Grading treatment system and method for urban road rainwater runoff

Abstract

The invention discloses an urban road rainwater runoff stage treatment system and method, and relates to the technical field of urban rainwater pollution treatment. In order to solve the problems that an existing system is difficult to monitor multi-index water quality in real time, the pollution degree is accurately recognized, and distribution control lacks dynamics, the system is provided with a multi-source spectrum water quality monitor at the tail end of a pipe network, turbidity, COD, TOC and heavy metal concentration are monitored in real time, rainfall intensity data are accessed into a control center, the pollutant load instantaneous rate L is calculated through a weighted summation formula, and the real-time monitoring of the pollution degree is achieved. A controllable flow distribution module in the flow distribution well distributes the rainwater to an enhanced treatment unit, a biological aerated filter system or a horizontal flow sedimentation tank, wherein the enhanced treatment unit comprises a rotational flow sand setting unit, an electrochemical heavy metal degradation unit and a chemical flocculation unit. The system is mainly used for realizing quality-divided efficient treatment and resource recycling of rainwater and guaranteeing the safety of a receiving water body.

Description

Technical Field

[0001] The present invention relates to the technical field of urban rainwater pollution control, and more particularly to a system and method for hierarchical treatment of urban road rainwater runoff. Background Art

[0002] In the field of urban rainwater pollution control technology, the existing urban road rainwater runoff treatment system has the following technical difficulties: First, it's difficult to accurately identify the extent of rainwater runoff pollution. Existing systems lack the real-time monitoring capabilities for multiple indicators, including turbidity, COD, TOC, and heavy metal concentrations, making it impossible to implement targeted treatment measures based on varying levels of pollution. This is due to the limitations of traditional monitoring technology, which makes it difficult to achieve simultaneous, real-time collection and analysis of multiple indicators, resulting in inaccurate assessments of rainwater pollution levels.

[0003] Second, stormwater diversion control strategies lack dynamism and scientific rationality. Traditional systems fail to fully integrate multiple factors, such as the real-time load of the pipe network, the water level of the regulating reservoir, and the real-time load of the sewage treatment plant. This makes diversion strategies difficult to adapt to complex actual operating conditions. This is due to a lack of comprehensive processing and analysis capabilities for multi-source data, making it impossible to dynamically optimize and adjust diversion strategies. Consequently, it is difficult to make scientific and reasonable diversion decisions when faced with varying rainfall intensities and pipe network loads.

[0004] Third, the control of chemical dosage during the treatment process is not precise enough. The existing system is unable to adjust the chemical dosage in real time according to the water quality of the treatment unit, resulting in low chemical utilization efficiency and potentially affecting the treatment effect. This is because there is a lack of real-time monitoring and feedback mechanism for treated water quality, making it difficult to establish a precise control relationship between water quality indicators and chemical dosage. As a result, chemical dosage cannot be adjusted in time according to changes in water quality during the treatment process. Summary of the Invention

[0005] An object of the present invention is to provide a system for grading and treating rainwater runoff from urban roads, comprising: Multi-source spectral water quality monitor, installed at the end of the rainwater collection network, monitors the turbidity, COD, TOC and heavy metal concentration of rainwater runoff in real time; The control center receives rain monitoring data from the multi-source spectral water quality monitor, accesses rainfall intensity data from rainfall forecast information, and calls the historical biological toxicity database; The control center uses the weighted summation formula to calculate the instantaneous rate of pollutant load L , the formula is: ,in, w i For the i The weight coefficient of the parameter, f i ( xi ) is the i A standardized data mapping function with parameters, n is the number of parameters, including real-time monitoring of turbidity, COD, TOC, heavy metal concentrations, rainfall intensity, and historical biological toxicity data; Multiple diversion wells are set up at the branch nodes of the rainwater collection network. A controllable diversion module is installed inside each diversion well. The controllable diversion module is connected to the command output terminal of the control center and performs rainwater distribution operations according to the hierarchical instructions of the control center; The inlet of the enhanced treatment unit is connected to the first outlet of the diversion well, and the cyclone sand settling treatment unit, the electrochemical heavy metal degradation treatment unit and the chemical flocculation treatment unit are sequentially arranged inside; a biological aeration filter system, the inlet of which is connected to the second outlet of the diversion well; a horizontal flow sedimentation tank, the inlet of which is connected to the third outlet of the diversion well; Among them, when L >When the first threshold is reached, the controllable diversion module distributes rainwater to the enhanced treatment unit; when L When the value is less than the second threshold, the controllable diversion module distributes the rainwater to the biological aeration filter system; When the second threshold ≤ L When the value is less than or equal to the first threshold, the controllable diversion module distributes the rainwater to the horizontal flow sedimentation tank.

[0006] Preferably, the control center includes an adaptive weight model that dynamically adjusts weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters based on a Bayesian inference framework. The dynamic adjustment process includes: Set initial weight coefficients for real-time rain monitoring, rainfall intensity, and historical biotoxicity parameters; According to the current rainfall intensity R t Rainfall intensity compared with similar historical rainfall events R h The absolute difference Δ R and the standard deviation of historical rainfall intensity s Calculating the matching factor ; Using the formula Correct the initial weight coefficient of the historical biological toxicity parameters, where is the weight coefficient of the historical biological toxicity parameter at the current moment, is the weight coefficient of the historical biological toxicity parameter at the previous moment, α is the preset correction factor for the historical biological toxicity parameter, | α ∣≤0.3, when Δ R ≤ s hour α>0, when ΔR> s hour α <0; The weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters were scaled so that their sum was 1.

[0007] Preferably, a collaborative control component is further included, which includes: The heavy metal concentration monitoring module is installed at the outlet of the electrochemical heavy metal degradation treatment unit to detect the residual heavy metal concentration in real time; The parameter collaborative controller is connected to the heavy metal concentration monitoring module and the reagent dosing system in the chemical flocculation treatment unit. The parameter collaborative controller receives the residual heavy metal concentration data and implements a hierarchical control strategy: When the residual heavy metal concentration is ≤ the safety threshold, the chemical dosing system maintains the dosage of chemical flocculants; When the residual heavy metal concentration is greater than the safety threshold, the parameter collaborative controller performs the following operations: Calculate the concentration value Δ at which the residual heavy metal concentration exceeds the safety threshold C , according to Δ C In the preset concentration range, select the corresponding dosage increase ratio m %, where the preset concentration range is m The corresponding relationship of % is: i). If 0μg / L<ΔC≤5μg / L, then m %=10%; ii). If 5μg / L<ΔC≤10μg / L, then m %=20%; iii). If 10μg / L<ΔC≤20μg / L, then m %=40%; iv). If ΔC>20μg / L, then m %=60%; Increase the amount of flocculant to 1+ of the original amount m % times.

[0008] Preferably, the parameter collaborative controller is also connected to the adaptive weight model. Before executing the hierarchical control strategy, the parameter collaborative controller corrects the dosage of the flocculant based on the instantaneous rate of the pollutant load in advance. The adaptive weight model is configured to execute: Instantaneous rate of pollutant load based on real-time calculation L t , generates instantaneous rate predictions of pollutant loads through built-in time series prediction algorithms : Will L t and history NThe instantaneous rate data of pollutant loads at consecutive time steps constitute a time series; Use the time series prediction algorithm to make a single-step prediction of the sequence and output the predicted value at the next acquisition moment , the time series prediction algorithm is any one of the sliding average model, autoregressive integrated sliding average model or long short-term memory network; Through the control center t +1 moment to calculate the actual pollutant load instantaneous rate monitoring value L t+1 : Comparison of predicted values Compared with the actual monitoring value L t+1 The mean absolute error ε, μ is used to calculate the prediction confidence coefficient μ=e -λε , λ is the error sensitivity coefficient, ranging from 0.5 to 1.0; The instantaneous rate prediction of pollutant load is multiplied by k The benchmark lead compensation is calculated, where: k It is the preset compensation coefficient, ranging from 0.1 to 0.3; Multiply the reference lead compensation by the prediction credibility coefficient to generate the effective lead compensation ΔE; According to ΔE, increase the flocculant dosage to 1+ΔE times of the original dosage.

[0009] Preferably, the effluent from the biological aerated filter system is treated by an ultrafiltration membrane and then enters the purified water storage tank; The urban road rainwater runoff classification treatment system further includes: The global optimization module is connected to the control center, purified water storage tank, sewage treatment plant, and rainfall forecast information, and is configured as follows: Real-time access to the current water level of the purified water storage tank, the real-time excess treatment capacity of the sewage treatment plant, and rainfall intensity; Dynamically adjust the first and second thresholds based on the above parameters: When the real-time surplus processing capacity of the sewage treatment plant is higher than the preset capacity, a first threshold lowering instruction is generated, and the first threshold is multiplied by a lowering coefficient, where 0.7≤lowering coefficient<1.0; When the rainfall intensity exceeds the rainstorm threshold and the current water level of the purified water storage tank reaches or exceeds the water level warning value, a second threshold increase instruction is generated, and the second threshold is multiplied by the increase coefficient, where 1.0<increase coefficient≤1.5.

[0010] Preferably, it also includes: The performance monitoring module obtains the electrochemical degradation electrode loss rate and ultrafiltration membrane transmembrane pressure difference of the enhanced treatment unit in real time; The state constraint layer connects the performance monitoring module and the global optimization module and is configured to perform: receiving a threshold adjustment instruction generated by a global optimization module; If a first threshold value downward adjustment instruction is received and the current electrode loss rate is greater than the loss rate upper limit, the downward adjustment instruction is discarded and an electrode loss warning is triggered; otherwise, the downward adjustment instruction is allowed to be executed; If a second threshold increase instruction is received and the real-time ultrafiltration membrane transmembrane pressure difference is greater than the pressure critical value, the increase range is limited so that the adjusted second threshold does not exceed the initial threshold; otherwise, the instruction is allowed to be executed according to the increase coefficient.

