A Watershed Dual-Threshold Collaborative Management Method Based on Big Data Processing

Abstract

This invention relates to the field of watershed risk management, specifically to a watershed dual-threshold collaborative management method based on big data processing. The method includes: first, acquiring multi-source monitoring data and generating an event state set; then, constructing a dynamic risk propagation network and extracting propagation characteristics, recovery characteristics, and the effectiveness of management measures; further determining the environmental risk threshold boundary, the watershed climate resilience threshold boundary, the dual-threshold state, and the cumulative resilience loss; determining the propagation blocking minimum cut set and the recovery control minimum set based on counterfactual simulation, generating collaborative management instructions, and executing single-layer updates. This invention can improve the continuity of watershed risk identification, the targeting of management actions, and the stability of feedback correction.

Description

Technical Field

[0001] This invention relates to the field of watershed risk management, specifically to a watershed dual-threshold collaborative management method based on big data processing. Background Technology

[0002] As watershed management shifts from single-factor pollution control to multi-factor collaborative management, the industry's requirements for integrated risk identification, process control, and dynamic scheduling are constantly increasing. Existing technologies mostly adopt static zoning, single-threshold early warning, or single-time scheduling methods, which usually focus on risk assessment at a certain point in time. They are difficult to continuously identify risk propagation paths, changes in resilience, and the cumulative impact of multiple disturbances. This can easily lead to problems such as delayed threshold determination, scattered control actions, and untimely feedback and correction, which restrict the refined, coordinated, and sustainable implementation of watershed management. Summary of the Invention

[0003] This invention provides a watershed dual-threshold collaborative management and control method based on big data processing, which is used to at least solve the problem of how to simultaneously identify the watershed risk expansion boundary and resilience boundary under continuous disturbance conditions, and generate executable collaborative management and control actions accordingly.

[0004] This invention provides a watershed dual-threshold collaborative management and control method based on big data processing, the method comprising: Acquire multi-source monitoring data of the target watershed, divide the multi-source monitoring data into risk response units, event segments, and spatiotemporal alignment, and generate an event state set; A dynamic risk propagation network is constructed based on the event state set, and the propagation characteristics, recovery characteristics, and effectiveness of control measures are determined based on the dynamic risk propagation network; Based on the characteristics of propagation and recovery, the environmental risk threshold boundary, the watershed climate resilience threshold boundary, the dual threshold state, and the cumulative amount of resilience loss are determined. Based on the dual threshold state, the effectiveness of control measures, and the cumulative amount of resilience loss, counterfactual simulations are performed on candidate control actions to determine the minimum cut set for propagation blocking and the minimum set for recovery control. Based on at least one of the minimum cut set for propagation blocking and the minimum set for recovery control, collaborative control instructions are generated, and the parameters of the dynamic risk propagation network or the threshold judgment parameters are updated in a single layer according to the physical residual after the control is executed.

[0005] In one possible implementation, the multi-source monitoring data includes meteorological monitoring data, hydrological monitoring data, water quality monitoring data, pollution discharge record data, dispatching and operation data, and shoreline remote sensing interpretation data; the event state set includes pollution input load, transport capacity, shoreline storage status, ecological buffer margin, engineering storage margin, and receptor exposure sensitivity.

[0006] In one possible implementation, a dynamic risk propagation network is constructed based on the event state set, including: extracting event nodes from the event state set according to the time of state change; sequentially associating event nodes according to a preset propagation time window to form a propagation sub-chain; and constructing a dynamic risk propagation network with risk response units as nodes and the propagation relationship between risk response units as edges.

[0007] In one possible implementation, the propagation characteristics, recovery characteristics, and effectiveness of control measures are determined based on a dynamic risk propagation network, including: performing time-series correlation analysis based on propagation subchains to determine the propagation intensity of each propagation relationship; determining the effectiveness of control measures corresponding to each propagation relationship based on quality throughput consistency verification and spatial distribution consistency verification; determining propagation characteristics based on propagation intensity; and determining recovery characteristics based on changes in the event state set during the propagation and recovery phases.

[0008] In one possible implementation, propagation characteristics include the length of the propagation front, the number of continuously affected risk response units, and the number of active propagation relationships; recovery characteristics include the shoreline saturation level and the recovery rate of the monitoring section.

[0009] In one possible implementation, the environmental risk threshold boundary, the watershed climate resilience threshold boundary, and the dual-threshold state are determined based on propagation and recovery characteristics. This includes: determining the environmental risk threshold boundary by performing connectivity analysis based on changes in the propagation front length, the number of continuously affected risk response units, and the number of active propagation relationships; determining the watershed climate resilience threshold boundary by determining whether the recovery time of the monitoring section and the shoreline segment corresponding to the risk response unit exceeds a preset recovery window; and determining the dual-threshold state based on the environmental risk threshold boundary and the watershed climate resilience threshold boundary.

[0010] In one possible implementation, the cumulative amount of resilience loss is determined based on propagation and recovery characteristics, including: after each event segment ends, determining the amount of unrecovered buffer based on the shoreline storage state, ecological buffer margin, and engineering regulation margin, and accumulating the unrecovered buffer to obtain the cumulative amount of resilience loss.

[0011] In one possible implementation, counterfactual simulations are performed on candidate control actions to determine the minimum cut set for propagation blocking and the minimum set for recovery control. This includes: simulating the impact of a single candidate control action and a combination of candidate control actions on the degree of environmental risk exceeding limits, recovery time, and cumulative resilience loss, respectively, wherein the degree of environmental risk exceeding limits is used to characterize the extent to which environmental risk exceeds the environmental risk threshold boundary; identifying candidate control actions that reduce at least two of the following: the degree of environmental risk exceeding limits, the recovery time, and the cumulative resilience loss; and determining the minimum cut set for propagation blocking and the minimum set for recovery control based on the target control actions.

[0012] In one possible implementation, the dual-threshold states include a security maintenance state, a propagation interception state, a resilience hardening state, and a cascading instability state. Cooperative control instructions are generated based on at least one of the propagation blocking minimum cut set and the recovery control minimum set, including: generating a maintenance control instruction when the dual-threshold state is a security maintenance state; generating a prevention control instruction based on the propagation blocking minimum cut set when the dual-threshold state is a propagation interception state; generating a recovery control instruction based on the recovery control minimum set when the dual-threshold state is a resilience hardening state; and generating a joint control instruction based on the propagation blocking minimum cut set and the recovery control minimum set when the dual-threshold state is a cascading instability state.

[0013] In one possible implementation, the physical residuals include quality flux residuals, propagation arrival time residuals, and recovery time residuals. Based on the physical residuals after control implementation, the dynamic risk propagation network parameters or threshold determination parameters are updated in a single layer, including: if the quality flux residual exceeds a preset flux threshold, only the pollution input allocation parameters or control measure effectiveness parameters are updated; if the propagation arrival time residual exceeds a preset time threshold, only the propagation delay parameters are updated; and if the recovery time residual exceeds a preset recovery threshold, only the threshold determination parameters are updated.

