A sewage treatment system data analysis method and system

Abstract

The application provides a sewage treatment system data analysis method and system, relates to the sewage treatment data monitoring and analysis technical field, and has the technical scheme points that: by acquiring monitoring information and flow information of an inlet end and an outlet end, identifying an inlet load change event from the inlet end monitoring information, and calculating a delay time of the change transmission to the outlet end according to the flow information, the real response event is positioned in the outlet end monitoring information based on the delay time, and correlation analysis is carried out, so that the matching of front-end and rear-end data fully considers the physical delay caused by the hydraulic retention time, avoids the time sequence misplacement, and the analysis accuracy is improved, has the advantages that the physical delay of the inlet load change transmission to the outlet end can be accurately identified and compensated, and the accuracy and reliability of the correlation analysis of the front-end and rear-end monitoring data are effectively improved.

Description

Technical Field

[0001] This application relates to the field of wastewater treatment data monitoring and analysis technology, and more specifically, to a wastewater treatment system data analysis method and system. Background Technology

[0002] In the operation and management of wastewater treatment systems, it is typically necessary to perform correlation analysis on influent online monitoring data, effluent online monitoring data, influent flow data, and related process operation data to assess pollutant removal efficiency, the impact of load shocks, and the causes of effluent anomalies. Since wastewater enters the system from the influent end and undergoes multiple treatment stages such as biological treatment and sedimentation, changes in influent water quality or load do not immediately reflect at the effluent end. Instead, they require a physical transfer process related to the system's hydraulic retention time. Therefore, determining the true time correspondence between changes in influent load and effluent response is a key factor affecting the accuracy of the analysis when performing front-end and back-end data analysis.

[0003] In conventional analysis, to simplify the pairing process of front-end and back-end data, static alignment methods such as fixed delays, empirical retention times, or the same statistical window are typically used. For example, data from the influent end at a certain time period can be compared with effluent data shifted by a fixed time interval, based on the design hydraulic retention time of the wastewater treatment plant. Alternatively, at coarser time scales such as shift statistics or daily statistics, influent and effluent data within the same statistical period can be used as corresponding data for removal rate calculation and anomaly analysis. These methods can reflect the overall treatment status of the system to a certain extent under relatively stable influent flow and pollution load conditions.

[0004] However, in real-world operating scenarios such as rainy season water inflow, sudden changes in pipeline inflow, or fluctuations in upstream discharge, the influent load may exhibit rapid or multi-stage fluctuations. After a load fluctuation occurs at the influent end, this fluctuation needs to pass through treatment stages such as biological treatment units before it is transmitted to the effluent end, and the actual hydraulic retention time varies with the influent flow rate. Therefore, the true response time window at the effluent end depends on the actual flow state and the hydraulic transmission process within the system at that time.

[0005] In this situation, if a fixed delay, empirical window, or the same statistical window is still used for front-end and back-end matching, the estimated static alignment window may deviate from the actual response window, resulting in the online monitoring data of the effluent used for correlation analysis not being able to truly reflect the back-end response to the corresponding changes in influent load.

[0006] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0007] The purpose of this application is to provide a data analysis method and system for a wastewater treatment system, which can accurately identify and compensate for the physical delay in the transmission of changes in influent load to the effluent end, and effectively improve the accuracy and reliability of the correlation analysis of front-end and back-end monitoring data.

[0008] This application provides a data analysis method for a wastewater treatment system, including: Acquire first monitoring information to characterize the change of water quality at the inlet of the wastewater treatment system over time, second monitoring information to characterize the change of water quality at the outlet of the wastewater treatment system over time, and flow information to characterize the change of influent flow rate over time. Identify target change events that characterize changes in influent load from the primary monitoring information; Based on the flow information at the time the target change event occurs, calculate the delay information used to characterize the time span required for the target change event to propagate to the outlet. Based on the delay information, the target response information used to characterize the actual backend response corresponding to the target change event is located in the second monitoring information, and the target change event and the target response information are correlated and analyzed.

[0009] Furthermore, after identifying target change events from the initial monitoring information, the process also includes: Determine the internal change stage of target change events; When a target change event is determined to contain multiple valid and relatively independent influent load change phases, the target change event is broken down into multiple sub-change events; Each sub-change event is treated as an independent object for calculating the corresponding delay information and locating the corresponding target response information.

[0010] Furthermore, the determination of internal change stages and the breakdown of target change events are carried out in the following ways: Determine the internal change stage of the target change event, including: Based on the sequence of pollutant indicators in the first monitoring information corresponding to the target change event, at least one local change segment is determined. The local change segment is a continuous period in the sequence where the change direction is consistent and the change amplitude exceeds the first threshold. Adjacent local change segments with the same direction of change and a time interval less than the second threshold are merged. When multiple local change segments still exist after merging, the target change event is determined to contain multiple effective and relatively independent influent load change stages; When a target change event is determined to contain multiple valid and relatively independent influent load change phases, the target change event is broken down into multiple sub-change events, including: The time interval of the target change event is divided into multiple sub-change events based on the merged local change segments. A representative timestamp and a representative inflow rate are determined for each sub-change event. The representative timestamp is the moment when the cumulative change in the corresponding time period of the sub-change event reaches half of the total change in that segment, and the representative inflow rate is the instantaneous flow rate at the representative timestamp in the flow information.

[0011] Further, the corresponding delay information is calculated, including: Delay information is determined based on the pre-established mapping relationship between hydraulic residence time and flow rate, as well as the flow rate information at the time of sub-change events.

[0012] Furthermore, target response information is located for sub-change events, including: Based on the delay information, candidate response time periods are determined from the second monitoring information; Based on the consistency between the changing trend of the second monitoring information and the changing trend of the sub-changing events within the candidate response period, the target response information is determined from the candidate response period.

[0013] Furthermore, after locating the target response information, it also includes: Based on the physical process of transmitting changes in influent load to the effluent monitoring point and establishing a stable response, the time interval corresponding to the target response information is divided into a first time region and a second time region along the time axis. The first time region is later than the second time region in time sequence. Monitoring data from the first time zone is used for the first type of correlation analysis, and monitoring data from the second time zone is used for the second type of correlation analysis. The first type of correlation analysis includes the calculation of pollutant removal efficiency, and the first type of correlation analysis does not use monitoring data from the second time zone.

[0014] Furthermore, it also includes: In response to the detection of abnormal information indicating water quality anomalies at the outlet, the system determines the historical time period corresponding to the abnormal information based on the delay information, searches for identified target change events within the historical time period, and performs anomaly attribution analysis based on the characteristics of the found target change events.

[0015] Furthermore, anomaly attribution analysis is performed based on the characteristics of the identified target change events, including: Based on the characteristics of the identified target change events, determine whether there is an abnormal influent load; The pollutant removal efficiency is calculated based on the influent pollutant index corresponding to the target change event and the effluent pollutant index corresponding to the target response information associated with it within a historical time period. When there is an abnormal influent load and the pollutant removal efficiency is lower than the preset reference removal efficiency, the root cause of the abnormal information is determined to be the decrease in treatment efficiency caused by the influent load shock. When there is an abnormal influent load and the pollutant removal efficiency is not lower than the preset reference removal efficiency, an attribution result is generated to indicate that the abnormal information is related to the abnormal influent load. When there is no abnormal influent load and the pollutant removal efficiency is lower than the preset reference removal efficiency, the root cause of the abnormal information is determined to be an internal process abnormality.

[0016] Furthermore, it also includes: Obtain matching result statistics, which are used to characterize the degree of matching failure in front-end and back-end matching based on the delay information determined by the mapping relationship and multiple sub-change events; When the statistical information of the matching result meets the reference offset judgment condition, a parameter calibration instruction is generated. The parameter calibration instruction is used to indicate that the mapping relationship is calibrated.

[0017] A wastewater treatment system data analysis system, used to execute the aforementioned wastewater treatment system data analysis method, includes: The acquisition module is used to acquire first monitoring information to characterize the change of water quality at the inlet of the sewage treatment system over time, second monitoring information to characterize the change of water quality at the outlet of the sewage treatment system over time, and flow information to characterize the change of influent flow rate over time. The identification module is used to identify target change events that characterize changes in influent load from the first monitoring information; The calculation module is used to calculate the delay information, which characterizes the time span required for the target change event to be transmitted to the outlet, based on the flow information when the target change event occurs. The location association module is used to locate the target response information, which is used to characterize the actual backend response corresponding to the target change event, in the second monitoring information based on the delay information, and to perform correlation analysis between the target change event and the target response information.

