Dynamic multi-objective optimization control method for sewage treatment process based on neural network

Abstract

The invention provides a dynamic multi-objective optimization control method for a sewage treatment process based on a neural network. The dynamic multi-objective optimization control method comprises the following steps: obtaining characteristic variables of effluent quality and energy consumption in the sewage treatment process; based on the echo state network, constructing a to-be-optimized effluent quality model and a to-be-optimized energy consumption model according to the characteristic variables; the dynamic multi-objective optimization algorithm is used for optimizing the to-be-optimized effluent quality model and the to-be-optimized energy consumption model to obtain a target effluent quality model and a target energy consumption model; determining the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration in the target effluent quality model and the target energy consumption model; and based on a control formula, according to the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration, performing dynamic tracking control on the sewage treatment process by using a single-neuron adaptive controller. The method solves the problems that in the prior art, dynamic optimization control is difficult to achieve in the sewage treatment process, and the energy consumption and the operation cost of the sewage treatment process are increased.

Description

Technical Field

[0001] The invention relates to the technical field of sewage treatment, and in particular to a dynamic multi-objective optimization control method for a sewage treatment process based on a neural network. Background Art

[0002] The urban sewage treatment process involves complex biochemical reactions, which are dynamic, nonlinear and uncertain. The core goal of the sewage treatment process is to remove or reduce harmful substances in sewage to a safe level to ensure that the effluent quality meets environmental protection standards. With the increase in sewage discharge and increasingly stringent emission standards, the cost of sewage treatment continues to increase. Therefore, for the sewage treatment process, how to ensure that the effluent quality meets the standards while reducing operating energy consumption is a major task and challenge.

[0003] In order to adapt to the dynamic sewage treatment process, improve sewage treatment efficiency and save operating costs, the dynamic optimization control method of the sewage treatment process came into being. In order to verify the effectiveness of different optimization control methods in the sewage treatment process, the International Water Quality Association and the European Union Scientific and Technical Cooperation Organization jointly developed the Benchmark Simulation Model No. 1 (BSM1) for sewage treatment. BSM1 consists of a biochemical reaction tank and a secondary sedimentation tank. The biochemical reaction tank includes 5 units, namely the first partition, the second partition, the third partition, the fourth partition and the fifth partition. The first two units are anoxic zones, where denitrification reactions mainly occur, and the last three units are aerobic zones, where nitrification reactions mainly occur. The sewage is clarified in the secondary sedimentation tank, the supernatant flows out, and part of the lower sludge flows into the anoxic zone through external reflux, and the other part is discharged as residual sludge. In the optimization control process, the accurate description of the dynamic mapping relationship between effluent water quality, operating energy consumption and its key variables is the basis and prerequisite for realizing the optimization control of the sewage treatment process. In the prior art, it is difficult to achieve dynamic optimization control in the sewage treatment process. In addition, in the actual sewage treatment process, the changes in the dissolved oxygen concentration in the fifth zone and the nitrate nitrogen concentration in the second zone will directly affect the nitrification and denitrification processes, and then affect the energy consumption and operating costs of the sewage treatment process, resulting in an increase in the energy consumption and operating costs of the sewage treatment process.

[0004] The present invention conducts a dynamic optimization control experiment of sewage treatment based on the BSM1 model, dynamically obtains the optimal set values ​​of the dissolved oxygen concentration in the fifth partition and the nitrate nitrogen concentration in the second partition, and uses a single neuron adaptive controller to achieve effective tracking control of the optimized set values, wherein the dissolved oxygen concentration in the fifth partition of the aerobic zone is expressed as S O The nitrate nitrogen concentration in the second zone of the anaerobic zone is expressed as S NO , the oxygen transfer coefficient K of the fifth zone L a5 is S OThe control parameters of the fifth partition to the second partition are Q a YesS NO control parameters. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a dynamic multi-objective optimization control method for a sewage treatment process based on a neural network. The present invention solves the problem that it is difficult to achieve dynamic optimization control of the sewage treatment process in the prior art and that the energy consumption and operating costs of the sewage treatment process are increased.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A dynamic multi-objective optimization control method for a sewage treatment process based on a neural network, comprising:

