Intelligent method and system for river emergency environmental event warning

Abstract

The application discloses a river sudden environmental event early warning intelligent method and system, relates to the technical field of water environment monitoring, environmental emergency and intelligent early warning, and comprises a river network monitoring diagram of upstream and downstream monitoring nodes, risk source nodes and sensitive receptor nodes; online water quality, hydrological and risk source fingerprint data are acquired; upstream and downstream anomaly vectors are generated; a propagation time delay window and an attenuation coefficient are generated according to the length of a river section, flow velocity and flow rate; the upstream anomaly vector is translated in time and corrected in amplitude, and then the propagation consistency is calculated with the downstream anomaly vector; when the propagation consistency reaches a threshold value, only the candidate risk sources that are reachable by a hydrodynamic path are subjected to fingerprint matching, and a graded early warning is output in combination with the expected time window of a pollution group reaching a sensitive receptor and the influence level. The application can reduce single-point false positives, improve the accuracy of sudden water pollution event identification, source tracing and sensitive receptor early warning.

Description

Technical Field

[0001] This invention relates to the fields of water environment monitoring, environmental emergency response and intelligent early warning technology, and in particular to an intelligent method and system for early warning of sudden environmental events in rivers. Background Technology

[0002] River emergencies are typically characterized by sudden occurrence, uncertain pollutant types, diffusion rates highly influenced by hydrological conditions, and short response windows. For key areas such as downstream of industrial parks, cross-boundary sections, upstream of drinking water sources, downstream of tailings ponds, and river sections near hazardous chemical transport routes, relying solely on fixed-section single-point exceedance alarms or manual experience-based assessments can easily lead to false alarms, missed alarms, delayed warnings, and difficulties in locating pollution sources.

[0003] Existing automatic water quality monitoring systems can typically acquire data such as pH, dissolved oxygen, conductivity, turbidity, ammonia nitrogen, permanganate index, total organic carbon, ultraviolet absorbance, and fluorescence spectrum in real time. However, in the event of a sudden environmental incident, a sudden change in a single indicator may be caused by sensor drift, equipment maintenance, rainfall runoff, algal activity, or local disturbances, and does not necessarily indicate a pollution accident.

[0004] Furthermore, pollutants exhibit patterns of migration, dilution, and attenuation along the direction of river flow. If the early warning model does not consider the propagation delay and attenuation changes of upstream anomalies reaching downstream sections, it may misclassify unrelated upstream and downstream anomalies as the same pollution event, or fail to provide sufficient advance warning before sensitive downstream receptors are affected.

[0005] Existing pollutant fingerprinting technology can match suspected risk sources through a combination of multiple indicators, fluorescence spectral features, or historical emission sample features. However, if fingerprint matching is performed only on all risk sources without considering river hydrodynamic paths, it is easy to mismatch risk sources that have no hydraulic connectivity.

[0006] Therefore, it is necessary to propose an intelligent early warning method and system suitable for sudden environmental events in rivers. This method uses hydrodynamic migration constraints as a prerequisite for pollution plume identification. After determining whether the anomalies in the upstream and downstream belong to the same pollution plume, it limits the scope of candidate risk sources for pollutant fingerprint matching and combines the expected impact time of sensitive receptors to provide graded early warning. Summary of the Invention

[0007] Technical problems to be solved The technical problem this invention aims to solve is that existing early warning methods for sudden environmental incidents in rivers tend to rely on single-point anomalies or simple upstream-downstream differences, failing to fully utilize the propagation delay and attenuation patterns of pollution plumes migrating along the river, resulting in high false alarm rates, large areas of suspected pollution source investigation, and insufficient advance warning for downstream sensitive receptors.

