Precise aeration method for sewage treatment plant based on prediction model and multi-objective optimization

Abstract

The invention discloses an accurate aeration method for a sewage treatment plant based on a prediction model and multi-objective optimization, which comprises the following steps: taking an aeration air volume value as an initial sample, and inputting the aeration air volume value into the prediction model to obtain a corresponding effluent index prediction value; based on a multi-objective optimization algorithm, solving and generating an optimal solution set of the aeration air volume value; weight distribution is carried out on the water outlet indexes, and the optimal solution of the aeration air volume value is obtained through a good and inferior solution distance method; according to the method, multiple performance indexes in the aeration regulation and control process are considered and optimized, the overall process operation effect and the system stability are remarkably improved, subjective preference and objective data factors are comprehensively considered through a multi-index decision-making method, final decision-making of aeration control is achieved in an auxiliary mode, and accordingly accurate aeration is achieved. According to the invention, the operation mode that the effluent reaches the standard and is discharged through excessive aeration in a traditional sewage treatment plant is overturned, the operation cost is greatly reduced, and the method has a good application prospect.

Description

Technical Field

[0001] This invention relates to the field of chemical process control technology, specifically to a precise aeration method for wastewater treatment plants based on predictive models and multi-objective optimization. Background Technology

[0002] Wastewater treatment plants, as a crucial component of urban infrastructure, bear the key responsibility of purifying municipal sewage and industrial wastewater. Their operational efficiency directly impacts water quality and ecological security. While my country's wastewater treatment infrastructure is now nearly complete, with a treatment rate approaching 100%, most wastewater treatment plants suffer from low levels of automation and relatively rudimentary operation and management. They often rely on excessive energy and chemical consumption to ensure stable effluent quality, which is inconsistent with my country's development goals in recent years. In particular, aeration tanks in municipal wastewater treatment plants can account for up to 75% of total energy consumption for aeration. Furthermore, due to the limitations of traditional control methods, many biological treatment units suffer from over-aeration, resulting in excessive aeration energy consumption. Therefore, achieving precise aeration is an effective strategy for reducing costs and increasing efficiency in wastewater treatment plants.

[0003] Existing aeration control technologies for wastewater treatment plants can be mainly divided into two types: manual experience control and PID control. Under manual experience control, operators manually adjust the process parameters of the aeration equipment based on their accumulated operational experience to regulate the treatment process. PID (Proportional-Integral-Derivative) control, due to its stable control effect and simple implementation, has become the most widely used automatic control strategy in wastewater treatment plants and a commonly adopted control method by engineers. It includes a proportional control unit, an integral control unit, and a derivative control unit. This method detects the deviation between the setpoint and the actual measured value of the target variable and generates corresponding adjustment quantities based on proportional, integral, and derivative operations, respectively. These adjustment quantities are then weighted and superimposed to form the final manipulated variable, used to achieve dynamic regulation and stable control of the system.

[0004] However, the effectiveness of manual experience-based control highly depends on the individual experience and professional judgment of operators, lacking standardization and consistency. Furthermore, its response is relatively delayed; when faced with sudden load fluctuations or complex changes in operating conditions, manual adjustments exhibit a certain delay, making it difficult to match the actual needs of the system in a timely and accurate manner. Manual experience-based control requires long-term on-duty deployment of highly experienced operators, increasing human resource costs and creating a management burden related to personnel training and handover. Moreover, this approach struggles to comprehensively consider multi-variable and multi-objective optimization needs in real time, limiting the potential for further optimization of wastewater treatment systems in areas such as energy consumption reduction and effluent quality improvement.

