CN117762012A - Knowledge and data hybrid driving urban sewage treatment process optimization control method - Google Patents

Abstract

The invention provides a knowledge and data mixed driving urban sewage treatment process optimization control method, which realizes the dissolution of oxygenS _O And nitrate nitrogenS _NO The optimal control of the concentration achieves the purposes of improving the water quality of the effluent and reducing the energy consumption of operation. The method designs a multi-objective optimization model of the urban sewage treatment process, and establishes an optimization model of the effluent quality and the operation energy consumption by using a radial basis function; the designed knowledge-driven method for solving the optimal control set point of the urban sewage treatment process realizes the rapid solving of the optimal set point, and improves the speed and the precision of the optimal set point of the process variable; and a proportional-integral-derivative controller is used for controlling the process variable to track and optimize the set value in real time, so that stable tracking control is realized.

Description

Knowledge and data hybrid driving urban sewage treatment process optimization control method

Technical Field

Aiming at the problem of optimizing operation of the urban sewage treatment process, the invention designs a knowledge and data hybrid-driven urban sewage treatment process optimizing control method, establishes an urban sewage treatment process optimizing control target model based on a radial basis function, designs a knowledge-driven urban sewage treatment process optimizing control set value solving method to obtain an optimizing set point, and uses a proportional-integral-differential controller to track and control the optimizing set point. The invention not only belongs to the water research field, but also belongs to the intelligent optimization control field, and has great significance in ensuring the water quality of effluent and reducing energy consumption at the same time, and high-efficiency stable operation of the urban sewage treatment process.

Background

With the increasing population of cities, the scale of urban sewage treatment is gradually increased, and the problem of urban domestic sewage treatment is increasingly prominent. The urban sewage treatment can purify water quality, improve ecological environment, promote water resource circulation and has important significance for urban development. As a typical high-energy-consumption industry, urban sewage treatment plants have increasingly high energy consumption for operation in order to ensure the quality of effluent. In order to improve the running effect and efficiency of urban sewage treatment, an optimal control strategy has been widely applied to the urban sewage treatment process.

The aim of the optimization control of the urban sewage treatment process is to ensure that the effluent quality reaches the standard and reduce the running energy consumption. However, the operation mechanism of the urban sewage treatment process is complex, and the effluent quality and the operation energy consumption have strong conflict. Therefore, balancing the relation between the water quality of the effluent and the operation energy consumption, and reducing the operation energy consumption while ensuring that the water quality of the effluent reaches the standard is an important research topic. In the process of establishing the sewage quality and energy consumption model, as the dynamics of the urban sewage treatment process have complex nonlinearity, the mechanism model is difficult to accurately express and optimize the target model. Therefore, it is of great importance to accurately describe the optimization objective of the urban sewage treatment process by adopting a data-based modeling method. In addition, multiple objectives in the municipal wastewater treatment process conflict with each other, affecting the quality of the solution of the optimization set point. Meanwhile, the goal is dynamic change in the urban sewage treatment process, and the speed of solving the optimal set point greatly influences the performance of optimal control. Therefore, the design of the optimal control method based on knowledge and data not only can improve the quality and speed of solving the optimal set point, reduce the energy consumption while guaranteeing the water quality of the effluent, but also is a key for the stable and efficient operation of the sewage treatment process.

According to the invention, through analyzing the characteristics of the urban sewage treatment process, the running performance index comprising the effluent quality and the running energy consumption is established, the knowledge-driven method for solving the optimal control set value of the urban sewage treatment process is designed, the speed and the precision for solving the optimal set value are improved, and the efficient and stable running of the sewage treatment process is realized.

Disclosure of Invention

The invention provides an optimization control method for the urban sewage treatment process driven by mixed knowledge and data, which reduces energy consumption while guaranteeing the quality of effluent. The method designs an urban sewage treatment process optimization control target model based on a radial basis function in the urban sewage treatment process, designs a knowledge-driven urban sewage treatment process optimization control set value solving method, and uses a proportional-integral-derivative controller to control a process variable to track an optimization set value in real time.

