Sewage treatment process control system and method based on neural network

Abstract

The invention discloses a sewage treatment process control system and method based on a neural network, and relates to the technical field of sewage treatment.The system can collect and preprocess input and output data of a sewage treatment device in real time through a data collection module, and a neural network model of the sewage treatment process is established through a model establishment module; a neural network structure and parameters are optimized through the model optimization module, output data of the sewage treatment device in the sewage treatment process are predicted through the data prediction module, and a control strategy is designed through the device control module so as to control various control variables of the sewage treatment device; intelligent, automatic and optimal control of the sewage treatment process is realized, the sewage treatment efficiency and quality are improved, and the sewage treatment cost and risk are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of sewage treatment, and in particular relates to a sewage treatment process control system and method based on a neural network. Background Art

[0002] Wastewater treatment refers to the process of treating wastewater through physical, chemical, and biological methods to achieve specified water quality standards for reuse or discharge. Wastewater treatment is a crucial measure for protecting water resources, improving the water environment, and promoting a circular economy, and is therefore crucial for achieving sustainable development. To effectively control and optimize the wastewater treatment process, a comprehensive neural network-based wastewater treatment process control system is required to enable real-time data collection, intelligent modeling, precise prediction, effective control, and fault detection of wastewater treatment plant input and output data.

[0003] Currently, my country has established a certain scale of sewage treatment process control systems in some sewage treatment plants. However, these systems have some problems and shortcomings, mainly manifested in the following aspects: 1. The sewage treatment process control system cannot accurately describe and predict the sewage treatment process. 2. The sewage treatment process control system cannot achieve comprehensive and detailed analysis of the sewage treatment process. 3. The sewage treatment process control system cannot effectively optimize and improve the sewage treatment process.

[0004] Accurately controlling the sewage treatment process and collecting sewage treatment data in real time to continuously optimize sewage treatment efficiency are important means to ensure the stability and efficiency of the sewage treatment process. Summary of the Invention

[0005] To address the above issues, the present invention provides a sewage treatment process control system based on a neural network. This system can achieve efficient, stable, and intelligent control of the sewage treatment process, improve the quality and efficiency of sewage treatment, and reduce operating costs and environmental impact. The system includes the following modules: A data acquisition module is used to collect input and output data of the sewage treatment device, including the flow rate, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content of the sewage, and pre-process the data; Model building module, used to build a neural network model of the sewage treatment process based on the pre-processed data; Model optimization module, which is used to optimize the neural network structure and parameters using particle swarm optimization to obtain the optimized neural network model; A data prediction module is used to predict the output data of the sewage treatment device during the sewage treatment process based on the optimized neural network model; The device control module is used to design a control strategy based on the prediction results of the data prediction module; The sewage treatment device has a plurality of actuators, and is used to adjust the control variables of each actuator according to a control strategy.

[0006] The beneficial effect of this implementation scheme is that, by utilizing the powerful nonlinear fitting ability and adaptive learning ability of the neural network, the dynamic characteristics and complex laws of the sewage treatment process can be accurately described and predicted, thereby achieving optimized control of the sewage treatment process, improving the water quality of the effluent and the efficiency of sewage treatment, and reducing energy consumption and emissions.

[0007] In the best implementation, the system solves the technical problem of how to effectively obtain relevant data of the sewage treatment process and perform necessary processing on the data in order to facilitate the establishment and optimization of the neural network model.

[0008] In a preferred embodiment, the data acquisition module of the system specifically includes: Sensors for measuring input and output data from wastewater treatment plants; A data converter, used to convert the data measured by the sensor into a digital signal and send it to the data processor; The data processor is used to preprocess the data sent by the data converter, including filtering, normalization and feature extraction, to obtain preprocessed data.

[0009] In the best implementation, the data acquisition module of this system can effectively obtain relevant data of the sewage treatment process and perform necessary processing on the data to facilitate the establishment and optimization of the neural network model.

[0010] Under optimal implementation conditions, this system solves the technical problem of how to effectively model and predict the dynamic characteristics and complex laws of the sewage treatment process.

[0011] In a preferred embodiment, the neural network model of this system is a recurrent neural network based on a long short-term memory network, the number of its input layer nodes is the dimension of the input data of the sewage treatment device, the number of its output layer nodes is the dimension of the output data of the sewage treatment device, the number of its hidden layer nodes is the average of the number of input layer nodes and the number of output layer nodes, and its activation function is a hyperbolic tangent function.

