A method and system for detecting river water pollution

Abstract

This invention discloses a method and system for detecting river water pollution, relating to the field of river water pollution detection, for accurately identifying real pollution events. The method includes: acquiring indirect water parameters related to upstream discharge activities; analyzing short-term variation patterns of the indirect water parameters; comparing the short-term variation patterns with preset historical normal patterns to identify persistent changes where the short-term variation patterns differ from the historical normal patterns; correlating the persistent changes with the time information of the discharge activities to obtain correlation results; judging potential discharge risks based on persistent changes and correlation results; and adjusting downstream monitoring parameters and water quality anomaly judgment thresholds in response to potential discharge risks.

Description

Technical Field

[0001] This invention relates to the field of river water pollution detection, and more particularly to a method and system for river water pollution detection. Background Technology

[0002] Current river water pollution detection systems often struggle to effectively integrate information from different sources in complex water environments, especially when pollutants transform in time and space. This makes it difficult to accurately identify pollution events and their sources.

[0003] For example, a company specializing in fine chemicals production in an upstream industrial park introduced a novel synthesis process. This new process generates a specific organic intermediate compound during production. This compound exhibits relatively low biotoxicity when discharged, and its chemical structure means that it does not immediately show significant anomalies in conventional water quality monitoring, such as using existing indicators like chemical oxygen demand (COD), ammonia nitrogen, and total phosphorus. The sensors and built-in analytical logic of existing monitoring systems cannot directly and instantly detect the presence of this specific organic intermediate compound. Although the company ensures that conventional indicators meet national discharge standards after wastewater treatment, trace amounts of this specific organic intermediate compound still enter the river with the treated effluent.

[0004] When this specific organic intermediate compound enters a river, it does not remain chemically inert. In the complex natural aquatic environment of the river, it begins to undergo a series of slow chemical and biological transformation processes. These degradation processes do not directly generate another "standard" pollutant that can be directly identified by existing sensors, but rather produce a series of secondary metabolites. Some of these secondary metabolites have high biodegradability and gradually consume dissolved oxygen in the water during their degradation, thus indirectly causing a decrease in dissolved oxygen levels downstream; others may have stronger reducing or oxidizing properties, thus making a cumulative contribution to the chemical oxygen demand (COD) of the water or altering the pH value of the water. More complexly, some secondary metabolites may have a certain complexing ability, capable of binding with trace heavy metal ions already present in the riverbed sediment or water, changing the occurrence form of these heavy metals from a relatively stable insoluble state to a more easily migratable soluble state, thereby causing abnormal fluctuations in the concentration of specific heavy metal ions detected at downstream monitoring points, even though these heavy metals do not directly originate from the emissions of the fine chemical plant.

[0005] Because this transformation process within the river channel requires a certain amount of time and distance, water quality data at monitoring stations immediately adjacent to the fine chemical plant's discharge outlet typically appear normal, or show only extremely weak and imperceptible fluctuations. The real anomalies often appear in downstream areas, farther from the discharge outlet, sometimes even across several monitoring stations. In these cases, the detected anomaly is no longer a sharp spike in the concentration of a single pollutant within a short period, but rather a more diffuse, longer-lasting, and "complex" anomaly involving multiple water quality indicators. For example, the system might observe a sustained, slow increase in chemical oxygen demand (COD) values ​​at multiple downstream stations, but the increase might not be significant enough to trigger a high-intensity alarm; simultaneously, dissolved oxygen levels might slightly decrease, and pH levels might fluctuate irregularly. These changes are often gradual, lacking a clear "breakout point," making it difficult for the system to correlate them with specific sudden pollution events.

[0006] Faced with such complex anomalies caused by endogenous transformation, existing analytical logic is mainly designed based on the direct correlation logic of "point source emission - downstream diffusion - concentration increase". It excels at identifying rapid changes in high-intensity, short-term, single or a few pollutant indicators and directly linking them to specific upstream discharge outlets. However, for "secondary pollution" formed by the slow transformation of upstream "precursors" within the river channel, the logic cannot establish an effective causal chain. The system may detect anomalies in downstream chemical oxygen demand, dissolved oxygen, or pH, but because these anomalies do not directly originate from the "standard" pollutant emission characteristics of a known discharge outlet, and their occurrence has significant lag and diffusion, the system often cannot accurately trace them back to the fine chemical plant. Sometimes, the system may even misjudge these slow, complex changes as ecological fluctuations within the river channel itself, or attribute them to other unrelated environmental factors, leading to "underreporting" or "misjudging" of real pollution events. Summary of the Invention

[0007] This invention provides a method for detecting river water pollution, aiming to solve the problems of current river water pollution detection systems in complex water environments, especially when facing situations where pollutants transform in time and space. These systems struggle to effectively integrate information from different sources, leading to difficulties in accurately identifying pollution events and their sources. Furthermore, the systems may misjudge slow, complex changes as ecological fluctuations in the river itself, or attribute them to other unrelated environmental factors, resulting in "underreporting" or "misjudging" of real pollution events.

[0008] Firstly, this application discloses a method for detecting river water pollution, including: Obtain indirect water parameters related to upstream discharge activities; Analyze the short-term variation patterns of indirect water parameters; Compare short-term change patterns with pre-defined historical normal patterns to identify persistent changes where short-term change patterns differ from historical normal patterns; The correlation results are obtained by linking persistent changes with the time information of emission activities; Based on persistent changes and correlation results, assess potential emission risks; In response to potential emission risks, downstream monitoring parameters and water quality anomaly detection thresholds were adjusted.

[0009] This technical solution can identify persistent changes by analyzing short-term variation patterns of indirect water parameters and comparing them with historical normal patterns. These changes can then be correlated with the timing of emission activities to determine potential emission risks and adjust downstream monitoring parameters and water quality anomaly judgment thresholds. This effectively solves the problems of existing technologies, such as difficulty in accurately identifying pollution events and their sources, and the tendency for "missed reports" or "false judgments."

[0010] Furthermore, based on persistent changes and correlations, potential emission risks are assessed, including: Identify interference signals from non-target emission sources; Cross-validation of non-target emission source characteristics was performed on persistent changes to obtain cross-validation results; Based on the cross-validation and correlation results, potential emission risks are assessed.

[0011] This technical solution enables the identification of interference signals from non-target emission sources and the performance of feature cross-validation, thereby more accurately assessing potential emission risks and avoiding misjudging interference from non-target emission sources as pollution events.

[0012] Based on the above, the deviation of the correlation evolution pattern is correlated with the time information of emission activities, including: Obtain hydrological information for the current river section; Based on hydrological information, calculate the material transport time from the discharge point to the monitoring point; Set a relevant time window based on the material transport time; The occurrence time of persistent changes is compared with the associated time window to determine whether persistent changes fall within the associated time window.

[0013] This technical solution allows for the setting of reasonable correlation time windows by taking into account hydrological information and material transport time, thereby more accurately linking persistent changes with emission activities and improving the accuracy of source tracing.

[0014] Furthermore, cross-validation of non-target emission source characteristics is performed on persistent changes, including: Collect emission signals generated by non-target emission sources during operation; Extract the characteristic fingerprint of the emission signal; The characteristic fingerprint is compared with the characteristic fingerprint of historical non-target emission sources to identify whether the cleaning agent composition, cleaning process or emission characteristics of non-target emission sources have changed. In response to changes in the composition of cleaning agents, cleaning processes, or emission characteristics of non-target emission sources, the characteristic fingerprint database of non-target emission sources is updated. Using the updated feature fingerprint database of non-target emission sources, cross-validation of non-target emission source features is performed on persistent changes.

