A method and system for advanced purification of wastewater plant effluent
Patent Information
- Application Number
- CN202610817162.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-08
- Publication Date
- 2026-09-15
Smart Images
Figure CN122748824A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of environmental remediation and water treatment technology, and in particular to a method and system for deep purification of wastewater treatment plant effluent. Background Technology
[0002] Traditional urban wastewater treatment plants, whose core objective is to achieve compliant discharge, typically produce effluent (tailwater) that, even if meeting the Class A standard of the "Discharge Standard of Pollutants for Urban Wastewater Treatment Plants," still contains certain amounts of inorganic nitrogen, phosphorus, and trace amounts of new pollutants such as pharmaceuticals and personal care products. Directly discharging this effluent into receiving water bodies such as natural lakes via pumping stations still poses risks such as eutrophication and disruption of the ecological balance. Constructed wetlands, as an ecological treatment method, are widely used for the advanced treatment of wastewater effluent and serve as a buffer zone between wastewater treatment plants and natural water bodies. However, existing technologies still have the following limitations:
[0003] 1. Limited Function: Most constructed wetlands for wastewater treatment only serve a purification and buffering function, and their comprehensive social and environmental benefits, such as public education, ecological landscape, and biodiversity enhancement, are not adequately developed.
[0004] 2. Low resource utilization rate: The reuse path of effluent from traditional sewage treatment plants is singular, failing to form a closed-loop resource utilization model including urban miscellaneous uses, ecological water replenishment, and landscape water use;
[0005] 3. Low land use efficiency: Traditional constructed wetland designs require a large area, making them difficult to promote and apply in urban sewage treatment plant areas where land resources are scarce;
[0006] 4. System disconnect: Wastewater treatment plants and constructed wetlands are designed and operated independently, lacking coordinated optimization, resulting in high overall energy consumption and weak resistance to shock loads.
[0007] Therefore, in order to solve the problems of existing technologies, it is necessary to propose a comprehensive solution that can deeply integrate wastewater treatment processes with ecological water purification processes, and at the same time utilize intelligent control systems based on real-time monitoring to achieve cross-unit resource utilization and dynamic optimization of pollution reduction across "plant-station-wetland", forming a comprehensive solution that integrates water resource recycling and environmental service functions. Summary of the Invention
[0008] The technical problem to be solved by the present invention is to provide a method and system for deep purification of wastewater treatment plant effluent, addressing the shortcomings of the prior art.
[0009] The technical solution of this invention to solve the above-mentioned technical problems is as follows: A low-carbon ecological wastewater deep purification system, the system comprising a wastewater treatment plant unit, a hydraulic control unit, an artificial wetland composite unit, and a central intelligent control platform connected to the aforementioned units, wherein:
[0010] The effluent outlet of the wastewater treatment plant unit is connected to the inlet of the hydraulic control unit. The effluent from the wastewater treatment plant unit is buffered through the built-in regulating tank via the hydraulic control unit.
[0011] The outlet of the hydraulic control unit is connected to the inlet of the constructed wetland composite unit, and the constructed wetland composite unit performs deep purification of the wastewater treatment plant effluent based on the synergistic mechanism of plant-microorganism-filler.
[0012] The central intelligent control platform is used to predict the risk of water quality exceeding standards based on the water volume data fed back by the hydraulic control unit and the water quality data fed back by the artificial wetland composite unit, and to adjust the inflow rate of the hydraulic control unit and control the amount of water purification chemicals based on the prediction results.
[0013] The central intelligent control platform is also used to map the wetland effluent prediction deviation into a predictive carbon source addition adjustment amount based on a nonlinear scaling function, and to allocate carbon sources in the biochemical treatment section of the wastewater treatment plant unit by integrating the carbon source addition adjustment amount benchmark value, the influent water quality monitoring feedback adjustment amount, and the predictive carbon source addition adjustment amount.
