A sewage treatment intelligent control method and system based on a hybrid communication architecture

Abstract

The application discloses a sewage treatment intelligent control method and system based on a mixed communication architecture, and belongs to the technical field of sewage treatment. By constructing a multi-protocol layered mixed communication architecture, full-link data acquisition and instruction issuing between field sensors, PLCs, edge servers and execution layers are realized, and redundancy switching is guaranteed. An intelligent prediction control module with LSTM and GPC fusion is deployed on the edge server side. LSTM is used to realize multi-step advanced prediction of key effluent water quality. The prediction result is input into GPC for constraint rolling optimization solution to generate optimal control instructions and issue execution. Through the mixed design of communication protocols and the deep fusion of prediction control algorithms, the sewage treatment process realizes the collaborative optimization of control accuracy, system real-time performance and economy, significantly improves the robustness of control, and provides a complete solution for the intelligent upgrading and fine operation of sewage treatment plants.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment technology, and more specifically to a wastewater treatment intelligent control method and system based on a hybrid communication architecture. Background Technology

[0002] Water is an indispensable resource for human survival. With increasing human demand for water, water resources are becoming increasingly scarce. Therefore, wastewater recycling is crucial. Wastewater treatment plants, as the primary means of treating urban domestic sewage and industrial wastewater, directly determine the degree to which effluent quality meets standards through the effective control of their treatment processes. Traditional wastewater treatment processes rely on manual control, adjusting chemical dosages based on experience to find a balance between effluent quality indicators and chemical concentrations. However, this control method suffers from significant uncertainty and instability due to changes in the operating environment and operators. While classic control algorithms adjust control quantities promptly based on feedback errors, stable set parameters cannot respond promptly to fluctuations in the influent wastewater load.

[0003] In recent years, with the steady improvement of the level of intelligentization in industrial processes, the control of key water quality in wastewater treatment has evolved into adaptive closed-loop control that utilizes machine learning and intelligent algorithms to optimize the control process. However, the selection of neural network models based on machine learning is equally crucial. Simple neural networks offer fast computation and strong model interpretability, but they are prone to overfitting and instability. Deep neural networks can handle complex data features, but their computation is complex, their model interpretability is poor, and a single control process lacking timely prediction of key indicators can lead to feedback lag and weak system anti-interference capabilities.

[0004] Therefore, how to propose an intelligent control method and system for wastewater treatment based on a hybrid communication architecture, so as to achieve synergistic optimization of control accuracy, system real-time performance and economy in the wastewater treatment process and improve the robustness of control, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a smart control method and system for wastewater treatment based on a hybrid communication architecture, which aims to overcome the problems of delayed prediction of key parameters, isolation of control units, and poor data coordination in existing wastewater treatment control.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: On one hand, this invention discloses an intelligent control method for wastewater treatment based on a hybrid communication architecture, comprising the following steps: Monitoring data during the wastewater treatment process is collected by on-site sensors and transmitted to the area PLC; the monitoring data received by the area PLC is uploaded to the main PLC; the main PLC then uploads the received monitoring data to the edge server. The edge server preprocesses the monitoring data to obtain standardized time-series data; The edge server inputs standardized time-series data into the long short-term memory neural network and outputs a multi-step advance prediction sequence of key effluent water quality indicators for multiple future sampling periods. The edge server has a built-in intelligent module. Based on the multi-step advance prediction value sequence, the current process state and the target setpoint, the intelligent module uses a generalized predictive control algorithm to solve the constrained optimization problem in each control cycle to obtain the optimal control quantity sequence. The first control quantity in the optimal control quantity sequence is used as the control command for the current control cycle and the control command is sent to the main PLC. The main PLC sends control commands to the execution layer devices to control the sewage treatment process.

[0007] Preferably, the hybrid communication architecture includes: A data acquisition link is constructed between the field sensors and the area PLC using the Modbus RTU protocol. A data aggregation link is built between the area PLC and the main PLC using the PROFINET protocol; The data transmission link between the main PLC and the edge server is built using the OPC UA protocol; The main PLC and the execution layer devices use the OPC UA protocol as the main communication link, and add the Modbus TCP protocol as a redundant communication link to construct the control command sending link.