[0014] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows: By constructing event state sets and dynamic risk propagation networks, continuous identification of risk propagation processes was achieved; by jointly determining environmental risk threshold boundaries and watershed climate resilience threshold boundaries, collaborative differentiation between propagation expansion and insufficient recovery was achieved; and by using counterfactual simulation, minimum set screening, and single-layer parameter update techniques, targeted generation and feedback correction of control actions were achieved. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of the method of the present invention; Figure 2 This is a statistical chart of rainfall and flow rate in a specific embodiment of the present invention; Figure 3 This is a comparison diagram of water quality response at key cross sections in a specific embodiment of the present invention; Figure 4 This is a schematic diagram of the dynamic risk propagation network strength in a specific embodiment of the present invention; Figure 5 This is a statistical chart showing the collaborative control effect in a specific embodiment of the present invention. Detailed Implementation

[0016] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0017] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0018] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0019] In this invention, "big data processing" does not simply refer to a large volume of data, but rather to the unified access, cleaning, verification, spatiotemporal registration, correlation analysis, and dynamic updating of multi-source, heterogeneous, and continuously changing data. By structurally integrating data from different sources, frequencies, and spatial locations, dispersed monitoring information can be transformed into state information that can be directly used for analysis and decision-making, thus providing a consistent data foundation for subsequent propagation identification, threshold determination, and collaborative management. Based on this, this invention further constructs an event state set around the multi-source monitoring data of the target watershed, and on this basis, forms a dynamic risk propagation network, dual threshold determination results, and collaborative management instructions, realizing continuous analysis and dynamic control of watershed processes.

[0020] like Figure 1 As shown, a watershed dual-threshold collaborative management method based on big data processing is proposed, which includes: Acquire multi-source monitoring data of the target watershed, divide the multi-source monitoring data into risk response units, event segments, and spatiotemporal alignment, and generate an event state set; In this embodiment, multi-source monitoring data generated in the target watershed within a preset monitoring period is first acquired. Based on river connectivity, shoreline morphology, pollution inflow location, and scheduling control boundaries, the target watershed is divided into multiple risk response units. Risk response units characterize analysis units with similar transport conditions and shoreline buffering conditions within the same spatial area. Subsequently, the multi-source monitoring data is segmented into event segments according to rainfall events, sustained low-flow events, sudden pollution events, or centralized scheduling events. Data with inconsistent collection frequencies are uniformly converted to the same time step. Data with different spatial locations are uniformly mapped to the corresponding risk response units. After time and spatial alignment, state information that can directly participate in subsequent risk propagation analysis is extracted to generate an event state set. The event state set serves as the basic input for subsequent dynamic risk propagation network construction and threshold determination.

[0021] Multi-source monitoring data includes meteorological monitoring data, hydrological monitoring data, water quality monitoring data, pollution discharge record data, dispatching and operation data, and shoreline remote sensing interpretation data; the event state set includes pollution input load, transport capacity, shoreline storage status, ecological buffer margin, engineering regulation margin, and receptor exposure sensitivity.

[0022] In one embodiment, the multi-source monitoring data is further limited to six categories of basic data in this step. The new limitation is to clarify the data source and the composition of the status information, so that the formation process of the event status set has directly implementable input boundaries. This limitation is used to avoid missing status items when the collection standards of different regions and departments are inconsistent, and to ensure that a unified data structure is used for subsequent propagation analysis.

[0023] Meteorological monitoring data is used to characterize external meteorological disturbances during an event, and may include rainfall, rainfall duration, temperature, wind speed, and evaporation. Hydrological monitoring data is used to characterize water transport conditions, and may include flow rate, water level, flow velocity, and cross-sectional water carrying capacity. Water quality monitoring data is used to characterize pollution changes, and may include chemical oxygen demand (COD), ammonia nitrogen, total phosphorus, dissolved oxygen, and turbidity. Pollution discharge record data is used to characterize point source or centralized discharge processes, and may include discharge location, discharge time period, discharge flow rate, and concentration of major pollutants. Dispatch and operation data is used to characterize human control actions, and may include gate opening and closing status, pump station operation status, storage facility activation status, and dispatch start and end times. Shoreline remote sensing interpretation data is used to characterize the spatial state of the shoreline zone, and may include shoreline wet range, vegetation cover changes, exposed shoreline length, and shoreline erosion traces. In practical applications, the six types of basic data can be resampled according to a uniform time step, and missing values ​​can be supplemented using valid values ​​from nearby time periods or interpolated from adjacent monitoring points. Then, each type of data can be mapped to the corresponding risk response unit according to the monitoring location.

[0024] After mapping, an event state set is formed around each event segment. The pollution input load in the event state set, determined based on pollution discharge records and water quality monitoring data, reflects the amount of pollution entering the target unit during the event. Transport capacity, determined based on hydrological monitoring data and scheduling operation data, reflects the transmission conditions of pollution within the current unit and between adjacent units. Shoreline storage status, determined based on shoreline remote sensing interpretation data and water level changes, reflects the temporary storage and occupation of incoming water and pollutants by the shoreline. Ecological buffer margin, determined based on shoreline vegetation cover changes, water quality changes, and historical baseline status, reflects the current reduction capacity that the shoreline ecosystem can still bear. Engineering regulation margin, determined based on scheduling operation data and facility design capacity, reflects the remaining available regulation space of regulation facilities in the current period. Receptor exposure sensitivity, determined based on the correspondence between receptor locations such as drinking water intake points, aquaculture areas, and ecologically sensitive areas and pollution input loads, reflects the differences in the tolerance of different risk response units to pollution disturbances. After completing the above processing, the six types of status information of each risk response unit in each event segment can be written into the event status set according to a unified field, and the event status set can be output to subsequent steps for direct use in the construction of the dynamic risk propagation network.

[0025] A dynamic risk propagation network is constructed based on the event state set, and the propagation characteristics, recovery characteristics, and effectiveness of control measures are determined based on the dynamic risk propagation network; In this embodiment, after obtaining the event state set, the spatial analysis is based on risk response units, and the temporal analysis is based on event segments. The state change processes within the event state set are correlated and organized to construct a dynamic risk propagation network that reflects pollution migration, shoreline retention, and scheduling impacts. After establishing the dynamic risk propagation network, the propagation connections, directions, and strengths between risk response units are further identified. Combined with the state changes during the propagation and recovery phases, propagation and recovery characteristics are determined. Simultaneously, based on the consistency between the dynamic risk propagation network and actual monitoring results, the reduction effect of different control measures on the propagation process is determined, yielding the effectiveness of the control measures. The propagation characteristics, recovery characteristics, and control measure effectiveness serve as direct inputs for subsequent environmental risk threshold boundary determination, watershed climate resilience threshold boundary determination, and the generation of coordinated control actions.

[0026] Constructing a dynamic risk propagation network based on an event state set includes: extracting event nodes from the event state set according to the time of state change; sequentially associating event nodes according to a preset propagation time window to form a propagation sub-chain; and constructing a dynamic risk propagation network with risk response units as nodes and the propagation relationships between risk response units as edges.

