This dissertation describes a new high performance, high switching frequency modulation method for power inverter applications. The new method, adaptive decision tree modulation, is a generalized method of inverter modulation that uses an inductive inference method to create decision tree modulators. The decision tree modulators emulate modulation strategies that are optimal with respect to a given performance measure. Adaptive decision tree modulation offers a general design approach because an analytical solution of the specified performance measure is not required by the method. Existing inverter modulation methods do not offer the high performance or generality provided by the new method. Adaptive decision tree modulators have been demonstrated on simulation and hardware test beds and provide performance that is superior to existing discrete pulse modulation methods. |