[0011] Preferably, it also includes: Redundant traffic protection components, including: Fault diagnosis module, real-time monitoring of the operating status of the controllable shunt module; Bypass switching unit, connecting the inlet of the horizontal flow sedimentation tank and the end of the rainwater collection network; Among them, when the fault diagnosis module detects that the controllable diversion module fails, it automatically activates the bypass switching unit to force all rainwater runoff into the horizontal flow sedimentation tank, and at the same time sends an equipment fault alarm to the control center.

[0012] Preferably, the adaptive weight model includes a burst pollution identification layer that performs: Real-time calculation of the sudden change gradient of turbidity, COD, TOC and heavy metal concentration of rainwater runoff output by multi-source spectral water quality monitor; When any of the above mutation gradients exceeds the maximum fluctuation threshold in the historical biological toxicity correlation parameters, a spectral anomaly index is generated, and the maximum fluctuation threshold is based on the 95th percentile value of historical data statistics; The dynamic weight allocation layer responds to the spectral anomaly index and forces the weights of the pollutant load instantaneous rate calculation to be assigned to the real-time rainwater monitoring data, while ignoring the rainfall forecast information and historical biotoxicity correlation parameters; At this time, if the instantaneous rate of pollutant load exceeds 80% of the first threshold, the control center will determine that L >First threshold.

[0013] Preferably, the global optimization module is provided with a depth processing allocation module, which performs: Real-time acquisition of the water level of the purified water storage tank and the transmembrane pressure difference of the ultrafiltration membrane; Construct the water level-membrane pressure synergistic constraint function: When the water level in the regulating reservoir is greater than the warning water level and the transmembrane pressure difference of the ultrafiltration membrane is greater than 90% of the critical value, the control center dynamically increases the second threshold; The controllable diversion module distributes rainwater to the aerated biological filter system according to the formula Q = β× ( Smax - S 当前 ) Current limiting is performed, wherein Q To allow for flow diversion, β is the attenuation coefficient, S max is the maximum safe capacity of the storage tank. S 当前 is the real-time water level; Excess rainwater is switched to the horizontal flow sedimentation tank through a controllable diversion module.

[0014] A treatment method of the urban road rainwater runoff classification treatment system is provided, comprising the following steps: S1. Real-time monitoring and data acquisition: A multi-source spectral water quality monitor installed at the end of the rainwater collection network monitors the turbidity, COD, TOC, and heavy metal concentrations of rainwater runoff in real time. Synchronously access real-time rainfall intensity in rainfall forecast information; S2. Dynamic calculation of pollutant load instantaneous rate L: Standardize real-time rain monitoring data, rainfall intensity, and historical biotoxicity correlation parameters and remove outliers; Dynamically adjust weight coefficients based on the Bayesian inference framework: Set initial weights for real-time monitoring data, rainfall intensity, and historical biotoxicity parameters; According to the current rainfall intensity R t Rainfall intensity compared with similar historical rainfall events R h The absolute difference Δ R and the standard deviation of historical rainfall intensity s Calculating the matching factor ; Using the formula Correct the initial weight coefficient of the historical biological toxicity parameters, where is the weight coefficient of the historical biological toxicity parameter at the current moment, is the weight coefficient of the historical biological toxicity parameter at the previous moment, α is the preset correction factor for the historical biological toxicity parameter, | α ∣≤0.3, when Δ R ≤ s hour α >0, when ΔR> s hour α <0; The weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters were scaled so that the sum was 1; The instantaneous rate of pollutant load L is calculated using the weighted summation formula: ,in: w i is the dynamically adjusted weight coefficient of the i-th data source, f i ( x i ) is the standardized data mapping function of the i-th data source, and n is the total number of data sources; S3, hierarchical diversion control: When L>the first threshold, the rainwater is distributed to the enhanced treatment unit, which sequentially undergoes cyclone sand settling treatment, electrochemical heavy metal degradation treatment, and chemical flocculation treatment; When L < the second threshold, the rainwater is distributed to the biological aeration filter system for treatment. The effluent of the biological aeration filter system is treated by ultrafiltration membrane and then input into the purified water storage tank. When the second threshold ≤ L ≤ the first threshold, the rainwater is allocated to the advection sedimentation tank for treatment; S4. Dynamic optimization threshold: Real-time access to the current water level of the purified water storage tank, the real-time excess treatment capacity of the sewage treatment plant, and rainfall intensity; The first and second thresholds are dynamically adjusted based on the following conditions: When the real-time surplus processing capacity of the sewage treatment plant is higher than the preset capacity, a first threshold lowering instruction is generated: the first threshold is multiplied by a lowering coefficient, where 0.7≤lowering coefficient<1.0; When the rainfall intensity exceeds the rainstorm threshold and the current water level of the purified water storage tank reaches or exceeds the water level warning value, a second threshold increase instruction is generated: the second threshold is multiplied by the increase coefficient, where 1.0 < increase coefficient ≤ 1.5; The updated first threshold and second threshold are applied in real time to the hierarchical diversion control in step S3.

[0015] The present invention has at least the following beneficial effects: The present invention uses a multi-source spectral water quality monitor to achieve real-time synchronous monitoring of turbidity, COD, TOC and heavy metal concentrations, providing multi-dimensional data support for the quantification of pollution load. L The hierarchical control strategy accurately divides rainwater into high-pollution, conventional pollution and clean stages, and drives the controllable diversion module to distribute it to the enhanced treatment unit, aerated biological filter or horizontal flow sedimentation tank according to the degree of pollution. This design breaks through the "one-size-fits-all" treatment mode of the traditional system, and requires highly polluted rainwater to undergo three-stage enhanced treatment of cyclone sand settling-electrochemical degradation-chemical flocculation to ensure the efficient removal of toxic substances such as heavy metals; clean rainwater is given priority to enter the aerated biological filter for resource reuse, reducing energy consumption and drug consumption.

[0016] The present invention adopts a Bayesian reasoning framework through an adaptive weight model to dynamically correct the weights of historical biotoxicity parameters and quantify the similarity between current rainfall and historical events through a matching factor. This mechanism effectively solves the problem of historical data lag and makes the weight distribution more in line with actual rainfall characteristics.

[0017] The present invention utilizes a collaborative control component to provide real-time feedback on the effluent quality of the electrochemical unit through a heavy metal concentration monitoring module, and implements a four-level gradient compensation strategy: the flocculant dosage is increased by 10%-60% according to the range of the residual heavy metal concentration ΔC. This design breaks through the traditional empirical dosing model and achieves a precise match between the dosage of the agent and the degree of residual pollution.

[0018] The present invention generates pollutant load prediction values ​​based on a time-series prediction algorithm, quantifies the prediction accuracy through a credibility coefficient, and calculates the advance compensation amount ΔE accordingly. This mechanism enables the flocculant dosage to be increased to 1+ΔE times in advance before a sudden increase in pollutant load, thus solving the control lag problem.

[0019] The present invention utilizes a global optimization module to link three data sources: the water level of the regulating reservoir, the load of the sewage treatment plant, and the rainfall intensity, to achieve dynamic adjustment of the threshold value: when the sewage treatment plant has a high surplus treatment capacity, the first threshold value is lowered to increase the frequency of enhanced treatment and make full use of the idle treatment capacity; when there is heavy rain and the water level of the regulating reservoir exceeds the warning level, the second threshold value is raised to reduce the amount of rainwater entering deep treatment and alleviate the overflow risk.

[0020] The present invention utilizes a state constraint layer to obtain the electrode loss rate and membrane pressure difference in real time through the performance monitoring module, establishing a linkage mechanism between the hardware status and the optimization instruction: when the first threshold down-adjustment instruction is triggered, if the electrode loss rate exceeds the limit, the instruction is discarded and an alarm is issued to prevent overload and damage to the electrochemical unit; when the second threshold is increased, if the membrane pressure difference is greater than 90% of the critical value, the increase range is limited to protect the integrity of the ultrafiltration membrane.

[0021] This invention utilizes a redundant diversion component to ensure that, in the event of a controllable diversion module failure, all rainwater is forced into the horizontal flow sedimentation tank within 0.5 seconds via a bypass switching unit, ensuring uninterrupted system operation. This also triggers an equipment fault alarm, locates the fault point, and prompts maintenance. This mechanism reduces the risk of system failure.

[0022] The present invention utilizes the sudden pollution identification layer to generate a spectral anomaly index through mutation gradient detection. The dynamic weight allocation layer immediately forces the calculation weights to be assigned to the real-time monitoring data, ignoring rainfall forecasts and historical parameters. When L>80% of the first threshold, the high pollution stage is directly determined. The response time is less than 3 seconds. This mechanism has a high recognition accuracy for sudden pollution events such as industrial leaks and traffic accidents.

[0023] The present invention utilizes a deep treatment distribution module to limit the amount of rainwater entering the aerated biological filter through a water level-membrane pressure collaborative constraint function and a diversion flow formula. Excess rainwater is switched to the horizontal flow sedimentation tank to avoid overpressure operation of the ultrafiltration membrane under high water level conditions. This strategy reduces the membrane fouling rate.

[0024] The processing method provided by the present invention realizes full-process adaptive control through the coordination of three stages: dynamic weight calculation, emergency response to pipeline sediment disturbance and real-time optimization of thresholds.