[0005] PID control is problematic because the activated sludge process in the biochemical treatment unit simultaneously performs carbon removal, nitrogen removal, and phosphorus removal. This involves the synergistic effects of various functional microorganisms and comprises multiple chemical, biological, and physical subprocesses with significant interactions between them. This results in a complex overall treatment process characterized by nonlinearity, multiple couplings, and strong interference. Given that PID control is essentially a linear control strategy, it struggles to effectively address these complexities, leading to insufficient adaptability when directly applied to activated sludge treatment. Furthermore, existing PID control methods typically fail to comprehensively evaluate the carbon removal, nitrogen removal, and phosphorus removal subprocesses during aeration regulation, lacking a systematic consideration of the interactions between these subprocesses. This makes it difficult to simultaneously address multi-objective optimization needs during regulation, limiting further improvements in overall process performance. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides a precise aeration method for wastewater treatment plants based on a predictive model and multi-objective optimization.

[0007] A precise aeration method for wastewater treatment plants based on predictive models and multi-objective optimization includes the following steps:

[0008] Multiple aeration volume values ​​from wastewater treatment plants are taken as initial samples and input into the effluent prediction model to obtain the predicted effluent index value corresponding to each aeration volume value.

[0009] Based on a multi-objective optimization algorithm, with the predicted effluent index and aeration volume as optimization objectives, the optimization solution set of the aeration volume is generated by dynamically adjusting the aeration volume.

[0010] For each aeration volume value in the optimized solution set, weights are assigned to the corresponding effluent indicators. The superior-inferior solution distance method is used to calculate and sort the weighted effluent indicators, and the optimal solution for the aeration volume value is selected.

[0011] Explanation: The above method obtains predicted effluent indicators from an initial sample of aeration volume using a predictive model, providing a foundation for subsequent optimization. Then, a multi-objective optimization algorithm is used to dynamically adjust the aeration volume to generate an optimized solution set, fully considering multiple objective factors and improving the comprehensiveness and rationality of the optimization. Finally, by assigning weights to the effluent indicators corresponding to each aeration volume in the optimized solution set and using the superior-inferior solution distance method to calculate the ranking, the optimal solution is selected. The method as a whole, through a multi-step organic combination, can accurately and efficiently determine the optimal aeration volume value that meets actual needs, contributing to scientific decision-making and refined management of aeration control in wastewater treatment plants. Specifically, this invention, by introducing a model predictive control strategy, achieves data-driven and automated aeration control decision-making, reducing reliance on the personal experience and subjective judgment of operators, improving the consistency and standardization of control, and reducing the management burden of human resource allocation and training handover. By adopting a multi-objective optimization algorithm and a multi-index decision-making method, the aeration strategy can be dynamically adjusted in real time according to changes in system load, significantly shortening the response time compared to manual adjustment and improving the control accuracy and stability under complex operating conditions. By introducing a multi-objective optimization algorithm, the effluent TN, COD, TP, and NH4+ are comprehensively considered during the aeration control process. + It can systematically balance the interaction between various sub-processes by controlling multiple key water quality indicators such as -N, breaking through the limitations of traditional PID control's single-objective optimization, and achieving comprehensive optimization of energy consumption and effluent water quality.

[0012] Furthermore, the inputs to the above-mentioned effluent prediction model are aeration volume, influent water quality data, process data, environmental data, and wastewater treatment unit data, and the output is the predicted effluent index data.

[0013] Note: The above method, by clearly defining the prediction model as an effluent prediction model and refining its input and output elements, comprehensively covers all kinds of key factors affecting the wastewater treatment effect, enabling the model to accurately capture variable correlations under complex operating scenarios and greatly improving the accuracy of prediction.

[0014] Furthermore, the method for generating an optimized solution set of aeration volume values ​​based on a multi-objective optimization algorithm, with the predicted effluent index as the optimization objective, and by dynamically adjusting the aeration volume values, includes:

[0015] First, reference data points are generated based on multiple initial samples and their corresponding predicted effluent indicators. Then, the distance between each initial sample and the reference data points is calculated, and the initial sample is associated with the nearest reference data point.

[0016] Multiple initial samples are sorted based on Pareto dominance theory, and each initial sample is assigned a priority based on dominance relationship; based on the sorting results, initial samples for generating offspring samples are selected; based on the selected initial samples for generating offspring samples, offspring samples are generated by simulating binary crossover and polynomial mutation methods.