[0008] Technical solution To address the aforementioned technical problems, this invention provides an intelligent method for early warning of sudden environmental events in rivers, comprising the following steps: S1. Construct a river network monitoring map, which includes upstream monitoring nodes, downstream monitoring nodes, risk source nodes located between the upstream and downstream monitoring nodes or within the upstream catchment area, and sensitive receptor nodes located downstream of the downstream monitoring nodes, connected according to the direction of water flow. S2. Obtain online water quality data and hydrological data of the upstream and downstream monitoring nodes, as well as risk source fingerprint data of the risk source nodes. The online water quality data includes at least three of the following: conductivity, turbidity, ammonia nitrogen, total organic carbon, ultraviolet absorbance, or fluorescence spectral data. The hydrological data includes at least the river length, flow velocity, and flow rate. S3. Based on the historical valid values ​​and current values ​​of the online water quality data, generate anomaly vectors for the upstream monitoring node and the downstream monitoring node, respectively. The anomaly vectors include the standardized anomaly amplitude, anomaly rise rate and anomaly duration of multiple water quality indicators relative to the dynamic baseline. S4. Based on the river section length, flow velocity, and flow rate, generate hydrodynamic migration constraint parameters between the upstream monitoring node and the downstream monitoring node. The hydrodynamic migration constraint parameters include a propagation delay window and an attenuation coefficient. S5. The anomaly vector of the upstream monitoring node is shifted in time according to the propagation delay window, and the amplitude of the shifted anomaly vector is corrected according to the attenuation coefficient to obtain the migration-corrected anomaly vector. S6. Calculate the similarity between the migration correction anomaly vector and the anomaly vector of the downstream monitoring node, and use the similarity as the consistency of upstream and downstream anomaly propagation; when the consistency of upstream and downstream anomaly propagation is lower than the propagation consistency threshold, reduce the early warning contribution of the corresponding downstream monitoring node or determine it as an anomaly of a different pollution cluster. S7. When the consistency of the upstream and downstream abnormal propagation reaches the propagation consistency threshold, the candidate risk source node that can be reached by the hydrodynamic path is determined in reverse according to the propagation delay window, and a pollutant fingerprint is formed according to the online water quality data of the downstream monitoring node. The pollutant fingerprint is matched only with the risk source fingerprint data of the candidate risk source node to obtain the risk source fingerprint matching degree. S8. Based on the propagation delay window, the attenuation coefficient, and the hydrological data of the river section from the downstream monitoring node to the sensitive receptor node, calculate the expected arrival time window of the pollution plume to the sensitive receptor node, and determine the impact level of the sensitive receptor based on the receptor type, protection level, water intake status, or ecological sensitivity level of the sensitive receptor node. S9. Calculate the event risk value based on the consistency of upstream and downstream abnormal propagation, the matching degree of risk source fingerprint, the influence level of sensitive receptor, and the expected arrival time window. When the event risk value reaches the warning threshold, output the warning information including the suspected risk source node, the expected arrival time window, the influence level of sensitive receptor, the warning level, and the emergency sampling point.

[0009] This invention also provides an intelligent early warning system for sudden environmental events in rivers, comprising: The river network monitoring map construction module is used to construct a river network monitoring map that includes upstream monitoring nodes, downstream monitoring nodes, risk source nodes, and sensitive receptor nodes, and to record the water flow connectivity between each node; The data access module is used to acquire online water quality data and hydrological data of the upstream and downstream monitoring nodes, as well as risk source fingerprint data of the risk source nodes; An anomaly vector generation module is used to generate anomaly vectors for the upstream and downstream monitoring nodes based on the historical valid values ​​and current values ​​of online water quality data. The anomaly vectors include the standardized anomaly amplitude, anomaly rise rate, and anomaly duration of multiple water quality indicators relative to the dynamic baseline. The hydrodynamic migration constraint module is used to generate propagation delay windows and attenuation coefficients based on river length, flow velocity, and flow rate. The propagation consistency calculation module is used to perform time shifting on the anomaly vector of the upstream monitoring node according to the propagation delay window, and to perform amplitude correction on the time-shifted anomaly vector according to the attenuation coefficient to obtain the migration correction anomaly vector, and to calculate the similarity between the migration correction anomaly vector and the anomaly vector of the downstream monitoring node to obtain the propagation consistency of upstream and downstream anomalies. The candidate risk source screening and fingerprint matching module is used to determine the candidate risk source nodes that can be reached by the hydrodynamic path in reverse according to the propagation delay window when the consistency of the upstream and downstream abnormal propagation reaches the propagation consistency threshold, and to match the pollutant fingerprint formed by the downstream monitoring node only with the risk source fingerprint data of the candidate risk source nodes to obtain the risk source fingerprint matching degree. The sensitive receptor impact assessment module is used to calculate the expected arrival time window of the pollution plume to the sensitive receptor node based on the propagation delay window, attenuation coefficient, and river section hydrological data from the downstream monitoring node to the sensitive receptor node, and to determine the sensitive receptor impact level based on the receptor type, protection level, water intake status, or ecological sensitivity level of the sensitive receptor node. The early warning output module is used to calculate the event risk value based on the consistency of upstream and downstream abnormal propagation, the matching degree of risk source fingerprint, the influence level of sensitive receptor, and the expected arrival time window. When the event risk value reaches the early warning threshold, it outputs the suspected risk source node, the expected arrival time window, the influence level of sensitive receptor, the early warning level, and the emergency sampling point.

[0010] The present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the above-mentioned intelligent method for early warning of sudden environmental events in rivers.

[0011] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned intelligent method for early warning of sudden environmental events in rivers.

[0012] Beneficial effects Compared with existing technologies, this invention provides an intelligent method and system for early warning of sudden environmental events in rivers, which has the following beneficial effects: 1. This invention first uses hydrodynamic migration constraints to determine whether upstream and downstream anomalies conform to the law of pollution cloud propagation along water flow, rather than directly relying on single-point exceedance or upstream-downstream difference alarms, which can reduce false alarms caused by equipment anomalies, local disturbances and natural fluctuations.

[0013] 2. This invention performs time shifting on the upstream anomaly vector according to the propagation delay window and amplitude correction according to the attenuation coefficient, and then calculates the propagation consistency with the downstream anomaly vector, which can more accurately identify the migration relationship of the same pollution plume between the upstream and downstream sections.

