Energy-saving optimization control method and system for sewage treatment aeration system based on multi-source information fusion

Abstract

The invention relates to the field of sewage treatment, and discloses an energy-saving optimization control method and system for a sewage treatment aeration system based on multi-source information fusion, and the method comprises the steps: firstly, collecting the quality and quantity of inlet water, the environment of a biological tank and an external electricity price signal in real time; then, dynamically calculating a dissolved oxygen set value through a threshold weight rule model based on water inlet parameters, generating a preliminary control instruction by taking the dissolved oxygen set value as a feedforward target, and meanwhile, performing feedback fine tuning in combination with an actual dissolved oxygen value in the pool; a multi-target reward function fusing water quality, energy consumption and stability targets is constructed, the weight of the multi-target reward function is dynamically adjusted according to the real-time electricity price, and an economically optimal dissolved oxygen set value correction instruction is decided through optimization; according to the method, advanced accurate control of aeration requirements and online optimization of operation cost are realized, the problems of response lagging and high energy consumption of traditional control are effectively solved, the stability of effluent quality is remarkably improved, and energy consumption is reduced.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment, and in particular to an energy-saving optimization control method and system for wastewater treatment aeration systems based on multi-source information fusion. Background Technology

[0002] In activated sludge wastewater treatment processes, the aeration system is the core unit of aerobic biological treatment, and its energy consumption accounts for approximately 50%-70% of the total energy consumption of the entire wastewater treatment plant. Therefore, the control precision and operating efficiency of the aeration system are directly related to the stable attainment of effluent quality standards and the overall operating cost of the plant.

[0003] Currently, most wastewater treatment plants use a single-loop PID control strategy based on a fixed dissolved oxygen (DO) setpoint in their aeration systems. This strategy typically involves installing an online DO monitoring instrument at the end of the aerobic zone of the biological treatment tank. The controller compares the detected actual DO value with a manually set fixed value (usually 1.5-2.5 mg / L) and adjusts the fan speed or valve opening using a PID algorithm to maintain the actual DO value near the setpoint.

[0004] However, this method suffers from severe lag in response and poor ability to cope with shock loads, resulting in a hydraulic retention time of several hours for wastewater in the biological tank. When influent flow and water quality (such as COD and ammonia nitrogen) fluctuate drastically (shock load), pollutants must reach the downstream DO monitoring point before causing a change in DO value, which the controller then adjusts accordingly. This results in a lag of several hours throughout the system. Consequently, the biological tank remains in an "unhealthy" state for an extended period during shock loads, potentially leading to fluctuations or even exceedances in effluent water quality.

[0005] Furthermore, the control is inefficient, resulting in significant energy waste, as the fixed DO setpoint cannot adapt to dynamic changes in the influent load. Maintaining a high DO setpoint at night or during low-load periods leads to over-aeration, wasting a large amount of electrical energy on unnecessary oxygen supply; during peak daytime load periods, the fixed DO setpoint may be insufficient to meet the oxygen demand for microbial degradation of pollutants, resulting in inadequate treatment. Summary of the Invention

[0006] To address at least one of the aforementioned technical problems, this invention proposes an energy-saving optimization control method and system for wastewater treatment aeration systems based on multi-source information fusion.

[0007] To achieve the above objectives, this invention proposes an energy-saving optimization control method and system for wastewater treatment aeration systems based on multi-source information fusion, the specific technical solution of which is as follows.

[0008] An energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion includes the following steps: S1. Real-time data acquisition steps: acquire influent water quality parameters, influent flow rate parameters, environmental parameters inside the biological tank, and external electricity price signals; S2, Dynamic dissolved oxygen setpoint calculation step: Based on the influent water quality parameters and influent flow rate parameters, the dynamic dissolved oxygen setpoint is calculated in real time through a preset dynamic calculation model; S3, Feedforward-feedback composite control step: The dynamic dissolved oxygen setpoint is used as the feedforward control target to generate preliminary fan control commands. At the same time, based on the deviation between the actual dissolved oxygen measurement value in the environmental parameters of the biological tank and the dynamic dissolved oxygen setpoint, the preliminary fan control commands are finely adjusted to generate the final fan control commands. S4. Economic optimization step: Under the premise of ensuring that the effluent water quality meets the standards, dynamically adjust the control parameters in the feedforward-feedback composite control step. S5. Execution step: Send the optimized final fan control command to the fan frequency converter to adjust the aeration volume.

[0009] Furthermore, the dynamic dissolved oxygen setpoint calculation step specifically includes: Based on the influent COD concentration and influent flow rate, the influent COD load is calculated using the following formula: COD_load = COD Flow 0.001; Wherein, COD is the influent COD concentration, in mg / L; Flow is the influent flow rate, in m³ / h; COD_load is the COD load, in kg / h; Based on the COD_load and combined with the influent ammonia nitrogen concentration (NH3_N), water temperature (T), and influent total nitrogen concentration (TN), the dynamic dissolved oxygen setpoint (DO_set) is calculated using a threshold-weighted rule model. The calculation formula is as follows: DO_set = BASE_DO + W_cod + W_nh + W_temp + W_tn; Wherein, the influent ammonia nitrogen concentration NH3_N, water temperature T and influent total nitrogen concentration TN are all parameters collected in real time through step S1; Where BASE_DO is the preset base dissolved oxygen value; W_cod, W_nh, W_temp, and W_tn are the weight values ​​corresponding to COD load, ammonia nitrogen concentration, water temperature, and total nitrogen concentration, respectively. Each weight value is obtained from the preset weight array by comparing the real-time value of the corresponding parameter with the preset threshold array; The calculated DO_set is subjected to amplitude limiting to satisfy: MIN_DO ≤ DO_set ≤ MAX_DO; MIN_DO and MAX_DO are the preset lower and upper limits of dissolved oxygen settings.

[0010] Furthermore, the method for determining the weight values ​​W_cod, W_nh, W_temp, and W_tn is as follows: The actual value of the parameter X is compared with a set of increasing thresholds T1, T2, T3, and the weight index idx is determined according to the following formula: If X ≥ T3, then idx = 3 If X ≥ T2, then idx = 2 If X ≥ T1, then idx = 1 Otherwise, idx = 0 Based on the weight index idx, retrieve the weight value W = W[idx] from the corresponding weight arrays W0, W1, W2, W3.