[0025] Other advantages, objectives and features of the present invention will be reflected in part from the following description and will be understood by those skilled in the art through study and practice of the present invention. DETAILED DESCRIPTION

[0026] The present invention is further described in detail below with reference to the embodiments so that those skilled in the art can implement the invention with reference to the description.

[0027] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0028] According to one implementation of the present invention, the method includes: Multi-source spectral water quality monitor, installed at the end of the rainwater collection network, monitors the turbidity, COD, TOC and heavy metal concentration of rainwater runoff in real time; The control center receives rain monitoring data from the multi-source spectral water quality monitor, accesses rainfall intensity data from rainfall forecast information, and calls the historical biological toxicity database; The control center uses the weighted summation formula to calculate the instantaneous rate of pollutant load L , the formula is: ,in, w i For the i The weight coefficient of the parameter, f i ( x i ) is the i A standardized data mapping function with parameters, n is the number of parameters, including real-time monitoring of turbidity, COD, TOC, heavy metal concentrations, rainfall intensity, and historical biological toxicity data; Multiple diversion wells are set up at the branch nodes of the rainwater collection network. A controllable diversion module is installed inside each diversion well. The controllable diversion module is connected to the command output terminal of the control center and performs rainwater distribution operations according to the hierarchical instructions of the control center; The inlet of the enhanced treatment unit is connected to the first outlet of the diversion well, and the cyclone sand settling treatment unit, the electrochemical heavy metal degradation treatment unit and the chemical flocculation treatment unit are sequentially arranged inside; a biological aeration filter system, the inlet of which is connected to the second outlet of the diversion well; a horizontal flow sedimentation tank, the inlet of which is connected to the third outlet of the diversion well; Among them, when L >When the first threshold is reached, the controllable diversion module distributes rainwater to the enhanced treatment unit; when L When the value is less than the second threshold, the controllable diversion module distributes the rainwater to the biological aeration filter system; When the second threshold ≤ L When the value is less than or equal to the first threshold, the controllable diversion module distributes the rainwater to the advection sedimentation tank; Among them, for the multi-source spectral water quality monitor, it is installed at the end of the rainwater collection network, and can monitor the turbidity, COD, TOC and heavy metal concentration of rainwater runoff in real time. The monitor can use a commercially available integrated spectrometer analyzer, such as a device that includes an optical probe to measure turbidity, an ultraviolet absorption module to measure COD and TOC, and an electrochemical sensor to detect heavy metals. The shell material can be 316L stainless steel or IP67 waterproof engineering plastic to ensure corrosion resistance and waterproof performance. When working, the multi-source spectral water quality monitor continuously collects rainwater samples, outputs various data in real time through built-in spectral analysis technology, and transmits them to the control center through wired or wireless communication. Among them, the calculation of the instantaneous rate L of pollutant load involves parameter weights. The initial weights can be set to 0.5 for real-time rainwater monitoring data weight, 0.3 for rainfall intensity weight, and 0.2 for historical biological toxicity parameter weight. The parameter standardization process can adopt the minimum-maximum normalization method, and the formula is ,in x min and x max Determined based on historical data sets; The control center receives rain monitoring data and accesses rainfall intensity data in rainfall forecast information, and uses the weighted summation formula Calculate the instantaneous rate of pollutant load L, where w i For the i The weight coefficient of the parameter, f i ( x i ) is the i A standardized data mapping function with parameters, nis the number of parameters, including rain monitoring, rainfall intensity and historical biological toxicity. The control center can use an industrial programmable logic controller or an embedded microprocessor system, and its data processing module runs an adaptive weight algorithm. When working, the control center accesses rainfall forecast information through the public meteorological service API, integrates real-time monitoring data, rainfall intensity and historical biological toxicity parameters, and dynamically adjusts the weight coefficient based on the Bayesian reasoning framework. The calculated L value is used for graded judgment: the first threshold is set to 50kg / h, and the second threshold is set to 10kg / h. When L>50kg / h, it is a high pollution stage, when L<10kg / h, it is a clean stage, and when 10kg / h≤L≤50kg / h, it is a normal pollution stage. Multiple diversion wells are equipped with controllable diversion modules, and the controllable diversion modules are connected to the control center to perform graded treatment according to L. The diversion wells are installed at the branch nodes or key confluence points of the rainwater pipe network. The controllable diversion modules can use electric ball valves or pneumatic butterfly valves and are connected to the control center through the narrowband Internet of Things. When L>50kg / h, the controllable diversion module distributes rainwater to the enhanced treatment unit, which includes a cyclone sand settling treatment unit, an electrochemical degradation heavy metal treatment unit and a chemical flocculation treatment unit arranged in sequence; when L<10kg / h, it is distributed to Aerated biological filter system; when 10kg / h≤L≤50kg / h, it is distributed to the horizontal flow sedimentation tank. The diversion well can be made of concrete or stainless steel to ensure structural strength and durability. When in operation, the controllable diversion module receives instructions from the control center and switches the valve according to the L value to direct the rainwater to the corresponding treatment unit: highly polluted rainwater first undergoes cyclonic sand settling to remove particulate impurities, then undergoes electrochemical degradation of heavy metals, and finally undergoes chemical flocculation and precipitation; clean rainwater enters the aerated biological filter for treatment and reuse; conventional polluted rainwater is initially purified in the horizontal flow sedimentation tank; Through the multi-source spectral water quality monitor to achieve real-time multi-index monitoring, combined with the dynamic calculation of the control center and the hierarchical control of the diversion well, it is possible to accurately identify the degree of rainwater pollution and implement differentiated treatment. This solution can effectively solve the problems of traditional systems that are difficult to monitor multi-index water quality in real time and the lack of dynamic diversion control, and achieve efficient treatment and resource reuse of rainwater by quality, ensuring the safety of receiving water bodies, while optimizing the allocation of treatment resources and reducing energy consumption and chemical consumption. According to another implementation of the present invention, the control center includes an adaptive weight model that dynamically adjusts the weight coefficients of rain monitoring, rainfall intensity, and historical biotoxicity parameters based on a Bayesian inference framework. The dynamic adjustment process includes: Set initial weight coefficients for real-time rain monitoring, rainfall intensity, and historical biotoxicity parameters; According to the current rainfall intensity R t Rainfall intensity compared with similar historical rainfall events R h The absolute difference ΔR and the standard deviation of historical rainfall intensity s Calculating the matching factor ; Using the formula Correct the initial weight coefficient of the historical biological toxicity parameters, where is the weight coefficient of the historical biological toxicity parameter at the current moment, is the weight coefficient of the historical biological toxicity parameter at the previous moment, α is the preset correction factor for the historical biological toxicity parameter, | α ∣≤0.3, when Δ R ≤ s hour α >0, when ΔR> s hour α <0; The weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters were scaled so that the sum was 1; Among them, for the initial weight setting, the adaptive weight model of the control center needs to set the initial weight coefficients of real-time rain monitoring, rainfall intensity, and historical biotoxicity parameters. The initial weight can be set with reference to the empirical value of historical data. For example, the weight of real-time rain monitoring data is set to 0.5, the weight of rainfall intensity is set to 0.3, and the weight of historical biotoxicity parameters is set to 0.2. The control center can use an industrial programmable logic controller or an embedded microprocessor system, whose data processing module has a built-in adaptive weight algorithm. The hardware carrier of the model can be an aluminum alloy chassis or a flame-retardant plastic shell to ensure heat dissipation and safety. It is assembled in the central processing unit of the control center. When working, the initial weight parameters are stored in the register of the control center as the reference value for dynamic adjustment. The parameter setting method is based on the statistical analysis of local historical rainfall and water quality monitoring data. For example, water quality data of similar rainfall events in the past five years are collected for weight calibration. In terms of matching factor calculation, the current rainfall intensity of the model R t Rainfall intensity compared with similar historical rainfall events R h The absolute difference Δ R and the standard deviation of historical rainfall intensity s Calculating the matching factor The historical rainfall intensity data is stored in the database of the control center and can be obtained by querying the rainfall data of the same period in the past three years. The calculation module can use the floating point operation unit in the control center to receive the rainfall intensity data in real time and perform exponential calculations. During the working process, the system obtains the current rainfall intensity in real time. R t , and similar rainfall intensities in historical data R h Compare and calculate Δ R=| R t - R h |, and call the standard deviation of historical rainfall intensity s , the matching factor is calculated by exponential function or ,This factor is used to quantify the similarity between current rainfall and historical events, and provide a basis for weight correction; In the dynamic correction and scaling mechanism of weight coefficient, the formula is used Correct the initial weight coefficient of the historical biological toxicity parameters, where is the weight coefficient of the historical biological toxicity parameter at the current moment, is the weight coefficient of the historical biological toxicity parameter at the previous moment, α is the preset correction factor for the historical biological toxicity parameter, | α ∣≤0.3, when Δ R ≤ s hour α >0, when ΔR> s hour α <0, after correction, the weight coefficients of the three types of parameters are scaled proportionally to a total of 1. The value of the correction coefficient α can be set according to local rainfall characteristics. For example, in rainy areas, α can be set to 0.2, and in less rainy areas, it can be set to 0.15. The scaling algorithm can use a normalization formula to ensure that the sum of the weights is 1. When working, the system selects the sign of α according to the comparison result of ΔR and σ, substitutes it into the formula to calculate the weight of the current historical biological toxicity parameter, and then normalizes it with the weights of other parameters to generate a dynamically adjusted weight vector for the calculation of the instantaneous rate L of pollutant load; By dynamically adjusting the weight coefficients through the Bayesian inference framework, the weight distribution can be optimized in real time based on the similarity between current rainfall and historical events, effectively solving the lag problem of historical data and making the calculation of the instantaneous rate of pollutant load more in line with actual rainfall characteristics. This mechanism improves the accuracy of pollution level identification and provides a more scientific decision-making basis for diversion control, thereby optimizing the targetedness and effectiveness of rainwater classification treatment and ensuring the stable operation of the treatment system under different rainfall conditions.

