Electric mobility in view of green growth A synthetic information system on HPC for the global car population Sarah Wolf, Global Climate Forum with Steffen Fürst, Andreas Geiges, Jette von Postel International Conference on Synthetic Populations Lucca, 23.2.2017
Motivation, on the one hand h+p://www.flickr.com/photos/lhirlimann/6511042715/ World Economic Forum, 2016 Google FT, 16.6.2016 transition in the auto industry... transition to Green Growth? 2
Green Growth Environment many actors expectations incentives Economy climate policy & investment impulse re-coordinate expectations investment -> technical progress -> growth -> expectations -> investment ->... 3
Cars 4
Reverse the trend? 5
Motivation, on the other hand pilot studies Health Habits, Green Growth, Global Urbanization GSS: stakeholder dialogues and data-driven simulation modelling of complex systems for decision support CoeGSS: enhance GSS modelling through High Performance Computing and Data Analytics (HPC, HPDA) enable use of higher resolution data sets allow models to grow in complexity / towards global scales facilitate deeper analysis of larger sets of output data from model simulation runs 6
Methods consider non-deterministic dynamics, open future take a systemic perspective complex system, multiscale, feedbacks, path-dependency,... draw on the framework of extended evolution random mutations of genes & natural selection (fitness) + regulatory networks: sequence and intensity of activation of genes + niches: environment, shaped in turn by species evolution of technologies: regulatory networks & niches restrain the space of possibilities for system evolution agent-based modelling on HPC micro level description of the system on the computer to observe macro level in simulations 7
The car centered global system consumers local environment: e.g., charging infrastructure, regulation (cities, states) local interaction: e.g., congestion, accidents social influence, in networks with hubs, clusters, assortativity, community and hierarchical structures employment in car industry and component suppliers manufacturers operating in a global market example Volkswagen sales 2016: Europe 41%, US 6%, China 39% regulation ( regulatory networks ) example China: - plans for manufacturers EV shares 8% in 2018, 12% in 2020 - megacities: quota policy to control car ownership growth: Beijing lottery, Shanghai auction, Guanzhou hybrid of these two 8
The car centered global system and its environment regulation ( niches ) laws and regulation for admission insurance standards (safety, emissions) => inertia of the system open system geopolitical environment: petroleum supply, climate policy,... oil industry as a niche for internal combustion engine cars prices shape the space of possibilities e.g., petroleum, carbon emissions, batteries 9
The model begin on the demand side spatially explicit data: population per grid cell, cars per 1000 people per country; data for 2005-2015, scenarios for 2016-2025 10
Scenarios: example total numbers of cars as estimated according to a model by Dargay, Gately, and Sommer (2007) based on GDP per capita, population density and urbanisation country specific saturation level input data (estimates) from IMF, WB, UN
The model begin on the demand side spatially explicit data: population per grid cell, cars per 1000 people per country; data for 2005-2015, scenarios for 2016-2025 green (EV for now) and brown cars diffusion model, dynamics based on innovation & imitation - innovation happens stochastically, higher income leads to higher probability - imitiation: EVs as an indicator of EV-friendly infrastructure deterministic and stochastic version of the model working towards synthetic population of households, later also firms optimizing agents, interaction networks, environment 12
Implementation model implementation in Pandora HPC framework for agent-based models developed at Barcelona Supercomputing Center advantages: twin interface Python (rapid prototyping) or C++ (parallelization), reads raster files, output supports GIS tools (maps), hides MPI layer from modeller disadvantages: spatial partitioning problematic for our global maps and for agent networks beyond spatial proximity CoeGSS working on a graph-based ABM framework for HPC interested? talk with us! 13
Output: global maps 14
Example: green cars share Share of BEV cars in 2025 5% 10% 15% 20-27% 15
Evolution of new BEV car numbers: aggregate calibration New BEV cars bought in year 3e+05 Nr. of cars 2e+05 Source simulation study 1e+05 0e+00 2010 2015 2020 2025 Year 16
Large differences between countries market shares 2016: Norway 29%, Netherlands 6%, China 1% Source: ICCT blog 17
Battery Electic Vehicles (BEV) sales 150000 Policy parameter calibration in progress China: 2 US: 1 Italy: 0.8 Japan: 0.5 Norway: 20 match parameter with policies/ prices allow for decreasing numbers value 100000 50000 country China data China simulated Italy data Italy simulated Japan data Japan simulated Norway data Norway simulated United States data United States simulated 0 2009 2010 2011 2012 2013 2014 2015 step 18
(Interactive) map 19
THANK YOU! Questions? 20