[0018] As can be seen from the above, the wastewater treatment system data analysis method and system provided in this application acquires monitoring information and flow information from the inlet and outlet ends, identifies influent load change events from the inlet end monitoring information, calculates the delay time for the change to be transmitted to the outlet end based on the flow information, and then locates the actual response event in the outlet end monitoring information based on the delay time and performs correlation analysis. This ensures that the matching of front-end and back-end data fully considers the physical delay caused by hydraulic retention time, avoids time sequence misalignment, and thus improves the accuracy of analysis. It has the advantages of accurately identifying and compensating for the physical delay in the transmission of influent load changes to the outlet end, and effectively improving the accuracy and reliability of correlation analysis of front-end and back-end monitoring data. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the wastewater treatment system data analysis method proposed in this application.

[0020] Figure 2 This is a schematic diagram of the overall process of the wastewater treatment system data analysis method provided in the embodiments of this application.

[0021] Figure 3 This is a schematic diagram illustrating the time axis matching and verification of influent load change events and effluent response information in this application.

[0022] Figure 4 This is a schematic diagram illustrating the reverse time extrapolation used to determine historical time periods during anomaly attribution analysis in this application. Detailed Implementation

[0023] The following description, in conjunction with the technical solution of this application, provides a clearer and more complete explanation of the relevant content. It should be noted that the embodiments described herein are only a part of the implementation methods of this application, and not all of them. Other implementation methods obtained by those skilled in the art based on the embodiments of this application without creative effort should also fall within the protection scope of this application. Furthermore, in the description of this application, the terms "first," "second," etc., are mainly for distinction and should not be construed as indicating relative importance.

[0024] Reference Figure 1 This application proposes a data analysis method for wastewater treatment systems, the method comprising: S1. Obtain first monitoring information to characterize the change of water quality at the inlet of the sewage treatment system over time, second monitoring information to characterize the change of water quality at the outlet of the sewage treatment system over time, and flow information to characterize the change of influent flow rate over time.

[0025] S2. Identify target change events from the first monitoring information to characterize changes in influent load.

[0026] S3. Based on the flow information at the time the target change event occurs, calculate the delay information used to characterize the time span required for the target change event to be transmitted to the outlet.

[0027] S4. Based on the delay information, locate the target response information in the second monitoring information to characterize the actual backend response corresponding to the target change event, and perform correlation analysis between the target change event and the target response information.

[0028] In this application, three basic information sources are first obtained. The first monitoring information comes from the online monitoring point of the influent, which includes continuous time series data of major pollutant indicators such as chemical oxygen demand and ammonia nitrogen. Each data record corresponds to a collection timestamp and the indicator value at that moment, which can continuously reflect the real change process of the influent water quality over time.

[0029] The second monitoring information comes from the online monitoring points at the effluent outlet, which also includes continuous time series data of the corresponding indicators, reflecting the changes in effluent water quality after the entire treatment process.

[0030] The flow information comes from the influent flow meter, which records the instantaneous influent flow value at each collection moment in the form of a continuous time series.

[0031] These three types of data are organized according to the process sequence and a unified timeline, forming the data foundation for all subsequent analysis steps. Flow information is included as a basic input because the hydraulic residence time under variable flow conditions is essentially determined by the influent flow rate. Without flow information, it is impossible to establish a dynamic correspondence between changes in influent load and their transmission delay. This is precisely what distinguishes this application from conventional processing methods that directly rely on timestamps.

[0032] After acquiring the above data, the events that truly represent changes in the influent load, i.e., target change events, will be identified from the continuous changes in influent water quality recorded in the first monitoring information. The basic idea for identification is to analyze the influent time series segment by segment, calculate the rate of change of pollutant concentration or influent load in each time period, and mark the time periods where the rate of change exceeds the normal fluctuation range as the time periods when the influent load changes.

[0033] The rate of change can be calculated by dividing the difference between the end and beginning values ​​of a time period by the length of that time period. The threshold can be determined based on the statistical distribution of the wastewater treatment plant's historical operating data. For example, during periods of stable influent flow, the rate of change of pollution indicators typically remains at a low level. However, during periods of rapid load increase caused by rainy season water flow, the rate of change will significantly exceed this level. Thus, the starting and ending windows of load ramp-up can be extracted from the continuous time series as target change events.

[0034] A target change event can be a complete load increase process, such as a rainfall event causing the influent chemical oxygen demand (COD) to rise rapidly from a low concentration to a high concentration and remain at a high level; or it can be a load decrease process, such as a rainfall event causing the influent concentration to gradually return to normal levels. Whether it is an increase or a decrease, as long as the magnitude and rate of change indicate that the influent load state has undergone a substantial change, it will be treated as an independent event object in subsequent treatment.

[0035] For each identified target change event, the actual time span required for the event to travel from the influent monitoring point to the effluent monitoring point needs to be calculated, i.e., the delay information. The calculation of the delay information is based on the influent flow rate at the time the target change event occurs.

[0036] Specifically, a mapping relationship between hydraulic retention time and influent flow rate can be established first, based on the wastewater treatment plant's design data and previous calibration operation records. This mapping relationship can take the form of a correspondence table, which provides the average hydraulic retention time obtained from actual measurements or calculations under different influent flow rate ranges.

[0037] When it is necessary to estimate the delay of a target change event, the instantaneous influent flow rate during the event is retrieved. The effective hydraulic residence time from the influent to the effluent under that flow rate condition is obtained through table lookup and linear interpolation. This effective residence time is used as the delay information corresponding to the target change event. The delay information reflects approximately how long it takes for a load change at the influent to be transmitted to the effluent as a real process load and gradually reflected in the effluent indicators under the current influent flow rate. This delay information is dynamic and changes with flow rate, rather than being a fixed empirical constant. Therefore, compared to conventional analysis methods that assume a constant residence time, this embodiment can more realistically track the differences in the rate of load transmission under different flow rate conditions.

[0038] After obtaining the delay information, the target response information can be located in the second monitoring information to find the data that truly reflects the backend response of the target change event.

[0039] Reference Figure 3 The time reference for the positioning process is the time position of the target change event in the water inlet time series. Based on this time position, the time length indicated by the delay information is pushed forward to find a basically aligned data window in the water outlet time series as a candidate response interval.

[0040] Considering the inherent uncertainty in delay estimation and the influence of axial diffusion and advection velocity distribution during actual transmission, an outward extension can be made on the time axis based on the candidate response interval by expanding it forward and backward. The expansion range can be a certain proportion of the delay information combined with the minimum expansion duration constraint, thereby ensuring that the real backend response data is included in this expanded interval.

[0041] Subsequently, the changing trends of effluent water quality indicators were examined within the expanded interval. If the target change event is a load increase event at the influent end, then the effluent pollutant concentration should show a corresponding upward trend within the effluent data window that actually receives that load, i.e., the two trends are consistent; similarly, if it is a load decrease event, the corresponding effluent data window should show a corresponding downward trend.

[0042] Only when the direction of change of the effluent index within the candidate interval is consistent with the direction of change at the influent end is the data within that interval officially determined as the target response information. If the trend directions within the candidate interval do not match, the window position is re-tested with a finite number of shifts to avoid matching failure due to delay estimation bias.

[0043] After establishing the correspondence between target change events and target response information, correlation analysis can be conducted. The correlation analysis can include at least the calculation of pollutant removal efficiency and a preliminary assessment of the source of the anomaly.

[0044] For example, the average value of the pollution index within the influent window corresponding to the target change event is used as the representative value of the influent, and the average value of the pollution index within the target response information window is used as the representative value of the effluent to calculate the pollutant removal rate of this batch of influent.

[0045] Since the influent and effluent data at this time physically correspond to the same batch of water that has undergone the complete treatment process, rather than misaligned data that may correspond to different batches of water after simple alignment based on timestamps, the removal rate obtained is closer to the actual treatment capacity of the biochemical treatment unit and the entire process system under the load impact.

