TELKOMNIKA, Vol.11, No. 7, July 2013, pp. 3498 ~ 3507
e-ISSN: 2087-278X
3498
Received January 30, 2013; Revised March 16, 2013; Accepted March 26, 2013
The Power Unit Coordinated Control via Uniform
Differential Evolution Algorithm
Zain Zaharn*
1
, Ruifeng Shi
2
, Xiangjie Liu
3
School of Control and Computer Engineering, North China Electric Power University
Beinong Road 2, Changping District, Beijing, 02206, China
*Corresponding author, e-mail: zainozahran@gmail.com
Abstract
This paper modified the differential evolution (DE) algorithm adaptively to solve the power unit
coordinated control (PUCC) problem. It was modified in two aspects: 1) a uniform initialization with a
controlling zone factor (m), 2) a regular mutation process, to develop an effective searching mechanism
and improve the performance of the basic DE algorithm. A numerical case study was employed to verify
the performance of our proposed uniform differential evolution (UDE) algorithm, by contrast to the basic
DE, genetic algorithm (GA), and particle swarm optimization (PSO) algorithm. The experimental simulation
results show that our proposed UDE algorithm has outperformed the other comparative algorithms, which
demonstrate the effectiveness and efficiency of the new algorithm.
Keywords: power unit coordinated control problem, differential evolution algorithm, uniform differential
evolution algorithm, genetic algorithm, particle swarm optimization algorithm
Copyright © 2013 Universitas Ahmad Dahlan. All rights reserved.
1. Introduction
The PUCC problem is an interesting task for many researchers during these decades.
The power unit is composed of a boiler-turbine system coupled to an electric power generator.
The boiler-turbine configuration is a multivariable, nonlinear, and time varying-system with
complex operation due to the uncertainities and high settling time. So, the main goal of the
boiler-turbine control is to justify the generator output power to maintain a high response to the
load fluctuation, while keeping the steam pressure and temperature within the permissible
values. For simplification, the power unit might be considered as three-input three-output
system, in which the essential inputs are the boiler firing rate, position of the steam valve, and
the feeding water, while the outputs are the electric power, steam pressure, and water level
deviation. Controlling of this system can be based on the stored thermal energy in the boiler or
on the boiler-turbine governor as two alternative conventional techniques. As the first technique
is slow with high stability, the latter is fast to follow the load variations but might be unstable.
The coordinated control is an integration of the two conventional techniques to result in a stable
control system with high level of response. Then the power unit is to be optimized under
considerations of minimizing the load tracking error, injecting fuel, steam losses, and feeding
water. There are different published works in the literature, through which the PUCC problem is
solved according to the mentioned considerations. In [1], it was solved traditionally to prevent
cyclic damage to the plant by designing prototype load controllers to adjust the flow of air and
fuel to the boiler and flow of steam to the turbine. In [2], the solution was based on coordinated
control with pressure set point scheduling by finding a single solution with the preferences of the
problem. In [3-5], it was solved by modern heuristic algorithms such as GA, DE, PSO, and
Pareto multi-objective optimization (PMOO) techniques, depending on a reference governor and
optimization unit. In [6, 7], a multi-agent system was presented to solve this problem, through
which a computer software programs work independently in a number of stages to establish the
required coordinated control. In [8], a nonlinear multivariable power plant coordinated control by
constrained predictive scheme was created to maintain the thermal constrained during load
fluctuations. In [9], a new control strategy was used to control the power plant by Smith`s
predictor to overcome the time-lag influences.
The DE algorithm was proposed in 1996 by Storn and Price [10] as a new evolutionary
algorithm, which has a simple structure and efficient solutions. Thus, its applications are