[0014] 3. This invention performs pollutant fingerprint matching only on candidate risk source nodes that are reachable by hydrodynamic paths and have matching propagation delays, which can narrow down the scope of suspected pollution sources and reduce the probability of mismatch with irrelevant risk sources.

[0015] 4. This invention combines the predicted arrival time window of the pollution plume to the sensitive receptor with the graded early warning output of the sensitive receptor's impact level, enabling sensitive targets such as drinking water intakes, cross-boundary sections, and nature reserves to obtain more targeted early warnings.

[0016] 5. This invention can update the dynamic baseline, propagation delay window, attenuation coefficient, fingerprint database, and early warning threshold based on laboratory verification, on-site inspection, and handling results, adapting to different seasons, different hydrological conditions, and different watershed characteristics. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the intelligent early warning method for sudden environmental events in rivers according to the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of the river network monitoring map of the present invention.

[0019] Figure 3 This is a schematic diagram of the anomaly vector generation process of the present invention.

[0020] Figure 4 This is a schematic diagram illustrating the consistency judgment of upstream and downstream anomaly propagation based on hydrodynamic migration constraints according to the present invention.

[0021] Figure 5 This is a schematic diagram of the candidate risk source screening and pollutant fingerprint matching process of the present invention.

[0022] Figure 6 This is a schematic diagram of the sensitive receptor impact grading and early warning output process of the present invention.

[0023] Figure 7 This is a schematic diagram of the module structure of the intelligent early warning system for sudden environmental events in rivers according to the present invention. Detailed Implementation

[0024] The embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Where there is no conflict, the technical features in the following embodiments can be combined with each other.

[0025] Example 1: Overall Method Flow Reference Appendix Figure 1 This embodiment provides an intelligent method for early warning of sudden environmental incidents in rivers. This method is applicable to river sections downstream of industrial parks, upstream of drinking water sources, cross-administrative region junctions, downstream of tailings ponds, river sections near hazardous chemical transport channels, and river sections near tributary confluences.

[0026] Step S1: Constructing a river network monitoring map Reference Appendix Figure 2 Construct a river network monitoring map G=(V,E), where G is the river network monitoring map, V is the set of nodes, and E is the set of river segment edges.

[0027] The node set V includes upstream monitoring nodes, downstream monitoring nodes, risk source nodes, and sensitive receptor nodes. Upstream and downstream monitoring nodes can be fixed automatic water quality monitoring stations, tributary confluence monitoring points, mobile sampling points, or temporary emergency monitoring points. Risk source nodes can be industrial enterprise discharge outlets, sewage treatment plant discharge outlets, stormwater discharge outlets, tailings ponds, hazardous chemical storage sites, wharves, bridges, hazardous chemical transport routes, or historical accident sites. Sensitive receptor nodes can be drinking water source intakes, cross-administrative region junctions, nature reserves, important aquaculture areas, farmland irrigation intakes, or landscape water body replenishment points.

[0028] Each river segment in the set E records the river segment length, flow direction, average gradient, cross-sectional shape, velocity range, discharge range, tributary inflow information, and dam scheduling information.

[0029] Step S2: Obtain multi-source data Acquire online water quality data and hydrological data from upstream and downstream monitoring nodes, as well as risk source fingerprint data from risk source nodes.

[0030] Online water quality data can include pH, where pH is acidity or alkalinity; DO, where DO is dissolved oxygen; and multiple parameters such as conductivity, turbidity, ammonia nitrogen, total organic carbon, ultraviolet absorbance, fluorescence spectrum, petroleum hydrocarbons, heavy metals, total phosphorus, and total nitrogen.

[0031] UV254 represents the ultraviolet absorbance of 254 nanometers, used to characterize the absorption characteristics of some organic matter in water. Fluorescence spectral data can include fluorescence peak positions, fluorescence peak intensities, and fluorescence responses under different combinations of excitation and emission wavelengths.

[0032] Hydrological data include river length, flow velocity, flow rate, water level, water temperature, tributary inflow, and dam / sluice gate operation status.

[0033] Risk source fingerprint data can come from historical emission monitoring, enterprise emission outlet sampling, laboratory testing, fluorescence spectroscopy analysis, or historical accident records.

[0034] Step S3: Generate anomaly vectors Reference Appendix Figure 3 A dynamic baseline is generated for each water quality indicator at each monitoring node.

[0035] Let x_i,k(t) be the current monitoring value of the k-th water quality indicator at time t of the i-th monitoring node, where i is the monitoring node number, k is the water quality indicator number, and t is the monitoring time.

[0036] Let B_i,k(t) be the dynamic baseline of the i-th monitoring node at time t for the k-th water quality indicator. The dynamic baseline can be generated based on historical valid monitoring values ​​under the same season, time period, similar flow rate, and similar rainfall conditions.

