SBIR Phase II: Improving fleet operational metrics through service optimization with automated learning of vehicle energy performance models for zero-emission public transport

点击翻译按钮获取中文摘要

This Small Business Innovation Research Phase (SBIR) II project will research and validate an internet-of-things (IOT) platform to help commercial fleets transition to zero-emission vehicles (ZEVs). The ZEV transition is the primary solution to the decarbonization of the transportation sector, which is the largest emitter of greenhouse gases in the US. The project focuses on transit agencies, with the ultimate goal of lowering both operating costs and capital costs of their ZEV fleets. Coupled with the current funding support by federal, state and local governments to transit agencies to purchase ZEVs, this project could accelerate the decarbonization of the US transit fleet. A 50% transition of the US transit fleet to ZEVs will reduce nearly 200 million metric tons of carbom dioxide (CO2) equivalent, providing cleaner air quality and reducing urban noise pollution, particularly in low-income communities that rely more heavily on transit services for their transportation needs. The addressable market of Transportation Management Systems will grow from $8.8 billion in 2020 to $27.48 billion in 2028. Demonstrating success in the transit segment will enable the replication of this approach to other fleet segments like school bus fleets, last-mile and mid-mile delivery fleets, and long-haul trucking fleets. The intellectual merit of this project is the design and implementation of an artificial intelligence software platform to automatically learn predictive vehicle models of transit ZEVs and provide recommendation services to transit agencies. The Phase II project has three integrated goals. The first goal is the development of energy prediction algorithms which are scalable and highly accurate. Transit ZEV fleets have stochastic load changes, high sensitivity to operator driving style and high variation of battery size, weight and driving range, even for similar vehicles. These challenges will be addressed by developing automated learning techniques built on algorithms developed in Phase I, which use contextualized data from ZEV stops and trips. The second goal is to validate the prediction accuracy via pilots with ZEV fleets providing scheduled bus services. The final goal is development of real-time, scalable, fleet optimization algorithms which optimize daily assignment and charge management of ZEV fleets. Chance-constrained optimization will be merged with predictive control theory to address scalability and real-time performance of the resulting optimization algorithms. These recommendations will, if successful, demonstrate highly accurate predictions of charge usage, a substantial increase in ZEV fleet utilization, and a reduction of transit ZEV fleet operating costs.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.