[0015] This technical solution can ensure the accuracy and timeliness of cross-validation by dynamically updating the feature fingerprint database of non-target emission sources, effectively eliminating the interference of changes in non-target emission sources on pollution judgment.

[0016] In some preferred embodiments, the step of updating the characteristic fingerprint database of non-target emission sources in response to changes in the cleaning agent composition, cleaning process, or emission characteristics of the non-target emission sources includes: Calculate the statistical characteristics and trends of the current emission signal; The statistical characteristics and trends of the current emission signal are compared with the most recent stable fingerprint stored in the feature fingerprint database; When the comparison results show that there is a persistent difference between the statistical characteristics or trend of the current emission signal and the most recent stable fingerprint that exceeds the preset slight deviation range, it is determined that the characteristics of the non-target emission source have undergone a gradual change. In response to gradual changes, the fingerprint representation in the feature fingerprint database is corrected by weighted averaging based on the statistical characteristics and trends of the current emission signals. Verify the consistency between the corrected fingerprint and the historical fingerprint version; When the corrected fingerprint passes the consistency verification, the corrected fingerprint is marked as a new stable fingerprint, and the feature fingerprint database is updated.

[0017] This technical solution can ensure that the updates to the feature fingerprint database reflect the real changes of non-target emission sources and maintain the stability of the fingerprints through a progressive correction and consistency verification mechanism, thus avoiding frequent erroneous updates.

[0018] More specifically, the corrected fingerprint will be verified to be consistent with historical fingerprint versions, including: Extract multiple dimensions of the current fingerprint from the emission signal. These multiple dimensions include the instantaneous fluctuation amplitude of the signal, the duration, the intensity of specific frequency components, and the rate of change of these features over time. Calculate the difference in multiple dimensions of features between the current fingerprint and each historical fingerprint version to identify the target historical fingerprint version with the highest similarity to the current fingerprint; Determine whether the difference between the current fingerprint and the target historical fingerprint version is within the preset normal evolution range; When the difference is within the normal evolution range, the current fingerprint is marked as normal evolution, and the current fingerprint marked as normal evolution is added to the historical fingerprint version library to update the evolution trajectory of historical fingerprints. When the difference exceeds the normal evolution range, the current fingerprint is marked as an abnormal drift and an alarm is triggered.

[0019] This technical solution enables more precise judgment on whether changes in the characteristics of non-target emission sources are normal evolution or abnormal drift through multi-dimensional feature analysis and consistency verification with historical fingerprint versions, thereby improving the accuracy of early warning.

[0020] Preferably, the target historical fingerprint version is determined, including: Calculate the similarity between the current fingerprint and each historical fingerprint version; When a historical fingerprint version exists in the historical fingerprint version library with a similarity higher than a preset similarity threshold to the current fingerprint, the historical fingerprint version with the highest similarity to the current fingerprint will be used as the target historical fingerprint version.

[0021] This technical solution enables the rapid and accurate identification of the most relevant historical fingerprint versions through similarity calculation and threshold filtering, thereby improving the efficiency of consistency verification.

[0022] In one implementation, calculating the similarity between the current fingerprint and each historical fingerprint version includes: For any historical fingerprint version, calculate the similarity between the current fingerprint and the historical fingerprint version according to the similarity formula; The similarity formula is: Match_Score=1 / (1+Sum((Actual_C_i-Predicted_C_i)2)); Where Match_Score is the similarity score; Actual_C_i is the concentration of the i-th detected indicator; and Predicted_C_i is the concentration of the i-th indicator in the historical fingerprint version.

[0023] This technical solution can quantify the similarity between the current fingerprint and historical fingerprint versions through a clear similarity calculation formula, providing an objective basis for subsequent judgments.

[0024] As a technological improvement, in response to potential emission risks, downstream monitoring parameters and water quality anomaly detection thresholds have been adjusted, including: Increase the monitoring frequency of dissolved oxygen, chemical oxygen demand, pH value, and specific heavy metals downstream, and lower the threshold for judging water quality anomalies.

[0025] This technical solution enables the dynamic adjustment of downstream monitoring strategies based on the assessment of potential emission risks, thereby optimizing resource allocation and improving sensitivity to specific pollutants.

[0026] Secondly, this application also discloses a river water pollution detection system, which includes: The acquisition module is used to acquire indirect water parameters related to upstream discharge activities; The analysis module is used to analyze short-term variation patterns of indirect water parameters; The comparison module is used to compare short-term change patterns with preset historical normal patterns to identify persistent changes where there are differences between short-term change patterns and historical normal patterns. The correlation module is used to correlate persistent changes with time information of emission activities to obtain correlation results; The judgment module is used to assess potential emission risks based on continuous changes and correlation results; The processing module adjusts downstream monitoring parameters and water quality anomaly detection thresholds in response to potential emission risks.

[0027] This technical solution enables the implementation of various functions of river water pollution detection methods through modular design, providing an efficient and accurate pollution detection and early warning solution.

[0028] Beneficial effects The river water pollution detection method disclosed in this application acquires indirect water parameters related to upstream discharge activities and analyzes their short-term change patterns. These patterns are then compared with preset historical normal patterns to identify persistent changes that deviate from the expected patterns. Subsequently, these persistent changes are correlated with the time information of discharge activities to obtain correlation results. Finally, based on the persistent changes and correlation results, potential discharge risks are assessed, and in response to these risks, downstream monitoring parameters and water quality anomaly detection thresholds are adjusted.

[0029] This method effectively solves the problem in existing technologies that, in complex aquatic environments, especially when pollutants undergo transformation in time and space, it is difficult to effectively integrate information from different sources, leading to difficulties in accurately identifying pollution events and their sources. Specifically, existing systems often fail to detect in a timely manner the "secondary pollution" formed by the slow transformation of "precursors" discharged from upstream within the river channel. They are prone to misjudging these slow, complex changes as ecological fluctuations within the river channel itself, or attributing them to other unrelated environmental factors, thus leading to "underreporting" or "misjudgment" of real pollution events.

[0030] This application, by focusing on "indirect water parameters" and "short-term change patterns," can capture subtle, gradual changes during pollutant transformation, rather than relying solely on large fluctuations in direct pollutant concentrations. By comparing these changes with "historical normal patterns," "persistent changes" can be effectively identified, avoiding misjudging normal ecological fluctuations or weak, non-persistent disturbances as pollution. More importantly, by correlating "persistent changes" with "temporal information of emission activities," this application can establish a causal chain of "precursor emission - in-river transformation - secondary pollution manifestation," enabling effective source tracing even when pollution signals exhibit temporal and spatial lag and dispersion. Finally, based on the identified potential emission risks, downstream monitoring parameters and water quality anomaly judgment thresholds are dynamically adjusted, allowing the monitoring system to specifically improve its sensitivity to particular pollutants and optimize monitoring resource allocation. This significantly enhances the early warning capability and response efficiency for complex water pollution events, effectively avoiding the problems of "missed reporting" or "false judgment" in existing technologies, achieving unexpected technical results. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of a river water pollution detection method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of another method for detecting river water pollution provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a river water pollution detection system provided in an embodiment of the present invention. Detailed Implementation

[0032] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0033] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0034] Traditional river pollution detection systems often struggle to effectively integrate information from different sources in complex aquatic environments, particularly when pollutants transform over time and space. This hinders the accurate identification of pollution events and their sources. These systems frequently misjudge subtle changes or fail to detect less noticeable pollutants, leading to unreliable alerts and impacting emergency response efficiency. To address this, this application proposes a river pollution detection method. This method acquires indirect water parameters related to upstream discharge activities, analyzes their short-term variation patterns, and compares them with historical normal patterns to identify persistent changes. Subsequently, these changes are correlated with the temporal information of discharge activities to assess potential discharge risks. Downstream monitoring parameters and water quality anomaly thresholds are then responsively adjusted to more accurately and promptly identify and respond to water pollution events.

