Design of dynamic adaptive fuzzy neural networks with applications in fault diagnosis and short-term load forecasting
Juan Du · 2011
Soft Computing became a formal Computer Science area of study in the early 1990's.It deals with imprecision, uncertainty, partial truth, and approximation to achieve tractability, robustness and low-cost solution.Fuzzy systems and neural networks have been regarded as the main branches of Soft Computing.In the classic fuzzy system approach, fuzzy rules are determined by domain experts and remain unchanged during the learning.In order to overcome this problem, it is desirable to develop an objective approach to automate the modeling process based on numerical training data for fuzzy systems.Towards this end, the neuralnetwork-based fuzzy systems, called fuzzy neural systems, exhibit great potential in the flexible adaptability to changes.This kind of potential is derived from the learning and adaptive capability of neural networks.Numerous research works have been dedicated to the development of theory and design of systems and algorithms for specific applications.Although those works have demonstrated the exceptional intelligent capability for computing and learning of hybrid systems, the determination of the number of fuzzy rules and identification of neural network structures are still open issues.More specifically, the number of fuzzy rules is fixed and neural network structures cannot be adjusted automatically.Therefore, the adaptive structure identification method is desirable to achieve better system performance.In this thesis, a new adaptive fuzzy neural learning algorithm, namely Dynamic Adaptive Fuzzy Neural Networks (DAFNN) for a nonlinear Multi-Input Multi-Output (MIMO) dynamic system is presented.The proposed DAFNN based on iii