A Robust Control Method for Urban Wastewater Treatment Process Based on Adaptive Type II Fuzzy Neural Network

Abstract

This invention proposes a robust control method for urban wastewater treatment processes based on an adaptive type-II fuzzy neural network. Addressing the problem of difficulty in precisely controlling variables in urban wastewater treatment processes due to external disturbances, this method achieves robust tracking control of dissolved oxygen and nitrate nitrogen concentrations under external disturbances. The proposed control method uses an adaptive type-II fuzzy neural network to predict the dynamic model of the urban wastewater treatment process, constructs a disturbance observer to estimate external disturbances, and designs a robust controller to precisely control the dissolved oxygen and nitrate nitrogen concentrations, thus achieving safe and stable operation of the wastewater treatment process under external disturbances.

Description

Technical Field

[0001] This invention designs a robust control method for urban wastewater treatment processes based on an adaptive type-II fuzzy neural network. It uses the adaptive type-II fuzzy neural network to predict changes in dissolved oxygen and nitrate nitrogen concentrations, and estimates external disturbances through a disturbance observer. This solves the problem of inaccurate control of urban wastewater treatment processes due to external disturbances, and achieves robust tracking control of urban wastewater treatment processes. Therefore, this invention belongs to both the field of water research and the field of intelligent control. Background Technology

[0002] With continuous population growth and increasing demand for freshwater resources, water scarcity has become a global problem that urgently needs to be addressed. The widespread establishment of urban wastewater treatment plants can realize the recycling of water resources and alleviate the water shortage problem; the control technologies employed in these plants have been significantly developed. Urban wastewater treatment process control technologies can ensure that wastewater meets discharge standards after treatment, reduce pollutant emissions, and avoid harm to the environment and human health. Therefore, researching and promoting effective wastewater treatment process control technologies is of great significance for improving the operational level and economic efficiency of wastewater treatment plants.

[0003] Currently, the most commonly used process in wastewater treatment is the activated sludge process. This process is a complex nonlinear dynamic biochemical reaction involving multiple reactions such as ammoniation, nitrification, and denitrification. The two main controlled variables are the dissolved oxygen concentration in the fifth biological reaction tank and the nitrate nitrogen concentration in the second biological reaction tank. Dissolved oxygen promotes the degradation of ammonia nitrogen during nitrification, while nitrate nitrogen, as an intermediate product, is ultimately decomposed into nitrogen gas through denitrification. Although these processes effectively remove organic matter, in practical controller design, they are often affected by external disturbances such as weather changes, flow fluctuations, and water quality fluctuations. These external disturbances can lead to a decline in the performance of the control system, making the control process more complex and difficult. Therefore, maintaining the robustness of wastewater treatment process control under external disturbances has become an urgent problem to be solved.

[0004] This invention designs a robust control method for urban wastewater treatment processes based on an adaptive type-II fuzzy neural network. It establishes a prediction model for dissolved oxygen concentration and nitrate nitrogen concentration based on the type-II fuzzy neural network, updates the prediction model parameters using urban wastewater treatment process data, and designs an external disturbance estimator based on the tracking error of dissolved oxygen concentration and nitrate nitrogen concentration to achieve robust tracking control of urban wastewater treatment processes under external disturbances, meeting the needs of actual wastewater treatment plants. Summary of the Invention

[0005] This invention provides a robust control method for urban wastewater treatment processes based on an adaptive type-II fuzzy neural network. This method uses a prediction model based on a type-II fuzzy neural network to predict changes in dissolved oxygen concentration and nitrate nitrogen concentration. It then uses a gradient descent algorithm to update the parameters of the prediction model and designs a robust controller based on a perturbation observer, thereby achieving robust and precise control of dissolved oxygen concentration and nitrate nitrogen concentration.