[0017] The initial and offspring samples are merged, and the predicted effluent index is used as the optimization objective to iterate and find the optimal solution set for the aeration volume value. The iteration is completed when the number of iterations reaches the maximum value.

[0018] Explanation: The multi-objective optimization algorithm described above demonstrates excellent performance in wastewater treatment aeration control. By generating reference data points from initial samples and predicted effluent indicators, and associating these samples, a scientific benchmark is established for subsequent optimization. This effectively distinguishes the characteristics of different samples and accurately identifies the optimization direction. By employing Pareto dominance theory to sort and prioritize samples, a clear hierarchical division of samples is achieved, ensuring that individuals with greater optimization potential are selected from among numerous samples, thus improving the algorithm's optimization efficiency. Simulated binary crossover and polynomial mutation are used to generate offspring samples, enriching sample diversity. Finally, the initial and offspring samples are merged and iteratively optimized, fully utilizing the advantages of historical information and newly generated samples to continuously approach the optimal solution until the maximum number of iterations is reached, ensuring full convergence of the algorithm. Ultimately, a high-quality optimized solution set of aeration volume values ​​is obtained, providing reliable technical support for wastewater treatment plants to achieve efficient and precise aeration control, and helping to improve wastewater treatment efficiency and operational stability.

[0019] Furthermore, the process involves ranking multiple initial samples based on Pareto dominance theory, assigning a priority to each initial sample based on dominance relationships. The predicted effluent indicators of the initial samples with higher priority are better than or equal to the predicted effluent indicators of other initial samples. The lower the predicted effluent indicator is to the pollutant limit, the better the predicted effluent indicator is.

[0020] Note: The above method ensures an optimal or equivalent balance among multiple objectives (such as predicted values ​​for water quality indicators) by assigning high-priority initial samples. This approach helps find solutions that satisfy multiple objectives in complex decision-making environments, improving the practicality and effectiveness of the prediction model.

[0021] Furthermore, the step of selecting initial samples for generating offspring samples based on the sorting results is to select multiple initial samples with high priority.

[0022] Note: The above method can ensure the high-quality evolution of offspring samples and drive the optimized solution set of aeration air volume to continuously approach the ideal state.

[0023] Furthermore, the predicted values ​​of the effluent indicators include the predicted values ​​of total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen.

[0024] Note: Total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen in effluent are core indicators for measuring wastewater treatment effectiveness, covering different types of pollutants such as organic matter, nitrogen and phosphorus nutrients.

[0025] Furthermore, the predicted effluent indicators in the high-priority initial samples are better than or equal to the predicted effluent indicators in the low-priority initial samples. The lower the predicted effluent indicator is than the pollutant limit, the better the predicted effluent indicator is. The optimization target of the predicted effluent indicator is: among all the predicted effluent indicators corresponding to the aeration volume values, obtain a set of aeration volume values ​​in which the predicted values ​​of total nitrogen, chemical oxygen demand, total phosphorus, ammonia nitrogen, and aeration volume are all less than or equal to the predicted effluent indicators corresponding to other aeration volume values. That is, the predicted effluent indicators corresponding to the aeration volume values ​​in the optimized solution set are all less than or equal to the predicted effluent indicators corresponding to other aeration volume values ​​in the initial samples.

[0026] Explanation: The above method accurately constructs a clear standard for multi-objective optimization in wastewater treatment aeration control, integrating key water quality indicators such as total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen into a unified optimization framework, ensuring that the optimization process comprehensively considers all indicators; by pursuing a set of aeration volume values ​​that make the above indicators better than or equal to other schemes, it provides a clear direction for the algorithm to search for the optimal solution, avoiding the pitfalls of falling into local optima or unilaterally pursuing a single indicator.

[0027] Furthermore, the method for assigning weights to the effluent indicators corresponding to each aeration volume value in the optimized solution set of aeration volume values ​​is a combined weighted method formed by the analytic hierarchy process and the entropy weight method.