[0035] The following specific embodiments will provide a detailed description and explanation of a river water pollution detection method provided in this application.

[0036] Reference Figure 1 This invention provides a method for detecting river water pollution, comprising the following steps: S1, obtain indirect water parameters related to upstream discharge activities.

[0037] Indirect water parameters can be indicators that reflect the comprehensive properties of water bodies, such as dissolved oxygen, chemical oxygen demand, pH value, conductivity, and turbidity.

[0038] Specifically, these parameters can be obtained in various ways. For example, real-time monitoring can be conducted using various water quality sensors deployed in the upstream areas of the river. The sensors transmit the collected data to a data processing center via wired or wireless networks. In addition, these parameters can also be obtained through regular manual sampling and laboratory analysis, or by combining satellite remote sensing, drone patrols, and other technologies to obtain large-scale apparent parameters of water bodies, and then obtaining indirect water body parameters through model inversion.

[0039] S2. Analyze the short-term variation patterns of indirect water parameters.

[0040] The analysis of short-term variation patterns aims to capture the fluctuation characteristics of water parameters over shorter time scales. For example, time series analysis methods, such as moving averages, exponential smoothing, and wavelet analysis, can be used to identify instantaneous fluctuations, periodic changes, or trend changes in parameters. Statistical methods, such as calculating the mean, variance, skewness, and kurtosis of parameters, can also be used to quantify their short-term distribution characteristics. For instance, a sliding time window can be set, within which the mean and standard deviation of dissolved oxygen can be calculated to reflect its stability and volatility over the past few hours.

[0041] S3. Compare the short-term change pattern with the preset historical normal pattern to identify persistent changes where the short-term change pattern differs from the historical normal pattern.

[0042] Among them, the historical normal pattern is established based on long-term and stable water quality monitoring data, representing the behavior of water body parameters under conditions of no pollution or normal environmental disturbance.

[0043] The comparison process can employ various statistical or machine learning methods. For example, distance metrics (such as Euclidean distance and Mahalanobis distance) can be calculated between the current short-term variation pattern and the historical normal pattern, or anomaly detection algorithms (such as Isolation Forest, Support Vector Machine, and Neural Network) can be used to determine whether the current pattern deviates from the normal range. Persistent variation refers to differences that are not instantaneous or accidental, but rather persist for a period of time, indicating the possible existence of a stable source of disturbance. For example, if the short-term mean of a water body parameter is significantly higher than its historical normal mean for several consecutive hours or days, it can be identified as persistent variation.

[0044] S4. Correlate persistent changes with time information of emission activities to obtain the correlation results.

[0045] The timing information for emission activities can include data on the emission time, cycle, and amount from potential pollution sources such as upstream industrial enterprises, agricultural non-point sources, and municipal wastewater treatment plants. The correlation process aims to determine whether the identified persistent changes correspond to a specific emission activity over time. For example, if an industrial enterprise discharges wastewater during a specific period, and downstream water parameters show persistent changes for a period afterward, a correlation can be considered to exist between the two. This correlation can be achieved through methods such as time window matching and cross-correlation analysis.

[0046] S5. Based on continuous changes and correlation results, assess potential emission risks.

[0047] The assessment of potential emission risks is a comprehensive decision-making process that combines abnormal changes in water parameters with the temporal correlation of emission activities.

[0048] If a persistent change is significant and strongly correlated with a particular emission activity, it is considered high-risk; if the change is not significant or the correlation is weak, it is considered low-risk or no-risk. For example, if a continuous decrease in dissolved oxygen is detected, and this decrease highly overlaps with the nighttime emission period of an upstream chemical plant, a high potential emission risk can be identified.

[0049] S6. In response to potential emission risks, adjust downstream monitoring parameters and water quality anomaly judgment thresholds.

[0050] When a potential emission risk is identified, the monitoring strategies at downstream monitoring stations can be adjusted accordingly. For example, the monitoring frequency for specific pollutants can be increased, or more sensitive sensors can be activated for detection. Simultaneously, the water quality anomaly detection threshold can be dynamically adjusted to make it more sensitive to potential risks. For instance, when a potential heavy metal emission risk is identified, the anomaly detection threshold for heavy metal concentrations at downstream monitoring points can be appropriately lowered to detect trace amounts of pollution earlier.

[0051] Specifically, increase the monitoring frequency of downstream dissolved oxygen, chemical oxygen demand, pH value, and specific heavy metals, and lower the threshold for judging water quality anomalies.

[0052] For example, if the abnormal threshold for dissolved oxygen is 5 mg / L under normal circumstances, this threshold may be adjusted to 4 mg / L after a potential emission risk is identified, in order to reflect the tolerance for water quality changes under specific risk scenarios.

[0053] The river water pollution detection method proposed in this application acquires indirect water parameters related to upstream discharge activities, analyzes their short-term change patterns, and compares them with preset historical normal patterns. This allows for the identification of persistent changes that deviate from historical normal patterns. This method can effectively capture even weak or gradual anomalous signals in water bodies caused by upstream discharge activities, overcoming the excessive reliance of traditional systems on rapid changes in a single indicator.

[0054] By correlating identified persistent changes with temporal information about emission activities, this method can establish causal chains between anomalies and potential pollution sources, thereby more accurately assessing potential emission risks. This contrasts sharply with existing technologies that struggle to trace complex downstream anomalies back to specific upstream emission sources. For example, when specific organic intermediate compounds emitted by an upstream fine chemical plant slowly transform in a river, causing complex and gradual anomalies in downstream dissolved oxygen, chemical oxygen demand, or pH, this method can link them to the plant's emission activities through temporal correlation, avoiding misjudgments or underreporting.

[0055] Furthermore, this method can dynamically adjust downstream monitoring parameters and water quality anomaly detection thresholds in response to potential emission risks. This adaptive mechanism enables the monitoring system to optimize monitoring resource allocation based on actual risk conditions, improving the timeliness and accuracy of early warnings. For example, when a potential risk is identified for a specific pollutant, the system can automatically increase the monitoring frequency for that pollutant and lower its anomaly detection threshold, thus issuing an early warning before the pollutant concentration reaches traditional alarm levels. This significantly enhances the ability to detect secondary and delayed pollution, effectively compensating for the shortcomings of existing systems in dealing with complex and transformative pollution events, thereby improving the overall efficiency and reliability of river water pollution detection.

[0056] In some of the embodiments described above in this application, potential emission risks are determined based on continuous changes and correlation results. However, in practical applications, the river water environment is complex and there may be interference from non-target emission sources, which may lead to inaccurate judgment of potential emission risks or even false alarms.

[0057] In one possible design, such as Figure 2 As shown, in order to determine potential emission risks based on persistent changes and associated results, this application may further include the following steps: S101. Identify interference signals from non-target emission sources.