[0029] According to another implementation of the present invention, a collaborative control component is further included, which includes: The heavy metal concentration monitoring module is installed at the outlet of the electrochemical heavy metal degradation treatment unit to detect the residual heavy metal concentration in real time; The parameter collaborative controller is connected to the heavy metal concentration monitoring module and the reagent dosing system in the chemical flocculation treatment unit. The parameter collaborative controller receives the residual heavy metal concentration data and implements a hierarchical control strategy: When the residual heavy metal concentration is ≤ the safety threshold, the chemical dosing system maintains the dosage of chemical flocculants; When the residual heavy metal concentration is greater than the safety threshold, the parameter collaborative controller performs the following operations: Calculate the concentration value Δ at which the residual heavy metal concentration exceeds the safety threshold C , according to Δ C In the preset concentration range, select the corresponding dosage increase ratio m %, where the preset concentration range is m The corresponding relationship of % is: i). If 0μg / L<ΔC≤5μg / L, then m %=10%; ii). If 5μg / L<ΔC≤10μg / L, then m %=20%; iii). If 10μg / L<ΔC≤20μg / L, then m %=40%; iv). If ΔC>20μg / L, then m %=60%; Increase the amount of flocculant to 1+ of the original amount m % times; Among them, the collaborative control component includes a heavy metal concentration monitoring module (located at the outlet of the electrochemical degradation heavy metal treatment unit) and a parameter collaborative controller. The parameter collaborative controller receives the residual heavy metal concentration data and executes a hierarchical control strategy: when the residual heavy metal concentration is ≤ the safety threshold (0.5 mg / L), the flocculant dosage is maintained; when it exceeds the standard, the dosage is increased according to the ΔC range (10%-60%). As for the heavy metal concentration monitoring module, it is located at the outlet of the electrochemical degradation heavy metal treatment unit to detect the residual heavy metal concentration in real time. The heavy metal concentration monitoring module can be a commercially available Online heavy metal analyzers, such as those equipped with electrochemical sensors or spectral analysis probes, can measure the concentrations of heavy metal ions such as lead and cadmium. The housing material of the monitor can be made of 316L stainless steel to ensure corrosion resistance. It is installed on the pipe 1 meter away from the outlet of the electrochemical degradation heavy metal treatment unit. The probe insertion depth is 1 / 3 of the pipe diameter to ensure measurement accuracy. When working, the heavy metal concentration monitoring module continuously collects effluent samples and transmits the data to the parameter collaborative controller via 4~20mA current signal or RS485 communication. The sampling frequency is once per minute. The parameter collaborative controller is connected to the heavy metal concentration monitoring module and the chemical dosing system in the chemical flocculation treatment unit to receive residual heavy metal concentration data. The parameter collaborative controller can adopt an industrial programmable logic controller or an embedded control unit, which is installed in the cabinet of the control center and communicates with the heavy metal concentration monitoring module and the chemical dosing system through a shielded cable. The controller has a built-in hierarchical control strategy program. When the monitoring data is received, it is immediately compared with the preset safety threshold. The safety threshold can be set to 0.5 mg / L. This value is set based on local environmental protection emission standards or water quality requirements of the receiving water body and is manually entered into the controller through the human-machine interface; When the hierarchical control strategy is executed, when the residual heavy metal concentration is ≤0.5mg / L, the chemical dosing system maintains the dosage of the chemical flocculant; when the residual concentration is >0.5mg / L, the excess value ΔC is calculated, and the corresponding dosage increase ratio m% is selected according to the preset concentration range of ΔC, specifically: when 0μg / L<ΔC≤5μg / L, m%=10%; when 5μg / L<ΔC≤10μg / L, m%=20%; when 10μg / L<ΔC≤20μg / L, m%=4 0%; when ΔC>20μg / L, m%=60%. The chemical dosing system can use a precision metering pump with a polyethylene drug storage tank, installed next to the chemical flocculation treatment unit. The pump flow adjustment range is 10~100L / h. When working, the controller determines m% according to ΔC and increases the flocculant dosage to 1+m% times the original dosage. For example, if the original dosage is 20L / h and ΔC=8μg / L, m%=20%, and the new dosage is 20×(1+20%)=24L / h. Through real-time monitoring of heavy metal concentrations and graded linkage of reagent dosage, precise control of chemical flocculation treatment is achieved. This mechanism breaks through the limitations of the traditional empirical dosage model and directly matches the reagent dosage with the degree of pollution residue. While ensuring the heavy metal removal effect, it reduces reagent waste, lowers treatment costs, improves the economy and reliability of system operation, and ensures that the effluent water quality meets the emission standard requirements.

[0030] According to another implementation of the present invention, the parameter collaborative controller is simultaneously connected to an adaptive weight model. Before executing the hierarchical control strategy, the parameter collaborative controller advance-corrects the flocculant dosage based on the instantaneous rate of pollutant load. The adaptive weight model is configured to execute: Instantaneous rate of pollutant load based on real-time calculation L t , generates instantaneous rate predictions of pollutant loads through built-in time series prediction algorithms : Will L t and history NThe instantaneous rate data of pollutant loads at consecutive time steps constitute a time series; Use the time series prediction algorithm to make a single-step prediction of the sequence and output the predicted value at the next acquisition moment , the time series prediction algorithm is any one of the sliding average model, autoregressive integrated sliding average model or long short-term memory network; Through the control center t +1 moment to calculate the actual pollutant load instantaneous rate monitoring value L t+1 : Comparison of predicted values Compared with the actual monitoring value L t+1 The mean absolute error ε, μ is used to calculate the prediction confidence coefficient μ=e -λε , λ is the error sensitivity coefficient, ranging from 0.5 to 1.0; The instantaneous rate prediction of pollutant load is multiplied by k The benchmark lead compensation is calculated, where: k It is the preset compensation coefficient, ranging from 0.1 to 0.3; Multiply the reference lead compensation by the prediction credibility coefficient to generate the effective lead compensation ΔE; According to ΔE, increase the flocculant dosage to 1+ΔE times of the original dosage; Among them, for the application of time series prediction algorithm, the parameter collaborative controller is connected to the adaptive weight model, based on the instantaneous rate of pollutant load calculated in real time L t , generates forecast values ​​through built-in time series forecasting algorithm , the time series is given by L t It is composed of L data of N consecutive historical time steps, where N can be set to 10~20, and is adjusted according to data stability. The prediction algorithm can use a sliding average model, an autoregressive integral moving average model (ARIMA) or a long short-term memory network (LSTM). The window size of the sliding average model can be set to 5~10 time steps. The controller hardware can use an industrial-grade programmable logic controller or an embedded computing unit, which is installed in the cabinet of the control center and connected to the adaptive weight model through a data interface. When working, the system stores the historical L data in the controller database, constructs the input matrix according to the time series, performs single-step prediction through the selected algorithm, and outputs the predicted value at the next acquisition moment. , the sampling interval can be set to 10~15 minutes; In terms of the prediction reliability coefficient calculation, by comparing the predicted values Compared with the actual monitoring value L t+1 The mean absolute error ε, using the formula μ = e-λε Calculate the credibility coefficient μ, where λ is the error sensitivity coefficient, with a value range of 0.5~1.0. The specific setting can be based on the local water quality fluctuation characteristics. For example, in areas with large water quality fluctuations, λ is set to 0.8. The error calculation module can be integrated into the data processing unit of the controller, receiving the predicted value and the actual monitoring value in real time, and generating ε and μ through arithmetic operations. During operation, after obtaining the actual monitoring value at time t+1, the system calculates the absolute error between the two and averages them, and substitutes them into the exponential function to obtain μ. This coefficient is used to quantify the reliability of the prediction results and provide a weight basis for the calculation of the compensation amount; In the mechanism of advance compensation amount generation and agent dosage adjustment, the benchmark advance compensation amount is determined by the predicted value. Multiply by the preset compensation coefficient k get, k The value range is 0.1~0.3, for example k =0.2, the effective advance compensation amount ΔE is the benchmark advance compensation amount multiplied by μ, and the flocculant dosage is finally increased to 1+ΔE times of the original dosage. The agent dosing system can use a precision metering pump with a polyethylene storage tank, which is installed next to the chemical flocculation treatment unit. The flow adjustment range of the pump is 10-100L / h. When working, the controller generates a control signal according to ΔE to drive the metering pump to adjust the dosage acceleration rate. For example, if the original dosage is 30L / h, when ΔE=0.15, the new dosage is 30×(1+0.15)=34.5L / h, and the compensation coefficient k The setting method is based on historical data calibration, by comparing different k The optimal value is selected based on the dosage effect and treatment cost under the given value. By combining the time series prediction algorithm with the credibility coefficient, advance prediction of pollutant load changes is achieved, so that the flocculant dosage can be adjusted in advance before the pollutant load suddenly increases. This mechanism effectively solves the problem of control lag in traditional systems, reduces fluctuations in treatment effects caused by load fluctuations, improves the accuracy of agent addition, reduces agent consumption while ensuring treatment effects, and optimizes the economy and stability of system operation.