[0037] In one implementation, the dynamic baseline is adopted using the following formula: B_i,k(t)=median{x_i,k(τ)|τ∈Ω(t)}; Where median is the median function, τ is the historical monitoring time, and Ω(t) is the set of historical times that have the same season, the same time period, and similar hydrological conditions as the current time t.

[0038] Let z_i,k(t) be the standardized anomaly amplitude of the i-th monitoring node at time t for the k-th water quality indicator. The standardized anomaly amplitude can be calculated using the following formula: z_i,k(t)=|x_i,k(t)-B_i,k(t)| / (MAD_i,k(t)+ε); Where MAD_i,k(t) is the median absolute deviation of the i-th monitoring node for the k-th water quality index at time t, and ε is a small positive number to avoid the denominator being zero.

[0039] Let r_i,k(t) be the abnormal rate of increase of the i-th monitoring node with respect to the k-th water quality index at time t, which can be calculated using the following formula: r_i,k(t)=[z_i,k(t)-z_i,k(t-Δt)] / Δt; Where Δt is the time interval between two adjacent monitoring times.

[0040] Let d_i,k(t) be the duration of the abnormality of the i-th monitoring node for the k-th water quality indicator at time t, which is the duration of the standardized abnormal amplitude continuously exceeding the abnormal threshold.

[0041] Therefore, an anomaly vector Y_i(t) is constructed, where Y_i(t) is the anomaly vector of the i-th monitoring node at time t: Y_i(t)=[z_i,1(t),r_i,1(t),d_i,1(t),...,z_i,K(t),r_i,K(t),d_i,K(t)]; Where K represents the number of water quality indicators involved in the early warning calculation.

[0042] Step S4: Generate hydrodynamic migration constraint parameters Reference Appendix Figure 4 For the river segment from upstream monitoring node u to downstream monitoring node v, calculate the propagation delay window and attenuation coefficient.

[0043] Let D_u,v be the length of the river segment from upstream monitoring node u to downstream monitoring node v, v_u,v(t) be the average flow velocity of the corresponding river segment at time t, and τ_u,v(t) be the reference propagation delay. The reference propagation delay can be calculated using the following formula: τ_u,v(t)=D_u,v / (v_u,v(t)+ε_v); Where ε_v is a small positive number to avoid zero flow velocity.

[0044] Considering the effects of flow velocity fluctuations, flow rate changes, dam scheduling, and tributary inflows, a propagation delay window T_u,v(t) is set, where T_u,v(t) is the propagation delay window from upstream monitoring node u to downstream monitoring node v: T_u,v(t)=[τ_u,v(t)-δ_u,v(t),τ_u,v(t)+δ_u,v(t)]; Wherein, δ_u,v(t) is the time delay tolerance, which can be determined based on the velocity fluctuation amplitude, the flow rate change amplitude, or historical abnormal propagation error.

[0045] Let α_u,v(t) be the attenuation coefficient from upstream monitoring node u to downstream monitoring node v. The attenuation coefficient can be calculated using the following formula: α_u,v(t)=exp[-λ_c·τ_u,v(t)]·Q_u(t) / (Q_v(t)+ε_Q); Where exp is the natural exponential function, λ_c is the decay parameter of pollutant category c, Q_u(t) is the upstream flow rate, Q_v(t) is the downstream flow rate, and ε_Q is a small positive number to avoid the denominator being zero.

[0046] For scenarios where it is difficult to determine the type of pollutant, the initial attenuation coefficient can be estimated using conductivity or conservative indicators, and then corrected based on laboratory verification results or historical event samples.

[0047] Step S5: Generate migration correction anomaly vector The anomaly vector of the upstream monitoring node is shifted according to the propagation delay window and the amplitude is corrected according to the attenuation coefficient to obtain the migration-corrected anomaly vector.

[0048] Let Y_hat_u→v(t) be the migration-corrected anomaly vector of the upstream monitoring node u at time t after propagation to the downstream monitoring node v, where Y_hat is the migration-corrected anomaly vector, which can be expressed as: Y_hat_u→v(t)=α_u,v(t)·Y_u(t-τ'); Where τ' is the propagation delay value that maximizes the similarity between upstream and downstream anomalies within the propagation delay window T_u,v(t), and Y_u(t-τ') is the upstream anomaly vector shifted according to the propagation delay τ'.

[0049] Step S6: Calculate the consistency of upstream and downstream anomaly propagation Calculate the similarity between the migration correction anomaly vector Y_hat_u→v(t) and the downstream monitoring node anomaly vector Y_v(t), and use it as the consistency of anomaly propagation between upstream and downstream.

[0050] Let C_u,v(t) represent the consistency of upstream and downstream anomaly propagation between upstream monitoring node u and downstream monitoring node v at time t. Cosine similarity can be used for calculation: C_u,v(t)=cos(Y_hat_u→v(t),Y_v(t)); Where cos is the cosine similarity function.

[0051] In another implementation, correlation coefficient similarity or dynamic time warping similarity can also be used. When C_u,v(t) is higher than the propagation consistency threshold, and the trends of abnormal peak order, abnormal duration, and abnormal amplitude change meet the consistency conditions, the probability of determining that the upstream and downstream anomalies belong to the same contamination cluster is relatively high.