[0012] Under optimal implementation conditions, the neural network model of this system can fully utilize the memory capacity of the long short-term memory network and the timing capacity of the recurrent neural network to effectively model and predict the dynamic characteristics and complex laws of the sewage treatment process.

[0013] Under optimal implementation conditions, this system solves the technical problem of how to optimize the effluent quality and sewage treatment efficiency.

[0014] In a preferred embodiment, the device control module of the system specifically includes: The target value setting submodule is used to set the target values of the outlet water flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content; The deviation acquisition submodule is used to obtain the predicted values of the outlet water flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content based on the prediction results of the data prediction module, compare the predicted values with the target values, and obtain the deviation and deviation change rate of each of the above output variables; The fuzzy reasoning submodule is used to fuzzify, reason and defuzzify the deviation and the deviation change rate using the fuzzy reasoning system to obtain the opening of the water inlet valve, the opening of the water outlet valve, the speed of the aerator and the adjustment amount of the sludge return ratio; The device adjustment submodule uses the adjustment amount as a control signal and sends it to various actuators of the sewage treatment device to adjust various control variables of the sewage treatment device.

[0015] In a preferred implementation, the device control module of this system can design a reasonable control strategy based on the prediction results of the neural network model to control various control variables of the sewage treatment device to optimize the water quality of the effluent and the efficiency of sewage treatment.

[0016] Under the best implementation conditions, the system solves the technical problem of how to accurately adjust various control variables of the sewage treatment device to achieve optimal control of the sewage treatment process.

[0017] In a preferred embodiment, the adjustment amounts of various control variables are calculated using the following formula: , in is the adjustment amount of each control variable, including the adjustment amount of the water inlet valve opening, the adjustment amount of the water outlet valve opening, the adjustment amount of the aerator speed and the adjustment amount of the sludge return ratio. is the deviation of each output variable, Refers to deviation The differential of , which represents the rate of change of the deviation, 、 and are the proportional, integral and differential coefficients respectively.

[0018] In a preferred embodiment, the device regulation submodule of the system can utilize a proportional-integral-derivative (PID) controller to precisely regulate various control variables, enabling it to quickly track the prediction results of the neural network model, thereby achieving optimized control of the sewage treatment process.

[0019] The present invention also provides a sewage treatment process control method based on a neural network, which can achieve efficient, stable and intelligent control of the sewage treatment process, improve the quality and efficiency of sewage treatment, and reduce operating costs and environmental impact. The method includes the following steps: Collecting input and output data of the sewage treatment plant, including sewage flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content, and preprocessing the data; Based on the pre-processed data, a neural network model of the sewage treatment process is established; The particle swarm algorithm is used to optimize the neural network structure and parameters to obtain the optimized neural network model; According to the optimized neural network model, the output data of the sewage treatment device during the sewage treatment process is predicted; According to the prediction results of the data prediction module, a control strategy is designed to control the control variables of each actuator in the sewage treatment device.

[0020] The beneficial effect of this implementation scheme is that, by utilizing the powerful nonlinear fitting ability and adaptive learning ability of the neural network, the dynamic characteristics and complex laws of the sewage treatment process can be accurately described and predicted, thereby achieving optimized control of the sewage treatment process, improving the water quality of the effluent and the efficiency of sewage treatment, and reducing energy consumption and emissions.

[0021] Under optimal implementation conditions, this system solves the technical problem of how to improve the prediction accuracy and generalization ability of the neural network model, thereby improving the control effect of the sewage treatment process.