[0028] Explanation: By using the combined weighting method described above to assign weights to the effluent indicators corresponding to each aeration volume value in the optimized solution set, subjective judgment and objective data are combined to ensure the objectivity and scientific nature of the weight allocation, thereby improving the accuracy and reliability of the optimization results.

[0029] Furthermore, the aeration volume value is related to the treated water volume. When adjusting the range of the aeration volume value, it is necessary to ensure that the air-to-water ratio at the aeration volume value is 5-12. The predicted values ​​of total nitrogen in the effluent are 0.1-15 mg / L, the predicted values ​​of chemical oxygen demand in the effluent are 0.1-50 mg / L, the predicted values ​​of total phosphorus in the effluent are 0.1-1 mg / L, and the predicted values ​​of ammonia nitrogen in the effluent are 0.1-5 mg / L.

[0030] Note: The above method clearly provides the numerical range of aeration volume and predicted values ​​of key effluent indicators, providing clear and quantifiable boundary constraints for the operation of the wastewater treatment aeration control system. This allows the control algorithm and optimization process to operate within a clear and reasonable numerical range, improving the efficiency and accuracy of the algorithm.

[0031] Furthermore, when the method is applied to an aerobic biochemical treatment unit for municipal sewage or industrial wastewater, the aeration volume and influent water quality data in the input of the effluent prediction model are data from the aerobic biochemical treatment unit, and the process data, environmental data, and sewage treatment unit data in the input of the effluent prediction model are all upstream data of the aerobic biochemical treatment unit. The predicted effluent index value refers to the predicted effluent index value of the aerobic biochemical treatment unit.

[0032] The beneficial effects of this invention are:

[0033] This invention's method uses a predictive model to obtain predicted values ​​of effluent indicators from an initial sample of aeration volume, providing a foundation for subsequent optimization. Then, a multi-objective optimization algorithm is employed to dynamically adjust the aeration volume to generate an optimized solution set, fully considering multiple objective factors and improving the comprehensiveness and rationality of the optimization. Finally, by assigning weights to the effluent indicators corresponding to each aeration volume in the optimized solution set and using the superior-inferior solution distance method to calculate and rank them, the optimal solution is selected. The method as a whole, through a multi-step organic combination, can accurately and efficiently determine the optimal aeration volume value that meets actual needs, contributing to scientific decision-making and refined management of aeration control in wastewater treatment plants. Specifically, according to environmental engineering principles, aeration operations in the biochemical treatment unit have a significant impact on key water quality indicators such as total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen in the effluent. Adjusting the aeration volume indirectly regulates the growth of various functional microorganisms and the pollutant degradation process by changing the dissolved oxygen concentration in the aerobic tank, thereby affecting various effluent indicators. By introducing a multi-objective optimization algorithm, the comprehensive impact on multiple biochemical processes is considered simultaneously during aeration control, enabling balanced optimization of multiple performance indicators and further improving overall process operation and system stability. In summary, this invention overturns the traditional wastewater treatment plant operation mode of ensuring effluent compliance through excessive aeration, significantly reducing operating costs and demonstrating promising application prospects. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of the multi-objective optimization algorithm used in the embodiments of the present invention. Detailed Implementation

[0036] To further illustrate the methods and effects of this invention, the technical solution of this invention will be clearly and completely described below in conjunction with experiments.

[0037] In light of the aforementioned background information, aeration volume is a critical parameter in practical applications such as wastewater treatment, directly affecting multiple water quality indicators (such as total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen in the effluent). However, adjusting the aeration volume often leads to interrelationships and conflicts among these water quality indicators. For example, increasing the aeration volume may reduce the ammonia nitrogen concentration in the effluent, but it may also cause changes in other indicators such as total nitrogen in the effluent, and may even increase operating costs.

[0038] Therefore, there is no single aeration volume value that can simultaneously optimize all water quality indicators. Instead, it is necessary to find a set of aeration volume values ​​that allow each water quality indicator to reach a relatively balanced "optimal" state under certain constraints. This is the Pareto optimal solution set in multi-objective optimization. Decision-makers then determine the optimal aeration volume value based on this optimal solution set. The specific details are as follows:

[0039] Example 1: A precise aeration method for wastewater treatment plants based on predictive models and multi-objective optimization, comprising the following steps:

[0040] S1. Take multiple aeration volume values ​​from the sewage treatment plant as initial samples and input them into the effluent prediction model to obtain the predicted effluent index value corresponding to each aeration volume value.

