LSMS2010 AND ICSEE 2010
Pareto-optimal solutions based multi-objective particle swarm
optimization control for batch processes
Li Jia
•
Dashuai Cheng
•
Min-Sen Chiu
Received: 28 October 2010 / Accepted: 1 June 2011
Ó Springer-Verlag London Limited 2011
Abstract In order to maximize the amount of the final
product while reducing the amount of the by-product in
batch process, an improved multi-objective particle swarm
optimization based on Pareto-optimal solutions is proposed
in this paper. A novel diversity preservation strategy that
combines the information of distance and angle into simi-
larity judgment is employed to select global best and thus
the convergence and diversity of the Pareto front is guar-
anteed. As a result, enough Pareto solutions are distributed
evenly in the Pareto front. To test the effectiveness of the
proposed algorithm, some benchmark functions are used
and a comparison with its conventional counterparts is
made. Furthermore, the algorithm is applied to two clas-
sical batch processes. The results show that the quality at
the end of each batch can approximate the desire value
sufficiently and the input trajectory converges, thus verify
the efficiency and practicability of the proposed algorithm.
Keywords Batch process Multi-objective
Pareto-optimal solutions Particle swarm optimization
1 Introduction
Recently, batch processes have been used increasingly in the
production of low volume and high value added products,
such as special polymers, special chemicals, pharmaceuti-
cals, and heat treatment process for metallic or ceramic
products [1]. For the purpose of deriving the maximum
benefit from batch process, it is important to optimize the
operation policy of batch process. Therefore, optimal con-
trol of batch process is of great strategically importance. In
batch process, production objective changes dynamically
with the dynamic market demand characterized by the
presence of multiple optimal objectives which often conflict
with each other. This raises the question how to effectively
search the feasible design region for optimal solutions and
simultaneously satisfy multiple constraints.
In the field of multiple optimizations, there rarely exists
a unique global optimal solution but a Pareto solution set.
Thus, the key is how to find out the satisfied solutions from
Pareto solution set. To solve it, the traditional and simple
method is to combine the multi-objective functions into
single objective function, and then the mature methods
used in single objective optimization problem can be
employed. But the drawback of this method is that only one
solution can be available at each time, which makes it
difficult for finding out enough satisfied solutions distrib-
uting in the Pareto solution set.
Recently, the particle swarm optimization (PSO) algo-
rithm has showed its powerful capability for multi-objective
optimization, which is a stochastic global optimization
approach derived from the feeding simulation of fish school
or bird flock [2]. In some literatures, several strategies have
been used into PSO to solve a wide range of complex multi-
objective optimal problems. Unfortunately, there are still
some deficiencies in the selection of fitness function and
global optimal solution (gbest). That is to say, the key of
multi-objective optimal control of batch process is to solve
the problem of selecting fitness function and global optimal
solution. In general, some kinds of functions can be chosen
L. Jia (&) D. Cheng
Shanghai Key Laboratory of Power Station Automation
Technology, Department of Automation,
College of Mechatronics Engineering and Automation,
Shanghai University, Shanghai 200072, China
e-mail: jiali@staff.shu.edu.cn
M.-S. Chiu
Faculty of Engineering, National University of Singapore,
Singapore, Singapore
123
Neural Comput & Applic
DOI 10.1007/s00521-011-0659-6