[0046] Furthermore, within this correlation analysis framework, when it is necessary to trace the cause of a particular effluent anomaly, one can start from the moment of the anomaly and trace back the corresponding time period of the delay information. This allows for locating the target change event that occurred at that time within the historical timeframe of the influent. It then checks whether this event includes an abnormal load shock and, combined with whether the removal rate calculated for the corresponding time period decreased, distinguishes whether the anomaly is due to a decrease in treatment efficiency caused by an influent load shock, is related to an abnormal influent load, or is more likely due to an internal process anomaly. This approach is more rigorous in terms of causal logic than simply comparing the effluent data at the moment of the anomaly with the influent data at the same time before drawing conclusions.

[0047] Understandably, through the above process, this embodiment breaks away from the conventional approach where front-end and back-end data are aligned solely based on a unified timestamp, and instead establishes a dynamic alignment mechanism constrained by physical transmission delay. Under this mechanism, each change in influent load has the opportunity to find a back-end response that corresponds to its actual time, preventing forced mismatch due to pre-bound timestamps. This helps improve the accuracy of wastewater treatment system operation status assessment and anomaly diagnosis under fluctuating influent conditions during the rainy season.

[0048] In actual operating conditions where water inflow fluctuates rapidly during the rainy season, the water quality and flow data at the influent end of wastewater treatment plants often exhibit extremely complex fluctuation patterns. The target change events identified through preliminary identification methods such as the aforementioned sliding window may simultaneously contain multiple relatively independent influent load change phases within their corresponding time periods.

[0049] For example, the influent flow rate may initially rise rapidly, then experience a brief plateau or slight decline due to changes in the flow rate into the pipeline network, before rising sharply again; or the changes in influent flow rate and pollutant concentration may not be synchronized, with the flow rate increasing but the concentration decreasing with a lag. In such cases, if a target change event containing multiple stages is treated as a single object, the calculated single representative timestamp cannot accurately reflect the true physical starting point of each independent stage's transmission to the downstream.

[0050] This mixture of internal objects can cause the estimated delay information to deviate from the actual physical transmission process. Consequently, when locating the response information at the water outlet, it is easy to mismatch the changes in water intake from different batches with the back-end response, resulting in distortion of the correlation analysis results.

[0051] In response, this application proposes that after identifying the target change event from the first monitoring information, the method further includes: determining the internal change stage of the target change event; when the target change event is determined to contain multiple valid and relatively independent influent load change stages, splitting the target change event into multiple sub-change events; and treating each sub-change event as an independent object for calculating the corresponding delay information and locating the corresponding target response information.

[0052] Specifically, determining the internal change stages of a target change event is to confirm whether the event represents a single physical change process or comprises multiple local change stages that should be handled separately. During this determination process, the first monitoring information sequence within the corresponding time period of the target change event is analyzed to observe the continuity of its change direction and the magnitude of the change.

[0053] When multiple valid and relatively independent influent load change stages are found within an event, it indicates that the event is in a complex change state. Here, "valid and relatively independent" means that each change stage can be distinguished in time, separated in terms of physical causality, and that its change magnitude is sufficient to establish an observable independent response in the subsequent biochemical treatment unit.

[0054] Conversely, if the event contains only brief sampling noise, minor swings, or short-term disturbances insufficient to establish an independent physical response, these minor fluctuations will not be considered valid independent phases, thus avoiding misjudging normal signal jitter as multiple independent water inrush events.

[0055] When a target change event is determined to contain multiple valid and relatively independent inflow load change phases, the target change event will be divided into multiple sub-change events. Each sub-change event corresponds to a sub-interval within the time interval of the original target change event. The sub-change events are arranged sequentially in time, with no overlap or only negligible edge overlap.

[0056] Through this decomposition, the originally mixed coarse-grained target change events are transformed into a consistent set of fine-grained sub-change events. Each sub-change event inherits some basic attributes of the original target change event, such as the location of the influent monitoring point and the type of indicator, but its representative timestamp and representative flow information are redefined based on the sub-change event's own time interval to ensure that these parameters accurately reflect the transmission focus of that single change phase. This purification process complements and enhances the aforementioned identification steps, helping to improve the purity of the object and representing a further optimization based on the initial identification process.

[0057] Treating each sub-change event as an independent object for calculating the corresponding delay information and locating the corresponding target response information means that subsequent dynamic delay estimation and backend window positioning will be performed independently for each sub-change event with a single physical meaning, thereby avoiding calculations based on a mixed overall window.

[0058] In this way, each sub-change event has its own independent representative timestamp and corresponding flow status, and the calculated delay information can accurately reflect the real physical time of the transmission of the single change process to the backend. When locating target response information at the water outlet, a clear and crosstalk-free causal relationship can be established, effectively reducing matching conflicts, data omissions, or data overlap caused by the mixing of different batches of data within the event.

[0059] This pre-emptive object purification mechanism can significantly improve the accuracy of data correlation analysis in wastewater treatment systems under complex influent fluctuation conditions, providing a more reliable data foundation for subsequent pollutant removal efficiency calculations and anomaly attribution analysis.

[0060] After initially identifying target change events using methods such as sliding windows, the complexity of water inflow conditions during the rainy season often results in multiple physically independent inflow load change processes being mixed within a single time window. Treating this mixed window directly as a single object for subsequent processing would distort the representative timestamps, thus affecting the accuracy of delay estimation.

[0061] In response, this application further proposes a specific implementation method for internal refinement and reorganization of target change events. In some preferred embodiments, the determination of internal change stages and the splitting of target change events are performed as follows: the determination of internal change stages of target change events includes: based on the sequence of pollutant indicators in the first monitoring information corresponding to the target change event, determining at least one local change segment, where the local change segment is a continuous time period in the sequence where the change direction is consistent and the change magnitude exceeds a first threshold; merging adjacent local change segments with the same change direction and a time interval less than a second threshold; when multiple local change segments still exist after merging, determining that the target change event contains multiple effective and relatively independent influent load change stages; when determining that the target change event contains multiple effective and relatively independent influent load change stages, splitting the target change event into multiple sub-change events, including: dividing the time interval of the target change event into multiple sub-change events according to the merged local change segments, and determining a representative timestamp and a representative influent flow rate for each sub-change event, where the representative timestamp is the moment when the cumulative change amount in the corresponding time period of the sub-change event reaches half of the total change amount of the segment, and the representative influent flow rate is the instantaneous flow rate at the representative timestamp in the flow information.

[0062] Specifically, during rainy season storms, inflow and pollutant concentrations often exhibit a complex pattern of multiple overlapping peaks. For example, an initial rainstorm may cause a rapid increase in inflow, followed by a slight decrease or stabilization due to brief changes in the rhythm of pipeline inflow, and then a rapid increase again due to subsequent rainfall.

[0063] In this scenario, the initially identified target change events actually encompass two or more relatively independent load impact processes. If only the overall characteristics of the entire window are relied upon to extract a uniform representative timestamp, this timestamp is likely to be biased towards only one impact, failing to accurately reflect the transmission starting point of other impacts. This impurity of the front-end object will directly propagate backward, causing the subsequently calculated dynamic dwell time to fail to accurately correspond to any of the actual physical change processes.

[0064] To address this issue, it is first necessary to delve into the target change event and identify the actual localized change segments. Specifically, this involves obtaining continuous sampling sequences of pollutant indicators from the first monitoring data within the time interval of the target change event, such as the concentration values ​​of chemical oxygen demand (COD) or ammonia nitrogen, and calculating the differences between adjacent sampling points.

[0065] By iterating through these differences, the direction of change for each sampling interval can be marked. When the difference of a certain indicator remains the same sign for several consecutive sampling intervals, that is, the direction of change remains consistent, and the cumulative change in this continuous period exceeds a preset first threshold, the continuous period can be identified as a local change segment.

[0066] It is worth noting that the first threshold is not arbitrarily chosen, but rather derived from the historical operating data of the wastewater treatment plant. For example, the upper quartile, a preset high quartile, or a statistical upper limit can be determined based on the normal fluctuation range and used as the first threshold to distinguish between changes in effective load and normal background fluctuations.

[0067] The purpose of this setup is to separate effective load changes with actual process impacts from conventional sampling noise or minor background fluctuations. Simultaneously, to prevent individual fluctuations from being misidentified as change segments, a minimum duration requirement can be set for local change segments, such as requiring at least three consecutive sampling intervals. This, combined with the first threshold, ensures the reliability of the identification results. In actual monitoring data, there may be minor reverse fluctuations. Therefore, when determining whether the change direction remains consistent, a small tolerance range can be introduced. If the absolute value of the reverse difference is less than a very small proportion of the normal fluctuation range, and then quickly returns to the original direction, it can still be considered that the direction remains consistent, without interrupting the continuity of the local change segment.

