An intelligent control method for leachate treatment of a waste transfer station

Abstract

The application discloses an intelligent control method for leachate treatment of a garbage transfer station, and relates to the technical field of sewage control.The method comprises the following steps: S1, collecting leachate impact monitoring data, and performing pretreatment on the leachate impact monitoring data; S2, generating an impact response residual matrix and a source end pulse matrix, performing lag translation on the source end pulse matrix according to the maximum lag sampling number, generating three types of lag dictionaries, and combining a water conservation residual value and a group set regression model to output lag inversion results; S3, generating a compression reconstruction matrix, a flushing reconstruction matrix and a unloading reconstruction matrix, calculating three types of contribution values through source-by-source removal and error increment calculation, and generating an impact source attribution mark; and S4, decomposing a segment inflow, combining a regulation and storage residual volume and a membrane segment protection state to output an intelligent control result of leachate treatment. The method solves the problem that, in the existing leachate treatment of a garbage transfer station, impact sources are difficult to attribute due to lag aliasing, thereby causing inaccurate regulation and storage discharge, chemical addition and aeration control.

Description

Technical Field

[0001] This invention relates to the field of wastewater control technology, specifically to an intelligent control method for leachate treatment at waste transfer stations. Background Technology

[0002] With the development of technologies such as waste sorting, municipal solid waste compression and transfer, and centralized wastewater collection within waste transfer stations, leachate treatment at waste transfer stations is gradually shifting from single-equipment start-up and shutdown control to multi-parameter linkage control. Existing treatment systems typically deploy online monitoring instruments for liquid level, flow rate, water quality, and pressure around equalization tanks, biological treatment units, coagulation and chemical dosing units, and membrane treatment units. Automatic control algorithms are used to dynamically adjust the operating parameters of booster pumps, chemical dosing pumps, aeration fans, and membrane sections to improve the continuity and stability of the leachate treatment process. In the area of ​​automatic control for landfill leachate, existing technologies are beginning to incorporate collaborative control, predictive control, and intelligent optimization algorithms into the treatment process.

[0003] For example, application CN121537120B discloses a control method for landfill leachate treatment based on collaborative control. This method acquires real-time monitored water quality parameter data streams from the landfill leachate treatment system, including pH value in the pretreatment stage, COD concentration in the main treatment stage, and membrane pressure difference in the advanced treatment stage. These data streams are input into a collaborative control framework. Primary adjustment instructions are generated by dynamically correcting PID controller parameters using a nonlinear gain function. Then, a reinforcement learning strategy network is used to evaluate and generate strategy optimization instructions. Finally, both are input into a grey prediction model to output a collaborative control signal with feedforward compensation. Based on this signal, the coagulant dosage, aeration intensity, and membrane module operating pressure are dynamically adjusted to form a closed-loop control circuit.

[0004] For example, application CN118348777B discloses a control method for landfill leachate treatment based on cooperative control. By improving the PID joint filter and the Actor-Critic structure in reinforcement learning to construct cooperative control, the algorithm's policy learning efficiency and the stability of the control system are improved. The outputs of each cooperative control are used as inputs to the coupled object and the grey prediction auxiliary controller, respectively, thus constructing four feedback controls to obtain a high-efficiency landfill leachate treatment control that satisfies nonlinearity and time-varying characteristics. The improved PID introduces a nonlinear variation function, and the controller's gain parameter changes with the control error, improving the algorithm's adaptability. The reinforcement learning value function considers advantage learning, reducing the Q-value estimation of non-optimal actions to widen the gap between optimal and non-optimal state values.

[0005] However, although existing technologies can coordinate the regulation of chemical dosing, aeration, and membrane parameters in landfill leachate treatment through PID control, reinforcement learning, and grey prediction, their control is largely based on real-time water quality parameters, membrane pressure differential, and control errors themselves. This makes it difficult to identify the hysteretic cascading relationships of different impact sources within the equalization tank, especially in scenarios involving concentrated collection and transportation at landfill transfer stations, compression and drainage, in-station flushing, and vehicle unloading. When compression and drainage, flushing influent, and unloading loads collectively cause changes in liquid level, water quality, and membrane state, existing control methods tend to directly translate short-term fluctuations into external discharge, chemical dosing, or aeration adjustments. This lack of back-attribution of impact sources and separate control criteria leads to inaccurate regulation of external discharge, chemical dosing, and aeration.

[0006] Therefore, in response to the above problems, there is an urgent need for an intelligent control method for leachate treatment in waste transfer stations. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides an intelligent control method for leachate treatment in waste transfer stations. This method solves the problem that the delayed and overlapping impact sources in existing leachate treatment processes are difficult to attribute, leading to inaccurate control of storage, release, chemical dosing, and aeration.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an intelligent control method for leachate treatment in a waste transfer station, comprising: S1, collecting leachate impact monitoring data, performing anomaly removal, missing data completion, and time alignment processing on the leachate impact monitoring data, and outputting pre-processed leachate impact monitoring data; S2, identifying candidate impact response intervals based on the pre-processed leachate impact monitoring data, generating an impact response residual matrix and a source-end pulse matrix, performing hysteresis shift on the source-end pulse matrix according to the maximum hysteresis sampling number, and generating corresponding compression discharge, flushing water inlet, and unloading loads. The three types of lag dictionaries are used to output lag inversion results by combining the water conservation residual value and the grouped nested regression model with signed constraints; S3, based on the lag inversion results, compression reconstruction matrix, flushing reconstruction matrix and unloading reconstruction matrix are generated, and three types of contribution values ​​are obtained by source removal and error increment calculation, and impact source attribution labels are generated to output counterfactual attribution results; S4, based on the counterfactual attribution results, the attribution fragments are merged, the influent volume of the fragment is decomposed into compression control volume, flushing control volume and unloading control volume, and the leachate treatment intelligent control results are output by combining the storage remaining volume and membrane segment protection status.

[0009] Furthermore, the specific steps for collecting infiltration impact monitoring data and performing anomaly removal, missing data completion, and time alignment processing on the infiltration impact monitoring data are as follows: The infiltration impact monitoring data includes vehicle entry timestamp, vehicle unloading mass value, compressor start timestamp, compressor stop timestamp, compressor operating current value, compressor hydraulic pressure value, flushing pump start timestamp, flushing pump stop timestamp, flushing water instantaneous flow rate value, equalization tank level value, equalization tank influent instantaneous flow rate value, equalization tank effluent instantaneous flow rate value, influent chemical oxygen demand concentration value, and influent ammonia nitrogen concentration value. The data collected includes influent suspended solids concentration, influent turbidity, influent pH, influent conductivity, influent oil concentration, dissolved oxygen concentration in the biological treatment tank, oxidation-reduction potential in the biological treatment tank, influent pressure before the membrane, permeate pressure after the membrane, and permeate conductivity of the membrane system. For the collected infiltration shock monitoring data, the Hample filter algorithm is used to identify instantaneous instrument jump values ​​and remove abnormal sampling points. The piecewise linear interpolation algorithm is used to complete short-term missing records and align sampling times for the infiltration shock monitoring data, outputting the preprocessed infiltration shock monitoring data.

[0010] Furthermore, the specific steps for identifying candidate impact response intervals based on pre-processed infiltration impact monitoring data are as follows: Read the pre-processed infiltration impact monitoring data, and compose a multi-channel response sequence from the liquid level value of the equalization tank, the instantaneous flow rate value of the equalization tank influent, the concentration value of chemical oxygen demand influent, the concentration value of ammonia nitrogen influent, the concentration value of suspended solids influent, the turbidity value of influent, the pH value of influent, the conductivity value of influent, and the oil concentration value of influent. Use the Bayesian online change point detection algorithm to identify the change points in the multi-channel response sequence, and determine the sampling interval between two adjacent change points as the candidate impact response interval.

[0011] Further, the specific steps for generating the impact response residual matrix and the source pulse matrix are as follows: For each candidate impact response interval, read N consecutive sampling points before the start of the candidate impact response interval, calculate the median of each data in the multi-channel response sequence, and obtain the interval baseline vector; take the sampling points in the candidate impact response interval as matrix rows, and take each data in the multi-channel response sequence as matrix columns, subtract the baseline value corresponding to the same matrix column in the interval baseline vector from the data value corresponding to each matrix column in each sampling point, and obtain the impact response residual matrix; construct the source pulse matrix with the unified sampling time as the row index, write the product of the compressor running current value and the compressor hydraulic pressure value from the compressor start time stamp to the compressor stop time stamp into the compressor discharge column, write the instantaneous flow rate value of the flushing water from the flushing pump start time stamp to the flushing pump stop time stamp into the flushing water inlet column, write the vehicle unloading mass value corresponding to the vehicle entry time stamp into the unloading load column, and write 0 in the remaining matrix positions.

[0012] Furthermore, the specific steps for generating three types of hysteresis dictionaries corresponding to compression discharge, flushing inlet, and unloading load by performing hysteresis shift on the source pulse matrix based on the maximum hysteresis sampling number are as follows: Calculate the ratio of the pipeline volume from the inlet of the regulating tank to the water quality sampling point to the instantaneous flow rate of the regulating tank inlet, and calculate the ratio of the effective volume of the regulating tank to the instantaneous flow rate of the regulating tank inlet. Divide the two ratios by a unified sampling period and round up to obtain two hysteresis sampling numbers. Take the larger of the two hysteresis sampling numbers as the maximum hysteresis sampling number. Shift the three columns of data in the source pulse matrix backward sequentially from 0 to the maximum hysteresis sampling number. Concatenate the shifted columns corresponding to each hysteresis sampling number to form the hysteresis dictionary for compression discharge, flushing inlet, and unloading load.

[0013] Furthermore, the specific steps for outputting the lag inversion results by combining the water conservation residual value and the signed-constrained nested regression model are as follows: Based on the liquid level value of the regulating tank, read the corresponding regulating tank volume value from the regulating tank level-volume calibration table; subtract the regulating tank volume value of the previous sampling point from the regulating tank volume value of the subsequent sampling point to obtain the volume change within the tank; subtract the instantaneous inflow rate of the regulating tank from the instantaneous outflow rate of the regulating tank and multiply by a uniform sampling period to obtain the theoretical net inflow rate within the tank; subtract the volume change within the tank from the theoretical net inflow rate to obtain the water conservation residual value; use the compression discharge lag dictionary, flushing inflow lag dictionary, and unloading load lag dictionary as explanatory variables of the nested regression model, use the impact response residual matrix as the matrix to be explained, add the absolute value of the water conservation residual value to the objective function of the nested regression model, and solve the signed-constrained nested regression model using the alternating direction multiplier method to obtain the compression discharge coefficient matrix. The system includes a flushing influent coefficient matrix and a discharge load coefficient matrix. The compression discharge coefficient matrix is ​​limited to non-negative values ​​in the corresponding channels for equalization tank level, equalization tank influent instantaneous flow rate, influent chemical oxygen demand (COD) concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, and influent turbidity. The flushing influent coefficient matrix is ​​limited to non-negative values ​​in the corresponding channels for equalization tank level and equalization tank influent instantaneous flow rate, and also limited to non-positive values ​​in the corresponding channels for influent COD concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, influent turbidity, and influent conductivity. The discharge load coefficient matrix is ​​limited to non-negative values ​​in the corresponding channels for influent grease concentration, influent COD concentration, and influent turbidity. The system outputs hysteresis inversion results, including the impact response residual matrix, compression discharge hysteresis dictionary, flushing influent hysteresis dictionary, discharge load hysteresis dictionary, compression discharge coefficient matrix, flushing influent coefficient matrix, and discharge load coefficient matrix.