[0031] According to another implementation of the present invention, the effluent from the biological aeration filter system is treated by an ultrafiltration membrane and then enters the purified water storage tank; The urban road rainwater runoff classification treatment system further includes: The global optimization module is connected to the control center, purified water storage tank, sewage treatment plant, and rainfall forecast information, and is configured as follows: Real-time access to the current water level of the purified water storage tank, the real-time excess treatment capacity of the sewage treatment plant, and rainfall intensity; Dynamically adjust the first and second thresholds based on the above parameters: When the real-time surplus processing capacity of the sewage treatment plant is higher than the preset capacity, a first threshold lowering instruction is generated, and the first threshold is multiplied by a lowering coefficient, where 0.7≤lowering coefficient<1.0; When the rainfall intensity exceeds the rainstorm threshold and the current water level of the purified water storage tank reaches or exceeds the water level warning value, a second threshold increase instruction is generated, and the second threshold is multiplied by the increase coefficient, where 1.0 < increase coefficient ≤ 1.5; Among them, for the effluent treatment of the biological aeration filter system, its effluent is treated by ultrafiltration membrane and then enters the purified water storage tank. The ultrafiltration membrane component can use a commercially available hollow fiber membrane system, made of polyvinylidene fluoride (PVDF) or polyethersulfone (PES), with a membrane pore size range of 0.01~0.1μm. It is assembled on the effluent pipe of the biological aeration filter, 1~2 meters away from the filter outlet, and connected through a flange. The purified water storage tank can be made of reinforced concrete or stainless steel. A liquid level sensor is installed in the tank to monitor the real-time water level. The sensor is installed on the tank wall 0.5 meters from the bottom. During operation, the effluent of the biological aeration filter enters the ultrafiltration membrane component through the pipe, is filtered under pressure, and the produced water flows into the storage tank, and the concentrated water returns to the front-end treatment unit. The global optimization module is connected to the control center, purified water storage tank, and sewage treatment plant respectively, and is connected to rainfall forecast information. The module can use an industrial computer system and be installed in the central control room. It communicates with each unit through a data interface. The hardware uses a standard server or edge computing device. The shell is an aluminum alloy chassis to ensure heat dissipation. The global optimization module is connected to the real-time surplus capacity interface of the sewage treatment plant, the water level sensor of the storage tank, and the meteorological API. The global optimization module obtains the current water level of the storage tank (through the set liquid level sensor), the real-time surplus processing capacity of the sewage treatment plant (through the set sewage plant data interface), and the rainfall intensity (through the meteorological API) in real time. The data is updated every 15 minutes to ensure timeliness. When working, the module integrates the three-source data, generates threshold adjustment instructions through the built-in algorithm, and transmits them to the control center for execution; The conditions for dynamically adjusting the thresholds are as follows: When the real-time surplus treatment capacity of the sewage treatment plant is higher than the preset capacity, the first threshold is multiplied by a downward adjustment coefficient K1 (0.7 ≤ K1 < 1.0, for example, K1 = 0.8); when the rainfall intensity exceeds the heavy rain threshold (such as 50 mm / h) and the water level of the storage tank reaches the warning value (such as 80% of the design water level), the second threshold is multiplied by an upward adjustment coefficient K2 (1.0 < K2 ≤ 1.5, for example, K2 = 1.2). The preset capacity can be set to 30% of the designed treatment capacity of the sewage treatment plant and is set in the module through the human-machine interface. The heavy rain threshold is determined based on local meteorological standards. The warning value of the water level in the storage tank is 80% of the tank height. During operation, the module continuously judges the conditions and generates an instruction when the conditions are met. For example, when the surplus treatment capacity of the sewage treatment plant reaches 40%, the first threshold is lowered from 50 kg / h to 50 × 0.8 = 40 kg / h; when the rainfall intensity is 60 mm / h and the water level of the storage tank reaches 80%, the second threshold is raised from 10 kg / h to 10 × 1.2 = 12 kg / h. The adjusted thresholds are fed back to the diversion control module in real time; Through the global optimization module, the water level of the storage tank, the load of the sewage treatment plant, and the rainfall intensity are linked to achieve dynamic adjustment of the thresholds. When the surplus treatment capacity of the sewage treatment plant is high, the first threshold is lowered to increase the frequency of enhanced treatment and make full use of idle resources; when there is heavy rain and the water level of the storage tank exceeds the warning level, the second threshold is raised to reduce the amount of deep treatment water and alleviate the overflow risk. This mechanism improves the adaptability of the system to external environmental changes, optimizes the allocation of treatment resources, ensures the stable operation of the system under different working conditions, and at the same time improves the efficiency of rainwater resource reuse.

[0032] According to another implementation manner of the present invention, it further includes: An efficiency monitoring module that obtains the electrode loss rate of the electrochemical degradation electrode and the transmembrane pressure difference of the ultrafiltration membrane of the enhanced treatment unit in real time; A state constraint layer, connecting the efficiency monitoring module and the global optimization module, is configured to execute: Receive the threshold adjustment instruction generated by the global optimization module; If a downward adjustment instruction for the first threshold is received and the current electrode loss rate > the upper limit of the loss rate, then discard the downward adjustment instruction and trigger an electrode loss warning; otherwise, allow the execution of the downward adjustment instruction; If an upward adjustment instruction for the second threshold is received and the real-time transmembrane pressure difference of the ultrafiltration membrane > the critical pressure-bearing value, then limit the upward adjustment range so that the adjusted second threshold does not exceed the initial threshold; otherwise, allow the execution of the instruction according to the upward adjustment coefficient; The performance monitoring module collects the electrode loss rate and ultrafiltration membrane pressure difference in real time, and constrains the threshold adjustment instructions. It actually obtains the electrochemical degradation electrode loss rate and ultrafiltration membrane transmembrane pressure difference of the enhanced processing unit in real time. The monitoring of the electrochemical degradation electrode loss rate can use a resistive sensor to calculate the loss degree by measuring the resistance change of the electrode material. The sensor is installed on the electrode frame of the electrochemical degradation heavy metal treatment unit, 5 to 10 mm away from the electrode surface; the monitoring of the ultrafiltration membrane transmembrane pressure difference can use a high-precision pressure sensor, which is installed on the inlet and outlet pipes of the ultrafiltration membrane assembly, within 1 meter of the membrane assembly to ensure measurement accuracy. The monitoring range of the electrode loss rate can be set to 0 to 100%, and the pressure threshold of the ultrafiltration membrane transmembrane pressure difference can be set to 0.3 MPa. The housing material of the sensor can be 316L stainless steel, and the cable sheath can be made of waterproof fluororubber to ensure corrosion resistance and waterproof performance. When working, the performance monitoring module collects data at a frequency of 1 time per minute and transmits it to the state constraint layer through a 4 to 20 mA current signal. The state constraint layer connects the performance monitoring module and the global optimization module to execute the constraints of the threshold adjustment instructions. The state constraint layer can adopt an industrial-grade programmable logic controller, which is installed in the cabinet of the control center and communicates with the performance monitoring module and the global optimization module through the data interface. When receiving the first threshold reduction instruction from the global optimization module, the state constraint layer reads the current electrode loss rate. If the loss rate is greater than 30% (the upper limit of the loss rate), the reduction instruction is discarded and the electrode loss warning is triggered; if the loss rate is ≤30%, the reduction instruction is allowed to be executed. When receiving the second threshold increase instruction, the state constraint layer reads the real-time ultrafiltration membrane transmembrane pressure difference. If the pressure difference is greater than 0.3MPa (the pressure critical value), the increase range is limited so that the adjusted second threshold does not exceed the initial threshold; if the pressure difference is ≤0.3MPa, the instruction is allowed to be executed according to the increase coefficient. The setting of the electrode loss rate upper limit of 30% is based on the service life curve of the electrode material and is determined through accelerated aging experiments; the transmembrane pressure difference critical value of 0.3MPa is the safe working pressure upper limit recommended by the ultrafiltration membrane manufacturer; Through the linkage between the performance monitoring module and the state constraint layer, a real-time feedback mechanism for hardware status and optimization instructions is established. When the electrode loss rate exceeds the limit or the transmembrane pressure difference of the ultrafiltration membrane is too high, the state constraint layer promptly intervenes with the threshold adjustment instruction to avoid overload damage to the electrochemical unit or damage to the integrity of the ultrafiltration membrane due to overpressure. This mechanism ensures the feasibility of the global optimization strategy while ensuring the safe operation of the equipment, extending the service life of key equipment and improving the reliability and stability of the system operation.