[0022] In a preferred embodiment, the optimization process of the neural network model of this embodiment specifically includes the following steps: Initialize the parameters of the particle swarm algorithm, including the number of particles, maximum number of iterations, inertia weight, learning factor, speed limit and position limit; A set of particles is randomly generated, each particle represents the structure and parameters of the neural network, including the number of input layer nodes, the number of hidden layer nodes, the number of output layer nodes, connection weights and biases; Calculate the fitness value of each particle, that is, the prediction accuracy of the neural network model, using the input data of the sewage treatment plant as the input of the neural network, and the output data of the sewage treatment plant as the expected output of the neural network. Calculate the mean square error between the output of the neural network and the expected output, and use it as the inverse of the fitness value; Update the individual optimal position and global optimal position of each particle, that is, record the position corresponding to the highest fitness value achieved by each particle in the historical iteration, and the position corresponding to the highest fitness value achieved by the entire particle group in the historical iteration; Update the speed and position of each particle. Use the inertia weight, learning factor, individual optimal position and global optimal position, and follow the update formula of the particle swarm algorithm to adjust the speed and position of each particle so that it searches in a more optimal direction. Determine whether the termination condition is met, that is, whether the maximum number of iterations is reached or the fitness value reaches the preset threshold. If so, output the neural network structure and parameters corresponding to the global optimal position as the optimized neural network model; if not, return to the third step and continue iteration.

[0023] Under optimal implementation conditions, the optimization process of the neural network model of this method can improve the prediction accuracy and generalization ability of the neural network model, so that it can better adapt to the changes and uncertainties of the sewage treatment process, thereby improving the control effect of the sewage treatment process. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Schematic diagram of the system structure in Example 1 of the present invention; Figure 2 Schematic diagram of the system structure in Example 2 of the present invention; Figure 3 Schematic diagram of the system structure in Example 3 of the present invention; Figure 4 This is a flow chart of the method in Example 4 of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solution, the technical solution is described in detail below in conjunction with the embodiments. The description in this section is only exemplary and explanatory and should not have any limiting effect on the scope of protection of the present invention.

[0026] Example 1: This embodiment provides a sewage treatment process control system based on a neural network, and its structural diagram is shown in FIG. Figure 1 As shown, it includes the following modules: The data acquisition module is used to collect input and output data of the sewage treatment device, including the flow rate, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content of the sewage, and pre-process the data. The function of this module is to obtain input and output data from the sewage treatment device. These data are important information of the sewage treatment process and reflect the characteristics of the sewage and the treatment effect. The module also pre-processes the data, including filtering, normalization and feature extraction, to remove noise and outliers in the data and improve the availability and effectiveness of the data.

[0027] The model building module is used to establish a neural network model of the sewage treatment process based on the pre-processed data. The function of this module is to establish a neural network model based on the pre-processed data. This model is an artificial intelligence technology that can simulate the dynamic changes of the sewage treatment process by learning the patterns and relationships in the data, providing a basis for data prediction and device control.

[0028] The model optimization module is used to optimize the neural network structure and parameters using the particle swarm algorithm to obtain the optimized neural network model. The function of this module is to optimize the neural network model using the particle swarm algorithm. The particle swarm algorithm is a heuristic optimization algorithm that can find the optimal neural network structure and parameters by simulating group behavior in nature, so that the neural network model can better fit the data, improve the accuracy and generalization ability of the model, and enable it to better adapt to the characteristics and changes of the sewage treatment process.

[0029] The data prediction module is used to predict the output data of the sewage treatment device during the sewage treatment process based on the optimized neural network model; the function of this module is to predict the output data of the sewage treatment device during the sewage treatment process based on the optimized neural network model. These output data include the flow rate, temperature, pH value, dissolved oxygen concentration and nitrogen and phosphorus content of the effluent, which are important indicators for evaluating the sewage treatment effect and are also an important basis for controlling the sewage treatment device; this module can predict future output data based on the current input data to provide a reference for device control.

[0030] The device control module is used to design a control strategy based on the prediction results of the data prediction module. The function of this module is to design a control strategy based on the prediction results of the data prediction module to control various control variables of the sewage treatment device. These control variables include the opening of the water inlet valve, the opening of the water outlet valve, the speed of the aerator and the sludge return ratio. They are important factors affecting the sewage treatment process and are also important means of regulating the sewage treatment effect. This module can calculate the appropriate control signal based on the deviation between the preset target value and the predicted value using a fuzzy inference system and a proportional-integral-differential controller, and send it to each actuator of the sewage treatment device, so that the sewage treatment device can operate according to the expected target and achieve the optimal sewage treatment effect. At the same time, it can also avoid or reduce failures and risks in the sewage treatment process.

[0031] The sewage treatment device has a plurality of actuators, and is used to adjust the control variables of each actuator according to a control strategy.