[0008] Preferably, when the main PLC sends out output control commands through the OPC UA main communication link, the heartbeat signal of the OPC UA link is monitored in real time. When the heartbeat signal is lost continuously for a preset number of times or the control deviation fed back by the actuator exceeds a preset threshold, the command is automatically switched to the Modbus TCP redundant communication link.

[0009] Preferably, the monitoring data is preprocessed to obtain standardized time-series data, including: First, outlier detection and removal are performed on the monitoring data. Then, time alignment is performed on the monitoring data after outlier removal. Finally, the time-aligned monitoring data is normalized to generate standardized time-series data.

[0010] Preferably, the long short-term memory neural network incorporates an attention mechanism to automatically identify and weight feature information of key historical moments that affect future water quality when processing time-series data.

[0011] Preferably, a generalized predictive control algorithm is used to solve the constrained optimization problem in each control cycle to obtain the optimal control quantity sequence, including: Using the multi-step forward prediction sequence output by the Long Short-Term Memory Neural Network, the current process state of the wastewater treatment process, and the target setpoint for effluent water quality as inputs, a finite-time domain optimization objective function is constructed with the objectives of minimizing tracking error and minimizing control increment changes. The optimization objective function is as follows: ; In the formula, The objective function value, For the present Predicting the future The key output values ​​of water quality indicators at any given time. These are the target values ​​for key effluent water quality indicators. To control the weighting coefficients, for Control increment at any time, For the minimum prediction time domain, For the maximum prediction time domain, To control the time domain.

[0012] Preferably, a smart control method for wastewater treatment based on a hybrid communication architecture further includes: Collect actual effluent water quality data of the wastewater treatment process after the execution layer equipment executes control commands; The actual effluent water quality data is compared with the predicted values ​​at the corresponding time points output by the long short-term memory neural network to calculate the prediction error. The prediction error is fed back to the generalized predictive control algorithm to correct the free response part of the prediction model inside the generalized predictive control algorithm online and adjust the weight allocation of the objective function. The system monitors the trend of prediction error in real time, and triggers online updates and retraining of the long short-term memory neural network when the prediction error continues to exceed the preset threshold.

[0013] On the other hand, the present invention also proposes an intelligent control system for wastewater treatment based on a hybrid communication architecture, comprising: The field layer includes field sensors used to collect monitoring data during the wastewater treatment process; The control layer, including the PLC, is used to receive monitoring data and control commands collected by the field layer and send the control commands to the execution layer; The edge computing layer, including edge servers, is used to receive monitoring data uploaded by the PLC; The intelligent prediction module, deployed at the edge computing layer, includes a long short-term memory neural network and a generalized predictive control algorithm, which is used to output control commands based on the received monitoring data. The execution layer, including dosing pumps and valves, is used to receive control commands and execute corresponding control actions. The application layer, including the human-computer interaction interface and system management platform, is used to display the operating status of the wastewater treatment process and provide remote monitoring and parameter configuration functions.

[0014] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a wastewater treatment intelligent control method and system based on a hybrid communication architecture. By constructing a multi-protocol layered communication network, it realizes efficient data acquisition, reliable transmission and redundant switching between field sensors, regional PLCs, main PLCs and edge servers and execution layer devices. At the same time, an intelligent predictive control module composed of a long short-term memory neural network and a generalized predictive control algorithm is deployed on the edge server side. LSTM is used to realize multi-step advance prediction of key effluent water quality indicators, and the prediction results are input into GPC for constrained rolling optimization solution to generate the optimal control command and send it to the execution layer. This invention achieves end-to-end coordination and redundancy assurance in data acquisition, aggregation, and command issuance through a hybrid communication architecture, improving the system's real-time performance and reliability. Deep integration with LSTM-GPC enables predictive control of water quality changes and precise optimization of reagent dosing, effectively reducing operating costs. Closed-loop feedback of prediction errors and online model updates enable the system to dynamically adapt to water quality fluctuations and external disturbances, enhancing control robustness and adaptability. This achieves synergistic optimization between control accuracy, system response, and economic operation, providing a complete solution for the intelligent and refined operation of wastewater treatment processes. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 A flowchart of the method provided by the present invention; Figure 2 This is a schematic diagram of the LSTM-GPC algorithm. Figure 3 This is a diagram of the overall system architecture. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] On the one hand, refer to Figures 1-2 This invention discloses an intelligent control method for wastewater treatment based on a hybrid communication architecture, comprising the following steps: 1. Collect monitoring data (influent chemical oxygen demand, ammonia nitrogen, flow rate, dissolved oxygen, pH, etc.) during the wastewater treatment process through on-site sensors, and transmit the monitoring data to the area PLC; upload the monitoring data received by the area PLC to the main PLC; the main PLC uploads the received monitoring data to the edge server.