[0068] After initially identifying multiple local variation segments, a merging process is needed to prevent excessive splitting. In actual water inflow processes, the same load impact may appear as a brief period of stability or a slight fluctuation in the data sequence due to short-term water flow disturbances or sensor sampling jitter during transmission and sampling. If the data is mechanically split according to the direction reversal, the process that originally belonged to the same physical event will be incorrectly divided into multiple independent events.

[0069] To address this, this solution introduces a merging rule based on a second threshold. Specifically, if two adjacent local change segments change in the same direction and the time interval between them is less than the second threshold, these two local change segments are merged into one, with its time interval covering the original two segments and maintaining the original direction of change. The second threshold can typically be set as several times the sampling interval, or selected in conjunction with the minimum duration of short-term flow disturbances in the biochemical pool. The merging process can be iteratively performed from left to right in chronological order until all adjacent segments no longer meet the merging conditions. This merging mechanism effectively filters out brief fluctuations in the same change process, ensuring the physical consistency of the front-end objects.

[0070] After merging, the number of remaining local change segments can be counted. If the number is still greater than one, it indicates that there are indeed multiple change processes with reversed direction or sufficiently long intervals within the target change event. In this case, it can be determined that the event contains multiple effective and relatively independent inflow load change stages, i.e., it is in a composite change state. Conversely, if the number is one, it indicates that the event belongs to a single change state, and it can proceed to subsequent steps directly with its original form or after retrieving representative information.

[0071] For target change events identified as being in a complex change state, they need to be broken down into multiple sub-change events. Specifically, the time interval of the original target change event is divided directly based on the time boundaries of the merged local change segments. Each sub-change event corresponds to a merged local change segment, with a clearly defined start and end time and dominant change direction.

[0072] More importantly, it is necessary to redetermine the representative timestamp and representative inflow rate for each sub-change event. In traditional processing methods, the moment with the largest change rate within the window is often taken as the representative timestamp. However, after the composite change is broken down, this method may not accurately reflect the overall transmission characteristics of the sub-segment.

[0073] In this scheme, the representative timestamp is defined as the moment when the cumulative change within the corresponding time period of the sub-change event reaches half of the total change in that period. The cumulative change here is not limited to the positive change in the direction of numerical increase, but is accumulated according to the dominant change direction of the sub-change event; for load-increasing sub-change events, the increase between adjacent sampling points is accumulated along the upward direction, and for load-decreasing sub-change events, the decrease between adjacent sampling points is accumulated along the downward direction.

[0074] In other words, when the cumulative change reaches half of the total change for that segment, it means the cumulative change amplitude reaches half of the total change amplitude for that segment. This avoids misunderstandings caused by the different signs of the numerical values ​​for the rising and falling segments. The reason for choosing the moment when the cumulative change amplitude exceeds half is that, in the process of physical hydraulic transmission, this moment roughly represents the center of gravity or midpoint of load transmission in that change stage. Compared to taking the start or end moment, using the center of gravity as a time representative can more reasonably reflect the time benchmark for the overall transmission of this batch of influent to the downstream.

[0075] In one specific implementation, assuming the total duration of a rising sub-event is forty minutes, comprising five discrete sampling points at ten-minute intervals, the total change in chemical oxygen demand (COD) concentration during this period is one hundred milligrams per liter, half of which is fifty milligrams per liter. By cumulatively adding the concentration increases between adjacent sampling points along the time axis, it is found that the cumulative change in the first twenty minutes is forty milligrams per liter, while in the sampling interval from the twentieth to the thirtieth minute, the cumulative change increases to sixty milligrams per liter, exceeding the fifty milligrams per liter threshold.

[0076] At this point, using linear interpolation, the precise moment when the cumulative change exactly reaches 50 mg / L is calculated to be the 25th minute, and this moment is precisely positioned as the representative timestamp. For a decreasing sub-change event, the change can also be accumulated along the decreasing direction in the same way, and the moment when the cumulative decrease reaches half of the total decrease in that segment is determined as the representative timestamp. Subsequently, the instantaneous flow rate value at the representative timestamp is directly read from the flow rate information as the representative influent flow rate for that sub-change event.

[0077] Through the above-mentioned refined judgment and reorganization process, a single target change event that might have contained multiple stages of change and mixed information has been transformed into a set of clearly defined and independent sub-change events. Each sub-change event now contains only one inflow load change stage with a clear direction of change and a single physical process, and has its own dedicated timestamp and flow value that are closer to the transmission focus of that stage.

[0078] This transition from complex to simple, from mixed to pure states ensures that each subsequent front-end input object entering the dynamic delay estimation stage strictly corresponds to an independent physical change process. This helps eliminate the root cause of errors propagating to subsequent delay calculations and window matching due to impurities in the front-end object. This allows the entire time-series correction method to maintain accurate correspondence between front-end and back-end events even when facing complex real-world conditions such as rapid fluctuations in inflow during the rainy season and multiple peak overlaps. This, in turn, improves the accuracy of pollutant removal rate calculations and the reliability of anomaly attribution.

[0079] In the actual operation of wastewater treatment systems, the influent flow rate can vary significantly due to rainy seasons or pipeline scheduling, which directly alters the actual hydraulic retention time of wastewater in treatment units such as biological treatment tanks. If a fixed empirical time difference is used to estimate the delay required for changes in the influent to be transmitted to the effluent, it often results in significant time positioning errors when the flow rate fluctuates drastically.

[0080] This is because under low-flow conditions, the wastewater flow velocity slows down, and the actual residence time is significantly extended; while under high-flow impact, the wastewater flow velocity accelerates, and the actual residence time is significantly shortened. If this dynamic change in physical transmission time is not accurately quantified, it will lead to a misalignment in the time sequence of front-end and back-end data, causing load changes that have not yet been transmitted to the back-end to be incorrectly compared with the current effluent results.

[0081] In some preferred approaches, the corresponding delay information is calculated, including: determining the delay information based on a pre-established mapping relationship between hydraulic residence time and flow rate, as well as the flow rate information at the time of the sub-change event.

[0082] To achieve the aforementioned process of dynamically determining delay information, it is first necessary to establish a mapping relationship between hydraulic retention time and flow rate. This mapping relationship reflects the actual physical time required for wastewater to travel from the influent monitoring point to the effluent monitoring point under different influent flow rate conditions. In establishing this mapping relationship, actual operating data of the wastewater treatment plant under different historical flow rate intervals can be collected, recording the measured hydraulic retention time from the influent to the effluent in each flow rate interval. Through data fitting or segmented correlation, a correspondence between flow rate values ​​and estimated retention time values ​​can be formed.

[0083] This correspondence can be stored in the form of a data table, such as a table containing the left endpoint of the flow interval, the right endpoint of the flow interval, and the estimated dwell time of the corresponding flow interval. It can also be stored in the form of a piecewise linear function or an interpolation curve.

[0084] It is worth noting that the reason why this mapping relationship needs to be established in advance is that the pool volume and internal flow pattern of a sewage treatment plant are relatively fixed, and the influent flow rate becomes the core variable that determines the hydraulic retention time. The mapping relationship calibrated through historical data can truly reflect the unique hydraulic transmission pattern of the plant.

[0085] After obtaining the refined sub-change events, the flow information at the time of occurrence is acquired for each sub-change event. Specifically, the representative timestamp corresponding to the sub-change event can be extracted, and the instantaneous influent flow value at that representative timestamp can be obtained from the reading sequence of the influent flow meter.

[0086] Subsequently, the instantaneous influent flow rate is used as input to look up a pre-established mapping relationship between hydraulic retention time and flow rate. If the mapping relationship exists in the form of a data table, the estimated hydraulic retention time that matches the instantaneous influent flow rate is calculated by looking up the table and combining it with linear interpolation.

[0087] During the long-term operation of the system, this mapping relationship can also be updated periodically based on new calibration data to maintain the accuracy of dynamic estimation.

