US 12,417,371 B2 · Optimal control method for wastewater treatment process based on self-adjusting multi-task particle swarm optimization

Abstract

An optimal control method for wastewater treatment process (WWTP) based on a self-adjusting multi-task particle swarm optimization (SA-MTPSO) algorithm belongs to the field of WWTP. To balance the relationship between the effluent water quality (EQ) and energy consumption (EC) and achieve optimization online quickly, the invention establishes a data-based multi-task optimization model for WWTP to describe the relationship between the control variables and EQ, EC. Then, the SA-MTPSO algorithm is adopted to solve the optimal set-points of the nitrate nitrogen and dissolved oxygen concentration for WWTP. The PID controller is used to track the optimal set-points, so as to reduce EC while ensuring EQ, and realize the online optimal control of WWTP.

An optimal control method for wastewater treatment process based on a self-adjusting multi-task particle swarm optimization algorithm, comprising the following steps:

(A) collecting, by a processor, process data:
dissolved oxygen is an important indicator of a reaction nitrification process, which is mainly detected by detection instrument of dissolved oxygen;
nitrate nitrogen determines a process of denitrification rate, which is mainly detected by detection instrument of nitrate nitrogen;
and the process data will be transmitted to a programmable logic controller by fieldbus;
the programmable logic controller converts a signal into digital or analog quantity and communicates with a master computer through a transmission protocol;
the master computer obtains real-time data from the wastewater treatment process and stores the obtained real-time data in a Structured Query Language server database;
(B) designing, by the processor, a data-based multi-task optimization model for wastewater treatment process:
the data-based multi-task optimization model for wastewater treatment process adopts a data-based method to describe relationship between optimization set-points and effluent quality, energy consumption, including an objective task model based on a kernel function and an auxiliary task model based on polynomial regression;
(i) establishing the objective task model of effluent quality and energy consumption based on a kernel function:
f
1
(
t
)
=
B
1
(
t
)
+
∑
z
=
1
1
⁢
0
W
1
,
z
(
t
)
⁢
H
1
,
z
(
t
)
(
1
)
f
2
(
t
)
=
B
2
(
t
)
+
∑
z
=
1
1
⁢
0
W
2
,
z
(
t
)
⁢
H
2
,
z
(
t
)
(
2
)
where ƒ1(t) is the objective task model of effluent quality at time t, and ƒ2(t) is the objective task model of energy consumption at time t;
B1(t) is an output offset of ƒ1(t) at time t, within [−2, 2], B2(t) is an output offset of ƒ2(t) at time t, within [−2, 2];
W1,z(t) is a weight of zth kernel function of ƒ1(t) at time t, within [−3, 3], W2,z(t) is a weight of zth kernel function of ƒ2(t) at time t, within [−3, 3];
H1,z(t) is kernel functions related to f1(t) for the wastewater treatment process, and H2,z(t) is kernel functions related to f2(t) for the wastewater treatment process:
H1,z(t)=e−∥x(t)-c1,z(t)∥2/2s1,z(t)2 (3)
H2,z(t)=e−∥x(t)-c2,z(t)∥2/2s2,z(t)2 (4)
where x(t)=[SNO(t), SO(t)] is an input variable at time t, SNO(t) is anaerobic end nitrate nitrogen concentration by detection instrument of dissolved oxygen at time t, within [0, 2], unit:
mg/L, and SO(t) is a concentration of dissolved oxygen at an aerobic terminal by detection instrument of nitrate nitrogen at time t, within [0, 3], unit:
mg/L;
T is transpose of a vector or a matrix;
c1,z(t)=[c1,z,1(t), c1,z,2(t)]T is a center of zth kernel function of ƒ1(t), and value ranges of c1,z,1(t) and c1,z,2(t) are [−1, 1], c2,z(t)=[c2,z,1(t), c2,z,2(t)]T is a center of zth kernel function of ƒ2(t), and value ranges of c2,z,1(t) and c2,z,2(t) are [−1,1];