The invention adopts the following technical scheme and implementation steps:

1. the knowledge and data mixed driving urban sewage treatment process optimization control method specifically comprises the following steps:

(1) Establishing an urban sewage treatment process optimization control target model based on radial basis function

Taking the running energy consumption and the effluent quality of the urban sewage treatment process as optimization targets, and establishing an urban sewage treatment process optimization control target model:

min J(t)＝[J ₁ (t),J ₂ (t)] (1)

wherein J (t) is an urban sewage treatment process optimization control target model at the moment t, J ₁ (t) is a model of the effluent quality of the urban sewage treatment process at the moment t, J ₂ (t) an energy consumption model for running the urban sewage treatment process at the moment t;

water quality model J ₁ (t) is:

wherein W is _1,h (t) the connection weight of the h kernel function of the water quality of the outlet water at the moment t and the initial value is [0,1]Random number x of (x) ₁ (t)＝[S _O (t),S _NO (t),S _NH (t),SS(t)]，S _O (t) dissolved oxygen concentration at time t, S _NO (t) is the nitrate nitrogen concentration at the moment t, S _NH (t) is ammonia nitrogen concentration at t moment, SS (t) is suspended matter concentration at t moment,the central value of the h kernel function of the water quality of the water discharged at the moment t is 0,1]Random number b of (b) _1,h (t) the width value of the h kernel function of the water quality of the outlet water at the moment t and the initial value is [0,1]Updating the effluent quality model J ₁ Parameters of (t):

wherein W is _1,h (t+1) is the connection weight of the h kernel function of the water quality of the effluent at the moment t+1,the central value of the h kernel function of the water quality of the water discharged at the time t+1, b _1,h (t+1) is the width value of the h kernel function of the water quality of the effluent at the moment t+1;

operating energy consumption model J ₂ (t) is:

wherein W is _2,h (t) the connection weight of the h kernel function of the energy consumption operated at the moment t and the initial value is [0,1 ]]Random number x of (x) ₂ (t)＝[S _O (t),S _NO (t),MLSS(t)]MLSS (t) is the concentration of the mixed suspension at time t,the central value of the h kernel function for the energy consumption of t time operation is 0,1]Random number b of (b) _2,h (t) the width value of the h kernel function of the operation energy consumption at the moment t and the initial value is [0,1 ]]Updating the running energy consumption model J ₂ Parameters of (t):

wherein W is _2,h (t+1) is the connection weight of the h kernel function of the running energy consumption at the time of t+1,the central value of the h kernel function of the energy consumption of the operation at the time t+1, b _2,h (t+1) is the width value of the h kernel function of the running energy consumption at the time t+1;

(2) Design knowledge driven method for solving optimization control set value in urban sewage treatment process

Establishing a knowledge-guided global optimal particle selection mechanism: setting the total iteration number of solving the optimized set value as kappa _max =500, particle population size Λ=50, position and velocity of particles are:

z _t,n (κ)＝[z _t,n,1 (κ),z _t,n,2 (κ)] (10)

v _t,n (κ)＝[v _t,n,1 (κ),v _t,n,2 (κ)] (11)

wherein z is _t,n (kappa) is the position vector from the nth particle to the kth generation at time t, z _t,n,1 (kappa) is z _t,n First component of (kappa) and z _t,n,2 (kappa) is z _t,n The second component of (kappa), v _t,n (kappa) is the velocity vector from the nth particle evolution to the kth generation at time t, v _t,n,1 (kappa) is v _t,n The first component of (kappa), v _t,n,2 (kappa) is v _t,n A second component of (κ); the individual optimal positions for each particle are calculated as:

wherein p is _t,n (kappa+1) the individual optimal position of the nth particle evolving to the kappa+1 generation at time t, p _t,n (kappa) is the individual optimal position at time t at which the nth particle evolved to the kth generation, z _t,n (kappa+1) is a position vector at time t for the nth particle to evolve to the kappa+1 generation, f representing the dominance and defined as:

if it isOr->Then

z _t,n (κ+1)fp _t,n (kappa) (13) wherein J ₁ (z _t,n (kappa+1)) is z _t,n (kappa+1) effluent quality, J ₂ (z _t,n (kappa+1)) is z _t,n Operational energy consumption (κ+1), J ₁ (p _t,n (kappa)) is p _t,n (kappa) effluent quality, J ₂ (p _t,n (kappa)) is p _t,n (κ) operational energy consumption; non-dominant solution of particle evolution to kth generation at time t is stored in archive A _t In (κ), A _t (κ)＝{z ^* _t,1 (κ),…,z ^* _t,d (κ),...,z ^* _t,D (κ)}，z ^* _t,d (κ) is the D-th non-dominant solution evolving from time t to the kth generation, d=50 is a _t Size of (κ);

global optimum position g _t (kappa) updating g for optimal position of all particles of population evolution to kth generation at time t _t (κ)：