[0019] (1) Each field sensor transmits real-time monitoring data to the corresponding area PLC via the Modbus RTU protocol to complete edge data acquisition.

[0020] (2) The regional PLC will transmit the field data collected from each location to the main PLC using the PROFINET industrial Ethernet communication protocol to realize the centralized aggregation of data from multiple regions.

[0021] (3) The main PLC uploads field data to the edge server through the OPC UA protocol. This step utilizes the semantic tag feature of OPC UA to enable the edge server to directly identify the monitoring point and parameter type corresponding to the data.

[0022] In this embodiment, the area PLC selected is the Siemens S7-1200 series PLC. This series of controllers supports the expansion of communication modules via RS485 interface to realize the master / slave function of Modbus RTU (Remote Terminal Unit) communication protocol. The S7-1200 series has a high cost-performance ratio and a broad market application base. Its programming and configuration are completed in the Siemens Totally Integrated Automation Platform, and it can achieve native and seamless engineering integration with the host PLC. In terms of cost and performance, it is positioned between the S7-200 SMART and the S7-1500 / ET200SP distributed I / O system, fully meeting the task requirements of periodic acquisition, preliminary processing, and uploading of field sensor data.

[0023] The main PLC is a Siemens S7-1500 series high-performance PLC. As the core node of the entire control system, the main PLC offers significantly improved performance compared to the S7-1200 series. In terms of processing speed, the processor adopts a multi-core architecture, enabling parallel processing of multi-region data aggregation, control command generation, and communication protocol conversion tasks. For scalability, it supports the integration of high-data acquisition modules and redundant communication modules, meeting future system expansion needs. Regarding reliability, it features hardware-level fault self-diagnosis capabilities, quickly locating communication or operational anomalies.

[0024] The communication network between the sensors and the area PLC uses the Modbus RTU communication protocol to establish a data transmission link. This link uses the RS485 bus standard for physical layer connection. The RS485 standard has differential signal transmission characteristics, which can effectively resist electromagnetic interference in industrial environments during long-distance transmission. Furthermore, the Modbus RTU protocol is a common open-source protocol in the industrial field, natively supported by most mainstream sensors in the field, eliminating the need for additional protocol conversion modules and significantly reducing the link construction cost. Through this link, the area PLC can collect real-time monitoring data such as temperature, pressure, and flow rate output from the sensors.

[0025] A master-slave communication link is established between the area PLCs and the master PLC using the PROFINET industrial Ethernet communication protocol. The master PLC is configured as a PROFINET master, and each area PLC is connected to the bus network as a slave. Since both the S7-1200 and S7-1500 PLCs are developed based on the TIA Portal software platform, the master station programming software can automatically identify all connected area slaves and directly map the I / O variables of each area PLC to the master station address space. The operator's read and write operations on the slave I / O variables are completely consistent with the operation of local I / O, eliminating the need to write additional protocol parsing or data conversion programs, greatly simplifying the system programming workload and improving communication response speed. Through this link, the area PLCs upload pre-screened sensor data to the master PLC at preset fixed time intervals.

[0026] 2. The edge server preprocesses the monitoring data to obtain standardized time-series data.

[0027] First, outlier detection and removal are performed on the monitoring data. Then, time alignment is performed on the monitoring data after outlier removal. Finally, the time-aligned monitoring data is normalized to generate standardized time-series data.

[0028] 3. The edge server inputs standardized time-series data into the long short-term memory neural network and outputs a multi-step advance prediction sequence of key effluent water quality indicators for multiple sampling periods in the future.

[0029] The Long Short-Term Memory Neural Network incorporates an attention mechanism to automatically identify and weight feature information of key historical moments that affect future water quality when processing time-series data.

[0030] The preprocessed standardized data is input into the LSTM network for model training. By learning the correlation between historical water quality data and effluent water quality, the model can accurately predict the effluent water quality of wastewater.