[0014] Furthermore, the specific steps for generating the compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix based on the hysteresis inversion results are as follows: Read the hysteresis inversion results, perform matrix multiplication between the compression discharge hysteresis dictionary and the compression discharge coefficient matrix to obtain the compression reconstruction matrix; perform matrix multiplication between the flushing inlet hysteresis dictionary and the flushing inlet coefficient matrix to obtain the flushing reconstruction matrix; perform matrix multiplication between the unloading load hysteresis dictionary and the unloading load coefficient matrix to obtain the unloading reconstruction matrix.

[0015] Further, the specific steps for obtaining three types of contribution values ​​through source-by-source removal and error increment calculation, generating impact source attribution labels, and outputting counterfactual attribution results are as follows: Add the three reconstruction matrices at the same sampling point and the same response channel to obtain the full-source reconstruction matrix; subtract the data values ​​of the same sampling point and the same response channel in the full-source reconstruction matrix from the impact response residual matrix, and add the absolute values ​​of the differences between each response channel at the same sampling point to obtain the full-source error value; subtract the compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix from the full-source reconstruction matrix to obtain the decompression matrix, defluxing matrix, and deunloading matrix; subtract the data values ​​of the same sampling point and the same response channel in the decompression matrix, defluxing matrix, and deunloading matrix from the impact response residual matrix, and add the absolute values ​​of the differences between each response channel at the same sampling point to obtain the decompression error value, defluxing error value, and deunloading error value; subtract the full-source error value from the decompression error value and divide the result by the negative... The value is set to zero to obtain the compression contribution value; the flushing contribution value is obtained by subtracting the total source error value from the defluxing error value and setting the negative value to zero; the unloading contribution value is obtained by subtracting the total source error value from the unloading error value and setting the negative value to zero; for the same sampling point, the impact source corresponding to the maximum value among the compression contribution value, flushing contribution value, and unloading contribution value is written into the dominant impact source label, and the impact source corresponding to the second largest value is written into the accompanying impact source label; for the same sampling point, the maximum contribution value is divided by the sum of the compression contribution value, flushing contribution value, and unloading contribution value and the sum of the minimum constant to obtain the attribution purity value; when the attribution purity value is less than the purity threshold, the corresponding sampling point is written into the composite impact label, and when the attribution purity value is greater than or equal to the purity threshold, the corresponding sampling point is written into the single source impact label; the counterfactual attribution result is output, which includes the compression contribution value, flushing contribution value, unloading contribution value, dominant impact source label, accompanying impact source label, composite impact label, and single source impact label.

[0016] Furthermore, based on the counterfactual attribution results, the specific steps for merging attribution fragments and decomposing the fragment influent volume into compression control volume, flushing control volume, and unloading control volume are as follows: Read the counterfactual attribution results and pre-processed infiltration shock monitoring data; merge sampling points with K consecutive sampling points that share the same dominant shock source marker and consistent composite shock marker into an attribution fragment; for each attribution fragment, divide the compression contribution value, flushing contribution value, and unloading contribution value by the sum of the three and the minimum constant, respectively, to obtain the compression control ratio, flushing control ratio, and unloading control ratio; multiply the instantaneous influent flow rate of the regulating tank at each sampling point within the attribution fragment by a uniform sampling period and sum them to obtain the fragment influent volume; multiply the fragment influent volume by the compression control ratio, flushing control ratio, and unloading control ratio, respectively, to obtain the compression control volume, flushing control volume, and unloading control volume.

[0017] Furthermore, the specific steps for outputting the intelligent control results of leachate treatment, combining the remaining storage capacity and the membrane protection status, are as follows: Read the current equalization tank volume value from the equalization tank level and volume calibration table based on the equalization tank level value; subtract the current equalization tank volume value from the equalization tank volume value corresponding to the upper safety limit level to obtain the remaining storage capacity; add the controlled water volume for compression and the controlled water volume for unloading to obtain the controlled water volume for high load; when the remaining storage capacity is greater than or equal to the controlled water volume for high load, the controlled flushing water volume is used as an allowable additional volume. For water discharge, when the remaining storage capacity is less than the high-load controlled water volume, the difference between the high-load controlled water volume and the remaining storage capacity is added to the flushing controlled water volume to obtain the permissible discharge volume; the membrane-side pressure difference is obtained by subtracting the membrane-side product water pressure from the membrane inlet pressure; when the membrane-side pressure difference is greater than the upper limit of membrane pressure difference, or the membrane system product water conductivity is greater than the upper limit of product water conductivity, the permissible discharge volume is divided by the control cycle and multiplied by the membrane protection reduction factor to obtain the protection discharge flow rate; when the membrane-side pressure difference is less than... When the permeate pressure equals the upper limit of the membrane differential pressure and the permeate conductivity of the membrane system is less than or equal to the upper limit of the permeate conductivity, the permissible discharge volume is divided by the control cycle to obtain the protected discharge flow rate. Based on the protected discharge flow rate, the booster pump control frequency value is read from the booster pump flow rate frequency calibration table. Based on the high-load controlled water volume, influent chemical oxygen demand concentration, influent suspended solids concentration, influent turbidity, influent oil concentration, and influent pH value, the target coagulant dosing flow rate and target flocculant dosing flow rate are read from the chemical dosing load calibration table. The system calculates the injection flow rate and target injection flow rate for acid and alkali reagents; it reads the target frequency value of the aeration blower from the aeration load frequency calibration table based on the controlled water volume, influent ammonia nitrogen concentration, dissolved oxygen concentration in the biological treatment tank, and oxidation-reduction potential of the biological treatment tank; and outputs the intelligent control results for leachate treatment, which include the booster pump control frequency value, target injection flow rate for coagulant, target injection flow rate for flocculant, target injection flow rate for acid and alkali reagents, target frequency value of the aeration blower, and protection release flow rate. Beneficial effects

[0018] The present invention has the following beneficial effects: (1) A smart control method for leachate treatment in a waste transfer station, which constructs the compression discharge, flushing water intake and unloading load as source-end pulse matrices respectively, and generates three types of hysteresis dictionaries according to the maximum hysteresis sampling number, so that the hysteresis process in the waste transfer station where the source-end operation occurs first and the water quality response occurs later can be expressed separately, reducing the source confusion caused by judging the impact state only based on the liquid level value and the instantaneous water quality value.

[0019] (2) A smart control method for leachate treatment in a waste transfer station, by adding water conservation residual values ​​to a grouped regression model with signed constraints, the solution of the compression discharge coefficient matrix, the flushing influent coefficient matrix and the unloading load coefficient matrix is ​​simultaneously constrained by the water quality response and the influent-outfluent relationship of the regulating tank, thus avoiding the impact source inversion shift caused by relying solely on statistical correlation and improving the credibility of the lag inversion results.

[0020] (3) A smart control method for leachate treatment in a waste transfer station, which performs source-by-source removal on the compression reconstruction matrix, flushing reconstruction matrix and unloading reconstruction matrix, and calculates the compression contribution value, flushing contribution value and unloading contribution value with error increment, so that the dominant impact source, accompanying impact source and composite impact state under the combined impact condition can be quantitatively distinguished, thereby improving the attribution stability under the scenario of overlapping compression discharge, flushing water inlet and unloading load.

[0021] (4) A smart control method for leachate treatment in a waste transfer station, which transforms counterfactual attribution results into controlled water volume for compression, controlled water volume for flushing, and controlled water volume for unloading, and combines the remaining storage volume with the membrane protection status to generate smart control results for leachate treatment, so that low-load flushing inlet water and high-load compression discharge and unloading load can obtain differentiated release, dosing and aeration control basis, and improve the matching of storage release and back-end treatment operation. Attached Figure Description

[0022] Figure 1 This is a flowchart of an intelligent control method for leachate treatment at a waste transfer station. Figure 2 Flowchart for generating source pulse hysteresis dictionary; Figure 3 This is a diagram showing the distribution of water volume to be controlled. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see Figures 1-3 This invention provides a technical solution: an intelligent control method for leachate treatment in a waste transfer station, comprising: S1, collecting leachate impact monitoring data, performing anomaly removal, missing data completion, and time alignment processing on the leachate impact monitoring data, and outputting pre-processed leachate impact monitoring data; S2, identifying candidate impact response intervals based on the pre-processed leachate impact monitoring data, generating an impact response residual matrix and a source-end pulse matrix, performing hysteresis shift on the source-end pulse matrix according to the maximum hysteresis sampling number, and generating three types of hysteresis corresponding to compression discharge, flushing inlet, and unloading loads. S3. Based on the lag inversion results, the compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix are generated. The three types of contribution values ​​are obtained by removing sources one by one and calculating error increments. Impact source attribution labels are generated and counterfactual attribution results are output. S4. Based on the counterfactual attribution results, the attribution segments are merged. The influent volume of the segment is decomposed into the controllable volume of compression, the controllable volume of flushing, and the controllable volume of unloading. The intelligent control results of leachate treatment are output by combining the remaining storage volume and the membrane segment protection status.

