Quantifying demand and weather uncertainty in power system models using the m out of n bootstrap
This paper introduces a novel approach to quantify demand weather uncertainty in power system models. Recent studies indicate that such sampling uncertainty, originating from demand weather time series inputs, leads to significant uncertainties in model outputs and should not be ignored, especially with increasing levels of weather-dependent renewable generation. However, established uncertainty quantification approaches fail in this context due to the computational resources and additional data required for Monte Carlo-based analysis. The methodology introduced in this paper quantifies demand weather uncertainty using a time series bootstrap scheme with shorter time series than the original, enhancing computational efficiency and avoiding the need for any additional data. It can be used both to quantify output uncertainty and to determine optimal sample lengths for prescribed confidence levels. Simulations are performed on three generation transmission expansion planning models and a simple test is introduced allowing users to determine whether estimated uncertainty bounds are valid. Code applying the methodology to the models used in this paper is provided as open-source software.
READ FULL TEXT