[0058] Specifically, identifying interference signals from non-target emission sources refers to recognizing signals that are not directly caused by upstream emission activities but may produce similar abnormal changes in water parameters. For example, these interference signals can originate from natural environmental changes (such as rainfall runoff or algal blooms), routine emissions from other legitimate sources, or drift of the monitoring equipment itself. The purpose of identifying these interference signals is to distinguish them from genuine upstream emission activities and avoid misjudgment.

[0059] S102. Perform cross-validation of non-target emission source characteristics on persistent changes to obtain cross-validation results.

[0060] Cross-validation of persistent changes using non-target emission source characteristics can be understood as comparing identified persistent water parameter changes with the characteristic fingerprints or patterns of known or pre-defined non-target emission sources. For example, by analyzing the chemical composition, physical properties, and time-series patterns of persistent changes, they can be matched with historical data or models of non-target emission sources (such as agricultural non-point source pollution, urban stormwater runoff, and routine emissions from small industrial wastewater treatment plants). The aim is to confirm whether persistent changes match the characteristics of non-target emission sources, thereby excluding or confirming that they are caused by non-target emission sources.

[0061] S103. Based on the cross-validation results and correlation results, determine the potential emission risks.

[0062] In practical applications, assessing potential emission risks based on cross-validation and correlation results means that, after confirming a correlation between persistent changes and the timing of emission activities, further cross-validation results are applied using characteristics of non-target emission sources. If the cross-validation results indicate that the persistent change is not caused by non-target emission sources, or that the contribution of non-target emission sources is low, then the assessment of potential emission risks can be determined with greater confidence. Conversely, if the cross-validation results strongly point to non-target emission sources, the assessment of potential emission risks can be reduced or eliminated.

[0063] This application's solution effectively distinguishes between genuine water pollution signals caused by upstream discharge activities and interference signals generated by non-target discharge sources by introducing steps to identify interference signals from non-target discharge sources and to cross-validate the characteristics of non-target discharge sources on persistent changes. Specifically, after initially identifying persistent changes in water parameters and correlating them with the time information of discharge activities, cross-validating these changes with the characteristics of non-target discharge sources can exclude or confirm the influence of non-target discharge sources. This avoids misjudging normal fluctuations or known interference caused by non-target discharge sources as upstream discharge risks, thereby improving the accuracy and reliability of potential discharge risk assessment.

[0064] Through the above technical solution, this application can significantly improve the accuracy of river water pollution detection and reduce the false alarm rate. Especially in the complex and ever-changing river environment, it can effectively distinguish between real pollution events and background disturbances, making subsequent adjustments to monitoring parameters and the setting of water quality anomaly judgment thresholds more targeted and effective, thereby optimizing the decision-making efficiency of water resource management and pollution prevention.

[0065] In some preferred embodiments, it is assumed that a river section is monitored for a continuous decrease in dissolved oxygen and a continuous increase in ammonia nitrogen concentration, and these changes are correlated with the production activity time information of an upstream industrial park during a specific period. Under the basic scheme, a potential emission risk might be directly identified. However, according to the optimized scheme of this application, interference signals from non-target emission sources are first identified before judging potential emission risks. For example, by analyzing historical data, it is found that the river section often experiences a decrease in dissolved oxygen at night due to large-scale algal blooms during specific seasons (such as the high-temperature period in summer), or a short-term increase in ammonia nitrogen due to upstream agricultural non-point source scouring. Next, the current monitored continuous changes in dissolved oxygen and ammonia nitrogen are cross-validated for non-target emission source characteristics. Specifically, changes in chlorophyll a concentration in the water can be analyzed to determine the intensity of algal activity, or specific pesticide residues can be analyzed to determine agricultural non-point source pollution. If the cross-validation results show that the current patterns of dissolved oxygen and ammonia nitrogen changes highly match the characteristic fingerprints of algal blooms or agricultural non-point source pollution, it can be determined that these changes are mainly caused by non-target emission sources, thereby reducing or eliminating the judgment that there is a potential emission risk from the upstream industrial park. Conversely, if the cross-validation results exclude the significant impact of non-target emission sources, it is possible to more confidently conclude that there is a potential emission risk in the upstream industrial park and adjust the downstream monitoring strategy accordingly.

[0066] In some embodiments described above in this application, persistent changes are correlated with temporal information of emission activities to obtain correlation results. However, in practice, if the physical transport process of pollutants in the river channel is not fully considered, the accuracy of this temporal correlation may be insufficient, thereby affecting the reliability of the assessment of potential emission risks.

[0067] In this regard, this application further proposes the following steps for linking the aforementioned deviations from the correlation evolution pattern with the temporal information of emission activities: S201. Obtain hydrological information for the current river section.

[0068] Obtaining hydrological information for the current river section refers to collecting real-time or near-real-time data related to river hydrodynamics, such as flow velocity, discharge, water level, channel width, and depth. This information is crucial for accurately simulating the migration and diffusion of pollutants in the water body, and its purpose is to provide basic data for subsequent material transport time calculations.

[0069] S202. Based on hydrological information, calculate the material transport time from the discharge point to the monitoring point.

[0070] Specifically, the time required for pollutants to reach downstream monitoring points from upstream emission sources can be estimated by using hydrodynamic models or empirical formulas, combined with acquired hydrological information. For example, one-dimensional or two-dimensional water quality models can be used to calculate this, taking into account the river's average flow velocity, river length, and potential diffusion effects. The aim is to accurately quantify the physical transport delay of pollutants in the river channel.

[0071] S203. Set the associated time window based on the material transport time.

[0072] This time window is typically centered on the material transport time, with a certain margin to cover the actual time range within which pollutants may arrive at the monitoring point. For example, it can be set as [transport time - deviation, transport time + deviation]. The purpose is to define a precise time period that corresponds to the physical process.

[0073] S204. Compare the occurrence time of the persistent change with the associated time window to determine whether the persistent change falls within the associated time window.

[0074] Specifically, it can be checked whether the start or significant time point of the monitored water parameters falls within a pre-set correlation time window. Only when the time of change and the discharge activity have a reasonable correspondence in terms of physical transport time are they considered to be correlated. The purpose is to eliminate interference factors due to time mismatch and improve the accuracy of the correlation.

[0075] This application's solution introduces the acquisition of current river section hydrological information and calculates the material transport time from the discharge point to the monitoring point based on this information. It then sets a precise correlation time window, enabling a more refined comparison between the occurrence time of persistent changes and the time information of discharge activities. It is precisely because the actual physical transport process of pollutants in the river channel is considered that the judgment of the time correlation between persistent changes and discharge activities becomes more scientific and accurate, effectively avoiding misjudgments or omissions caused by time misalignment.

[0076] Through the above technical solution, this application can significantly improve the accuracy and reliability of correlating persistent changes with emission activity time information. By accurately calculating the material transport time and setting the correlation time window, the interference of unrelated emission activities or natural background changes on the judgment can be effectively eliminated, making the identification of potential emission risks more accurate. This provides a more solid foundation for subsequent risk assessment and monitoring parameter adjustment, and avoids resource waste or environmental risk omissions caused by inaccurate time correlation.