s(z ^* _t,d (kappa)) is the non-dominant solution z of the evolution of the population to the kth generation at time t ^* _t,d The sum of the guided knowledge of (κ) calculates s (z ^* _t,d (κ)) is:

z ^* _t,d′ (kappa) is archive A _t The d' th non-dominant solution in (kappa),is a non-dominant solution z ^* _t,d′ Normalized fitness value of (κ) mth target, +.>Is a non-dominant solution z ^* _t,d Normalized fitness value of (κ) mth target, calculateThe method comprises the following steps:

wherein J is ^min _m (z ^* _t (kappa)) is archive A _t Minimum value of mth target fitness value of non-dominant solution in (κ), J ^max _m (z ^* _t (kappa)) is archive A _t The maximum value of the mth target fitness value of the non-dominant solution in (κ);

update mode of particle speed and position of knowledge decision: the development knowledge of all particles from population evolution to the kth generation at time t is:

wherein, tau is the recorded historical iteration times, tau is more than or equal to 1 and less than or equal to 5,individual optimal p for the nth particle to evolve to the kth- τ+1st generation at time t _t,n Normalized fitness value of (κ - τ+1) mth target, +.>Individual optimal p for the evolution of the nth particle to the kth- τ generation at time t _t,n Normalized fitness value of (κ - τ+1) mth target;

when E is _t (κ)≤E _t At (κ+1), the velocity of the particles is updated as:

v _t,n (κ+1)＝0.7v _t,n (κ)+0.5μ ₁ (p _t,n (κ)-z _t,n (κ))+0.5μ ₂ (g _t (κ)-z _t,n (κ)) (18)

wherein v is _t,n (kappa+1) is the velocity vector, mu, of the nth particle evolution to the kappa+1 generation at time t ₁ Random numbers are evolved for individuals and the value ranges are [0,1 ]]，μ ₂ Random numbers are evolved for the population and the value ranges are [0,1]；

When E is _t (κ)>E _t At (κ+1), the velocity of the particles is updated as:

v _t,n (κ+1)＝0.7v _t,n (κ)+0.5μ ₁ (p _t,n (κ)-z _t,n (κ))+0.5μ ₂ (g _t (κ)-z _t,n (κ))+0.5μ ₃ K _t,n (κ)(19)

wherein mu ₃ Knowledge random numbers are developed for population evolution and the value range is [0,1 ]]，K _t,n (κ)＝[p _t,n (κ)-p _t,n (κ-1),…,p _t,n (κ-4)-p _t,n (κ-5)]For time t particle n evolve toThe evolution direction difference vector of the individual optimal position of the kth generation for 5 successive generations;

the location of the particles is updated as:

z _t,n (κ+1)＝z _t,n (κ)+v _t,n (κ+1) (20)

wherein z is _t,n (kappa+1) is the position at which the nth particle evolved to the kappa+1 generation at time t;

detecting whether the evolution process reaches a stopping condition: if algebra of evolution kappa<κ _max The algebra kappa increases by 1 and updates the speed and position of the particles again; if algebra k=k _max Terminating the evolution process from the kth _max Archive A of the generation _t (κ _max )＝[z ^* _t,1 (κ _max ),…,z ^* _t,d (κ _max ),...,z ^* _t,D (κ _max )]Randomly selecting a solution as the optimal set value y of the process variable ^* (t)＝[S ^* _O (t),S ^* _NO (t)]，S ^* _O (t) is the optimized set value of dissolved oxygen in the aerobic reactor at the moment t, S ^* _NO (t) is a nitrate nitrogen optimization set value in the anoxic reactor at the moment t;

(3) Optimizing set value for tracking and controlling urban sewage treatment process

Solving the optimal set point by using multivariable proportional-integral-derivative control algorithm [ S ^* _O (t),S ^* _NO (t)]Tracking control is performed:

wherein Δu (t) = [ Δk ] _L a(t),ΔQ _a (t)] ^T For the operating variable matrix at time t, ΔK _L a (t) is the variation of the transfer coefficient of dissolved oxygen at time t, ΔQ _a (t) is the variation of the circulation flow rate at the time t, and C= [20,10]As a proportionality coefficient, l= [5,3]As an integral time constant, f= [2,1]E (t) =y as differential time constant ^* (t) -y (t) is a control error vector at time t, y (t) =[S _O (t),S _NO (t)]Is the actual dissolved oxygen concentration S at the time t _O (t) and nitrate nitrogen concentration S _NO (t)；

Adjusting the transfer coefficient of dissolved oxygen and the internal reflux amount by using Deltau (t):