[0032] This embodiment adopts a sewage treatment process control system based on a neural network, which can realize comprehensive, automated and intelligent control and management of the sewage treatment process, improve the efficiency and effect of sewage treatment, and reduce the cost and risk of sewage treatment.

[0033] Example 2: This embodiment provides a sewage treatment process control system based on a neural network. Based on Example 1, this embodiment specifically implements a data acquisition module and a neural network model.

[0034] like Figure 2 As shown in the figure, the data acquisition module consists of the following three parts: Sensors are used to measure the input and output data of sewage treatment plants, including the flow rate, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content of sewage. These data are important indicators for evaluating the effectiveness of sewage treatment and are also an important basis for controlling sewage treatment plants. Sensors include flow meters, thermometers, pH meters, dissolved oxygen meters, total nitrogen and total phosphorus analyzers, etc. They can be installed at the water inlet, outlet, or other key locations of sewage treatment plants to monitor the status and changes of sewage in real time. For example, a flow meter can measure the flow rate of sewage, a thermometer can measure the temperature of sewage, a pH meter can measure the pH value of sewage, a dissolved oxygen meter can measure the dissolved oxygen concentration in sewage, and a total nitrogen and total phosphorus analyzer can measure the nitrogen and phosphorus content in sewage.

[0035] The data converter converts the sensor's measured data into a digital signal and sends it to the data processor. Data converters include analog-to-digital converters, wireless transmitters, and data acquisition cards. They convert the sensor's analog or wireless signals into digital signals and transmit the data to the data processor via wired or wireless means for storage and analysis. For example, an analog-to-digital converter converts the sensor's analog signal into a digital signal, while a wireless transmitter sends the sensor's wireless signal to a data acquisition card. The data acquisition card then collects the data and stores it in a computer for use by the data processor.

[0036] The data processor is used to preprocess the data sent by the data converter, including filtering, normalization, and feature extraction, to obtain preprocessed data. The data processor includes data filters, data normalizers, and data feature extractors, which perform different processing on the data to improve its quality and effectiveness. For example, the data filter can filter the data to remove noise and interference, improving the signal-to-noise ratio and stability. The data normalizer can normalize the data to eliminate dimensional and scale differences in the data, improving the comparability and consistency of the data. The data feature extractor can extract features from the data to extract effective information and features, reducing the data's dimensionality and redundancy.

[0037] Furthermore, the neural network model is a recurrent neural network based on a long short-term memory network. The number of nodes in its input layer is the dimension of the input data of the sewage treatment device, the number of nodes in its output layer is the dimension of the output data of the sewage treatment device, the number of nodes in its hidden layer is the average of the number of nodes in the input layer and the number of nodes in the output layer, and its activation function is a hyperbolic tangent function. A specific implementation of the neural network model enables the system to effectively model and predict the sewage treatment process, improving the system's performance and accuracy.

[0038] Example 3: This embodiment provides a sewage treatment process control system based on a neural network. Based on Example 1, this embodiment specifically implements the device control module to achieve regulation and optimization of various control variables of the sewage treatment device.

[0039] like Figure 3 As shown, the device control module specifically includes the following four submodules: The target value setting submodule is used to set the target values of the outlet water flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content; The deviation acquisition submodule is used to obtain the predicted values of the effluent flow, temperature, pH value, dissolved oxygen concentration and nitrogen and phosphorus content based on the prediction results of the data prediction module, compare the predicted values with the target values, and obtain the deviations and deviation change rates of the above-mentioned output variables; this part calculates the deviations and deviation change rates of each output variable based on the prediction results of the data prediction module and the target values of the target value setting submodule to reflect the actual situation and control requirements of the sewage treatment process.

[0040] The fuzzy reasoning submodule is used to fuzzify, reason, and defuzzify the deviation and the rate of change of the deviation using the fuzzy reasoning system to obtain the opening of the water inlet valve, the opening of the water outlet valve, the speed of the aerator, and the adjustment amount of the sludge return ratio. This module specifically includes the following three steps: Fuzzification, this step converts the deviation and the rate of change of the deviation into fuzzy values, i.e., the degree of membership in a fuzzy set, to facilitate fuzzy reasoning. There are various fuzzification methods, such as triangular fuzzification, trapezoidal fuzzification, and Gaussian fuzzification, which can map the deviation and the rate of change of the deviation into fuzzy values between 0 and 1 based on the shape and characteristics of different fuzzy sets.