[0052] When a local anomaly occurs at a downstream monitoring node, but the corresponding upstream anomaly cannot be matched with the downstream anomaly after being corrected by the propagation delay window and attenuation coefficient, the system reduces the contribution of the downstream anomaly to the event risk value, or marks it as an anomaly of a different pollution cluster, a local disturbance anomaly, or an equipment anomaly pending verification.

[0053] Step S7: Candidate risk source screening and pollutant fingerprint matching Reference Appendix Figure 5 When the consistency of abnormal propagation in the upstream and downstream reaches the propagation consistency threshold, the system determines the candidate risk source node in reverse based on the propagation delay window.

[0054] First, there is a water flow connection path between the risk source node and the abnormal downstream monitoring node; Second, the risk source node is located within the upstream catchment area of ​​the abnormal downstream monitoring node; Third, the predicted propagation time from the risk source node to the downstream monitoring node and the time difference between the occurrence of the anomaly fall within the propagation delay window or its extended range.

[0055] Pollutant fingerprint matching is performed on candidate risk source nodes that meet the above conditions, instead of performing indiscriminate matching on all risk source nodes.

[0056] Let F_v(t) be the pollutant fingerprint vector of downstream monitoring node v at time t. The pollutant fingerprint vector may include the amplitude of changes in conductivity, ammonia nitrogen, total organic carbon, and ultraviolet absorbance, fluorescence peak position, fluorescence peak intensity, ratio between indicators, abnormal rise rate, and abnormal duration.

[0057] Let P_s be the risk source fingerprint vector of candidate risk source node s, where s is the candidate risk source node number. Calculate the risk source fingerprint matching degree M_s(t), where M_s(t) is the fingerprint matching degree of candidate risk source node s at time t: M_s(t) = sim_F(F_v(t), P_s); Where sim_F is the fingerprint similarity function, which can be cosine similarity, Euclidean distance transformation similarity, fluorescence peak matching similarity, or weighted similarity.

[0058] The system outputs suspected risk source nodes in descending order of risk source fingerprint matching degree. If multiple candidate risk source nodes have similar fingerprint matching degrees, the system prioritizes outputting risk source nodes with higher propagation delay window matching degree, higher historical risk level, or closer proximity to the abnormal river section.

[0059] Step S8: Sensitive receptors affect grading Reference Appendix Figure 6 The system calculates the estimated arrival time window of the pollution plume to the sensitive receptor node based on the hydrological data of the river section from the downstream monitoring node to the sensitive receptor node.

[0060] Let T_v,r(t) be the propagation delay window from downstream monitoring node v to sensitive receptor node r, where r is the sensitive receptor node number. Let ETA_r(t) be the estimated arrival time window of the pollution plume to sensitive receptor node r, where ETA is the abbreviation for Estimated Arrival Time. The estimated arrival time window can be obtained by summing the propagation delay windows of each segment along the downstream path.

[0061] The impact level I_r(t) of sensitive receptors can be determined based on receptor type, protection level, water intake status, ecological sensitivity level, and expected arrival time window. Higher initial impact levels can be set for drinking water source intakes, core areas of nature reserves, important aquaculture areas, and cross-administrative region junctions.

[0062] When the expected arrival time window is earlier than the preset treatment preparation time, or when the sensitive receptor node is in a water intake state, an ecologically sensitive period, or a high protection level state, the impact level of the sensitive receptor should be increased.

[0063] Step S9: Calculate the event risk value and output an early warning. Let R_v(t) be the event risk value of the downstream monitoring node v at time t. The event risk value can be calculated using the following formula: R_v(t)=σ(b_0+b_1C_u,v(t)+b_2M_s(t)+b_3I_r(t)+b_4H_v(t)-b_5E_v(t)); Where σ is the Sigmoid function, which maps input values ​​to a range of 0 to 1; b_0 is the bias parameter; b_1 to b_5 are weight parameters; C_u,v(t) is the consistency of upstream and downstream anomaly propagation; M_s(t) is the risk source fingerprint matching degree; I_r(t) is the sensitive receptor influence level; H_v(t) is the local anomaly intensity of downstream monitoring nodes; and E_v(t) is the penalty for equipment anomalies or low data credibility.

[0064] The Sigmoid function can be represented as: σ(x) = 1 / (1 + e^{-x}); Where x is the input variable of the function, and e is the natural constant.

[0065] When the event risk value reaches the warning threshold, the system outputs a warning message. The warning message includes the suspected risk source node, the expected arrival time window, the impact level of the sensitive receptor, the warning level, the abnormal indicators, the emergency sampling point, the mobile sampling path, and the verification monitoring indicators.

[0066] The warning levels can be divided into three levels: Attention, Warning, and Emergency. Attention level is used to remind people to strengthen monitoring; Warning level is used to initiate mobile sampling and on-site inspections; Emergency level is used to indicate that there may be an impact on drinking water intakes, nature reserves, or cross-boundary sections, requiring coordinated emergency response.