[0033] According to another implementation of the present invention, the method further includes: Redundant traffic protection components, including: Fault diagnosis module, real-time monitoring of the operating status of the controllable shunt module; Bypass switching unit, connecting the inlet of the horizontal flow sedimentation tank and the end of the rainwater collection network; When the fault diagnosis module detects that the controllable diversion module has failed, it automatically activates the bypass switching unit to force all rainwater runoff into the horizontal flow sedimentation tank and sends an equipment fault alarm to the control center. Among them, the redundant diversion protection component includes a fault diagnosis module (monitoring valve current / displacement) and a bypass switching unit (connecting the pipe network and the horizontal flow sedimentation tank). The bypass switching valve adopts a DN300-DN500 pneumatic gate valve with an action time of ≤0.5 seconds. When a valve failure is detected, it is forced to switch to the horizontal flow sedimentation tank within 0.5 seconds and an alarm signal is issued. For the fault diagnosis module, it actually monitors the operating status of the controllable diversion module in real time. The module can use a current sensor or a displacement sensor to monitor the current signal or valve opening of the controllable diversion module (such as an electric ball valve). The sensor shell material can be made of 316L stainless steel to ensure corrosion resistance. It is installed near the controllable diversion module in the diversion well, 10 to 20 cm away from the valve actuator, and connected to the control center through a shielded cable. The sampling frequency of fault diagnosis can be set to 10 times / second. When the monitored current is abnormal (such as exceeding 120% of the rated current) or the valve opening feedback does not match the instruction (the deviation exceeds 5%), it is determined that the controllable diversion module has failed. The specific value of the rated current is determined according to the model parameters of the selected valve, and the valve opening deviation threshold of 5% is determined through debugging; The bypass switching unit connects the inlet of the horizontal flow sedimentation tank to the end of the rainwater collection network. When the fault diagnosis module detects a failure, it automatically activates and diverts all rainwater runoff to the horizontal flow sedimentation tank, while sending an alarm signal. The bypass switching unit can use an electric butterfly valve or a pneumatic gate valve. The valve body material can be ductile iron, and the sealing ring is made of EPDM rubber. It is installed on the connecting pipe between the inlet pipe of the horizontal flow sedimentation tank and the end of the rainwater network, 1 to 2 meters away from the inlet of the horizontal flow sedimentation tank. The valve diameter is determined by the network flow, such as DN300 to DN500, and the actuation time is ≤0.5 seconds. During operation, the fault diagnosis module transmits the failure signal to the controller of the bypass switching unit. The controller drives the valve to fully open within 0.5 seconds and sends an equipment fault alarm to the control center via wired or wireless communication. The alarm information includes the fault type and location; Through the linkage between the fault diagnosis module and the bypass switching unit, emergency diversion can be achieved when the controllable diversion module fails. This mechanism ensures that the system can continue to operate in the event of equipment failure, avoiding the direct discharge of untreated rainwater. At the same time, it triggers an alarm to prompt maintenance, reducing the risk of system failure, improving the reliability and safety of the overall operation, and ensuring the continuity of rainwater runoff treatment on urban roads.

[0034] According to another implementation of the present invention, the adaptive weight model is provided with a burst pollution identification layer, which performs: Real-time calculation of the sudden change gradient of turbidity, COD, TOC and heavy metal concentration of rainwater runoff output by multi-source spectral water quality monitor; When any of the above mutation gradients exceeds the maximum fluctuation threshold in the historical biological toxicity correlation parameters, a spectral anomaly index is generated, and the maximum fluctuation threshold is based on the 95th percentile value of historical data statistics; The dynamic weight allocation layer responds to the spectral anomaly index and forces the weights of the pollutant load instantaneous rate calculation to be assigned to the real-time rainwater monitoring data, while ignoring the rainfall forecast information and historical biotoxicity correlation parameters; At this time, if the instantaneous rate of pollutant load exceeds 80% of the first threshold, the control center will determine that L > first threshold; Among them, for the calculation of mutation gradient, the sudden pollution identification layer of the adaptive weight model calculates the mutation gradient of turbidity, COD, TOC and heavy metal concentration output by the multi-source spectral water quality monitor in real time. The mutation gradient can be obtained by dividing the difference between adjacent time steps by the time interval. For example, using a 1-minute time interval, the calculation formula is: mutation gradient = | X t - X t-1 | / Δt, where X t is the parameter value at the current moment, X t-1 is the parameter value at the previous moment, Δt is the time interval (1 minute), and the maximum fluctuation threshold of the historical biological toxicity-related parameters is based on the 95th percentile value of the historical data statistics. For example, by collecting data from the past three years, the 95th percentile value of the mutation gradient of each parameter is calculated as the threshold value. The turbidity threshold may be 20NTU / min, and the COD threshold may be 15mg / L / min. The multi-source spectral water quality monitor is installed at the end of the rainwater collection network, and its data output interface is connected to the sudden pollution identification layer module of the control center. The module can use an industrial-grade data processing unit and be installed in the control center cabinet; In terms of spectral anomaly index generation and weight redistribution, when any mutation gradient exceeds the corresponding maximum fluctuation threshold, a spectral anomaly index is generated. The dynamic weight distribution layer responds to the index and forcibly distributes the weight of the pollutant load instantaneous rate calculation to the real-time rainwater monitoring data, ignoring the rainfall forecast and historical biological toxicity parameters. The generation logic of the spectral anomaly index is: when any parameter mutation gradient exceeds the historical 95th percentile threshold, the spectral anomaly index = 1; otherwise, the spectral anomaly index = 0. The spectral anomaly index triggers dynamic weight forced distribution. When the weight is redistributed, the weight of the real-time monitoring data is set to 1, and the weights of other parameters are set to 0. The dynamic weight distribution layer can be used as an algorithm module in the control center, running on an industrial programmable logic controller or embedded system, and connected to the adaptive weight model through a data bus. When working, the system continuously monitors the mutation gradient of each parameter and triggers weight redistribution when it exceeds the threshold to ensure that real-time data dominates the pollution load calculation. In the pollution degree mandatory judgment mechanism, if the instantaneous rate of pollutant load exceeds 80% of the first threshold, the control center determines that L> the first threshold. The first threshold can be set to 50kg / h, and 80% of it is 40kg / h. After the weight is redistributed, the system recalculates the L value. If L> 40kg / h, it is directly judged as a high pollution stage. The judgment logic is executed by the decision module of the control center. The module can use an industrial-grade computer and be installed in the central control room. It receives the weight distribution result and the L calculation value through the network interface. When working, the module compares the L value with 40kg / h in real time. When the conditions are met, it sends a high pollution stage instruction to the diversion system. The response time is ≤ 3 seconds; Through sudden gradient detection and forced weight redistribution, sudden pollution incidents such as industrial leaks and traffic accidents can be quickly identified. When pollutant indicators suddenly become abnormal, the system will prioritize determining the degree of pollution based on real-time monitoring data, shortening the response time and ensuring that highly polluted rainwater enters the enhanced treatment unit in a timely manner, avoiding the impact of sudden pollution on the receiving water body and improving the safety and reliability of the system in responding to emergency pollution incidents.

[0035] According to another implementation of the present invention, the global optimization module is provided with a depth processing allocation module, which performs: Real-time acquisition of the water level of the purified water storage tank and the transmembrane pressure difference of the ultrafiltration membrane; Construct the water level-membrane pressure synergistic constraint function: When the water level in the regulating reservoir is greater than the warning water level and the transmembrane pressure difference of the ultrafiltration membrane is greater than 90% of the critical value, the control center dynamically increases the second threshold; The controllable diversion module distributes rainwater to the aerated biological filter system according to the formula Q = β× ( S max - S 当前 ) for current limiting, whereQ To allow for flow diversion, β is the attenuation coefficient, S max is the maximum safe capacity of the storage tank. S 当前 is the real-time water level; Excess rainwater is switched to the horizontal flow sedimentation tank through the controllable diversion module; Among them, for data acquisition, the deep processing distribution module obtains the water level of the purified water storage tank and the transmembrane pressure difference of the ultrafiltration membrane in real time, and limits the water inlet of the aerated biological filter based on the water level-membrane pressure synergistic function. The water level monitoring of the storage tank can use an ultrasonic liquid level sensor, which is installed on the inner wall of the storage tank 0.5 meters from the bottom, with a measurement range of 0~10 meters; the transmembrane pressure difference of the ultrafiltration membrane can use a high-precision pressure sensor to monitor the transmembrane pressure difference of the ultrafiltration membrane, which is installed on the inlet and outlet pipes of the ultrafiltration membrane assembly, within 1 meter from the membrane assembly, with a measurement range of 0~0.6MPa. The housing material of the sensor can be 316L stainless steel, and the cable sheath can be waterproof fluororubber. The deep processing distribution module can use an industrial-grade data acquisition unit, which is installed in the control center cabinet and communicates with the sensor through the 485 bus. The data update frequency is once per minute. In the construction of water level-membrane pressure collaborative constraint function and threshold adjustment, when the water level of the regulating reservoir is greater than the warning water level (such as the maximum safe capacity of the regulating reservoir), S max When the water level exceeds 80% of the critical value (e.g., 0.3 MPa) and the transmembrane pressure difference of the ultrafiltration membrane is greater than 90% (i.e., 0.27 MPa) of the critical value (e.g., 0.3 MPa), the control center dynamically increases the second threshold. The initial value of the second threshold is set to 10 kg / h. The increase is determined according to the degree of deviation between the water level and the membrane pressure difference, for example, it is increased by 10% each time. The logic of the constraint function is executed by the built-in algorithm of the deep processing allocation module. The module can use an industrial computer, installed in the central control room, and connected to the global optimization module through a data interface. When working, the module continuously compares the water level and membrane pressure difference data. When the conditions are met, it generates a threshold increase instruction and transmits it to the control center for execution. In the rainwater diversion and flow limiting calculation and switching mechanism, the controllable diversion module allocates the amount of rainwater to the aerated biological filter according to the formula Q = β× ( S max - S 当前 ) for current limiting, where Q To allow for flow diversion, β is the attenuation coefficient (range 0.5~0.8, for example 0.6), S 当前For real-time water level, excess rainwater is switched to the horizontal flow sedimentation tank through the controllable diversion module. The controllable diversion module can use an electric butterfly valve, which is installed on the connecting pipe between the diversion well and the aerated biological filter. The valve body material is ductile iron, and the drive motor response time is ≤10 seconds. When working, the module adjusts the valve opening according to the calculated Q value, for example S max =1000m 3 , S 当前 =850m 3 , when β=0.6, Q=0.6×(1000-850)=90m 3 / h, the excess part is directed to the horizontal flow sedimentation tank, Through the water level-membrane pressure collaborative constraint function and the diversion flow calculation formula, dynamic flow limiting of the aerated biological filter inlet is achieved. When the water level in the regulating reservoir is too high and the transmembrane pressure difference of the ultrafiltration membrane is close to the critical value, the system automatically raises the second threshold and limits the diversion flow to prevent the ultrafiltration membrane from operating at overpressure under high load conditions. At the same time, excess rainwater is switched to the horizontal flow sedimentation tank to balance the load of each treatment unit. This mechanism effectively reduces the membrane fouling rate, extends the service life of the ultrafiltration membrane, ensures the stable operation of the system under extreme conditions such as heavy rain, and improves the overall treatment efficiency and safety.