[0041] Inference: This step is used to perform fuzzy reasoning based on the fuzzified deviation and deviation change rate, as well as pre-set fuzzy rules, to obtain the fuzzy values of the adjustment amounts of each control variable. There are many methods of fuzzy reasoning, including maximum-minimum reasoning, maximum product reasoning, etc. According to different fuzzy operations and fuzzy implication principles, the fuzzy rules are reasoned to obtain the fuzzy values of the adjustment amounts of each control variable. For example, if the fuzzy rule is "If the deviation is large and the deviation change rate is positive, then the adjustment amount of the water inlet valve opening is increased", the maximum-minimum reasoning method can be used to perform a minimum operation on the fuzzy values of the deviation and the deviation change rate to obtain the strength of the rule. Then, the strength of the rule and the fuzzy value of the adjustment amount of the water inlet valve opening can be maximized to obtain the fuzzy value of the adjustment amount of the water inlet valve opening.

[0042] Defuzzification, a step used to convert the fuzzy values of the adjustment variables of each control variable into actual values—percentages between 0 and 100—to facilitate device adjustment. There are various defuzzification methods, including the centroid method, the average maximum method, and the median method. These methods convert the fuzzy values of the adjustment variables of each control variable into actual values based on different calculation and evaluation criteria.

[0043] The device adjustment submodule transmits the adjustment value as a control signal to the various actuators of the sewage treatment plant to adjust various control variables of the sewage treatment plant. These control variables include the opening of the water inlet valve, the opening of the water outlet valve, the speed of the aerator, and the sludge return ratio. An actuator is a device that receives a control signal and changes the control variable based on the control signal, such as an electric valve or motor. A control signal is an instruction to adjust the control variable, such as the opening or speed.

[0044] In this embodiment, a commonly used industrial control method, namely PID controller, is used to calculate the value of the control variable. The principle of PID controller is to comprehensively calculate the value of the control variable based on the current deviation, the accumulated deviation in the past and the predicted deviation in the future to achieve fast response, stability and accuracy of the control system. The three parameters of PID controller, namely the proportional coefficient , integral coefficient and differential coefficients , all need to be optimized and adjusted according to the characteristics and requirements of the control system.

[0045] Specifically, the adjustment amounts of the above-mentioned control variables are calculated using the following formula: , in is the adjustment amount of each control variable, including the adjustment amount of the water inlet valve opening, the adjustment amount of the water outlet valve opening, the adjustment amount of the aerator speed and the adjustment amount of the sludge return ratio. is the deviation of each output variable, is the value of the past cumulative deviation, Refers to deviation The differential of , which represents the rate of change of the deviation, 、 and are the three parameters of the PID controller.

[0046] The value of the adjustment is composed of three parts: Proportional control , indicating that the adjustment amount is proportional to the current deviation, which can make the control system quickly approach the target value; Integral control term , indicating that the adjustment amount is proportional to the past accumulated deviation and can eliminate the steady-state error; Derivative control term , indicating that the adjustment amount is proportional to the rate of change of the deviation, which can predict the trend of the control system, suppress overshoot or oscillation, and improve the stability of the control system; By adding these three parts together and comprehensively utilizing various information of the control system, the optimized control of the control system can be achieved.

[0047] Example 4: This embodiment provides a sewage treatment process control method based on neural network, and its flow chart is as follows: Figure 4 As shown, the following steps are included: S1, collects input and output data of the sewage treatment plant, including sewage flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content, and preprocesses the data. The purpose of preprocessing is to remove noise, outliers, missing values, etc. in the data to improve the quality and usability of the data. There are many preprocessing methods, such as smoothing, filtering, interpolation, normalization, etc., which can be selected and combined according to the characteristics and needs of the data.

[0048] S2. Based on the pre-processed data, a neural network model of the sewage treatment process is established. A neural network model is a nonlinear dynamic system composed of multiple neurons. It can model and predict the sewage treatment process by learning the patterns and relationships in the data. There are many types of neural network models, such as feedforward neural networks, feedback neural networks, recursive neural networks, etc., which can be selected and designed according to the characteristics and needs of the sewage treatment process.