[0031] Using preprocessed multidimensional time-series data as input feature vectors, the model leverages the recurrent connection structure and gating mechanism within the LSTM to capture the complex nonlinear dynamic characteristics and long-term dependencies in the wastewater treatment process. The model output is a sequence of multi-step forward predictions of key effluent water quality indicators for the next N sampling periods.

[0032] 4. The edge server has a built-in intelligent module. Based on the multi-step advance prediction value sequence, the current process state and the target setpoint, the intelligent module uses a generalized predictive control algorithm to solve the constrained optimization problem in each control cycle to obtain the optimal control quantity sequence. The first control quantity in the optimal control quantity sequence is used as the control command for the current control cycle, and the control command is transmitted back to the main PLC through the OPC UA protocol.

[0033] This embodiment adds a Modbus TCP (Transmission Control Protocol) communication channel as a redundant network to improve the reliability of the communication link between the main PLC and the actuator. When the heartbeat signal of the OPC UA link between the main PLC and the actuator is lost three times consecutively, or the control deviation value fed back by the actuator exceeds a preset threshold, it automatically switches to the Modbus TCP redundant channel to ensure the continuous transmission of control commands and avoid the actuator from stopping or malfunctioning due to communication interruption.

[0034] The effluent water quality prediction data output by the LSTM is transmitted to the GPC network. Based on the deviation between the prediction results and the water quality standard threshold, the GPC network performs prediction calculations for control variables such as chemical dosage and stirring rate, forming a targeted control strategy. Specifically: Using the multi-step forward prediction sequence output by the Long Short-Term Memory Neural Network, the current process state of the wastewater treatment process, and the target setpoint for effluent water quality as inputs, a finite-time domain optimization objective function is constructed with the objectives of minimizing tracking error and minimizing control increment changes. The optimization objective function is as follows: ; In the formula, The objective function value, For the present Predicting the future The key output values ​​of water quality indicators at any given time. These are the target values ​​for key effluent water quality indicators. To control the weighting coefficients, for Control increment at any time, For the minimum prediction time domain, For the maximum prediction time domain, To control the time domain.

[0035] 5. The main PLC sends control commands to the execution layer equipment (mainly composed of key execution equipment such as dosing pumps, blowers, and valves) to control the sewage treatment process.

[0036] 6. Collect actual effluent water quality data of the sewage treatment process after the execution layer equipment executes the control command, compare the actual effluent water quality data with the predicted value at the corresponding time output by the long short-term memory neural network, and calculate the prediction error.

[0037] The prediction error is fed back to the generalized predictive control algorithm to correct the free response part of the prediction model inside the algorithm online and adjust the weight allocation of the objective function.

[0038] The optimal control quantity u(t) calculated by GPC is sent to the corresponding devices in the execution layer through a hybrid communication network. At the new sampling time, the system collects the actual effluent water quality y(t+1) and compares it with the corresponding predicted value of LSTM at the previous time. LSTM ( t +1| t The prediction error e(t+1) is obtained by comparing the two values. This error is fed back to the GPC to correct the free response part of its internal prediction model online, thereby enhancing the controller's robustness to model mismatch and unknown disturbances.

[0039] The system monitors the trend of prediction error in real time, and triggers online updates and retraining of the long short-term memory neural network when the prediction error continues to exceed the preset threshold.

[0040] This embodiment leverages the multi-step prediction capabilities of LSTM networks to anticipate water quality trends and proactively adjust control parameters before water quality deviates from standards, achieving predictive regulation and avoiding water quality exceedances caused by reactive treatment. GPC's precise optimization algorithm calculates the minimum reagent dosage and equipment energy consumption required to achieve the desired standard, avoiding redundant operation issues common in traditional control systems. The closed-loop feedback mechanism and rolling optimization strategy enable the system to automatically overcome the effects of unpredictable disturbances such as model errors, influent water quality fluctuations, and equipment aging, maintaining stable control performance even under complex operating conditions.

[0041] 7. The edge server pushes the entire data processing information to the client via the OPC UA protocol, providing users with an intuitive process monitoring interface.

[0042] 8. Users can view real-time data and control status through the client. When special working conditions occur, manual intervention commands can be issued. The commands are transmitted to the execution layer through the edge server and the main PLC to achieve optimized control through human-machine collaboration, ensuring the stability and safety of the process under abnormal or special working conditions.