[0025] Specifically, the steps for collecting infiltration shock monitoring data and performing anomaly removal, missing data completion, and time alignment processing on the infiltration shock monitoring data are as follows: The infiltration shock monitoring data includes vehicle entry timestamp, vehicle unloading mass value, compressor start timestamp, compressor stop timestamp, compressor operating current value, compressor hydraulic pressure value, flushing pump start timestamp, flushing pump stop timestamp, flushing water instantaneous flow rate value, equalization tank level value, equalization tank influent instantaneous flow rate value, equalization tank effluent instantaneous flow rate value, influent chemical oxygen demand concentration value, influent ammonia nitrogen concentration value, influent suspended solids concentration value, influent turbidity value, influent pH value, influent conductivity value, influent grease concentration value, and biological treatment tank data. Dissolved oxygen concentration, oxidation-reduction potential of the biological treatment tank, inlet water pressure of the membrane, post-membrane permeate pressure, and permeate conductivity of the membrane system; among these, the vehicle entry timestamp is collected through the vehicle identification gate, and is used to characterize the recording time of the source-side unloading event corresponding to the vehicle unloading mass value. The vehicle unloading mass value is collected through the weighbridge terminal, and is obtained by subtracting the vehicle exiting mass value from the vehicle entry weighing value. The compressor start timestamp, compressor stop timestamp, and compressor operating current value are collected through the compressor control cabinet. The compressor hydraulic pressure value is collected through a pressure transmitter installed on the compressor hydraulic pipeline. The flushing pump start timestamp and flushing pump stop timestamp are collected through the flushing pump control cabinet. The instantaneous flow rate of the wash water is collected by an electromagnetic flow meter in the flushing water pipeline; the liquid level in the equalization tank is collected by an ultrasonic level meter in the equalization tank; the instantaneous influent and effluent flow rates of the equalization tank are collected by electromagnetic flow meters in the equalization tank inlet and effluent pipelines, respectively; the influent chemical oxygen demand (COD), ammonia nitrogen, suspended solids, turbidity, pH, conductivity, and oil concentration are collected by an online water quality analyzer at the equalization tank inlet; the dissolved oxygen concentration in the biological treatment tank is collected by a dissolved oxygen sensor in the biological treatment tank; the oxidation-reduction potential (ORP) in the biological treatment tank is collected by an ORP sensor in the biological treatment tank; and the inlet pressure and outlet pressure of the membrane are collected by an online water quality analyzer at the membrane inlet. The pressure transmitters before and after the membrane collect data, and the conductivity value of the membrane system's permeate is collected by the conductivity sensor at the permeate end of the membrane system. The data acquisition period for continuous numerical instruments is set to 10s to 60s, preferably 30s. The original output period of the online water quality analyzer is 60s to 300s, preferably 120s. The water quality data output by the online water quality analyzer is entered into a unified sampling time sequence according to the original output period. The missing unified sampling time between two adjacent original output times is filled in by a piecewise linear interpolation algorithm. The vehicle entry time stamp, compressor start time stamp, compressor stop time stamp, flushing pump start time stamp, and flushing pump stop time stamp are recorded using an event-triggered method with a time resolution of not less than 1s.The unit for vehicle unloading mass is kilogram; the unit for compressor operating current is ampere; the unit for compressor hydraulic pressure, membrane inlet water pressure, and membrane outlet permeate pressure is kilopascal; the unit for flushing water instantaneous flow rate, equalization tank inlet instantaneous flow rate, and equalization tank effluent instantaneous flow rate is cubic meter per hour; the unit for equalization tank level is meter; the units for influent chemical oxygen demand (COD), ammonia nitrogen, suspended solids, oil, and dissolved oxygen concentrations in the biological treatment tank are milligrams per liter; the unit for influent turbidity is turbidity unit; the unit for influent pH is dimensionless; the units for influent conductivity and membrane system permeate conductivity are microsiemens per centimeter; and the unit for biological treatment tank oxidation-reduction potential is milligrams per centimeter. After data acquisition, the continuous numerical instrument data are sorted in ascending order according to the data acquisition timestamp, and a one-to-one correspondence is established based on the equipment number, sampling channel number, and data acquisition timestamp, so that the liquid level, flow rate, water quality, pressure, and biochemical operating status at the same unified sampling time can be read synchronously. For the acquired infiltration shock monitoring data, the Hample filter algorithm is used to perform instrument instantaneous jump value identification and abnormal sampling point removal processing on the continuous numerical instrument data in the infiltration shock monitoring data. Specifically, a sliding detection window is set for the continuous numerical instrument data, with a sliding detection window length of 5 to 11 sampling points, preferably 7 sampling points, and the sliding detection window covers no less than 2 minutes and The sampling duration is no more than 10 minutes; within each sliding detection window, the absolute deviation between the median and the middle value of the window is calculated, and the absolute value of the difference between the current sampling point value and the median value of the window is divided by the absolute deviation of the median to obtain the instantaneous jump discrimination value; when the instantaneous jump discrimination value is greater than the jump threshold, the current sampling point is marked as an abnormal sampling point and removed from the corresponding sampling channel; the jump threshold ranges from 3 to 5, preferably 3.5, and is determined based on continuous numerical instrument data from no less than 7 historical stable operating days. Specifically, the instantaneous jump discrimination value distribution of each sampling channel is calculated, and the larger value between the 99th percentile and 3 is used as the jump threshold; vehicle entry timestamp, compressor start-up The timestamps, compressor stop timestamps, flushing pump start timestamps, and flushing pump stop timestamps are used as event timestamp data. They are not involved in the numerical elimination calculation of the Hanpuer filtering algorithm, but only in the subsequent unified sampling time mapping. A piecewise linear interpolation algorithm is used to perform short-term missing record completion and unified sampling time alignment processing on the continuous numerical instrument data after the abnormal sampling point elimination processing is completed. Specifically, a unified sampling time sequence is first established based on the unified sampling period. The unified sampling period is 10s to 60s, preferably 30s. The unified sampling period is determined according to the acquisition period of the instantaneous flow rate of the regulating tank inlet, the original output period of the online water quality analyzer, and the control execution period, and a sampling period not greater than the control execution period is selected.Next, missing intervals are identified in the continuous numerical instrument data after the abnormal sampling point removal process. When the length of the missing interval is less than the short-term missing length threshold, the previous and next valid sampling points of the missing interval are read, and a linear interpolation relationship is established based on the time difference and numerical difference between the two valid sampling points. The interpolation result is written into the unified sampling time within the missing interval. The short-term missing length threshold is 2 to 6 unified sampling periods, preferably 3 unified sampling periods. The short-term missing length threshold is determined based on the communication buffer packet loss duration, the online instrument response duration, and the unified sampling period, so that short-term communication interruptions can be filled in and long-term missing channels do not participate in the subsequent candidate impulse response interval identification and impulse response residual matrix construction. When the length of the missing interval is greater than or equal to the short-term missing length threshold... When the sampling channel is missing for a long period, no interpolation completion value is generated. For the vehicle arrival time stamp, compressor start time stamp, compressor stop time stamp, flushing pump start time stamp, and flushing pump stop time stamp, the unified sampling time with the smallest absolute time difference is found, and the corresponding event record is mapped to the unified sampling time. When the same unified sampling time corresponds to multiple vehicle arrival time stamps, the unloading quality value of the corresponding vehicles is accumulated and written to the unified sampling time, so that the vehicle unloading quality value, compressor operating current value, compressor hydraulic pressure value, and flushing water instantaneous flow rate value can be written to the source pulse matrix according to the unified time index. After completing the abnormal sampling point removal, short-term missing record completion, and unified sampling time alignment processing, the preprocessed infiltration impact monitoring data is output.

[0026] In this implementation plan, by limiting the data acquisition sources, measurement units, sampling cycles, event timestamp mapping, abnormal sampling point removal, short-term missing record completion, and unified sampling time alignment of the infiltration shock monitoring data, the vehicle unloading mass value, compressor operating current value, compressor hydraulic pressure value, flushing water instantaneous flow rate value, equalization tank liquid level value, equalization tank influent instantaneous flow rate value, water quality concentration data, and membrane pressure data can form a computable continuous data foundation under the same time index. This reduces the impact of instrument glitch, communication interruption, and asynchronous multi-source sampling on the identification of candidate shock response intervals, and provides time-consistent, source-clear, and numerically reliable input data for subsequent source pulse matrix construction, shock response residual matrix generation, and hysteresis inversion result solution.

[0027] Specifically, the steps for identifying candidate impact response intervals based on pre-processed infiltration impact monitoring data are as follows: Read the pre-processed infiltration impact monitoring data and assemble a multi-channel response sequence from the following values: equalization tank level, instantaneous influent flow rate, influent chemical oxygen demand (COD) concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, influent turbidity, influent pH, influent conductivity, and influent oil concentration. The multi-channel response sequence uses a unified sampling time as the row index and the equalization tank level, instantaneous influent flow rate, COD concentration, ammonia nitrogen concentration, suspended solids concentration, turbidity, pH, conductivity, and oil concentration as the response channels. A nine-dimensional response vector is formed at a unified sampling time. Before performing change point identification, M consecutive sampling points are read for each response channel to form a detection reference window, where M is between 10 and 30, preferably 20. The length of the detection reference window is determined based on the unified sampling period and the duration of one compression and discharge at the waste transfer station, ensuring that the detection reference window covers a response background of not less than 5 minutes and not more than 15 minutes. The median and median absolute deviation of each response channel within the detection reference window are calculated respectively. The data value of the corresponding response channel at each unified sampling time is subtracted from the median of the same response channel, and then divided by the sum of the median absolute deviation and the minimum constant of the same response channel to obtain the response vector for change point detection. The minimum constant is taken as... to Preferred To avoid the divisor failing due to a median absolute deviation of 0, a Bayesian online change point detection algorithm is used to identify change points in the multi-channel response sequence. Specifically, the response vector for change point detection is input into the Bayesian online change point detection algorithm in the order of uniform sampling time. The posterior probability of the current uniform sampling time continuing the previous running length and the posterior probability of the current uniform sampling time changing are calculated using a multivariate Gaussian observation model. The multivariate Gaussian observation model uses a normal-inverse Wieshard conjugate prior initialization. The prior mean vector is taken as the mean vector of the response vector for change point detection within the historical stable running days. The prior scale matrix is ​​taken as the covariance matrix of the response vector for change point detection within the historical stable running days. The prior degrees of freedom are the number of response channels plus 2. The prior strength coefficient is 1 to 5, preferably 2. The historical stable running days are no less than 7 natural days. A constant hazard function is used to control the prior probability of change points. The value of the constant hazard function is 1 / L, where L is 20 to 80, preferably 40. L is determined based on the median number of sampling points in adjacent impact response intervals within the historical morning peak collection period. When the current uniform sampling... When the posterior probability of a change in the sampling time is greater than the change point probability threshold, the current unified sampling time is written into the change point set. The change point probability threshold is between 0.60 and 0.85, preferably 0.70. The change point probability threshold is determined based on the manually labeled start and end times of the impact over at least 7 historical operating days. Specifically, candidate thresholds within the range of 0.60 to 0.85 are traversed, and the candidate threshold that maximizes the weighted sum of the detection rate of the manually labeled impact start point and the false detection rejection rate during non-impact periods is selected as the change point probability threshold. When the interval between two adjacent change points is less than the minimum interval length, change points with higher posterior probabilities are retained and change points with lower posterior probabilities are deleted. The minimum interval length is between 3 and 8 unified sampling periods, preferably 5 unified sampling periods, to avoid instrument residual jitter being divided into multiple excessively short response intervals. The start and end unified sampling times of the multi-channel response sequence are written into the change point set as interval boundaries, and the change point set is arranged in ascending order of unified sampling time. The sampling interval between two adjacent change points is determined as the candidate impact response interval.

[0028] In this implementation scheme, by unifying the liquid level value of the equalization tank, the instantaneous flow rate value of the equalization tank influent, and the relevant data of the influent water quality into a multi-channel response sequence, and identifying the change points based on a unified sampling time, the candidate impact response interval can simultaneously reflect the common abrupt change characteristics of water volume change, water quality concentration change, and oil load change. This reduces missegmentation caused by single-channel fluctuations, improves the stability of the impact response boundary under the scenario of overlapping compression discharge, flushing influent, and unloading loads, and provides a clear and temporally continuous response interval foundation for subsequent impact response residual matrix construction, source-end pulse matrix alignment, and hysteresis inversion result solution.