[0049] S3, using the particle swarm algorithm to optimize the neural network structure and parameters to obtain the optimized neural network model. The particle swarm algorithm is a global optimization algorithm based on swarm intelligence. It can find the optimal solution for the neural network structure and parameters by simulating the self-organizing behavior of bird flocks or fish schools, thereby improving the prediction accuracy and generalization ability of the neural network model. The optimization process of the particle swarm algorithm specifically includes the following steps: S31, initializing the parameters of the particle swarm algorithm, including the number of particles, the maximum number of iterations, the inertia weight, the learning factor, the speed limit, and the position limit; these parameters can be set based on experience or experiments, and can also be adjusted according to the dynamic situation of the optimization process; S32, randomly generating a set of particles, each particle representing the structure and parameters of the neural network, including the number of input layer nodes, the number of hidden layer nodes, the number of output layer nodes, connection weights and biases; these particles can be randomly generated within a certain range, or generated based on some prior knowledge or heuristic rules; S33, calculating the fitness value of each particle, that is, the prediction accuracy of the neural network model, using the input data of the sewage treatment plant as the input of the neural network, using the output data of the sewage treatment plant as the expected output of the neural network, calculating the mean square error between the output of the neural network and the expected output, and taking it as the inverse of the fitness value; the larger the fitness value, the higher the prediction accuracy of the neural network model, and the closer it is to the optimal solution; S34, update the individual optimal position and global optimal position of each particle, that is, record the position corresponding to the highest fitness value achieved by each particle in the historical iteration, and the position corresponding to the highest fitness value achieved by the entire particle swarm in the historical iteration; the individual optimal position and the global optimal position can be used as the search direction and target of the particle to guide the particle to move to a more optimal area; S35, updates the speed and position of each particle. Using the inertia weight, learning factor, individual optimal position, and global optimal position, according to the particle swarm algorithm update formula, the speed and position of each particle are adjusted to search in a more optimal direction. The speed and position update can maintain the diversity and exploratory nature of the particles, while also enhancing the convergence and development of the particles. S36, determine whether the termination condition is met, that is, whether the maximum number of iterations is reached or the fitness value reaches a preset threshold. If so, output the neural network structure and parameters corresponding to the global optimal position as the optimized neural network model; if not, return to S33 and continue iterating.

[0050] S4, based on the optimized neural network model, predict the output data of the sewage treatment device during the sewage treatment process; use the input data of the sewage treatment device collected in real time as the input of the optimized neural network model to obtain the output of the neural network model, that is, the predicted value of the output data of the sewage treatment device; these predicted values can be used as the basis of the control strategy to control various control variables of the sewage treatment device.

[0051] S5. Based on the prediction results of the data prediction module, a control strategy is designed to control the control variables of each actuator in the sewage treatment device. The purpose of the control strategy is to make the output data of the sewage treatment device reach the expected target while taking into account the operating cost and safety of the sewage treatment device. There are many control strategy methods, such as fuzzy control, adaptive control, intelligent control, etc., which can be selected and designed according to the characteristics and needs of the sewage treatment process.

[0052] This embodiment provides a sewage treatment process control method based on a neural network. The method improves the efficiency and quality of the sewage treatment process through the following aspects: By collecting and pre-processing the input and output data of the sewage treatment device, real-time information of the sewage treatment process can be obtained, providing data support for the establishment and optimization of the neural network model; By establishing and optimizing the neural network model, accurate modeling and prediction of the sewage treatment process can be achieved, providing a theoretical basis for the design of control strategies; By designing and implementing control strategies, dynamic adjustment of various control variables of the sewage treatment plant can be achieved so that the output data of the sewage treatment plant can reach the expected target while taking into account the operating cost and safety of the sewage treatment plant.

Claims

1. A sewage treatment process control system based on neural network, characterized in that: Includes the following modules: A data acquisition module is used to collect input and output data of the sewage treatment device, including the flow rate, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content of the sewage, and pre-process the data; Model building module, used to build a neural network model of the sewage treatment process based on the pre-processed data; Model optimization module, which is used to optimize the neural network structure and parameters using particle swarm optimization to obtain the optimized neural network model; A data prediction module is used to predict the output data of the sewage treatment device during the sewage treatment process based on the optimized neural network model; The device control module is used to design a control strategy based on the prediction results of the data prediction module; The sewage treatment device has a plurality of actuators, and is used to adjust the control variables of each actuator according to a control strategy.