[0041] The inputs to the above effluent prediction model are aeration volume, influent water quality data, process data, environmental data, and wastewater treatment unit data, and the output is the predicted effluent index data.

[0042] The predicted values ​​for effluent indicators include the predicted values ​​for total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen.

[0043] The aforementioned water discharge prediction model comprises a data sharing layer, an attention network layer, and an output layer connected in sequence. The data sharing layer describes the temporal dependencies between the various inputs in the water discharge prediction model. The attention network layer extracts the inputs related to each water discharge indicator from the data sharing layer. The output layer fits the corresponding water discharge indicator data based on the water discharge indicators in the multi-task attention layer and the extracted inputs. The water discharge prediction model is implemented in PyTorch and / or TensorFlow.

[0044] S2. Based on a multi-objective optimization algorithm, with the predicted effluent index and aeration volume as optimization objectives, the optimization solution set of the aeration volume is generated by dynamically adjusting the aeration volume.

[0045] The optimization objective of the above-mentioned effluent index prediction values ​​refers to obtaining a set of aeration volume values ​​among all the effluent index prediction values ​​corresponding to aeration volume values, such as the predicted values ​​of total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen in the effluent, and the aeration volume value, all of which are less than or equal to the predicted effluent index values ​​corresponding to other aeration volume values. In other words, the predicted effluent index values ​​corresponding to the aeration volume values ​​in the optimized solution set are all less than or equal to the predicted effluent index values ​​corresponding to other aeration volume values ​​in the initial sample.

[0046] Specifically: In the above optimization algorithm, optimization is achieved by using an optimization objective function, which includes four effluent objective functions and one air volume objective function. The effluent objective functions include effluent total nitrogen, effluent chemical oxygen demand, effluent total phosphorus and effluent ammonia nitrogen. The air volume objective function is the aeration air volume. The comprehensive optimization function in the optimization algorithm is expressed as the following formula (1).

[0047]

[0048] In Equation (1), the goal is to minimize the aeration volume x that minimizes each objective function. The normalized form of each objective function in Equation (1) is shown in Equation (2) below.

[0049]

[0050] in Represents one of five objective functions. This represents the index value corresponding to the objective function. These represent the minimum and maximum values ​​of the corresponding indicator data, respectively; the above normalization is to prevent adverse effects caused by differences in data dimensions.

[0051] The multi-objective optimization algorithm is linked with the effluent prediction model. It is mainly responsible for finding the optimal aeration volume under the current state based on the prediction results of the effluent prediction model. Since the aeration volume parameters of the aeration tank have an important impact on the carbon removal, phosphorus removal and nitrogen removal processes, the multi-objective optimization algorithm needs to consider multiple optimization objectives at the same time to find the optimal aeration volume. The final optimization result is a set of Pareto optimal solutions.

[0052] The specific process is as follows: Figure 2 As shown, the method for generating an optimized solution set of aeration air volume values ​​based on the optimization algorithm includes:

[0053] S2-1. First, based on multiple initial samples and corresponding predicted values ​​of effluent indicators, reference data points are generated. Then, the distance between each initial sample and the reference data points is calculated, and the initial sample is associated with the nearest reference data point. The Das-Dennis method is used to uniformly distribute 120 reference data points in the three-dimensional target space and map them onto a normalized hyperplane.

[0054] S2-2. Based on Pareto dominance theory, sort multiple initial samples, and assign priority to each initial sample based on dominance relationship; select initial samples for generating offspring samples according to the sorting results; generate offspring samples according to the selected initial samples for generating offspring samples by simulating binary crossover and polynomial mutation methods (based on the basic mechanism of genetic algorithm).