[0008] Obtain characteristic variables of effluent water quality and energy consumption in the sewage treatment process; wherein the characteristic variables include: dissolved oxygen concentration, nitrate nitrogen concentration, suspended solids concentration, ammonia nitrogen concentration and inlet flow rate;

[0009] Based on the echo state network, the effluent water quality model to be optimized and the energy consumption model to be optimized are constructed according to the characteristic variables;

[0010] Based on the dynamic multi-objective optimization algorithm of two-space prediction, the effluent water quality model to be optimized and the energy consumption model to be optimized are optimized to obtain the target effluent water quality model and the target energy consumption model;

[0011] Determine the optimal dissolved oxygen concentration and optimal nitrate nitrogen concentration in the target effluent water quality model and the target energy consumption model;

[0012] Based on the control formula, a single neuron adaptive controller is used to perform dynamic tracking control of the sewage treatment process according to the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration.

[0013] Preferably, the energy consumption model is:

[0014]

[0015] f1(x(t)) is the energy consumption model at time t, and the input variable is xt=S O t S NO t,MLSS t S NH t,Q in (t], where S O (t) is the dissolved oxygen concentration at time t, S NO (t) is the nitrate nitrogen concentration at time t, MLSS(t) is the suspended solids concentration at time t, S NH (t) is the ammonia nitrogen concentration at time t, Q in(t) is the water inflow rate at time t, h(·) is the sigmoid activation function, and is the input weight matrix, W1 and W2 are the internal weight matrices of the reserve pool, and is the output weight matrix, H1 and H2 are the internal state matrices, s1(t) and s2(t) are the echo state matrices at time t, and Y1 and Y2 are the target output matrices.

[0016] Preferably, the effluent water quality model is:

[0017]

[0018] Among them, f2(x(t)) is the outlet water quality model at time t.

[0019] Preferably, the effluent water quality model to be optimized and the energy consumption model to be optimized are optimized based on the dynamic multi-objective optimization algorithm of two-space prediction to obtain the target effluent water quality model and the target energy consumption model, including:

[0020] Set the population size to N∈[50,150] and the maximum number of iterations to G max ∈[40,50];

[0021] Taking the energy consumption model to be optimized and the effluent water quality model to be optimized as the optimization objectives, and the dissolved oxygen concentration in the fifth partition and the nitrate nitrogen concentration in the second partition of the BSM1 model as the decision variables, the dynamic multi-objective optimization algorithm based on two-space prediction is used to initialize the population, calculate the objective function value, and obtain the target energy consumption model and the target effluent water quality model.

[0022] Preferably, the population is initialized based on a dynamic multi-objective optimization algorithm based on two-space prediction, the objective function value is calculated, and the target energy consumption model and the target effluent water quality model are obtained, including:

[0023] Get the environment vector;

[0024] Normalizing the environment vector to obtain a normalized vector;

[0025] Determine whether the Euclidean distance between the normalized vectors is greater than the environmental change threshold. If so, use a dynamic multi-objective optimization algorithm based on two-space prediction to calculate the optimized setting values ​​of dissolved oxygen and nitrate nitrogen concentrations to obtain the target energy consumption model and the target effluent water quality model. If not, predict the initial population at the next moment based on the initial population at the current moment and the initial population at the previous moment to obtain the initial population clusters at each moment, and optimize the initial population clusters at each moment based on the decomposed multi-objective evolutionary algorithm until the multi-objective evolutionary algorithm reaches the maximum number of iterations to obtain the target energy consumption model and the target effluent water quality model.

[0026] Preferably, the initial population at the next moment is predicted based on the initial population at the current moment and the initial population at the previous moment, and the initial population groups at each moment are obtained, including:

[0027] In the decision space, according to the initial population at the current moment and the initial population at the previous moment in the decision space, half of the initial population groups are obtained;

[0028] In the target space, the other half of the initial population is obtained according to the initial population of the target space at the current moment and the initial population at the previous moment;

[0029] The half of the initial seed clusters and the other half of the initial seed clusters are merged to obtain the initial seed clusters at each moment.

