广义预测控制算法的改进

Improvement of Generalized Predictive Control Algorithm

将广义预测控制算法的各环节整合在一起,提出了根据模型参数和过程输入输出值直接计算未来时间的闭环预测值和单位阶跃响应曲线的新算法。在新算法中无需显式求解Diophantine方程,仅需要使用矢量相加减、点乘和移位运算等基本算子,易于在PLC中实现。新算法只需要存储未来时刻的闭环预测值及单位阶跃响应曲线数据,存储数据较现有方法明显减少。利用GPC模型导出了与DMC相似的闭环预测式,可以利用GPC算法的在线模型辨识功能计算DMC算法的闭环观测值及阶跃响应曲线,将这两种方法融合起来,使DMC算法也具备在线修正过程模型的能力。通过例子对算法的正确性进行了验证。

A new algorithm was proposed to calculate closed loop predictive values in the future area and unit step response curve according to process inputs and outputs. GPC algorithm was simplified by considering several calculat- ing sections altogether. The new algorithm is only composed of several basic operators including vector add and sub- tract, vector dot product and shifting, and so on. The new algorithm does not solve Diophantine equation explicitly and can be realized in PLC easily. The new algorithm only need to store two vectors about closed loop predictive val- ues in the future area and unit step response curve, and has smaller data storage demand than that in reported litera- tures. GPC algorithm and DMC algorithm have similar formulas of closed loop predictive values. The two algorithms can be fused by this means： GPC algorithm can be used to calculate closed loop predictive values in the future area and unit step response curve in real time; on the other hand, DMC algorithm can be used to calculate control values. An example is used to verify the new algorithm.