[0029] Specifically, the steps for generating the impact response residual matrix and the source-end pulse matrix are as follows: For each candidate impact response interval, read N consecutive sampling points before the start of the candidate impact response interval, and calculate the median of each data point in the multi-channel response sequence to obtain the interval baseline vector; where N is between 10 and 30, preferably 20, and N is determined based on the unified sampling period and the duration of stable influent background at the waste transfer station, so that the N consecutive sampling points cover the interval background for no less than 5 minutes and no more than 15 minutes; when reading the N consecutive sampling points, first check whether there is a long-term missing channel marker in each response channel. When any response channel has a long-term missing channel marker in the candidate impact response interval and the interval background, the corresponding candidate impact... The response interval is used to construct the impact response residual matrix. When the number of valid sampling points before the starting point of the candidate impact response interval is less than N, all valid sampling points before the starting point of the candidate impact response interval are used as baseline sampling points, and the number of baseline sampling points is written into the interval baseline vector record. The median is used as the interval baseline value because the instantaneous flow rate, chemical oxygen demand (COD), ammonia nitrogen, suspended solids, turbidity, and oil concentration of the regulating tank influent are prone to isolated peaks before and after the morning peak collection. The median can reduce the pull of isolated peaks on the interval background level, so that the subsequent residuals reflect the true deviation of the candidate impact response interval from the background before the interval. The sampling points within the candidate impact response interval are used as the matrix rows. Using the data from the multi-channel response sequence as matrix columns, the baseline value corresponding to the same matrix column in the interval baseline vector is subtracted from the data value corresponding to each matrix column at each sampling point to obtain the impact response residual matrix. Each row of the impact response residual matrix corresponds to a unified sampling time, and each column corresponds sequentially to the equalization tank level, the instantaneous influent flow rate, the influent chemical oxygen demand (COD) concentration, the influent ammonia nitrogen concentration, the influent suspended solids concentration, the influent turbidity, the influent pH value, the influent conductivity value, and the influent oil concentration. Matrix elements represent the deviation of the corresponding unified sampling time and the corresponding response channel from the interval baseline vector. When the influent pH value is involved in the residual calculation, it is directly used as the acid-base response channel, and the baseline value is subtracted in the same way. The baseline value corresponding to the influent pH value in the interval baseline vector allows the direction of acid-base change and the water quality load change to participate in subsequent inversion in the same residual matrix. The source pulse matrix is ​​constructed with the unified sampling time as the row index. The product of the compressor operating current value and the compressor hydraulic pressure value within the compressor start time stamp to the compressor stop time stamp is written into the compressor discharge column. The instantaneous flow rate value of the flushing water within the flushing pump start time stamp to the flushing pump stop time stamp is written into the flushing inlet column. The vehicle unloading mass value corresponding to the vehicle entry time stamp is written into the unloading load column. The remaining matrix positions are written as 0. Among them, the row index of the source pulse matrix and the impact response residual matrix adopt the same unified sampling time sequence. The column indexes are the compression discharge column, the flushing inlet column, and the unloading load column, respectively.In the compression and discharge column, the compressor operating range is first determined based on the compressor start-up and stop timestamps. Then, the compressor operating current and hydraulic pressure values ​​corresponding to each uniform sampling time within the operating range are read, and their product is written as the compression and discharge intensity value into the compression and discharge column. The compression and discharge intensity value is used to characterize the discharge driving force under the combined action of compressor load and hydraulic compression intensity. The compression and discharge intensity value is not directly equivalent to the actual compression and discharge volume. In specific implementation, a calibration relationship between the compression and discharge intensity value and the measured compression and discharge volume is established based on historical compression operation records. The historical compression operation records include the compressor operating current and compressor hydraulic pressure values ​​within the same compression operation period. The measured compressed liquid discharge volume is obtained from the integral value of the flow rate in the collection pipeline corresponding to the compression operation period. The calibration relationship is established using a least squares regression model, with the product of the compressor operating current value and the compressor hydraulic pressure value as the independent variable and the measured compressed liquid discharge volume as the dependent variable, to obtain the calibration coefficient and calibration intercept of the compressed liquid discharge volume. When the compressed liquid discharge intensity value is written into the compressed liquid discharge column, the compressed liquid discharge column is used to express the compressed liquid discharge impact intensity characterized by the compressor operating current value and the compressor hydraulic pressure value, and checked by the compressed liquid discharge volume calibration coefficient and calibration intercept, rather than directly limiting the product of the compressor operating current value and the compressor hydraulic pressure value to the discharge volume. In the flushing water inlet column... The instantaneous flow rate of flushing water corresponding to each unified sampling time from the flushing pump start time to the flushing pump stop time is written into the flushing water inlet column, and the position of the flushing water inlet column matrix after the flushing pump stop time is written as 0; in the unloading load column, the vehicle entry time is mapped to the most recent unified sampling time, and the corresponding vehicle unloading quality value is written into the unloading load column. When there are multiple vehicle entry time stamps for the same unified sampling time, the corresponding vehicle unloading quality values ​​are accumulated and written into the unloading load column; the vehicle unloading quality value corresponding to the vehicle entry time stamp is written into the unloading load column in the form of an event pulse because the duration of unloading after the garbage transport vehicle enters the station lags behind the pipeline transmission and the mixing in the regulating pool The lag and water quality response lag are shorter, and the vehicle unloading mass value is represented as a concentrated load event at the source input. In specific implementation, the single-point event pulse corresponding to the vehicle unloading mass value is not directly used as the water quality response occurrence time. Instead, the unloading load column is successively shifted backward according to the number of lag samples from 0 to the maximum lag, forming an unloading load lag dictionary. This allows the vehicle unloading mass value to participate in the grouped lasso regression model solution under multiple lag samples, thereby characterizing the delayed impact of the unloading event on the equalization tank level, equalization tank influent instantaneous flow rate, influent chemical oxygen demand concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, influent turbidity, influent pH, influent conductivity, and influent oil concentration.The source-side pulse matrix uses three columns of data to express the source-side intensity before the compressed discharge, flushing influent, and unloading load enter the equalization tank. This allows subsequent hysteresis translation to establish a time correspondence between the source-side operation process and the water quantity, water quality, and grease response within the candidate impact response interval.

[0030] In this implementation scheme, the multi-channel response sequence within the candidate impact response interval is converted into an impact response residual matrix relative to the interval baseline vector. The compressor operating current value and the compressor hydraulic pressure value are combined to form the compression discharge intensity value, flushing water instantaneous flow rate value, and vehicle unloading mass value, which are then uniformly written into the source pulse matrix. This allows the response deviations of the regulating tank level value, regulating tank influent instantaneous flow rate value, influent chemical oxygen demand concentration value, influent ammonia nitrogen concentration value, influent suspended solids concentration value, influent turbidity value, influent pH value, influent conductivity value, and influent grease concentration value to establish the same time index relationship with the source-side intensity of compression discharge, flushing water influent, and unloading load. This reduces the interference of interval background level differences on impact response identification and provides a matrix-based data foundation for subsequent maximum hysteresis sampling number shifting, three-type hysteresis dictionary generation, and signed constraint grouped regression solution, where the response side and source side correspond to each other.

[0031] Specifically, the steps for generating three types of hysteresis dictionaries corresponding to the compression discharge, flushing water inlet, and unloading loads by performing a hysteresis shift on the source pulse matrix based on the maximum hysteresis sampling number are as follows: Figure 2As shown, the ratio of the pipeline volume from the inlet of the regulating tank to the water quality sampling point to the instantaneous flow rate of the regulating tank inlet is calculated, as is the ratio of the effective volume of the regulating tank to the instantaneous flow rate of the regulating tank inlet. Both ratios are divided by a unified sampling period and rounded up to obtain two lag sampling numbers. The larger of the two lag sampling numbers is taken as the maximum lag sampling number. The pipeline volume from the inlet of the regulating tank to the water quality sampling point is calculated based on the inner diameter of the regulating tank inlet pipeline, the pipeline length, and the pipeline bend correction factor. The pipeline bend correction factor is between 1.05 and 1.30, preferably 1.15. The effective volume of the regulating tank is obtained by subtracting the volume corresponding to the lowest operating liquid level and the volume corresponding to the upper safety limit liquid level from the regulating tank liquid level volume calibration table. The pipeline volume from the inlet of the regulating tank to the water quality sampling point... The units for both the pipeline volume and the effective volume of the regulating tank are cubic meters. The instantaneous influent flow rate of the regulating tank is converted to cubic meters per second before being included in the ratio calculation, and the unified sampling period is converted to seconds. The units for the ratio of pipeline volume to the instantaneous influent flow rate of the regulating tank and the ratio of the effective volume of the regulating tank to the instantaneous influent flow rate of the regulating tank are seconds. The two ratios are divided by the unified sampling period to obtain the dimensionless number of sampling points, so that the two lag sampling numbers represent the number of sampling points in the unified sampling time sequence. When the instantaneous influent flow rate of the regulating tank is less than the minimum effective influent flow rate threshold, the most recent instantaneous influent flow rate of the regulating tank that is greater than the minimum effective influent flow rate threshold before the current unified sampling time is read and included in the calculation. The minimum effective influent flow rate threshold is taken from 0.05 m³ / h to 0.20 m³ / h, preferably 0.The minimum effective influent flow rate threshold of 10 m³ / h is determined based on the lower limit of the range of the electromagnetic flowmeter in the regulating tank influent pipeline and the historical average noise level of the empty pipe, to avoid abnormal amplification of the maximum lag sampling number when the instantaneous influent flow rate of the regulating tank approaches zero. An upper limit for the maximum lag sampling number is set, ranging from 20 to 80, preferably 40. This upper limit is determined based on the median interval between the sampling points from the vehicle entry time stamp, compressor start time stamp, flushing pump start time stamp to the point of change in the multi-channel response sequence during the historical morning peak collection period. When the calculated maximum lag sampling number is greater than the upper limit, the upper limit is used as the maximum lag sampling number. The three columns of data in the source pulse matrix are sequentially shifted backward from 0 to the maximum lag sampling number, and the shifted columns corresponding to each lag sampling number are concatenated to form the compression discharge lag dictionary, flushing influent lag dictionary, and unloading load lag dictionary. Specifically, when the lag sampling number is 0, the original time sequence position of the corresponding column in the source pulse matrix is ​​retained. When the number of lag samples is r, each row of data in the corresponding column is shifted backward by r unified sampling times. Data that exceeds the unified sampling time of the candidate impact response interval after shifting is not written into the dictionary, and the matrix position vacated at the beginning position after shifting is written with 0. The compression discharge lag dictionary is obtained by splicing all the shifted columns formed by the compression discharge column under 0 to the maximum number of lag samples in ascending order of lag sample number. The flushing water lag dictionary is obtained by splicing all the shifted columns formed by the flushing water column under 0 to the maximum number of lag samples in ascending order of lag sample number. The unloading load lag dictionary is obtained by splicing all the shifted columns formed by the unloading load column under 0 to the maximum number of lag samples in ascending order of lag sample number. The row index of the three types of lag dictionaries is consistent with the row index of the impact response residual matrix. Each column of the three types of lag dictionaries corresponds to a source-side intensity sequence under a specific number of lag samples, so that the transmission delay of the compression discharge, flushing water, and unloading load from before entering the regulating tank to the water quality sampling point can participate in the subsequent constraint inversion through different lag columns. .

[0032] In this implementation scheme, by uniformly converting pipeline transmission delay, equalization tank mixing delay, and source-side intensity time series into three types of lag dictionaries arranged according to a unified sampling time, the response propagation process of compression discharge, flushing inlet, and unloading load after entering the equalization tank can be expressed in the form of a calculable lag column. This avoids the time mismatch caused by direct alignment of the source-end pulse matrix and the impact response residual matrix, reduces the interference of the instantaneous inlet flow rate of the equalization tank being close to zero, the transmission distance difference, and the mixing effect of the tank volume on the lag estimation, and provides a basis for explanatory variables with clear time delay boundaries and source-side type separation for subsequent water conservation residual value constraints and signed constraint grouped regression models.

