基于自适应神经模糊推理系统的连续搅拌反应釜系统的改进广义预测控制

连续搅拌反应釜(continuously stirred tank reactor,CSTR)是典型的非线性、大滞后的化工对象。为克服传统控制方法难以建立其机理模型的难点,基于对实际工业现场采集得到的大量输入输出数据,提出采用自适应神经模糊推理系统(adaptive-network-based fuzzy inference system, ANFIS)的方法,得到CSTR系统较理想的输入输出关系规则库,即CSTR系统的T-S模糊模型。最后,通过改进的广义预测控制(Jin’s generalized predictive control, JGPC)算法对CSTR系统的浓度进行控制,并与常规的广义预测控制(generalized predictive control,GPC)算法、比例-积分-微分(proportion integral differential,PID)控制算法进行对比,仿真结果显示,JGPC算法的控制效果优于GPC算法和PID控制算法,证明了方法的有效性。

Improved Generalized Predictive Control for CSTR System Based On ANFIS

Continuous stirred reactor is a typical non-linear, large-lag chemical object. In order to overcome the difficulty of establishing the mechanism model by traditional control methods, this paper proposes an adaptive neuro-fuzzy inference system based on a large number of input and output data collected from actual industrial sites, and obtains an ideal input-output relationship of the continuous stirred reactor system. The rule base, the TS fuzzy model of the continuous stirred reactor system. Finally, the concentration of the continuous stirred reactor system is controlled by the improved generalized predictive control algorithm, and compared with the conventional generalized predictive control algorithm and PID control algorithm. The simulation results show that the improved generalized predictive control algorithm has better control effect than GPC. The algorithm and PID control algorithm prove the effectiveness of the proposed method.