[0011] Furthermore, the economic optimization step is achieved by constructing and optimizing a multi-objective reward function R, the expression of which is: R = W1·R_quality + W2·R_energy + W3·R_stability; Among them, R_quality is the water quality reward item, R_energy is the energy consumption reward item, R_stability is the stability reward item, and W1, W2, and W3 are dynamically adjusted weight coefficients. The formula for calculating the water quality incentive R_quality is as follows: R_quality = - Σᵢ [λᵢ × max(0, C_{i,out} - C_{i,standard})² ]; Wherein, λᵢ is the weighting coefficient of the i-th water quality indicator, C_{i,out} is the effluent concentration of the i-th indicator, and C_{i,standard} is the emission standard limit of the i-th indicator. The formula for calculating the energy consumption bonus item R_energy is as follows: R_energy = - P_total / P_baseline; Where P_total is the real-time total power consumption of the aeration system, and P_baseline is the baseline power consumption; The formula for calculating the stability reward R_stability is as follows: R_stability = - Σⱼ |DO_j - DO_{target,j}|; Where DO_j is the actual dissolved oxygen value at the j-th monitoring point, and DO_{target,j} is the target dissolved oxygen value at the j-th monitoring point; The weighting coefficients W1, W2, and W3 are dynamically adjusted based on real-time electricity price signals. The weight of W2 is increased during peak electricity price periods, and the weights of W1 and W3 are increased during off-peak electricity price periods. The weighting coefficient W2 increases when the real-time electricity price is higher than the average electricity price by a set percentage, and the weighting coefficients W1 and / or W3 increase when the real-time electricity price is lower than the average electricity price by a set percentage.

[0012] Furthermore, the parameters of the dynamic calculation model are updated online through a parameter adaptation mechanism; The parameter adaptation mechanism is triggered when the model's prediction error exceeds a set threshold, and the model parameter vector θ is updated according to the gradient descent method. The update formula is as follows: θ_{t+1} = θ_t - η × ∇J(θ_t); Where θ_t is the model parameter vector at time t, η is the learning rate, J(θ) is the performance evaluation function constructed with the prediction error, and ∇J(θ_t) is the gradient of the performance function at θ_t.

[0013] Furthermore, it also includes fault diagnosis and self-recovery steps: The formula for calculating the anomaly score of sensor data is: Anomaly_score = |X_measured - X_expected| / σ; Where X_measured is the actual measured value of the sensor, X_expected is the expected value based on the process model or adjacent sensor data, and σ is the standard deviation of the historical data of this parameter; When the abnormal score exceeds the set threshold, it is determined to be a sensor malfunction, triggering the self-recovery mechanism; The self-recovery mechanism includes: activating a backup sensor, and / or estimating the fault point data through interpolation based on measurements from adjacent normal sensors, using the following estimation formula: X_fault = (X_adjacent1 + X_adjacent2) / 2; And reduce the weight of the faulty area in control decisions.

[0014] Furthermore, a cloud-edge collaborative architecture is adopted for implementation; The cloud server layer is responsible for periodically collecting historical operational data and refitting and optimizing the parameters in the dynamic calculation model based on regression analysis. The regression model is in the following form: DO_set= β0 + β1×NH3_N + β2×COD_load + β3×T + β4×TN + β5×Flow; Where β0 is the intercept, and β1 to β5 are the regression coefficients, which are obtained by least squares fitting. The edge computing layer is responsible for executing the steps of the method described in any one of claims 1 to 6 in real time, and receiving optimized model parameters from the cloud server layer.

[0015] Furthermore, the feedback fine-tuning is achieved through a PID controller, with the input being the dissolved oxygen deviation e(t) = DO_set - DO_measured, and the output u(t) calculated using the following formula: u(t) = K_p e(t) + K_i ∫e(τ)dτ + K_d (de(t) / dt); Where K_p, Ki, and K_d are the proportional, integral, and differential coefficients, respectively.

[0016] Furthermore, following the execution steps, there are also effect evaluation and parameter adjustment steps: Real-time monitoring of effluent water quality indicators and system energy consumption; Based on the monitoring results, evaluate the performance of the current control strategy; Based on the performance evaluation results, the control parameters in the feedforward-feedback composite control step or the economic optimization step are dynamically adjusted.

[0017] A wastewater treatment aeration energy-saving optimization control system based on the method of any one of claims 1-9, comprising: The data perception layer includes: The influent monitoring unit is installed at the inlet of the biological tank to collect influent water quality parameters and influent flow parameters in real time. It includes at least an influent COD online analyzer, an influent ammonia nitrogen online analyzer, an influent total nitrogen online analyzer, and an influent flow meter. The process monitoring unit is set up inside the biological tank to collect environmental parameters in the biological tank in real time. It includes at least a water temperature sensor, a pH sensor, and multiple dissolved oxygen sensors distributed at the beginning, middle and end of the aerobic tank. The energy consumption and external data interface unit is used to collect wind turbine energy consumption data and receive real-time electricity price signals from the external power grid. The intelligent control layer is communicatively connected to the data sensing layer. The intelligent control layer includes a programmable logic controller (PLC) that stores a computer program. When executed by a processor, the computer program is used for: Based on the influent water quality parameters and influent flow rate parameters from the data sensing layer, the dynamic dissolved oxygen setpoint is calculated in real time through a preset dynamic calculation model. The initial fan control command is generated using the dynamic dissolved oxygen setpoint as the feedforward control target. At the same time, the initial fan control command is fine-tuned based on the deviation between the actual dissolved oxygen measurement value from the process monitoring unit and the dynamic dissolved oxygen setpoint, and the final fan control command is generated. Under the premise of ensuring that the effluent water quality meets the standards, the control strategy of the aeration system is adjusted according to the real-time electricity price signal from the energy consumption and external data interface unit. The execution layer, which is communicatively connected to the intelligent control layer, includes a fan frequency converter and / or an air regulating valve. It is used to receive the final fan control command issued by the intelligent control layer and adjust the fan speed and valve opening to control the aeration volume input to the biological tank.