[0077] In some preferred embodiments, it is assumed that there is a potential discharge source and a water quality monitoring station upstream of a certain river segment. When the monitoring station detects a continuous change in indirect water parameters, the system first acquires hydrological information such as the current flow velocity and flow rate of the river segment. For example, if the river segment is 10 kilometers long and the average flow velocity is 0.5 m / s, the calculated material transport time is approximately 5.5 hours. Based on this, the system can set a correlation time window, for example, from 5 to 6 hours after the discharge activity. Subsequently, the system compares the start time of the detected continuous change with this correlation time window. If the continuous change occurs 5.2 hours after the discharge activity, it is determined that it falls within the correlation time window, thereby enhancing the confidence of the correlation between the continuous change and the upstream discharge activity, and providing a more accurate time basis for subsequent assessment of potential discharge risks.

[0078] In some embodiments described above, this application proposes a method for determining potential emission risks based on persistent changes and correlation results. This involves identifying interference signals from non-target emission sources and performing cross-validation of non-target emission source characteristics on persistent changes. However, in practical applications, the cleaning agent composition, cleaning process, or emission characteristics of non-target emission sources are not static; they may undergo gradual changes or sudden adjustments over time. If the system relies on fixed historical characteristic fingerprints for cross-validation, when the actual characteristics of the non-target emission source change, the original characteristic fingerprints will no longer accurately reflect its current state. This may cause the system to misjudge normal changes in non-target emission sources as potential pollution risks, resulting in false alarms and affecting the accuracy and reliability of risk assessment.

[0079] In response, this application further proposes a step for cross-validating the characteristics of non-target emission sources for the aforementioned persistent changes, which includes: S301. Collect emission signals generated by non-target emission sources during operation.

[0080] Specifically, monitoring equipment can be deployed at or near the discharge outlet of a non-target emission source to acquire various emission data generated during normal operation, either in real time or periodically. These emission signals may include, but are not limited to, water quality parameters (e.g., pH, conductivity, dissolved oxygen, chemical oxygen demand, specific ion concentrations), flow rate, temperature, and other process parameters related to the emission activity. The aim is to obtain the real-time emission characteristics of the non-target emission source as a basis for subsequent analysis and comparison.

[0081] S302, Extract the characteristic fingerprint of the emission signal.

[0082] Specifically, the collected emission signals can be preprocessed and feature-engineered to extract key information that uniquely or significantly characterizes the emission characteristics of the non-target emission source. These feature fingerprints can be a set of statistics (e.g., mean, variance, kurtosis, skewness), frequency domain features (e.g., the intensity of specific frequency components obtained through Fourier transform), time domain features (e.g., autocorrelation function or waveform pattern), or abstract feature vectors learned through machine learning models. The aim is to transform the complex raw signal into a concise and discriminative digital representation for subsequent storage, comparison, and identification.

[0083] S303. Compare the characteristic fingerprint with the characteristic fingerprint of historical non-target emission sources to identify whether the cleaning agent composition, cleaning process or emission characteristics of the non-target emission sources have changed.

[0084] Specifically, the extracted feature fingerprints can be compared with pre-stored feature fingerprints representing the historical stable emission states of the non-target emission source. Various comparison techniques can be employed, such as calculating the distance metric between two fingerprints (e.g., Euclidean distance, Mahalanobis distance), assessing their similarity (e.g., cosine similarity), or performing pattern matching based on classifiers or clustering algorithms. This comparison can determine whether there are significant differences between the current emission characteristics and historical characteristics, thereby inferring whether the cleaning agent composition, cleaning process, or overall emission characteristics of the non-target emission source have changed. For example, changes in the cleaning agent composition may lead to changes in the concentration of specific chemical substances, adjustments to the cleaning process may affect the instantaneous flow rate or periodicity of emissions, and changes in emission characteristics may manifest as a combined drift of multiple parameters.

[0085] S304. In response to changes in the composition of cleaning agents, cleaning processes, or emission characteristics of non-target emission sources, update the characteristic fingerprint database of non-target emission sources.

[0086] Specifically, when the comparison results confirm a significant change in the characteristics of a non-target emission source, the system will correct or replace the corresponding fingerprint stored in the feature fingerprint database based on the feature fingerprint extracted from the new emission signal. This update process can employ various strategies, such as directly replacing the old fingerprint with the new one, or using weighted averaging, sliding window averaging, or other methods to fuse the new and old fingerprints to smoothly reflect the gradual changes in characteristics. The aim is to ensure that the feature fingerprint database always accurately reflects the latest emission characteristics of the non-target emission source, avoiding misjudgments caused by using outdated fingerprints for comparison.

[0087] S305. Use the updated feature fingerprint database of non-target emission sources to perform cross-validation of non-target emission source features for persistent changes.

[0088] Specifically, after the feature fingerprint database is updated, these latest and more accurate feature fingerprints of non-target emission sources can be used to re-verify the previously identified "persistent changes." This step aims to eliminate false alarms caused by changes in the characteristics of the non-target emission sources themselves, ensuring that only anomalous changes truly related to upstream target emission activities are identified as potential emission risks.

[0089] This application's solution effectively addresses the problem of decreased cross-validation accuracy in traditional methods when the characteristics of non-target emission sources change by introducing a mechanism for dynamically updating the non-target emission source characteristic fingerprint database. Specifically, by continuously collecting emission signals from non-target emission sources and extracting their characteristic fingerprints, the system can monitor in real time whether there are deviations in the cleaning agent composition, cleaning process, or emission characteristics of these emission sources. Once a significant change is detected, the characteristic fingerprint database is updated promptly, ensuring that the stored fingerprints always represent the latest and most accurate emission characteristics of the non-target emission sources. Therefore, when performing non-target emission source characteristic cross-validation on persistent changes identified in river water bodies, the system can use the most accurate reference data for comparison, effectively distinguishing signal fluctuations caused by changes in the non-target emission sources themselves from pollution events truly caused by upstream target emission activities. This adaptive fingerprint database management mechanism makes the cross-validation process more robust and accurate.

[0090] Through the above technical solution, this application can significantly improve the accuracy and reliability of river water pollution detection. Since the characteristics of non-target emission sources are dynamically changing, traditional fixed fingerprint databases are prone to misjudgments. This application ensures that the cross-validation benchmark remains consistent with the actual situation by monitoring and dynamically updating the feature fingerprint database of non-target emission sources in real time. This not only reduces false alarms caused by changes in the non-target emission sources themselves, avoiding unnecessary resource waste and erroneous decisions, but also enables the system to more accurately identify genuine potential emission risks, thus providing a more reliable basis for timely countermeasures.

[0091] In some preferred embodiments, a specific example is given below. Suppose there is an industrial park upstream of a certain river section, which contains a factory A that regularly cleans its equipment, and its emissions are considered a non-target emission source. Initially, the system collects the emission signals of factory A during normal cleaning and extracts its characteristic fingerprints, such as the concentration curves of specific organic substances in the cleaning wastewater, pH fluctuation ranges, etc., and stores them in a non-target emission source characteristic fingerprint database.

[0092] After a period of time, in order to improve cleaning efficiency, Factory A switched to a new type of cleaning agent. This new agent had a different chemical composition than the previous one, causing slight but continuous changes in certain water quality parameters of its wastewater discharge (e.g., conductivity or the concentration of a specific ion). The detection method of this application continuously collects the discharge signals from Factory A and extracts new characteristic fingerprints. When the new characteristic fingerprints are compared with historical fingerprints, the system identifies persistent differences in indicators such as conductivity or the concentration of a specific ion that exceed normal fluctuation ranges, thus determining that the composition of Factory A's cleaning agent has changed.