K _L a(t+1)＝K _L a(t)+△K _L a(t) (22)

Q _a (t+1)＝Q _a (t)+△Q _a (t) (23)

wherein K is _L a (t+1) is the dissolved oxygen transfer coefficient, K, at time t+1 _L a (t) is the dissolved oxygen transfer coefficient at time t, Q _a (t+1) is the internal reflux amount at time t+1, Q _a (t) is the internal reflux amount at time t;

the input of the urban sewage treatment control system at the time t is the change delta K of the dissolved oxygen transfer coefficient _L a (t) and the amount of change Δq of the internal reflux amount _a (t) by operating the dissolved oxygen transfer coefficient K _L a (t) realizes the set value S of the concentration of the dissolved oxygen ^* _O Tracking control of (t) by operating the internal reflux quantity Q _a (t) realizing the set value S of the concentration of the paranitronitrogen ^* _NO Tracking control of (t), the dissolved oxygen concentration is adjusted to S ^* _O (t) the nitrate nitrogen concentration is adjusted to S ^* _NO (t)。

Drawings

FIG. 1 shows the dissolved oxygen concentration S of the present invention _O Optimizing a control effect diagram and an error diagram;

FIG. 2 shows the nitrate nitrogen concentration S of the present invention _NO Optimizing a control effect diagram and an error diagram.

Detailed Description

1. The knowledge and data mixed driving urban sewage treatment process optimization control method specifically comprises the following steps:

(1) Establishing an urban sewage treatment process optimization control target model based on radial basis function

Taking the running energy consumption and the effluent quality of the urban sewage treatment process as optimization targets, and establishing an urban sewage treatment process optimization control target model:

min J(t)＝[J ₁ (t),J ₂ (t)] (24)

wherein J (t) is an urban sewage treatment process optimization control target model at the moment t, J ₁ (t) is a model of the effluent quality of the urban sewage treatment process at the moment t, J ₂ (t) an energy consumption model for running the urban sewage treatment process at the moment t;

water quality model J ₁ (t) is:

wherein W is _1,h (t) the connection weight of the h kernel function of the water quality of the outlet water at the moment t and the initial value is [0,1]Random number x of (x) ₁ (t)＝[S _O (t),S _NO (t),S _NH (t),SS(t)]，S _O (t) dissolved oxygen concentration at time t, S _NO (t) is the nitrate nitrogen concentration at the moment t, S _NH (t) is ammonia nitrogen concentration at t moment, SS (t) is suspended matter concentration at t moment,the central value of the h kernel function of the water quality of the water discharged at the moment t is 0,1]Random number b of (b) _1,h (t) the width value of the h kernel function of the water quality of the outlet water at the moment t and the initial value is [0,1]Updating the effluent quality model J ₁ Parameters of (t):

wherein W is _1,h (t+1) is the connection weight of the h kernel function of the water quality of the effluent at the moment t+1,the central value of the h kernel function of the water quality of the water discharged at the time t+1, b _1,h (t+1) is the width value of the h kernel function of the water quality of the effluent at the moment t+1;

operating energy consumption model J ₂ (t) is:

wherein W is _2,h (t) the connection weight of the h kernel function of the energy consumption operated at the moment t and the initial value is [0,1 ]]Random number x of (x) ₂ (t)＝[S _O (t),S _NO (t),MLSS(t)]MLSS (t) is the concentration of the mixed suspension at time t,the central value of the h kernel function for the energy consumption of t time operation is 0,1]Random number b of (b) _2,h (t) the width value of the h kernel function of the operation energy consumption at the moment t and the initial value is [0,1 ]]Updating the running energy consumption model J ₂ Parameters of (t):

wherein W is _2,h (t+1) is the connection weight of the h kernel function of the running energy consumption at the time of t+1,the central value of the h kernel function of the energy consumption of the operation at the time t+1, b _2,h (t+1) is the width value of the h kernel function of the running energy consumption at the time t+1;

(2) Design knowledge driven method for solving optimization control set value in urban sewage treatment process

