A Novel Dissolve Oxygen Control Method Based on Fuzzy Neural
Network
Jinchao Xu
1,2
, Cuili Yang
1,2
, Junfei Qiao
1,2
1. Faculty of Information Technology, Beijing University of Technology, Beijing 100124
2. Beijing Key Laboratory Computational Intelligence and Intelligent System, Beijing 100124
E-mail: winadream@163.com
Abstract: In the wastewater treatment processes (WWTPs), the oxygen transfer is a nonlinear and large time-delay process,
which makes the dissolved oxygen (DO) concentration difficult to control. In this paper, a novel kind of control method based on
fuzzy neural network (FNN) is proposed for controlling DO concentration. The parameters of the neural network were adjusted
online through the gradient descent algorithm to get the minimum error. Finally, the simulation results in Benchmark Simulation
Model No.1 (BSM1) show that the proposed fuzzy neural network controller has better adaptability than some other existing
methods.
Key Words: Fuzzy Neural Network, Dissolved Oxygen Control, Wastewater Treatment Processes
1 Introduction
The control of biological wastewater treatment processes
(WWTPs) is very complex since the biological and
biochemical processes are strongly interrelated, the
dissolved oxygen (DO) concentration is one of the most
important control parameters in the WWTPs [1]. If the DO
concentration is too low, the sludge bulking will happen and
the effectiveness of treatment will reduce. If the DO
concentration is too high, the settlement of suspended solids
will be worse, and the energy consumption will be high [2-3].
In order to ensure the indicators of effluent satisfy the
standard for WWTPs, the DO concentration must be
controlled within a certain normal range.
In the past, there are many works for controlling DO
concentration in WWTPs, the classical
Proportional-Integral-Differential (PID) controllers is
chosen in most of the DO control systems due to their simple
design and application. Wahab [4] proposed the
multivariable PID controller for the WWTPs to realize the
control of the DO, meanwhile the parameters of the PID
controller were optimized. However, when significant
nonlinearities, uncertainties and disturbances are present
within the system, the performance of the PID control will
degrade subsequently. To avoid performance reduction in
DO control, Vrecko [5] proposed a hybrid strategy that
combined PID and feed-forward control methods to control
the DO concentration. Furthermore, Zhang [6] proposed an
adaptive genetic algorithm to tune the parameters of PID
controller, which makes good performance for DO
concentration control. Although these methods have
achieved good performance, they are not good enough to
adapt to the nonlinear characteristics of the wastewater
treatment process.
Recently, researchers have become more interested in
employing fuzzy control strategies for complex nonlinear
*This work was supported by the National Natural Science Foundation
of China under Grants 61533002 and 61603012, Beijing Science and
Technology Project under Grant Z141100001414005, Beijing Municipal
Education Commission Foundation under Grant KM201710005025.
system. Wang [7] proposed a fuzzy expert control method
based on the predictive model, which effectively reduces the
large delay effect of the system and obtains better control
quality. Belchior [8] proposed an adaptive fuzzy control
strategy into the wastewater treatment Benchmark
Simulation Model No.1 (BSM1) to control the DO
concentration and get a better control effect. However, the
traditional fuzzy control depends on the input dimension of
the system in a large extent, the larger were the input
dimension, the more the fuzzy rules and the more complex
were the fuzzy inference process, so high computation
complexity are involved.
On the other hand, since neural network control method
has good self-learning and adaptive capacity, many
attentions are paid, especially the fuzzy neural network
(FNN). The FNN neural network takes advantage of the
self-learning, self-adaptive and strong fault tolerance
features of neural network, and combine with the fuzzy
system characteristics, which makes it not only closed to the
human experience, but also have strong self-adaptive. Thus
many researchers applied the FNN into the WWTPs to solve
the DO concentration problem. Hu [9] proposed a fuzzy
neural network to control DO concentration in the varying
parameters activated sludge systems. The results show that
the FNN control system has not only achieved better control
effect, but also acquired better dynamic performance. Liu
[10] proposed adaptive fuzzy neural network controller to
realize the control of DO concentration in activated sludge
model. The results show that the controller can not only
adjust subjection function on-line, but also have great
robustness. Fu [11] proposed the Takagi-Sugeno (T-S) fuzzy
neural network to control the DO concentration in the BSM1,
the results show that better adaptability and robustness are
achieved in this control system.
In this paper, a novel FNN controller is constructed and
applied into DO concentration control in the WWTPs. The
rest of this paper is organized as follows. Section 2 briefly
describes the activated sludge wastewater treatment BSM1.
Section 3 presents the control framework of FNN. Section 4
Proceedings of the 36th Chinese Control Conference
Jul
26-28, 2017, Dalian, China
4363