[0014] Furthermore, the constructed wetland composite unit includes a vertical flow constructed wetland and an ecological stabilization pond. The vertical flow constructed wetland and the ecological stabilization pond achieve step-by-step purification of wastewater through a composite series structure. The effluent from the vertical flow constructed wetland is introduced into the ecological stabilization pond via a bottom water collection and distribution system, forming a series process of "preliminary purification - deep stabilization." The vertical flow constructed wetland consists of wetland plants, multi-layered packing materials, and a bottom water collection and distribution system, used for preliminary purification of wastewater treatment plant effluent. The ecological stabilization pond contains various plants and aquatic animals to stabilize water quality and remove trace pollutants from the water through plant absorption and microbial synergistic degradation mechanisms, ultimately achieving effluent reuse.
[0015] Furthermore, a recycled water pumping station is also installed at the end of the ecological stabilization pond. The recycled water pumping station is used to lift the final effluent to an ecological water replenishment system for replenishing urban rivers or landscape water bodies, as well as urban miscellaneous water uses.
[0016] Furthermore, the central intelligent control platform is also used to determine a preset adjustment benchmark value based on the following formula:
[0017] ;
[0018] Among them, Q in TN indicates the influent flow rate of the plant area. in Indicates total nitrogen (TN) in the influent. targetThe target total nitrogen in the effluent is represented by k1, which represents the preset theoretical carbon-to-nitrogen ratio for denitrification, and ΔCOD. internal C represents the amount of carbon source that denitrifying bacteria can utilize in the wastewater. source This indicates the COD equivalent concentration of the added carbon source.
[0019] Furthermore, the central intelligent control platform is also used to determine the influent water quality monitoring feedback adjustment amount by using a fuzzy PID controller to dynamically adjust the PID parameters by taking the real-time monitoring deviations of ammonia nitrogen, nitrate nitrogen and COD as fuzzy inputs and combining them with fuzzy rules, based on the real-time monitoring values of ammonia nitrogen, nitrate nitrogen and COD.
[0020] Furthermore, the central intelligent control platform is also used to predict wetland outflow based on a stacked LSTM long short-term memory network, wherein: the LSTM long short-term memory network takes as input the carbon-nitrogen ratio at the wetland inlet, the chemical oxygen demand at the wetland inlet, total nitrogen and total phosphorus, wetland water temperature, wetland water level, inflow rate, dissolved oxygen, pH, and seasonal encoding used to capture the nonlinear effect of temperature on microbial activity; the LSTM long short-term memory network includes an input layer, an LSTM layer 1 containing 64 hidden units and capable of returning the complete sequence, a Dropout layer for preventing overfitting, an LSTM layer 2 containing 32 hidden units and only returning the state at the last time step, a fully connected layer 1 containing 16 neurons and using the ReLU activation function, and a fully connected layer 2 containing 1 neuron and using the linear activation function.
[0021] Furthermore, the LSTM long short-term memory network determines the training loss function for wetland water discharge prediction based on a weighted sum of the mean square error of MSE, the mean absolute error of MAE, and the first-order difference loss that reflects the difference between the rate of change of the predicted value and the true value.
[0022] Furthermore, the wetland effluent prediction bias is mapped to a predictive carbon source addition adjustment amount based on the following nonlinear scaling function:
[0023] ΔQ ff,wetland (t) = K ff,eff (t) × Η( ) × Q base (t);
[0024] in, τ(t) represents the prediction bias of wetland outflow. × ζ 水温 (t)×ζ 水位 (t) represents the non-dynamic threshold fluctuation of adaptive temperature and water level changes in wetlands. ζ represents the basic threshold for the non-dynamic zone of wetlands. 水温 (t) represents the water temperature correction factor, ζ 水位(t) represents the water level correction factor, K ff, eff (t)=K ff, base ×(1+η×σ 常态 (t) represents the uncertainty compensation gain, K ff, base η represents the basic predictive feedback gain coefficient, η represents the uncertainty compensation coefficient, and σ represents the uncertainty compensation coefficient. 常态 (t) represents the standard deviation under normal operating conditions, Q base (t) represents the basic carbon source input acceleration rate.
[0025] Furthermore, to automatically optimize the uncertainty compensation gain, the following penalty function for exceeding the effluent standard is set:
[0026] Penalty 超标 = ;
[0027] in, Indicates the predicted value of wetland outflow. Indicates wetland effluent discharge standards, This indicates that the wetland's outflow exceeded the warning line. This indicates the preset penalty coefficient.