Establishing a knowledge-guided global optimal particle selection mechanism: setting the total iteration number of solving the optimized set value as kappa _max =500, particle population size Λ=50, position and velocity of particles are:

z _t,n (κ)＝[z _t,n,1 (κ),z _t,n,2 (κ)] (33)

v _t,n (κ)＝[v _t,n,1 (κ),v _t,n,2 (κ)] (34)

wherein z is _t,n (kappa) is the position vector from the nth particle to the kth generation at time t, z _t,n,1 (kappa) is z _t,n First component of (kappa) and z _t,n,2 (kappa) is z _t,n The second component of (kappa), v _t,n (kappa) is the velocity vector from the nth particle evolution to the kth generation at time t, v _t,n,1 (kappa) is v _t,n The first component of (kappa), v _t,n,2 (kappa) is v _t,n A second component of (κ); the individual optimal positions for each particle are calculated as:

wherein p is _t,n (kappa+1) the individual optimal position of the nth particle evolving to the kappa+1 generation at time t, p _t,n (kappa) is the individual optimal position at time t at which the nth particle evolved to the kth generation, z _t,n (kappa+1) is a position vector at time t for the nth particle to evolve to the kappa+1 generation, f representing the dominance and defined as:

if it isOr->Then

z _t,n (κ+1)fp _t,n (kappa) (36) wherein J ₁ (z _t,n (kappa+1)) is z _t,n (kappa+1) effluent quality, J ₂ (z _t,n (kappa+1)) is z _t,n Operational energy consumption (κ+1), J ₁ (p _t,n (kappa)) is p _t,n (kappa) effluent quality, J ₂ (p _t,n (kappa)) is p _t,n (κ) operational energy consumption; non-dominant solution of particle evolution to kth generation at time t is stored in archive A _t In (κ), A _t (κ)＝{z ^* _t,1 (κ),…,z ^* _t,d (κ),...,z ^* _t,D (κ)}，z ^* _t,d (κ) is the D-th non-dominant solution evolving from time t to the kth generation, d=50 is a _t Size of (κ);

global optimum position g _t (kappa) updating g for optimal position of all particles of population evolution to kth generation at time t _t (κ)：

s(z ^* _t,d (kappa)) is the non-dominant solution z of the evolution of the population to the kth generation at time t ^* _t,d The sum of the guided knowledge of (κ) calculates s (z ^* _t,d (κ)) is:

z ^* _t,d′ (kappa) is archive A _t The d' th non-dominant solution in (kappa),is a non-dominant solution z ^* _t,d′ Normalized fitness value of (κ) mth target, +.>Is a non-dominant solution z ^* _t,d Normalized fitness value of (κ) mth target, calculateThe method comprises the following steps:

wherein J is ^min _m (z ^* _t (kappa)) is archive A _t Minimum value of mth target fitness value of non-dominant solution in (κ), J ^max _m (z ^* _t (kappa)) is archive A _t The maximum value of the mth target fitness value of the non-dominant solution in (κ);

update mode of particle speed and position of knowledge decision: the development knowledge of all particles from population evolution to the kth generation at time t is:

wherein, tau is the recorded historical iteration times, tau is more than or equal to 1 and less than or equal to 5,individual optimal p for the nth particle to evolve to the kth- τ+1st generation at time t _t,n Normalized fitness value of (κ - τ+1) mth target, +.>Individual optimal p for the evolution of the nth particle to the kth- τ generation at time t _t,n Normalized fitness value of (κ - τ+1) mth target;

when E is _t (κ)≤E _t At (κ+1), the velocity of the particles is updated as:

v _t,n (κ+1)＝0.7v _t,n (κ)+0.5μ ₁ (p _t,n (κ)-z _t,n (κ))+0.5μ ₂ (g _t (κ)-z _t,n (κ)) (41)

wherein v is _t,n (kappa+1) is the velocity vector, mu, of the nth particle evolution to the kappa+1 generation at time t ₁ Evolving random numbers for individuals and taking valuesIn the range of [0,1 ]]，μ ₂ Random numbers are evolved for the population and the value ranges are [0,1]；

When E is _t (κ)>E _t At (κ+1), the velocity of the particles is updated as:

v _t,n (κ+1)＝0.7v _t,n (κ)+0.5μ ₁ (p _t,n (κ)-z _t,n (κ))+0.5μ ₂ (g _t (κ)-z _t,n (κ))+0.5μ ₃ K _t,n (κ)(42)

wherein mu ₃ Knowledge random numbers are developed for population evolution and the value range is [0,1 ]]，K _t,n (κ)＝[p _t,n (κ)-p _t,n (κ-1),…,p _t,n (κ-4)-p _t,n (κ-5)]The evolution direction difference vector of 5 successive generations for the individual optimal positions from the particle n to the kth generation at the moment t;

the location of the particles is updated as:

z _t,n (κ+1)＝z _t,n (κ)+v _t,n (κ+1) (43)

wherein z is _t,n (kappa+1) is the position at which the nth particle evolved to the kappa+1 generation at time t;

