【作者】 冷欣；

【导师】 朱齐丹；

【作者基本信息】 哈尔滨工程大学 ， 控制理论与控制工程， 2009， 博士

【摘要】 增压锅炉在船舶蒸汽动力装置中具有举足轻重的地位,具有尺寸小、重量轻、功率大、经济性以及高可靠性等优点。为推动船舶蒸汽动力的发展,提高船用增压锅炉的自动化水平,开展增压锅炉先进控制方法研究是发展增压锅炉在船舶行业安全、稳定、高效运行的迫在眉睫的研究课题。预测控制作为一种性能优越的控制算法,对模型精度要求不高,在过程控制应用中显示了良好的性能,本文以实际科研项目为课题研究背景,以预测控制理论为基础,围绕船用增压锅炉蒸发系统,以减少算法计算量、提高汽包水位控制算法的实时性为目标,开展了以下的研究工作：首先,从建立船用增压锅炉系统仿真平台角度出发,全面分析了船用增压锅炉蒸发系统的工作机理,简化其物理模型,利用汽水两相的质量平衡方程和能量平衡方程建立了蒸发系统的动态机理模型,仿真结果反映了汽包水位在不同工况下的动态响应,符合基本规律和过程机理,为后续的船用增压锅炉汽包水位预测控制方法研究奠定了基础。其次,为提高船用增压锅炉汽包水位控制系统的快速性,对串级广义预测控制算法进行了改进,设计了船用增压锅炉汽包水位的串级GGPC(灰色广义预测控制)-PI控制器。该控制器采用简单的灰色预测模型,与广义预测控制中的CARIMA模型相比减小了计算量,实现了对汽包水位特定工况的有效控制。该方法综合了预测控制、串级控制、灰色理论和PID控制的优点,且方法简单易行。再次,根据船用增压锅炉大范围变工况的运行特点,提出了汽包水位的RBF神经网络动态补偿多模型预测控制策略,即在典型工作点建立固定模型并结合RBF神经网络动态补偿模型误差,控制器及时实现模型切换,以适应动态特性的变化。该方法充分考虑到单一固定模型不能适应全局大范围工况变化的特点,预测控制与RBF神经网络的结合极大地提高了系统的鲁棒性与稳定性。最后,针对船用增压锅炉汽包水位非线性、时变和环境不确定等特性,设计了基于T-S模糊多模型的汽包水位滑模预测控制器。采用具有万能逼近特性的T-S模糊模型作为预测模型,利用滑模变结构的强鲁棒性改善预测控制特性,仿真实验表明所设计的非线性预测控制策略解决了汽包水位非线性不确定性的控制难点,且具有良好的鲁棒性。更多还原

【Abstract】 Supercharged boiler plays an important role in marine steam power plant and it has the advantages of small size, light weight, big power, economy and high reliability and so on. To promote the development of marine steam power and improve the automation level of supercharged boiler, it is urgent to carry out advanced control method, which will ensure supercharged boiler safety and stability operation in the shipbuilding industry. As a superior control algorithm, predictive control shows good performance in process control applications. The thesis relies on the actual scientific research for the topic research background and is based on the evaporation system of marine supercharged boiler. For reducing the amount of algorithm and improving the real-time behavior, the following research is carried out:Firstly, from the establishment of supercharged boiler system simulation platform point of view, working mechanism of the evaporation system is comprehensively analyzed. Simplifying the physical model, dynamic mechanism model is established by using mass balance and energy balance equation of steam and water two-phase. The simulation results reflect dynamic response in the different conditions and accord with basic law and process mechanism for providing the foundation on the follow-up study of drum water level control method.Secondly, to improve marine supercharged boiler drum water level system’s rapidity, the cascade GPC control algorithm has been improved and a practical cascade generalized predictive control algorithm based on grey predictive model is put forward. The controller uses a simple gray forecasting model which reduces the amount of calculation compared to CARIMA model of generalized predictive control and realizes effective control of a particular condition for drum water level. This method combines the advantages of the predictive control, cascade control, gray theory and PID control, and the method is simple and easily used in real systems.Thirdly, according to the operational characteristics of a wide range of variable conditions, multi-model predictive control strategy based RBF neural network dynamic compensation is proposed. The fixed models are established on typical operating points and the RBF neural network model is used to compensate the model error. The models are swiched in time to adapt to the changes of dynamic characteristics. The method fully considers that a single fixed model can not adapt to the changes in a wide range of operating conditions.The combination with predictive control and RBF neural network greatly improves the system robustness and stability.Finally, for the drum water level with the non-linear, time-varying and uncertain environment, sliding mode variable structure predictive control strategy based on T-S fuzzy model is designed. The method uses T-S fuzzy model for approaching the nonlinear model of the system and applies the strong robustness of sliding variable structure control to enhance the control behavior of predictive control. The simulation results reflect the new nonlinear predictive control strategy has a very good robustness.更多还原