monitoring information in the multi-source observation data and the area source parameters in the watershed operation benchmark data package, predict the area source pollutant inflow load and the estimated amount of greenhouse gas emissions from farmland for each time period, so as to identify the peak period of pollutant inflow load and the peak period of greenhouse gas emissions.

[0218] By combining rainfall intensity and agricultural non-point source surface parameters, the runoff scouring effect within each discrete time step is calculated, and the discrete time step with the highest load value is extracted as the peak period of pollutant inflow load into the river.

[0219] Meanwhile, the system estimates the change curves of methane and other gases produced by farmland affected by rainfall and flooding, and extracts the discrete time step with the highest emission equivalent as the peak period of greenhouse gas emissions.

[0220] Step 707: Determine whether the time interval between the peak period of pollutant inflow into the river and the peak period of greenhouse gas emissions is within a preset threshold.

[0221] Obtain the center timestamps of the two peak periods mentioned above, and calculate the absolute span difference between them on the time axis. The system extracts a preset time interval threshold constant for comparison. The preset threshold provides a benchmark scale for judging the physical overlap effect between the two, such as determining whether the interval exceeds a certain number of time spans.

[0222] Step 708: If yes, then determine that the two types of peaks are in the same time window, and collaboratively generate a joint control sequence that includes pre-rain interception, rain-induced slow drainage, and post-rain low-disturbance drainage, and configure it as a segmented control command for surface sources.

[0223] If the system detects that the absolute span difference is less than a preset threshold during the judgment process, the system identifies a high degree of physical resonance and overlap between the intense influx of non-point source pollutants into the river and the concentrated release of greenhouse gases from farmland. At this time, the system calls the collaborative intervention template to generate a three-stage time-series control action, which includes physical interception of pollutants in ditches before the rain, slowing down the discharge by reducing the flow rate of water pumps during the peak period of rainfall, and conducting low-disturbance drainage after the reoxygenation conditions improve in the later stage of rainfall. This action is encapsulated as a segmented control command for non-point sources, thereby mitigating the high-intensity superposition impact of pollution and carbon emissions from the perspective of spatial watershed management.

[0224] Establishing a collaborative hedging evaluation system and a point and non-point source compensation and transfer mechanism that incorporates environmental capacity benchmarks alleviates the limitations of unidirectional scheduling that relies solely on hydraulic dilution, and is conducive to promoting a balanced trade-off between positive environmental benefits and secondary rebound costs. Furthermore, when reservoir scheduling is limited, it can smoothly transfer emission reduction tasks to wastewater treatment plants or non-point source pollution control windows with high marginal synergistic efficiency, which is conducive to achieving long-term and stable multi-objective governance effects of watershed pollution reduction and carbon reduction.

[0225] If not, then generate a separate area source interception control command for the peak period of pollutant inflow into the river and configure it as an area source segmented control command.

[0226] Example 8 further explains the process of conducting rolling simulation and evaluation verification of the collaborative control instruction package.

[0227] In one possible implementation, the following steps are included:

[0228] Step 801: Execute a full-process prediction simulation using the collaborative control instruction package. If the simulation evaluation result exceeds the preset rebound threshold or engineering executable constraint, trigger a dynamic correction rule to intervene in the collaborative control instruction package.

[0229] The previously generated collaborative control instruction package is used as input parameters and loaded into the virtual simulation environment constructed from the basin operation benchmark data package. The virtual simulation environment uses the water quality response coefficient matrix and sediment interface activation parameters as the core inference engine, and gradually superimposes the control actions of each execution unit according to the preset rolling time step to calculate the evolution trajectory of water quality concentration, endogenous release and greenhouse gas emissions at the control section at each time step.

[0230] Based on the preset rolling time step, the evolution of each execution unit's actions in the spatial pipeline network and river topology is simulated, and the water quality improvement, endogenous phosphorus release, and greenhouse gas emissions during the evolution process are extracted.

[0231] Preset rebound thresholds are introduced as the baseline for safety assessment. For example, the rebound threshold for endogenous phosphorus is set at 10% of the total phosphorus target control capacity of the control section, and the rebound threshold for greenhouse gases is set at 5% of the baseline carbon emissions for a single cycle in the watershed. The extracted simulation data are then compared with the above thresholds.

[0232] When any simulated data exceeds the corresponding set rebound threshold, or when the virtual evolution state violates engineering executable constraints such as reservoir water supply level, the system determines that the current instruction package has a risk of being executed, and then activates the built-in dynamic correction rules to take over the current instruction data.