[0006] The present invention adopts the following technical solution and implementation steps:

[0007] 1. A robust control method for urban wastewater treatment processes based on adaptive type-II fuzzy neural networks, characterized by establishing a dynamic prediction model for urban wastewater treatment processes based on adaptive type-II fuzzy neural networks, constructing an external disturbance estimator for urban wastewater treatment processes based on a disturbance observer, and designing a robust controller based on adaptive type-II fuzzy neural networks and a disturbance observer to achieve robust tracking control of dissolved oxygen concentration and nitrate nitrogen concentration under external disturbances in urban wastewater treatment processes, comprising the following steps:

[0008] (1) Establish a dynamic prediction model for urban wastewater treatment process based on adaptive type-II fuzzy neural network.

[0009] The kinetic models for dissolved oxygen and nitrate nitrogen concentrations in the urban wastewater treatment process are as follows:

[0010]

[0011]

[0012]

[0013] Among them, K L a5(t) represents the oxygen transfer coefficient of the fifth zone of the biochemical reactor at time t, Q a (t) represents the internal recirculation flow rate of the second section of the biochemical reactor at time t, Q k (t) represents the component flow rate of the k-th partition in the biochemical reactor at time t, S O,k (t) represents the dissolved oxygen concentration in the k-th partition of the biochemical reaction tank at time t, S NO,k (t) represents the nitrate nitrogen concentration in the kth zone of the biochemical reactor at time t, k = 1, 2, ..., 5, g1(t) represents the change in dissolved oxygen concentration in the fifth zone of the biochemical reactor at time t, g2(t) represents the change in nitrate nitrogen concentration in the second zone of the biochemical reactor at time t, d1(t) represents the external disturbance to dissolved oxygen concentration at time t, and d2(t) represents the external disturbance to nitrate nitrogen concentration at time t.

[0014] An adaptive type-2 fuzzy neural network was designed to predict dissolved oxygen and nitrate nitrogen concentrations. The adaptive type-2 fuzzy neural network consists of an input layer, a membership function layer, an activation layer, a consequent layer, and an output layer, as detailed below:

[0015] Input layer: This layer consists of 4 neurons:

[0016] z(t)=[z1(t),z2(t),z3(t),z4(t)] T (4)

[0017] Where z(t) represents the input vector of the type II fuzzy neural network at time t, z1(t) represents the error between the set value and the actual measured value of dissolved oxygen concentration at time t, z2(t) represents the error between the set value and the actual measured value of nitrate nitrogen concentration at time t, z3(t) represents the rate of change of the error between the set value and the actual measured value of dissolved oxygen concentration at time t, z4(t) represents the rate of change of the error between the set value and the actual measured value of nitrate nitrogen concentration at time t, and T represents the transpose of the matrix;

[0018] Membership function layer: This layer has M neurons, and the output of each neuron is:

[0019]

[0020]

[0021] Among them, z i (t) represents the i-th variable in the input vector z(t) of the input layer. m ij (t) represents the lower bound of the output values ​​of neurons in the membership function layer from the i-th input neuron to the j-th activation layer neuron at time t. This represents the upper bound of the output values ​​of neurons in the membership function layer from the i-th input neuron to the j-th activation layer neuron at time t. c ij (t) represents the lower bound of the center value of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t. σ represents the upper bound of the center value of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t. ij (t) represents the width of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t, where i represents the number of inputs to the type II fuzzy neural network and j represents the number of activation layer neurons in the type II fuzzy neural network, i = 1, 2, 3, 4, j = 1, 2, ..., M;

[0022] Activation layer: This layer has M neurons, and the upper bound and lower bound output of each neuron are represented as follows:

[0023]

[0024]

[0025] Among them, f j (t) represents the lower bound output of the j-th activated neuron at time t. This represents the upper bound of the output of the j-th activated neuron at time t;

[0026] The consequent layer: This layer has 4 neurons, and the output of each neuron is represented as:

[0027]

[0028]

[0029]

[0030] in, y o (t) represents the lower bound of the output of the consequent layer neuron of the o-th output neuron at time t. This represents the upper bound of the output of the consequent layer neuron of the o-th output neuron at time t. Let represent the weight coefficients of the o-th output neuron, the i-th input neuron, and the j-th activation layer neuron at time t. This represents the weight bias of the o-th output neuron and the j-th activation layer neuron at time t. Let $t$ represent the consequent weights between the $o$-th output neuron and the $j$-th activation layer neuron at time $t$, where $o = 1, 2$.