[0018] The present invention has the following advantages: This invention achieves advanced and precise control of aeration demand by constructing a dynamic dissolved oxygen setpoint model based on influent water quality and quantity feedforward, and combining it with feedback fine-tuning from real-time dissolved oxygen measurement. This effectively overcomes the response lag problem caused by hydraulic retention time in traditional control systems, significantly improving the system's ability to cope with shock loads and the stability of effluent water quality. Furthermore, by introducing an economic optimization decision-making mechanism based on a multi-objective reward function and real-time electricity price signals, the system can intelligently balance energy consumption and accuracy to reduce electricity costs during peak electricity price periods and strengthen control to improve water quality assurance during off-peak electricity price periods, while ensuring that effluent meets standards, thus achieving overall optimization of operating costs. In addition, integrated fault self-diagnosis and self-recovery functions ensure the long-term reliable operation of the system, while the cloud-edge collaborative parameter continuous optimization architecture enables the system to adapt to different plant conditions and operating conditions. In summary, the present invention ultimately achieved multiple beneficial effects in actual testing, including a significant reduction in energy consumption of the aeration system, a steady increase in the effluent compliance rate, and a substantial reduction in the frequency of equipment adjustments. It completed the synergistic optimization of the wastewater treatment process in terms of energy saving, stability, quality improvement, and cost reduction. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall logic of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] Example 1: This embodiment provides an energy-saving optimization control method and system for wastewater treatment aeration systems based on multi-source information fusion, such as... Figure 1 As shown, the system includes a data perception layer, an intelligent control layer, and an execution layer.

[0022] Since the method relies on running on the system, the method will now be explained further in conjunction with the system. The specific steps are as follows: Step S1: Real-time data acquisition.

[0023] The data perception layer is responsible for collecting information from multiple sources in real time, specifically including: Influent water quality and flow parameters: Real-time data are collected using online analytical instruments and flow meters installed on the main influent pipe of the biological treatment tank, including influent chemical oxygen demand (COD) concentration (mg / L), influent ammonia nitrogen (NH3-N) concentration (mg / L), influent total nitrogen (TN) concentration (mg / L), and instantaneous influent flow rate (m³ / h). Data acquisition frequency is no less than once per minute to ensure the capture of dynamic changes in the influent load.

[0024] Environmental parameters in the biological tank: Water temperature (T, unit: °C), pH value, and at least three dissolved oxygen (DO) concentration values ​​(unit: mg / L) distributed at the beginning, middle and end of the aerobic tank are collected in real time by a sensor array installed in the aerobic tank.

[0025] External electricity price signal: Real-time time-of-use electricity price signal (unit: yuan / kWh) is received from the smart meter or energy management platform of the power grid company via a data interface.

[0026] Step S2: Calculation of dynamic dissolved oxygen setpoint.

[0027] This step is the core of feedforward control, aiming to predict and set reasonable DO target values ​​in advance based on the influent load, overcoming system lag. It involves performing the following calculations through an intelligent control layer, such as a high-performance PLC or edge computing gateway: The technical principle behind the calculation of the dynamic dissolved oxygen setpoint in this invention lies in the significant linear correlation between dissolved oxygen demand and feedforward variables such as influent water quality and quantity. Based on this, a prediction model is first established through multiple linear regression analysis of historical operating data: In the formula: DO_optimal: is the optimal dissolved oxygen value predicted by the model that matches the current influent load, in mg / L.

[0028] : This is the intercept term of the regression model, representing the baseline dissolved oxygen level, in mg / L.

[0029] NH3_N: The concentration of ammonia nitrogen in the influent, in mg / L.

[0030] : is the regression coefficient of influent ammonia nitrogen concentration NH3_N, reflecting the contribution of ammonia nitrogen to dissolved oxygen demand.

[0031] COD_load: The influent COD load, calculated from COD concentration (mg / L) × Flow (m³ / h) × 0.001, in units of kg / h.

[0032] : is the regression coefficient of influent COD load (COD_load).

[0033] T: Water temperature in the biological treatment tank, in °C.

[0034] : is the regression coefficient of water temperature T, which is usually positive to compensate for the effect of reduced oxygen solubility at high temperatures.

[0035] TN: Total nitrogen concentration in influent, in mg / L.

[0036] : is the regression coefficient of total nitrogen concentration (TN) in the influent.

[0037] Flow: Instantaneous influent flow rate, in m³ / h.

[0038] : is the regression coefficient of influent flow rate.

[0039] To ensure the stable and efficient operation of this model in industrial embedded controllers (such as PLCs), it needs to be transformed into a rule-based form with fast query speed and strong anti-interference capability. Therefore, this invention further provides a rule model implementation method based on threshold weights, as described below. The parameters (threshold and weight) of this rule model are derived from the regression model analysis above, and the two are technically unified.

[0040] Engineering model construction method: To transform the above continuous regression model into a discrete rule model suitable for execution by an embedded controller, the following steps are used to determine the rule parameters: Calculate the influent COD load: Rule model calculation formula: Based on the parameters obtained by the above method, the following rule model calculation formula based on threshold weight is constructed for real-time calculation of dynamic DO settings: COD_load is measured in kg / h and reflects the rate at which organic pollutants enter the biological system.

[0041] First, data preparation and regression analysis: Historical datasets covering various operating conditions were collected, including influent NH3_N, COD_load, T, TN, Flow, and corresponding optimal DO values. Through multiple linear regression fitting, a set of regression coefficients β0, β1, β2, β3, β4, β5 and a regression equation were obtained.

[0042] Subsequently, the variable is segmented (thresholds are determined): For each feedforward variable (such as NH3_N), based on its historical data distribution and process understanding, its value range is divided into several continuous intervals, such as [low, medium, high, extremely high]. The boundary points of the intervals constitute the threshold array. For example, for NH3_N, the threshold array could be [20, 30, 40] (unit: mg / L).

[0043] Finally, calculate the interval weights (determine the weights): For each interval of a variable (e.g., NH3_N), using the regression equation obtained in step a, calculate the increment of the predicted DO value of all sample points within that interval relative to BASE_DO (i.e., β1 × NH3_N), and obtain the average of this increment. Use this average as the weight value corresponding to that interval. Repeat this process for all variables to obtain their respective weight arrays.

[0044] Based on the parameters obtained by the above method, the following threshold-weighted rule model calculation formula is constructed for real-time calculation of dynamic DO settings: BASE_DO is the preset baseline dissolved oxygen value, such as 2.0 mg / L.

[0045] W_cod, W_nh, W_temp, and W_tn represent the adjustment weights for four key parameters: COD load, ammonia nitrogen concentration, water temperature, and total nitrogen concentration.

[0046] The weights are determined as follows (taking W_cod as an example): The real-time COD load (COD_load) is compared with a set of preset incremental thresholds [T1, T2, T3] (e.g., [500, 1000, 1500] kg / h) to determine a weighted index idx: If COD_load ≥ T3, then idx = 3 If COD_load ≥ T2, then idx = 2 If COD_load ≥ T1, then idx = 1 Otherwise, idx = 0 Subsequently, based on idx, the corresponding weight value W_cod = W[idx] is retrieved from the preset weight array [W0, W1, W2, W3] (e.g., [0.1, 0.3, 0.5, 0.8]). The weights W_nh, W_temp, and W_tn for ammonia nitrogen, water temperature, and total nitrogen are obtained using the exact same logic, employing their respective threshold and weight arrays. These threshold and weight arrays are based on the inventor's first and second phase field practice analysis and can be periodically updated via a cloud-based optimization module.