σ1,z(t) is a width of the zth kernel function of ƒ1(t) at time t, within [0, 2], σ2,z(t) is a width of the zth kernel function of ƒ2(t) at time t, within [0, 2];
(ii) training parameters c(t), σ(t) and W(t) of the objective task model for wastewater treatment process:
c
⁡
(
t
+
1
)
=
c
⁡
(
t
)
-
α
1
⁢
∂
e
1
⁢
u
(
t
)
∂
c
⁡
(
t
)
(
5
)
σ
⁡
(
t
+
1
)
=
σ
⁡
(
t
)
-
α
2
⁢
∂
e
1
⁢
u
(
t
)
∂
σ
⁡
(
t
)
(
6
)
W
⁡
(
t
+
1
)
=
W
⁡
(
t
)
-
α
3
⁢
∂
e
1
⁢
u
(
t
)
∂
W
⁡
(
t
)
(
7
)
where c(t)=[c1,z(t), c2,z(t)] is a center of a kernel function for the objective task model at time t, and c(t+1)=[c1,z(t+1), c2,z(t+1)] is a center of the kernel function for the objective task model at time (t+1);
σ(t)=[σ1,z(t), σ2,z(t)] is a width of the kernel function for the objective task model at time t, and σ(t+1)=[σ1,z(t+1), σ2,z(t+1)] is a width of the kernel function for the objective task model at time (t+1);
W(t)=[W1,z(t), W2,z(t)] is a weight of the kernel function for the objective task model at time t, and W(t+1)=[W1,z(t+1), W2,z(t+1)] is a weight of the kernel function for the objective task model at time (t+1);
α1 is a learning rate of a kernel function center for the objective task model within [0, 1], α2 is a learning rate of a kernel function width for the objective task model within [0, 1], α3 is a learning rate of the kernel function weight for the objective task model within [0, 1];
e1u(t)=y1(t)−yu(t) is a prediction error of the objective task model at time t, y1(t)=[ƒ1(t), ƒ2 (t)] is an output of the objective task model at time t, yu(t)=[EQ (t), EC(t)] is an actual output value of wastewater treatment process at time t, EQ(t) is an actual effluent quality value of wastewater treatment process at time t, EC(t) is an actual energy consumption value of wastewater treatment process at time t;
(iii) establishing an auxiliary task model of effluent quality and energy consumption in wastewater treatment process based on the following polynomial regression:
f
3
(
t
)
=
b
1
(
t
)
+
∑
q
=
0
2
A
1
,
q
(
t
)
⁢
(
x
⁡
(
t
)
)
q
(
8
)
f
4
(
t
)
=
b
2
(
t
)
+
∑
q
=
0
2
A
2
,
q
(
t
)
⁢
(
x
⁡
(
t
)
)
q
(
9
)
where ƒ3(t) is an auxiliary task model of effluent quality at time t, and ƒ4(t) is an auxiliary task model of energy consumption at time t;
b1(t) is an output offset of ƒ3(t) at time t, within [−2, 2], b2(t) is an output offset of ƒ4(t) at time t, within [−2, 2];
A1,q(t) is qth coefficient of ƒ3(t) at time t, within [−3, 3], A2,q(t) is qth coefficient of ƒ4(t) at time t, within [−3, 3];
(iv) training polynomial coefficient A (t) of an auxiliary task model for wastewater treatment process:
A
⁡
(
t
+
1
)
=
A
⁡
(
t
)
-
α
4
⁢
∂
e
2
⁢
u
(
t
)
∂
A
⁡
(
t
)
(
10
)
where A(t)=[A1,q(t), A2,q(t)] is polynomial coefficient of the auxiliary task model for wastewater treatment process at time t, and A(t+1)=[A1,q(t+1), A2,q(t+1)] is polynomial coefficient of the auxiliary task model for wastewater treatment process at time (t+1);
α4 is a learning rate of polynomial coefficient for the auxiliary task model within [0, 1];
e2u(t)=y2(t)−yu(t) is a prediction error of the auxiliary task model at time t, and y2(t)=[ƒ3(t),ƒ4(t)] is an output of the auxiliary task model at time t;
(v) establishing multi-task optimization model for wastewater treatment process:
minimize
⁢
F
⁡
(
t
)
=
[
F
1
(
t
)
,
F
2
(
t
)
]
(
11
)
{
F
1
(
t
)
=
[
f
1
(
t
)
,
f
2
(
t
)
]
F
2
(
t
)
=
[
f
3
(
t
)
,
f
4
(
t
)
]
(
12
)
where, F(t) is the multi-task optimization model of wastewater treatment process at time t, F1(t) is the objective task model of wastewater treatment process at time t, and F2(t) is the auxiliary task model of wastewater treatment process at time t;