[0031] Output layer: This layer has 2 neurons, and the output of each neuron is represented as:

[0032]

[0033]

[0034] Among them, y 1 (t) represents the predicted dissolved oxygen concentration in the fifth zone at time t, y 2 (t) represents the predicted value of nitrate nitrogen concentration in the second partition at time t, η 1 (t) represents y at time t. 1 The lower bound of the output scale of (t), η 2 (t) represents y at time t. 2 The lower bound of the output scale of (t), 0 < η 1 (t)<1, 0<η 2 (t)<1;

[0035] Define the error function E(t):

[0036]

[0037] in, y represents the expected output of the type II fuzzy neural network at time t. o (t) represents the actual output of the type II fuzzy neural network at time t;

[0038] Design a type II fuzzy neural network with parameter η. o (t), and Adaptive update law:

[0039]

[0040]

[0041]

[0042] Where, η o (t+1) represents the proportion of the lower bound of the output of the consequent layer neuron of the 0th output layer neuron at time t+1. Let represent the weight coefficients of the o-th output neuron, the i-th input neuron, and the j-th activation layer neuron at time t+1. This represents the weight bias of the o-th output neuron and the j-th activation layer neuron at time t+1;

[0043] (2) Construct an external disturbance estimator for urban wastewater treatment process based on a disturbance observer.

[0044] A perturbation observer is constructed to estimate the external perturbations d1(t) and d2(t) of the urban wastewater treatment process at time t, specifically as follows:

[0045]

[0046]

[0047] in, Let represent the estimated value of the external disturbance d1(t) at time t. Let d2(t) represent the estimated value of the external disturbance at time t, λ represent the attenuation coefficient of the disturbance observer, and k represent the value of the external disturbance at time t. e Indicates the gain of the perturbation observer;

[0048] (3) Design a robust controller based on an adaptive type-II fuzzy neural network and a disturbance observer.

[0049] ① At the initial time t=1 of the control action, the initial values ​​of the parameters of the adaptive type II fuzzy neural network are randomly generated;

[0050] ②The predicted value of the change in dissolved oxygen concentration in the fifth zone is calculated according to formulas (4)-(13). Predicted values ​​of nitrate nitrogen concentration changes in the second zone

[0051] ③ Solve the parameter update law of the adaptive type II fuzzy neural network prediction model according to formulas (14)-(17);

[0052] ④ Calculate the estimated values ​​of external disturbances in the urban wastewater treatment process according to formulas (18)-(19). and

[0053] ⑤ Calculate the robust control law u1(t) of dissolved oxygen concentration at time t and the robust control law u2(t) of nitrate nitrogen concentration at time t, as follows:

[0054]

[0055]

[0056] Where sgn(·) represents the sign function;

[0057] ⑥ Increment time t by 1. If t < 200, return to step ②. If t = 200, then give the output values ​​u1(t) and u2(t) of the robust controller at the current time.

[0058] (4) Robust tracking control of dissolved oxygen concentration and nitrate nitrogen concentration

[0059] Dissolved oxygen concentration is controlled by a robust control law u1(t) to obtain the control quantity at the corresponding time, and nitrate nitrogen concentration is controlled by a robust control law u2(t) to obtain the control quantity at the corresponding time. The control quantities are then applied to the urban wastewater treatment process to achieve robust tracking control of the controlled variables, dissolved oxygen concentration and nitrate nitrogen concentration.

[0060] The inventiveness of this invention is mainly reflected in:

[0061] (1) In view of the problem that it is difficult to establish an accurate mathematical model for urban sewage treatment process, this invention designs an adaptive type II fuzzy neural network to predict unknown dynamic models and uses the gradient descent method to update the weight parameters of the prediction model, which has good adaptive ability.

[0062] (2) This invention adopts an external disturbance estimator for urban sewage treatment process based on a disturbance observer to ensure the anti-interference capability of the control system, solve the problem that dissolved oxygen concentration and nitrate nitrogen concentration are difficult to control accurately due to external disturbances, and achieve good robust performance. Attached Figure Description

[0063] Figure 1This is a structural diagram of the controller of the present invention.