[0047] Amplitude limiting: To ensure the stability of the biochemical process and the safety of the equipment, the calculated DO_set is limited: MIN_DO and MAX_DO are the lower and upper limits allowed by the process, for example, set to 1.5 mg / L and 4.0 mg / L respectively.

[0048] Step S3: Feedforward-feedback composite control step.

[0049] This step combines the speed of feedforward with the precision of feedback to generate the final control command.

[0050] Feedforward control: The dynamic DO_set calculated in step S2 is used as the feedforward control target. Based on the DO_set and a preset baseline DO-airflow relationship curve, the controller directly generates a preliminary fan frequency or airflow control command. This is equivalent to adjusting the aeration intensity in advance before changes in the influent load affect the DO in the tank.

[0051] Feedback fine-tuning: Simultaneously, the system continuously monitors the average value DO_measured from multiple DO sensors within the biological tank. The deviation e(t) = DO_set - DO_measured from the target value DO_set is calculated. An anti-integral saturation PID controller is used to process this deviation, outputting a feedback correction Δu(t): Where K_p, K_i, and K_d are the tuned controller parameters.

[0052] Command synthesis: The feedforward control command is added to the feedback correction amount Δu(t) to generate the final fan control command, such as the target frequency or target power, thereby achieving precise and rapid adjustment of the aeration volume.

[0053] Step S4, Economic Optimization Step.

[0054] This step aims to incorporate real-time electricity price signals and economic objectives into control decisions, seeking the optimal total operating cost while ensuring water quality meets standards. Its core is running a rolling optimization decision-maker based on a multi-objective reward function, which executes the following algorithm at fixed intervals (e.g., every 5-15 minutes): Decision-making cycle and data preparation The optimization decision engine starts at a set period T_opt (e.g., every 5 minutes). Each time it starts, the following real-time data is collected: Current real-time electricity price: Price_current.

[0055] Current water parameters: NH3_N_in, COD_in, TN_in, Flow, T.

[0056] The current dynamic DO setting value is DO_set_ff (from S2).

[0057] The current average operating power of the wind turbine is P_blower.

[0058] Average effluent quality during the previous decision-making cycle (e.g., from online instruments or soft measurement models): .

[0059] Dynamic weight calculation The weights of each item in the incentive function are adjusted according to the real-time electricity price to reflect the economic priority at different times: Calculate the electricity price ratio factor , where Price_average is the daily average electricity price.

[0060] Dynamically set weights: Water quality weight W1: f1 is the decay function. When k_price > 1.2 (peak electricity price), W1 takes a lower value (e.g., 0.3); when k_price < 0.8 (low electricity price), W1 takes a higher value (e.g., 0.5).

[0061] Energy consumption weight f2 is an enhancing function; W2 takes a higher value (e.g., 0.5) during peak electricity price periods and a lower value (e.g., 0.2) during off-peak electricity price periods.

[0062] Stability weight W3: Or, it can be set separately according to similar rules.

[0063] Generate DO setpoint candidate set and control command prediction Generate a candidate set: centered on the current dynamic setpoint DO_set_ff, within the allowable process range. (e.g., Δ=0.5 mg / L), generate a set of candidate DO setpoints in step size δ (e.g., 0.1 mg / L). .

[0064] Predictive control command: For each DO_candidate_i, based on the current influent load, use the baseline DO-airflow relationship curve in the feedforward control module to predict the required initial fan frequency F_feedforward_i.

[0065] Predicting the Actual DO Trajectory: Based on the current state of the biological tank and influent conditions, a simplified first-order hysteresis process model is used to predict the possible value of the average DO in the biological tank within a future decision cycle after applying F_feedforward_i, DO_predicted_i. This model can be expressed as: , where τ is the pool time constant identified based on historical data, and DO_current is the current measured DO.

[0066] Multi-objective reward function calculation For each candidate solution i, calculate its comprehensive reward value R_i: Water quality bonus item R_quality_i: In the formula, the subscript j is synonymous with i in claim 4, both representing the serial number of the water quality index. Predicted effluent concentration: A simplified steady-state or quasi-steady-state effluent water quality prediction model is used.

[0067] For example, for the removal of ammonia nitrogen, a simplified formula based on microbial reaction kinetics can be used: Where k_nit is the nitrification rate coefficient and SRT is the sludge age (considered a known or slowly changing parameter). Similarly, predictive relationships for other indicators can be established.

[0068] Calculate the penalty: Where j represents different water quality indicators.

[0069] Energy reward item R_energy_i: Predicted wind turbine power consumption: , where f_power is the power-frequency characteristic function of the wind turbine.

[0070] Calculate the reward: .

[0071] Stability reward item R_stability_i: Calculate DO fluctuations: Where Δ_max is the maximum allowable single adjustment amount. This penalty for adjustments that deviate too much from the current target promotes smooth operation.

[0072] Overall Rewards: .

[0073] Optimal decision and instruction output Select the candidate DO setting that yields the highest overall reward value: .

[0074] Safety and smooth processing: make sure .

[0075] Rate limiting is applied to DO_set_optimal to prevent excessive jumps in set values ​​between adjacent cycles.

[0076] Output: DO_set_optimal is passed as the output of the economic optimization module to the feedforward-feedback composite control step (step S3) to replace or correct the original dynamic DO setting value, thereby affecting the final wind turbine control command.

[0077] To enable those skilled in the art to better understand and implement this method, an example is presented below: Assuming the current electricity price is peak (W1=0.3, W2=0.5, W3=0.2), DO_set_ff = 2.8 mg / L. The candidate set is {2.5, 2.6, 2.7, 2.8, 2.9, 3.0} mg / L.

[0078] For a candidate value of 2.7 mg / L: the predicted effluent ammonia nitrogen is 1.2 mg / L (meets the standard), the predicted fan power consumption is 45 kW, and the stability penalty is -0.1. The calculated values ​​are R_quality = -0, R_energy = -0.9, R_stability = -0.1, and the overall R = 0.3. 0 +0.5 (-0.9) + 0.2 (-0.1) = -0.47.

[0079] For a candidate value of 2.5 mg / L: the predicted effluent ammonia nitrogen is 1.5 mg / L (meets the standard), the predicted fan power consumption is 42 kW, and the stability penalty is -0.3. The calculated R = -0.36.