[0043] On the other hand, this invention also proposes an intelligent control system for wastewater treatment based on a hybrid communication architecture, comprising: an application layer, an edge computing layer, intelligent modules, a control layer, an execution layer, and a field layer. The application layer is the human-computer interface and system management; the edge computing layer mainly includes a storage database and edge servers; the intelligent modules mainly include LSTM (Long Short-Term Memory) and GPC (Generalized Predictive Control); the control layer mainly includes PLC (Programmable Logic Controller); the execution layer includes control equipment such as dosing pumps and valves; and the field layer includes various water quality monitoring devices (field sensors). The overall system architecture diagram is shown below. Figure 3 As shown.

[0044] Specifically, this embodiment unifies data from different PLCs and sensors to a common time base through edge layer computation. The LSTM soft measurement module predicts key wastewater effluent quality based on historical window data (influent chemical oxygen demand, ammonia nitrogen, flow rate, dissolved oxygen, pH, etc.), while the GPC approximates the target setpoint based on the predicted effluent quality, simultaneously solving for the optimal control quantity in the future control time domain with the goal of minimizing control action changes. The optimized dynamic setpoint is then sent to the PLC through a reliable control network. The PLC quickly and accurately tracks this change. The system continuously performs rolling optimization based on new data generated after execution. Simultaneously, model performance monitoring can be established; when the prediction error continues to increase, online updates of the LSTM model are triggered.

[0045] The main functions of each module in this embodiment can be analyzed in detail as follows: (1) The field layer mainly consists of various online hardware monitoring devices such as sensors, flow meters, and turbidity meters. As the basic data acquisition and sensing system, the field layer collects water quality and quantity data as the raw input values ​​for LSTM soft measurement. The accuracy and real-time performance of the data directly determine the effectiveness of the upper-level intelligent predictive control system.

[0046] (2) The control layer consists of main PLC, area PLCs, and other devices. The control layer is the bridge connecting field devices and the upper-level intelligent predictive control system. The area PLC is mainly responsible for all sensor data, while the main PLC mainly receives optimization instructions from the edge layer to avoid excessive instruction delay.

[0047] (3) The execution layer mainly consists of key execution equipment such as dosing pumps, blowers, and valves. The execution layer mainly receives instructions from the control layer and precisely adjusts the dosage, aeration rate, and return ratio, thereby changing the biochemical reaction environment of the wastewater.

[0048] (4) The intelligent prediction module mainly consists of an LSTM data-driven prediction model and a GPC model-driven optimization control. The LSTM module uses historical and real-time data to perform multi-step forward predictions of key effluent water quality indicators, providing a future reference benchmark for the GPC. The GPC module, based on the LSTM prediction results, calculates the optimal control increment sequence that minimizes cost and ensures effluent compliance within a given timeframe by solving a constrained quadratic programming problem in each control cycle. The first value of the optimal sequence is then converted into a specific control command and sent to the main PLC. Therefore, the intelligent predictive control model, combining the data-driven prediction model with the model-driven optimization control strategy, enables online updating and correction prediction of changes in water quality characteristics.

[0049] (5) Edge computing layer, including database and edge server. The database is used to store historical and real-time data, while the edge server is the physical carrier for deploying intelligent algorithms. This module mainly processes the raw data uploaded from the field layer and provides computing resources and operating environment for LSTM and GPC.

[0050] (6) The application layer mainly consists of a mobile management platform, a mobile monitoring terminal, and a wastewater treatment control system. The wastewater treatment control system is used for human-machine interaction, and it can graphically display the operating status of the wastewater treatment process, water quality trends, equipment status, alarm information, etc. The mobile monitoring terminal supports remote access and mobile monitoring. The system management platform is mainly used for user permission management, parameter configuration, report generation, historical data query, and other operations.

[0051] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0052] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart control method for wastewater treatment based on a hybrid communication architecture, characterized in that, Includes the following steps: Monitoring data during the wastewater treatment process is collected by on-site sensors and transmitted to the area PLC; the monitoring data received by the area PLC is uploaded to the main PLC; the main PLC then uploads the received monitoring data to the edge server. The edge server preprocesses the monitoring data to obtain standardized time-series data; The edge server inputs standardized time-series data into the long short-term memory neural network and outputs a multi-step advance prediction sequence of key effluent water quality indicators for multiple future sampling periods. The edge server has a built-in intelligent module. Based on the multi-step advance prediction value sequence, the current process state and the target setpoint, the intelligent module uses a generalized predictive control algorithm to solve the constrained optimization problem in each control cycle to obtain the optimal control quantity sequence. The first control quantity in the optimal control quantity sequence is used as the control command for the current control cycle and the control command is sent to the main PLC. The main PLC sends control commands to the execution layer devices to control the sewage treatment process.