[0030] Preferably, determining the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration in the target effluent water quality model and the target energy consumption model comprises:

[0031] Obtain the dissolved oxygen concentration and optimal nitrate nitrogen concentration solution set in the target effluent water quality model and the target energy consumption model;

[0032] Based on the inflection point calculation formula, the inflection point in the solution set of dissolved oxygen concentration and optimal nitrate nitrogen concentration is calculated;

[0033] The optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration are determined according to the inflection point.

[0034] Preferably, the inflection point calculation formula is:

[0035]

[0036] The solution with the largest Dis value is the inflection point, where a, b, and c are coefficients determined by the two boundary points in the non-dominated solution set.

[0037] Preferably, the single neuron adaptive controller is:

[0038]

[0039] K1 is the neuron scale coefficient, satisfying K1>0, w i (t) is the weight coefficient.

[0040] Preferably, the control formula is:

[0041]

[0042] Among them, γ i >0 is the learning rate of the neuron, v1(t), v2(t), v3(t) are the inputs of the neuron, Δu(t)=[ΔK L a5(t),ΔQa (t)] T , ΔK L a5(t) is the change in oxygen transfer coefficient of the fifth subzone at time t, ΔQ a (t) is the change in the reflux flow at time t, e(t) = z * (t)-z(t) is the control error between the actual output and the optimized setting value, z * (t) = [S O * (t),S NO * (t)] is the optimal setting value of dissolved oxygen concentration and nitrate nitrogen concentration at time t, z(t) = [S O (t),S NO (t)] is the actual output dissolved oxygen concentration and nitrate nitrogen concentration at time t.

[0043] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0044] The invention provides a neural network-based dynamic multi-objective optimization control method for a sewage treatment process, comprising: obtaining characteristic variables of effluent water quality and energy consumption of the sewage treatment process; wherein the characteristic variables include: dissolved oxygen concentration, nitrate nitrogen concentration, suspended solids concentration, ammonia nitrogen concentration and inlet flow rate; constructing an effluent water quality model to be optimized and an energy consumption model to be optimized according to the characteristic variables based on an echo state network; optimizing the effluent water quality model to be optimized and the energy consumption model to be optimized based on a dynamic multi-objective optimization algorithm of two-space prediction to obtain a target effluent water quality model and a target energy consumption model; determining an optimal dissolved oxygen concentration and an optimal nitrate nitrogen concentration in the target effluent water quality model and the target energy consumption model; and performing dynamic tracking control of the sewage treatment process using a single neuron adaptive controller based on a control formula according to the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration. The invention can not only calculate the optimal set values ​​of the dissolved oxygen concentration and the nitrate nitrogen concentration in real time, but also realize accurate tracking of the optimal set values ​​of the dissolved oxygen concentration and the nitrate nitrogen concentration. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0046] Figure 1 A flow chart of a dynamic multi-objective optimization control method for a sewage treatment process based on a neural network provided by an embodiment of the present invention;

[0047] Figure 2 The overall structure diagram of the dynamic optimization control system provided by the embodiment of the present invention;

[0048] Figure 3 A diagram showing the dissolved oxygen concentration setting value and tracking result provided by an embodiment of the present invention;

[0049] Figure 4 A dissolved oxygen concentration tracking control error diagram provided by an embodiment of the present invention;

[0050] Figure 5 A graph showing the nitrate nitrogen concentration setting value and tracking result provided by an embodiment of the present invention;

[0051] Figure 6 This is a nitrate nitrogen concentration tracking control error diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] The purpose of the present invention is to provide a dynamic multi-objective optimization control method for a sewage treatment process based on a neural network. The present invention solves the problem in the prior art that it is difficult to achieve dynamic optimization control of the sewage treatment process and that the energy consumption and operating cost of the sewage treatment process increase.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] like Figure 1 As shown, the present invention provides a dynamic multi-objective optimization control method for a sewage treatment process based on a neural network, comprising:

[0056] Step 100: Obtain characteristic variables of effluent quality and energy consumption in the sewage treatment process; wherein the characteristic variables include: dissolved oxygen concentration, nitrate nitrogen concentration, suspended solids concentration, ammonia nitrogen concentration and inlet flow rate;