detecting whether the evolution process reaches a stopping condition: if algebra of evolution kappa<κ _max The algebra kappa increases by 1 and updates the speed and position of the particles again; if algebra k=k _max Terminating the evolution process from the kth _max Archive A of the generation _t (κ _max )＝[z ^* _t,1 (κ _max ),…,z ^* _t,d (κ _max ),...,z ^* _t,D (κ _max )]Randomly selecting a solution as the optimal set value y of the process variable ^* (t)＝[S ^* _O (t),S ^* _NO (t)]，S ^* _O (t) is the optimized set value of dissolved oxygen in the aerobic reactor at the moment t, S ^* _NO (t) is a nitrate nitrogen optimization set value in the anoxic reactor at the moment t;

(3) Optimizing set value for tracking and controlling urban sewage treatment process

Solving the optimal set point by using multivariable proportional-integral-derivative control algorithm [ S ^* _O (t),S ^* _NO (t)]Tracking control is performed:

wherein Δu (t) = [ Δk ] _L a(t),ΔQ _a (t)] ^T For the operating variable matrix at time t, ΔK _L a (t) is the variation of the transfer coefficient of dissolved oxygen at time t, ΔQ _a (t) is the variation of the circulation flow rate at the time t, and C= [20,10]As a proportionality coefficient, l= [5,3]As an integral time constant, f= [2,1]E (t) =y as differential time constant ^* (t) -y (t) is a control error vector at time t, y (t) = [ S ] _O (t),S _NO (t)]Is the actual dissolved oxygen concentration S at the time t _O (t) and nitrate nitrogen concentration S _NO (t)；

Adjusting the transfer coefficient of dissolved oxygen and the internal reflux amount by using Deltau (t):

K _L a(t+1)＝K _L a(t)+△K _L a(t) (45)

Q _a (t+1)＝Q _a (t)+△Q _a (t) (46)

wherein K is _L a (t+1) is the dissolved oxygen transfer coefficient, K, at time t+1 _L a (t) is the dissolved oxygen transfer coefficient at time t, Q _a (t+1) is the internal reflux amount at time t+1, Q _a (t) is the internal reflux amount at time t;

the input of the urban sewage treatment control system at the time t is the change delta K of the dissolved oxygen transfer coefficient _L a (t) and the amount of change Δq of the internal reflux amount _a (t) by operating the dissolved oxygen transfer coefficient K _L a (t) realizes the set value S of the concentration of the dissolved oxygen ^* _O Tracking control of (t) by operating the internal reflux quantity Q _a (t) realizing the set value S of the concentration of the paranitronitrogen ^* _NO Tracking control of (t), the dissolved oxygen concentration is adjusted to S ^* _O (t) the nitrate nitrogen concentration is adjusted to S ^* _NO (t)。

Claims (1)

1. A knowledge and data hybrid driving urban sewage treatment process optimization control method is characterized by comprising the following steps of: establishing an urban sewage treatment process optimization control target model based on a radial basis function, designing a knowledge-driven method for solving an urban sewage treatment process optimization control set value, and tracking and controlling the urban sewage treatment process optimization set value, wherein the method specifically comprises the following steps of:

(1) Establishing an urban sewage treatment process optimization control target model based on radial basis function

Taking the running energy consumption and the effluent quality of the urban sewage treatment process as optimization targets, and establishing an urban sewage treatment process optimization control target model:

minJ(t)＝[J ₁ (t),J ₂ (t)] (1)

wherein J (t) is an urban sewage treatment process optimization control target model at the moment t, J ₁ (t) is a model of the effluent quality of the urban sewage treatment process at the moment t, J ₂ (t) an energy consumption model for running the urban sewage treatment process at the moment t;

water quality model J ₁ (t) is:

wherein W is _1,h (t) the connection weight of the h kernel function of the water quality of the outlet water at the moment t and the initial value is [0,1]Random number x of (x) ₁ (t)＝[S _O (t),S _NO (t),S _NH (t),SS(t)]，S _O (t) dissolved oxygen concentration at time t, S _NO (t) is the nitrate nitrogen concentration at the moment t, S _NH (t) is ammonia nitrogen concentration at t moment, SS (t) is suspended matter concentration at t moment,the central value of the h kernel function of the water quality of the water discharged at the moment t is 0,1]Random number b of (b) _1,h (t) the width value of the h kernel function of the water quality of the outlet water at the moment t and the initial value is [0,1]Updating the effluent quality model J ₁ Parameters of (t):