[0027] In one embodiment, this stage is further defined as three consecutive steps: event node extraction, propagation sub-chain formation, and network construction. The purpose of this setup is to first convert discrete monitoring data into a continuously analyzable propagation process, and then map the propagation process onto the propagation relationships between risk response units, avoiding the problems of loose correlation and unclear direction that occur when directly using the original time series for analysis.

[0028] In practice, the pollution input load, transport capacity, shoreline storage status, ecological buffer capacity, engineering storage capacity, and receptor exposure sensitivity are first read from the event state set within each event segment, and each state quantity is sequentially scanned along the time axis. When any state quantity exceeds a preset change threshold, it is determined to be a state change moment. The state change threshold can be preset according to the accuracy of the monitoring equipment and the historical fluctuation range. For example, a significant increase in pollution input load within adjacent time steps, a significant change in transport capacity under scheduling, or a high occupancy level of shoreline storage in a short period of time can all trigger event node extraction. After an event node is formed, the corresponding risk response unit, occurrence time, change type, and change magnitude are recorded. Subsequently, the event nodes are sequentially associated according to a preset propagation time window. The preset propagation time window is used to limit the reasonable transmission time of upstream state changes to downstream state changes. If the preceding and following event nodes satisfy a temporal relationship, and their corresponding risk response units are directly related in terms of river connectivity, shoreline exchange, or scheduling control, then the two are sequentially associated as part of the same propagation process. For multiple event nodes that consecutively meet the association conditions, they are connected in chronological order to form a propagation sub-chain. The propagation sub-chain is used to characterize the process of pollution or hydrodynamic impacts being transmitted from the initial risk response unit to downstream or adjacent risk response units in an event segment. After the propagation sub-chain is formed, a dynamic risk propagation network is established using all risk response units as network nodes and the propagation connections corresponding to adjacent event nodes in the propagation sub-chain as network edges.

[0029] The direction of network edges is determined according to the chronological order of event nodes. The starting point of a network edge corresponds to the upstream or preceding risk response unit of the propagation impact, and the ending point corresponds to the downstream or subsequent risk response unit of the propagation impact. If the same pair of risk response units forms propagation links multiple times in different event segments, the propagation links are retained, and the frequency and duration of occurrence are recorded. Through this processing, the dynamic risk propagation network can not only reflect the spatial propagation path but also the dynamic changes in propagation relationships within event segments, providing a unified structural basis for subsequent analysis of propagation characteristics, recovery characteristics, and the effectiveness of control measures.

[0030] The dynamic risk propagation network is used to determine propagation characteristics, recovery characteristics, and the effectiveness of control measures. This includes: conducting time-series correlation analysis based on propagation subchains to determine the propagation intensity of each propagation relationship; determining the effectiveness of control measures corresponding to each propagation relationship based on quality and throughput consistency checks and spatial distribution consistency checks; determining propagation characteristics based on propagation intensity; and determining recovery characteristics based on changes in the event state set during the propagation and recovery phases.

[0031] In one embodiment, this stage is further defined as three processing steps: calculating the propagation intensity, identifying the effectiveness of control measures, and extracting propagation and recovery characteristics. The purpose of introducing this limitation is to further transform the dynamic risk propagation network from structural relationships into measurable analytical results, so that propagation relationships can not only be identified, but also sorted, compared, and used for subsequent threshold determination.

[0032] In practice, temporal correlation analysis is first performed based on the propagation subchain. This analysis uses the time difference between adjacent event nodes, the magnitude of state changes, and the connectivity between risk response units within the propagation subchain as its foundation, calculating the propagation intensity for each propagation relationship. Propagation intensity can be understood as the actual driving force of a propagation relationship on downstream state changes within a unit event segment. If the pollution input load increment corresponding to the preceding event node is large, and the subsequent event node shows a significant response within the preset propagation time window, the propagation intensity of the corresponding propagation relationship increases; conversely, if the preceding event node changes significantly but the subsequent event node responds weakly, the propagation intensity of the corresponding propagation relationship decreases. After calculating the propagation intensity, quality flux consistency checks and spatial distribution consistency checks are performed on each propagation relationship.

[0033] The quality flux consistency check compares whether the pollution output from the risk response unit at the starting point of the propagation relationship and the pollution change received by the risk response unit at the ending point of the propagation relationship maintain a reasonable correspondence within the same event segment. If the difference between the two is within a preset tolerance range, the propagation relationship is considered valid in terms of quality transfer. The spatial distribution consistency check compares whether the affected area predicted by the dynamic risk propagation network is consistent with the changes in monitoring sections, shoreline remote sensing interpretation results, and the actual affected area in the event segment. If the predicted affected area and the actual affected area can maintain a match in spatial location and direction of expansion, the propagation relationship is considered valid in terms of spatial distribution. For propagation relationships that pass both types of checks, the changes in propagation intensity before and after the implementation of control measures are analyzed by combining the gate opening and closing, pump station operation, storage facility activation, and shoreline artificial intervention recorded in the scheduling operation data, thereby determining the effectiveness of the corresponding control measures. The more significant the reduction in propagation intensity caused by the control measures, the higher the effectiveness of the control measures.

[0034] Subsequently, propagation features are extracted based on the propagation intensity distribution of all propagation relationships. These features primarily reflect the extent and activity of the propagation process within the current event segment. Recovery features, however, are not directly calculated from propagation relationships but are extracted by combining changes in the event state set during the propagation and recovery phases. Specifically, the shoreline storage status, water quality changes, and hydrological changes at each time step during the end of the propagation phase and the recovery phase are compared to identify the rate of state decline and the duration of recovery, thereby obtaining the recovery features. Through this processing, the dynamic risk propagation network outputs not only the propagation path structure but also the propagation intensity, effectiveness of control measures, propagation features, and recovery features that can directly lead to the next stage of the judgment process.

[0035] The propagation characteristics include the length of the propagation front, the number of continuously affected risk response units, and the number of active propagation relationships; the recovery characteristics include the shoreline saturation level and the recovery rate of the monitoring section.

[0036] In one embodiment, this stage further defines the specific composition of the propagation and recovery characteristics to clarify the actual meaning and value method of each characteristic in subsequent threshold determination. With this definition, the propagation and recovery characteristics are no longer merely conceptual but are constrained into state indicators that can be directly calculated and compared. Propagation characteristics include the propagation front length, the number of continuously affected risk response units, and the number of active propagation relationships. The propagation front length characterizes the farthest distance the propagation impact extends along the main propagation direction of the watershed in a single event segment.

[0037] Specifically, the initial affected risk response unit can be used as the starting point, and the farthest risk response unit with a continuous effective propagation relationship can be used as the ending point. The propagation front length is calculated based on the river path length or the cumulative connection distance between risk response units. The number of continuously affected risk response units is used to characterize the continuous spatial expansion range of the propagation process. Specifically, starting from the initial risk response unit, the search proceeds downstream or to adjacent risk response units along the active propagation relationship, counting the number of risk response units that remain continuously affected within the same event segment. If there is a spatial or temporal breakpoint between two risk response units, the counting restarts.

