现代控制方法对系统模型精度要求较高，但其又难以精确建立，为了摆脱模型的束缚，无模型控制方法应运而生。无模型自适应控制(MFAC)仅利用受控系统的I/O信息，通过引入伪梯度向量(或伪Jacobi矩阵)和伪阶数，实现受控系统的参数和结构自适应控制。 基本MFAC算法可以用于大多数过程，为了适应一些特殊的控制问题，保证闭环稳定，本文针对非线性时滞离散系统对基本MFAC算法进行改进，提出改进型无模型自适应控制。给出了SISO，MISO，MIMO非线性时滞离散系统基于紧格式、偏格式、全格式的线性化方法，即用动态线性时变模型来替代一般非线性时滞离散系统；设计了一系列无模型自适应控制器；对SISO非线性时滞离散系统基于紧格式、偏格式、全格式的改进型MFAC方法给出理论推导和稳定性分析。 通过一些典型过程控制系统仿真及实验，将改进型MFAC算法与基本MFAC算法及常规PID算法的控制结果进行比对研究，结果表明改进型MFAC算法优于基本MFAC算法及常规PID控制，有更好的抗滞后特性和动态特性，且具有较好的鲁棒性，证实改进型MFAC方案是合理可行的。 最后，将改进型MFAC算法应用于锅炉给水调节系统，它是一个有迟延、无自衡的环节，传统控制效果不理想，应用本文提出的改进型MFAC算法后控制效果比较理想。

As we all know that the modern control methords require higher accuracy mathematical model but which is difficult to set up. In order to get ride of the shackles of system mathematical model, the Model Free Control methord comes into being. With a newly introduced pseudo-gradient (or pseudo-Jacobi matrix) and pseudo-order, Model Free Adaptive Control only uses I/O information of control system, can achieve the parameter and structure adaptive. Standard MFAC algorithm can be used for most process. In order to adapt special problems and ensure stability, this paper improved the standard algorithm for lag nonlinear discrete system, proposed improved MFAC algorithm. Explain the linearization method of SISO, MISO, MIMO lag nonlinear discrete system based on tight format, partial format and full format, using dynamic linear time-varying model to replace the gengral one. Design series of MFA controller and introduce rigorous theoretical proof and stability analysis for SISO lag nonlinear discrete system based on tight format, partial format and full format. By compared the simulation and esperimental results of typical process control system with the improved MFAC, standard MFAC and PID algorithm, it shows that the improved MFAC has superior anti-lag, dynamic characteristics and better robustness. Confirmed the new scheme is feasible. Finally, the improved MFAC algorithm is used in boiler feed water control system, which is a self-unbalance part and less controllable. The control results are not satisfactory with the traditional methords. After applicating the improved MFAC algorithm, it could achieve better control effect.