[0033] Specifically, the steps for outputting the lag inversion results by combining the water conservation residual value and the signed-constrained grouped regression model are as follows: Based on the liquid level value of the regulating tank, the corresponding regulating tank volume value is read from the regulating tank level-volume calibration table. The regulating tank level-volume calibration table is established through verification of on-site geometric measurements, segmented water injection calibration, and as-built measurement data. The table fields include the regulating tank level value and the regulating tank volume value. Specifically, the on-site geometric measurements are used to obtain the tank length, tank width, tank bottom elevation, and effective water depth; the segmented water injection calibration is used to record... The liquid level and cumulative injection volume of the equalization tank at different water injection stages, along with the as-built measurement data, are used to verify the tank's structural dimensions. When the liquid level in the equalization tank is between two adjacent calibrated levels, linear interpolation is used to calculate the corresponding equalization tank volume. The change in tank volume is obtained by subtracting the volume of the equalization tank at the previous sampling point from the volume of the equalization tank at the later sampling point. The theoretical net inflow rate is obtained by subtracting the instantaneous inflow rate of the equalization tank from the instantaneous outflow rate of the equalization tank and multiplying the result by a uniform sampling period. The instantaneous inflow rate and the instantaneous outflow rate of the equalization tank are used to calculate the theoretical net inflow rate. The unit of the flow rate value is cubic meters per hour. The sampling period is converted to hours before calculation to ensure that the unit of the theoretical net inflow and the change in volume of the pool is cubic meters. The change in volume of the pool is subtracted from the theoretical net inflow to obtain the water conservation residual value, which is used to characterize the deviation between the integral result of the inflow and outflow of the regulating pool and the result of the change in liquid level and volume. The lag dictionaries of compression discharge, flushing inflow, and unloading load are used as explanatory variables in the nested regression model, and the impact response residual matrix is ​​used as the matrix to be explained. The residual values ​​of water conservation are added to the objective function of the nested regression model. The alternating direction multiplier method is used to solve the signed constraint nested regression model, obtaining the compression discharge coefficient matrix, flushing inlet coefficient matrix, and unloading load coefficient matrix. Specifically, the compression discharge lag dictionary, flushing inlet lag dictionary, and unloading load lag dictionary are concatenated column-wise to form the regression input matrix, making the row index of the regression input matrix consistent with the row index of the impact response residual matrix; where, it is assumed that the candidate impact response interval contains T uniform sampling times, and the impact response residual matrix... The dimension is T×P, where P represents the number of response channels. In this embodiment, P is 9, corresponding to the liquid level in the regulating tank, the instantaneous flow rate of the influent to the regulating tank, the concentration of chemical oxygen demand in the influent, the concentration of ammonia nitrogen in the influent, the concentration of suspended solids in the influent, the turbidity of the influent, the pH value of the influent, the conductivity value of the influent, and the concentration of oil in the influent, respectively. Let the maximum number of hysteresis samples be R, then the compression discharge hysteresis dictionary... , Flushing water inlet delay dictionary Dictionary of unloading load lag The dimension of each matrix is ​​T×(R+1), and the compression and drainage coefficient matrix is... Flushing inlet water coefficient matrix and unloading load coefficient matrix The dimension of each is (R+1)×P, making , and All dimensions are T×P, and can be correlated with the shock response residual matrix. Element-by-element difference calculations are performed. The objective function of the grouped-loop regression model is a weighted sum of the residual squared terms, the grouped-loop penalty term, and the water conservation constraint term. The residual squared term is the sum of the squared differences between the impact response residual matrix and the products of the compression discharge hysteresis dictionary and the compression discharge coefficient matrix, the flushing inlet hysteresis dictionary and the flushing inlet coefficient matrix, and the unloading load hysteresis dictionary and the unloading load coefficient matrix. The grouped-loop penalty term is the sum of the grouped-loop penalty coefficients multiplied by the sum of the L2 norms of the coefficient groups corresponding to each impact source. The water conservation constraint term is the sum of the water conservation constraint coefficients multiplied by the sum of the absolute values ​​of the water conservation residuals at each unified sampling time after normalization by the water residual scaling value. The objective function is as follows: ;in, Represents the residual matrix of the impact response. A dictionary representing the hysteresis of compression and drainage. Dictionary indicating lag in flushing water inlet. Dictionary indicating unloading load lag. This represents the compression and discharge coefficient matrix. This represents the flushing inlet water coefficient matrix. This represents the unloading load coefficient matrix. This represents the coefficient group corresponding to the same lag sample number in the compression discharge lag dictionary. This represents the coefficient group corresponding to the same hysteresis sample number in the flushing inlet hysteresis dictionary. This represents the coefficient group corresponding to the same lag sample number in the unloading load lag dictionary. , , These represent the sets of coefficients corresponding to the compression discharge lag dictionary, the flushing water inlet lag dictionary, and the unloading load lag dictionary, respectively. This represents the water conservation residual value at the t-th uniform sampling time. This represents the scaled value of the water quantity residual. This represents the group lasso penalty coefficient. The water conservation constraint coefficient is used; the water residual scale value is taken as the median of the absolute values ​​of the water conservation residual values ​​within the historical stable operating days, which is used to convert the water conservation residual values ​​into dimensionless residual quantities; the grouping cable penalty coefficient is taken from 0.01 to 0.20, preferably 0.08, and is determined based on manually labeled impact source data of no less than 7 historical operating days. Specifically, the candidate values ​​are traversed at intervals of 0.01 within the range of values, and each candidate value is substituted into the objective function to solve the problem. The candidate value that maximizes the weighted sum of the identification accuracy of the compression discharge, flushing water inlet, and unloading load is selected as the grouping cable penalty coefficient; the water conservation constraint coefficient is taken from 0.10 to 1. The water conservation constraint coefficient is preferably 0.40. It is determined through cross-validation using historical operating day data. Specifically, candidate values ​​are iterated at 0.05 intervals within the range of 0.10 to 1.00. Each candidate value is substituted into the objective function for solution. The candidate value that maximizes the weighted sum of the accuracy of manually labeled impact source identification and the water conservation residual suppression rate is selected as the water conservation constraint coefficient. When using the alternating direction multiplier method, the compression discharge coefficient matrix, flushing inlet coefficient matrix, and unloading load coefficient matrix are first initialized to zero. Then, the coefficient matrices, symbolic projection variables, and Lagrange multipliers are alternately updated until the change in the objective function between two adjacent iterations is less than 0.40. And the original residual norm is less than The maximum number of iterations is set to 200. Specifically, the compression discharge coefficient matrix is ​​limited to non-negative values ​​in the corresponding channels for equalization tank level, equalization tank influent instantaneous flow rate, influent chemical oxygen demand (COD) concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, and influent turbidity. The flushing influent coefficient matrix is ​​limited to non-negative values ​​in the corresponding channels for equalization tank level and equalization tank influent instantaneous flow rate, and also limited to non-positive values ​​in the corresponding channels for influent COD concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, influent turbidity, and influent conductivity. The unloading load coefficient matrix is ​​limited to non-positive values ​​in the corresponding channels for influent grease concentration, influent COD concentration, and influent COD concentration. The turbidity values ​​and influent turbidity values ​​in the corresponding channels are limited to non-negative values; the sign constraint is achieved through projection operation in the alternating direction multiplier method. When the coefficients limited to non-negative values ​​are less than 0, they are set to 0, and when the coefficients limited to non-positive values ​​are greater than 0, they are set to 0, so that the water quantity and water quality response directions corresponding to the compression discharge, flushing influent, and unloading load are consistent with the physical mechanism of the leachate treatment process of the waste transfer station; the output lag inversion results include the impact response residual matrix, the lag dictionary of compression discharge, the lag dictionary of flushing influent, the lag dictionary of unloading load, the coefficient matrix of compression discharge, the coefficient matrix of flushing influent, and the coefficient matrix of unloading load.

[0034] In this implementation scheme, by incorporating the equalization tank level and volume calibration table, water conservation residual values, three types of hysteresis dictionaries, and the impact response residual matrix into a grouped regression model with signed constraints, the hysteresis response relationships corresponding to compression discharge, flushing influent, and unloading loads are simultaneously constrained by water balance, source grouping sparsity, and water quality response direction. This reduces the confusion of impact sources caused by simply relying on the correlation of multi-channel responses, and enables the compression discharge coefficient matrix, flushing influent coefficient matrix, and unloading load coefficient matrix to more stably represent the actual impact of different impact sources on the equalization tank level, the instantaneous flow rate of the equalization tank influent, and the influent water quality related data. This provides a more physically consistent hysteresis inversion basis for subsequent source-by-source removal, error increment calculation, and counterfactual attribution results generation.

[0035] Specifically, the steps for generating the compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix based on the hysteresis inversion results are as follows: Read the hysteresis inversion results, perform matrix multiplication between the compression discharge hysteresis dictionary and the compression discharge coefficient matrix to obtain the compression reconstruction matrix; where each row of the compression discharge hysteresis dictionary corresponds to a uniform sampling time, and each column corresponds to a compression discharge intensity sequence under a hysteresis sampling number; each row of the compression discharge coefficient matrix corresponds to a hysteresis column in the compression discharge hysteresis dictionary, and each column corresponds to a response channel in the impact response residual matrix; the number of columns in the compression discharge hysteresis dictionary... The compression reconstruction matrix, obtained after matrix multiplication, has the same number of rows and columns as the compression discharge coefficient matrix. The elements in the compression reconstruction matrix represent the deviation in the reconstruction response caused by compression discharge at the corresponding unified sampling time and within the corresponding response channel. The flushing inlet hysteresis dictionary and the flushing inlet coefficient matrix are multiplied to obtain the flushing reconstruction matrix. Each row of the flushing inlet hysteresis dictionary corresponds to a unified sampling time, and each column corresponds to a sequence of instantaneous flushing water flow rates under a given hysteresis sampling number. Each row of the flushing inlet coefficient matrix corresponds to the flushing inlet... A hysteresis column in the hysteresis dictionary corresponds to a response channel in the impact response residual matrix. The flushing reconstruction matrix obtained after matrix multiplication represents the deviation of the reconstructed response caused by the flushing influent to the equalization tank level, instantaneous influent flow rate, influent chemical oxygen demand concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, influent turbidity, influent pH, influent conductivity, and influent grease concentration. The unloading load hysteresis dictionary and the unloading load coefficient matrix are multiplied to obtain the unloading reconstruction matrix. Each row of the unloading load hysteresis dictionary corresponds to a unified sample. At any given time, each column corresponds to a sequence of vehicle unloading mass values ​​under a lag sampling number. Each row of the unloading load coefficient matrix corresponds to a lag column in the unloading load lag dictionary, and each column corresponds to a response channel in the impact response residual matrix. The unloading reconstruction matrix obtained after matrix multiplication is used to represent the deviation of the reconstruction response formed by the unloading load on each response channel. The compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix all use the row index and column index of the impact response residual matrix, so that the three types of reconstruction matrices can be added, subtracted, and error increment calculated under the same unified sampling time and the same response channel.

[0036] In this implementation scheme, by converting the compression discharge hysteresis dictionary, flushing influent hysteresis dictionary, and unloading load hysteresis dictionary into compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix, respectively, the impact of compression discharge, flushing influent, and unloading load on each response channel can be quantified and expressed separately at a unified sampling time. This avoids the problem of difficulty in distinguishing the three types of impact sources after they are mixed and superimposed in the impact response residual matrix. It also enables the response deviations of the regulating tank level, regulating tank influent instantaneous flow rate, influent chemical oxygen demand concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, influent turbidity, influent pH, influent conductivity, and influent grease concentration to be separated by source. This provides a directly comparable reconstructed response basis for subsequent source-by-source removal, full-source error value calculation, compression removal error value calculation, flushing removal error value calculation, unloading removal error value calculation, and generation of the three types of contribution values.