[0028] The beneficial effects of this invention are as follows: by coupling the sewage treatment plant unit, the hydraulic control unit, and the constructed wetland composite unit system, and utilizing the water volume regulation and kinetic energy enhancement effect of the hydraulic control unit, combined with the synergistic effect of physical filtration, chemical adsorption, and biodegradation of the constructed wetland composite unit, the constructed wetland becomes an ecological buffer zone for sewage treatment plant effluent discharged into urban natural rivers and lakes, achieving deep purification and stabilization of sewage treatment plant effluent, while helping to restore and enhance the ecological function of the receiving water body. Under the premise of not significantly increasing investment and operating costs, it increases the function of deep sewage treatment and has significant ecological benefits. It is suitable for the ecological regeneration of urban sewage treatment plant effluent and the buffer protection of natural water bodies when effluent replenishment and reuse are required. Attached Figure Description
[0029] Figure 1 This is a schematic diagram of the structure of a low-carbon ecological wastewater deep purification system disclosed in this application;
[0030] Figure 2 This is a schematic diagram illustrating the effects of vertical flow constructed wetlands and ecological stabilization ponds within a constructed wetland complex. Detailed Implementation
[0031] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.
[0032] like Figure 1As shown, this application discloses a low-carbon ecological wastewater deep purification system, which includes a wastewater treatment plant unit, a hydraulic control unit (i.e., a pumping station), an artificial wetland composite unit, and a central intelligent control platform connected to the aforementioned units, wherein:
[0033] The effluent outlet of the wastewater treatment plant unit is connected to the inlet of the hydraulic control unit. The hydraulic control unit then buffers fluctuations in the effluent flow from the wastewater treatment plant unit through a built-in regulating tank.
[0034] Specifically, the regulating tank mentioned here mainly serves to homogenize the flow rate, transforming the unstable pulsed effluent from the wastewater treatment plant unit into a constant flow rate, thereby ensuring that the subsequent constructed wetland unit receives a stable hydraulic load. The purpose of buffering water flow fluctuations is to prevent hydraulic shocks from disrupting the stable operation of the subsequent ecological treatment system, and to avoid insufficient hydraulic retention time or uncontrolled carbon source addition due to sudden changes in flow rate, thus ensuring that the effluent quality meets the standards.
[0035] The outlet of the hydraulic control unit is connected to the inlet of the constructed wetland composite unit, and the constructed wetland composite unit performs deep purification of the wastewater treatment plant effluent based on the synergistic mechanism of plants, microorganisms and fillers.
[0036] For details, please refer to Figure 2 This application utilizes a vertical flow constructed wetland. Through the adsorption of sediment and the denitrification by functional bacteria within the wetland, reactive nitrogen and trace organic pollutants such as antibiotics in the wastewater treatment plant effluent are further removed, reducing the risk of eutrophication and resistance, and maintaining the ecological health of the water body. Simultaneously, the abundant submerged and emergent plants can remove emerging pollutants such as microplastics from the water. If the construction area allows, a surface flow constructed wetland module can be added, along with automatic water quality and quantity monitoring facilities installed in the wetland water body to monitor the influent and effluent water quality parameters and quantity in real time. The data is then fed back to a central intelligent control platform to adjust operating parameters such as carbon source dosage and hydraulic control unit valve opening, ensuring that the effluent water quality consistently meets standards.
[0037] Specifically, this application also employs an ecological stabilization pond, which contains a variety of plants and aquatic animals (such as shellfish and fish) to form a micro-ecosystem, further stabilizing water quality and removing trace pollutants.
[0038] The central intelligent control platform is used to predict the risk of water quality exceeding standards based on the water volume data fed back by the hydraulic control unit and the water quality data fed back by the constructed wetland composite unit, and to adjust the influent flow rate of the hydraulic control unit and control the dosage of water purification chemicals based on the prediction results.
[0039] Specifically, this application will construct a dynamic early warning model based on historical operating data. When the model predicts that the effluent quality is close to the threshold for exceeding the standard, it will automatically control the frequency of the carbon source dosing pump and increase the amount of easily degradable carbon source added to enhance denitrification. At the same time, this application will also use newly added operating data to incrementally learn the model and dynamically correct the prediction parameters to adapt to slowly changing operating conditions such as plant growth and packing blockage in the effluent constructed wetland, ensuring that the effluent quality consistently meets the standards.
