Alternative fuel buses have great potential to reduce life-cycle energy use and criteria pollutant emissions from urban and rural transit fleets. However, market penetration of alternative fuel transit buses in the U.S. is currently below 50%. There are a number of barriers that discourages switching from traditional diesel vehicles. This study develops an analytical framework to automatically generate optimal plans for fleet operations and conversion based on fleet-specific characteristics. The goal of the framework is to allow agencies to minimize the fleet-wide energy use, life-cycle CO2 emissions, and/or economic costs when transitioning to alternative fuels. Energy use are modeled using advanced tools based on real-world second-by-second operations data. Machine learning models are developed by using vehicle-specific and operating-related features, and then applied to assess the fleet-wide energy use. Cost changes of adopting alternative fuel vehicles are evaluated comprehensively. The framework includes three sets of optimization models, focusing on optimizing existing fleet operations, fleet electrification, and long-term fleet conversion. Decisions are optimized for annual fleet purchase/replacement of different vehicle types, vehicle-route assignment, vehicle-depot assignment, charging scheduling, and/or charging station/depot location selection. The proposed models are applied to the bus fleet from Metropolitan Atlanta Rapid Transit Authority (MARTA), the largest transit agency in Metro Atlanta.