[0088] In this way, the delay information is transformed from a fixed empirical value that does not consider the current flow conditions into a dynamic estimate that matches the actual influent flow rate at the time of the current sub-change event. Since the flow rates at the time of different sub-change events may differ, each sub-change event can obtain personalized delay information reflecting its true transmission speed. This dynamic query process based on a physical mapping model makes the delay estimate closer to the actual hydraulic retention time of the wastewater treatment system under the current influent load. This helps reduce time positioning deviations caused by flow fluctuations, thereby providing a more accurate time center benchmark for locating the target response window in the subsequent backend data and reducing the risk of overall misalignment between front-end and back-end data.

[0089] Although the dynamic delay information corresponding to each sub-change event has been calculated in the preceding steps, if it is used as a fixed time offset to directly extract the back-end effluent data, in actual rainy season conditions, due to estimation errors in the hydraulic model itself, short-circuiting phenomena in the pool, and the discreteness of sensor sampling time, the directly extracted time window often deviates from the actual physical response window. This deviation can easily lead to data that has not yet generated a response or data caused by other independent interference being mistakenly identified as the response of the current inflow event, thus causing correlation misalignment. To address this, this application proposes a further positioning verification mechanism.

[0090] Locating target response information for sub-change events includes: determining candidate response time periods in the second monitoring information based on delay information; and determining target response information from the candidate response time periods based on the consistency between the change trend of the second monitoring information and the change trend of the sub-change events within the candidate response time periods.

[0091] In the specific implementation process, when determining the candidate response period based on the delay information, the expected arrival time indicated by the delay information can be used as the center, and a certain time range can be extended to the front and back sides to define the candidate response period.

[0092] For example, the time span can be extended by 15 percent before and after the expected arrival time, with a minimum time width of 20 minutes, thus defining a limited search area on the timeline of the back-end water effluent monitoring data. The purpose of this candidate response period is to narrow down the search scope of the back-end data, providing a reasonable analytical boundary for subsequent trend verification. This helps avoid aimless blind matching within massive historical data, thereby keeping the verification workload within a reasonable range.

[0093] After defining the candidate response period, it is necessary to observe the change trend of the effluent water quality indicators over time during this period and compare it with the change trend of the influent load of the front-end sub-event. The change trend of the influent load of the front-end sub-event can be determined based on the change direction and magnitude of the pollutant indicators in the first monitoring information corresponding to the sub-event.

[0094] This comparison is made because, under the complex operating conditions of the rainy season, fluctuations in the effluent data may originate from multiple different upstream events or internal process disturbances. Specifically, the direction of change in the pollutant concentration sequence within the candidate time period can be calculated. If the upstream event is a continuous increase in the influent load, then, under the laws of physical transfer, the corresponding pollutant concentration in the downstream effluent should also show a gradual upward trend after the residence time.

[0095] By comparing the direction or pattern of change between the two, it is possible to determine whether the changes in the backend data constitute a reasonable physical response to the frontend water ingress event. This verification based on physical response patterns helps to identify independent fluctuations or noise data that, although falling within the candidate time period, are not related in terms of process logic, thereby avoiding mistaking irrelevant interference signals as a real response to the current water ingress impact.

[0096] When determining the final target response information based on the consistency results, if the trend of backend data changes within the candidate response period is consistent with the trend of frontend terminal change events, then the data within that period can be confirmed as the target response information.

[0097] If the trend of backend data is inconsistent with that of the frontend within the initially defined candidate response period, the candidate period can be shifted forward or backward for a re-search. For example, a fixed time step can be shifted each time, with a maximum number of shifts limited. If a window with a consistent trend cannot be found after multiple shifts, the sub-change event can be marked as temporarily unmatched and recorded in the analysis log for subsequent manual verification, avoiding forced association with incorrect backend data.

[0098] Through this positioning mechanism that includes trend verification and fault-tolerant search, the correlation status of front-end and back-end data is transformed from an unverified state that simply depends on time offset to a reliable correlation status that is confirmed by physical response trends. This provides a data foundation with high physical credibility for subsequent removal rate calculation and anomaly attribution.

[0099] In some preferred approaches, after locating the target response information, the process further includes: based on the physical process of transmitting changes in influent load to the effluent monitoring point and establishing a stable response, dividing the time interval corresponding to the target response information into a first time region and a second time region along the time axis, with the first time region being later than the second time region in chronological order; using the monitoring data in the first time region for a first type of correlation analysis, and using the monitoring data in the second time region for a second type of correlation analysis, wherein the first type of correlation analysis includes the calculation of pollutant removal efficiency, and the first type of correlation analysis does not use the monitoring data in the second time region.

[0100] In the aforementioned steps, a backend data window, namely the target response information, has been determined to match the front-end influent change event through delay information and trend verification. Because the biological treatment tank of the wastewater treatment system has a large volume and a long process chain, the transmission of influent load changes to the effluent monitoring point and the establishment of a complete response is a gradual physical process.

[0101] Within the time interval corresponding to the target response information, data closer to the start time often represents a transitional phase where the influent change has just been transmitted to the effluent monitoring point and the effluent water quality has not yet fully responded. This portion of the monitoring data cannot fully reflect the final impact of the influent load change. Directly using all monitoring data from the entire time interval without differentiation for subsequent calculations may introduce bias.

[0102] To this end, this application further distinguishes and processes the data properties and available status within the target response information. When dividing the first time zone and the second time zone, the start and end timestamps of the target response information can be obtained, and the division point can be determined based on the physical transmission characteristics of the water inflow change.

[0103] In one optional implementation, the time interval can be divided according to a fixed ratio. For example, the first 1 / 3 of the time interval can be designated as the second time zone, and the remaining 2 / 3 as the first time zone. In another optional implementation, the time interval can be dynamically divided based on the expected transmission time of the water inflow change and the inflection point of the actual backend monitoring data trend. The time period before the trend inflection point can be designated as the second time zone, and the time period after the trend inflection point as the first time zone. Through this division, the target response information is transformed from a single data window without distinguishing internal states into a structured data object whose internal structure is clearly divided into different response stages.

[0104] The first time zone corresponds to the data area where the inflow change has been fully transmitted and the downstream status has stabilized. The second time zone corresponds to the data area where the inflow change has just begun to be transmitted to the downstream monitoring points but has not yet fully propagated. Based on this distinction in physical state, different usage labels can be configured for the monitoring data in the two zones.

[0105] The first type of correlation analysis mainly refers to analysis and processing that requires steady-state response data. Besides pollutant removal efficiency calculations, the first type of correlation analysis can also include load shock tolerance assessment and sludge activity assessment. When performing pollutant removal efficiency calculations, the representative influent value can be taken as the average of all influent monitoring index values ​​within the influent change window, while the representative effluent value is only taken as the average of the corresponding effluent monitoring index values ​​within the first time period.

[0106] The removal rate calculated in this way is based on truly stable response data, which can more accurately reflect the system's actual processing capacity under that load condition. Without internal zone division, in scenarios where the influent load rises rapidly, the calculated removal rate is prone to being artificially inflated because the effluent data in the first half of the window has not yet reflected the water quality deterioration; conversely, when the load drops rapidly, the calculated removal rate is prone to being artificially understated. By excluding unstable data in the second time zone, these distortions can be effectively avoided, contributing to improved accuracy in pollutant removal efficiency calculations.

[0107] The second type of correlation analysis mainly refers to analysis and processing that can be performed using transitional state data or full-process data. This type of correlation analysis can include response establishment rate assessment, transmission delay verification, and detection of abnormal fluctuations during the transition period. When performing a response establishment rate assessment, the response rate and buffering capacity of the biochemical treatment unit to shock loads can be evaluated by monitoring the rate of change of effluent indicators within a second time zone and the time required to reach stability.

[0108] If the response time is significantly longer than the historical normal level, it may indicate that the activity of the biochemical system is inhibited or that short-circuiting is present.

[0109] Through the above processing, we can further introduce data availability differentiation for micro-transmission processes, so that subsequent correlation analysis is based on purer data that better reflects the true stable state.

[0110] In the daily operation of wastewater treatment systems, ensuring stable effluent quality meets standards is the core requirement. However, when abnormalities such as excessive pollutant concentrations are detected at the online monitoring points for the effluent, operators typically instinctively check the influent data at the same time, attempting to find the cause of the anomaly.