[0040] The central intelligent control platform is also used to map the wetland effluent prediction deviation into a predictive carbon source addition adjustment amount based on a nonlinear scaling function, and to allocate carbon sources in the biochemical treatment section of the wastewater treatment plant unit by integrating the carbon source addition adjustment amount benchmark value, the influent water quality monitoring feedback adjustment amount, and the predictive carbon source addition adjustment amount.
[0041] As can be seen from the above, the low-carbon ecological wastewater deep purification system disclosed in this application couples a wastewater treatment plant unit, a hydraulic control unit, and an artificial wetland composite unit system. By utilizing the water volume regulation and kinetic energy enhancement effect of the hydraulic control unit, combined with the synergistic effect of physical filtration, chemical adsorption, and biodegradation of the artificial wetland composite unit, the artificial wetland becomes an ecological buffer zone for wastewater treatment plant effluent discharged into urban natural rivers and lakes. This achieves deep purification and stabilization of wastewater treatment plant effluent, while also helping to restore and enhance the ecological function of the receiving water body. Under the premise of not significantly increasing investment and operating costs, it adds the function of deep wastewater treatment and has significant ecological benefits. It is suitable for the ecological regeneration of urban wastewater treatment plant effluent and the buffer protection of natural water bodies when effluent replenishment and reuse are required.
[0042] In one embodiment, the constructed wetland complex unit includes a vertical flow constructed wetland and an ecological stabilization pond. The vertical flow constructed wetland and the ecological stabilization pond achieve step-by-step purification of wastewater through a composite series structure. The effluent from the vertical flow constructed wetland is introduced into the ecological stabilization pond via a bottom water collection and distribution system, forming a series process of "preliminary purification - deep stabilization." The vertical flow constructed wetland consists of wetland plants, multi-layered packing materials, and a bottom water collection and distribution system, used for preliminary purification of wastewater treatment plant effluent. The ecological stabilization pond contains various plants and aquatic animals to stabilize water quality and remove trace pollutants from the water through plant absorption and microbial synergistic degradation mechanisms, ultimately achieving effluent reuse.
[0043] For details, please refer to Figure 2The vertical flow constructed wetland adopts a composite structure of "multi-layer packing material + plant root system + water collection and distribution system" to achieve preliminary purification. In this structure, sewage seeps down evenly from the top, undergoes physical filtration through the packing material layer and biofilm adhesion, and is then collected through the bottom water collection and distribution system to avoid short-circuiting. The roots of wetland plants penetrate deep into the packing material layer and secrete oxygen to create an aerobic-anaerobic microenvironment. Nitrifying bacteria in the aerobic zone convert ammonia nitrogen into nitrate, while denitrifying bacteria in the anaerobic zone reduce nitrate into nitrogen. At the same time, the plants directly absorb nutrients such as nitrogen and phosphorus, achieving preliminary removal of TN and TP.
[0044] For details, please refer to Figure 2 The ecological stabilization pond achieves deep stabilization and reuse through a "plant-animal-microorganism" ecosystem. Floating-leaved plants (such as water lilies), submerged plants (such as Vallisneria natans), and emergent plants (such as cattails) are planted within the pond. Floating-leaved plants release oxygen through photosynthesis, maintaining dissolved oxygen levels and promoting the degradation of residual organic matter by aerobic microorganisms. The roots of submerged plants adsorb phosphorus, and emergent plants further absorb nitrogen. Aquatic animals in the pond (such as filter-feeding shellfish and zooplankton) consume suspended solids and algae to reduce water turbidity. Microorganisms (such as heterotrophic bacteria and algae) decompose recalcitrant organic matter, while the algae-bacteria symbiotic system adsorbs phosphorus, further reducing total phosphorus. Finally, the deeply purified effluent can be reused for landscape irrigation and greening, achieving resource recycling.
[0045] In one embodiment, a recycled water pumping station is also provided at the end of the ecological stabilization pond. The recycled water pumping station is used to lift the final effluent to an ecological water replenishment system for replenishing urban rivers or landscape water bodies, as well as urban miscellaneous water uses.
