An integrated approach coupling behavior, buildings and transportation into Agent Base Models: A San Francisco Use Case

Abstract:

This presentation will highlight preliminary results of a high-impact use case, which lies at the nexus of NREL’s expertise in sustainable transportation and grid-interactive efficiency. The case study includes accurate knowledge of load patterns on the urban electric grid, which is key to designing and using electric power efficiently.

Existing research typically estimates the load patterns of building systems, based on aggregate-level building features like building age and occupancy by building type. However, this research ignores that human activities demanding electricity vary even within the same type of building. For example, residential buildings house people with different household sizes, different cooking or heating habits, and different working schedules. Existing research also miss estimates of new charging demands (e.g., growth in electric vehicles (EVs). These omissions could lead to an inaccurate estimation of the impact of electricity use patterns and peaks on the electric grid.

A key goal of this project has been to fill this gap by embedding NREL’s engineering-based, agent-based models (ABM) in the context of behavioral science, and thereby understanding how societal factors and human behavior affect transportation and building loads. The ABM framework integrates building and transportation systems and connects people’s activities in different buildings with transportation system simulations. It also simulates people’s heterogeneous occupancy plans (i.e., arriving and leaving times) in different buildings, with different EV charging demands. Inside buildings, ABM analyses electricity usage (e.g., lighting, heating, etc.) by population group, identified by key sociodemographic, economic and built-environment characteristics.

In a proof-of-concept for the proposed framework, we applied a typology approach to cluster population groups with different activity patterns using data from the American Time Use Survey within a case study area, San Francisco County. Within the case study area, we used our model to generate daily activity schedules for different populations. Then, using the traffic simulation tool, BEAM, we took the generated activity schedules and road network to simulate mobility patterns and updated activity schedules with traffic conditions and EV charging. Finally, we calculated total electricity consumption with updated building occupancy, in-building activities and EV charging.

The proposed framework will enable “agents” to cross disciplinary boundaries and is expected to estimate electricity loads on the electric grid more accurately. In so doing, it will inform power system management designs by considering the complex spatiotemporal patterns of behavior and energy use within cities.