[0233] Step 802, the dynamic correction rules include: reducing the unit time water level change amplitude of the scheduling action and increasing the segmented release interval, or removing the control instructions corresponding to the execution unit that generates the boundary index, and transferring and allocating the pollution reduction task to the control instructions with higher ratios according to the ratio between the pollution reduction contribution and carbon emission increment of the remaining available control instructions.

[0234] The system initiates an intervention procedure for the instruction sequence that triggers the out-of-bounds alarm.

[0235] Under the first intervention path, when the alarm source is traced back to the excessive sediment disturbance caused by the large-volume release of a specific reservoir, the system reads the dispatch instruction, extracts the water level drop parameters, and reduces the amplitude setting value per unit time according to the preset step size.

[0236] Simultaneously, the time interval between the low-disturbance sub-actions after the action is broken down is lengthened. Under the second intervention path, when simply adjusting the hydraulic parameters cannot restore it to within the safe threshold, or when the modified hydraulic action cannot meet the basic water quality improvement target, the system directly deletes the control command of the execution unit from the sequence.

[0237] The system searches for other available execution entities within the same period, performs a descending index based on the marginal synergy efficiency calculated in the previous period, extracts the upstream sewage treatment plants or non-point source control nodes with the highest marginal synergy efficiency, converts the missing reduction tasks into the corresponding reagent addition increments or rainwater temporary storage volumes, and adds them to the instruction parameters of the above-mentioned alternative nodes.

[0238] Step 803: Use the revised collaborative management instruction package to re-simulate and evaluate until all constraints are met, and then output an executable collaborative management scheme.

[0239] The system obtains the collaborative control instruction package after the above intervention procedure reconstruction, and reloads it into the virtual simulation environment for a new round of full-process state simulation.

[0240] The above "deduction-comparison-correction" calculation chain is executed cyclically. After each iteration, the system checks whether the environmental and engineering constraints are met.

[0241] When all evaluation indicators obtained from the current simulation are detected to be within the preset safety thresholds and no engineering red lines have been triggered, the system terminates the simulation iteration loop. The system locks the current combination of control parameters, encapsulates it into a data file that can be read and executed by external physical devices, and outputs it as the final executable collaborative management and control scheme.

[0242] Example 9 provides a non-real-time closed-loop parameter optimization method, which describes the underlying parameter update steps based on execution feedback after outputting an executable collaborative management and control scheme.

[0243] In one possible implementation, the following steps are included:

[0244] Step 901: Obtain the measured feedback data after the execution of the executable collaborative control scheme, and compare it with the corresponding prediction results of each evaluation object stored in the prediction stage to calculate the prediction deviation.

[0245] After the executable collaborative control plan is issued to the physical execution end and a complete control cycle has passed, the system reads the measured feedback data for that cycle in batches through the data acquisition interface. The measured feedback data includes the cross-sectional pollutant concentration reported by the online water quality monitoring network, the dissolved oxygen concentration at the bottom layer returned by the physical detection device, and the greenhouse gas release rate estimated by the gas capture device. The system retrieves the corresponding prediction results for the same cycle from the historical prediction database. The system performs deviation calculations for each evaluation object:

[0246] ;

[0247] Among them, D k,t The prediction deviation of the k-th evaluation object within the t-th regulation cycle. For the measured feedback data observations of the k-th evaluation object, For the corresponding predicted values, The set allowable deviation reference value is ε, which is a preset minimum positive number to prevent the denominator from being zero.

[0248] Step 902: Attribution classification of prediction bias is performed to identify the dominant error sources causing the bias. The specific determination rules are as follows:

[0249] Calculate the prediction deviation D for each evaluation object. k,t The samples were then grouped by category to obtain the mean deviation D for hydrodynamic categories. hydro Mean deviation of sediment type D sed Greenhouse gas type deviation mean D ghg Attribution bias mean D attr ;

[0250] The category corresponding to the largest mean of the four types of deviations mentioned above is determined as the dominant source of error for the current cycle. The dominant source of error includes at least one of the following: hydrodynamic response error, sediment release error, greenhouse gas factor error, and attribution confidence error.

[0251] Among them, the mean deviation of hydrodynamic type D hydro To control the prediction bias of pollutant concentration at cross-sections; mean bias D for sediment types. sed The prediction bias is for the bottom-level total phosphorus concentration; the mean bias for greenhouse gases is D. ghgThe prediction bias for greenhouse gas emissions at the water-air interface; mean attribution bias D attr It is obtained by calculating the deviation between the measured contribution ratio of each source and the predicted contribution ratio.

[0252] All calculated prediction bias values ​​are aggregated, and clustering and classification algorithms are executed.

[0253] Analyze the characteristic variable domains in which various deviations occur. For example, when deviations are concentrated in the relationship between water level / flow rate and concentration diffusion, they are identified as hydrodynamic response errors.