[0046] In one embodiment, this application also effectively collects rainwater within the constructed wetland composite unit through a corresponding rainwater pipe network and introduces it into the constructed wetland treatment process, incorporating it into automatic flow monitoring. At the same time, water quality monitoring ensures that the constructed wetland treatment stage is relatively stable, realizing the resource utilization of rainwater from collection to recycling.
[0047] In one embodiment, this application also designs and constructs the artificial wetland complex as a wetland park, connecting the flora and fauna of the ecosystem through facilities such as boardwalks, viewing platforms, etc. Furthermore, this application establishes a water purification science education area within the unit, and dynamically displays the entire process of "plant-station-wetland" units from "sewage" to "clean water" and then to "ecological water use" through a data intelligent control system and digital twin model, combining public environmental education and recreational functions.
[0048] In one embodiment, the central intelligent control platform is further configured to determine a preset adjustment benchmark value based on the following formula:
[0049] ;
[0050] Among them, Q in TN indicates the influent flow rate of the plant area. in Indicates total nitrogen (TN) in the influent. target The target total nitrogen in the effluent is represented by k1, which represents the preset theoretical carbon-to-nitrogen ratio for denitrification, and ΔCOD. internal C represents the amount of carbon source that denitrifying bacteria can utilize in the wastewater. source This indicates the COD equivalent concentration of the added carbon source.
[0051] Specifically, This represents the difference between the total nitrogen in the wastewater influent and the target total nitrogen in the effluent per unit time, multiplied by the influent flow rate. This represents the total nitrogen that needs to be removed through denitrification. Multiplying this total nitrogen by the preset theoretical carbon-to-nitrogen ratio coefficient k1 for denitrification yields the theoretically required total carbon source for the denitrification process. Subtracting the amount of carbon source ΔCOD that can be utilized by denitrifying bacteria from this theoretical total carbon source yields the final amount. internal This yields the amount of carbon source that needs to be replenished from the external source. Finally, the amount of carbon source that needs to be replenished from the external source is divided by the COD equivalent concentration C of the added carbon source. source The required external carbon source flow rate per unit time is obtained, which is the preset adjustment baseline value. In summary, this application calculates the required external carbon source replenishment amount by using the logic of "total nitrogen to be removed × theoretical carbon-nitrogen ratio coefficient - amount of endogenous available carbon source", and then divides it by the carbon source concentration to obtain the addition flow rate. This allows for precise control of the external carbon source addition, ensuring denitrification efficiency, avoiding excessive addition costs, and ensuring stable effluent total nitrogen compliance.
[0052] In one embodiment, the central intelligent control platform is also used to determine the influent water quality monitoring feedback adjustment amount by using a fuzzy PID controller to dynamically adjust the PID parameters by taking the real-time monitoring deviations of ammonia nitrogen, nitrate nitrogen and COD as fuzzy inputs and combining them with fuzzy rules, based on the real-time monitoring values of ammonia nitrogen, nitrate nitrogen and COD.
[0053] Specifically, the central intelligent control platform first collects real-time monitoring values of influent ammonia nitrogen, nitrate nitrogen, and COD, and calculates their deviations from the set values. Then, this deviation is quantified into fuzzy linguistic variables (such as "negative large," "negative medium," "zero," "positive small," etc.) as input to the fuzzy controller. Next, a fuzzy rule base is constructed based on wastewater treatment process experience. Fuzzy inference maps the input deviations to PID parameter adjustments, and defuzzification yields precise parameter adjustment values, updating the PID parameters. Finally, the adjusted PID parameters are used to calculate the influent water quality monitoring feedback adjustment amount. This adjustment amount is used to dynamically adjust the influent flow rate or carbon source addition strategy to stabilize the influent COD of the wastewater treatment plant's biological treatment section and constructed wetland unit within the set range.
[0054] In one embodiment, the central intelligent control platform is also used to predict wetland outflow based on a stacked LSTM long short-term memory network, wherein: the LSTM long short-term memory network takes as input the carbon-nitrogen ratio at the wetland inlet, the chemical oxygen demand at the wetland inlet, total nitrogen and total phosphorus, wetland water temperature, wetland water level, inflow rate, dissolved oxygen, pH, and seasonal encoding used to capture the nonlinear effect of temperature on microbial activity; the LSTM long short-term memory network includes an input layer, an LSTM layer 1 containing 64 hidden units and capable of returning a complete sequence, a Dropout layer for preventing overfitting, an LSTM layer 2 containing 32 hidden units and only returning the state at the last time step, a fully connected layer 1 containing 16 neurons and using the ReLU activation function, and a fully connected layer 2 containing 1 neuron and using a linear activation function.