[0036] A treatment method of the urban road rainwater runoff classification treatment system is provided, comprising the following steps: S1. Real-time monitoring and data acquisition: A multi-source spectral water quality monitor installed at the end of the rainwater collection network monitors the turbidity, COD, TOC, and heavy metal concentrations of rainwater runoff in real time. Synchronously access real-time rainfall intensity in rainfall forecast information; S2. Dynamic calculation of pollutant load instantaneous rate L: Standardize real-time rain monitoring data, rainfall intensity, and historical biotoxicity correlation parameters and remove outliers; Dynamically adjust weight coefficients based on the Bayesian inference framework: Set initial weights for real-time monitoring data, rainfall intensity, and historical biotoxicity parameters; According to the current rainfall intensity R t Rainfall intensity compared with similar historical rainfall events R h The absolute difference Δ R and the standard deviation of historical rainfall intensity s Calculating the matching factor ; Using the formula Correct the initial weight coefficient of the historical biological toxicity parameters, where is the weight coefficient of the historical biological toxicity parameter at the current moment, is the weight coefficient of the historical biological toxicity parameter at the previous moment, α is the preset correction factor for the historical biological toxicity parameter, | α ∣≤0.3, when Δ R ≤ s hour α >0, when ΔR> s hour α <0; The weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters were scaled so that the sum was 1; The instantaneous rate of pollutant load L is calculated using the weighted summation formula: ,in: w i is the dynamically adjusted weight coefficient of the i-th data source, f i ( x i ) is the standardized data mapping function of the i-th data source, and n is the total number of data sources; S3, hierarchical diversion control: When L>the first threshold, the rainwater is distributed to the enhanced treatment unit, which sequentially undergoes cyclone sand settling treatment, electrochemical heavy metal degradation treatment, and chemical flocculation treatment; When L < the second threshold, the rainwater is distributed to the biological aeration filter system for treatment. The effluent of the biological aeration filter system is treated by ultrafiltration membrane and then input into the purified water storage tank. When the second threshold ≤ L ≤ the first threshold, the rainwater is allocated to the advection sedimentation tank for treatment; S4. Dynamic optimization threshold: Real-time access to the current water level of the purified water storage tank, the real-time excess treatment capacity of the sewage treatment plant, and rainfall intensity; The first and second thresholds are dynamically adjusted based on the following conditions: When the real-time surplus processing capacity of the sewage treatment plant is higher than the preset capacity, a first threshold lowering instruction is generated: the first threshold is multiplied by a lowering coefficient, where 0.7≤lowering coefficient<1.0; When the rainfall intensity exceeds the rainstorm threshold and the current water level of the purified water storage tank reaches or exceeds the water level warning value, a second threshold increase instruction is generated: the second threshold is multiplied by the increase coefficient, where 1.0 < increase coefficient ≤ 1.5; Applying the updated first threshold and second threshold to the hierarchical diversion control in step S3 in real time; Among them, real-time monitoring and data acquisition: A multi-source spectral water quality monitor installed at the end of the rainwater collection network, 1 meter from the outlet, collects turbidity, COD, TOC, and heavy metal concentrations of rainwater runoff in real time. This monitor can use commercially available devices that integrate optical probes, UV absorption modules, and electrochemical sensors. The housing is made of 316L stainless steel or IP67 waterproof engineering plastic to ensure corrosion resistance and waterproof performance. Monitoring data is transmitted to the control center via a 4-20mA current signal or wireless communication in a 2-minute cycle. Simultaneously, the control center accesses rainfall forecast information in real time through the public meteorological service API, obtaining real-time rainfall intensity data with an update frequency of 15 minutes. The control center can use an industrial programmable logic controller (PLC) or embedded microprocessor system, installed in the central control room, to ensure the real-time and stable data processing. Dynamic calculation of pollutant load instantaneous rate L: First, the real-time rain monitoring data, rainfall intensity and historical biotoxicity correlation parameters were standardized using the minimum-maximum normalization formula. ,in x min and x max Based on the local historical data set (such as turbidity range 0~1000NTU), and using the 3σ principle to eliminate outliers, the weight coefficients are dynamically adjusted based on the Bayesian reasoning framework: the initial weight of real-time monitoring data is set to 0.5, the weight of rainfall intensity is set to 0.3, and the weight of historical biological toxicity parameters is set to 0.2; according to the current rainfall intensity R t Similar rainfall intensity in history R h The absolute difference ΔR and the historical standard deviation σ are calculated by the formula Calculate the matching factor; when ΔR≤σ, the historical biological toxicity parameter correction coefficient α is 0.2, and when ΔR>σ, α is -0.15, through the formula Correct the weights and finally scale them so that the sum of the weights is 1, using the weighted sum formula Calculate the instantaneous rate of pollutant load L, where w i is the dynamically adjusted weight, f i ( x i ) is the standardized mapping function, n=3, Hierarchical diversion control and threshold dynamic optimization: When L>50kg / h, the controllable diversion module distributes rainwater to the enhanced treatment unit, which performs cyclone sand settling, electrochemical degradation of heavy metals and chemical flocculation treatment in sequence; when L<10kg / h, it is distributed to the aerated biological filter system, and the effluent is treated by the ultrafiltration membrane and then input into the purified water storage tank; when 10kg / h≤L≤50kg / h, it is distributed to the horizontal flow sedimentation tank. The controllable diversion module in the diversion well can use an electric ball valve or a pneumatic butterfly valve, which is installed at the branch node of the rainwater pipe network and connected to the control unit through the narrowband Internet of Things. When the central connection and dynamic optimization threshold are used, the water level of the regulating reservoir (monitored by an ultrasonic liquid level sensor installed on the reservoir wall 0.5 meters from the bottom), the real-time surplus treatment capacity of the sewage treatment plant, and the rainfall intensity are obtained in real time: when the surplus treatment capacity of the sewage treatment plant exceeds 30% of the design capacity, the first threshold is multiplied by 0.8 and adjusted downward; when the rainfall intensity exceeds 50mm / h and the water level of the regulating reservoir reaches 80% of the warning value, the second threshold is multiplied by 1.2 and adjusted upward. The updated threshold is applied to the diversion control module in real time to ensure that the diversion strategy adapts to changes in working conditions; This treatment method achieves precise control of rainwater classification treatment through real-time monitoring of multi-source data, dynamic weight calculation and threshold optimization. It can effectively deal with rainwater runoff with different degrees of pollution. While ensuring the safety of receiving water bodies, it improves the efficiency of rainwater resource reuse and optimizes the energy consumption and chemical consumption of the treatment system.

[0037] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. Urban road rainwater runoff classification treatment system, characterized by: include: Multi-source spectral water quality monitor, installed at the end of the rainwater collection network, monitors the turbidity, COD, TOC and heavy metal concentration of rainwater runoff in real time; The control center receives rain monitoring data from the multi-source spectral water quality monitor, accesses rainfall intensity data from rainfall forecast information, and calls the historical biological toxicity database; The control center uses the weighted summation formula to calculate the instantaneous rate of pollutant load L , the formula is: ,in, w i For the i The weight coefficient of the parameter, f i ( x i ) is the i A standardized data mapping function with parameters, n is the number of parameters, including real-time monitoring of turbidity, COD, TOC, heavy metal concentrations, rainfall intensity, and historical biological toxicity data; Multiple diversion wells are set up at the branch nodes of the rainwater collection network. A controllable diversion module is installed inside each diversion well. The controllable diversion module is connected to the command output terminal of the control center and performs rainwater distribution operations according to the hierarchical instructions of the control center; The inlet of the enhanced treatment unit is connected to the first outlet of the diversion well, and the cyclone sand settling treatment unit, the electrochemical heavy metal degradation treatment unit and the chemical flocculation treatment unit are sequentially arranged inside; a biological aeration filter system, the inlet of which is connected to the second outlet of the diversion well; a horizontal flow sedimentation tank, the inlet of which is connected to the third outlet of the diversion well; Among them, when L >When the first threshold is reached, the controllable diversion module distributes rainwater to the enhanced treatment unit; when L When the value is less than the second threshold, the controllable diversion module distributes the rainwater to the biological aeration filter system; When the second threshold ≤ L When the value is less than or equal to the first threshold, the controllable diversion module distributes the rainwater to the horizontal flow sedimentation tank.

2. The urban road rainwater runoff classification treatment system according to claim 1, characterized in that: The control center includes an adaptive weighting model that dynamically adjusts the weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters based on a Bayesian inference framework. The dynamic adjustment process includes: Set initial weight coefficients for real-time rain monitoring, rainfall intensity, and historical biotoxicity parameters; According to the current rainfall intensity R t Rainfall intensity compared with similar historical rainfall events R h The absolute difference Δ R and the standard deviation of historical rainfall intensity σ Calculating the matching factor ; Using the formula Correct the initial weight coefficient of the historical biological toxicity parameters, where is the weight coefficient of the historical biological toxicity parameter at the current moment, is the weight coefficient of the historical biological toxicity parameter at the previous moment, α is the preset correction factor for the historical biological toxicity parameter, | α ∣≤0.3, when Δ R ≤ σ hour α >0, when ΔR> σ hour α <0; The weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters were scaled so that their sum was 1.

