Abstract

AbstractIn this chapter, a step-by-step application of calibrating an agent-based model is presented. In particular, an agent-based model for small-scale PV adoption was calibrated on the historical data for the small-scale solar PV capacity additions that took place in Greece from January 2010 to February 2013. The process of the model calibration allowed to (a) quantify and take into consideration uncertainties that are related to the characteristics and the decision-making criteria of the agents (i.e. independent PV power producers), in contrast to the more obvious uncertainties, such as technology costs, and (b) use the calibration results to explore the plausible—given the historical data—behaviour of the potential PV adopters in Greece under the new net-metering scheme (in effect as of mid-2015).

Highlights

  • A common approach to validating the extent at which an Agent-based models (ABMs) is realistic is through calibration using historical data

  • This process is an opportunity to deal with model uncertainty, which is the specific type of uncertainty that stems from the fact that there exist many variations of an ABM that are all plausible under the same set of historical data

  • In the application presented in this chapter, a Gaussian process (GP) emulator—the STatistical approximation-based modEl EMulator (STEEM) of the Technoeconomics of Energy Systems Laboratory (TEESlab)—was fitted on the results from a number of parameter combinations at which the ABM was run

Read more

Summary

Introduction

Models try to narrow the differences between decision-makers’ thinking, reasoning, representation and computing (Doukas 2013). Agent-based models (ABMs) fall into the second choice (Flamos 2016). This means that developing a new ABM is an interesting endeavour, but unless we. A common approach to validating the extent at which an ABM is realistic is through calibration using historical data If done right, this process is an opportunity to deal with model uncertainty, which is the specific type of uncertainty that stems from the fact that there exist many variations of an ABM that are all plausible under the same set of historical data. 4 presents the details of fitting a Gaussian process emulator on the ABM results and Sect.

The ABM for the Diffusion of Small-Scale Solar PV
The Concept of Emulators
Gaussian Processes for Regression
Þ σ2 ðx1Þ f
Benefits of Using Gaussian Processes as Emulators
Fitting the GP Emulator
Diagnostics
Sensitivity Analysis
The Patient Rule Induction Method
Calibration and Extrapolation Results
Discussion
Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call