[0055] Specifically, LSTM layer 1 receives the input and processes the sequence data step-by-step through a gating mechanism of 64 hidden units (forget gate, input gate, output gate), returning the complete sequence (preserving the hidden state at each time step). The Dropout layer randomly discards some neuron outputs with a probability of 0.2 to prevent overfitting. LSTM layer 2 receives the output of the Dropout layer and further extracts long-term dependency features through 32 hidden units, returning only the hidden state at the last time step (capturing key information at the end of the sequence). Fully connected layer 1 maps the output of LSTM layer 2 to a 16-dimensional space and introduces non-linearity through the ReLU activation function to enhance the model's ability to fit complex relationships. Fully connected layer 2 compresses the 16-dimensional features into 1-dimensional space and uses a linear activation function to output the predicted value, thus achieving an end-to-end mapping from multi-dimensional features to the target predicted value.
[0056] In one embodiment, during training, the application uses the Adam optimizer to adjust the weights, trains the model using historical data, and ultimately predicts the wetland effluent quality in real time, providing a forward-looking adjustment basis for the intelligent control platform.
[0057] In one embodiment, the LSTM (Long Short-Term Memory) network determines a training loss function for wetland outflow prediction based on a weighted sum of the mean squared error of MSE (mean squared error), the mean absolute error of MAE (mean absolute error), and a first-order difference loss that reflects the difference between the predicted and actual values.
[0058] Specifically, the training loss function is a weighted sum of the mean squared error (MSE), the mean absolute error (MAE), and the first-order difference loss, expressed by the following formula: ,in, , Let i be the actual wetland effluent value of the i-th sample. For the corresponding predicted value, N is the total number of samples; it should be noted that the mean square error of MSE amplifies the prediction error by squaring, making it more sensitive to large deviations. Here, the mean absolute error (MAE) is not sensitive to outliers and can more robustly reflect the average deviation between the predicted and the actual values. ,in, It is the first difference of the true value (reflecting the rate of change of the true value). The first difference of the predicted value, By ensuring consistency between the trend of predicted values and actual values, excessive fluctuations in predicted values or incorrect trends can be avoided. The preset weighting coefficients satisfy... The loss function can be adjusted according to actual needs to balance the impact of different losses. Through multi-dimensional constraints (absolute error, squared error, and trend), the LSTM model can simultaneously optimize prediction accuracy, robustness, and trend consistency during training, thereby improving the reliability of wetland outflow prediction.
[0059] In one embodiment, wetland effluent prediction bias is mapped to predictive carbon source addition adjustment based on the following nonlinear scaling function:
[0060] ΔQ ff,wetland (t) = K ff,eff (t) × Η( ) × Q base (t);
[0061] in, τ(t) represents the prediction bias of wetland outflow. × ζ 水温 (t)×ζ 水位 (t) represents the non-dynamic threshold fluctuation of adaptive temperature and water level changes in wetlands. ζ represents the basic threshold for the non-dynamic zone of wetlands. 水温 (t) represents the water temperature correction factor, ζ 水位 (t) represents the water level correction factor, K ff, eff (t)=K ff, base ×(1+η×σ 常态 (t) represents the uncertainty compensation gain, K ff, base η represents the basic predictive feedback gain coefficient, η represents the uncertainty compensation coefficient, and σ represents the uncertainty compensation coefficient. 常态 (t) represents the standard deviation under normal operating conditions, Q base (t) represents the basic carbon source input acceleration rate.
[0062] In one embodiment, to automatically optimize the uncertainty compensation gain, the following penalty function for exceeding the effluent standard is set:
[0063] Penalty 超标 = ;
[0064] in, Indicates the predicted value of wetland outflow. Indicates wetland effluent discharge standards, This indicates that the wetland's outflow exceeded the warning line. This indicates the preset penalty coefficient.
