Initial problem modeling
We would like to evaluate charging infrastructure across various scenarios. The end goal is for Czechia to have optimal EV charging station deployment with optimizing the following criteria (in arbitrary order, and criteria may be revised):
- charging infrastructure cost
- electric energy distribution grid
- electric energy production
- transition to EVs
- satisfaction with charging infrastructure
High level
Individuals from population are the source of mobility. Part of which is individual mobility and its subset which may be individual electro-mobility. Determinants of the mobility are the socioeconomic and demographic values of the individuals. And outcome of this is a desire to charge their vehicle to be able to go on with their daily lives.
One way of tackling this issue is a agent-based simulation. In style of MatSim (W. Axhausen, Horni, and Nagel 2016).
Agents are a way to model those individuals. In our context an aspects of interest to model are: daily activities, modal choice, EV ownership, daily activities from which mobility pattern arise.
As obtaining the log of activities of all individuals in Czechia is not feasible for lots of reasons the limitiations is also it being gorunded in reality. But in our interest is also evaluating of future scenarios, where electromobility will be more prominent (todo: source). One approach is to create a syntehtic list of activities and agent behaviours.
In those modeled activities is included the charging behaviour, charging station occupation. From this the usage of current charging infrastructure, grid impacts, satisfaction with charging infrastructure can be observed.
The generative model requires a synthetic population. And may look something like: \[ \begin{aligned} P &= \{1,\dots,N\} \\ \mathbb{P}(X) &= \{X_1, \dots, X_n \} \\ Individual_i &\sim \mathbb{P}(X) & i \in \{1,\dots,S\} \\ Population & = \{Individual_1, \dots, Individual_S\} \\ Desired\: activities_i &\sim Individual_i & i \in \{1,\dots,S\} \\ Simulation^* & = argmax_{schedule \in Schedules}\;\; Simulation() \\ Realized\: activities_i &= argmax_{schedule \in Schedules}\;\;Utility(schedule) \end{aligned} \]
where \(\mathbb{P}(X)\) is a multivariate distribution over our populations deomgraphic and socioeconomic values. N is the number of variables in \(\mathbb{P}\). \(S\) is the sampled size of population. Schedules is all the possible schedules for all S agents.
Available data determines method utilized for obtaining the synthetic population. Used data are either census (geographically) aggregated statistics, travel surveys, public use microdata sample (PUMS), or cross clasification tables , colectively characterized as partial views of the joint distribution of the real population (Farooq et al. 2013).
Other related literature:
(Farooq et al. 2013; Borysov, Rich, and Pereira 2019; Sun and Erath 2015)