Data-Driven Agent-Based Modeling, with Application to Rooftop Solar Adoption

摘要: Agent-based modeling is commonly used for studying complex system properties emergent from interactions among many agents. We present a novel data-driven agent-based framework applied to forecasting individual and aggregate residential rooftop solar adoption in San Diego county. Our first step learn model of agent behavior combined data characteristics property assessment. then construct an simulation with the learned embedded artificial agents, proceed validate it using holdout sequence collective decisions. demonstrate that resulting successfully forecasts trends provides meaningful quantification uncertainty about its predictions. utilize our optimize two classes policies aimed at spurring adoption: one subsidizes cost adoption, another gives away free systems low-income households. find optimal derived latter class are significantly more efficacious, whereas similar current California Solar Initiative incentive scheme appear have limited impact on overall trends.