Using a Local Framework Combining Principal Component Regression and Monte Carlo Simulation for Uncertainty and Sensitivity Analysis of a Domestic Energy Model in Sub-City Areas

Abstract

Domestic energy modelling is complex, in terms of user input and the approach used to define the model; therefore, there is an increase in the sources of uncertainties. Previous efforts to perform sensitivity and uncertainty analyses have focused on national energy models, while in this research, the objective is to extend traditional sensitivity analysis and use a local framework combining principal component regression and Monte Carlo Simulation. Therefore, in our method the total amount of the energy output’s variance is decomposed, in relative terms, according to the contribution of the different predictor parameters. Our framework provides compelling evidence that local area characteristics are important in energy modelling and those national and regional indexes and values may not properly reflect the local conditions, resulting in programmes and interventions that will be sub-optimal. Furthermore, our uncertainty methodology uses a three dimensional integrative taxonomy and a concept map. The concept map identified concrete terminal causes of uncertainty within the taxonomic framework of sources, issues, sub-issues and a further abstraction of those quantities in terms of accuracy and precision. Understanding uncertainties in this way provides a possible framework for modellers, policy makers and data collectors to improve practice in key areas and to reduce uncertainty.

Keywords

• concept map
• cities
• Monte Carlo Simulation
• neighbourhood urban energy modelling
• principal component regression
• sensitivity analysis
• uncertainty taxonomy