[0038] The number of active propagation relationships is used to characterize the scale of propagation relationships actually involved in the current event segment. Specifically, propagation relationships whose propagation intensity reaches a preset effective threshold are identified as active propagation relationships, and the number of active propagation relationships is counted. Recovery characteristics include shoreline storage saturation and monitoring section recovery rate. Shoreline storage saturation reflects the level of space occupancy of the shoreline zone within the current event segment that can be used to temporarily store incoming water and pollutants. Specifically, it can be determined by comparing the shoreline storage state with the preset shoreline storage capacity; when the current occupancy is close to the preset shoreline storage capacity, the shoreline storage saturation increases. Monitoring section recovery rate reflects how quickly water quality or hydrodynamic indicators return to their pre-event stable state after entering the recovery phase. Specifically, during the recovery phase, water quality and hydrological monitoring data corresponding to the monitoring section are read sequentially over time, the reduction rate of deviation between adjacent time steps is calculated, and then the reduction rates of multiple time steps are averaged to obtain the monitoring section recovery rate. A higher recovery rate at the monitoring section indicates a faster decline in the recovery phase; a lower recovery rate indicates a longer recovery phase. By uniformly extracting the propagation front length, the number of continuously affected risk response units, the number of active propagation relationships, the shoreline saturation level, and the monitoring section recovery rate, the propagation and recovery process of the current event segment can be fully characterized from five aspects: expansion range, expansion continuity, propagation activity, shoreline carrying capacity, and recovery speed. This provides a clear, stable, and reusable analytical basis for subsequent dual-threshold determination.

[0039] Based on the characteristics of propagation and recovery, the environmental risk threshold boundary, the watershed climate resilience threshold boundary, the dual threshold state, and the cumulative amount of resilience loss are determined. In this embodiment, after the propagation and recovery features have been extracted, the process proceeds to the dual-threshold determination stage. This stage first identifies the spatial extent and continuity of the current event segment based on the propagation features, and then identifies the receding rate and buffer recovery status of the target watershed in the later stages of the event based on the recovery features. This allows for the determination of the environmental risk threshold boundary and the watershed climate resilience threshold boundary, respectively. After obtaining these two threshold boundaries, the propagation and recovery states of the current event segment are compared and judged to form a dual-threshold state. Simultaneously, at the end of each event segment, the remaining shoreline storage capacity, ecological buffer capacity, and engineering regulation capacity are summarized to form a cumulative resilience loss. The environmental risk threshold boundary, the watershed climate resilience threshold boundary, the dual-threshold state, and the cumulative resilience loss collectively serve as the basis for selecting subsequent coordinated control actions.

[0040] Based on the propagation and recovery characteristics, the environmental risk threshold boundary, the watershed climate resilience threshold boundary, and the dual-threshold state are determined, including: determining the environmental risk threshold boundary by conducting connectivity analysis based on changes in the propagation front length, the number of continuously affected risk response units, and the number of active propagation relationships; determining the watershed climate resilience threshold boundary by determining whether the recovery time of the monitoring section and the shoreline segment corresponding to the risk response unit exceeds the preset recovery window; and determining the dual-threshold state based on the environmental risk threshold boundary and the watershed climate resilience threshold boundary.

[0041] In one embodiment, this stage further defines the environmental risk threshold boundary, the watershed climate resilience threshold boundary, and the dual-threshold state. The new limitation is that the dual-threshold determination is based on two directly verifiable judgment processes: propagation connectivity and recovery time. This makes the threshold boundary no longer dependent on a single score, but on the actual propagation range and recovery process within the event segment.

[0042] In practice, the process begins by sequentially reading the propagation front length, the number of continuously affected risk response units, and the number of active propagation relationships, using event segments as units, and generating corresponding change sequences. Connectivity analysis is then performed on these sequences. The core of connectivity analysis is determining whether the propagation impact remains confined to a local area or has formed a continuous expansion. If the propagation front length continuously increases within adjacent time steps, the number of continuously affected risk response units increases synchronously, and the number of active propagation relationships reaches a preset active threshold, then the current propagation process is determined to have formed a continuous expansion trend, and the starting time satisfying the continuous expansion condition is defined as the environmental risk threshold boundary. If the propagation front length begins to stabilize, the number of continuously affected risk response units no longer increases, and the number of active propagation relationships falls below the preset active threshold, then the corresponding time is taken as the fallback position of the environmental risk threshold boundary. After identifying the environmental risk threshold boundary, the recovery time of the monitoring section and the corresponding shoreline segment of the risk response unit is then assessed.

[0043] The recovery time is calculated from the start of the recovery phase of the event segment until the main water quality indicators at the monitoring section return to the preset stable range, and the main water storage state of the shoreline section falls back to the preset recovery range. If either the recovery time of the monitoring section or the recovery time of the shoreline section exceeds the preset recovery window, the watershed climate resilience threshold boundary is determined to have been triggered; if both are within the preset recovery window, the watershed climate resilience threshold boundary is determined not to have been triggered. After identifying the two types of boundaries, the propagation state and recovery state of the current event segment are combined for judgment. When neither the environmental risk threshold boundary nor the watershed climate resilience threshold boundary is triggered, the current event segment is in a safe maintenance state; when the environmental risk threshold boundary is triggered but the watershed climate resilience threshold boundary is not triggered, the current event segment is in a propagation interception state; when the environmental risk threshold boundary is not triggered but the watershed climate resilience threshold boundary is triggered, the current event segment is in a resilience reinforcement state; when both the environmental risk threshold boundary and the watershed climate resilience threshold boundary are triggered simultaneously, the current event segment is in a cascading instability state. This process allows dual-threshold states to directly correspond to different types of subsequent control actions, making it easier to distinguish between propagation and insufficient recovery.

[0044] Based on the propagation and recovery characteristics, the cumulative amount of resilience loss is determined, including: after the end of each event segment, the amount of unrecovered buffer is determined based on the shoreline storage status, ecological buffer capacity, and engineering regulation capacity, and the unrecovered buffer is accumulated to obtain the cumulative amount of resilience loss.

[0045] In one embodiment, this stage further limits the amount of accumulated resilience loss. The new limitation is that the amount of accumulated resilience loss is defined as the total amount of buffer capacity that has not been recovered after the end of each event segment, so that the recovery gap of the watershed under the action of continuous events can be continuously recorded, rather than making a one-time judgment based solely on the current state.

[0046] In practice, after each event segment ends, the shoreline storage status, ecological buffer capacity, and engineering storage capacity corresponding to the current risk response unit are first read and compared with the pre-set recovery benchmark for that risk response unit under non-event conditions. The recovery benchmark can be obtained during a stable monitoring period without significant rainfall, centralized dispatch, or sudden pollution, and is used to characterize the available buffer level of the target watershed under normal conditions. If the current shoreline storage status is still higher than the occupancy level corresponding to the recovery benchmark, the excess is determined as the shoreline unrestored buffer volume; if the current ecological buffer capacity is lower than the available level corresponding to the recovery benchmark, the difference is determined as the ecological unrestored buffer volume; if the current engineering storage capacity is lower than the available storage space corresponding to the recovery benchmark, the difference is determined as the engineering unrestored buffer volume.