[0254] When the deviations are concentrated in the positive correlation range between total phosphorus concentration and bottom water temperature and low oxygen, they are identified as sediment release errors; when the deviations are concentrated in the methane equivalent estimation stage, they are identified as greenhouse gas factor errors.

[0255] When the deviation manifests as persistent inversion of multi-sensor data or uninterrupted conflicting data, it is identified as attribution confidence error. Through this classification operation, the system locates the root node causing prediction distortion.

[0256] Step 903: Based on the identified dominant error sources, update the underlying operating parameters for the next control cycle accordingly.

[0257] A dynamic parameter optimization interface is configured to map the identified dominant error types to specific underlying model parameter arrays. This step provides the system with a mechanism for self-correction based on operational experience, specifically including the following four parallel update execution paths.

[0258] In the first execution path, when the dominant error source is hydrodynamic response error, the water quality response coefficient of the control section to the regulation action is updated accordingly. The system extracts the actual value of the scheduled discharge flow and the measured concentration variation within the regulation cycle, and uses a historical calibration algorithm to refit the regression matrix, covering the elements of the old linear response coefficient matrix in the basin operation benchmark data package, thereby improving the fitting degree of the dilution diffusion prediction for the next cycle.

[0259] In the second execution path, when the dominant error source is sediment release error, the sediment release sensitivity parameters and interface activation weights used to calculate endogenous pollution rebound are updated accordingly. Measured sediment flow velocity and dissolved oxygen status are extracted, and the bias gradient descent algorithm is applied to adjust the corresponding weight coefficients in the interface activation degree calculation formula. Simultaneously, newly acquired static release data from the laboratory or field are read and used to replace the pre-stored sediment phosphorus release sensitivity coefficient constant.

[0260] In the third execution path, when the dominant error source is greenhouse gas factor error, the potential greenhouse gas emission flux and equivalent conversion parameters are updated accordingly. The system extracts the latest measurement data of the water-gas interface, modifies the basic methane emission constant or temperature sensitivity coefficient in the empirical estimation formula, and calibrates the underlying carbon emission conversion mapping relationship.

[0261] In the fourth execution path, when the dominant error source is attribution confidence error, the confidence threshold used to determine the confidence scenario type is increased accordingly. By extracting the historical calculation sequence of L1 norm attribution confidence, and discovering that a large amount of abnormal data bypasses the interception system, a direct instruction is sent to the gating decision module to increase the original confidence threshold boundary, for example, from 0.6 to 0.7. In the next cycle, the security access control that triggers direct control actions is tightened, enhancing the system's anti-interference defenses.

Claims

1. A data processing method for coordinated management and control of watershed pollution reduction and carbon reduction for river and lake restoration, characterized in that, include: Acquire basic watershed data and establish a watershed operation benchmark data package that includes spatial correlations, pollution source accounting parameters, and scheduling engineering constraints; Collect multi-source observation data for the current regulation cycle, combine it with the basin operation benchmark data package to perform multi-source attribution consistency verification, obtain attribution credibility and determine the current management and control credibility scenario type; Based on multi-source observation data, the environmental risk types of control sections are identified, and candidate control action sets are generated by combining the reliable control scenario types with the scheduling engineering constraints in the basin operation benchmark data package. Predict the water quality improvement benefits after the execution of each action in the candidate set of regulatory actions, assess the degree of sediment interface activation caused by each action, and calculate the risk of endogenous pollution rebound and greenhouse gas rebound induced by interface disturbance. By combining the benefits of water quality improvement, the risk of rebound of endogenous pollution, the risk of rebound of greenhouse gases, and the credibility of attribution, the synergistic feasibility index of each candidate regulatory action is calculated. Based on collaborative feasibility indicators, control instructions for scheduling projects, point sources or area sources are selected and configured, and integrated to obtain a collaborative management and control instruction package. Perform rolling simulations and evaluations of the collaborative management instruction package, and output an executable collaborative management solution.

2. The method according to claim 1, characterized in that, In the process of combining the baseline data package of the watershed operation to perform multi-source attribution consistency verification, obtaining attribution credibility, and determining the current management and control credibility scenario type, the specific steps include: Extract the online monitoring contribution ratio corresponding to the multi-source observation data and the load simulation contribution ratio corresponding to the basin operation baseline data package respectively; The sum of the absolute deviations between the online monitoring contribution ratio and the load simulation contribution ratio of various pollution sources is calculated, and the result is normalized based on the theoretical maximum value of the deviation to obtain the attribution confidence. The attribution credibility is compared with the pre-configured credibility threshold to classify the current control cycle's control credibility scenario type into a credible control scenario, a control scenario pending review, or an untrustworthy control scenario.

