复杂的工业大系统,利用传统的模型预测控制(MPC)方法对其实施鲁棒性控制时,不可避免地产生因计算复杂而带来巨大的计算负荷问题.首先利用离散正交小波变换(DWT)的多尺度分解技术和解相关能力,将时域中得到的问题变换到多尺度城中,以简化问题的复杂性,并依此在多尺度域中建立起相应的动态模型;然后利用建立的多尺度模型给出一种能并行执行的多尺度模型预测控制算法,以确保系统的鲁棒性和稳定性;最后利用计算机仿真实验结果比较了传统的MPC方法和新的多尺度模型预测控制(MSMPC)方法在性能上的差异.
In practical application, aiming at large-scale complex industry systems, to implement robust control of them , a very large control horizon is usually needed by using the traditional model predictive control (MPC) methods. Therefore, the high computational complexity and large computational burden are to be solved. For the above cases, the paper firstly uses the mulitscale decomposition technique and decorrelation capability from discrete wavelet transforms to transform the original problems from time domain into multiscale domain to reduce the complexity and establish a multiscale model. Secondly, a new multiscale parallel MPC algorithm is proposed based on the multiscale model. The new algorithm possesses some good characteristics. It reduces the complexity and increases running speed of algorithm, Besides, it can ensure robust stability. Finally, the difference in performance between traditional model predictive control (MPC) and miltiscale model predictive control (MSMPC) can be clearly shown by simulation result.