[0067] Step S10: Feedback Update The system receives laboratory verification results, on-site inspection results, risk source verification results, and handling results, and updates the dynamic baseline, propagation delay window, attenuation coefficient, propagation consistency threshold, early warning threshold, and risk source fingerprint data based on the feedback results.

[0068] If an alert is confirmed to be a genuine sudden environmental event, the event's anomaly vector, propagation path, propagation delay, attenuation coefficient, pollutant fingerprint, suspected source, and handling results are written into the historical event database. If an alert is confirmed to be due to equipment anomaly, natural fluctuation, or local disturbance, the sample is treated as a false alarm sample and used to optimize the equipment anomaly penalty and propagation consistency threshold.

[0069] Example 2: System Structure Reference Appendix Figure 7 This embodiment provides an intelligent early warning system for sudden environmental events in rivers. The system includes a river network monitoring map construction module, a data access module, an anomaly vector generation module, a hydrodynamic migration constraint module, a propagation consistency calculation module, a candidate risk source screening and fingerprint matching module, a sensitive receptor impact assessment module, an early warning output module, and a feedback update module.

[0070] The river network monitoring map construction module is used to construct a river network monitoring map that includes upstream monitoring nodes, downstream monitoring nodes, risk source nodes, and sensitive receptor nodes, and to record the water flow connectivity between each node.

[0071] The data access module is used to acquire online water quality data, hydrological data, risk source fingerprint data, meteorological data, laboratory verification data, and on-site inspection data.

[0072] The anomaly vector generation module is used to generate anomaly vectors for upstream and downstream monitoring nodes based on the historical valid values ​​and current values ​​of online water quality data.

[0073] The hydrodynamic migration constraint module is used to generate propagation delay windows and attenuation coefficients based on river length, flow velocity, and flow rate.

[0074] The propagation consistency calculation module is used to time-shift the upstream anomaly vector according to the propagation delay window and correct its amplitude according to the attenuation coefficient to obtain the migration-corrected anomaly vector. Then, the similarity between the migration-corrected anomaly vector and the downstream anomaly vector is calculated to obtain the propagation consistency of upstream and downstream anomalies.

[0075] The candidate risk source screening and fingerprint matching module is used to determine candidate risk source nodes based on hydrodynamic path accessibility and propagation delay window when the propagation consistency reaches the propagation consistency threshold, and to match pollutant fingerprints only with the risk source fingerprint data of the candidate risk source nodes.

[0076] The sensitive receptor impact assessment module is used to calculate the expected arrival time window of the pollutant plume to the sensitive receptor node and determine the sensitive receptor impact level based on receptor type, protection level, water intake status, or ecological sensitivity level.

[0077] The early warning output module is used to calculate the event risk value and output the early warning level, suspected risk source node, expected arrival time window, sensitive receptor impact level and emergency sampling point when the event risk value reaches the early warning threshold.

[0078] The feedback update module is used to update the dynamic baseline, propagation delay window, attenuation coefficient, propagation consistency threshold, early warning threshold, or risk source fingerprint data based on laboratory verification results, on-site inspection results, or handling results.

[0079] Example 3: Typical Application Scenarios In a downstream river section of an industrial park, the system detected a simultaneous increase in conductivity, ammonia nitrogen, and ultraviolet absorbance at downstream monitoring nodes. The anomaly vector generation module calculated the downstream anomaly vector based on the dynamic baseline.

[0080] The hydrodynamic migration constraint module obtains a propagation delay window of 25 to 40 minutes based on the river length, real-time flow velocity, and flow rate between the upstream and downstream monitoring nodes, and calculates the attenuation coefficient.

[0081] The propagation consistency calculation module performs time shifting and attenuation correction on the abnormal vector from the upstream monitoring node 30 minutes ago. It finds that the migrated and corrected abnormal vector has a high similarity with the downstream abnormal vector in terms of abnormal peak order, duration and amplitude change trend, and determines that the upstream and downstream abnormal propagation consistency has reached the threshold.

[0082] The candidate risk source screening and fingerprint matching module determines the two risk source nodes reachable by the hydrodynamic path based on the propagation delay window. Then, it matches the downstream pollutant fingerprint with the risk source fingerprint data of the two candidate risk source nodes and outputs the risk source node with the highest matching degree as the suspected risk source node.

[0083] The sensitive receptor impact assessment module further determined that the contamination plume might reach the downstream drinking water intake within one hour, and that the intake was in operation, thus raising the sensitive receptor impact level. The early warning output module classified the event as an emergency and output recommended emergency sampling locations and mobile sampling routes.

[0084] Example 4: Parameter Setting Example The propagation consistency threshold can be set from 0.65 to 0.85 based on historical event samples. For upstream river sections of drinking water sources, a lower threshold can be used to improve early warning sensitivity; for ordinary sections with more equipment disturbances, a higher threshold can be used to reduce false alarms.

