International Journal of Automation and Power Engineering Vol. 1 Iss. 8, November 2012 179 Identification and Adaptive Neural Network Control of the Speed of Marine Diesel Engine Shi Yong College of Power and Energy Engineering, Harbin Engineering University, Harbin 150001, China sy_hit@163.com Abstract With characteristics of non-linear and time-varied, so it is difficult for a marine Diesel Engine to be controlled with traditional PID controller. An adaptive controller based on back-propagation (BP) neural network and Wiener model identifier was put forwarded to tune PID parameters for marine diesel engine speed control system. In the controller, a Wiener neural network structure was applied to identify Wiener model of diesel engine nonlinear model. The weights in the Wiener neural network are adjusted with backward-propagation methods, and those weights stand for the parameters of the Wiener model. In order to satisfy the different work conditions, the adaptive controller is improved via introducing relative error in target evaluation function of the BP neural network. Moreover, with the sensitivity function of diesel engine output with respect to its input obtained by the WNN identifier, the convergence speed of optimizing PID parameters is improved. A simulated test on a diesel engine demonstrated that the adaptive controller improved control performance over the traditional PID control. Keywords Marine Diesel Engine; BP Neural Network; Speed Control; Wiener Neural Network; Adaptive Controller Introduction Speed governing system of diesel engine is an important component of diesel engine electronic control system. At present, the traditional PID controller is one of the popular controllers in diesel engine speed governing system, since its design is simple and does not require detailed knowledge of the system dynamics [1]. However, diesel engine is a non-linear and time-varied system, and it is difficult to determine the appropriate PID parameters when various uncertainties and nonlinearities exist. Moreover, once control parameters established, it is difficult to be adjusted online. Numerous research papers focused on adaptive PID control [2], self-tuning PID control [3, 4], etc. In recent years, the combination of the PID control and neural network has become a new direction for intelligent controller, attracting many researchers [5-10]. The self-tuning PID can use the BP neural network to tune PID parameters online in accordance with change of load and conditions automatically. However, because partial derivative of diesel engine output with respect to its input cannot be calculated directly, convergence speed of the BP neural network tuning PID parameters is not fast. In order to increase convergence speed, it is necessary to identify the model of diesel engine, and obtain sensitivity function of diesel engine output with respect to its input. As to nonlinear system identification, Wiener models are widely used for modeling of processes [11-14], which describe a diesel engine model as a linear dynamic block in cascade with a nonlinear static gain. In order to achieve a certain performance or adaptive capability, neural networks were integrated into Wiener model to form Wiener neural network [15, 16]. However, a diesel engine is a time delay system, so time delay must be considered in Wiener neural network. In this paper, an adaptive controller is proposed, which includes a Wiener model identifier and a PID controller tuned by back-propagation (BP) neural network (BP-PID). The Wiener model identifier is a Wiener-type neural network, which is used to identify diesel engine model and to provide approximate sensitivity function of the diesel engine model online to update the PB-PID controller. Moreover, considering different work conditions, relative error is introduced in target evaluation function of the BP neural network. The adaptive controller was tested on identifying and controlling a diesel engine, and the experiment confirmed the effectiveness of the proposed method. The Control Model of Marine Diesel Engine From a control point of view, two important paths