[0093] In response to this change, the system automatically updates the fingerprint representation of factory A in the non-target emission source feature fingerprint database. For example, a weighted average method can be used to fuse the new feature fingerprint with the old fingerprint to form a new fingerprint that reflects the current emission characteristics. Subsequently, when river monitoring points detect continuous changes in water parameters, the system will use the updated fingerprint database reflecting the latest emission characteristics of factory A for comparison when performing cross-validation of non-target emission source features. In this way, even if the emission characteristics of factory A change, the system can accurately identify its emission signal as normal fluctuations of a non-target emission source, avoiding misjudgment as a pollution event caused by upstream target emission activities, thereby improving the accuracy of potential emission risk assessment.

[0094] In some embodiments described above, a scheme is proposed to compensate timestamps using a local time compensation factor for data points within the transient variation range of the internal timer device. However, if the local time compensation factor fails to adequately reflect the dynamic characteristics of the transient variation, such as highly irregular, nonlinear, and transient fluctuations in clock drift, it may lead to insufficient compensation accuracy and affect the accuracy of data alignment. Therefore, this application further proposes a method for dynamically calculating the local time compensation factor to more accurately process data within the transient variation range.

[0095] In some embodiments described above, this application proposes updating the feature fingerprint database of non-target emission sources. However, during implementation, if the cleaning agent composition, cleaning process, or emission characteristics of the non-target emission source undergo gradual and slow changes, traditional direct update mechanisms may fail to effectively capture these subtle evolutions, or may be overly sensitive, leading to frequent fluctuations in the fingerprint database, thereby affecting the accuracy and stability of non-target emission source identification. If these problems are not addressed, the feature fingerprint database may not accurately reflect the true state of the non-target emission source, thus affecting the reliability of potential emission risk assessment.

[0096] In response to changes in the composition of cleaning agents, cleaning processes, or emission characteristics of non-target emission sources, this application further proposes steps for updating the characteristic fingerprint database of non-target emission sources, including: S401. Calculate the statistical characteristics and trends of the current emission signal.

[0097] Specifically, data processing can be performed on emission signals from non-target emission sources monitored in real time to extract key statistics, such as average, variance, peak value, intensity of specific frequency components, and the rate or direction of change of these statistics over time. These characteristics can comprehensively reflect the instantaneous state and dynamic evolution of the current emission signal.

[0098] S402. Compare the statistical characteristics and trends of the current emission signal with the most recent stable fingerprint stored in the feature fingerprint database.

[0099] The statistical characteristics and trends of the current emission signal are compared with the most recent stable fingerprint stored in the feature fingerprint database. The purpose of this comparison is to assess the degree of deviation between the current emission signal and the known stable state. The most recent stable fingerprint represents the typical characteristics of the non-target emission source under normal operating conditions.

[0100] S403. When the comparison results show that there is a persistent difference between the statistical characteristics or trend of the current emission signal and the most recent stable fingerprint that exceeds the preset slight deviation range, it is determined that the characteristics of the non-target emission source have undergone a gradual change.

[0101] "Persistence" means that the difference is not accidental or instantaneous, but has been maintained for a period of time or appears in multiple consecutive sampling periods; "Preset weak deviation range" is used to filter out normal random fluctuations or measurement errors, ensuring that only significant and stable changes are identified as gradual changes.

[0102] S404. In response to gradual changes, the fingerprint representation in the feature fingerprint database is corrected by weighted averaging based on the statistical characteristics and trends of the current emission signal.

[0103] The weighted average correction method allows new emission signal features to be incorporated into the existing fingerprint representation with certain weights, thereby achieving smooth fingerprint updates, avoiding drastic changes in the fingerprint due to single data fluctuations, and gradually reflecting the true evolution of emission sources.

[0104] S405. Verify the consistency between the corrected fingerprint and the historical fingerprint version.

[0105] Specifically, the consistency between the corrected fingerprint and the historical fingerprint version can be verified based on the following steps: 1. Extract multiple dimensional features of the current fingerprint from the emission signal.

[0106] Among them, multiple dimensional features include the instantaneous fluctuation amplitude of the signal, the duration, the intensity of specific frequency components, and the rate of change of these features over time; Among these features, the instantaneous fluctuation amplitude of the signal refers to the maximum range of change of the signal within a short period of time, reflecting the instantaneous intensity of the emission; the duration refers to the length of time the signal exceeds a certain threshold, characterizing the persistence of the emission event; the intensity of specific frequency components can be obtained through signal processing techniques such as Fourier transform, used to identify periodic or specific emission characteristics; and the rate of change of these features over time reflects the dynamic evolution trend of the emission process. By extracting these multi-dimensional features, a richer and more accurate fingerprint representation can be constructed, thereby more accurately describing the characteristics of non-target emission sources.

[0107] 2. Calculate the difference in multiple dimensions of features between the current fingerprint and each historical fingerprint version to identify the target historical fingerprint version with the highest similarity to the current fingerprint.

[0108] The degree of dissimilarity can be calculated using various mathematical methods, such as Euclidean distance, Mahalanobis distance, cosine similarity, or dynamic time warping (DTW). By calculating the degree of dissimilarity between the current fingerprint and every historical fingerprint version stored in the feature fingerprint database, the similarity or dissimilarity between them can be quantified. Identifying the target historical fingerprint version with the highest similarity to the current fingerprint aims to find the nearest "reference point" on the historical evolution path of the current fingerprint, so as to subsequently determine whether the current change belongs to normal evolution or abnormal drift.

[0109] 3. Determine whether the difference between the current fingerprint and the target historical fingerprint version is within the preset normal evolution range.

[0110] The normal evolution range refers to the permissible degree of difference between the characteristic fingerprint and historical fingerprint versions when a non-target emission source is operating normally or undergoing known and predictable gradual changes. This range can be preset and dynamically adjusted through historical data analysis, statistical modeling (e.g., the 3σ principle based on Gaussian distribution), or machine learning methods (e.g., anomaly detection models). Determining whether the difference is within the normal evolution range is a key step in distinguishing between normal evolution and anomalous drift.

[0111] When the difference is within the normal evolution range, the current fingerprint is marked as normal evolution, and the current fingerprint marked as normal evolution is added to the historical fingerprint version library to update the evolution trajectory of historical fingerprints. When the difference exceeds the normal evolution range, the current fingerprint is marked as an abnormal drift and an alarm is triggered.

[0112] Understandably, when the difference exceeds the normal evolution range, it indicates that the current fingerprint change may not originate from the normal gradual evolution of non-target emission sources, but rather from anomalies such as sensor malfunction, data acquisition errors, sudden changes in non-target emission sources, or the presence of new, unknown interference sources. Marking this as an abnormal drift and triggering an alarm aims to promptly alert operators or system administrators to the anomaly, enabling manual verification, equipment maintenance, or further analysis. This prevents inaccurate fingerprint data from being included in the feature fingerprint database, thereby maintaining the reliability of the entire detection system.

[0113] In some preferred embodiments, a specific example is given below. Suppose there is a non-target emission source upstream of a certain river section, such as a wastewater treatment plant in an industrial park, which performs equipment cleaning at specific times, generating emission signals with specific characteristics. To verify the consistency of the characteristic fingerprint of this non-target emission source, firstly, multiple dimensions of the current fingerprint are extracted from the wastewater treatment plant's emission signal. For example, these may include the cleaning agent composition (obtained through spectral analysis), the instantaneous fluctuation range of the emission flow rate, the duration of the cleaning process, and the rate of change of the concentration of specific pollutants (such as chemical oxygen demand and ammonia nitrogen) over time.

