STatistical approximation-based modEl EMulator (STEEM)

Welcome to STEEM, the STatistical approximation-based modEl EMulator

STEEM is a data-driven framework, developed as toolbox to facilitate fast model approximations (model emulator). Data-driven means that simulation models’ outputs are mapped to inputs and used to train a statistical model. The train model is then used to make many fast output approximations given new inputs, without the need to know the complex, and usually computationally heavy, mathematical landscape governing the original simulation models.

STEEM performs exploratory analysis on the large number of inputs/outputs set (simulation scenarios) to visualize sustainable transition pathways that lead to the desired policy outcome (simulation outputs) with limited sensitivity to external forces (simulation inputs). This functionality makes it a valuable toolbox for the creation of adaptive policy pathways. To do so, STEEM uses a heuristic algorithm called Patient Rule Induction Method (PRIM). More information on the algorithm behind STEEM can be found in TRANSrisk Deliverable 7.3 and in the following scientific publication:

Papadelis S., Flamos A. (2019) An Application of Calibration and Uncertainty Quantification Techniques for Agent-Based Models. In: Doukas H., Flamos A., Lieu J. (eds) Understanding Risks and Uncertainties in Energy and Climate Policy. Springer, Cham 

The major contributions of STEEM, in the context of both TRANSrisk and the wider modelling community, are:

  • Fast/interactive model approximations. This can help establish a tight loop between stakeholders and the modelling teams, thus making it possible to identify options and scenarios in a timely fashion that stakeholders either reject or explore further in terms of costs or impacts.
  • Quantification of the uncertainties governing the modelling assumptions. Through the relevance-based learning capability of the STEEM, users can evaluate what the most valuable data for the given task are.

Online Version

STEEM is available online in the following link:  http://transrisk-steem.tex.unipi.gr:31312

In this application STEEM is used to emulate and perform exploratory analysis of simulation inputs and outputs of the Agent-based Technology AdOption Model (ATOM), developed by the TechnoEconomics of Energy Systems laboratory (TEESlab) at the University of Piraeus.

ΑΤΟΜ simulates the decision of agents (i.e., Greek consumers) to invest in small-scale solar PV (i.e., up to 10kWp), considering supporting policies for the diffusion of PV, and external factors affecting the decision of agents. In the application in the link, 4 policy support schemes are considered:

  • Net-metering (as currently in operation in Greece),
  • Self-consumption with 30% subsidy of the battery’s cost,
  • Self-consumption with 50% subsidy of the battery’s cost,
  • Self-consumption with 65% subsidy of the battery’s cost.

The 4 external factors affecting the agents’ decision to invest, are:

  • The retail price of electricity,
  • The residential electricity demand,
  • The investment cost of a battery storage system, and
  • The investment cost of a small-scale PV system.

STEEM performs exploratory analysis on 500 scenarios (combinations of external factors) to derive 500 investment decisions (total PV capacity addition in MW) for each policy support scheme. The final outcome of this modeling example is an adaptation map, showing alternative policy pathways leading to the 2025 and 2030 small-scale solar PV capacity targets in Greece, respecting the intermediate target milestones (pathways follow the target trajectory with maximum ±20% deviation).

An easy to use toolbox

  • Visit transrisk-steem.tex.unipi.gr:31312
  • Press “Start Over”, wait for the “updating” notification to be replaced by “Dash” and then refresh the webpage. This procedure ensures correct results.

  • Choose a policy for simulation by moving the range-sliders:  

  • Every next policy should start from the year after a previous policy has ended  

  • Discover dead ends and alternative pathways

  • Hover for performance, target and cost info< >Please wait for the “updating” notification to be replaced by “Dash” to view your updated map  

  • Please wait for the “updating” notification to be replaced by “Dash” to view your updated map

  • Please do not try to reverse a simulation (i.e., undo a simulation from 2019 to 2020). Instead press “Start Over”, wait for the “updating” notification to be replaced by “Dash”, refresh your browser’s webpage and then simulate again.

Users with a simulation model, willing to perform exploratory analysis on their data and produce adaptive policy pathways, are welcome to contact us at teeslab@unipi.gr.

Go to top