[0064] Figure 2 This is a graph showing the dissolved oxygen concentration tracking results of the present invention.

[0065] Figure 3 This is a graph showing the dissolved oxygen concentration tracking error results of the present invention.

[0066] Figure 4 This is a graph showing the nitrate nitrogen concentration tracking results of the present invention.

[0067] Figure 5 This is a graph showing the tracking error results of nitrate nitrogen concentration according to the present invention. Detailed Implementation

[0068] 1. A robust control method for urban wastewater treatment processes based on adaptive type-II fuzzy neural networks, characterized by establishing a dynamic prediction model for urban wastewater treatment processes based on adaptive type-II fuzzy neural networks, constructing an external disturbance estimator for urban wastewater treatment processes based on a disturbance observer, and designing a robust controller based on adaptive type-II fuzzy neural networks and a disturbance observer to achieve robust tracking control of dissolved oxygen concentration and nitrate nitrogen concentration under external disturbances in urban wastewater treatment processes, comprising the following steps:

[0069] (1) Establish a dynamic prediction model for urban wastewater treatment process based on adaptive type-II fuzzy neural network.

[0070] The kinetic models for dissolved oxygen and nitrate nitrogen concentrations in the urban wastewater treatment process are as follows:

[0071]

[0072]

[0073]

[0074] Among them, K L a5(t) represents the oxygen transfer coefficient of the fifth zone of the biochemical reactor at time t, Q a (t) represents the internal recirculation flow rate of the second section of the biochemical reactor at time t, Q k (t) represents the component flow rate of the k-th partition in the biochemical reactor at time t, S O,k (t) represents the dissolved oxygen concentration in the k-th partition of the biochemical reaction tank at time t, S NO,k(t) represents the nitrate nitrogen concentration in the kth zone of the biochemical reactor at time t, k = 1, 2, ..., 5, g1(t) represents the change in dissolved oxygen concentration in the fifth zone of the biochemical reactor at time t, g2(t) represents the change in nitrate nitrogen concentration in the second zone of the biochemical reactor at time t, d1(t) represents the external disturbance to dissolved oxygen concentration at time t, and d2(t) represents the external disturbance to nitrate nitrogen concentration at time t.

[0075] An adaptive type-2 fuzzy neural network was designed to predict dissolved oxygen and nitrate nitrogen concentrations. The adaptive type-2 fuzzy neural network consists of an input layer, a membership function layer, an activation layer, a consequent layer, and an output layer, as detailed below:

[0076] Input layer: This layer consists of 4 neurons:

[0077] z(t)=[z1(t),z2(t),z3(t),z4(t)] T (25)

[0078] Where z(t) represents the input vector of the type II fuzzy neural network at time t, z1(t) represents the error between the set value and the actual measured value of dissolved oxygen concentration at time t, z2(t) represents the error between the set value and the actual measured value of nitrate nitrogen concentration at time t, z3(t) represents the rate of change of the error between the set value and the actual measured value of dissolved oxygen concentration at time t, z4(t) represents the rate of change of the error between the set value and the actual measured value of nitrate nitrogen concentration at time t, and T represents the transpose of the matrix;

[0079] Membership function layer: This layer has 10 neurons, and the output of each neuron is:

[0080]

[0081]

[0082] Among them, z i (t) represents the i-th variable in the input vector z(t) of the input layer, mm ij (t) represents the lower bound of the output values ​​of neurons in the membership function layer from the i-th input neuron to the j-th activation layer neuron at time t. cj represents the upper bound of the output values ​​of neurons in the membership function layer from the i-th input neuron to the j-th activation layer neuron at time t. ij (t) represents the lower bound of the center value of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t. σ represents the upper bound of the center value of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t. ij(t) represents the width of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t, where i represents the number of inputs to the type II fuzzy neural network and j represents the number of activation layer neurons in the type II fuzzy neural network, i = 1, 2, 3, 4, j = 1, 2, ..., 10;