2. The sewage treatment process control system based on neural network according to claim 1 is characterized in that: The data acquisition module specifically includes: Sensors for measuring input and output data from wastewater treatment plants; A data converter, used to convert the data measured by the sensor into a digital signal and send it to the data processor; The data processor is used to preprocess the data sent by the data converter, including filtering, normalization and feature extraction, to obtain preprocessed data.

3. The sewage treatment process control system based on neural network according to claim 1, characterized in that: The neural network model is a recurrent neural network based on a long short-term memory network, the number of its input layer nodes is the dimension of the input data of the sewage treatment device, the number of its output layer nodes is the dimension of the output data of the sewage treatment device, the number of its hidden layer nodes is the average of the number of input layer nodes and the number of output layer nodes, and its activation function is a hyperbolic tangent function.

4. The sewage treatment process control system based on neural network according to claim 1, characterized in that: The device control module specifically includes: The target value setting submodule is used to set the target values of the outlet water flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content; The deviation acquisition submodule is used to obtain the predicted values of the outlet water flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content based on the prediction results of the data prediction module, compare the predicted values with the target values, and obtain the deviation and deviation change rate of each of the above output variables; The fuzzy reasoning submodule is used to fuzzify, reason and defuzzify the deviation and the deviation change rate using the fuzzy reasoning system to obtain the opening of the water inlet valve, the opening of the water outlet valve, the speed of the aerator and the adjustment amount of the sludge return ratio; The device adjustment submodule uses the adjustment amount as a control signal and sends it to various actuators of the sewage treatment device to adjust various control variables of the sewage treatment device.

5. The sewage treatment process control system based on neural network according to claim 4 is characterized in that: The adjustment amounts of the various control variables are calculated using the following formula: ， in is the adjustment amount of each control variable, including the adjustment amount of the water inlet valve opening, the adjustment amount of the water outlet valve opening, the adjustment amount of the aerator speed and the adjustment amount of the sludge return ratio. is the deviation of each output variable, Refers to deviation The differential of , which represents the rate of change of the deviation, 、 and are the proportional, integral and differential coefficients respectively.

6. A sewage treatment process control method based on neural network, characterized in that: The following steps are involved: Collecting input and output data of the sewage treatment plant, including sewage flow, temperature, pH value, dissolved oxygen concentration, and nitrogen and phosphorus content, and preprocessing the data; Based on the pre-processed data, a neural network model of the sewage treatment process is established; The particle swarm algorithm is used to optimize the neural network structure and parameters to obtain the optimized neural network model; According to the optimized neural network model, the output data of the sewage treatment device during the sewage treatment process is predicted; According to the prediction results of the data prediction module, a control strategy is designed to control the control variables of each actuator in the sewage treatment device.

7. The sewage treatment process control method based on neural network according to claim 6 is characterized in that: The method of optimizing the neural network structure and parameters using the particle swarm algorithm to obtain an optimized neural network model specifically includes the following steps: Initialize the parameters of the particle swarm algorithm, including the number of particles, maximum number of iterations, inertia weight, learning factor, speed limit and position limit; A set of particles is randomly generated, each particle represents the structure and parameters of the neural network, including the number of input layer nodes, the number of hidden layer nodes, the number of output layer nodes, connection weights and biases; Calculate the fitness value of each particle, that is, the prediction accuracy of the neural network model, using the input data of the sewage treatment plant as the input of the neural network, and the output data of the sewage treatment plant as the expected output of the neural network. Calculate the mean square error between the output of the neural network and the expected output, and use it as the inverse of the fitness value; Update the individual optimal position and global optimal position of each particle, that is, record the position corresponding to the highest fitness value achieved by each particle in the historical iteration, and the position corresponding to the highest fitness value achieved by the entire particle group in the historical iteration; Update the speed and position of each particle. Use the inertia weight, learning factor, individual optimal position and global optimal position, and follow the update formula of the particle swarm algorithm to adjust the speed and position of each particle so that it searches in a more optimal direction. Determine whether the termination condition is met, that is, whether the maximum number of iterations is reached or the fitness value reaches the preset threshold. If so, output the neural network structure and parameters corresponding to the global optimal position as the optimized neural network model; if not, return to the third step and continue iteration.

Images