[0111] However, due to the large volume of the biological treatment tank and the long process chain, the influent needs several hours of hydraulic retention time to reach the effluent. This means that an abnormality in the effluent at the current moment is often actually caused by the influent load surge several hours earlier. If the same timestamp is used directly to compare the data from the front and back ends, it will forcibly associate the normal load at the front end, which has not yet reached the effluent, with the current effluent abnormality, or mistakenly identify the current normal influent as the cause of the abnormality. This time sequence misalignment will not only lead to misjudgment of the root cause of the abnormality, but may also mislead the direction of process adjustments, causing operators to take incorrect countermeasures, thereby causing further deterioration of the effluent quality.

[0112] To address this issue, in order to identify historical influent load change events that may correspond to effluent anomalies in time under abnormal operating conditions, and to distinguish between influent load factors and internal process factors, thereby avoiding misattribution caused by blind comparison, this application further proposes a reverse backtracking mechanism based on physical transmission delay. In response to the detection of abnormal information characterizing water quality anomalies at the effluent end, the mechanism determines the historical time period corresponding to the abnormal information in time based on the delay information, searches for identified target change events within the historical time period, and performs anomaly attribution analysis based on the characteristics of the found target change events.

[0113] Specifically, during the continuous acquisition of data sequences from online monitoring points at the effluent outlet, various water quality indicators at the effluent end are monitored in real time. A normal threshold range can be set for the effluent indicators, which can be determined comprehensively based on the statutory discharge standards of the wastewater treatment plant and the fluctuation range during historical normal operation.

[0114] When the real-time collected value or short-term moving average of the effluent indicator exceeds the normal threshold range, an anomaly backtracking marker can be triggered, generating anomaly information. This anomaly information can record the specific timestamp of the anomaly occurrence and the type of anomaly indicator involved.

[0115] The short-time moving average is introduced for judgment because online monitoring instruments may occasionally generate instantaneous noise or spike signals. Short-time smoothing can effectively filter out these occasional interferences, ensure the authenticity of the triggered abnormal information, and avoid frequent false alarms that bring additional psychological burden and ineffective troubleshooting work to the operators.

[0116] Reference Figure 4 After generating the anomaly information, the dynamic delay estimation capability established in the previous steps needs to be used for reverse extrapolation along the time axis. Specifically, the delay information used for reverse backtracking can be determined based on the anomaly time corresponding to the anomaly information and the flow information before that anomaly time, combined with the pre-established mapping relationship between hydraulic residence time and flow rate.

[0117] The delay information used for reverse backtracking can be estimated by the inflow rate near the abnormal moment and its preceding time period; when the flow rate changes relatively smoothly, the instantaneous inflow rate near the abnormal moment can be used for table lookup or interpolation; when the flow rate changes drastically, the flow rate sequence within a preset time range before the abnormal moment can be used to calculate the corresponding estimated residence time or candidate residence time range.

[0118] Subsequently, using the anomalous moment recorded in the anomaly information as the time endpoint, a reverse calculation is performed by subtracting the time span represented by the delay information from the time axis, thereby locating a historical center moment or a candidate historical center moment. Considering the hydraulic diffusion effect and inherent errors in model estimation during actual hydraulic transmission, a certain margin of time can be added forward and backward from this historical center moment, such as ±15 or ±20 minutes, to ultimately determine a historical time period that has a real physical connection with the anomaly information in terms of time sequence.

[0119] The above approach narrows the scope of the search for upstream causes of anomalies from inlet data contemporaneous with the anomaly to historical time periods at the inlet that could actually lead to the anomaly, thus laying the foundation for accurate matching in the future.

[0120] After determining the historical time period, the identified target change events will be searched within that time period. The previous steps have already performed a sliding window analysis on the data at the inlet end to identify the target change events that characterize changes in the inlet load, and the characteristic information of these events, such as start time, end time, change type, and load change magnitude, will be recorded in the corrected correlation analysis dataset.

[0121] Therefore, the dataset can be directly accessed to search for target change events with overlapping or highly consistent time intervals within historical periods. If one or more target change events are found within that historical period, their feature information is extracted. This search method fully utilizes the refined event identification results already completed in the preceding steps, avoiding blind secondary searches in massive amounts of raw historical data, and helps improve the computational efficiency and response speed of anomaly backtracking.

[0122] After identifying the corresponding target change event, anomaly attribution analysis is performed based on the characteristics of the identified target change event. Specifically, the extracted event characteristics, such as whether the event is a short-term rapid fluctuation or whether the increase in influent pollution load exceeds the system's tolerance limit, are combined to determine whether there was a significant influent load anomaly at that time.

[0123] Simultaneously, it will also call upon the pollutant removal efficiency data calculated in the previous steps for this target change event. If it is determined that there was indeed an abnormal influent load at that time, and the pollutant removal efficiency during that period was significantly lower than the reference removal efficiency under normal operating conditions, it can be preliminarily inferred that the effluent abnormality is due to a decrease in treatment efficiency caused by an influent load shock.

[0124] If it is determined that there is an abnormal influent load at that time, but the pollutant removal efficiency remains at a normal level, it indicates that the removal capacity of the biological treatment unit has not significantly decreased. However, the abnormality may still be related to excessively high influent load, insufficient effluent compliance margin, or influent peak exceeding the system design capacity. Therefore, attribution results can be generated to indicate that the abnormality is related to the abnormal influent load.

[0125] If no obvious abnormality in the influent load was detected at the time, but the pollutant removal efficiency was lower than the reference removal efficiency under normal operating conditions, the root cause of the abnormality can be attributed to process abnormalities within the biological treatment unit or abnormal sludge-water separation in the secondary sedimentation tank.

[0126] After completing the above analysis, the historical time period obtained from reverse location, the characteristics of the target change event found, the removal efficiency judgment result, and the final attribution result will be packaged together with the water output index data at the time of the anomaly to form an anomaly backtracking analysis record and stored in the corrected correlation analysis dataset.

[0127] Through the above approach, the diagnostic status is transformed from merely knowing that an anomaly occurred in the effluent at a certain moment with no clear cause, to being able to determine whether the anomaly is related to a specific historical influent change event on the physical timeline, and to obtain a root cause conclusion with clear direction. This transforms the anomaly information from an isolated alarm signal into a traceable event that can be correlated with upstream disturbance sources or internal process conditions.

[0128] Based on this accurate traceability information, operators can quickly and specifically take process adjustment measures, such as adjusting the sludge return ratio in advance or adding emergency agents, thereby significantly improving the sewage treatment system's ability to handle abnormal operating conditions and effectively avoiding misattribution and blind investigation caused by misaligned timing.

[0129] In some preferred approaches, anomaly attribution analysis is performed based on the characteristics of the identified target change events, including: determining whether there is an influent load anomaly based on the characteristics of the identified target change events; obtaining the pollutant removal efficiency calculated based on the influent pollutant index corresponding to the identified target change event and the effluent pollutant index corresponding to the associated target response information within a historical time period; when there is an influent load anomaly and the pollutant removal efficiency is lower than a preset reference removal efficiency, determining the root cause of the anomaly information as a decrease in treatment efficiency caused by influent load shock; when there is an influent load anomaly and the pollutant removal efficiency is not lower than a preset reference removal efficiency, generating attribution results indicating the correlation between the anomaly information and the influent load anomaly; when there is no influent load anomaly and the pollutant removal efficiency is lower than a preset reference removal efficiency, determining the root cause of the anomaly information as an internal process anomaly.

[0130] After tracing back to identify historical target change events that may have caused effluent anomalies, the specific root cause of the anomaly often remains unclear. At this point, the exceedance of effluent indicators may be due to a sudden surge in the influent load exceeding the treatment limit of the biological treatment tank, or it may be caused by process malfunctions such as decreased sludge activity or deterioration of sludge-water separation in the secondary sedimentation tank.

[0131] Blindly attributing causes based solely on experience can easily lead to operators taking incorrect intervention measures, such as incorrectly restricting water intake when internal equipment malfunctions, or failing to adjust internal recirculation and aeration rates in a timely manner during water inrush.

[0132] In response, this application proposes a root cause differentiation mechanism based on both influent characteristics and treatment efficiency, thereby transforming general anomaly analysis into a diagnostic process with clear direction.

[0133] Specifically, when determining whether there is an abnormal influent load, the change characteristics of the influent flow rate and major pollutant indicators corresponding to the target change event can be extracted for comprehensive evaluation. For example, the rate of change of influent flow rate or pollutant indicators such as chemical oxygen demand and ammonia nitrogen within the event window can be calculated and compared with the statistical thresholds of the historical rainy season fluctuation period of the wastewater treatment plant.

