DC-DC Converter Based Power Management In Clean Energy Systems

As energy crisis become urgent, clean energy (wind energy, solar energy, fuel cell etc.) has attracted more attention. In power management of clean energy systems, the output of energy generation varies aperiodically. Thus, it is vital important for DC-DC converters to regulate voltage of energy generation system to a desired value so that load can work well. In this sense, accurate prediction of energy generation and proper modeling, observing and control schemes for DC-DC converters are essential to build a practical clean energy system. Some innovative work have been down on these two problems, which are listed below.● Observability criterion of large signal models of DC-DC converters is solved in this thesis. We proposed a general framework for observability analysis and ob-server design. Derived from structured bilinear systems (SBLS) and graph theory, a novel method with small computational complexity is proposed to verify ob-servability properties of the boost converter and the superbuck converter. Further-more, a full state nonlinear observer is designed based on the port-Hamiltonian system form of the large signal model of a DC-DC converter, which is easy to find the Lyapunov function and the condition of observer error convergence. The observer is applied to the boost converter and the observation of the inductor current tracks the actual value well.● Markov Chain (MC) models are widely used in power prediction. Classification of time series data to construct MC states plays a key role in MC models. A Spectral-analysis-based K-means Clustering (SKC) method is presented to dis-cretize data set containing few variables. Experimental results show that clusters distribute more properly than the traditional Equal-interval Classification (EC) method and the Spectral Clustering approach. Based on the SKC method, wind power prediction by a MC Transition-Probability-Matrix (MC-TPM) performs better than the one based on an EC approach. Moreover, the convergence prop-erty of transition probabilities has been discovered and proved, which points out the limitation of MC models.