[0057] Specifically, the dynamic characteristics and operation data of the sewage treatment process are analyzed to obtain the key characteristic variables that affect the effluent quality and energy consumption of the sewage treatment process: dissolved oxygen concentration S O , nitrate nitrogen concentration S NO , suspended solids concentration MLSS, ammonia nitrogen concentration S NH and water inflow rate Q in ;

[0058] Step 200: Based on the echo state network, construct an effluent water quality model to be optimized and an energy consumption model to be optimized according to characteristic variables;

[0059] Step 300: Based on the dynamic multi-objective optimization algorithm of two-space prediction, the effluent water quality model to be optimized and the energy consumption model to be optimized are optimized to obtain the target effluent water quality model and the target energy consumption model;

[0060] Step 400: Determine the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration in the target effluent water quality model and the target energy consumption model;

[0061] Step 500: Based on the control formula, according to the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration, a single neuron adaptive controller is used to perform dynamic tracking control of the sewage treatment process.

[0062] Further, such as Figure 2 As shown, the energy consumption model is:

[0063]

[0064] f1(x(t)) is the energy consumption model at time t, and the input variable is xt=S O t S NO t,MLSS t S NH t,Q in (t], where S O (t) is the dissolved oxygen concentration at time t, S NO (t) is the nitrate nitrogen concentration at time t, MLSS(t) is the suspended solids concentration at time t, S NH (t) is the ammonia nitrogen concentration at time t, Q in (t) is the water inflow rate at time t, h(·) is the sigmoid activation function, and is the input weight matrix, W1 and W2 are the internal weight matrices of the reserve pool, they are randomly generated before training and remain unchanged once generated. and is the output weight matrix, H1 and H2 are the internal state matrices, s1(t) and s2(t) are the echo state matrices at time t, and Y1 and Y2 are the target output matrices.

[0065] Furthermore, the effluent water quality model is:

[0066]

[0067] Among them, f2(x(t)) is the outlet water quality model at time t.

[0068] Furthermore, based on the dynamic multi-objective optimization algorithm of two-space prediction, the effluent water quality model to be optimized and the energy consumption model to be optimized are optimized to obtain the target effluent water quality model and the target energy consumption model, including:

[0069] Set the population size to N∈[50,150] and the maximum number of iterations to G max ∈[40,50];

[0070] Taking the energy consumption model to be optimized and the effluent water quality model to be optimized as the optimization objectives, and the dissolved oxygen concentration in the fifth zone and the nitrate nitrogen concentration in the second zone as the decision variables, a dynamic multi-objective optimization algorithm based on two-space prediction is used to initialize the population, calculate the objective function value, and obtain the target energy consumption model and the target effluent water quality model.

[0071] Specifically, the dynamic multi-objective optimization algorithm based on two-space prediction is equivalent to the prediction strategy based on two spaces in the attached figure.

[0072] Specifically, the population size is set to N = 100 and the maximum number of iterations is Gmax = 40;

[0073] Set the energy consumption model and effluent water quality model as optimization objectives, and the decision variables are the dissolved oxygen concentration in the fifth partition and the nitrate nitrogen concentration in the second partition, initialize the population, and calculate the objective function value;

[0074] Furthermore, based on the dynamic multi-objective optimization algorithm of two-space prediction, the population is initialized, the objective function value is calculated, and the target energy consumption model and the target effluent water quality model are obtained, including:

[0075] Get the environment vector;

[0076] Normalizing the environment vector to obtain a normalized vector;

[0077] Determine whether the Euclidean distance between the normalized vectors is greater than the environmental change threshold. If so, use a dynamic multi-objective optimization algorithm based on two-space prediction to calculate the optimized setting values ​​of dissolved oxygen and nitrate nitrogen concentrations to obtain the target energy consumption model and the target effluent water quality model. If not, predict the initial population at the next moment based on the initial population at the current moment and the initial population at the previous moment to obtain the initial population clusters at each moment, and optimize the initial population clusters at each moment based on the decomposed multi-objective evolutionary algorithm until the multi-objective evolutionary algorithm reaches the maximum number of iterations to obtain the target energy consumption model and the target effluent water quality model.