[0085] The preset preparation time can be set to 30 minutes, one hour, two hours, or other durations, depending on the local emergency response capabilities. If the expected arrival time of the pollution plume is shorter than the preset preparation time, the system will raise the warning level.

[0086] Risk source fingerprint data can be updated every six months or after changes occur in the enterprise's production process, wastewater treatment process, or discharge destination. For historically high-risk enterprises, tailings ponds, and hazardous chemical storage sites, the fingerprint database update frequency can be increased.

[0087] This invention can be deployed on ecological and environmental supervision platforms, water environment automatic monitoring platforms, industrial park environmental risk supervision platforms, drinking water source early warning platforms, and emergency command platforms for sudden environmental incidents. Its data sources are compatible with existing automatic water quality monitoring stations, hydrological stations, online monitoring equipment at enterprise discharge outlets, mobile sampling terminals, and laboratory testing systems, without requiring large-scale replacement of existing hardware systems.

[0088] This invention enables intelligent auxiliary decision-making for anomaly detection, pollution plume propagation assessment, suspected source screening, sensitive receptor impact prediction, and emergency sampling site deployment before and after sudden environmental events in rivers. It has a clear engineering implementation path and industrial application value.

Claims

1. An intelligent method for early warning of sudden environmental incidents in rivers, characterized in that, Includes the following steps: S1. Construct a river network monitoring map, which includes upstream monitoring nodes, downstream monitoring nodes, risk source nodes located between the upstream and downstream monitoring nodes or within the upstream catchment area, and sensitive receptor nodes located downstream of the downstream monitoring nodes, connected according to the direction of water flow. S2. Obtain online water quality data and hydrological data of the upstream and downstream monitoring nodes, as well as risk source fingerprint data of the risk source nodes. The online water quality data includes at least three of the following: conductivity, turbidity, ammonia nitrogen, total organic carbon, ultraviolet absorbance, or fluorescence spectral data. The hydrological data includes at least the river length, flow velocity, and flow rate. S3. Based on the historical valid values ​​and current values ​​of the online water quality data, generate anomaly vectors for the upstream monitoring node and the downstream monitoring node, respectively. The anomaly vectors include the standardized anomaly amplitude, anomaly rise rate and anomaly duration of multiple water quality indicators relative to the dynamic baseline. S4. Based on the river section length, flow velocity, and flow rate, generate hydrodynamic migration constraint parameters between the upstream monitoring node and the downstream monitoring node. The hydrodynamic migration constraint parameters include a propagation delay window and an attenuation coefficient. S5. The anomaly vector of the upstream monitoring node is shifted in time according to the propagation delay window, and the amplitude of the shifted anomaly vector is corrected according to the attenuation coefficient to obtain the migration-corrected anomaly vector. S6. Calculate the similarity between the migration correction anomaly vector and the anomaly vector of the downstream monitoring node, and use the similarity as the consistency of upstream and downstream anomaly propagation; when the consistency of upstream and downstream anomaly propagation is lower than the propagation consistency threshold, reduce the early warning contribution of the corresponding downstream monitoring node or determine it as an anomaly of a different pollution cluster. S7. When the consistency of the upstream and downstream abnormal propagation reaches the propagation consistency threshold, the candidate risk source node that can be reached by the hydrodynamic path is determined in reverse according to the propagation delay window, and a pollutant fingerprint is formed according to the online water quality data of the downstream monitoring node. The pollutant fingerprint is matched only with the risk source fingerprint data of the candidate risk source node to obtain the risk source fingerprint matching degree. S8. Based on the propagation delay window, the attenuation coefficient, and the hydrological data of the river section from the downstream monitoring node to the sensitive receptor node, calculate the expected arrival time window of the pollution plume to the sensitive receptor node, and determine the impact level of the sensitive receptor based on the receptor type, protection level, water intake status, or ecological sensitivity level of the sensitive receptor node. S9. Calculate the event risk value based on the consistency of upstream and downstream abnormal propagation, the matching degree of risk source fingerprint, the influence level of sensitive receptor, and the expected arrival time window. When the event risk value reaches the warning threshold, output the warning information including the suspected risk source node, the expected arrival time window, the influence level of sensitive receptor, the warning level, and the emergency sampling point.

2. The intelligent early warning method for river emergencies according to claim 1, characterized in that, Step S3 includes: A dynamic baseline is generated based on the historical effective values ​​of the same monitoring node under the same season, time period, and similar hydrological conditions; The standardized anomaly amplitude is calculated based on the degree of deviation between the current online water quality data and the dynamic baseline; The rate of increase of anomalies is calculated based on the standardized changes in the amplitude of anomalies over continuous monitoring time. The duration of the anomaly is calculated based on the duration during which the standardized anomaly amplitude continuously exceeds the anomaly threshold. The standardized abnormal amplitude, abnormal rise rate, and abnormal duration are combined to form the abnormal vector.