[0083] Activation layer: This layer has 10 neurons. The upper bound and lower bound output of each neuron are represented as follows:

[0084]

[0085]

[0086] —where f j (t) represents the lower bound output of the j-th activated neuron at time t, f j (t) represents the upper bound output of the j-th activated neuron at time t;

[0087] The consequent layer: This layer has 4 neurons, and the output of each neuron is represented as:

[0088]

[0089]

[0090]

[0091] Among them, y o (t) represents the lower bound of the output of the consequent layer neuron of the o-th output neuron at time t. This represents the upper bound of the output of the consequent layer neuron of the o-th output neuron at time t. Let represent the weight coefficients of the o-th output neuron, the i-th input neuron, and the j-th activation layer neuron at time t. This represents the weight bias of the o-th output neuron and the j-th activation layer neuron at time t. Let $t$ represent the consequent weights between the $o$-th output neuron and the $j$-th activation layer neuron at time $t$, where $o = 1, 2$.

[0092] Output layer: This layer has 2 neurons, and the output of each neuron is represented as:

[0093]

[0094]

[0095] Among them, y 1 (t) represents the predicted dissolved oxygen concentration in the fifth zone at time t, y 2 (t) represents the predicted value of nitrate nitrogen concentration in the second partition at time t, η 1(t) represents y at time t. 1 The lower bound of the output scale of (t), η 2 (t) represents y at time t. 2 The lower bound of the output scale of (t), 0 < η 1 (t)<1, 0<η 2 (t)<1;

[0096] Define the error function E(t):

[0097]

[0098] in, y represents the expected output of the type II fuzzy neural network at time t. o (t) represents the actual output of the type II fuzzy neural network at time t;

[0099] Design a type II fuzzy neural network with parameter η. o (t), and Adaptive update law:

[0100]

[0101]

[0102]

[0103] Where, η o (t+1) represents the proportion of the lower bound of the output of the consequent layer neuron of the 0th output layer neuron at time t+1. Let represent the weight coefficients of the o-th output neuron, the i-th input neuron, and the j-th activation layer neuron at time t+1. This represents the weight bias of the o-th output neuron and the j-th activation layer neuron at time t+1;

[0104] (2) Construct an external disturbance estimator for urban wastewater treatment process based on a disturbance observer.

[0105] A perturbation observer is constructed to estimate the external perturbations d1(t) and d2(t) of the urban wastewater treatment process at time t, specifically as follows:

[0106]

[0107]

[0108] in, Let represent the estimated value of the external disturbance d1(t) at time t. Let d2(t) represent the estimated value of the external disturbance at time t, λ represent the attenuation coefficient of the disturbance observer, and k represent the value of the external disturbance at time t.e Indicates the gain of the perturbation observer;

[0109] (3) Design a robust controller based on an adaptive type-II fuzzy neural network and a disturbance observer.

[0110] ① At the initial time t=1 of the control action, the initial values ​​of the parameters of the adaptive type II fuzzy neural network are randomly generated;

[0111] ②The predicted value of the change in dissolved oxygen concentration in the fifth zone is calculated according to formulas (25)-(34). Predicted values ​​of nitrate nitrogen concentration changes in the second zone

[0112] ③ Solve the parameter update law of the adaptive type II fuzzy neural network prediction model according to formulas (35)-(38);

[0113] ④ According to formulas (39)-(40), the attenuation coefficient of the disturbance observer λ=15, and the gain k e =0.1, calculate the estimated value of external disturbances in the urban wastewater treatment process. and

[0114] ⑤ Calculate the robust control law u1(t) of dissolved oxygen concentration at time t and the robust control law u2(t) of nitrate nitrogen concentration at time t, as follows:

[0115]

[0116]

[0117] Where sgn(·) represents the sign function;

[0118] ⑥ Increment time t by 1. If t < 200, return to step ②. If t = 200, then give the output values ​​u1(t) and u2(t) of the robust controller at the current time.