(C) solving, by the processor, optimal set-points based on self-adjusting multi-task particle swarm optimization embedded in the master computer:
through knowledge transfer between an objective task and an auxiliary task, the self-adjusting multi-task particle swarm optimization algorithm achieves accelerated convergence of the objective task;
(i) setting total iteration times of self-adjusting multi-task particle swarm optimization as τmax=500, particle swarm size as N=100, the number of tasks as K=2, and initialize external archive U(0) as empty set;
(ii) an optimization objective of self-adjusting multi-task particle swarm optimization is the data-based multi-task optimization model for wastewater treatment process:
min F(t)=[F1(t), F2(t)];
(iii) position information of particles t(τ)=[SNOt(t), SOt(t)] is used as input in τ iteration at time t;
fitness and skill factors of particles are calculated, and particles are divided into different groups according to the skill factors;
fitness ranking of the particles are carried out;
(iv) evaluating knowledge validity carried by particles:
E
t
(
t
)
=
∑
m
=
1
20
(
F
t
m
(
t
)
-
F
_
t
(
t
)
)
⁢
(
g
t
m
(
t
)
-
g
t
_
(
t
)
)
s
⁡
(
F
t
(
t
)
)
⁢
s
⁡
(
g
t
(
t
)
)
(
13
)
where Et(τ) is the knowledge validity of τth iteration at time t, Fmt(t)=[Fm1,t(τ), Fm2t(τ)] is multi-task optimization fitness of mth particle in τiteration at time t, Fm1t(τ) is objective task fitness of the mth particle in τ iteration at time t, and Fm2t(τ) is auxiliary task fitness of the mth particle in τ iteration at time t;
Ft(τ)=[F1t(τ), F2t(τ)] is a mean of fitness for multi-task optimization in τ iteration at time t, F2t(τ) is a mean of fitness for the objective task in τ iteration at time t, F2t(τ) is a mean of fitness for the auxiliary task in τ iteration at time τ;
Ft(τ)=[F1t(τ), F2t(τ)] is a fitness of multitask optimization in τ iteration at time t, F1t(τ) is a fitness of the objective task in τ iteration at time t, F2t(τ) is the fitness of the auxiliary task in t iteration at time t;
gt(τ) is the distance between an objective task particle and a global optimal particle of the auxiliary task in τ iteration at time t, gmt(τ) is a distance between mth objective task particle and the global optimal particle of the auxiliary task in r iteration at time t, gt(τ) is a mean distance between the objective particle and the global optimal particle of the auxiliary task in τ iteration at time t;
(v) designing knowledge transfer strategy:
p
tg
*
(
t
)
=
{
p
tJ
(
t
)
E
t
(
t
)
≤
E
t
(
t
-
1
)
p
tJ
(
t
-
1
)
E
t
(
t
)
>
E
t
(
t
-
1
)
(
14
)
where P*tg(τ)=[p*tg,1(τ), p*tg,2(τ)] is a knowledge transfer term in τ iteration at time t, PtJ(τ)=[ptJ,1(τ), ptJ,2(τ)] is a global optimal solution of the auxiliary task in τ iteration at time t, PtJ(τ−1)=[ptJ,1(τ−1), ptJ,2(T−1)] is the global optimal solution of the auxiliary task in (τ−1) iteration at time t, and J is a solution with highest knowledge validity in a Pareto optimal solution set of the auxiliary task at time t:
J=argminj∈{1,2, . . . M}{Etj(t)} (15)
where Etj(τ) is the knowledge validity of Jth particle in the Pareto optimal solution set of the auxiliary task in t iteration at time t;
a particle velocity update formula:
vti(t+1)=wvti(t)+c1r1(pti(t))−xti(t)) +c2r2(Ptg(t)−xti(t))+c3r3(p*tg(t)−xti(t)) (16)
where vti(τ+1) is a velocity of ith particle in (τ+1) iteration at time t, and vti(τ) is the velocity of the ith particle in τ iteration at time t, xti(τ+1) is a position of the ith particle in (τ+1) iteration at time t, and xti(τ) is a position of the ith particle in τ iteration at time t, Pti(τ) is an individual optimal position of the ith particle in τ iteration at time t, and Ptg(τ) is a global optimal position of the ith particle in τ iteration at time t, ω=0.8 is an inertia weight;