[0055] Specifically, the above-mentioned initial samples are sorted based on Pareto dominance theory. Each initial sample is assigned a priority based on dominance relationship. The predicted effluent index values ​​of the initial samples with higher priority are better than or equal to (here, better than or equal to means that the pollution index is less than or equal to) the predicted effluent index values ​​of other initial samples. The initial samples used to generate offspring samples are selected according to the sorting results by selecting multiple initial samples with higher priority.

[0056] S2-3. Merge the previous initial sample and offspring sample, and use the predicted value of the effluent index as the optimization target to enter the above objective function formula (1) for iterative optimization until the number of iterations reaches the maximum value of 50 to 200 times (set to 200 in this embodiment of the invention) to complete the iteration and obtain the optimized solution set of aeration air volume value.

[0057] The specific explanations of S2-1 to S2-3 above, in conjunction with the embodiments of the present invention, are as follows:

[0058] First, reference data points are defined to evaluate the diversity of the solution set. The coordinates of the generated reference data points will not change during subsequent optimization. Then, the algorithm first connects the reference data points to the origin of the normalized target space, forming a set of reference directions. For each solution in the population, the algorithm calculates its distance to each reference direction and associates each solution with the reference direction with the shortest distance and its corresponding reference data point.

[0059] After establishing the mapping relationship between initial sample individuals and reference data points, the initial sample is first divided into non-dominated levels according to fast non-dominated sorting, and initial samples for generating offspring samples are selected from the best to worst according to the level. When adding all solutions of a certain level would make the size of the parent population larger than the sample size limit, a niche preservation operation based on reference data points is performed. This operation first counts the number of individuals associated with each reference data point in the current initial samples (which are less than N), called the niche count. Subsequently, when the algorithm continues to select initial samples, it prioritizes selecting individuals associated with the reference data point with the lowest current niche count, until the number of initial samples equals N, thereby maintaining population diversity.

[0060] After the initial sample selection is completed, the algorithm performs crossover and mutation operations to generate the next generation of offspring samples. The simulated binary crossover and polynomial mutation methods specifically involve exchanging two or more aeration air volume values ​​once. The mutation operation refers to randomly modifying the air volume value of the current parent individual according to a certain probability.

[0061] The parent population and the generated offspring population are merged into a new population. Finally, the algorithm iteratively repeats the above steps, such as associating solutions with reference data points, fast non-dominated sorting, and niche preservation operations, until the termination condition is met, and gives the final Pareto approximate solution set.

[0062] S2-4. Based on the above, in another implementation method of this invention, energy consumption cost and process stability are added as optimization objectives of the multi-objective optimization algorithm; specifically as follows:

[0063] The objective function expression for energy consumption cost is as follows (3):

[0064]

[0065] In equation (3), P 风机 Q represents the aeration power. 曝气 (t) represents the aeration air volume value, and Δt represents the time period for each prediction; it should be understood that the lower the energy cost, the better for the solution.

[0066] The objective function for process stability is expressed as follows (4):

[0067]

[0068] In equation (4), Q satisfies 曝气 ∈[Q min Q max ];

[0069] It should be understood that the objective function for process stability is to prevent the aeration volume from changing too drastically. In other words, it limits the difference between the aeration volume proposed in the current optimization and the aeration volume at the previous moment. Here, the f3 value represents the range of change in aeration volume. The lower the value, the better the operational stability and the smaller the adjustment range. The higher the value, the larger the adjustment range and the poorer the process stability. Drastic adjustments to the aeration volume may damage the blower equipment or shorten its service life. It can be said that the higher the stability of the optimization adjustment, the better. In other words, the lower the objective function value, the better.

[0070] Specifically, the aeration volume value is related to the treatment volume. When adjusting the range of the aeration volume value, it is necessary to ensure that the air-to-water ratio at the aeration volume value is 5-12. The predicted values ​​of total nitrogen in the effluent are 0.1-15 mg / L, the predicted values ​​of chemical oxygen demand in the effluent are 0.1-50 mg / L, the predicted values ​​of total phosphorus in the effluent are 0.1-1 mg / L, and the predicted values ​​of ammonia nitrogen in the effluent are 0.1-5 mg / L.

