Analytics for Sustainable Logistics: Fuel Efficiency and Hydrogen Planning

As the global community moves to decarbonize the modern economy, freight-based emissions—comprising about 8% of global emissions—pose a major challenge to meeting the Paris Agreement’s goal of limiting global warming to 1.5 ∘C to 2 ∘C by 2100. One category of solutions seeks to improve current operational systems and processes, often in data-rich environments. A second category involves leveraging emerging technologies toward zero-emissions transportation, often in the absence of detailed operating data and established procedures. In this thesis, we address problems in both categories, employing data-driven machine learning approaches to enhance fuel efficiency in existing operations as well as stochastic optimization to plan the deployment of hydrogen production facilities in future operations. In Chapter 2, we employ data-driven techniques to explore the drivers of fuel efficiency in the logistics sector in an effort to improve the current operating practices of Rivigo—a large logistics provider in India. Historically, driver training and vehicle maintenance have been the two main areas of focus in the literature to enhance the fuel efficiency of road freight. The advent of telematics, however, provides more granular visibility into fuel consumption, giving rise to opportunities to assess a broader array of fuel-saving interventions. Using data from a large logistics provider in India, we find that driver training and vehicle maintenance can have a limited impact on fuel efficiency. In contrast, we find a more significant effect of driving behaviors, measured through speed and acceleration variables, and of the road infrastructure, captured by the road section of the vehicle. We bring all these fuel efficiency interventions into a unified comparative framework based on machine learning and optimization. Results suggest find that driver training and vehicle maintenance have a smaller overall impact than other interventions such as speed-limiting devices and route optimization. These findings suggest that current practices and policies need to capture an array of interventions to improve fuel efficiency in logistics. Next, Chapter 3 optimizes the deployment of solar-hydrogen systems. Recent advances in vehicle technologies offer the promise to replace conventional gasoline-powered engines with zero-emission technologies based on hydrogen fuel. To be successful, however, hydrogen-powered vehicles need supporting fueling infrastructure to enable their large-scale deployment in logistics fleets. Moreover, to truly make a dent toward decarbonization, hydrogen needs to be produced from carbon-free electricity sources. In response to these two interrelated challenges, we propose an optimization approach to support the deployment of joint solar-hydrogen systems. Specifically, we formulate a two-stage stochastic mixed-integer model to optimize the location of solar power plants, hydrogen production facilities and a supporting distribution infrastructure, along with subsequent operations pertaining to production and distribution decisions in response to hydrogen demand. We implement the model using real-world data from Dordogne, France. The model’s solution mirrors the current solution planned in practice, but differs by inducing a higher electrolyzer capacity and by distributing this capacity across two locations. Results suggest that the proposed optimization approach can provide significant cost reductions along with reductions in demand shortages, as compared to existing solutions implemented in practice, alternative benchmarks, and deterministic optimization approaches.S.M