15h30 - 15h50
Recent Advance in Machine Learning-based Short-term Load Forecasting
With the increasing adoption of renewable energy generation and electric devices, electric load forecasting, especially short-term load forecasting (STLF), is becoming more and more important. The widespread adoption of smart meters makes it possible to utilize complex machine learning models for both aggregated load and single-home residential load forecasting. Although machine learning models, especially deep learning models, have shown impressive success in different areas, including short-term electric load forecasting, such models require a large amount of training data. For many real-world load forecasting cases, we may not have enough training data to learn a reliable forecasting model, and the data distribution may not be stationary, which will make the classical machine learning models can not perform well. In this presentation, I will introduce our recent work on leveraging advanced machine learning techniques such as transfer learning, graph neural networks, reinforcement learning, and boosting to further improve the load forecasting performance.
15h50 - 16h10
Combining Bayesian optimization with analytical dynamic traffic models to tackle dynamic simulation-based transportation optimization problems
Cities and companies alike often resort to the use of detailed, dynamic, and stochastic traffic simulators to tackle their transportation optimization problems. Methods from the field of simulation-based optimization are often used. However, their use to tackle problems with time-dependent decision variables is limited. In this talk, we present recent work that formulates an analytical, differentiable, and compute efficient dynamic traffic model, and integrates it within a Bayesian optimization (BO) framework to tackle high-dimensional dynamic transportation problems. We present validation results on small network problems, as well as a large-scale traffic signal control case study for Midtown Manhattan, New York City. We discuss how the proposed approach enhances the scalability of BO methods.
16h10 - 16h30
Dynamic routing and charging policies for battery-equipped electric vehicles with partial information
Electric vehicles (EVs) are crucial in developing sustainable transportation systems. Although technological advances increase their attractiveness, several barriers to their mass adoption still exist. In this respect, the EV's limited driving autonomy remains a significant issue. As a result, driving between a distant origin and destination may necessitate several stops at Charging Stations (CSs). A fair amount of uncertainty exists with respect to the status of public CSs, which are primarily used by private vehicles. Namely, if an EV arrives at a CS and finds it occupied, the amount of time it needs to wait is uncertain. In such situations, an EV may choose to wait or deviate to another CS. In line with recent technological advancements, we expand the literature on EV shortest path problems with uncertainty by considering a real-time binary indicator of the CS status, namely, busy or vacant. Such additional information facilities better route planning. We propose a real-time re-optimization approach that incorporates indicators to constantly update the optimal route to the destination. Our preliminary results show that our suggested framework considerably outperforms techniques with no real-time information provided to the planner in terms of overall travel duration. Notably, we demonstrate that effectively incorporating indicators reduces the waiting time significantly.
16h30 - 16h50
ThisClimateDoesNotExist.com: Visualising climate change impacts on street photos with AI
Climate change is a major threat to humanity, and the actions required to prevent its catastrophic consequences include changes in both policy-making and individual behaviour. However, taking action requires understanding the effects of climate change, even though they may seem abstract and distant. Projecting the potential consequences of extreme climate events such as flooding in familiar places can help make the abstract impacts of climate change more concrete and encourage action. In this talk, I will present ThisClimateDoesNotExist.com, a website that generates photorealistic visualisations of extreme climate events onto user-chosen street photos. I will motivate the reasons that led us to develop such project as a contribution to raise climate change awareness, the details of the generative deep learning methods behind the algorithm and preliminary results on how these images impact the willingness to adapt the behaviour of users.
16h50 - 17h10
When Shared Autonomous Electric Vehicles Meet Microgrids: Citywide Energy-Mobility Orchestration
We develop a cross-disciplinary analytics framework to understand citywide mobility-energy synergy. In particular, we investigate the potential of shared autonomous electric vehicles (SAEVs) for improving the self-sufficiency and resilience of solar-powered urban microgrids. We develop a space-time-energy network representation of SAEVs and formulate linear program models to incorporate operational decisions interconnecting the mobility and energy systems. To preventatively ensure microgrid resilience, we also propose an “N − 1” resilience-constrained fleet dispatch problem to cope with microgrid outages. Combining eight data sources of New York City, our results show that 80,000 SAEVs in place of the current ride-sharing mobility assets can improve the microgrid self-sufficiency by 1.45% (benchmarked against the case without grid support) mainly via the spatial transfer of electricity, which complements conventional VGI. Scaling up the SAEV fleet size to 500,000 increases the microgrid self-sufficiency by 8.85% mainly through temporal energy transfer, which substitutes conventional VGI. In addition, microgrid resilience can be enhanced by SAEVs. The SAEV fleet operator can further maintain the resilience of pivotal microgrid areas at their maximum achievable level with no more than a 1% increase in the fleet repositioning trip length.