3. The urban road rainwater runoff classification treatment system according to claim 2, characterized in that: Also included is a collaborative control component, which includes: The heavy metal concentration monitoring module is installed at the outlet of the electrochemical heavy metal degradation treatment unit to detect the residual heavy metal concentration in real time; The parameter collaborative controller is connected to the heavy metal concentration monitoring module and the reagent dosing system in the chemical flocculation treatment unit. The parameter collaborative controller receives the residual heavy metal concentration data and implements a hierarchical control strategy: When the residual heavy metal concentration is ≤ the safety threshold, the chemical dosing system maintains the dosage of chemical flocculants; When the residual heavy metal concentration is greater than the safety threshold, the parameter collaborative controller performs the following operations: Calculate the concentration value Δ at which the residual heavy metal concentration exceeds the safety threshold C , according to Δ C In the preset concentration range, select the corresponding dosage increase ratio m %, where the preset concentration range is m The corresponding relationship of % is: i). If 0μg / L<ΔC≤5μg / L, then m %=10%; ii). If 5μg / L<ΔC≤10μg / L, then m %=20%; iii). If 10μg / L<ΔC≤20μg / L, then m %=40%; iv). If ΔC>20μg / L, then m %=60%; Increase the amount of flocculant to 1+ of the original amount m % times.

4. The urban road rainwater runoff classification treatment system according to claim 3, characterized in that: The parameter collaborative controller is also connected to the adaptive weight model. Before executing the hierarchical control strategy, the parameter collaborative controller corrects the flocculant dosage based on the instantaneous rate of pollutant load. The adaptive weight model is configured to perform: Instantaneous rate of pollutant load based on real-time calculation L t , generates instantaneous rate predictions of pollutant loads through built-in time series prediction algorithms : Will L t and history N The instantaneous rate data of pollutant loads at consecutive time steps constitute a time series; Use the time series prediction algorithm to make a single-step prediction of the sequence and output the predicted value at the next acquisition moment , the time series prediction algorithm is any one of the sliding average model, autoregressive integrated sliding average model or long short-term memory network; Through the control center t +1 moment to calculate the actual pollutant load instantaneous rate monitoring value L t+1 : Comparison of predicted values Compared with the actual monitoring value L t+1 The mean absolute error ε, μ is used to calculate the prediction confidence coefficient μ=e -λε , λ is the error sensitivity coefficient, ranging from 0.5 to 1.0; The instantaneous rate prediction of pollutant load is multiplied by k The benchmark lead compensation is calculated, where: k It is the preset compensation coefficient, ranging from 0.1 to 0.3; Multiply the reference lead compensation by the prediction credibility coefficient to generate the effective lead compensation ΔE; According to ΔE, increase the flocculant dosage to 1+ΔE times of the original dosage.

5. The urban road rainwater runoff classification treatment system according to claim 2, characterized in that: The effluent from the biological aerated filter system is treated by ultrafiltration membrane and then enters the purified water storage tank; The urban road rainwater runoff classification treatment system further includes: The global optimization module is connected to the control center, purified water storage tank, sewage treatment plant, and rainfall forecast information, and is configured as follows: Real-time access to the current water level of the purified water storage tank, the real-time excess treatment capacity of the sewage treatment plant, and rainfall intensity; Dynamically adjust the first and second thresholds based on the above parameters: When the real-time surplus processing capacity of the sewage treatment plant is higher than the preset capacity, a first threshold lowering instruction is generated, and the first threshold is multiplied by a lowering coefficient, where 0.7≤lowering coefficient<1.0; When the rainfall intensity exceeds the rainstorm threshold and the current water level of the purified water storage tank reaches or exceeds the water level warning value, a second threshold increase instruction is generated, and the second threshold is multiplied by the increase coefficient, where 1.0<increase coefficient≤1.

5.

6. The urban road rainwater runoff classification treatment system according to claim 5, characterized in that: Also includes: The performance monitoring module obtains the electrochemical degradation electrode loss rate and ultrafiltration membrane transmembrane pressure difference of the enhanced treatment unit in real time; The state constraint layer connects the performance monitoring module and the global optimization module and is configured to perform: receiving a threshold adjustment instruction generated by a global optimization module; If a first threshold lowering instruction is received and the current electrode loss rate is greater than the upper limit of the loss rate, the lowering instruction is discarded and an electrode loss warning is triggered; Otherwise, the downward adjustment instruction is allowed to be executed; If a second threshold increase instruction is received and the real-time ultrafiltration membrane transmembrane pressure difference is greater than the pressure critical value, the increase range is limited so that the adjusted second threshold does not exceed the initial threshold; otherwise, the instruction is allowed to be executed according to the increase coefficient.

7. The urban road rainwater runoff classification treatment system according to claim 1, characterized in that: Also includes: Redundant traffic protection components, including: Fault diagnosis module, real-time monitoring of the operating status of the controllable shunt module; Bypass switching unit, connecting the inlet of the horizontal flow sedimentation tank and the end of the rainwater collection network; Among them, when the fault diagnosis module detects that the controllable diversion module fails, it automatically activates the bypass switching unit to force all rainwater runoff into the horizontal flow sedimentation tank, and at the same time sends an equipment fault alarm to the control center.

8. The urban road rainwater runoff classification treatment system according to claim 2, characterized in that: The adaptive weight model sets up a burst pollution identification layer, which performs: Real-time calculation of the sudden change gradient of turbidity, COD, TOC and heavy metal concentration of rainwater runoff output by multi-source spectral water quality monitor; When any of the above mutation gradients exceeds the maximum fluctuation threshold in the historical biological toxicity correlation parameters, a spectral anomaly index is generated, and the maximum fluctuation threshold is based on the 95th percentile value of historical data statistics; The dynamic weight allocation layer responds to the spectral anomaly index and forces the weights of the pollutant load instantaneous rate calculation to be assigned to the real-time rainwater monitoring data, while ignoring the rainfall forecast information and historical biotoxicity correlation parameters; At this time, if the instantaneous rate of pollutant load exceeds 80% of the first threshold, the control center will determine that L >First threshold.

9. The urban road rainwater runoff classification treatment system according to claim 5, characterized in that: The global optimization module sets up a depth processing allocation module, which performs: Real-time acquisition of the water level of the purified water storage tank and the transmembrane pressure difference of the ultrafiltration membrane; Construct the water level-membrane pressure synergistic constraint function: When the water level in the regulating reservoir is greater than the warning water level and the transmembrane pressure difference of the ultrafiltration membrane is greater than 90% of the critical value, the control center dynamically increases the second threshold; The controllable diversion module distributes rainwater to the aerated biological filter system according to the formula Q = β× ( S max - S 当前 ) for current limiting, where Q To allow for flow diversion, β is the attenuation coefficient, S max is the maximum safe capacity of the storage tank. S 当前 is the real-time water level; Excess rainwater is switched to the horizontal flow sedimentation tank through a controllable diversion module.

10. The processing method based on the urban road rainwater runoff classification processing system according to claim 5 is characterized in that: The following steps are involved: S1. Real-time monitoring and data acquisition: A multi-source spectral water quality monitor installed at the end of the rainwater collection network monitors the turbidity, COD, TOC, and heavy metal concentrations of rainwater runoff in real time. Synchronously access real-time rainfall intensity in rainfall forecast information; S2. Dynamic calculation of pollutant load instantaneous rate L: Standardize real-time rain monitoring data, rainfall intensity, and historical biotoxicity correlation parameters and remove outliers; Dynamically adjust weight coefficients based on the Bayesian inference framework: Set initial weights for real-time monitoring data, rainfall intensity, and historical biotoxicity parameters; According to the current rainfall intensity R t Rainfall intensity compared with similar historical rainfall events R h The absolute difference Δ R and the standard deviation of historical rainfall intensity σ Calculating the matching factor ; Using the formula Correct the initial weight coefficient of the historical biological toxicity parameters, where is the weight coefficient of the historical biological toxicity parameter at the current moment, is the weight coefficient of the historical biological toxicity parameter at the previous moment, α is the preset correction factor for the historical biological toxicity parameter, | α ∣≤0.3, when Δ R ≤ σ hour α >0, when ΔR> σ hour α <0; The weight coefficients of rainwater monitoring, rainfall intensity, and historical biotoxicity parameters were scaled so that the sum was 1; The instantaneous rate of pollutant load L is calculated using the weighted summation formula: ,in: w i is the dynamically adjusted weight coefficient of the i-th data source, f i ( x i ) is the standardized data mapping function of the i-th data source, and n is the total number of data sources; S3, hierarchical diversion control: When L>the first threshold, the rainwater is distributed to the enhanced treatment unit, which sequentially undergoes cyclone sand settling treatment, electrochemical heavy metal degradation treatment, and chemical flocculation treatment; When L < the second threshold, the rainwater is distributed to the biological aeration filter system for treatment. The effluent of the biological aeration filter system is treated by ultrafiltration membrane and then input into the purified water storage tank. When the second threshold ≤ L ≤ the first threshold, the rainwater is allocated to the advection sedimentation tank for treatment; S4. Dynamic optimization threshold: Real-time access to the current water level of the purified water storage tank, the real-time excess treatment capacity of the sewage treatment plant, and rainfall intensity; The first and second thresholds are dynamically adjusted based on the following conditions: When the real-time surplus processing capacity of the sewage treatment plant is higher than the preset capacity, a first threshold lowering instruction is generated: the first threshold is multiplied by a lowering coefficient, where 0.7≤lowering coefficient<1.0; When the rainfall intensity exceeds the rainstorm threshold and the current water level of the purified water storage tank reaches or exceeds the water level warning value, a second threshold increase instruction is generated: the second threshold is multiplied by the increase coefficient, where 1.0 < increase coefficient ≤ 1.5; The updated first threshold and second threshold are applied in real time to the hierarchical diversion control in step S3.