[0071] S2-5. When the predicted value of the effluent water quality index is greater than the limit, a penalty term is added to the objective function value of the effluent water quality, which is the objective function value corresponding to the current aeration volume value multiplied by the penalty score.

[0072] The penalty score P satisfies: P = C·max{M - M0, 0}

[0073] Where M is the predicted value of the indicator, M0 is the range limit, and C is the penalty factor (value between 1000 and 10000) to force the avoidance of exceeding the limit;

[0074] Since multi-objective optimization problems do not inherently possess a unique optimal solution, optimization algorithms typically output a set of solutions that approximate Pareto optimal solutions after optimization. These solutions cannot be directly used for the final control of aeration equipment. Therefore, it is necessary to introduce an appropriate decision-making method to select the final control solution from this solution set that best meets the operational needs and actual conditions of the wastewater treatment plant, thereby completing the final control action in the aeration optimization process. This scheme introduces a multi-index decision-making method, comprehensively considering subjective preferences and objective data factors, to assist in achieving the final decision-making for aeration control, thus realizing automated and real-time optimization and control of aeration volume; details are as follows in S3.

[0075] S3. Assign weights to the effluent indicators corresponding to each aeration volume value in the optimized solution set of aeration volume values, calculate and sort the effluent indicators after weighting using the superior-inferior solution distance method, and select the optimal solution of aeration volume value.

[0076] The weighting method is a combined weighting method consisting of the analytic hierarchy process (AHP) and the entropy weighting method; specifically, it includes:

[0077] The judgment matrix of each optimization objective is constructed by the analytic hierarchy process, and the subjective weight w1 is determined by the consistency test; the objective weight w2 is calculated by the entropy weight method based on the information entropy of the distribution of each objective indicator.

[0078] The weight w is obtained through linear combination. z w z =αw1+(1-α)w2; α can take the value 0.6;

[0079] Based on weight w z The calculation of the superior and inferior solution distance method includes:

[0080] Weighted Euclidean distance calculation: (distance from the i-th solution to the positive ideal solution) positive distance

[0081] (Distance from the i-th solution to the negative ideal solution) Negative distance

[0082] In the formula, j represents the number of optimization objectives, and x represents the number of optimization objectives. ij Let x be the value of the i-th scheme on the j-th index. j + Let x be the optimal value of the i-th scheme on the j-th metric; j - Let be the worst value of the i-th scheme on the j-th metric;

[0083] Calculate each aeration air volume value C i The closer to 1, the better, according to C i Sort the values ​​and select the optimal aeration air volume.

[0084] For example, the technical effectiveness of this solution was tested based on historical data from a wastewater treatment plant in southern China. Compared to actual data, the method in this solution achieved an average reduction of 44.74% in aeration volume, with the optimized average aeration volume reduced to 1282.70 m³ / s. 3 The aeration rate was reduced by approximately 5% per hour, and the overall exceedance rate of effluent indicators was reduced by nearly 5%. Overall, this solution achieves cost reduction and efficiency improvement in the aeration process.

[0085] Example 2: The method is applied to an aerobic biochemical treatment unit for municipal sewage or industrial wastewater. The aeration volume and influent water quality data in the input of the effluent prediction model are data from the aerobic biochemical treatment unit. The process data, environmental data, and sewage treatment unit data in the input of the effluent prediction model are all upstream data of the aerobic biochemical treatment unit. The predicted effluent index value refers to the predicted effluent index value of the aerobic biochemical treatment unit.

[0086] In summary, since multi-objective optimization problems cannot minimize all objective functions and the objective functions conflict with each other, minimizing one objective function may lead to the deterioration of another. Therefore, this embodiment of the invention adopts multi-objective optimization to obtain a set of non-dominated solutions that are not inferior to (i.e., less than or equal to) any other solution in all objective functions. Then, based on the combination weighting + multi-index decision method, i.e., the TOPSIS method, the solution that best meets the decision-maker's needs is automatically selected from the non-dominated solution set according to subjective and objective conditions as the final actual aeration air volume value.