[0047] After determining the three types of unrecovered buffer quantities, they are first aggregated within a single risk response unit to obtain the unit's unrecovered buffer quantity; then, they are aggregated across all risk response units to obtain the unrecovered buffer quantity for the current event segment. Subsequently, the unrecovered buffer quantity for the current event segment is written into the resilience loss accumulation record to form the resilience loss accumulation. If all risk response units recover to the recovery baseline range by the end of the current event segment, the unrecovered buffer quantity corresponding to the current event segment is recorded as zero, and the resilience loss accumulation does not increase. If the interval between multiple consecutive event segments is short, and the unrecovered buffer quantity formed at the end of the previous event segment has not yet been eliminated, the unrecovered buffer quantity formed at the end of the subsequent event segment continues to be added to the resilience loss accumulation record. In this way, the resilience loss accumulation can continuously reflect the buffer capacity occupancy of the target watershed after multiple disturbances. Subsequently, in the collaborative control phase, the resilience loss accumulation can be directly used as the basis for determining scheduling priority and recovery priority. When the resilience loss accumulation is high, it indicates that there is still a significant recovery gap in the target watershed after the preceding event. In this case, even if the current propagation range has not significantly expanded, recovery-type control actions should be prioritized to prevent continuous instability after being disturbed again in a short period of time. Through this process, the dual threshold determination results and the recovery burden under continuous events are unified into the same judgment system, which enables subsequent control actions to be more in line with the actual state of the target watershed.

[0048] Based on the dual threshold state, the effectiveness of control measures, and the cumulative amount of resilience loss, counterfactual simulations are performed on candidate control actions to determine the minimum cut set for propagation blocking and the minimum set for recovery control. Based on at least one of the minimum cut set for propagation blocking and the minimum set for recovery control, collaborative control instructions are generated, and the parameters of the dynamic risk propagation network or the threshold judgment parameters are updated in a single layer according to the physical residual after the control is executed.

[0049] In this embodiment, after obtaining the dual-threshold state, the effectiveness of control measures, and the cumulative amount of resilience loss, the collaborative control phase begins. This phase first constructs candidate control actions around the current event segment. Without changing the current monitoring input, counterfactual simulations are performed on each candidate control action to identify action combinations that can simultaneously suppress propagation and shorten the recovery process. Subsequently, the minimum cut set for propagation blocking and the minimum set for recovery control are determined based on the counterfactual simulation results, and corresponding collaborative control instructions are generated in conjunction with the current dual-threshold state. After the collaborative control instructions are executed, the difference between the actual monitoring results and the predicted results is compared to form physical residuals. Only the parameters of the corresponding category are updated at a single level, ensuring that the propagation analysis and threshold determination of subsequent event segments remain consistent with the current state of the target watershed.

[0050] Counterfactual simulations are performed on candidate control actions to determine the minimum cut set for propagation blocking and the minimum set for recovery control. This includes: simulating the impact of single and combined candidate control actions on the degree of environmental risk exceeding limits, recovery time, and cumulative resilience loss, where the degree of environmental risk exceeding limits is used to characterize the extent of environmental risk exceeding the environmental risk threshold boundary; identifying candidate control actions that reduce at least two of the following: the degree of environmental risk exceeding limits, the recovery time, and the cumulative resilience loss; and determining the minimum cut set for propagation blocking and the minimum set for recovery control based on the target control actions.

[0051] In one embodiment, this stage further defines the counterfactual simulation, target action screening, and minimum set determination process. The purpose of this design is to ensure that candidate control actions undergo comparable evaluation within the same event context before formal execution, avoiding insufficient propagation suppression or improper configuration of recovery actions due to action selection based solely on experience.

[0052] In practice, candidate control actions are first listed under the dual-threshold state corresponding to the current event segment, based on existing scheduling capabilities, shoreline conditions, and emission control conditions. Candidate control actions may include adjusting emission periods, adjusting gate opening, adjusting pump station start / stop, pre-emptying storage facilities, activating shoreline interception zones, releasing shoreline storage space, ecological water replenishment, and switching water intake to key recipients. Candidate control actions can be single actions or combinations of two or more single actions. Subsequently, starting from the propagation characteristics, recovery characteristics, dual-threshold state, effectiveness of control measures, and cumulative resilience loss at the end of the current event segment, counterfactual simulations are performed on each single candidate control action and each combination of candidate control actions. During counterfactual simulations, the monitoring data basis of the current event segment is not changed; only the control conditions corresponding to the candidate control actions are replaced in the dynamic risk propagation network, and the activity of the propagation path, the scope of risk expansion, and changes in the recovery process are recalculated.

[0053] After the simulation, the changes in the degree of environmental risk exceeding the limit, recovery time, and cumulative resilience loss were obtained. The degree of environmental risk exceeding the limit is used to characterize the extent to which the current propagation state exceeds the environmental risk threshold boundary. If the propagation front shortens, the number of continuously affected risk response units decreases, or the number of active propagation relationships decreases after the implementation of control actions, the degree of environmental risk exceeding the limit decreases. If the time required for the monitoring section and shoreline segment to return to a stable range is shortened after the implementation of control actions, the recovery time is shortened. If the amount of unrecovered buffer is reduced at the end of the event after the implementation of control actions, the cumulative resilience loss decreases. Subsequently, each candidate control action is screened. If at least two of the following conditions are met—a decrease in the degree of environmental risk exceeding the limit, a shortening of the recovery time, and a decrease in the cumulative resilience loss—the candidate control action is determined as the target control action. Target control actions are further classified according to their target object and scope of action. Target control actions that directly affect the propagation chain and can cut off or weaken key propagation relationships are included in the screening scope of the propagation blocking minimum cut set; target control actions that directly affect the recovery process and can shorten the duration of the recovery phase or reduce the amount of unrecovered buffer are included in the screening scope of the recovery control minimum set. When determining the minimum cut set for propagation blocking, priority is given to retaining target control actions that can cut off the most active propagation relationships with the fewest actions; when determining the minimum set for recovery control, priority is given to retaining target control actions that can significantly shorten recovery time and significantly reduce the amount of unrecovered buffer with the fewest actions. Through this process, a large number of candidate control actions can be compressed into a small set of key actions that can be directly used for scheduling execution, providing a clear basis for the generation of subsequent collaborative control instructions.

[0054] The dual-threshold states include a safety maintenance state, a propagation interception state, a resilience hardening state, and a cascading instability state. Coordinated control instructions are generated based on at least one of the propagation blocking minimum cut set and the recovery control minimum set, including: generating a maintenance control instruction when the dual-threshold state is a safety maintenance state; generating a prevention control instruction based on the propagation blocking minimum cut set when the dual-threshold state is a propagation interception state; generating a recovery control instruction based on the recovery control minimum set when the dual-threshold state is a resilience hardening state; and generating a joint control instruction based on the propagation blocking minimum cut set and the recovery control minimum set when the dual-threshold state is a cascading instability state.