[0078] Specifically, the environment vector is set to [MLSS(t),S NH (t)], where MLSS(t) is the concentration of suspended solids in the effluent at the current moment, S NH(t) is the current effluent ammonia nitrogen concentration; the optimization cycle is set to T = 2, that is, environmental testing is performed every 2 hours; the environmental vector needs to be normalized, and when the Euclidean distance between the environmental vectors is greater than the set environmental change threshold hour, If it is set to 0.1, it indicates that the environment has changed. When the environment has changed, the dynamic multi-objective optimization algorithm based on two-space prediction is used to obtain the optimal set values ​​of dissolved oxygen and nitrate nitrogen concentrations. When the environment has not changed, the decomposition-based multi-objective evolutionary algorithm MOEA / D is directly used to optimize the initial population.

[0079] The initial population is optimized using the decomposition-based multi-objective evolutionary algorithm MOEA / D to obtain a set of Pareto optimal solutions. This patent selects the weighted sum method as the decomposition method of MOEA / D.

[0080] The weighted summation method mainly assigns a weight value λ to each sub-problem i =(λ i1 ,λ i2 ),i∈1,...,N, and λ i1 +λ i2 =1, the weight value can be used to transform the multi-objective optimization problem into the following problem:

[0081]

[0082] By changing λ, a set of Pareto optimal solutions can be obtained.

[0083] Determine whether the algorithm has reached the set maximum number of evolutionary iterations G max If it is reached, the iterative evolution process is terminated; if it is not reached, the optimization continues.

[0084] Furthermore, the initial population at the next moment is predicted based on the initial population at the current moment and the initial population at the previous moment, and the initial population groups at each moment are obtained, including:

[0085] In the decision space, according to the initial population at the current moment and the initial population at the previous moment in the decision space, half of the initial population groups are obtained;

[0086] In the target space, the other half of the initial population is obtained according to the initial population of the target space at the current moment and the initial population at the previous moment;

[0087] The half of the initial seed clusters and the other half of the initial seed clusters are merged to obtain the initial seed clusters at each moment.

[0088] Specifically, the initial population of the new environment consists of two parts; first, in the decision space, 50% of the initial population is generated by establishing a denoising autoencoder prediction model; second, in the target space, the desired target vector is mapped back to the decision space by establishing an inverse model to generate the remaining 50% of the initial population.

[0089] Establish a denoising autoencoder model: When the environment changes, the Pareto optimal solution obtained in environment t and t-1 is used to predict the initial population at environment t+1. The Pareto optimal solutions of environment t and t-1 are represented by POS t and POS t-1 , randomly select POS t and POS t-1 The Pareto optimal solutions of half of them can be expressed as and Where N t Represents the number of solutions, then, using the denoising autoencoder for prediction, set As the input P of the denoising autoencoder, As the output Q, M is from arrive The learning mapping M is calculated as follows:

[0090] M=(QP T )(PP T ) -1 ;

[0091] Where T is the matrix transposition operation, the Pareto optimal solution at time t+1 in the environment can be expressed as:

[0092]

[0093] Finally, the Pareto optimal solution based on the autoencoder model is obtained, which is expressed as Pop dec .

[0094] Establishing the inverse model for demapping: POS t and POS t-1 The remaining half of the solution is expressed as and Then, calculate and The objective function values ​​are expressed as and use and Directly predict individuals in the target space. First, calculate and The center point F t and F t-1, then, make predictions for individuals in the target space:

[0095]

[0096] Where δ is the variance perturbation, and the perturbation of the variance is to expand the search area and make the new population closer to the true point.

[0097] Given that The predicted value is a set of target vectors, so we need to Mapping back to the decision space to obtain a new Pareto optimal solution. Since the radial basis function neural network is easy to train and has good nonlinear fitting ability, it can map any complex nonlinear relationship. Therefore, the present invention uses the radial basis function neural network to construct a nonlinear inverse model.