[0119] (4) Robust tracking control of dissolved oxygen concentration and nitrate nitrogen concentration

[0120] Dissolved oxygen concentration is controlled by a robust control law u1(t) to obtain the control quantity at the corresponding time, and nitrate nitrogen concentration is controlled by a robust control law u2(t) to obtain the control quantity at the corresponding time. The control quantities are then applied to the urban wastewater treatment process to achieve robust tracking control of the controlled variables, dissolved oxygen concentration and nitrate nitrogen concentration. Figure 2 Displays dissolved oxygen concentration values ​​during the wastewater treatment process. X-axis: Time (days); Y-axis: Dissolved oxygen concentration (mg / L). The black dashed line represents the setpoint dissolved oxygen concentration, and the black solid line represents the actual dissolved oxygen concentration. The error between the setpoint and the actual dissolved oxygen concentration is shown below. Figure 3X-axis: time, in days; Y-axis: dissolved oxygen concentration error value, in milligrams per liter. Figure 4 Displays the nitrate nitrogen concentration values ​​during the wastewater treatment process. X-axis: time (days); Y-axis: nitrate nitrogen concentration (mg / L). The black dashed line represents the setpoint nitrate nitrogen concentration, and the black solid line represents the actual nitrate nitrogen concentration. The error between the setpoint and the actual nitrate nitrogen concentration is shown below. Figure 5 X-axis: time, in days; Y-axis: nitrate nitrogen concentration error, in milligrams per liter; experimental results demonstrate the effectiveness of this method.