[0055] In one embodiment, this stage further defines the correspondence between the dual threshold states and the collaborative control instructions. This definition is introduced to ensure that the action outputs under different risk states have fixed branch paths, preventing inconsistent action selections for the same event segment under different judges or different scheduling periods.

[0056] In practice, the dual-threshold states corresponding to the current event segment are first read, and then divided into four states: safety maintenance, propagation interception, resilience reinforcement, and cascading instability. The safety maintenance state indicates that the environmental risk has not yet significantly expanded, and the recovery process has not exceeded the preset recovery window; at this time, the target watershed as a whole is within a controllable range. The propagation interception state indicates that the environmental risk has shown an expanding trend, but the recovery capacity is not yet significantly insufficient; at this time, the priority is to cut off the propagation relationship that continues to expand. The resilience reinforcement state indicates that the environmental risk expansion has not yet reached a high level, but the recovery process has clearly lagged behind; at this time, the priority is to replenish the buffer capacity and shorten the recovery time. The cascading instability state indicates that propagation expansion and insufficient recovery occur simultaneously; at this time, it is necessary to control both the propagation process and the recovery process simultaneously. After determining the dual-threshold states, collaborative control instructions are generated according to a fixed mapping relationship. In the safety maintenance state, maintenance control instructions are generated.

[0057] Maintaining control instructions primarily aim to prevent significant changes to the current dispatch boundaries and continue basic monitoring, routine patrols, and lightweight early warnings. This approach ensures the target watershed remains under stable control without adding additional dispatch burden. In the propagation interception state, preventative control instructions are generated based on the propagation blocking minimum cut set. These instructions focus on the front end and key nodes of the propagation chain and may include restricting discharge periods at key outlets, adjusting gate flow rhythms, activating interception facilities ahead of schedule, reducing the input intensity of some tributaries, and switching local storage paths. In the resilience reinforcement state, recovery control instructions are generated based on the recovery control minimum set. These instructions focus on replenishing buffer capacity during the recovery phase and may include increasing ecological water replenishment, releasing shoreline storage space, extending the retention time of storage facilities, reducing secondary disturbances during the recovery phase, and implementing short-term protective dispatching for sensitive areas. In the cascading instability state, joint control instructions are generated based on both the propagation blocking minimum cut set and the recovery control minimum set. These joint control instructions include both preventative and recovery actions, executed in the order of first blocking high-risk propagation relationships and then replenishing key recovery capabilities. In actual output, collaborative control instructions can be written as a set of scheduling instructions that includes the action object, execution order, start and end time, execution scope, and review conditions, and sent directly to the scheduling platform or management terminal so that relevant personnel can implement them according to unified rules. Through this one-to-one correspondence between state and action, the collaborative control phase can quickly switch to the appropriate control path under different risk conditions.

[0058] Physical residuals include quality flux residuals, propagation arrival time residuals, and recovery time residuals. Based on the physical residuals after control implementation, the parameters of the dynamic risk propagation network or threshold judgment parameters are updated in a single layer, including: when the quality flux residual exceeds the preset flux threshold, only the pollution input allocation parameters or control measure effectiveness parameters are updated; when the propagation arrival time residual exceeds the preset time threshold, only the propagation delay parameters are updated; when the recovery time residual exceeds the preset recovery threshold, only the threshold judgment parameters are updated.

[0059] In one embodiment, this stage further defines the physical residual identification and single-layer update process. This limitation is introduced to ensure that the feedback results after the collaborative control action can directly affect the subsequent model update, but the update scope is restricted to the single parameter level corresponding to the error source, thereby avoiding the simultaneous drift of multiple types of parameters caused by a single error.

[0060] In practice, after the coordinated control instructions are issued and executed, multi-source monitoring data is continuously collected. The actual monitoring results after execution are compared with the predicted results obtained from counterfactual simulations before execution to form physical residuals. Physical residuals include mass flux residuals, propagation arrival time residuals, and recovery time residuals. Mass flux residuals reflect the difference between the predicted pollution transfer rate and the actual monitored pollution transfer rate. Propagation arrival time residuals reflect the difference between the predicted time for the propagation impact to reach a monitoring section or risk response unit and the actual arrival time. Recovery time residuals reflect the difference between the predicted duration of the recovery phase and the actual duration of the recovery phase.

[0061] After calculating the physical residuals, a single-layer update is performed according to the residual category. When the mass flux residual exceeds the preset flux threshold, only the pollution input allocation parameter or the control measure effectiveness parameter is updated. If the actual pollution transmission rate after execution is higher than the predicted value, it indicates that the pollution input allocation is too low or the reduction effect of some control measures is overestimated. In this case, the corresponding parameters are corrected according to the actual monitoring results. If the actual pollution transmission rate after execution is lower than the predicted value, it indicates that the pollution input allocation is too high or the reduction effect of some control measures is underestimated. Similarly, the parameters are corrected according to the actual results. When the propagation arrival time residual exceeds the preset time threshold, only the propagation lag parameter is updated. If the actual arrival time is earlier than the predicted value, the propagation lag of the corresponding propagation relationship is shortened; if the actual arrival time is later than the predicted value, the propagation lag of the corresponding propagation relationship is extended. When the recovery time residual exceeds the preset recovery threshold, only the threshold determination parameter is updated.

[0062] If the actual recovery process takes longer than the predicted value, the recovery window judgment boundary should be tightened appropriately; if the actual recovery process is shorter than the predicted value, the recovery window judgment boundary should be widened appropriately. Throughout the update process, only one parameter level is allowed to be adjusted at a time, without simultaneously adjusting the propagation relationship, control measure effectiveness, and threshold judgment parameters. After the update is completed, the corrected parameters are rewritten into the dynamic risk propagation network or threshold judgment module and used for the analysis of the next event segment. This allows subsequent propagation analysis and control decisions to gradually align with the actual operation of the target watershed, while maintaining clear parameter sources and a well-defined update path, facilitating long-term continuous use.

[0063] In one specific embodiment, a typical plain river network basin was used as the target, with a main stream length of approximately 40 kilometers. Six risk response units were established, namely, the upstream section, urban section, industrial section, confluence section, downstream wetland section, and water intake section. The monitoring period was continuous for 72 hours, with a time step of 1 hour. Three rainfall monitoring points, two hydrological monitoring sections, and three water quality monitoring sections were set up, and pollution discharge records and dispatch operation records were connected. Shoreline remote sensing interpretation data were acquired daily, converted to hourly values, and mapped to the six risk response units.

[0064] The following is a set of key raw data. The rainfall event saw two main rainfall peaks: a moderate to heavy rainfall occurred around the 13th hour, and a heavy rainfall occurred at the 36th hour, with a peak rainfall intensity of 18.37 mm / h. The downstream hydrological section's flow rate peaked at 354.49 cubic meters per second between the 26th and 30th hours, and then declined. Figure 2 This is a statistical graph of rainfall and flow rate. Rainfall is derived from the hourly cumulative rainfall data from rainfall monitoring points, and flow rate is derived from the hourly average value from downstream hydrological sections. Figure 2 The peak rainfall occurred in the 36th hour, and the peak flow occurred in the 27th hour, reflecting the lag response relationship under the combined effect of confluence and scheduling.

