基于数据与知识的模糊系统设计与应用研究

发布时间：2018-04-24 20:20

本文选题：数据与知识驱动 + 单输入规则模块　；参考：《山东建筑大学》2017年硕士论文

【摘要】：自从模糊集合的概念出现以来,由于其能够充分地利用人类或专家知识来处理系统中存在的各种不确定性,使其在各个研究领域中得到了越来越多的应用。特别是模糊逻辑系统,已广泛应用于建模和控制领域。然而,在系统建模或控制过程中,当输入变量维数较高时,模糊逻辑系统不可避免地面临规则爆炸问题,在这种情况下,很难实现模糊规则的建立以及系统参数的优化。为解决以上问题,本文提出了基于数据与知识的模糊系统的设计方法,其主要研究内容如下:首先,详细介绍了单输入规则模块加权模糊推理系统的结构及其单调性性能,并在此基础上,提出了一种基于数据与知识的单输入规则模块加权模糊推理系统的设计方法。该方法在嵌入知识的基础上,运用基于数据的参数学习策略对系统的参数进行优化。将该方法应用于热舒适性预测,仿真和比较结果证明了该方法对于热舒适性预测的有效性,并且比一些其他现有方法表现得更好。其次,研究了单输入规则模块加权模糊推理系统的扩展结构—函数形单输入规则模块模糊推理系统(FSIRM-FIS),并在此基础上加入神经结构,提出了一种函数型单输入规则模块加权神经模糊系统(FSIRMNFS),此系统结合了FSIRM-FIS和神经网络两者的优点。同时,为了得到系统的最小训练误差和最佳参数,提出了一种基于最小二乘法的参数学习算法。将提出的FSIRMNFS及其参数学习算法应用于小时风速预测,仿真和比较结果验证了该系统对于小时风速预测的有效性。最后,提出了一种数据驱动的二型模糊逻辑系统的构建方法。首先通过自适应神经模糊推理系统(ANFIS)构建一型模糊逻辑系统,再通过集成构建的多个一型模糊逻辑系统得到二型模糊逻辑系统。将此方法构造的二型模糊逻辑系统应用于风速预测问题,并与常用的BPNN和ANFIS方法进行比较,仿真和比较结果表明,所提出的方法在达到类似性能的同时,大大减少了训练时间。随着数据量的爆炸性增长,该方法还有效地减少了二型模糊逻辑系统的建模时间。
[Abstract]:Since the concept of fuzzy set appeared, it has been applied more and more in various research fields because of its ability to make full use of human or expert knowledge to deal with all kinds of uncertainties in the system. Especially fuzzy logic systems have been widely used in modeling and control fields. However, in the process of system modeling or control, when the dimension of input variables is high, the fuzzy logic system inevitably faces the problem of rule explosion. In this case, it is difficult to realize the establishment of fuzzy rules and the optimization of system parameters. In order to solve the above problems, this paper presents a design method of fuzzy system based on data and knowledge. The main research contents are as follows: firstly, the structure and monotonicity of the weighted fuzzy reasoning system with single input rule module are introduced in detail. On this basis, a design method of a single input rule modular weighted fuzzy reasoning system based on data and knowledge is proposed. On the basis of embedding knowledge, the parameter learning strategy based on data is used to optimize the parameters of the system. The method is applied to thermal comfort prediction. The simulation and comparison results show that the proposed method is effective in predicting thermal comfort and is better than some other existing methods. Secondly, the extended structure of the single input rule module weighted fuzzy inference system, the function form single input rule module fuzzy inference system, is studied, and the neural structure is added to the system. A functional single-input rule modular weighted neurofuzzy system (FSIRMNFS) is proposed, which combines the advantages of FSIRM-FIS and neural network. At the same time, in order to obtain the minimum training error and optimal parameters, a parameter learning algorithm based on least square method is proposed. The proposed FSIRMNFS and its parameter learning algorithm are applied to hourly wind speed prediction. The simulation and comparison results show that the proposed system is effective for hourly wind speed prediction. Finally, a data-driven fuzzy logic system is proposed. First, a type of fuzzy logic system is constructed by adaptive neural fuzzy inference system (ANFIS), and then a type 2 fuzzy logic system is obtained by integrating multiple types of fuzzy logic systems. The type 2 fuzzy logic system constructed by this method is applied to the wind speed prediction problem and compared with the usual BPNN and ANFIS methods. The simulation and comparison results show that the proposed method achieves similar performance and greatly reduces the training time. With the explosive growth of data volume, the modeling time of type 2 fuzzy logic system is reduced effectively.
【学位授予单位】：山东建筑大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O159;TP181

【相似文献】