[0080] The comparison revealed that the R value of 2.5 mg / L (-0.36) was higher than that of 2.7 mg / L (-0.47). Therefore, the algorithm tends to select the lower 2.5 mg / L as the optimization setting during peak electricity prices in order to save energy.

[0081] Step S5: Execute the steps.

[0082] The intelligent control layer will send the final fan control commands, which are obtained through feedforward, feedback and economic optimization, to the fan frequency converter (VFD) and air conditioning valve in the execution layer via industrial real-time Ethernet, such as Profinet, Ethernet / IP or fieldbus.

[0083] The fan inverter adjusts the fan motor speed according to the received frequency or speed command, thereby changing the air supply.

[0084] The air conditioning valve adjusts its opening synchronously to optimize the characteristics of the air distribution pipeline.

[0085] The execution layer devices feed back the actual operating status, such as frequency, current, and valve opening, to the intelligent control layer, forming a closed loop.

[0086] Step S6, Fault Diagnosis and Self-Recovery Steps (in parallel with the control loop).

[0087] To improve system reliability, the following online fault handling mechanism is designed: Fault detection: For critical sensor data (such as DO value), calculate its anomaly score: Where X_measured is the actual measured value of the sensor; X_expected is the expected value calculated based on the process model, or the mean of the measured values ​​of adjacent sensors of the same type; σ is the standard deviation of the historical data of this parameter. When Anomaly_score continuously exceeds the set threshold (e.g., 3.0), the sensor is determined to be faulty or severely drifting.

[0088] Self-recovery mechanism: Once a fault is detected, the self-recovery mechanism is triggered immediately. Data replacement: If a backup sensor is configured, the data is switched to the backup sensor. If not, an estimation is performed using interpolation based on adjacent normal sensor data. For example, for a faulty DO sensor j, the data replacement is as follows: Control weight reduction: In the control algorithm, reduce the weight of the control loop in the area where the faulty sensor is located or temporarily put it into manual mode to avoid erroneous data causing the system to go out of control.

[0089] Alarm notification: Simultaneously send a maintenance alarm containing the location and type of the fault to the host computer monitoring system to prompt maintenance personnel to intervene.

[0090] Step S7: Cloud-Edge Collaborative Optimization Steps.

[0091] To continuously improve model adaptability, a cloud-edge collaborative architecture is adopted: Real-time control is achieved at the edge layer: The intelligent control layer located on-site, such as a PLC or edge gateway, is responsible for executing all real-time data acquisition, calculation and control tasks from S1 to S6 above, ensuring the immediacy and reliability of control.

[0092] Model training and optimization are performed in the cloud layer: The cloud server collects historical operating data such as water quality, load, DO, and energy consumption uploaded from the edge layer daily or weekly.

[0093] Model parameter update: The cloud server uses the exact same process as the dynamic DO setting value calculation model construction method described in step S2 to re-optimize the threshold array and weight array of the rule model. The specific steps are as follows: Regression Analysis: Using the latest uploaded historical data, a multiple linear regression analysis was performed to obtain the updated regression model coefficients. The specific expression is as follows: In the formula, the meaning of each parameter is consistent with the definition in step S2: DO_optimal is the predicted optimal dissolved oxygen value (mg / L). , where NH3_N is the influent ammonia nitrogen concentration (mg / L), COD_load is the influent COD load (kg / h), T is the water temperature (°C), TN is the influent total nitrogen concentration (mg / L), and Flow is the influent flow rate (m³ / h). to These are the regression coefficients for each variable.

[0094] For example, fitting based on new data might yield the following specific equation: The coefficient of determination R² = 0.90 for this equation indicates that the model has a high degree of interpretability.

[0095] Reverse parsing mapping: Based on the variable segmentation (determining the threshold) and interval weight calculation (determining the weight) methods disclosed in step S2, the above continuous regression model is transformed into a new threshold-weight rule table.

[0096] Specifically: Determine the threshold array: Analyze the distribution characteristics of each variable (NH3N, COD_load, T, TN) in the new dataset, and determine the new threshold boundary points according to the same or adjusted segmentation principle (such as maintaining four levels: low, medium, high, and extremely high).

[0097] Calculate the weight array: For each variable, substitute all sample points within each interval defined by its new threshold array into the updated regression equation, calculate the DO increment corresponding to each sample point (i.e., the value of the corresponding term of the variable in the regression equation), and then take the average of the DO increments of all samples within the interval as the new weight value of the interval.

[0098] Generate parameter package: Organize the obtained new threshold array and weight array into a rule table with the same format as Table 1 in Example 2, and use it as the optimized parameter package.

[0099] Step S8: Effect Evaluation and Parameter Adjustment To continuously ensure and improve system performance, this invention also includes an effect evaluation and parameter adjustment mechanism. The system monitors key effluent water quality indicators (such as NH3-N and TN) and the electricity consumption per ton of water in the aeration system in real time, and generates performance reports periodically (e.g., daily). The current performance data is compared and evaluated with historical benchmarks or set targets. If the evaluation finds room for performance optimization under specific operating conditions (such as high energy consumption or increased water quality fluctuations), the system can automatically or, with administrator confirmation, dynamically fine-tune control parameters such as the PID parameters (K_p, Ki, K_d) in the feedforward-feedback composite control step (S3) or the baseline weights (W1_base, W2_base) in the economic optimization step (S4), thereby ensuring that the system continuously maintains its optimal operating state.

[0100] Example 2: Parameter examples of applying the method of this invention to a specific dataset To demonstrate the specific application of the method of this invention, the following will use the operational data from our plant's on-site test from August to October 2025 as an example, applying the dynamic dissolved oxygen setpoint calculation model construction method described in step S2 to show a set of specific rule model parameters. These parameters have proven the effectiveness of this invention in practice.

[0101] Data Preparation and Regression Analysis: First, historical datasets containing influent NH3_N, COD concentrations, flow rates, water temperature, TN, and corresponding stable operating DO values ​​for the specified period were collected. Through multiple linear regression analysis, the regression equation was obtained as follows: The coefficient of determination for this equation is R² = 0.893.

[0102] Variable segmentation (threshold determination) and weight calculation: According to the method described in step S2, the historical data of each variable are analyzed to determine its threshold range, and the weight value corresponding to each range is calculated. The final parameter set is shown in Table 1.

[0103] Table 1: Calculation Parameters for Dynamic DO Setpoints (Based on Phase II Test Data).

