摘要
模型算法控制(MAC)是预测控制的一种, 在惯性迟延对象的控制中有很好的应用。但模型算法控制存在稳态误差及控制作用约束性差等问题。Hopfield网络是神经网络的一种重要形式, 在求解有约束二次优化问题中有很好的应用。首先将模型算法控制转化为有约束二次约束优化问题, 然后与标准形式相比较, 得到Hopfield网络的参数, 并用其求解有约束的模型控制算法预测控制序列。这种新方法被应用到电厂中延时和惯性较大的主汽温串级控制系统中。内回路采用比例控制器, 主回路的控制作用由Hopfield网络优化计算得出。与简单的串级控制相比, 它的超调量小, 鲁棒性强, 控制作用平滑, 对控制作用输出可以进行约束。对电厂主汽温系统的仿真研究证明了这种控制方式的有效性。
MAC is one kind of predictive control methods, and has found many applications in controlling objects with a big inertia and delay. However, the MAC has some defects such as stability error and bad constraint of the control output. Hopfield network is suitable for solving the quadratic constrained optimization problem. In this paper, MAC was converted into a quadratic constrained optimization problem and a Hopfield network was developed to solve the control output series with constraints. The new control method was used in the control of main steam temperature system. The inner control loop was implemented by proportional controller. The output of the main control loop can be obtained by using the Hopfield network. The new method has the merits of small over-tuning, strong robustness, smooth control output and the constraints of the control output. The simulation of the steam temperature system has convinced that new control method has an impressive control effect.
出处
《现代电力》
2005年第2期50-52,共3页
Modern Electric Power