3. The intelligent early warning method for river emergencies according to claim 1, characterized in that, Step S4 includes: The baseline propagation delay is calculated based on the river length and flow velocity. Based on the range of flow velocity fluctuations, the range of flow rate changes, or the dam scheduling status, the reference propagation delay is extended to obtain a propagation delay window; The attenuation coefficient is determined based on upstream flow, downstream flow, tributary inflow, and pollutant type.

4. The intelligent early warning method for river emergencies according to claim 1, characterized in that, The similarity in step S6 is one of cosine similarity, correlation coefficient similarity, or dynamic time warping similarity; When the migration correction anomaly vector and the anomaly vector of the downstream monitoring node all meet the consistency conditions in terms of peak order, anomaly duration, and anomaly amplitude change trend across multiple water quality indicators, it is determined that the upstream and downstream anomaly propagation consistency has reached the propagation consistency threshold.

5. The intelligent method for early warning of sudden environmental events in rivers according to claim 1, characterized in that, The pollutant fingerprint in step S7 includes at least three of the following: conductivity change amplitude, ammonia nitrogen change amplitude, total organic carbon change amplitude, ultraviolet absorbance change amplitude, fluorescence peak position, fluorescence peak intensity, ratio between indicators, abnormal rise rate, or abnormal duration. The risk source fingerprint data includes one or more of the following: industrial wastewater fingerprint, domestic sewage fingerprint, agricultural non-point source pollution fingerprint, oil pollution fingerprint, heavy metal pollution fingerprint, tailings dam leakage fingerprint, or transportation leakage fingerprint.

6. The intelligent early warning method for sudden environmental events in rivers according to claim 1, characterized in that, The sensitive receptor nodes in step S8 include one or more of the following: drinking water source intake, cross-administrative region junction, nature reserve, important aquaculture area, farmland irrigation intake, or landscape water body replenishment point; When the expected arrival time window is earlier than the preset treatment preparation time, or when the sensitive receptor node is in a water intake state, an ecologically sensitive period, or a high protection level state, the impact level of the sensitive receptor is increased.

7. The intelligent early warning method for sudden environmental events in rivers according to claim 1, characterized in that, Step S9 is followed by: Based on the suspected risk source nodes and emergency sampling points, generate mobile sampling paths and review monitoring indicators; Receive laboratory verification results, on-site inspection results, or handling results; Update the dynamic baseline, propagation delay window, attenuation coefficient, propagation consistency threshold, early warning threshold, or risk source fingerprint data based on the laboratory verification results, on-site inspection results, or handling results.

8. An intelligent early warning system for sudden environmental incidents in rivers, characterized in that, include: The river network monitoring map construction module is used to construct a river network monitoring map that includes upstream monitoring nodes, downstream monitoring nodes, risk source nodes, and sensitive receptor nodes, and to record the water flow connectivity between each node; The data access module is used to acquire online water quality data and hydrological data of the upstream and downstream monitoring nodes, as well as risk source fingerprint data of the risk source nodes; An anomaly vector generation module is used to generate anomaly vectors for the upstream and downstream monitoring nodes based on the historical valid values ​​and current values ​​of online water quality data. The anomaly vectors include the standardized anomaly amplitude, anomaly rise rate, and anomaly duration of multiple water quality indicators relative to the dynamic baseline. The hydrodynamic migration constraint module is used to generate propagation delay windows and attenuation coefficients based on river length, flow velocity, and flow rate. The propagation consistency calculation module is used to perform time shifting on the anomaly vector of the upstream monitoring node according to the propagation delay window, and to perform amplitude correction on the time-shifted anomaly vector according to the attenuation coefficient to obtain the migration correction anomaly vector, and to calculate the similarity between the migration correction anomaly vector and the anomaly vector of the downstream monitoring node to obtain the propagation consistency of upstream and downstream anomalies. The candidate risk source screening and fingerprint matching module is used to determine the candidate risk source nodes that can be reached by the hydrodynamic path in reverse according to the propagation delay window when the consistency of the upstream and downstream abnormal propagation reaches the propagation consistency threshold, and to match the pollutant fingerprint formed by the downstream monitoring node only with the risk source fingerprint data of the candidate risk source nodes to obtain the risk source fingerprint matching degree. The sensitive receptor impact assessment module is used to calculate the expected arrival time window of the pollution plume to the sensitive receptor node based on the propagation delay window, attenuation coefficient, and river section hydrological data from the downstream monitoring node to the sensitive receptor node, and to determine the sensitive receptor impact level based on the receptor type, protection level, water intake status, or ecological sensitivity level of the sensitive receptor node. The early warning output module is used to calculate the event risk value based on the consistency of upstream and downstream abnormal propagation, the matching degree of risk source fingerprint, the influence level of sensitive receptor, and the expected arrival time window. When the event risk value reaches the early warning threshold, it outputs the suspected risk source node, the expected arrival time window, the influence level of sensitive receptor, the early warning level, and the emergency sampling point.

9. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it implements the intelligent method for early warning of sudden environmental events in rivers as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the intelligent early warning method for sudden environmental events in rivers as described in any one of claims 1 to 7.

Images