2. The intelligent control method for wastewater treatment based on a hybrid communication architecture according to claim 1, characterized in that, Hybrid communication architecture includes: A data acquisition link is constructed between the field sensors and the area PLC using the Modbus RTU protocol. A data aggregation link is built between the area PLC and the main PLC using the PROFINET protocol; The data transmission link between the main PLC and the edge server is built using the OPC UA protocol; The main PLC and the execution layer devices use the OPC UA protocol as the main communication link, and add the Modbus TCP protocol as a redundant communication link to construct the control command sending link.

3. The intelligent control method for wastewater treatment based on a hybrid communication architecture according to claim 2, characterized in that, When the main PLC sends out control commands through the OPC UA main communication link, it monitors the heartbeat signal of the OPC UA link in real time. When the heartbeat signal is lost for a preset number of consecutive times or the control deviation fed back by the actuator exceeds a preset threshold, it automatically switches to the Modbus TCP redundant communication link to send commands.

4. The intelligent control method for wastewater treatment based on a hybrid communication architecture according to claim 1, characterized in that, The monitoring data is preprocessed to obtain standardized time-series data, including: First, outlier detection and removal are performed on the monitoring data. Then, time alignment is performed on the monitoring data after outlier removal. Finally, the time-aligned monitoring data is normalized to generate standardized time-series data.

5. The intelligent control method for wastewater treatment based on a hybrid communication architecture according to claim 1, characterized in that, Long Short-Term Memory (LSTM) neural networks introduce an attention mechanism to automatically identify and weight feature information of key historical moments that affect future water quality when processing time-series data.

6. The intelligent control method for wastewater treatment based on a hybrid communication architecture according to claim 1, characterized in that, A generalized predictive control algorithm is used to solve the constrained optimization problem in each control cycle, yielding the optimal control quantity sequence, including: Using the multi-step forward prediction sequence output by the Long Short-Term Memory Neural Network, the current process state of the wastewater treatment process, and the target setpoint for effluent water quality as inputs, a finite-time domain optimization objective function is constructed with the objectives of minimizing tracking error and minimizing control increment changes. The optimization objective function is as follows: ； In the formula, The objective function value, For the present Predicting the future The key output values ​​of the effluent water quality indicators at any given time, These are the target values ​​for key effluent water quality indicators. To control the weighting coefficients, for Control increment at any time, For the minimum prediction time domain, For the maximum prediction time domain, To control the time domain.

7. The intelligent control method for wastewater treatment based on a hybrid communication architecture according to claim 1, characterized in that, Also includes: Collect actual effluent water quality data of the wastewater treatment process after the execution layer equipment executes control commands; The actual effluent water quality data is compared with the predicted values ​​at the corresponding time points output by the long short-term memory neural network to calculate the prediction error. The prediction error is fed back to the generalized predictive control algorithm to correct the free response part of the prediction model inside the generalized predictive control algorithm online and adjust the weight allocation of the objective function. The system monitors the trend of prediction error in real time, and triggers online updates and retraining of the long short-term memory neural network when the prediction error continues to exceed the preset threshold.

8. A smart control system for wastewater treatment based on a hybrid communication architecture, characterized in that, include: The field layer includes field sensors used to collect monitoring data during the wastewater treatment process; The control layer, including the PLC, is used to receive monitoring data and control commands collected by the field layer and send the control commands to the execution layer; The edge computing layer, including edge servers, is used to receive monitoring data uploaded by the PLC; The intelligent prediction module, deployed at the edge computing layer, includes a long short-term memory neural network and a generalized predictive control algorithm, which is used to output control commands based on the received monitoring data. The execution layer, including dosing pumps and valves, is used to receive control commands and execute corresponding control actions. The application layer, including the human-computer interaction interface and system management platform, is used to display the operating status of the wastewater treatment process and provide remote monitoring and parameter configuration functions.

Images