[0065] Specifically, this application uses the aforementioned effluent exceeding penalty function as the objective function for iterative optimization calculation until the penalty function value is minimized or a preset convergence condition is met. The core logic of this piecewise penalty function is: no penalty for compliance, linear penalty for slight exceedances, and quadratic heavy penalty for severe exceedances. When determined in the first segment (compliant zone), i.e., when the predicted total nitrogen in the effluent is less than or equal to the discharge standard, the penalty value is set to 0. This signifies that as long as the water quality meets the standard, the system considers this a perfect operating state with no cost. When determined in the second segment (slight exceedance zone), i.e., when the total nitrogen in the effluent exceeds the discharge standard but is still within the warning line, this application uses a linear penalty method, setting the penalty value to increase linearly with the degree of exceedance. This allows for slight fluctuations within a certain range, preventing the optimization algorithm from overreacting. When determined in the third segment (severe exceedance zone), i.e., when the current total nitrogen in the effluent exceeds the warning line, this application uses a non-linear heavy penalty method. Once the warning line is exceeded, the system imposes a very large penalty, forcing the optimization algorithm to quickly adjust its parameters to avoid this situation.
[0066] This allows the function to be adaptively adjusted according to the fluctuations in operating conditions (such as seasonal changes and hydraulic load) during actual operation, and the carbon source addition gain can be appropriately increased when the prediction uncertainty increases.
[0067] In one embodiment, to strengthen the zero-tolerance policy for water quality exceeding standards, the aforementioned penalty function can be further upgraded to an exponential form for reinforcement learning scenarios, as optimized as follows:
[0068] Penalty 超标 = ;
[0069] in, The parameter representing the overscaling threshold (e.g., a maximum of 1.0 mg / L) allows the reinforcement learning function to provide significant exponential penalty feedback even for slight overscaling, with the penalty increasing faster as the overscaling increases. This helps prevent the agent from becoming immune to overscaling.
[0070] Finally, the uncertainty compensation gain K ff, eff (t) and the above over-limit penalty function Penalty 超标Through reinforcement learning, a closed-loop optimization relationship is ultimately formed between them, where K ff, eff The magnitude of (t) directly affects the amount of carbon source compensation, thereby changing the total nitrogen concentration in the constructed wetland effluent, and ultimately determining the value of the penalty term for exceeding the standard; while the reinforcement learning algorithm maximizes the reward function, the calculation form of which can be found in: Dynamically adjust K ff, eff (t) is brought to the optimal value so that carbon source cost and energy consumption are minimized in a synergistic manner while meeting the effluent standards.
[0071] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A low-carbon, ecological wastewater deep purification system, characterized in that, The system includes a wastewater treatment plant unit, a hydraulic control unit, an constructed wetland composite unit, and a central intelligent control platform connected to the aforementioned units, wherein: The effluent outlet of the wastewater treatment plant unit is connected to the inlet of the hydraulic control unit. The effluent from the wastewater treatment plant unit is buffered through the built-in regulating tank via the hydraulic control unit. The outlet of the hydraulic control unit is connected to the inlet of the constructed wetland composite unit, and the constructed wetland composite unit performs deep purification of the wastewater treatment plant effluent based on the synergistic mechanism of plant-microorganism-filler. The central intelligent control platform is used to predict the risk of water quality exceeding standards based on the water volume data fed back by the hydraulic control unit and the water quality data fed back by the artificial wetland composite unit, and to adjust the inflow rate of the hydraulic control unit and control the amount of water purification chemicals based on the prediction results. The central intelligent control platform is also used to map the wetland effluent prediction deviation into a predictive carbon source addition adjustment amount based on a nonlinear scaling function, and to allocate carbon sources in the biochemical treatment section of the wastewater treatment plant unit by integrating the carbon source addition adjustment amount benchmark value, the influent water quality monitoring feedback adjustment amount, and the predictive carbon source addition adjustment amount.