Claims

1. A precise aeration method for wastewater treatment plants based on predictive models and multi-objective optimization, characterized in that, Includes the following steps: Multiple aeration volume values ​​from wastewater treatment plants are taken as initial samples and input into the effluent prediction model to obtain the predicted effluent index value corresponding to each aeration volume value. Based on a multi-objective optimization algorithm, with the predicted effluent index and aeration volume as optimization objectives, the optimization solution set of the aeration volume is generated by dynamically adjusting the aeration volume. For each aeration volume value in the optimized solution set, weights are assigned to the corresponding effluent indicators. The superior-inferior solution distance method is used to calculate and sort the effluent indicators after weight assignment, and the optimal solution for the aeration volume value is selected.

2. The method as described in claim 1, characterized in that, The effluent prediction model takes aeration volume, influent water quality data, process data, environmental data, and wastewater treatment unit data as inputs and outputs predicted effluent index data.

3. The method as described in claim 2, characterized in that, The method based on a multi-objective optimization algorithm, which uses the predicted effluent index as the optimization objective and dynamically adjusts the aeration volume value to generate a set of optimized solutions for the aeration volume value, includes: First, based on multiple initial samples and their corresponding predicted effluent indicators, reference data points for the predicted effluent indicators are generated. Then, the distance between each initial sample and the reference data points is calculated, and the initial sample is associated with the nearest reference data point. Multiple initial samples are sorted based on Pareto dominance theory, and each initial sample is assigned a priority based on dominance relationship. According to the sorting results, initial samples for generating offspring samples are selected. Offspring samples are generated by simulating binary crossover and polynomial mutation methods based on the selected initial samples for generating offspring samples. The multiple initial samples and offspring samples are merged, and the predicted effluent index is used as the optimization objective to perform iterative optimization until the number of iterations reaches the maximum value, thus completing the iteration and obtaining the optimized solution set of aeration volume value.

4. The method as described in claim 3, characterized in that, The initial samples for generating offspring samples are selected based on the sorting results by selecting multiple initial samples with high priority.

5. The method as described in claim 4, characterized in that, The predicted values ​​of effluent indicators in the high-priority initial samples are better than or equal to the predicted values ​​of effluent indicators in the low-priority initial samples. The lower the predicted value of the effluent indicator is than the pollutant limit, the better the predicted value of the effluent indicator.

6. The method as described in claim 2, characterized in that, The predicted values ​​for effluent indicators include predicted values ​​for total nitrogen, chemical oxygen demand, total phosphorus, and ammonia nitrogen.

7. The method as described in claim 6, characterized in that, The predicted effluent index values ​​corresponding to the aeration volume values ​​in the optimized solution set are all less than or equal to the predicted effluent index values ​​corresponding to other aeration volume values ​​in the initial sample.

8. The method as described in claim 1, characterized in that, The method for assigning weights to the effluent indicators corresponding to each aeration volume value in the optimized solution set of aeration volume values ​​is a combined weighted method formed by the analytic hierarchy process and the entropy weight method.

9. The method as described in claim 6, characterized in that, The predicted values ​​for total nitrogen in the effluent range from 0.1 to 15 mg / L, chemical oxygen demand in the effluent range from 0.1 to 50 mg / L, total phosphorus in the effluent range from 0.1 to 1 mg / L, and ammonia nitrogen in the effluent range from 0.1 to 5 mg / L.

10. The application of the method as described in any one of claims 2 to 9, characterized in that, When the method is applied to an aerobic biochemical treatment unit for municipal sewage or industrial wastewater, the aeration volume and influent water quality data in the input of the effluent prediction model are data from the aerobic biochemical treatment unit, and the process data, environmental data, and sewage treatment unit data in the input of the effluent prediction model are upstream data of the aerobic biochemical treatment unit. The predicted effluent index value refers to the predicted effluent index value of the aerobic biochemical treatment unit.

Images