[0098] The implementation steps are as follows: and As input and output respectively for inverse model training, where i∈1,...,N t , so we need to establish two inverse models. and Divide into N t training data pairs. For the first inverse model, the training samples are For the second inverse model, the training samples are The Gaussian function is selected as the activation function of the radial basis function neural network, which is expressed as:

[0099]

[0100] Where σ>0 is the variance of the activation function, and C is the center of the Gaussian function. Therefore, the output of the radial basis function neural network is:

[0101]

[0102] where w j is the weight of the jth neuron.

[0103] Finally, use the trained inverse model to find The mapping solution in the decision space is regarded as the predicted solution based on the target space, denoted as Pop obj .

[0104] The Pop obtained from the decision space dec and the solution Pop required in the target space obj Merge to form the initial population of the environment at time t+1.

[0105] Furthermore, the determination of the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration in the target effluent water quality model and the target energy consumption model includes:

[0106] Obtain the dissolved oxygen concentration and optimal nitrate nitrogen concentration solution set in the target effluent water quality model and the target energy consumption model;

[0107] Based on the inflection point calculation formula, the inflection point in the solution set of dissolved oxygen concentration and optimal nitrate nitrogen concentration is calculated;

[0108] The optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration are determined according to the inflection point.

[0109] Specifically, the optimal setting value of the controller is selected:

[0110] The optimal solutions on the Pareto front do not dominate each other, and the specific selection of the optimal solution directly affects the optimization control performance of the sewage treatment process. After comprehensive consideration, the inflection point in the Pareto optimal solution is selected as the final set point of the controller. The present invention adopts a simple method to find the inflection point, which determines whether the solution is an inflection point by calculating the distance from each solution to the extreme value line. For the two-objective optimization problem, the extreme value line L is defined as follows:

[0111] L: a.f1+b.f2+c=0;

[0112] Among them, a, b and c can be determined by two extreme points in the non-dominated solution set. Therefore, for the solution x with coordinates (f1(x), f2(x)), its distance to the extreme line L can be calculated as:

[0113]

[0114] The inflection point represents the solution with the largest Dis value in the non-dominated solution set, that is, the solution corresponding to the maximum Dis value is selected as the optimal setting value, which can be expressed as z * (t) = [S O * (t),S NO * (t)].

[0115] Furthermore, tracking and controlling of dissolved oxygen concentration and nitrate nitrogen concentration:

[0116] After obtaining the optimal set values ​​of dissolved oxygen concentration and nitrate nitrogen concentration, the single neuron adaptive PID controller is used for tracking control. The dissolved oxygen concentration and nitrate nitrogen concentration are respectively determined by the change of oxygen transfer coefficient ΔK in the 5th partition. L a5(t) and the change of internal recirculation flow ΔQ a(t) is adjusted. For the single neuron adaptive control method, the present invention adopts a supervised Delta learning rule to design an incremental controller, and the controller (the single neuron adaptive controller) is defined as follows:

[0117]

[0118] K1 is the neuron scale coefficient, satisfying K1>0. i is the connection weight coefficient, and its (control formula) calculation formula is:

[0119]

[0120] Among them, γ i >0 is the learning rate of the neuron, v1(t), v2(t), v3(t) are the inputs of the neuron, Δu(t)=[ΔK L a5(t),ΔQ a (t)] T , ΔK L a5(t) is the change in oxygen transfer coefficient of the fifth subzone at time t, ΔQ a (t) is the change in the reflux flow at time t, e(t) = z * (t)-z(t) is the control error between the actual output and the optimized setting value, z * (t) = [S O * (t),S NO * (t)] is the optimal setting value of dissolved oxygen concentration and nitrate nitrogen concentration at time t, z(t) = [S O (t),S NO (t)] is the actual output dissolved oxygen concentration and nitrate nitrogen concentration at time t.