2. The system according to claim 1, characterized in that, The constructed wetland complex unit includes a vertical flow constructed wetland and an ecological stabilization pond, wherein: The vertical flow constructed wetland and the ecological stabilization pond achieve step-by-step purification of wastewater through a composite series structure. The effluent from the vertical flow constructed wetland is introduced into the ecological stabilization pond through a bottom water collection and distribution system, forming a series process of "preliminary purification - deep stabilization". The vertical flow constructed wetland consists of wetland plants, multi-layer packing material, and a bottom water collection and distribution system, and is used for the preliminary purification treatment of wastewater effluent from sewage treatment plants. The ecological stabilization pond contains a variety of plants and aquatic animals to stabilize water quality. Through plant absorption and microbial synergistic degradation, trace pollutants in the water are removed, ultimately enabling the reuse of the effluent.
3. The system according to claim 1, characterized in that, At the end of the ecological stabilization pond, a recycled water pumping station is also installed. The recycled water pumping station is used to lift the final effluent to an ecological water replenishment system for replenishing urban rivers or landscape water bodies, as well as urban miscellaneous water uses.
4. The system according to claim 1, characterized in that, The central intelligent control platform is also used to determine a preset adjustment benchmark value based on the following formula: ； Among them, Q in TN indicates the influent flow rate of the plant area. in Indicates total nitrogen (TN) in the influent. target The target total nitrogen in the effluent is represented by k1, which represents the preset theoretical carbon-to-nitrogen ratio for denitrification, and ΔCOD. internal C represents the amount of carbon source that denitrifying bacteria can utilize in the wastewater. source This indicates the COD equivalent concentration of the added carbon source.
5. The system according to claim 1, characterized in that, The central intelligent control platform is also used to determine the influent water quality monitoring feedback adjustment amount by using a fuzzy PID controller to dynamically adjust the PID parameters based on the real-time monitoring values of ammonia nitrogen, nitrate nitrogen, and COD, using the real-time monitoring deviations of ammonia nitrogen, nitrate nitrogen, and COD as fuzzy inputs and combining them with fuzzy rules.
6. The system according to claim 1, characterized in that, The central intelligent control platform is also used for wetland outflow prediction based on a stacked LSTM long short-term memory network, wherein: The LSTM long short-term memory network takes as input the carbon-to-nitrogen ratio at the wetland inlet, the chemical oxygen demand at the wetland inlet, total nitrogen and total phosphorus, wetland water temperature, wetland water level, inflow rate, dissolved oxygen, pH, and seasonal coding used to capture the nonlinear effect of temperature on microbial activity. The LSTM (Long Short-Term Memory) network includes an input layer, an LSTM layer 1 with 64 hidden units that can return the complete sequence, a Dropout layer to prevent overfitting, an LSTM layer 2 with 32 hidden units that only returns the state at the last time step, a fully connected layer 1 with 16 neurons and using the ReLU activation function, and a fully connected layer 2 with 1 neuron and using the linear activation function.
7. The system according to claim 6, characterized in that, The LSTM (Long Short-Term Memory) network determines the training loss function for wetland water discharge prediction based on a weighted sum of the mean square error (MSE), the mean absolute error (MAE), and the first-order difference loss, which reflects the difference between the predicted and actual values.
8. The system according to claim 1, characterized in that, The wetland effluent prediction bias is mapped to the predictive carbon source addition adjustment amount based on the following nonlinear scaling function: ΔQ ff,wetland (t) = K ff,eff (t) × Η( ) × Q base (t); in, τ(t) represents the prediction bias of wetland outflow. × ζ 水温 (t)×ζ 水位 (t) represents the non-dynamic threshold fluctuation of adaptive temperature and water level changes in wetlands. ζ represents the basic threshold for the non-dynamic zone of wetlands. 水温 (t) represents the water temperature correction factor, ζ 水位 (t) represents the water level correction factor, K ff, eff (t)=K ff, base ×(1+η×σ 常态 (t) represents the uncertainty compensation gain, K ff, base η represents the basic predictive feedback gain coefficient, η represents the uncertainty compensation coefficient, and σ represents the uncertainty compensation coefficient. 常态 (t) represents the standard deviation under normal operating conditions, Q base (t) represents the basic carbon source input acceleration rate.
9. The system according to claim 1, characterized in that, To automatically optimize the uncertainty compensation gain, the following penalty function for exceeding the effluent standard is set: Penalty 超标 = ; in, Indicates the predicted value of wetland outflow. Indicates wetland effluent discharge standards, This indicates that the wetland's outflow exceeded the warning line. This indicates the preset penalty coefficient.