[0037] Specifically, the following steps are taken to obtain three types of contribution values ​​through source-by-source removal and error increment calculation, generate impact source attribution labels, and output counterfactual attribution results: The three reconstruction matrices are added together at the same sampling point and the same response channel to obtain the full-source reconstruction matrix; The three reconstruction matrices include a compression reconstruction matrix, a flushing reconstruction matrix, and a discharge reconstruction matrix. The row index of the full-source reconstruction matrix is ​​consistent with the unified sampling time of the impact response residual matrix, and the column index of the full-source reconstruction matrix is ​​consistent with the response channel of the impact response residual matrix. Each element in the full-source reconstruction matrix represents the deviation of the reconstruction response formed on the corresponding response channel under the combined action of compression discharge, flushing inlet, and discharge loads; The impact response residual matrix is ​​then added together with the full-source reconstruction matrix at the same sampling point, The data values ​​of the same response channel are subtracted, and the absolute values ​​of the differences between the response channels at the same sampling point are added together to obtain the total source error value. The total source error value is calculated point-by-point at a unified sampling time. The response channels include the level of the regulating tank, the instantaneous flow rate of the regulating tank influent, the concentration of chemical oxygen demand (COD) in the influent, the concentration of ammonia nitrogen in the influent, the concentration of suspended solids in the influent, the turbidity of the influent, the pH value of the influent, the conductivity value of the influent, and the concentration of oil in the influent. The total source error value is used to represent the interpretation deviation of the impact response residual matrix when the three impact sources are retained. The compression reconstruction matrix, the flushing reconstruction matrix, and the unloading reconstruction matrix are subtracted from the total source reconstruction matrix to obtain the decompression matrix, the defluxing matrix, and the deunloading matrix. The decompression matrix represents the result of retaining the flushing influent. The counterfactual reconstruction results after removing the influence of compression and discharge load and the influence of compression and discharge liquid are presented. The defluxing matrix represents the counterfactual reconstruction results after retaining the influence of compression and discharge load and removing the influence of flushing water. The dedischarging matrix represents the counterfactual reconstruction results after retaining the influence of compression and discharge load and removing the influence of discharge load. The data values ​​of the same sampling point and the same response channel in the impact response residual matrix are subtracted from the data values ​​of the same sampling point and the same response channel in the decompression matrix, defluxing matrix, and dedischarging matrix, respectively. The absolute values ​​of the differences of each response channel at the same sampling point are added to obtain the decompression error value, defluxing error value, and dedischarging error value. The decompression error value, defluxing error value, and dedischarging error value are calculated point by point at a unified sampling time and are used to measure the removal of the corresponding impact. The degree of deviation of the post-source reconstruction result from the true impact response residual matrix is ​​measured. The compression contribution value is obtained by subtracting the total source error value from the decompression error value and setting negative values ​​to zero. The flushing contribution value is obtained by subtracting the total source error value from the defluxing error value and setting negative values ​​to zero. The unloading contribution value is obtained by subtracting the total source error value from the deunloading error value and setting negative values ​​to zero. Setting negative values ​​to zero is used to exclude sampling points where the error does not increase after removing a certain impact source, ensuring that the compression contribution value, flushing contribution value, and unloading contribution value only retain positive attribution values ​​that can increase the explained error. For the same sampling point, the impact source corresponding to the maximum value among the compression contribution value, flushing contribution value, and unloading contribution value is written into the dominant impact source label, and the impact source corresponding to the second largest value is written into the accompanying impact source label.Specifically, when the compression contribution value is at its maximum, the dominant impact source marker is written to the compression discharge; when the flushing contribution value is at its maximum, the dominant impact source marker is written to the flushing inlet; when the unloading contribution value is at its maximum, the dominant impact source marker is written to the unloading load, and the accompanying impact source marker is written to the impact source corresponding to the second largest contribution value using the same rules. When two contribution values ​​are tied for the maximum, the dominant impact source marker is determined according to the priority of compression discharge, unloading load, and flushing inlet, and the other impact source among the tied maximum values ​​is written to the accompanying impact source marker. When three contribution values ​​are tied for the maximum, the dominant impact source marker is written to the compression discharge, and the accompanying impact source marker is written to the flushing inlet. Record the unloading load; when there are ties for the second largest contribution value, determine the accompanying impact source marker according to the priority of unloading load, compression discharge, and flushing inlet; among them, compression discharge and unloading load have higher priority than flushing inlet because compression discharge and unloading load correspond to high organic load, high suspended solids load, and high grease load impacts, and are given priority in the calculation of subsequent high load control water volume, so that each unified sampling time forms a definite dominant source and accompanying source sequence marker; for the same sampling point, divide the largest contribution value by the sum of the compression contribution value, flushing contribution value, and unloading contribution value and the minimum constant to obtain the attribution purity value; where the minimum constant is taken as; to Preferred This is used to avoid the divisor failing when the sum of the compression contribution value, flushing contribution value, and unloading contribution value is 0. The closer the attribution purity value is to 1, the more concentrated the explanation of the current sampling point by a single impact source is; the lower the attribution purity value, the higher the degree of combined effect of compression drainage, flushing water inlet, and unloading load at the current sampling point. When the attribution purity value is less than the purity threshold, the corresponding sampling point is written as a composite impact marker; when the attribution purity value is greater than or equal to the purity threshold, the corresponding sampling point is written as a single-source impact marker. The purity threshold is between 0.55 and 0.75, preferably 0.65. The purity threshold is determined based on manually labeled impact source data from no less than 7 historical operating days. Specifically, candidate thresholds are traversed at intervals of 0.01 within the range of 0.55 to 0.75, and the candidate threshold that maximizes the weighted sum of the single-source impact recognition accuracy and the composite impact recognition accuracy is selected as the purity threshold. The counterfactual attribution result is output, which includes compression... The system includes contribution values, flushing contribution values, unloading contribution values, dominant impact source markers, accompanying impact source markers, composite impact markers, and single-source impact markers. Counterfactual attribution results are arranged according to a unified sampling time and maintain the same time index as the pre-treated percolation impact monitoring data. This allows subsequent attribution segment merging, segment influent calculation, and decomposition of the three types of controllable water volumes to directly read the attribution information of the corresponding sampling points. By writing impact sources with large error increments into the impact source attribution markers, the water volumes corresponding to compression discharge and unloading loads can be prioritized for inclusion in high-load controllable water volumes, while the water volumes corresponding to flushing influent can be prioritized for release. Furthermore, when the membrane-side pressure difference or membrane system permeate conductivity value is abnormal, the permissible release volume is reduced. This reduces transmembrane pressure fluctuations caused by high-load impacts directly entering the membrane segment, and reduces the cumulative increase in target coagulant, flocculant, and acid / alkali agent dosage values ​​due to short-term misjudgments.

[0038] In this implementation scheme, by performing full-source reconstruction and source-by-source removal comparison of the compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix, the explanatory effect of compression discharge, flushing influent, and unloading load on the impact response residual matrix can be transformed into a comparable error increment, avoiding misjudgment caused by judging the impact source solely based on the peak value of a single water quality index. At the same time, by distinguishing between single-source impacts and compound impacts through attribution purity values, the dominant impact source marker, accompanying impact source marker, compression contribution value, flushing contribution value, and unloading contribution value can form a stable correspondence at the same unified sampling time, providing a clear source and quantifiable contribution attribution basis for subsequent calculation of attribution fragment merging, fragment influent volume decomposition, compression control volume, flushing control volume, and unloading control volume.

[0039] Specifically, based on the counterfactual attribution results, the steps for merging attribution fragments and decomposing the influent volume into controlled water volume for compression, controlled water volume for flushing, and controlled water volume for unloading are as follows: Read the counterfactual attribution results and the pre-treated leachate impact monitoring data; merge sampling points with K consecutive sampling points having the same dominant impact source marker and consistent composite impact markers into an attribution fragment; where K is 3 to 8, preferably 5, and K is determined based on the unified sampling period and the shortest duration of the impact response at the waste transfer station, ensuring that the duration of the attribution fragment is not less than 2 minutes and not more than 8 minutes; when adjacent sampling points have the same dominant impact source marker but inconsistent composite impact markers, the attribution fragment is merged at the unified sampling time when the composite impact marker changes. The attribution segment is segmented so that single-source impact segments and composite impact segments are respectively included in the subsequent decomposition of the water volume to be controlled. When the length of the attribution segment is less than K sampling points, the attribution segment is merged with the previous attribution segment. The merging condition is that the dominant impact source markers of the two segments are the same and there are no long-term missing channel markers between them. For each attribution segment, the compression contribution value, flushing contribution value, and unloading contribution value are divided by the sum of the three and the minimum constant, respectively, to obtain the compression control ratio, flushing control ratio, and unloading control ratio. Among them, the compression contribution value, flushing contribution value, and unloading contribution value are first summed separately within the same attribution segment, and then the compression control ratio, flushing control ratio, and unloading control ratio are calculated. The minimum constant is taken as... to Preferred This is used to avoid the divisor failing when the sum of the three types of contribution values ​​is 0; when the sum of the compression contribution value, flushing contribution value, and unloading contribution value within the same attribution segment is 0, the compression control ratio, flushing control ratio, and unloading control ratio are all written to 0, and the attribution segment is marked as an invalid attribution segment; the instantaneous flow rate of the regulating tank inlet at each sampling point within the attribution segment is multiplied by the unified sampling period and then summed to obtain the segment inlet volume; whereby the instantaneous flow rate of the regulating tank inlet is converted from cubic meters per hour to cubic meters per second before calculation, and the unified sampling period is converted to seconds, so that the unit of the segment inlet volume is cubic meters; when there is a long-term missing channel marker within the attribution segment, the segment inlet volume is not calculated, and the corresponding attribution segment is excluded from subsequent control quantity generation; the segment The influent volume is multiplied by the compression control ratio, flushing control ratio, and unloading control ratio to obtain the controlled water volume for compression, flushing, and unloading. Among them, the controlled water volume for compression represents the high organic load water volume generated mainly by compression discharge and requiring delayed release; the controlled water volume for flushing represents the low concentration water volume generated mainly by flushing influent and which can be preferentially released; and the controlled water volume for unloading represents the water volume generated mainly by leachate corresponding to the vehicle unloading mass value and which needs to be included in the high load slow release along with the controlled water volume for compression. By decomposing the influent volume of the same attribution segment into three contribution values, the subsequent calculation of the remaining storage volume, the calculation of the permissible water release volume, and the membrane segment protection reduction can directly read the controlled water volume for compression, flushing, and unloading.

[0040] In this implementation plan, by converting counterfactual attribution results from sampling point-level attribution to attribution segment-level controllable water volume, the compression contribution value, flushing contribution value, and unloading contribution value are no longer limited to source judgment, but are further mapped to compression controllable water volume, flushing controllable water volume, and unloading controllable water volume, thereby enhancing the continuity between the impact source attribution results and subsequent regulation and release control. At the same time, the attribution segment merging rule can weaken the impact of single-point attribution jumps on segment influent decomposition, so that the segment influent volume formed by the instantaneous influent flow value of the regulating tank can be divided and measured according to the actual contribution of compression discharge, flushing influent, and unloading load, providing a control basis with clear sources and calculable volume for subsequent regulation and storage remaining volume matching, permissible external discharge volume determination, and membrane segment protection reduction.

