船用柴油机转速控制系统辨识及自适应控制技术仿真研究  

System Identification and Adaptive Speed Governing Control Simulation of Marine Diesel Engine

作　　者：石勇[1] 宋恩哲[1] 

机构地区：[1]哈尔滨工程大学动力与能源工程学院,哈尔滨150001

出　　处：《系统仿真学报》2014年第5期1078-1083,1100,共7页Journal of System Simulation

基　　金：国家自然科学基金(50909024);中央高校基本科研业务费专项资金(HEUCF130307)

摘　　要：船用柴油机具有非线性和时变性,而传统PID控制器无法优化参数,以适应柴油机不同工况下的调速控制要求。因此,为实现柴油机转速自适应控制,一种基于BP神经网络参数优化和Wiener神经网络模型辨识器的PID自整定控制器被提出。在该控制器中,柴油机非线性模型采用Wiener模型描述,并通过神经网络算法优化辨识。利用辨识的模型求出被控对象输出对控制输入变化的灵敏度信息,并用于优化PID参数,提高了参数收敛速度。此外,在该控制器的优化目标评价函数引入相对偏差,以满足不同转速下的控制误差。利用dSPACE开发了该控制技术的半物理仿真平台,仿真在50%额定负荷,柴油机转速交替从1100(RPM)变化到1500(RPM)过程下,采用自适应控制器,速度波动率小于0.81%,稳态响应时间小于6.5s,模型仿真验证该自适应控制可满足的柴油机转速控制要求。Marine diesel engine has the characteristics of non-linear and time-invariant. Traditional PID controller cannot optimize parameters to adapt to the governor of diesel engine according to different working conditions and control requirements. An adaptive controller based on back-propagation（BP） neural network and Wiener neural network was used to tune PID parameters for speed control system of marine diesel engine. In the controller, the model of diesel engine was expressed by Wiener model, and was identified with neural network method. The sensitivity of diesel engine output with respect to its input was obtained via Wiener model, and was used to optimize PID parameters. It improved the convergence speed of the model. Moreover, a relative error was used in target evaluation function of the BP neural network. A simulation platform based on dSPACE was developed, and a simulation using the adaptive controller was done, which was at 50% of rated load and the speed of the diesel engine alternately changed from 1100（RPM） to 1500（RPM）. In the simulation, speed fluctuation rate was less than 0.81%, and the steady state response time was less than 6.5s. The simulation demonstrated that the adaptive controller can meet the demand of diesel engine speed governing.

关 键 词：船用柴油机 BP神经网络 速度控制 Wiener神经网络 自适应控制器 

分 类 号：TK427[动力工程及工程热物理—动力机械及工程]