[0114] Next, the dissimilarity of these multi-dimensional features is calculated against each historical fingerprint version stored in the feature fingerprint database. For example, weighted Euclidean distance can be used to quantify the difference between the current fingerprint and historical fingerprints. By comparing the dissimilarity scores, the historical fingerprint version most similar to the current fingerprint is identified and selected as the target historical fingerprint version.

[0115] Then, it is determined whether the difference between the current fingerprint and the target historical fingerprint version is within a preset normal evolution range. For example, a statistical threshold can be set: if the change in certain dimensions of the current fingerprint (such as cleaning agent concentration or duration) compared with the target historical fingerprint version exceeds three standard deviations derived from historical data statistics, it is considered to have exceeded the normal evolution range.

[0116] If the difference is within the normal evolution range—for example, a slight adjustment in the cleaning agent composition but no change in the overall emission pattern—the current fingerprint is marked as normal evolution and added to the historical fingerprint version library to update the evolution trajectory of the wastewater treatment plant's cleaning emissions. This helps the system learn and adapt to normal process improvements or seasonal changes from non-target emission sources.

[0117] Conversely, if the difference exceeds the normal range of evolution, such as the detection of entirely new cleaning agent ingredients, a sudden and significant increase in emission flow with an abnormally prolonged duration, or drastic fluctuations in the concentration of a specific pollutant, the current fingerprint will be marked as an abnormal drift, and an alarm will be triggered immediately.

[0118] S406. When the corrected fingerprint passes the consistency verification, the corrected fingerprint is marked as a new stable fingerprint, and the feature fingerprint database is updated.

[0119] This application's solution effectively addresses the limitations of traditional update mechanisms in handling the slow evolution of non-target emission sources by introducing a mechanism for identifying gradual changes in the characteristics of non-target emission sources and a weighted average fingerprint correction method. Specifically, by continuously monitoring and calculating the statistical characteristics and trends of the current emission signal and comparing it with the most recent stable fingerprint, the system can accurately capture persistent differences exceeding the normal fluctuation range, thereby determining that a gradual change is occurring in the non-target emission source. Once this gradual change is identified, the system does not simply replace the old fingerprint but uses a weighted average method to smoothly correct it. This allows the fingerprint database to gradually adapt to the actual evolution of the emission source, avoiding misjudgments caused by instantaneous fluctuations. Furthermore, by verifying the consistency of the corrected fingerprint with historical fingerprint versions, the accuracy and reliability of fingerprint updates are further ensured, preventing abnormal data from interfering with the fingerprint database, thus maintaining the long-term stability and effectiveness of the non-target emission source characteristic fingerprint database.

[0120] Through the above technical solution, this application can significantly improve the adaptability and robustness of the non-target emission source characteristic fingerprint database. This solution can effectively identify and smoothly handle the gradual changes in non-target emission sources, avoiding fingerprint distortion or frequent false alarms that may result from simple update mechanisms. Therefore, the characteristic fingerprint database can more accurately reflect the true characteristics of non-target emission sources, thereby improving the accuracy of non-target emission source identification, reducing the false alarm rate, and enabling river water pollution detection systems to more reliably distinguish between target pollution events and the normal evolution of non-target emission sources. This, in turn, improves the accuracy of potential emission risk assessment and the effectiveness of downstream monitoring parameter adjustments.

[0121] In some preferred embodiments, it is assumed that a non-target emission source exists within an upstream industrial park, such as a factory that regularly cleans its equipment. Over a period of time, to improve cleaning efficiency, this factory gradually adjusts the formulation of its cleaning agents, resulting in slow and continuous changes in the concentration and emission pattern of certain trace components in its wastewater. Traditional fingerprint update mechanisms may fail to identify these changes in a timely manner because they are too subtle, or may ignore them because the changes do not reach a certain hard threshold, thus failing to update the feature fingerprint database in a timely manner and affecting the accuracy of subsequent cross-validation. The solution of this application can effectively address this situation. Specifically, the system continuously calculates the statistical characteristics (e.g., average concentration and fluctuation range of specific chemical substances) and trends of the factory's emission signals. When these characteristics are compared with the most recent stable fingerprint stored in the feature fingerprint database, even if the difference in a single comparison is small, if this difference persists and accumulates beyond a preset weak deviation range, the system will determine that the characteristics of the non-target emission source are undergoing a gradual change. In response to this gradual change, the system will not immediately replace the old fingerprints. Instead, it will adjust the corresponding fingerprint representation in the feature fingerprint database using a weighted average based on the statistical characteristics and trends of the current emission signals. For example, new emission features may be incorporated into the old fingerprint with a weight of 0.1, while the old fingerprint is retained with a weight of 0.9, thus achieving a smooth transition. Subsequently, the adjusted fingerprint will be verified for consistency with the plant's historical fingerprint versions to ensure that the adjusted fingerprint still conforms to the plant's emission patterns rather than exhibiting abnormal drift. Once verified, the adjusted fingerprint will be marked as a new stable fingerprint and updated in the feature fingerprint database. In this way, the feature fingerprint database can dynamically and smoothly adapt to the real evolution of non-target emission sources, ensuring the accuracy of non-target emission source identification and the overall robustness of the system.

[0122] Specifically, the above methods also include: S501. Calculate the similarity between the current fingerprint and each historical fingerprint version.

[0123] Specifically, calculating the similarity between the current fingerprint and each historical fingerprint version means that the system quantitatively compares the feature fingerprint extracted from the current emission signal with each historical fingerprint version stored in the feature fingerprint database to determine the degree of similarity between them. This similarity calculation aims to provide a quantitative basis for subsequently selecting the best-matching historical fingerprint version, ensuring that the selected historical version can truly reflect the evolution trend of the current fingerprint.

[0124] Similarity can be calculated using various mathematical methods, such as the reciprocal of Euclidean distance, cosine similarity, and Pearson correlation coefficient. The specific method chosen can be adjusted based on the type of fingerprint features and the needs of the actual application scenario.

[0125] Specifically, for any historical fingerprint version, the similarity between the current fingerprint and the historical fingerprint version can be calculated using the similarity formula; The similarity formula is: Match_Score=1 / (1+Sum((Actual_C_i-Predicted_C_i) 2 )); Where Match_Score is the similarity score; Actual_C_i is the concentration of the i-th detected indicator; and Predicted_C_i is the concentration of the i-th indicator in the historical fingerprint version.

[0126] S502. When there is a historical fingerprint version in the historical fingerprint version library that has a similarity higher than the preset similarity threshold with the current fingerprint, the historical fingerprint version with the highest similarity with the current fingerprint will be used as the target historical fingerprint version.

[0127] The proposed solution first calculates the similarity between the current fingerprint and all historical fingerprint versions, then filters based on a preset similarity threshold, and finally selects the historical fingerprint version with the highest similarity as the target. This effectively solves the problem of accurately selecting the most relevant reference benchmark from multiple historical versions during consistency verification. This method avoids the inefficiency of randomly selecting or traversing all historical versions for comparison, ensuring the relevance and accuracy of subsequent consistency verification, thereby improving the reliability of identifying changes in the characteristics of non-target emission sources.