3. The method according to claim 2, characterized in that, When a trusted control scenario is classified as an untrusted control scenario, the generation of control actions is suspended, the minimum intervention plan to maintain the current operating status of each project and pollution source is executed, and the minimum intervention plan is output as an executable collaborative control plan. At the same time, the frequency of cross-sectional water quality monitoring and online monitoring of sewage outlets will be increased during the current regulation cycle, and the increased monitoring data will be incorporated into the multi-source observation data for the next regulation cycle.

4. The method according to claim 1, characterized in that, Environmental risk types of control sections are identified based on multi-source observation data, including: Extract the flow quantiles, hourly dissolved oxygen data, water temperature data, effluent variation coefficients from upstream wastewater treatment plants, and the number of consecutive drought days in the basin from multi-source observation data. Based on the fact that the extracted parameters meet the preset quantitative criteria, the environmental risk type is specifically identified as at least one of the following: scheduling sensitive scenario, point source marginal control scenario, or rainfall surface source pulse scenario.

5. The method according to claim 1, characterized in that, Predict the water quality improvement benefits after the execution of each action within the candidate control action set, including: For each candidate regulatory action, the baseline predicted concentration when the regulatory action is not executed, and the predicted concentration after the regulatory action is executed are predicted respectively. Using the target limit values ​​corresponding to each water quality indicator as a dimensionless benchmark, the concentration difference between the basic predicted concentration and the action predicted concentration is normalized. The water quality improvement benefits corresponding to each candidate regulation action are obtained by weighted calculation based on the normalized concentration difference.

6. The method according to claim 1, characterized in that, Assess the degree of activation of the sediment interface caused by each action, including: For each candidate control action, obtain the corresponding estimated values ​​of dissolved oxygen, redox potential, and water temperature during the scheduled execution period. By combining the predicted flow velocity and water level fluctuations caused by each candidate control action, the state deviation relative to each environmental state reference value is calculated. The degree of sediment interface activation corresponding to each candidate regulatory action is quantified by weighting the deviations of each state.

7. The method according to claim 6, characterized in that, Calculate the risk of endogenous pollution rebound and greenhouse gas rebound induced by interface perturbation, including: Determine the sediment-sensitive area affected by each candidate regulatory action and the duration of the regulatory disturbance; The absolute release mass of the rebound is calculated as the risk of endogenous pollution rebound by multiplying the sensitive area of ​​sediment, the duration of regulation disturbance, and the degree of activation of sediment interface, and combining it with the pre-stored sediment phosphorus release sensitivity coefficient. Furthermore, by combining the pre-stored potential greenhouse gas emission fluxes and carbon dioxide equivalent conversion parameters, an equivalent conversion is performed to estimate the corresponding increase in carbon dioxide equivalent as the risk of greenhouse gas rebound.

8. The method according to claim 1, characterized in that, Calculate the synergistic feasibility indicators for each candidate regulatory action, including: The benefits of water quality improvement are treated as positive gain items, while the risks of endogenous pollution rebound, greenhouse gas rebound, and missing values ​​calculated based on attribution credibility are treated as negative penalty items. By introducing allowable reference values ​​corresponding to positive gain terms and each negative penalty term, the denominator of each index is normalized. By performing linear hedging summation on the normalized indicators, we obtain the synergistic feasibility indicators corresponding to each candidate regulatory action.

9. The method according to claim 8, characterized in that, After calculating the synergistic feasibility index corresponding to each candidate regulatory action, the following is also included: Set collaborative feasibility thresholds and prohibition thresholds; When the collaborative feasibility index is greater than the collaborative feasibility threshold, the corresponding candidate control action is marked as an action that can be directly executed. When the collaborative feasibility index is between the collaborative feasibility threshold and the prohibition threshold, the corresponding candidate control action is marked as an action that needs to be corrected in stages. When the collaborative feasibility index is lower than the prohibition threshold, the corresponding candidate control action will be marked as an action not recommended for execution.

10. The method according to claim 9, characterized in that, Configure control commands for point sources, including: When the control actions for the scheduling project do not meet the execution conditions, calculate the reduction in cross-sectional risk and the increase in carbon emissions resulting from the upstream sewage treatment plant's enhanced control operations. The baseline carbon emissions corresponding to each wastewater treatment plant are introduced, and the increase in carbon emissions is treated as a dimensionless relative increase. The marginal synergistic efficiency of each wastewater treatment plant is obtained by calculating the ratio between the reduction in cross-sectional risk and the increase in carbon emissions after dimensionless treatment. Based on the marginal synergistic efficiency, target wastewater treatment plants are selected and the control intensity is deduced to configure and generate point source low-carbon operation setpoints.

Images