项目名称: 离散时间非线性多步滑模预测控制理论研究
项目编号: No.61463029
项目类型: 地区科学基金项目
立项/批准年度: 2015
项目学科: 自动化技术、计算机技术
项目作者: 唐伟强
作者单位: 兰州理工大学
项目金额: 44万元
中文摘要: 滑模预测控制在高精度工程中有着广泛的应用前景,而采用线性预测方法对滑模动态进行预测往往满足不了高精度控制的要求。本项目拟将非线性预测方法与滑模控制相结合,力图建立一种基于滑模非线性多步预测的滑模控制新方法。基于稳定性分析确定终端滑模函数和状态的收敛边界,综合优化终端滑模函数的控制参数和系统采样周期选取;利用非线性终端滑模函数作为滑模动态的预测模型,考虑控制、状态和保持算法稳定性的滑模初始收缩约束;采用自适应信赖域序列二次规划对约束优化问题进行高效求解。通过对线性、非线性和时滞等不同系统进行研究,明确滑模非线性预测与控制精度之间的关系,分析其影响控制性能的机理,建立集控制与优化为一体的滑模预测控制分析与设计理论。本项目创新之处为利用连续时间域内具有有限时间收敛属性的非线性终端滑模函数作为滑模预测模型和考虑约束的非线性优化控制求解。研究成果将丰富滑模控制理论,为滑模预测控制应用开启新的途径。
中文关键词: 非线性;滑模预测控制;多步预测;序列二次规划;信赖域方法
英文摘要: Sliding mode predictive control has a broad prospect of applications in the high-precision engineering, however,the use of linear prediction method to predict the dynamics of the sliding mode often can not meet requirements of high-precision control. The project intends to combine nonlinear prediction method with sliding mode control, trying to establish a new sliding mode predictive control method based on sliding mode nonlinear multi-step prediction. Based on the stability analysis method the convergent boundaries of terminal sliding mode function and system state can be determined, so the control parameters of terminal sliding function and the sampling period can be optimally selected. Nonlinear terminal sliding function is adopted as the sliding mode dynamics prediction model, considering the control, state constraints and sliding mode initial contractive constraints that keeping stability. Using adaptive trust region sequential quadratic programming solves the constrained optimization problem efficiently. Research on linear systems, nonlinear systems and time-delay systems using the proposed approach, a relationship between the nonlinear sliding mode prediction and the control accuracy can be identified, analyzing the mechanism of its impact on the control performance, establishing sliding mode predictive control analysis and design theory integrated control and optimization. Innovations of this project are the use of nonlinear terminal sliding mode function with a finite time convergent property in continuous-time domain as the sliding mode prediction model and solving nonlinear optimization considering constraints. The research results will enrich sliding mode control theory, opening a new avenue for applications of sliding mode predictive control.
英文关键词: nonlinearity;sliding mode predictive control;multi-step prediction;sequential quadratic programming;trust region method