[0128] The above technical solution ensures that the selected historical fingerprint version is the most representative and relevant when verifying the consistency between the corrected fingerprint and the historical fingerprint version, thus significantly improving the accuracy and efficiency of the verification. This solution avoids misjudgments or omissions caused by selecting inappropriate historical versions, making the identification of gradual changes in the characteristics of non-target emission sources more accurate, thereby improving the robustness and reliability of the entire river water pollution detection method.

[0129] like Figure 3 As shown in the figure, this embodiment of the invention also provides a river water pollution detection system. The system includes: The acquisition module is used to acquire indirect water parameters related to upstream discharge activities; The analysis module is used to analyze short-term variation patterns of indirect water parameters; The comparison module is used to compare short-term change patterns with preset historical normal patterns to identify persistent changes where there are differences between short-term change patterns and historical normal patterns. The correlation module is used to correlate persistent changes with time information of emission activities to obtain correlation results; The judgment module is used to assess potential emission risks based on continuous changes and correlation results; The processing module adjusts downstream monitoring parameters and water quality anomaly detection thresholds in response to potential emission risks.

[0130] This application also provides a computer-readable storage medium. All or part of the processes in the above method embodiments can be executed by a computer program instructing related hardware. This program can be stored in the computer-readable storage medium, and when executed, it can include the processes of the above method embodiments. The computer-readable storage medium can be an internal storage unit of the task execution device (including a data sending end and / or a data receiving end) of any of the foregoing embodiments, such as the hard disk or memory of the task execution device. The computer-readable storage medium can also be an external storage device of the terminal device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the terminal device. Further, the computer-readable storage medium can include both the internal storage unit of the task execution device and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the task execution device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0131] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0132] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0133] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be covered within the scope of protection of this application.

Claims

1. A method for detecting river water pollution, characterized in that, include: Obtain indirect water parameters related to upstream discharge activities; Analyze the short-term variation patterns of the aforementioned indirect water parameters; The short-term change pattern is compared with a preset historical normal pattern to identify persistent changes where the short-term change pattern differs from the historical normal pattern. The persistent changes are correlated with the time information of emission activities to obtain the correlation results; Based on the persistent changes and the correlation results, potential emission risks are assessed; In response to the potential emission risks, downstream monitoring parameters and water quality anomaly detection thresholds are adjusted.

2. The method for detecting river water pollution according to claim 1, characterized in that, The step of determining potential emission risks based on the persistent changes and the correlation results includes: Identify interference signals from non-target emission sources; Cross-validation of non-target emission source characteristics is performed on the persistent changes to obtain cross-validation results; Based on the cross-validation results and the correlation results, potential emission risks are determined.

3. The method for detecting river water pollution according to claim 2, characterized in that, The step of associating the persistent changes with time information of emission activities includes: Obtain hydrological information for the current river section; Based on the hydrological information, calculate the material transport time from the discharge point to the monitoring point; Based on the material transport time, a related time window is set; The occurrence time of the persistent change is compared with the associated time window to determine whether the persistent change falls within the associated time window.

4. The method for detecting river water pollution according to claim 2, characterized in that, The cross-validation of non-target emission source characteristics for the persistent changes includes: Collect emission signals generated by non-target emission sources during operation; Extract the characteristic fingerprint of the emission signal; The characteristic fingerprint is compared with the characteristic fingerprint of historical non-target emission sources to identify whether the cleaning agent composition, cleaning process or emission characteristics of the non-target emission sources have changed. In response to changes in the cleaning agent composition, cleaning process, or emission characteristics of the non-target emission source, the feature fingerprint database of the non-target emission source is updated; The persistent changes are cross-validated using non-target emission source features through an updated feature fingerprint database of non-target emission sources.

5. The method for detecting river water pollution according to claim 4, characterized in that, The step of updating the feature fingerprint database of non-target emission sources in response to changes in the cleaning agent composition, cleaning process, or emission characteristics of the non-target emission source includes: Calculate the statistical characteristics and trends of the current emission signal; The statistical characteristics and trends of the current emission signal are compared with the most recent stable fingerprint stored in the feature fingerprint database; When the comparison results show that there is a persistent difference between the statistical characteristics or trend of the current emission signal and the most recent stable fingerprint that exceeds the preset slight deviation range, it is determined that the characteristics of the non-target emission source have undergone a gradual change. In response to the gradual change, the fingerprint representation in the feature fingerprint database is corrected by weighted averaging based on the statistical characteristics and changing trends of the current emission signal. Verify the consistency between the corrected fingerprint and the historical fingerprint version; When the corrected fingerprint passes the consistency verification, the corrected fingerprint is marked as a new stable fingerprint, and the feature fingerprint database is updated.

6. The method for detecting river water pollution according to claim 5, characterized in that, The process of verifying the consistency between the corrected fingerprint and historical fingerprint versions includes: Multiple dimensional features of the current fingerprint are extracted from the emission signal, including the instantaneous fluctuation amplitude of the signal, the duration, the intensity of specific frequency components, and the rate of change of these features over time. Calculate the difference in multiple dimensions of features between the current fingerprint and each historical fingerprint version to identify the target historical fingerprint version with the highest similarity to the current fingerprint; Determine whether the difference between the current fingerprint and the target historical fingerprint version is within a preset normal evolution range; When the difference is within the normal evolution range, the current fingerprint is marked as normal evolution, and the current fingerprint marked as normal evolution is included in the historical fingerprint version library to update the evolution trajectory of the historical fingerprint. When the difference exceeds the normal evolution range, the current fingerprint is marked as an abnormal drift, and an alarm is triggered.

7. The method for detecting river water pollution according to claim 6, characterized in that, Determining the target historical fingerprint version includes: Calculate the similarity between the current fingerprint and each historical fingerprint version; When a historical fingerprint version exists in the historical fingerprint version library with a similarity higher than a preset similarity threshold to the current fingerprint, the historical fingerprint version with the highest similarity to the current fingerprint is selected as the target historical fingerprint version.

8. The method for detecting river water pollution according to claim 7, characterized in that, The calculation of the similarity between the current fingerprint and each historical fingerprint version includes: For any historical fingerprint version, the similarity between the current fingerprint and the historical fingerprint version is calculated according to the similarity formula; The similarity formula is: Match_Score=1 / (1+Sum((Actual_C_i-Predicted_C_i) 2 )); Where Match_Score is the similarity score; Actual_C_i is the concentration of the i-th detected indicator; and Predicted_C_i is the concentration of the i-th indicator in the historical fingerprint version.

9. The method for detecting river water pollution according to claim 1, characterized in that, The adjustment of downstream monitoring parameters and water quality anomaly detection thresholds in response to the potential emission risk includes: Increase the monitoring frequency of dissolved oxygen, chemical oxygen demand, pH value, and specific heavy metals in the downstream area, and lower the threshold for judging water quality anomalies.

10. A river water pollution detection system, characterized in that, The system includes: The acquisition module is used to acquire indirect water parameters related to upstream discharge activities; The analysis module is used to analyze the short-term variation patterns of the indirect water body parameters; The comparison module is used to compare the short-term change pattern with a preset historical normal pattern to identify persistent changes in which the short-term change pattern differs from the historical normal pattern. The correlation module is used to correlate the persistent changes with the time information of emission activities to obtain the correlation results; The judgment module is used to determine potential emission risks based on the persistent changes and the correlation results; The processing module adjusts downstream monitoring parameters and water quality anomaly detection thresholds in response to the potential emission risks.

Images