[0104] Table 1: Calculation Parameters for Dynamic DO Setpoints (Based on Phase II Test Data) For example, when at a certain moment COD_load = 1200 kg / h, NH3_N = 35 mg / L, T = 26 °C, TN = 45 mg / L: COD_load ≥ 1000, corresponding to weight index 2, W_cod = 0.5 NH3_N ≥ 30, corresponding to weight index 2, W_nh = 0.5 For T ≥ 25, the corresponding weight index is 1, and W_temp = 0.1 TN ≥ 40, corresponding to weight index 2, W_tn = 0.3 DO setting value calculation and limit: The limit ranges were checked: MIN_DO = 1.5, MAX_DO = 4.0, and 3.4 mg / L was within the effective range. Therefore, the final DO_set = 3.4 mg / L. This value will take effect immediately as the feedforward control target.

[0105] Feedforward-feedback composite control loop The control loop primarily uses feedforward, with feedback as a secondary method for fine-tuning.

[0106] Feedforward channel: The controller has a built-in "DO-airflow" reference mapping table. Based on the calculated DO_set=3.4mg / L, the corresponding reference fan frequency FF_freq (e.g., 45 Hz) is obtained from the table and used as the initial control command.

[0107] Feedback Channel: The DO sensor array measures the average DO value in the aerobic tank in real time as DO_measured = 3.1 mg / L. The calculated deviation e(t) = 3.4 - 3.1 = 0.3 mg / L. This deviation is input to a PID controller resistant to integral saturation. After parameter tuning (e.g., K_p = 2.0, Ki = 0.05, K_d = 0.5), the PID outputs a frequency correction Δu = +1.5Hz.

[0108] Command synthesis: The final wind turbine frequency command F_final = FF_freq + Δu = 45 + 1.5 = 46.5Hz. This command is sent to the wind turbine inverter via the Profinet network.

[0109] Real-time decision-making in the economic optimization module The economic optimization module runs at a slightly lower frequency (e.g., every 5 minutes) to adjust the weights of the reward function.

[0110] Scenario A: Peak electricity price period. Assume the received real-time electricity price is 1.2 yuan / kWh, which is 150% of the daily average price of 0.8 yuan / kWh. The economic optimization module automatically adjusts the weights to: W_1=0.4 (water quality), W_2=0.5 (energy consumption), W_3=0.1 (stability). This means the control algorithm will tend to accept larger DO fluctuations (e.g., widening the PID control dead zone from ±0.2 mg / L to ±0.5 mg / L) to reduce frequent adjustments made by the wind turbine in pursuit of absolute stability, thereby saving electricity costs during peak electricity price periods.

[0111] Scenario B: Off-peak electricity price period. Real-time electricity price is 0.5 yuan / kWh. The module adjusts the weights to: W_1=0.5, W_2=0.2, W_3=0.3. At this time, the algorithm tightens control, pursuing more precise DO control and more stable aeration to provide the highest guarantee for water quality. Simultaneously, due to low energy consumption costs, the weight of energy conservation considerations is reduced. Example of a fault self-recovery mechanism. Suppose that the DO sensor (DO2) in the middle of the aerobic tank experiences a drift failure, with its reading abnormally low to 0.5 mg / L, while the readings of the adjacent first (DO1) and last (DO3) sensors are 2.8 mg / L and 3.0 mg / L, respectively. The expected value of the system is approximately 2.9 mg / L, and the historical standard deviation σ is 0.3.

[0112] Fault detection: The anomaly score is calculated as Anomaly_score = |0.5 - 2.9| / 0.3 = 8.0, which far exceeds the threshold of 3.0.

[0113] Self-recovery action: Data replacement: Immediately replace the fault value with the interpolation of adjacent sensors: DO2_corrected = (2.8 + 3.0) / 2 = 2.9 mg / L.

[0114] Weighting control: When averaging DO in the computation pool, the weight of DO2 is reduced from 1.0 to 0.2, mainly relying on DO1 and DO3.

[0115] Alarm: An alarm message “DO2 sensor suspected of malfunction, data redundancy strategy has been enabled” pops up on the host computer monitoring screen and is logged in the event log.

[0116] Example 3: The parameters of this invention also have self-learning capabilities, specifically achieved through cloud-edge collaboration.

[0117] Data Upload: Every 24 hours, the edge controller uploads the complete daily operating dataset, including influent parameters, calculated DO_set, actual DO value, effluent water quality, energy consumption, etc., and packages and encrypts it to the cloud server.

[0118] Cloud analytics: A parameter optimization program runs on a cloud server. It performs regression analysis on a dynamic DO setpoint model using the latest uploaded batch data. For example, it uses multiple linear regression to obtain the updated model coefficients. The cloud-based algorithm translates this continuous model into a new set of threshold-weight rules.

[0119] Parameter distribution and update: The cloud securely distributes the new rule table parameters to the edge controller in JSON format. The edge controller automatically receives and updates the local calculation parameters during the set maintenance window, such as 2:00 AM. After 8:00 AM the next day, the system will use the optimized rules to calculate the DO setting value, thereby continuously adapting to changes in water quality and microbial populations.

[0120] The end server runs a parameter optimization program, whose core algorithm is the same as the model building method described in step S2. It uses the latest uploaded batch data to re-perform regression analysis and rule parameter transformation, generating new threshold and weight arrays. This means that when this system is deployed to a new wastewater treatment plant, it is only necessary to collect the plant's historical operating data and obtain an initial parameter table suitable for the plant using the same method, without relying on the specific parameters in Example 2.

[0121] The system of this invention was implemented in a phase II trial at a wastewater treatment plant, and a comparative test was conducted for 30 days. The first 15 days used traditional PID control with a fixed DO setpoint, and the latter 15 days used the method of this invention. The performance indicators are compared below: Table 2: Comparison of performance with traditional PID control Example 4: To further enhance the adaptability of the dynamic DO setpoint model to short-term operating condition fluctuations, and as a supplement to cloud-based periodic batch optimization (see Example 3), this invention can also integrate a lightweight online adaptive fine-tuning mechanism for parameters into the computing core deployed in the edge intelligent control layer (such as a PLC). This mechanism is driven by real-time data from the feedforward-feedback composite control process to perform small-range, high-frequency calibration of the parameters of the dynamic DO setpoint model.

[0122] Mechanism triggering conditions The aforementioned online adaptive fine-tuning is not continuous but is triggered by specific conditions. The system monitors the effect of feedforward control in real time, specifically by comparing the deviation e_adapt between the dynamic DO setpoint (DO_set) and the actual DO measurement value (DO_stable) when the system reaches stability after feedback control. When the absolute value of e_adapt exceeds the preset fine-tuning trigger threshold Th_adapt, for example, 0.3 mg / L, for N consecutive control cycles (e.g., N=10, corresponding to approximately 5 minutes), the system determines that the current model has a persistent small deviation and then triggers a parameter adaptive fine-tuning process.