[0041] Specifically, the steps for outputting the intelligent control results of leachate treatment, combining the remaining storage capacity and the membrane protection status, are as follows: The current storage capacity is read from the storage tank level-volume calibration table based on the storage tank level value. The current storage capacity is then subtracted from the storage capacity corresponding to the upper safety limit level to obtain the remaining storage capacity. The storage tank level-volume calibration table is generated by verifying the segmented water injection calibration data and as-built measurement data of the storage tank. The table records a one-to-one correspondence between the storage tank level and the storage tank volume. When the storage tank level is between two adjacent calibration levels, linear interpolation is used to calculate the current storage capacity. The upper safety limit level is taken as 8 times the design overflow level of the storage tank. The safe upper limit of the liquid level is determined based on the equalization tank design drawings, the installation height of the level gauge, and the historical maximum influent impact level, ensuring that the remaining equalization tank volume reflects the space available for the current equalization tank to continue accepting high-load controlled water volume. The high-load controlled water volume is obtained by adding the compressed controlled water volume and the unloading controlled water volume. When the remaining equalization tank volume is greater than or equal to the high-load controlled water volume, the flushing controlled water volume is used as the permissible discharge volume. When the remaining equalization tank volume is less than the high-load controlled water volume, the difference between the high-load controlled water volume and the remaining equalization tank volume is added to the flushing controlled water volume to obtain the permissible discharge volume. The compressed controlled water volume and the unloading controlled water volume correspond to high organic load and high suspended solids. For leachate shocks with high material and oil loads, the equalization tank is prioritized for temporary storage and slow release. For flushing water volumes corresponding to low-concentration, high-flow shocks, the permissible discharge volume is prioritized in the calculation, ensuring the limited storage capacity of the equalization tank is used primarily to reduce high-load shocks. The membrane-side pressure difference is obtained by subtracting the post-membrane permeate pressure from the inlet water pressure. When the membrane-side pressure difference exceeds the upper limit of membrane pressure difference, or when the permeate conductivity exceeds the upper limit of permeate conductivity, the permissible discharge volume is divided by the control cycle and multiplied by the membrane protection reduction factor to obtain the protection discharge flow rate. When the membrane-side pressure difference is less than or equal to the upper limit of membrane pressure difference and the permeate conductivity is less than or equal to the upper limit of permeate conductivity, the permissible discharge volume is... The protected discharge flow rate is obtained by dividing the discharge volume by the control cycle. The upper limit of the membrane pressure difference is 45 kPa to 80 kPa, preferably 60 kPa, and is determined based on the membrane module operation manual, historical cross-membrane pressure difference records before cleaning, and stable permeate pressure records. The upper limit of the permeate conductivity is 800 μS / cm to 1500 μS / cm, preferably 1000 μS / cm, and is determined based on discharge control requirements and historical distribution of compliant permeate conductivity. The control cycle is 5 min to 30 min, preferably 10 min, and is determined based on the booster pump frequency adjustment response time, chemical mixing residence time, and aeration adjustment response time.The control period is an integer multiple of the unified sampling period, and includes multiple unified sampling periods. The unified sampling period is used for data alignment of infiltration shock monitoring, hysteresis inversion, and segment influent calculation. The control period is used to summarize the protection release flow rate, booster pump control frequency, coagulant target dosage flow rate, flocculant target dosage flow rate, acid and alkali agent target dosage flow rate, and aeration blower target frequency within the same control period. The membrane protection reduction factor is taken from 0.40 to 0.80, preferably 0.60. The membrane protection reduction factor is determined based on the excess range of membrane side pressure difference, the excess range of membrane system permeate conductivity, and historical membrane flux recovery records, so as to reduce the instantaneous release intensity entering the downstream when the membrane section experiences pressure increase or permeate conductivity increase; according to the protection... The protective discharge flow rate is read from the booster pump flow rate and frequency calibration table. This table is generated from the booster pump's on-site frequency conversion trial operation data and records a one-to-one correspondence between the protective discharge flow rate and the booster pump control frequency. When the protective discharge flow rate is between two adjacent calibrated flow rates, linear interpolation is used to read the booster pump control frequency, limiting it to between the minimum and maximum allowable operating frequencies. Based on the high-load controlled water volume, influent chemical oxygen demand (COD) concentration, influent suspended solids concentration, influent turbidity, influent oil concentration, and influent pH, the target dosing flow rates for coagulants and flocculants, as well as the acid / alkali ratio, are read from the chemical dosing load calibration table. The target dosing flow rate for chemicals is specified. The dosing load calibration table is generated from historical dosing operation records and beaker coagulation test data. The table uses the high-load controlled water volume, influent chemical oxygen demand (COD) concentration, influent suspended solids (SSD) concentration, influent turbidity, influent oil concentration, and influent pH as indexes to record the target dosing flow rates for coagulants, flocculants, and acid / alkali chemicals. The higher the high-load controlled water volume and the higher the influent COD, SSD, turbidity, and oil concentrations, the higher the corresponding chemical dosing flow rate. The adjustment direction of the target dosing flow rate for acid / alkali chemicals is determined by the influent pH value. The table also considers the compressed controlled water volume, influent ammonia nitrogen concentration, dissolved oxygen concentration in the biological treatment tank, and biological treatment... The redox potential value of the pool is read from the aeration load frequency calibration table to obtain the target frequency value of the aeration blower. This table is generated from historical biological pool operation data and aeration blower frequency adjustment test data. The table uses the controlled water volume, influent ammonia nitrogen concentration, dissolved oxygen concentration in the biological pool, and redox potential of the biological pool as indexes to record the target frequency value of the aeration blower. This ensures that the aeration adjustment volume matches the oxygen supply status of the biological pool before the high ammonia nitrogen shock corresponding to the compressed discharge enters the biological treatment process. The intelligent control results for leachate treatment are output, including the booster pump control frequency value, coagulant target dosage flow rate value, flocculant target dosage flow rate value, acid and alkali agent target dosage flow rate value, aeration blower target frequency value, and protection release flow rate value.Among them, the booster pump control frequency value is used to control the intensity of the discharge from the equalization tank; the target dosing flow rates of coagulant, flocculant, and acid / alkali agents are used to control the intensity of upstream coagulation and equalization; the target frequency value of the aeration blower is used to control the oxygen supply intensity of the biological treatment tank; and the protection discharge flow rate value is used to limit the instantaneous flow rate entering the downstream treatment unit.

[0042] In this implementation plan, by mapping the controlled volume of compressed water, the controlled volume of flushing water, and the controlled volume of unloading water to the remaining storage volume, the membrane protection status, the dosing load calibration table, and the aeration load frequency calibration table, respectively, the counterfactual attribution results can be directly converted into the control frequency value of the booster pump, the target dosing flow rate value of the coagulant, the target dosing flow rate value of the flocculant, the target dosing flow rate value of the acid and alkali agents, and the target frequency value of the aeration blower. This avoids the mismatch of the release rhythm caused by using a single flow threshold control after mixing the compressed discharge, flushing influent, and unloading load. The high load controlled water volume prioritizes the use of the remaining storage volume, and the controlled flushing water volume forms a controllable release basis. Furthermore, when the membrane side pressure difference and the conductivity value of the membrane system permeate are abnormal, the instantaneous influent intensity at the downstream end is reduced. This improves the storage stability, agent dosing matching, biochemical oxygen supply adaptability, and membrane section operation safety of the leachate in the waste transfer station under shock load conditions.

[0043] Table 1. Data on the release of uncontrolled water volume by compartment.

[0044] In this embodiment, taking the leachate treatment process during the continuous collection period of a waste transfer station in the morning peak as an example, four attribution segments formed from 08:00 to 08:40 are selected as the analysis objects. During this period, waste transport vehicles successively enter the station to unload, the compressor operates intermittently to generate compressed discharge, and the flushing water from the station floor and equipment enters the equalization tank in stages, causing continuous changes in the influent flow rate, water concentration, and membrane segment operating status of the equalization tank. The system performs hysteresis inversion and counterfactual attribution on the four attribution segments based on leachate impact monitoring data, obtains the controllable water volume corresponding to the compressed discharge, flushing water influent, and unloading load in each segment, and determines the compartment release results of different segments by combining the remaining regulation capacity of the equalization tank and the membrane segment protection status.

[0045] Table 1 is a data table of controlled water volume release by compartment, which lists the three types of controlled water volume, segment inflow, storage and interception water volume and membrane protection status corresponding to the four attribution segments. In the table, the controlled water volume for flushing in segment 1 is 1.30 m³, the controlled water volume for compression is 4.46 m³, the controlled water volume for unloading is 1.44 m³, and the segment inlet water volume is 7.20 m³; the controlled water volume for flushing in segment 2 is 5.57 m³, the controlled water volume for compression is 2.40 m³, the controlled water volume for unloading is 1.63 m³, and the segment inlet water volume is 9.60 m³; the controlled water volume for flushing in segment 3 is 1.68 m³, the controlled water volume for compression is 3.02 m³, the controlled water volume for unloading is 3.70 m³, and the segment inlet water volume is 8.40 m³; and the controlled water volume for flushing in segment 4 is 1.25 m³, the controlled water volume for compression is 3.74 m³, the controlled water volume for unloading is 2.81 m³, and the segment inlet water volume is 7.80 m³. Among them, the remaining storage capacity of segments 1 and 2 can cover the high-load controlled water volume formed by the controlled water volume of compression and unloading. Therefore, the stored water volume is marked as completely intercepted, and the membrane protection status is not triggered. The stored water volumes of segments 3 and 4 are 3.50 m³ and 2.20 m³, respectively, indicating that the remaining storage capacity is insufficient to cover the entire high-load controlled water volume, and the membrane protection status is triggered. It is necessary to further reduce the membrane protection based on the storage compartment. Table 1 corresponds to the diagram of the classified release of controlled water volume, which can reflect the differentiated control relationship between flushing inlet, high-load compression discharge, and unloading load in storage release and membrane protection.

[0046] like Figure 3 The diagram illustrates the process of transforming counterfactual attribution results into intelligent control results for leachate treatment. Segments 1 to 4 represent consecutive attribution segments. Each segment's column is composed of superimposed controlled volumes of flushing, compression, and unloading water, representing the three types of controlled water volumes after decomposition of the segment's influent volume by compression, flushing, and unloading contributions. The diagonally filled area represents the controlled volume of flushing water, the dotted area represents the controlled volume of compression water, and the cross-filled area represents the controlled volume of unloading water. The dashed line represents the regulation and retention line. When the remaining regulation and retention volume can accommodate high-load controlled water volumes, the diagram marks "complete retention," indicating that the controlled volumes of compression and unloading water are preferentially released by the regulation tank, while the controlled volume of flushing water is released along the permissible discharge direction. When the remaining regulation and retention volume is insufficient, the dashed line marks the boundary where high-load controlled water is retained; the high-load water above the dashed line, together with the controlled volume of flushing water, forms the permissible discharge volume. The red bars represent membrane protection trigger segments, indicating that the membrane side pressure difference or membrane system permeate conductivity value in segments 3 and 4 has reached the protection judgment condition. The permissible discharge volume needs to be further reduced to the protection discharge flow rate, thus forming a smart control result for leachate treatment that takes into account both regulation, slow release and membrane segment protection.

[0047] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0048] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A smart control method for leachate treatment in waste transfer stations, characterized in that, Includes the following steps: S1: Collect infiltration impact monitoring data, perform anomaly removal, missing data completion and time alignment processing on the infiltration impact monitoring data, and output the preprocessed infiltration impact monitoring data. S2 identifies candidate impact response intervals based on preprocessed infiltration impact monitoring data and generates impact response residual matrix and source pulse matrix. It performs hysteresis translation on the source pulse matrix according to the maximum hysteresis sampling number to generate three types of hysteresis dictionaries corresponding to compression discharge, flushing water inlet and unloading load. It also outputs hysteresis inversion results by combining water conservation residual value and a grouped nested regression model with sign constraints. S3 generates compression reconstruction matrix, flushing reconstruction matrix and unloading reconstruction matrix based on hysteresis inversion results. It obtains three types of contribution values ​​through source removal and error increment calculation, generates impact source attribution labels, and outputs counterfactual attribution results. S4, based on the counterfactual attribution results, merges the attribution segments, decomposes the influent volume into the controlled volume of compression, the controlled volume of flushing, and the controlled volume of unloading, and outputs the intelligent control results of leachate treatment by combining the remaining storage volume and the membrane segment protection status.

2. The intelligent control method for leachate treatment in a waste transfer station according to claim 1, characterized in that: The specific steps for collecting infiltration impact monitoring data and performing anomaly removal, missing data completion, and time alignment on the infiltration impact monitoring data are as follows: Data on infiltration shock monitoring was collected, including vehicle entry timestamp, vehicle unloading mass value, compressor start timestamp, compressor stop timestamp, compressor operating current value, compressor hydraulic pressure value, flushing pump start timestamp, flushing pump stop timestamp, flushing water instantaneous flow rate value, equalization tank level value, equalization tank influent instantaneous flow rate value, equalization tank effluent instantaneous flow rate value, influent chemical oxygen demand concentration value, influent ammonia nitrogen concentration value, influent suspended solids concentration value, influent turbidity value, influent pH value, influent conductivity value, influent oil concentration value, biological treatment tank dissolved oxygen concentration value, biological treatment tank oxidation-reduction potential value, membrane influent pressure value, membrane post-permeate pressure value, and membrane system permeate conductivity value. For the collected infiltration impact monitoring data, the Hanpuer filtering algorithm is used to identify the instantaneous jump value of the instrument and remove abnormal sampling points. The piecewise linear interpolation algorithm is used to complete the short-term missing records and align the sampling time, and output the preprocessed infiltration impact monitoring data.

