Diversity of urban energy system structure (UESS) was analyzed based on information entropy. To find out the evolution law of UESS and thus provide principles to coordinate and optimize the UESS, the correlative relationship model between UESS and its influencing factors can be developed using the partial least square (PLS) regression. Here we took Xiamen City as an example to analyze the evolution tendency and the influencing factors of diversity of energy supply structure (ESS) and energy demand structure (EDS). The results showed that, on the whole, the diversity of ESS in Xiamen was rising over time as a logistic curve from 1996 to 2003, with reducing speed from 1999 though. While in the period, EDS in Xiamen City didn't improve efficiently. Urbanization level, population, gross domestic product (GDP), and the third industrial structure had significantly affected ESS. Meanwhile, number of vehicles, GDP, population, per capita disposable income, and energy eco-efficiency had obvious influences on EDS. As for Xiamen City, the improvement of diversity of EDS should be especially emphasized.