wherein W is _1,h (t+1) is the connection weight of the h kernel function of the water quality of the effluent at the moment t+1,the central value of the h kernel function of the water quality of the water discharged at the time t+1, b _1,h (t+1) is the width value of the h kernel function of the water quality of the effluent at the moment t+1;

operating energy consumption model J ₂ (t) is:

wherein W is _2,h (t) the connection weight of the h kernel function of the energy consumption operated at the moment t and the initial value is [0,1 ]]Random number x of (x) ₂ (t)＝[S _O (t),S _NO (t),MLSS(t)]MLSS (t) is the concentration of the mixed suspension at time t,the central value of the h kernel function for the energy consumption of t time operation is 0,1]Random number b of (b) _2,h (t) the width value of the h kernel function of the operation energy consumption at the moment t and the initial value is [0,1 ]]Updating the running energy consumption model J ₂ Parameters of (t):

wherein W is _2,h (t+1) is the connection weight of the h kernel function of the running energy consumption at the time of t+1,the central value of the h kernel function of the energy consumption of the operation at the time t+1, b _2,h (t+1) is the width value of the h kernel function of the running energy consumption at the time t+1;

(2) Design knowledge driven method for solving optimization control set value in urban sewage treatment process

Establishing a knowledge-guided global optimal particle selection mechanism: setting the total iteration number of solving the optimized set value as kappa _max =500, particle population size Λ=50, position and velocity of particles are:

z _t,n (κ)＝[z _t,n,1 (κ),z _t,n,2 (κ)] (10)

v _t,n (κ)＝[v _t,n,1 (κ),v _t,n,2 (κ)] (11)

wherein z is _t,n (kappa) is the position vector from the nth particle to the kth generation at time t, z _t,n,1 (kappa) is z _t,n First component of (kappa) and z _t,n,2 (kappa) is z _t,n The second component of (kappa), v _t,n (kappa) is the velocity vector from the nth particle evolution to the kth generation at time t, v _t,n,1 (kappa) is v _t,n The first component of (kappa), v _t,n,2 (kappa) is v _t,n A second component of (κ); calculating individual best for each particleThe preferred positions are:

wherein p is _t,n (kappa+1) the individual optimal position of the nth particle evolving to the kappa+1 generation at time t, p _t,n (kappa) is the individual optimal position at time t at which the nth particle evolved to the kth generation, z _t,n (kappa+1) is a position vector at time t for the nth particle to evolve to the kappa+1 generation, f representing the dominance and defined as:

if it isOr->Then

z _t,n (κ+1)f p _t,n (kappa) (13) wherein J ₁ (z _t,n (kappa+1)) is z _t,n (kappa+1) effluent quality, J ₂ (z _t,n (kappa+1)) is z _t,n Operational energy consumption (κ+1), J ₁ (p _t,n (kappa)) is p _t,n (kappa) effluent quality, J ₂ (p _t,n (kappa)) is p _t,n (κ) operational energy consumption; non-dominant solution of particle evolution to kth generation at time t is stored in archive A _t In (κ), A _t (κ)＝{z ^* _t,1 (κ),…,z ^* _t,d (κ),...,z ^* _t,D (κ)}，z ^* _t,d (κ) is the D-th non-dominant solution evolving from time t to the kth generation, d=50 is a _t Size of (κ);

global optimum position g _t (kappa) updating g for optimal position of all particles of population evolution to kth generation at time t _t (κ)：

s(z ^* _t,d (kappa)) is the non-dominant solution z of the evolution of the population to the kth generation at time t ^* _t,d The sum of the guided knowledge of (κ) calculates s (z ^* _t,d (κ)) is:

z ^* _t,d′ (kappa) is archive A _t The d' th non-dominant solution in (kappa),is a non-dominant solution z ^* _t,d′ Normalized fitness value of (κ) mth target, +.>Is a non-dominant solution z ^* _t,d Normalized fitness value of (κ) mth target, calculateThe method comprises the following steps:

wherein J is ^min _m (z ^* _t (kappa)) is archive A _t Minimum value of mth target fitness value of non-dominant solution in (κ), J ^max _m (z ^* _t (kappa)) is archive A _t The maximum value of the mth target fitness value of the non-dominant solution in (κ);

update mode of particle speed and position of knowledge decision: the development knowledge of all particles from population evolution to the kth generation at time t is:

wherein, tau is the recorded historical iteration times, tau is more than or equal to 1 and less than or equal to 5,individual optimal p for the nth particle to evolve to the kth- τ+1st generation at time t _t,n Normalized fitness value of (κ - τ+1) mth target, +.>Individual optimal p for the evolution of the nth particle to the kth- τ generation at time t _t,n Normalized fitness value of (κ - τ+1) mth target;

when E is _t (κ)≤E _t At (κ+1), the velocity of the particles is updated as:

v _t,n (κ+1)＝0.7v _t,n (κ)+0.5μ ₁ (p _t,n (κ)-z _t,n (κ))+0.5μ ₂ (g _t (κ)-z _t,n (κ)) (18)

wherein v is _t,n (kappa+1) is the velocity vector, mu, of the nth particle evolution to the kappa+1 generation at time t ₁ Random numbers are evolved for individuals and the value ranges are [0,1 ]]，μ ₂ Random numbers are evolved for the population and the value ranges are [0,1]；

When E is _t (κ)>E _t At (κ+1), the velocity of the particles is updated as:

v _t,n (κ+1)＝0.7v _t,n (κ)+0.5μ ₁ (p _t,n (κ)-z _t,n (κ))+0.5μ ₂ (g _t (κ)-z _t,n (κ))+0.5μ ₃ K _t,n (κ) (19)

wherein mu ₃ Knowledge random numbers are developed for population evolution and the value range is [0,1 ]]，K _t,n (κ)＝[p _t,n (κ)-p _t,n (κ-1),…,p _t,n (κ-4)-p _t,n (κ-5)]The evolution direction difference vector of 5 successive generations for the individual optimal positions from the particle n to the kth generation at the moment t;

the location of the particles is updated as:

z _t,n (κ+1)＝z _t,n (κ)+v _t,n (κ+1) (20)

wherein z is _t,n (kappa+1) is the position at which the nth particle evolved to the kappa+1 generation at time t;

detecting whether the evolution process reaches a stopping condition: if algebra of evolution kappa<κ _max The algebra kappa increases by 1 and updates the speed and position of the particles again; if algebra k=k _max Terminating the evolution process from the kth _max Archive A of the generation _t (κ _max )＝[z ^* _t,1 (κ _max ),…,z ^* _t,d (κ _max ),...,z ^* _t,D (κ _max )]Randomly selecting a solution as the optimal set value y of the process variable ^* (t)＝[S ^* _O (t),S ^* _NO (t)]，S ^* _O (t) is the optimized set value of dissolved oxygen in the aerobic reactor at the moment t, S ^* _NO (t) is a nitrate nitrogen optimization set value in the anoxic reactor at the moment t;

(3) Optimizing set value for tracking and controlling urban sewage treatment process

Solving the optimal set point by using multivariable proportional-integral-derivative control algorithm [ S ^* _O (t),S ^* _NO (t)]Tracking control is performed:

wherein Δu (t) = [ Δk ] _L a(t),ΔQ _a (t)] ^T For the operating variable matrix at time t, ΔK _L a (t) is the variation of the transfer coefficient of dissolved oxygen at time t, ΔQ _a (t) is the variation of the circulation flow rate at the time t, and C= [20,10]As a proportionality coefficient, l= [5,3]As an integral time constant, f= [2,1]E (t) =y as differential time constant ^* (t) -y (t) is a control error vector at time t, y (t) = [ S ] _O (t),S _NO (t)]Is the actual dissolved oxygen concentration S at the time t _O (t) and nitrate nitrogen concentration S _NO (t)；

Adjusting the transfer coefficient of dissolved oxygen and the internal reflux amount by using Deltau (t):

K _L a(t+1)＝K _L a(t)+△K _L a(t) (22)

Q _a (t+1)＝Q _a (t)+△Q _a (t) (23)

wherein K is _L a (t+1) is the dissolved oxygen transfer coefficient, K, at time t+1 _L a (t) is the dissolved oxygen transfer coefficient at time t, Q _a (t+1) is the internal reflux amount at time t+1, Q _a (t) is the internal reflux amount at time t;

the input of the urban sewage treatment control system at the time t is the change delta K of the dissolved oxygen transfer coefficient _L a (t) and the amount of change Δq of the internal reflux amount _a (t) by operating the dissolved oxygen transfer coefficient K _L a (t) realizes the set value S of the concentration of the dissolved oxygen ^* _O Tracking control of (t) by operating the internal reflux quantity Q _a (t) realizing the set value S of the concentration of the paranitronitrogen ^* _NO Tracking control of (t), the dissolved oxygen concentration is adjusted to S ^* _O (t) the nitrate nitrogen concentration is adjusted to S ^* _NO (t)。