Claims

1. A robust control method for urban wastewater treatment processes based on an adaptive type-II fuzzy neural network, characterized in that, Includes the following steps: (1) Establish a dynamic prediction model for urban wastewater treatment process based on adaptive type-II fuzzy neural network. The kinetic models for dissolved oxygen and nitrate nitrogen concentrations in the urban wastewater treatment process are as follows: Among them, K L a5(t) represents the oxygen transfer coefficient of the fifth zone of the biochemical reactor at time t, Q a (t) represents the internal recirculation flow rate of the second section of the biochemical reactor at time t, Q k (t) represents the component flow rate of the k-th partition in the biochemical reactor at time t, S O,k (t) represents the dissolved oxygen concentration in the k-th partition of the biochemical reaction tank at time t, S NO,k (t) represents the nitrate nitrogen concentration in the kth zone of the biochemical reactor at time t, k = 1, 2, ..., 5, g1(t) represents the change in dissolved oxygen concentration in the fifth zone of the biochemical reactor at time t, g2(t) represents the change in nitrate nitrogen concentration in the second zone of the biochemical reactor at time t, d1(t) represents the external disturbance to dissolved oxygen concentration at time t, and d2(t) represents the external disturbance to nitrate nitrogen concentration at time t. An adaptive type-2 fuzzy neural network was designed to predict dissolved oxygen and nitrate nitrogen concentrations. The adaptive type-2 fuzzy neural network consists of an input layer, a membership function layer, an activation layer, a consequent layer, and an output layer, as detailed below: Input layer: This layer consists of 4 neurons: z(t)=[z1(t),z2(t),z3(t),z4(t)] T (4) Where z(t) represents the input vector of the type II fuzzy neural network at time t, z1(t) represents the error between the set value and the actual measured value of dissolved oxygen concentration at time t, z2(t) represents the error between the set value and the actual measured value of nitrate nitrogen concentration at time t, z3(t) represents the rate of change of the error between the set value and the actual measured value of dissolved oxygen concentration at time t, z4(t) represents the rate of change of the error between the set value and the actual measured value of nitrate nitrogen concentration at time t, and T represents the transpose of the matrix; Membership function layer: This layer has M neurons, and the output of each neuron is: Among them, z i (t) represents the i-th variable in the input vector z(t) of the input layer. m ij (t) represents the lower bound of the output values ​​of neurons in the membership function layer from the i-th input neuron to the j-th activation layer neuron at time t. This represents the upper bound of the output values ​​of neurons in the membership function layer from the i-th input neuron to the j-th activation layer neuron at time t. c ij (t) represents the lower bound of the center value of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t. σ represents the upper bound of the center value of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t. ij (t) represents the width of the membership function layer neurons from the i-th input neuron to the j-th activation layer neuron at time t, where i represents the number of inputs to the type II fuzzy neural network and j represents the number of activation layer neurons in the type II fuzzy neural network, i = 1, 2, 3, 4, j = 1, 2, ..., M; Activation layer: This layer has M neurons, and the upper bound and lower bound output of each neuron are represented as follows: in, f j (t) represents the lower bound output of the j-th activated neuron at time t. This represents the upper bound of the output of the j-th activated neuron at time t; The consequent layer: This layer has 4 neurons, and the output of each neuron is represented as: in, y o (t) represents the lower bound of the output of the consequent layer neuron of the o-th output neuron at time t. This represents the upper bound of the output of the consequent layer neuron of the o-th output neuron at time t. Let represent the weight coefficients of the o-th output neuron, the i-th input neuron, and the j-th activation layer neuron at time t. This represents the weight bias of the o-th output neuron and the j-th activation layer neuron at time t. Let $t$ represent the consequent weights between the $o$-th output neuron and the $j$-th activation layer neuron at time $t$, where $o = 1, 2$. Output layer: This layer has 2 neurons, and the output of each neuron is represented as: Among them, y 1 (t) represents the predicted dissolved oxygen concentration in the fifth zone at time t, y 2 (t) represents the predicted value of nitrate nitrogen concentration in the second partition at time t, η 1 (t) represents y at time t. 1 The lower bound of the output scale of (t), η 2 (t) represents y at time t. 2 The lower bound of the output scale of (t), 0 < η 1 (t)<1, 0<η 2 (t)<1; Define the error function E(t): in, y represents the expected output of the type II fuzzy neural network at time t. o (t) represents the actual output of the type II fuzzy neural network at time t; Design a type II fuzzy neural network with parameter η. o (t), and The adaptive update law: Where, η o (t+1) represents the proportion of the lower bound of the output of the consequent layer neuron of the 0th output layer neuron at time t+1. Let represent the weight coefficients of the o-th output neuron, the i-th input neuron, and the j-th activation layer neuron at time t+1. This represents the weight bias of the o-th output neuron and the j-th activation layer neuron at time t+1; (2) Construct an external disturbance estimator for urban wastewater treatment process based on a disturbance observer. A perturbation observer is constructed to estimate the external perturbations d1(t) and d2(t) of the urban wastewater treatment process at time t, specifically as follows: in, Let represent the estimated value of the external disturbance d1(t) at time t. Let d2(t) represent the estimated value of the external disturbance at time t, λ represent the attenuation coefficient of the disturbance observer, and k represent the value of the external disturbance at time t. e Indicates the gain of the perturbation observer; (3) Design a robust controller based on an adaptive type-II fuzzy neural network and a disturbance observer. ① At the initial time t=1 of the control action, the initial values ​​of the parameters of the adaptive type II fuzzy neural network are randomly generated; ②The predicted value of the change in dissolved oxygen concentration in the fifth zone is calculated according to formulas (4)-(13). Predicted values ​​of nitrate nitrogen concentration changes in the second zone ③ Solve the parameter update law of the adaptive type II fuzzy neural network prediction model according to formulas (14)-(17); ④ Calculate the estimated values ​​of external disturbances in the urban wastewater treatment process according to formulas (18)-(19). and ⑤ Calculate the robust control law u1(t) of dissolved oxygen concentration at time t and the robust control law u2(t) of nitrate nitrogen concentration at time t, as follows: Where sgn(·) represents the sign function; ⑥ Increment time t by 1. If t < 200, return to step ②. If t = 200, then give the output values ​​u1(t) and u2(t) of the robust controller at the current time. (4) Robust tracking control of dissolved oxygen concentration and nitrate nitrogen concentration Dissolved oxygen concentration is controlled by a robust control law u1(t) to obtain the control quantity at the corresponding time, and nitrate nitrogen concentration is controlled by a robust control law u2(t) to obtain the control quantity at the corresponding time. The control quantities are then applied to the urban wastewater treatment process to achieve robust tracking control of the controlled variables, dissolved oxygen concentration and nitrate nitrogen concentration.

Images