[0134] An abnormal influent load can be identified when the rate of change exceeds the upper quartile of the historical normal fluctuation distribution, the preset high quartile, or the statistical upper limit determined based on historical normal fluctuation data, or when the peak concentration of influent pollutants exceeds the upper limit of the historical normal operating range.

[0135] Simultaneously, it is necessary to obtain the pollutant removal efficiency calculated and stored based on the preceding correlation analysis steps within the historical time period identified by reverse location. To improve the robustness of single anomaly backtracking, representative values ​​of multiple batches of correlation analysis results within this historical time period can be read, such as the average removal rate calculated within the communicated status data area, which can serve as a basis for reflecting the overall processing efficiency during this period.

[0136] This application examines both influent load anomalies and pollutant removal efficiency simultaneously because they reflect the upstream load input status and the actual treatment efficiency of the treatment system, respectively. If both influent load anomalies and pollutant removal efficiency are lower than the preset reference removal efficiency, it indicates that the upstream load shock has affected the actual removal capacity of the treatment system. The root cause of the anomaly can be identified as the decrease in treatment efficiency caused by the influent load shock.

[0137] If there is an abnormal influent load, but the pollutant removal efficiency remains at a normal level, it indicates that the relative removal capacity of the biological treatment unit has not significantly decreased. However, since the peak concentration of pollutants in the influent or the total load may have exceeded the effluent compliance margin, the current abnormality may still be related to the abnormal influent load. Therefore, attribution results can be generated to indicate that the abnormal information is related to the abnormal influent load, rather than directly classifying it as an internal process abnormality.

[0138] If no obvious abnormality in the influent load is detected, but the pollutant removal efficiency decreases, it indicates that the decline in treatment efficiency is more likely caused by internal process factors such as sludge aging, improper dissolved oxygen control, slow accumulation of toxic substances, and abnormal sludge-water separation in the secondary sedimentation tank. The root cause of the abnormal information can be identified as an internal process abnormality.

[0139] The preset reference removal efficiency can be calculated based on the pollutant removal efficiency data of the wastewater treatment plant during stable operation in the dry season or under historical normal operating conditions. For example, the average value or a specific lower quantile value of the removal rate during normal operation can be taken, and it can be updated regularly in combination with historical intervals of environmental factors such as water temperature and sludge age.

[0140] Through the above-described scenario-based judgment, the internal state of the anomaly information changes from a vague alarm with an undetermined cause to a structured record carrying a clear attribution label. This state change allows downstream alarm modules and operation and maintenance suggestion modules to directly read the classification results, thereby providing operators with targeted process control guidance, helping to shorten the anomaly investigation time, and enabling the entire time series correction correlation analysis to form a closed loop at the application level from problem discovery to identifying the source of responsibility.

[0141] During the long-term operation of a wastewater treatment plant, physical changes often gradually occur inside the biological treatment tank, such as sludge accumulation, short-circuiting, or changes in the effective tank volume utilization efficiency. These physical changes cause the mapping relationship between flow rate and retention time, which was originally established based on historical calibration data, to gradually deviate from the actual physical hydraulic retention time.

[0142] If this already offset mapping relationship is used to calculate latency information, it will lead to a systematic deviation in the matching of front-end and back-end data windows. This will cause more and more front-end change events to fail to find consistent back-end responses within a reasonable shift range, thus affecting the accuracy of subsequent removal rate calculations and anomaly attribution.

[0143] In response, this application proposes to obtain matching result statistics, which are used to characterize the degree of matching failure of front-end and back-end matching based on the delay information determined by the mapping relationship and multiple sub-change events; when the matching result statistics meet the reference offset judgment condition, a parameter calibration instruction is generated, which is used to indicate that the mapping relationship is calibrated.

[0144] Specifically, the mapping relationship itself is a static data table, while the actual physical volume of the biochemical pool is a dynamically changing entity over time. To determine whether the static data table can accurately reflect the dynamic physical entity, reverse inference can be performed by examining the matching quality of the data table's output.

[0145] During the continuous process of matching front-end and back-end data, a large number of matching records will be accumulated. Every fixed verification cycle, such as every two hours, the most recent batch of front-end and back-end data window records that have been matched can be retrieved, and the matching process in these records can be reviewed and statistically analyzed.

[0146] The specific statistical method involves calculating the number of backend candidate windows in the most recent batch of records that reached a preset limit in the number of times they performed translation searches to find data with consistent trends, and then using this number as the proportion of the total number of matching windows in this batch as the statistical information of the matching results. This proportion can intuitively reflect the degree of failure of the current mapping relationship in guiding the matching.

[0147] When the mapping relationship is offset, the calculated estimated transmission time will be too large or too small overall, causing the center of the initially located candidate window to deviate from the actual backend response position. To compensate for this deviation, the matching logic will continuously shift the window forward or backward until the maximum number of shifts is reached.

[0148] After obtaining the statistical information of the matching results, it will be compared with the preset reference offset judgment conditions. For example, when the proportion of windows that have reached the upper limit of the above translation search exceeds 40%, it can be determined that the reference offset judgment conditions are met. At this time, a parameter calibration command can be generated. This parameter calibration command can be manifested as a pop-up prompt box on the operation monitoring interface, or an alarm record written to the operation log. Its content is mainly used to remind operators that the current mapping relationship may be offset, and it is recommended to recalibrate the actual pool capacity utilization efficiency and hydraulic retention time of the biological tank.

[0149] After generating the parameter calibration command, the current mapping relationship will continue to be used for subsequent matching processes until the operator completes the on-site recalibration and performs the confirmation operation on the interface. Only then will the updated mapping relationship be officially applied to the online analysis.

[0150] This approach avoids uncontrollable risks that automatic correction might introduce, ensuring the stability and traceability of the data analysis process. Through this dynamic self-checking mechanism, potential failures of core parameters can be detected and prompt for manual intervention in a timely manner, helping to break the chain of continuous accumulation of erroneous delay estimates and thus ensuring the accuracy of the dynamic delay estimation mechanism in long-term operation.

[0151] This application also proposes a wastewater treatment system data analysis system for performing the method steps described above, including an acquisition module, an identification module, a calculation module, and a location association module. The acquisition module acquires first monitoring information characterizing the change in influent water quality over time, second monitoring information characterizing the change in effluent water quality over time, and flow information characterizing the change in influent flow rate over time. The identification module identifies a target change event characterizing a change in influent load from the first monitoring information. The calculation module calculates a delay information characterizing the time span required for the target change event to propagate to the effluent end, based on the flow information at the time the target change event occurs. The location association module locates the target response information characterizing the actual backend response of the target change event in the second monitoring information based on the delay information, and performs association analysis between the target change event and the target response information.

[0152] Through the above scheme, the system can perform time-series correction correlation analysis on the load change events at the inlet end and the back-end response information corresponding to the hydraulic residence time at the front end. This avoids misjudgment of load transfer caused by directly aligning according to the collection timestamp when the inlet flow fluctuates rapidly, thereby improving the accuracy of subsequent removal rate calculation and anomaly source attribution.

[0153] In some specific implementations, refer to Figure 2This figure illustrates the overall flow of a wastewater treatment system data analysis method provided in an embodiment of this application. Simplified names are used to represent various information and processing steps in the figure, and the specific content of each processing step is as follows.

[0154] Figure 2 In the table, influent water quality, effluent water quality, and influent flow rate represent the first monitoring information, the second monitoring information, and the flow rate information, respectively. The first monitoring information is input into the event identification and segmentation stage; the second monitoring information is input into the response location and correlation analysis stage and the anomaly detection stage, respectively; the flow rate information is used to determine the representative influent flow rate of the sub-change event and serves as the basis for delay calculation.

[0155] Figure 2 The event identification and on-demand segmentation representation are as follows: Target change events are identified from the first monitoring information, and their internal change stages are determined. When a target change event contains multiple valid and relatively independent influent load change stages, it is segmented into multiple sub-change events, and the representative timestamp and representative influent flow rate of each sub-change event are determined. For single-segment change events, they proceed directly to subsequent processing in the aforementioned manner. The events in the figure include the identified target change events or the segmented sub-change events. Event information is provided to delay calculation, response location and correlation analysis, and anomaly attribution stages, respectively.