3. The intelligent control method for leachate treatment in a waste transfer station according to claim 2, characterized in that: The specific steps for identifying candidate impact response intervals based on preprocessed infiltration impact monitoring data are as follows: The pre-processed infiltration shock monitoring data were read, and a multi-channel response sequence was formed by combining the liquid level of the equalization tank, the instantaneous flow rate of the equalization tank influent, the concentration of chemical oxygen demand influent, the concentration of ammonia nitrogen influent, the concentration of suspended solids influent, the turbidity of influent, the pH value of influent, the conductivity value of influent, and the oil concentration value of influent. The change points in the multi-channel response sequence were identified by the Bayesian online change point detection algorithm, and the sampling interval between two adjacent change points was determined as the candidate shock response interval.

4. The intelligent control method for leachate treatment in a waste transfer station according to claim 3, characterized in that: The specific steps for generating the impulse response residual matrix and the source impulse matrix are as follows: For each candidate impact response interval, read N consecutive sampling points before the start of the candidate impact response interval, calculate the median of each data in the multi-channel response sequence, and obtain the interval baseline vector; take the sampling points in the candidate impact response interval as matrix rows, and take each data in the multi-channel response sequence as matrix columns, subtract the baseline value corresponding to the same matrix column in the interval baseline vector from the data value corresponding to each matrix column in each sampling point, and obtain the impact response residual matrix; A source pulse matrix is ​​constructed using the unified sampling time as the row index. The product of the compressor operating current value and the compressor hydraulic pressure value within the time stamp from the compressor start time stamp to the compressor stop time stamp is written into the compressor discharge column. The instantaneous flow rate value of the flushing water within the time stamp from the flushing pump start time stamp to the flushing pump stop time stamp is written into the flushing water inlet column. The vehicle unloading mass value corresponding to the vehicle arrival time stamp is written into the unloading load column. The remaining matrix positions are written with 0.

5. The intelligent control method for leachate treatment in a waste transfer station according to claim 4, characterized in that: The specific steps for performing a hysteresis shift on the source pulse matrix based on the maximum hysteresis sampling number to generate three types of hysteresis dictionaries corresponding to the compression discharge, flushing water inlet, and unloading loads are as follows: Calculate the ratio of the pipeline volume from the inlet of the equalization tank to the water quality sampling point to the instantaneous flow rate of the equalization tank inlet, and calculate the ratio of the effective volume of the equalization tank to the instantaneous flow rate of the equalization tank inlet. Divide the two ratios by a uniform sampling period and round up to obtain two lag sampling numbers. Take the larger of the two lag sampling numbers as the maximum lag sampling number. Shift the three columns of data in the source pulse matrix backward sequentially from 0 to the maximum lag sampling number. Concatenate the shifted columns corresponding to each lag sampling number to form a compression discharge lag dictionary, a flushing inlet lag dictionary, and a discharge load lag dictionary.

6. The intelligent control method for leachate treatment in a waste transfer station according to claim 5, characterized in that: The specific steps for outputting the lag inversion results using the combined water conservation residual value and the signed constraint-based nested regression model are as follows: The corresponding volume value of the equalization tank is read from the equalization tank volume calibration table based on the liquid level value of the equalization tank. The volume value of the equalization tank at the next sampling point is subtracted from the volume value of the equalization tank at the previous sampling point to obtain the change in the volume of the tank. The instantaneous inflow rate of the equalization tank is subtracted from the instantaneous outflow rate of the equalization tank and then multiplied by the uniform sampling period to obtain the theoretical net inflow rate of the tank. The change in the volume of the tank is subtracted from the theoretical net inflow rate of the tank to obtain the water conservation residual value. The lag dictionaries for compression discharge, flushing influent, and unloading load were used as explanatory variables in the nested cable regression model. The impact response residual matrix was used as the matrix to be explained. The absolute value of the water conservation residual was added to the objective function of the nested cable regression model. The alternating direction multiplier method was used to solve the signed constraint nested cable regression model, obtaining the compression discharge coefficient matrix, flushing influent coefficient matrix, and unloading load coefficient matrix. Among them, the compression discharge coefficient matrix is ​​used for the liquid level in the equalization tank, the instantaneous flow rate of the equalization tank influent, and the chemical oxygen demand of the influent. The oxygen concentration, influent ammonia nitrogen concentration, influent suspended solids concentration, and influent turbidity values ​​in the corresponding channels are limited to non-negative values. The flushing influent coefficient matrix is ​​limited to non-negative values ​​in the corresponding channels of equalization tank level and equalization tank influent instantaneous flow rate, and is limited to non-positive values ​​in the corresponding channels of influent chemical oxygen demand (COD), influent ammonia nitrogen concentration, influent suspended solids concentration, influent turbidity, and influent conductivity. The unloading load coefficient matrix is ​​limited to non-negative values ​​in the corresponding channels of influent grease concentration, influent COD concentration, and influent turbidity. The output lag inversion results include the impact response residual matrix, compression discharge lag dictionary, flushing water inlet lag dictionary, unloading load lag dictionary, compression discharge coefficient matrix, flushing water inlet coefficient matrix, and unloading load coefficient matrix.

7. The intelligent control method for leachate treatment in a waste transfer station according to claim 6, characterized in that: The specific steps for generating the compression reconstruction matrix, flushing reconstruction matrix, and unloading reconstruction matrix based on the hysteresis inversion results are as follows: Read the lag inversion results, perform matrix multiplication between the compression discharge lag dictionary and the compression discharge coefficient matrix to obtain the compression reconstruction matrix; perform matrix multiplication between the flushing water inlet lag dictionary and the flushing water inlet coefficient matrix to obtain the flushing reconstruction matrix; perform matrix multiplication between the unloading load lag dictionary and the unloading load coefficient matrix to obtain the unloading reconstruction matrix.

8. The intelligent control method for leachate treatment in a waste transfer station according to claim 7, characterized in that: The specific steps for obtaining three types of contribution values ​​through source-by-source removal and error increment calculation, generating impact source attribution labels, and outputting counterfactual attribution results are as follows: The three reconstruction matrices are added together at the same sampling point and the same response channel to obtain the full-source reconstruction matrix; Subtract the data values ​​of the same sampling point and the same response channel in the impact response residual matrix from the full-source reconstruction matrix, and add the absolute values ​​of the differences of each response channel at the same sampling point to obtain the full-source error value; Subtract the compression reconstruction matrix, flushing reconstruction matrix and unloading reconstruction matrix from the full source reconstruction matrix to obtain the decompression matrix, defluxing matrix and deunloading matrix; Subtract the data values ​​of the same sampling point and the same response channel from the impact response residual matrix and the decompression matrix, defluxing matrix, and deunloading matrix, respectively, and add the absolute values ​​of the differences of each response channel at the same sampling point to obtain the decompression error value, defluxing error value, and deunloading error value; Subtract the total source error value from the decompression error value and set negative values ​​to zero to obtain the compression contribution value; Subtracting the total source error value from the de-flushing error value and setting negative values ​​to zero yields the flushing contribution value; subtracting the total source error value from the de-discharge error value and setting negative values ​​to zero yields the discharge contribution value; for the same sampling point, the impact source corresponding to the maximum value among the compression contribution value, flushing contribution value, and discharge contribution value is written into the dominant impact source marker, and the impact source corresponding to the second largest value is written into the accompanying impact source marker. For the same sampling point, the attribution purity value is obtained by dividing the maximum contribution value by the sum of the compression contribution value, flushing contribution value, and unloading contribution value and the minimum constant. When the attribution purity value is less than the purity threshold, the corresponding sampling point is written into the composite impact marker; when the attribution purity value is greater than or equal to the purity threshold, the corresponding sampling point is written into the single-source impact marker. Output counterfactual attribution results, which include compression contribution value, flushing contribution value, unloading contribution value, dominant impact source marker, accompanying impact source marker, composite impact marker, and single-source impact marker.

9. The intelligent control method for leachate treatment in a waste transfer station according to claim 8, characterized in that: The specific steps for merging attribution fragments based on counterfactual attribution results and decomposing the influent fragment into controlled water volume for compression, controlled water volume for flushing, and controlled water volume for unloading are as follows: Read the counterfactual attribution results and preprocessed infiltration shock monitoring data, and merge the sampling points with the same K consecutive sampling points of the dominant shock source and the same composite shock mark into an attribution segment; for each attribution segment, divide the compression contribution value, flushing contribution value and unloading contribution value by the sum of the three and the minimum constant respectively to obtain the compression control ratio, flushing control ratio and unloading control ratio; multiply the instantaneous flow rate of the regulating tank influent at each sampling point in the attribution segment by a uniform sampling period and sum them to obtain the segment influent; multiply the segment influent by the compression control ratio, flushing control ratio and unloading control ratio respectively to obtain the compression control water volume, flushing control water volume and unloading control water volume.

10. The intelligent control method for leachate treatment in a waste transfer station according to claim 9, characterized in that: The specific steps for combining the remaining storage volume and membrane segment protection status to output the intelligent control results of leachate treatment are as follows: The current volume of the equalization tank is read from the equalization tank volume calibration table based on the liquid level value of the equalization tank. The volume of the equalization tank corresponding to the upper limit liquid level is subtracted from the current volume of the equalization tank to obtain the remaining volume of the equalization tank. The controlled water volume for compression and the controlled water volume for unloading are added together to obtain the controlled water volume for high load. When the remaining volume of the equalization tank is greater than or equal to the controlled water volume for high load, the controlled water volume for flushing is used as the permissible water volume for discharge. When the remaining volume of the equalization tank is less than the controlled water volume for high load, the difference between the controlled water volume for high load and the remaining volume of the equalization tank is added to the controlled water volume for flushing to obtain the permissible water volume for discharge. The membrane pressure difference is obtained by subtracting the permeate pressure from the inlet water pressure. When the membrane pressure difference is greater than the upper limit of the membrane pressure difference, or when the permeate conductivity of the membrane system is greater than the upper limit of the permeate conductivity, the permissible discharge volume is divided by the control cycle and multiplied by the membrane protection reduction factor to obtain the protective discharge flow rate. When the membrane pressure difference is less than or equal to the upper limit of the membrane pressure difference and the permeate conductivity of the membrane system is less than or equal to the upper limit of the permeate conductivity, the permissible discharge volume is divided by the control cycle to obtain the protective discharge flow rate. The control frequency value of the booster pump is read from the booster pump flow frequency calibration table based on the protected external flow rate value; the target dosing flow rates of coagulant, flocculant, and acid / alkali agents are read from the chemical dosing load calibration table based on the high load uncontrolled water volume, influent chemical oxygen demand concentration, influent suspended solids concentration, influent turbidity, influent oil concentration, and influent pH value; the target frequency value of the aeration blower is read from the aeration load frequency calibration table based on the compressed uncontrolled water volume, influent ammonia nitrogen concentration, dissolved oxygen concentration in the biological treatment tank, and oxidation-reduction potential in the biological treatment tank. The system outputs intelligent control results for leachate treatment, including the control frequency of the booster pump, the target dosing flow rate of the coagulant, the target dosing flow rate of the flocculant, the target dosing flow rate of the acid and alkali agents, the target frequency of the aeration fan, and the protection release flow rate.

Images