[0123] Fine-tuning algorithm and process This embodiment employs a parameter fine-tuning algorithm based on the gradient descent concept, with the core objective of reducing the aforementioned deviation e_adapt. The currently active weight arrays W_cod, W_nh, W_temp, and W_tn in the dynamic DO setpoint rule model (formula DO_set = BASE_DO + W_cod + W_nh + W_temp + W_tn) are considered as the parameter vector θ to be fine-tuned.

[0124] Define the performance evaluation function J(θ) as the square of the prediction bias, i.e.: Parameter updates are performed using gradient descent, and the update formula is as follows: in: θ_{old} and θ_{new} are the parameter vectors before and after fine-tuning, respectively.

[0125] η is the learning rate, a preset small positive number (e.g., 0.02) used to control the magnitude of a single adjustment and ensure smooth updates.

[0126] ∇J(θ_{old}) represents the gradient of the performance function J(θ) at θ_{old}. For regular models, due to their discreteness, the gradient component corresponding to each weight W_i can be approximated using the finite difference method. For example, for W_cod, the gradient component can be calculated as follows: Where δ is a small perturbation value (e.g., 0.01). This indicates the performance function value obtained by recalculating DO_set after only increasing W_cod by δ.

[0127] Protection and Constraint Logic To ensure the security and stability of online adaptive fine-tuning, the mechanism incorporates the following protection logic: Single-step magnitude limit: The absolute value of a single adjustment to any weight W_i shall not exceed the set limit Δ_max (e.g., 0.05) to prevent drastic changes in the model due to single data anomalies.

[0128] Overall range constraint: The fine-tuned parameters must ensure that the computed DO_set still satisfies the condition under any possible combination of inputs. Process constraints.

[0129] Reset mechanism: When the cloud server sends out a new parameter package that has been refitted and optimized (see Example 3), the edge control layer will automatically overwrite the existing parameters with the new parameters and clear all historical deviation data accumulated by the online fine-tuning mechanism, starting from a new starting point.

[0130] Function switch: This online fine-tuning function can be remotely enabled or disabled via host computer software, facilitating operation and maintenance management.

[0131] To facilitate better understanding by those skilled in the art, the specific workflow is as follows: Assuming the current COD_load is in the high range, the corresponding weight W_cod is 0.5. The system continuously monitors e_adapt, which averages +0.4 mg / L, meaning the model's predicted value is consistently high, exceeding the threshold Th_adapt = 0.3 mg / L, triggering fine-tuning.

[0132] Calculate current performance .

[0133] Assuming δ=0.01, calculate the performance after perturbation: If W_cod is adjusted to 0.51, the model's predicted DO_set will increase accordingly, potentially causing e_adapt to increase to 0.41. .

[0134] Approximate gradient: .

[0135] Update the weights (with η=0.02): .

[0136] Application range limitation: adjustment amount Therefore, the new weight is accepted.

[0137] Through this process, the model slightly lowered the DO setting value for the current high COD load condition, making it closer to the actual needs, thereby achieving online self-calibration of the parameters.

[0138] The intelligent aeration control method and system based on multi-source information fusion provided by this invention, through a complete technology chain of dynamic feedforward setting, feedback fine-tuning, economic optimization, fault self-recovery and cloud-edge collaborative evolution, significantly reduces the energy consumption of the aeration system, greatly improves the stability and compliance rate of effluent water quality, reduces the frequency of equipment operation, and further reduces operating costs by utilizing time-of-use electricity pricing policies, thus achieving synergistic optimization of energy saving, stability, quality improvement and cost reduction in the sewage treatment process.

[0139] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for energy-saving optimization control of a wastewater treatment aeration system based on multi-source information fusion, characterized in that, Includes the following steps: S1. Real-time data acquisition steps: acquire influent water quality parameters, influent flow rate parameters, environmental parameters inside the biological tank, and external electricity price signals; S2, Dynamic dissolved oxygen setpoint calculation step: Based on the influent water quality parameters and influent flow rate parameters, the dynamic dissolved oxygen setpoint is calculated in real time through a preset dynamic calculation model; S3, Feedforward-feedback composite control step: The dynamic dissolved oxygen setpoint is used as the feedforward control target to generate preliminary fan control commands. At the same time, based on the deviation between the actual dissolved oxygen measurement value in the environmental parameters of the biological tank and the dynamic dissolved oxygen setpoint, the preliminary fan control commands are finely adjusted to generate the final fan control commands. S4. Economic optimization step: Under the premise of ensuring that the effluent water quality meets the standards, dynamically adjust the control parameters in the feedforward-feedback composite control step. S5. Execution step: Send the optimized final fan control command to the fan frequency converter to adjust the aeration volume.

2. The energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion according to claim 1, characterized in that, The specific steps for calculating the dynamic dissolved oxygen setpoint include: Based on the influent COD concentration and influent flow rate, the influent COD load is calculated using the following formula: COD_load = COD Flow 0.001; Wherein, COD is the influent COD concentration, in mg / L; Flow is the influent flow rate, in m³ / h; COD_load is the COD load, in kg / h; Based on the COD_load and combined with the influent ammonia nitrogen concentration (NH3_N), water temperature (T), and influent total nitrogen concentration (TN), the dynamic dissolved oxygen setpoint (DO_set) is calculated using a threshold-weighted rule model. The calculation formula is as follows: DO_set = BASE_DO + W_cod + W_nh + W_temp + W_tn; Wherein, the influent ammonia nitrogen concentration NH3_N, water temperature T and influent total nitrogen concentration TN are all parameters collected in real time through step S1; Where BASE_DO is the preset base dissolved oxygen value; W_cod, W_nh, W_temp, and W_tn are the weight values ​​corresponding to COD load, ammonia nitrogen concentration, water temperature, and total nitrogen concentration, respectively. Each weight value is obtained from the preset weight array by comparing the real-time value of the corresponding parameter with the preset threshold array; The calculated DO_set is subjected to amplitude limiting to satisfy: MIN_DO ≤ DO_set ≤ MAX_DO; MIN_DO and MAX_DO are the preset lower and upper limits of dissolved oxygen settings.