[0121] Specifically, the output of the dynamic multi-objective optimization control system for sewage treatment process based on neural network is the actual dissolved oxygen S O and nitrate S NO The concentration value of Figure 3 This is a graph of dissolved oxygen concentration setting value and tracking results. The solid line is the optimized setting value of dissolved oxygen concentration, and the dotted line is the actual dissolved oxygen output concentration value. The horizontal axis is time, unit: day, and the vertical axis is dissolved oxygen concentration, unit: mg / L. Figure 4 This is the dissolved oxygen concentration tracking control error diagram, horizontal axis: time, unit: day, vertical axis: dissolved oxygen concentration error value, unit: mg / L, Figure 5 This is a graph of nitrate nitrogen concentration setting values ​​and tracking results. The solid line is the optimized setting value of nitrate nitrogen concentration, and the dotted line is the actual nitrate nitrogen output concentration value. The horizontal axis is time, unit: day, and the vertical axis is nitrate nitrogen concentration, unit: mg / L. Figure 6It is the nitrate nitrogen concentration tracking control error diagram, horizontal axis: time, unit: day, vertical axis: nitrate nitrogen concentration error value, unit: mg / L.

[0122] The beneficial effects of the present invention are as follows:

[0123] (1) According to the dynamic characteristics and operation data of the sewage treatment process, the present invention uses the echo state network to establish an energy consumption and effluent water quality model, and sets it as the optimization objective function. The optimal setting values ​​of dissolved oxygen concentration and nitrate nitrogen concentration are obtained through a dynamic multi-objective optimization algorithm based on two-space prediction; and the tracking control of dissolved oxygen concentration and nitrate nitrogen concentration is realized in combination with a single neuron adaptive controller, which solves the problem that the effluent water quality is difficult to meet the standard and the operation energy consumption is high;

[0124] (2) The present invention adopts an optimization control method that combines a dynamic multi-objective optimization algorithm based on two-space prediction and a single neuron adaptive controller to achieve optimization and tracking control of the set values ​​of dissolved oxygen concentration and nitrate nitrogen concentration, thereby reducing operating costs while ensuring that the effluent water quality meets the standards.

[0125] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0126] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A dynamic multi-objective optimization control method for sewage treatment process based on neural network, characterized in that: include: Obtain characteristic variables of effluent water quality and energy consumption in the sewage treatment process; wherein the characteristic variables include: dissolved oxygen concentration, nitrate nitrogen concentration, suspended solids concentration, ammonia nitrogen concentration and inlet flow rate; Based on the echo state network, the effluent water quality model to be optimized and the energy consumption model to be optimized are constructed according to the characteristic variables; Based on the dynamic multi-objective optimization algorithm of two-space prediction, the effluent water quality model to be optimized and the energy consumption model to be optimized are optimized to obtain the target effluent water quality model and the target energy consumption model; Determine the optimal dissolved oxygen concentration and optimal nitrate nitrogen concentration in the target effluent water quality model and the target energy consumption model; Based on the control formula, a single neuron adaptive controller is used to perform dynamic tracking control of the sewage treatment process according to the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration.

2. According to claim 1, a dynamic multi-objective optimization control method for sewage treatment process based on neural network is characterized in that: The energy consumption model is: f1(x(t)) is the energy consumption model at time t, and the input variable is xt=S O t S NO t,MLSS t S NH t,Q in (t], where S O (t) is the dissolved oxygen concentration at time t, S NO (t) is the concentration of nitrate nitrogen at time t, MLSS(t) is the concentration of suspended solids at time t, S NH (t) is the ammonia nitrogen concentration at time t, Q in (t) is the water inflow rate at time t, h(·) is the sigmoid activation function, and is the input weight matrix, W1 and W2 are the internal weight matrices of the reserve pool, W1 out and is the output weight matrix, H1 and H2 are the internal state matrices, s1(t) and s2(t) are the echo state matrices at time t, and Y1 and Y2 are the target output matrices.

3. According to claim 1, a dynamic multi-objective optimization control method for sewage treatment process based on neural network is characterized in that: The effluent water quality model is: Among them, f2(x(t)) is the outlet water quality model at time t.