[0156] Figure 2 The mapping relationship in the diagram represents a pre-established mapping between hydraulic residence time and flow rate. The delay calculation stage determines the delay information based on the flow rate information corresponding to the event and this mapping relationship. The delay information is provided to the response localization and correlation analysis stage and the anomaly attribution stage, respectively. Event information and delay information are different inputs; event information is not generated by the delay calculation stage.

[0157] Figure 2 The response location and correlation analysis stages receive event information, delay information, and second monitoring information. Based on the event's time location and delay information, candidate response periods are determined, and the target response information is located based on the trend consistency between the event and the effluent data within the candidate response periods. After locating the target response information, a first time region and a second time region can be divided as described above, and correlation analysis can be performed to obtain analytical results such as pollutant removal efficiency. The figure only shows the main processing stages; the specific methods for candidate time period determination, trend consistency matching, and time region division adopt the aforementioned implementation method.

[0158] Input the statistical information of matching results generated during response localization and correlation analysis. Figure 2The calibration and judgment process involves a parameter calibration instruction generated when the statistical information of the matching results meets the reference offset judgment criteria. This instruction prompts recalibration. The mapping relationship between hydraulic residence time and flow rate is only updated after the operator completes and confirms the on-site recalibration. The current mapping relationship is used until the new mapping relationship is confirmed. The update shown in the figure represents a mapping relationship update after manual confirmation, and does not indicate automatic online correction of the mapping relationship based on the matching results.

[0159] Figure 2 The anomaly detection stage detects water quality anomalies at the effluent end based on the second monitoring information and provides the detected anomaly information to the anomaly attribution stage. The anomaly attribution stage also receives event information, delay information, and pollutant removal efficiency obtained from correlation analysis. It determines the historical time period using the aforementioned reverse backtracking method, finds the corresponding historical target change events, and performs an influent load anomaly judgment and pollutant removal efficiency comparison. In the diagram, the lines represent the information provision relationships; branches with solid dots indicate that the same information is provided to different treatment stages; and arcs crossing lines indicate that the intersecting lines are not connected.

[0160] Figure 2 The attribution results include the following scenarios: when there is an abnormal influent load and the pollutant removal efficiency is lower than the preset reference removal efficiency, the attribution result for the decrease in treatment efficiency due to influent load shock is output; when there is an abnormal influent load and the pollutant removal efficiency is not lower than the preset reference removal efficiency, the attribution result for the abnormal information related to the abnormal influent load is output; when there is no abnormal influent load and the pollutant removal efficiency is lower than the preset reference removal efficiency, the attribution result for internal process abnormality is output. The above results are combined in the figure for simplification.

[0161] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. All modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A data analysis method for a wastewater treatment system, characterized in that, include: Acquire first monitoring information to characterize the change of water quality at the inlet of the wastewater treatment system over time, second monitoring information to characterize the change of water quality at the outlet of the wastewater treatment system over time, and flow information to characterize the change of influent flow rate over time. Identify target change events that characterize changes in influent load from the primary monitoring information; Based on the flow information at the time the target change event occurs, calculate the delay information used to characterize the time span required for the target change event to propagate to the outlet. Based on the delay information, the target response information used to characterize the actual backend response corresponding to the target change event is located in the second monitoring information, and the target change event and the target response information are correlated and analyzed.

2. The wastewater treatment system data analysis method according to claim 1, characterized in that, After identifying target change events from the initial monitoring information, the process also includes: Determine the internal change stage of target change events; When a target change event is determined to contain multiple valid and relatively independent influent load change phases, the target change event is broken down into multiple sub-change events; Each sub-change event is treated as an independent object for calculating the corresponding delay information and locating the corresponding target response information.

3. The wastewater treatment system data analysis method according to claim 2, characterized in that, The determination of internal change stages and the breakdown of target change events are carried out in the following ways: Determine the internal change stage of the target change event, including: Based on the sequence of pollutant indicators in the first monitoring information corresponding to the target change event, at least one local change segment is determined. The local change segment is a continuous period in the sequence where the change direction is consistent and the change amplitude exceeds the first threshold. Adjacent local change segments with the same direction of change and a time interval less than the second threshold are merged. When multiple local change segments still exist after merging, the target change event is determined to contain multiple effective and relatively independent influent load change stages; When a target change event is determined to contain multiple valid and relatively independent influent load change phases, the target change event is broken down into multiple sub-change events, including: The time interval of the target change event is divided into multiple sub-change events based on the merged local change segments. A representative timestamp and a representative inflow rate are determined for each sub-change event. The representative timestamp is the moment when the cumulative change in the corresponding time period of the sub-change event reaches half of the total change in that segment, and the representative inflow rate is the instantaneous flow rate at the representative timestamp in the flow information.

4. The wastewater treatment system data analysis method according to claim 2, characterized in that, Calculate the corresponding delay information, including: Delay information is determined based on the pre-established mapping relationship between hydraulic residence time and flow rate, as well as the flow rate information at the time of sub-change events.

5. The wastewater treatment system data analysis method according to claim 4, characterized in that, Locate the target response information for sub-change events, including: Based on the delay information, candidate response time periods are determined from the second monitoring information; Based on the consistency between the changing trend of the second monitoring information and the changing trend of the sub-changing events within the candidate response period, the target response information is determined from the candidate response period.

6. The wastewater treatment system data analysis method according to claim 5, characterized in that, After locating the target response information, it also includes: Based on the physical process of transmitting changes in influent load to the effluent monitoring point and establishing a stable response, the time interval corresponding to the target response information is divided into a first time region and a second time region along the time axis. The first time region is later than the second time region in time sequence. Monitoring data from the first time zone is used for the first type of correlation analysis, and monitoring data from the second time zone is used for the second type of correlation analysis. The first type of correlation analysis includes the calculation of pollutant removal efficiency, and the first type of correlation analysis does not use monitoring data from the second time zone.

7. The wastewater treatment system data analysis method according to claim 1, characterized in that, Also includes: In response to the detection of abnormal information indicating water quality anomalies at the outlet, the system determines the historical time period corresponding to the abnormal information based on the delay information, searches for identified target change events within the historical time period, and performs anomaly attribution analysis based on the characteristics of the found target change events.

8. The wastewater treatment system data analysis method according to claim 7, characterized in that, Perform anomaly attribution analysis based on the characteristics of the identified target change events, including: Based on the characteristics of the identified target change events, determine whether there is an abnormal influent load; The pollutant removal efficiency is calculated based on the influent pollutant index corresponding to the target change event and the effluent pollutant index corresponding to the target response information associated with it within a historical time period. When there is an abnormal influent load and the pollutant removal efficiency is lower than the preset reference removal efficiency, the root cause of the abnormal information is determined to be the decrease in treatment efficiency caused by the influent load shock. When there is an abnormal influent load and the pollutant removal efficiency is not lower than the preset reference removal efficiency, an attribution result is generated to indicate that the abnormal information is related to the abnormal influent load. When there is no abnormal influent load and the pollutant removal efficiency is lower than the preset reference removal efficiency, the root cause of the abnormal information is determined to be an internal process abnormality.

9. The wastewater treatment system data analysis method according to claim 4, characterized in that, Also includes: Obtain matching result statistics, which are used to characterize the degree of matching failure in front-end and back-end matching based on the delay information determined by the mapping relationship and multiple sub-change events; When the statistical information of the matching result meets the reference offset judgment condition, a parameter calibration instruction is generated. The parameter calibration instruction is used to indicate that the mapping relationship is calibrated.

10. A wastewater treatment system data analysis system, used to execute the wastewater treatment system data analysis method according to any one of claims 1 to 9, characterized in that, include: The acquisition module is used to acquire first monitoring information to characterize the change of water quality at the inlet of the sewage treatment system over time, second monitoring information to characterize the change of water quality at the outlet of the sewage treatment system over time, and flow information to characterize the change of influent flow rate over time. The identification module is used to identify target change events that characterize changes in influent load from the first monitoring information; The calculation module is used to calculate the delay information, which characterizes the time span required for the target change event to be transmitted to the outlet, based on the flow information when the target change event occurs. The location association module is used to locate the target response information, which is used to characterize the actual backend response corresponding to the target change event, in the second monitoring information based on the delay information, and to perform correlation analysis between the target change event and the target response information.

Images