产品介绍
本文为印度Rourkela国家技术研究所（作者：SATH YAM BONALA）的硕士论文，共86页。
近年来，将人工智能各个方面纳入自动控制系统设计和操作的努力，集中在诸如模糊逻辑、人工神经网络和专家系统等技术上。虽然LMS算法被认为是一种流行的系统辨识方法，但在许多情况下，采用LMS算法并不能实现精确的系统辨识。另一方面，人工神经网络（ANN）由于其良好的函数逼近能力被选择作为非线性系统辨识的一种合适的替代方法，即ANN能够在输入和输出空间之间产生复杂的映射。因此，人工神经网络可以用于复杂动力系统的建模，具有合理的精度。
计算机用于直接数字控制突出了最近朝着更有效和更高效的供暖、通风和空调（HVAC）控制方法发展的趋势。暖通空调领域强调了自我学习在建筑控制系统中的重要性，并鼓励进一步研究如何将最优控制和其他先进技术集成到此类系统的制定中。本文介绍了用于暖通空调系统辨识的功能链接人工神经网络（FLANN）、反向传播多层感知器（MLP）和称为情感BP和神经模糊方法的改进BP多层感知器。本文描述了多层感知器（MLP）神经网络、功能连接人工神经网络（FLANN））和ANFIS结构等非线性系统辨识方案的不同结构和学习算法。在使用MLP作为标识符的情况下，考虑了关于隐藏层选择和每层中节点的不同结构。值得注意的是，要实现MLP神经辨识器的正确拓扑，难点在于选择隐层的数目。为了克服这一点，在FLANN标识符中不需要隐藏层，而通过使用三角多项式（即cos（nπu）和sin（nπu））扩展输入，其中n=0,1,2,…。上述人工神经网络结构MLP、FLANN和神经模糊（ANFIS模型）已经得到了广泛的研究。
Recent efforts to incorporate aspects ofartificial intelligence into the design and operation of automatic controlsystems have focused attention on techniques such as fuzzy logic, artificialneural networks, and expert systems. Although LMS algorithm has been consideredto be a popular method of system identification but it has been seen in manysituations that accurate system identification is not achieved by employingthis technique. On the other hand, artificial neural network (ANN) has beenchosen as a suitable alternative approach to nonlinear system identificationdue to its good function approximation capabilities i.e. ANNs are capable ofgenerating complex mapping between input and output spaces. Thus, ANNs can beemployed for modeling of complex dynamical systems with reasonable degree ofaccuracy. The use of computers for direct digital control highlights the recenttrend toward more effective and efficient heating, ventilating, andair-conditioning (HVAC) control methodologies. The HVAC field has stressed theimportance of self learning in building control systems and has encouragedfurther studies in the integration of optimal control and other advancedtechniques into the formulation of such systems. In this thesis we describe thefunctional link artificial neural network (FLANN), Multi-Layer Perceptron (MLP)with Back propagation (BP) and MLP with modified BP called the emotional BP andNeuro fuzzy approaches for the HVAC System Identification. The thesis describesdifferent architectures together with learning algorithms to build neuralnetwork based nonlinear system identification schemes such as Multi-LayerPerceptron (MLP) neural network, Functional Link Artificial Neural Network(FLANN) and ANFIS structures. In the case of MLP used as an identifier, differentstructures with regard to hidden layer selection and nodes in each layer havebeen considered. It may be noted that difficulty lies in choosing the number ofhidden layers for achieving a correct topology of MLP neural identifier. Toovercome this, in the FLANN identifier hidden layers are not required whereasthe input is expanded by using trigonometric polynomials i.e. with cos(nπu) andsin(nπu), for n=0,1,2,…. The above ANN structures MLP, FLANN and Neuro-fuzzy(ANFIS Model) have been extensively studied.
1. 引言2. HVAC系统模型3. 基于人工神经网络的系统辨识4. 改进的后向传播和后向传播算法5. 自适应神经模糊推断系统6. 结论与未来工作建议