[0065] Regarding pollution input, one industrial discharge outlet is located in the industrial section, with a normal discharge flow of 2000 cubic meters per hour and a chemical oxygen demand (COD) concentration of 520 mg / L. During hours 10-16, an overflow occurred due to combined sewer overflow, causing the COD concentration to rise to 900 mg / L. One domestic discharge outlet is located in the urban section, with a normal discharge flow of 2700 cubic meters per hour and a COD concentration of 300 mg / L. During hours 32-40, a short-term load increase occurred, causing the COD concentration to rise by 150 mg / L. The non-point source pollution load is converted into a pollution load sequence based on rainfall intensity, using a base value of 400 kg / h, and superimposed with the rainfall term. The conversion relationship is that for every 1 mm / h increase in rainfall, the non-point source pollution load increases by 800 kg / h. After risk response unit segmentation, event segmentation, and spatiotemporal alignment, the above data forms an event state set, with fields including pollution input load, transport capacity, shoreline storage status, ecological buffer margin, engineering storage margin, and receptor exposure sensitivity. The event segments are divided into rainfall process, emission surge process, and scheduling process. This embodiment obtains two main event segments, covering hours 8 to 28 and hours 30 to 56, respectively.

[0066] In the phase of constructing a dynamic risk propagation network, the event state set is first scanned within each event segment, and the moments of state change are extracted as event nodes. Event node determination uses threshold rules; for example, an increase of more than 20% in pollution input load over consecutive hours, a change of more than 15% in transport capacity over consecutive hours, and an increase of more than 10% in shoreline storage over consecutive hours all trigger event node recording. Then, a propagation time window of 2 to 8 hours is set, and event nodes that satisfy the window constraint in time and have spatial connectivity are sequentially associated to form propagation sub-chains. A dynamic risk propagation network is then constructed using risk response units as nodes and propagation relationships as edges. Figure 4 This diagram illustrates the dynamic risk propagation network strength, using hour 42 as an example. The chemical oxygen demand (COD) concentration of each unit at that time is given below the nodes, and the arrow width is mapped from the propagation relationship strength. At hour 42, the COD concentrations were: upstream 25.4 mg / L, urban area 25.1 mg / L, industrial area 28.0 mg / L, inlet area 30.8 mg / L, downstream wetland area 33.5 mg / L, and water intake area 34.2 mg / L. It can be observed that the propagation relationship from the industrial area to the downstream wetland and water intake areas is more active, indicating a significant increase in network connectivity.

[0067] To facilitate the reproduction of key calculations by those skilled in the art, this embodiment employs mass flux consistency verification in the water quality response calculation. When converting the chemical oxygen demand (COD) concentration and cross-sectional flow rate at hydrological sections into pollution load, the following expression is used: The cross-sectional chemical oxygen demand concentration is expressed in milligrams per liter. Background chemical oxygen demand (COD) concentration, in milligrams per liter; LLL represents the pollution load entering the section, in kilograms per hour. The flow rate is expressed in cubic meters per second. The background chemical oxygen demand (COD) concentration is taken as 18 mg / L. Taking the 42nd hour as an example, without coordinated control measures, the flow rate at the intake section is 173.87 cubic meters per second, and the COD concentration at the intake section is 39.50 mg / L. The calculated pollution load entering the intake section is approximately (39.50 mg / L). 18) × 3.6 × 173.87 = 13456 kg / hour. After implementing coordinated management, the flow rate at the water intake section is 213.87 cubic meters per second, and the chemical oxygen demand (COD) concentration at the water intake section is 30.28 mg / L. The calculated pollution load entering the water intake section is approximately (30.28 mg / L). 18) × 3.6 × 213.87 = 9455 kg / hour. This result is used to verify the consistency between the strength of the propagation relationship and the effectiveness of control measures, avoiding the propagation network relying solely on correlation while ignoring quality conservation.

[0068] The propagation and recovery characteristics were extracted based on the dynamic risk propagation network. Propagation characteristics included the length of the propagation front, the number of consecutively affected risk response units, and the number of active propagation relationships. The threshold for determining impact was a chemical oxygen demand (COD) concentration greater than 28 mg / L. At the 42nd hour, without coordinated control measures, the affected units were continuously distributed from the industrial section to the water intake section, with 4 consecutively affected risk response units, a propagation front length of 25 km, and 3 active propagation relationships. Recovery characteristics included the shoreline saturation level and the recovery rate at the monitoring section. The recovery rate was calculated by the magnitude of the decrease in water quality deviation between adjacent hours during the recovery phase.

[0069] The threshold determination phase uses propagation and recovery characteristics as inputs. The environmental risk threshold boundary is determined using connectivity rules: the environmental risk threshold boundary is triggered when the propagation front length is not less than 20 km, the number of continuously affected risk response units is not less than 3, and the number of active propagation relationships is not less than 2. The watershed climate resilience threshold boundary is determined using recovery window rules: the watershed climate resilience threshold boundary is triggered when the recovery time of the monitoring section and shoreline segment exceeds 6 hours. The dual-threshold state is obtained by combining the two types of boundaries and is divided into a safe maintenance state, a propagation interception state, a resilience strengthening state, and a cascading instability state. Without coordinated control measures, a propagation interception state appears around the 42nd hour, and a resilience strengthening state appears during the recovery phase. After coordinated control measures are implemented, the duration of the propagation interception state is significantly shortened, and the resilience strengthening state does not appear.

[0070] The cumulative resilience loss was calculated at the end of each event segment. The shoreline storage status, ecological buffer capacity, and engineering storage capacity were compared with the recovery baseline established during the stable monitoring cycle to obtain the unrecovered shoreline buffer capacity, unrecovered ecological buffer capacity, and unrecovered engineering buffer capacity, which were then accumulated and recorded after being aggregated across the entire watershed. In this embodiment, the resilience loss for two event segments was calculated at the 24th and 56th hours. The cumulative resilience loss without coordinated control measures was 137.71, while the cumulative resilience loss with coordinated control measures was 101.45, a decrease of approximately 26.33%.

[0071] In the coordinated control phase, a set of candidate control actions is first constructed, including adjustments to emission periods, gate opening, pump station start / stop adjustments, pre-emption of storage facilities, activation of shoreline interception zones, ecological water replenishment, and switching of water intake for key recipients. Then, without changing the current monitoring input, counterfactual simulations are performed on both single and combined candidate control actions. The counterfactual simulation outputs three results: the degree of environmental risk exceeding limits, recovery time, and cumulative resilience loss. The degree of environmental risk exceeding limits characterizes the extent to which environmental risk exceeds the environmental risk threshold boundary; in this embodiment, the proportion of exceeding limits in the propagation front length is used as the characterization metric. Taking the 42nd hour as an example, without coordinated control, the propagation front length is 25 kilometers, the environmental risk threshold boundary corresponds to a propagation front length of 20 kilometers, and the degree of environmental risk exceeding limits is 0.25; after coordinated control, the propagation front length drops back to 0 kilometers, and the degree of environmental risk exceeding limits is 0. The selection rule is that at least two of the three results improve simultaneously. After screening, the minimum cut set for propagation blocking determined in this embodiment is the adjustment of emission time period and the activation of shoreline interception zone, while the minimum set for recovery control is the retention of ecological water replenishment and water storage facilities. The coordinated control command outputs branches according to dual threshold states: prevention control command is output in the propagation blocking state, recovery control command is output in the resilience reinforcement state, and joint control command is output in the cascade instability state.