3. The energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion according to claim 2, characterized in that, The weight values ​​W_cod, W_nh, W_temp, and W_tn are determined as follows: The actual value of the parameter X is compared with a set of increasing thresholds T1, T2, T3, and the weight index idx is determined according to the following formula: If X ≥ T3, then idx = 3 If X ≥ T2, then idx = 2 If X ≥ T1, then idx = 1 Otherwise, idx = 0 Based on the weight index idx, retrieve the weight value W = W[idx] from the corresponding weight arrays W0, W1, W2, W3.

4. The energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion according to claim 3, characterized in that, The economic optimization step is achieved by constructing and optimizing a multi-objective reward function R, the expression of which is: R = W1·R_quality + W2·R_energy + W3·R_stability; Among them, R_quality is the water quality reward item, R_energy is the energy consumption reward item, R_stability is the stability reward item, and W1, W2, and W3 are dynamically adjusted weight coefficients. The formula for calculating the water quality incentive R_quality is as follows: R_quality = - Σᵢ [λᵢ × max(0, C_{i,out} - C_{i,standard})² ]; Wherein, λᵢ is the weighting coefficient of the i-th water quality indicator, C_{i,out} is the effluent concentration of the i-th indicator, and C_{i,standard} is the emission standard limit of the i-th indicator. The formula for calculating the energy consumption bonus item R_energy is as follows: R_energy = - P_total / P_baseline; Where P_total is the real-time total power consumption of the aeration system, and P_baseline is the baseline power consumption; The formula for calculating the stability reward R_stability is as follows: R_stability = - Σⱼ |DO_j - DO_{target,j}|; Where DO_j is the actual dissolved oxygen value at the j-th monitoring point, and DO_{target,j} is the target dissolved oxygen value at the j-th monitoring point; The weighting coefficients W1, W2, and W3 are dynamically adjusted based on real-time electricity price signals. The weight of W2 is increased during peak electricity price periods, and the weights of W1 and W3 are increased during off-peak electricity price periods. The weighting coefficient W2 increases when the real-time electricity price is higher than the average electricity price by a set percentage, and the weighting coefficients W1 and / or W3 increase when the real-time electricity price is lower than the average electricity price by a set percentage.

5. The energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion according to claim 2, characterized in that, The parameters of the dynamic calculation model are updated online through a parameter adaptive mechanism; The parameter adaptation mechanism is triggered when the model's prediction error exceeds a set threshold, and the model parameter vector θ is updated according to the gradient descent method. The update formula is as follows: θ_{t+1} = θ_t - η × ∇J(θ_t); Where θ_t is the model parameter vector at time t, η is the learning rate, J(θ) is the performance evaluation function constructed with the prediction error, and ∇J(θ_t) is the gradient of the performance function at θ_t.

6. The energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion according to claim 1, characterized in that, It also includes fault diagnosis and self-recovery steps: The formula for calculating the anomaly score of sensor data is: Anomaly_score = |X_measured - X_expected| / σ; Where X_measured is the actual measured value of the sensor, X_expected is the expected value based on the process model or adjacent sensor data, and σ is the standard deviation of the historical data of this parameter; When the abnormal score exceeds the set threshold, it is determined to be a sensor malfunction, triggering the self-recovery mechanism; The self-recovery mechanism includes: activating a backup sensor, and / or estimating the fault point data through interpolation based on measurements from adjacent normal sensors, using the following estimation formula: X_fault = (X_adjacent1 + X_adjacent2) / 2; And reduce the weight of the faulty area in control decisions.

7. The energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion according to claim 1, characterized in that, Implemented using a cloud-edge collaborative architecture; The cloud server layer is responsible for periodically collecting historical operational data and refitting and optimizing the parameters in the dynamic calculation model based on regression analysis. The regression model is in the following form: DO_set= β0 + β1×NH3_N + β2×COD_load + β3×T + β4×TN + β5×Flow; Where β0 is the intercept, and β1 to β5 are the regression coefficients, which are obtained by least squares fitting. The edge computing layer is responsible for executing the steps of the method described in any one of claims 1 to 6 in real time, and receiving optimized model parameters from the cloud server layer.

8. The energy-saving optimization control method for wastewater treatment aeration systems based on multi-source information fusion according to claim 1, characterized in that, The feedback fine-tuning is achieved through a PID controller, with the input being the dissolved oxygen deviation e(t) = DO_set - DO_measured, and the output u(t) calculated using the following formula: u(t) = K_p e(t) + K_i ∫e(τ)dτ + K_d (de(t) / dt); Where K_p, Ki, and K_d are the proportional, integral, and differential coefficients, respectively.

9. The method according to claim 1, characterized in that, Following the execution steps, there are also effect evaluation and parameter adjustment steps: Real-time monitoring of effluent water quality indicators and system energy consumption; Based on the monitoring results, evaluate the performance of the current control strategy; Based on the performance evaluation results, the control parameters in the feedforward-feedback composite control step or the economic optimization step are dynamically adjusted.

10. A wastewater treatment aeration energy-saving optimization control system based on the method of any one of claims 1-9, characterized in that, include: The data perception layer includes: The influent monitoring unit is installed at the inlet of the biological tank to collect influent water quality parameters and influent flow parameters in real time. It includes at least an influent COD online analyzer, an influent ammonia nitrogen online analyzer, an influent total nitrogen online analyzer, and an influent flow meter. The process monitoring unit is set up inside the biological tank to collect environmental parameters in the biological tank in real time. It includes at least a water temperature sensor, a pH sensor, and multiple dissolved oxygen sensors distributed at the beginning, middle and end of the aerobic tank. The energy consumption and external data interface unit is used to collect wind turbine energy consumption data and receive real-time electricity price signals from the external power grid. The intelligent control layer is communicatively connected to the data sensing layer. The intelligent control layer includes a programmable logic controller (PLC) that stores a computer program. When executed by a processor, the computer program is used for: Based on the influent water quality parameters and influent flow rate parameters from the data sensing layer, the dynamic dissolved oxygen setpoint is calculated in real time through a preset dynamic calculation model. The initial fan control command is generated using the dynamic dissolved oxygen setpoint as the feedforward control target. At the same time, the initial fan control command is fine-tuned based on the deviation between the actual dissolved oxygen measurement value from the process monitoring unit and the dynamic dissolved oxygen setpoint, and the final fan control command is generated. Under the premise of ensuring that the effluent water quality meets the standards, the control strategy of the aeration system is adjusted according to the real-time electricity price signal from the energy consumption and external data interface unit. The execution layer, which is communicatively connected to the intelligent control layer, includes a fan frequency converter and / or an air regulating valve. It is used to receive the final fan control command issued by the intelligent control layer and adjust the fan speed and valve opening to control the aeration volume input to the biological tank.

Images