c1=0.25 is an individual experience acceleration constant, c2=0.25 is a social experience acceleration constant, c3=0.25 is a knowledge transfer term acceleration constant;
r1 is an individual experience random number within [0, 1], r2 is a social experience random number within [0, 1], r3 is a knowledge transfer item random number within [0, 1];
(vi) by comparing the individual optimal position Pti(τ) with a solution of an archive Φt(τ−1) in τ iteration at time t, the archive Φt(τ) is updated:
{
Φ
t
(
τ
)
=
Φ
t
(
τ
-
1
)
⋃
P
ti
(
τ
)
if
⁢
F
1
(
φ
t
(
τ
-
ι
)
)
⁢
…
⁢
F
1
(
P
ti
(
τ
)
)
Φ
t
(
τ
)
=
Φ
t
(
τ
-
1
)
else
,
ι
=
1
,
2
,
…
,
t
-
1
(
17
)
where ∪ is logic relationship “union”, Φt(τ−1)=[φt−1), φt(τ−2), φt(τ−t), . . . , φt(1)], φt(τ−1) is ith optimal solution of archives in τ iteration, F1(φt(τ−1)) is an objective tasks fitness of φt(τ−1), F1(Pti(τ)) is an objective tasks fitness of Pti(τ);
(vii) judging whether to stop iteration:
if current iteration number τ≥τmax, terminate the iteration process and go to step (viii);
otherwise, the iteration number τ increases by 1 and go back to step (iii);
(viii) randomly selecting a solution from archives Φt(τmax) as the optimal set-points at time t u*(t)=[S*NO(t), S*O(t)], where S*NO(t) is an optimal set-point of nitrate nitrogen at time t and S*O(t) is an optimal set-point of dissolved oxygen at time t;
(D) tracking, by the processor, control of the optimal set-points:
a multivariable proportion integral differential controller is used to track and control the optimal set-points, and a control signal is transmitted to the programmable logic controller though Object Linking and Embedding for Process Control of communication;
an analog signal is converted to a digital signal through an analog digital converter;
the digital signal is transferred to a frequency converter by fieldbus, to control a frequency of an oxygen supply pump and a reflux pump, and realize optimal control of effluent water quality and energy consumption for wastewater treatment process;
(i) the multivariable proportion integral differential controller performs tracking control on the optimal set-points SNO(t) and SO(t):
Δ
⁢
u
⁡
(
t
)
=
K
p
[
e
⁡
(
t
)
+
H
l
⁢
∫
0
t
e
⁡
(
t
)
⁢
dt
+
H
d
⁢
de
⁡
(
t
)
dt
]
(
18
)
where Δu(t)=[ΔQa(t), ΔKLa(t)]T is an operational variable matrix, ΔQa(t) is a change of a circulating flow, ΔKLa(t) is a change of an oxygen transfer coefficient;
Kp is a proportional coefficient, Hl is an integral coefficient, Hd is a differential coefficient;
e(t)=y* (t)−y(t) is a control error at time t, y*(t)=[S*NO(t), S*O(t)]T is the optimal set-points at time t, y*(t)=[SNO(t), S*O(t)]T is actual output values of SNO(t), SO(t) at time t;
(ii) adjusting the oxygen transfer coefficient of the dissolved oxygen and an internal return flow:
KLa(t+1)=KLa(t)+ΔKLa(t) (19)
Qa(t+1)=Qa(t)+ΔQa(t) (20)
where KLa(t+1) is the oxygen transfer coefficient of dissolved oxygen at time (t+1), KLa(t) is the oxygen transfer coefficient of the dissolved oxygen at time t;
Qa(t+1) is the internal return flow at time (t+1), Qa(t) is the internal return flow at time t;
when the frequency of the oxygen supply pump and the reflux pump is adjusted by the frequency converter, a concentration of nitrate nitrogen will be adjusted to S*NO(t) and the concentration of dissolved oxygen will be adjusted to S*O(t);
thus, the optimal control of effluent quality and energy consumption for wastewater treatment process is realized.

656 more elements

Optimal control method for wastewater treatment process based on self-adjusting multi-task particle swarm optimization

Life of this patent

Optimal control method for wastewater treatment process based on self-adjusting multi-task particle swarm optimization

Life of this patent

Abstract

Claims1