[0072] Figure 3 This is a comparison chart of water quality responses at key sections. Chemical oxygen demand (COD) concentrations are from the water quality monitoring section at the intake. Without coordinated control measures, the peak COD was 39.50 mg / L, and the duration exceeding 30 mg / L was 13 hours. After coordinated control measures were implemented, the peak COD was 35.16 mg / L, and the duration exceeding 30 mg / L was 5 hours. Recovery was defined as reaching 23 mg / L and maintaining it continuously for 6 hours. The recovery time without coordinated control measures was 11 hours, while the recovery time with coordinated control measures was 6 hours. Figure 5To illustrate the statistical effects of coordinated control, five indicators are summarized: peak chemical oxygen demand, number of hours exceeding limits, recovery time, cumulative toughness loss, and flux relative residual. The flux relative residual is calculated based on the mass flux residual, and was 5.48% before the single-layer update and 2.34% after the single-layer update, indicating that the parameter update can converge the model error without introducing multi-layer linkage adjustments.

[0073] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.

[0074] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A watershed dual-threshold collaborative management and control method based on big data processing, characterized in that, The method includes: Acquire multi-source monitoring data of the target watershed, and perform risk response unit division, event segment division, and spatiotemporal alignment on the multi-source monitoring data to generate an event state set; A dynamic risk propagation network is constructed based on the event state set, and propagation characteristics, recovery characteristics, and the effectiveness of control measures are determined based on the dynamic risk propagation network. Based on the propagation characteristics and the recovery characteristics, the environmental risk threshold boundary, the watershed climate resilience threshold boundary, the dual threshold state, and the cumulative amount of resilience loss are determined. Based on the dual threshold states, the effectiveness of the control measures, and the accumulated resilience loss, counterfactual simulations are performed on candidate control actions to determine the minimum cut set for propagation blocking and the minimum set for recovery control. Based on at least one of the minimum cut set for propagation blocking and the minimum set for recovery control, a collaborative control instruction is generated, and the parameters of the dynamic risk propagation network or the threshold judgment parameters are updated in a single layer according to the physical residual after the control execution.

2. The method according to claim 1, characterized in that, The multi-source monitoring data includes meteorological monitoring data, hydrological monitoring data, water quality monitoring data, pollution discharge record data, dispatching and operation data, and shoreline remote sensing interpretation data; The event state set includes pollution input load, transport capacity, shoreline storage status, ecological buffer margin, engineering storage margin, and receptor exposure sensitivity.

3. The method according to claim 1, characterized in that, The construction of a dynamic risk propagation network based on the event state set includes: Extract event nodes from the event state set based on the time of state change; The event nodes are sequentially associated according to a preset propagation time window to form a propagation sub-chain; The dynamic risk propagation network is constructed using the risk response units as nodes and the propagation relationships between the risk response units as edges.

4. The method according to claim 3, characterized in that, The determination of propagation characteristics, recovery characteristics, and the effectiveness of control measures based on the dynamic risk propagation network includes: Based on the propagation subchains, perform temporal correlation analysis to determine the propagation strength of each propagation relationship; The effectiveness of the control measures corresponding to each propagation relationship is determined based on the consistency verification of quality flux and spatial distribution. The propagation characteristics are determined based on the propagation intensity; The recovery characteristics are determined based on the changes in the event state set during the propagation and recovery phases.

5. The method according to claim 4, characterized in that, The propagation characteristics include the length of the propagation frontier, the number of continuously affected risk response units, and the number of active propagation relationships; The recovery characteristics include the shoreline saturation level and the recovery rate of the monitoring section.

6. The method according to claim 5, characterized in that, The step of determining the environmental risk threshold boundary, the watershed climate resilience threshold boundary, and the dual-threshold state based on the propagation characteristics and the recovery characteristics includes: Based on the changes in the propagation front length, the number of continuously affected risk response units, and the number of active propagation relationships, a connectivity analysis is performed to determine the environmental risk threshold boundary. The boundary of the watershed climate resilience threshold is determined based on whether the recovery time of the monitoring section and the shoreline section corresponding to the risk response unit exceeds the preset recovery window. The dual-threshold state is determined based on the environmental risk threshold boundary and the watershed climate resilience threshold boundary.

7. The method according to claim 2, characterized in that, Determining the cumulative amount of toughness loss based on the propagation characteristics and the recovery characteristics includes: After each event segment ends, the amount of unrestored buffer is determined based on the shoreline storage status, the ecological buffer margin, and the engineering storage margin, and the amount of unrestored buffer is accumulated to obtain the accumulated amount of resilience loss.

8. The method according to claim 1, characterized in that, The counterfactual simulation of candidate control actions to determine the minimum cut set for propagation blocking and the minimum set for recovery control includes: The effects of a single candidate control action and a combination of candidate control actions on the degree of environmental risk exceeding limits, recovery time, and the cumulative amount of resilience loss are simulated respectively, wherein the degree of environmental risk exceeding limits is used to characterize the extent to which environmental risk exceeds the environmental risk threshold boundary; Candidate control actions that achieve at least two of the following: reducing the degree of environmental risk exceeding the limit, shortening the recovery time, and reducing the cumulative amount of toughness loss, are identified as target control actions. The propagation blocking minimum cut set and the recovery control minimum set are determined based on the target control action.

9. The method according to claim 8, characterized in that, The dual-threshold states include a security maintenance state, a propagation interception state, a resilience hardening state, and a cascading instability state; the generation of collaborative control instructions based on at least one of the propagation blocking minimum cut set and the recovery control minimum set includes: When the dual threshold state is the safety maintenance state, a maintenance control instruction is generated; When the dual threshold state is the propagation interception state, a prevention and control instruction is generated based on the propagation blocking minimum cut set; When the dual-threshold state is the toughness-enhanced state, a recovery control instruction is generated based on the recovery control minimum set; When the dual threshold state is the cascaded instability state, a joint control command is generated based on the propagation blocking minimum cut set and the recovery control minimum set.

10. The method according to claim 1, characterized in that, The physical residuals include quality flux residuals, propagation arrival time residuals, and recovery time residuals; the step of updating the dynamic risk propagation network parameters or the threshold judgment parameters in a single layer based on the physical residuals after control execution includes: If the quality flux residual exceeds the preset flux threshold, only the pollution input allocation parameter or the control measure effectiveness parameter is updated. If the propagation arrival time residual exceeds a preset time threshold, only the propagation delay parameter is updated; if the recovery duration residual exceeds a preset recovery threshold, only the threshold determination parameter is updated.

Images