The location for biomass power plant is regarded as an effective way for convenient and economical feedstock delivery. The main purpose of this study was to develop new site selection methods that can determine the optimal sites for plant and satellite storages under China's specific delivery modes. This methodology was improved from a methodology based on site selection models for single type of biomass and straight-line delivery, and can be used for cases with multi-types of biomass and road delivery, which is more close to reality. A GIS (Geographic Information System) model based on ArcGIS 9.3 was developed, which can not only give the optimal number of storage sites, but also the optimal geographic distribution for the power plant and storage sites, as well as their sub-collection network. In the case of a 100 MWe Zhanjiang Biomass Power Plant, the GIS model applied decreased the overall logistics cost by 5% and 18% relative to the power plant logistics design and direct delivery mode, respectively. Optimization results by the GIS model were compared with a mathematical model developed for straight-line delivery. We found that the mathematical model was also applicable to the road delivery mode. The combination of the two models will be efficient for realistic biomass power plant logistics and optimization under specific delivery modes in China. (C) 2012 Elsevier Ltd. All rights reserved.