4. According to claim 1, a dynamic multi-objective optimization control method for sewage treatment process based on neural network is characterized in that: Based on the dynamic multi-objective optimization algorithm of two-space prediction, the effluent water quality model to be optimized and the energy consumption model to be optimized are optimized to obtain the target effluent water quality model and the target energy consumption model, including: Set the population size to N∈[50,150] and the maximum number of iterations to G max ∈[40,50]; Taking the energy consumption model to be optimized and the effluent water quality model to be optimized as the optimization objectives, and the dissolved oxygen concentration in the fifth zone and the nitrate nitrogen concentration in the second zone in the BSM1 model as the decision variables, the population is initialized based on the dynamic multi-objective optimization algorithm of two-space prediction, and the objective function value is calculated to obtain the target energy consumption model and the target effluent water quality model.

5. A dynamic multi-objective optimization control method for sewage treatment process based on neural network according to claim 4, characterized in that: Based on the dynamic multi-objective optimization algorithm of two-space prediction, the population is initialized, the objective function value is calculated, and the target energy consumption model and the target effluent water quality model are obtained, including: Get the environment vector; Normalizing the environment vector to obtain a normalized vector; Determine whether the Euclidean distance between the normalized vectors is greater than the environmental change threshold. If so, use a dynamic multi-objective optimization algorithm based on two-space prediction to calculate the optimized setting values ​​of dissolved oxygen and nitrate nitrogen concentrations to obtain the target energy consumption model and the target effluent water quality model. If not, predict the initial population at the next moment based on the initial population at the current moment and the initial population at the previous moment to obtain the initial population clusters at each moment, and optimize the initial population clusters at each moment based on the decomposed multi-objective evolutionary algorithm until the multi-objective evolutionary algorithm reaches the maximum number of iterations to obtain the target energy consumption model and the target effluent water quality model.

6. A neural network-based dynamic multi-objective optimization control method for sewage treatment process according to claim 5, characterized in that: Then, according to the initial population at the current moment and the initial population at the previous moment, the initial population at the next moment is predicted, and the initial population groups at each moment are obtained, including: In the decision space, according to the initial population at the current moment and the initial population at the previous moment in the decision space, half of the initial population groups are obtained; In the target space, the other half of the initial population is obtained according to the initial population of the target space at the current moment and the initial population at the previous moment; The half of the initial seed clusters and the other half of the initial seed clusters are merged to obtain the initial seed clusters at each moment.

7. The method for dynamic multi-objective optimization control of a sewage treatment process based on a neural network according to claim 3 is characterized in that: The step of determining the optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration in the target effluent water quality model and the target energy consumption model comprises: Obtain the dissolved oxygen concentration and optimal nitrate nitrogen concentration solution set in the target effluent water quality model and the target energy consumption model; Based on the inflection point calculation formula, the inflection point in the solution set of dissolved oxygen concentration and optimal nitrate nitrogen concentration is calculated; The optimal dissolved oxygen concentration and the optimal nitrate nitrogen concentration are determined according to the inflection point.

8. A neural network-based dynamic multi-objective optimization control method for sewage treatment process according to claim 7, characterized in that: The inflection point calculation formula is: The solution with the largest Dis value is the inflection point, where a, b, and c are coefficients determined by the two boundary points in the non-dominated solution set.

9. The method for dynamic multi-objective optimization control of sewage treatment process based on neural network according to claim 1 is characterized in that: The single neuron adaptive controller is: K1 is the neuron scale coefficient, satisfying K1>0, w i (t) is the weight coefficient.

10. The method for dynamic multi-objective optimization control of sewage treatment process based on neural network according to claim 1, characterized in that: The control formula is: Among them, γ i >0 is the learning rate of the neuron, v1(t), v2(t), v3(t) are the inputs of the neuron, Δu(t)=[ΔK L a5(t),ΔQ a (t)] T , ΔK L a5(t) is the change in oxygen transfer coefficient of the fifth subzone at time t, ΔQ a (t) is the change in the reflux flow at time t, e(t) = z * (t)-z(t) is the control error between the actual output and the optimized setting value, z * (t) = [S O * (t),S NO * (t)] is the optimal setting value of dissolved oxygen concentration and nitrate nitrogen concentration at time t, z(t) = [S O (t),S NO (t)] is the actual output dissolved oxygen concentration and nitrate nitrogen concentration at time t.

Images