Wavelet neural networks in nonlinear system modeling and motor drives

This work is focused on Wavelet Neural Networks (WNNs) in nonlinear system modeling and in particular its applications in advanced motor drives. This work mainly includes the following two parts.; In the first part, a WNN-based motor speed estimator is developed to demonstrate the advantage of WNNs in modeling dynamic nonlinear system rich of local nonlinearities and fast variations. Firstly, the network structures of WNN-based speed estimators are determined according to DC motor dynamic equations. Secondly, the WNN-based speed estimator models are constructed in the Virtual Test Bed computational environment to verify the estimator design and to test the model generalization property. Finally, the proposed speed estimators are validated through hardware tests for a brush DC motor and a brushless DC motor. The experimental results demonstrate that the speed estimator provides accurate speed outputs over a wide operating range including low-speed bands and transient processes. In addition, these estimators have concise network structures. The WNN-based speed estimator can be used as a virtual sensor in high performance speed control of DC motors.; The second part concentrates on developing a novel WNN-based multi-resolution modeling approach with a hierarchical structure and progressive accuracy. The proposed WNN-based modeling approach possesses two unique characteristics. Firstly, a multi resolution system model, having an output corresponding to each resolution, is developed from a coarser approximation to a finer representation by adding more details progressively. Secondly, the model at a low resolution is compatible with the model at a high resolution, which means that the well trained WNNs used in a low resolution can be directly incorporated into a high resolution without any modification. This modeling approach provides a generic model for various applications with flexible accuracy and complexity requirements. Different users can activate different resolutions according to their requirements during model utilization. At the same time, this compatible modeling structure avoids a large amount of modeling repetition and thus lightens the computational burden. Two practical applications are used as examples to demonstrate the implementation of the proposed modeling approach including a lithium-ion battery and a nonlinear resistor in the electrical field. The study results also indicate that the proposed modeling approach is promising for many engineering applications.