How Sampling and Averaging Historical Solar and Wind Data Can Distort Resource Adequacy

Abstract

Capacity planning models and resource adequacy assessments have often relied on averaging and sampling techniques that disregard important reasonably expected interactions of weather-based resources. We provide a method to capture the economic value of information and reliability risk from using inadequate sample data to design sustainable systems with high renewable generation. Analysis of long run portfolio cost and sources of uncertainty shows as much as a 16% system cost increase with a 38-fold increase in expected unserved energy when average renewable outputs are modeled rather than a 10 year hourly coincident sample, which illustrates the pitfalls of averaging data and ignoring temporal interdependencies. Investment recommendations can significantly differ depending on which years and how many years are included in the analysis. We show that selecting the wrong year can increase system costs by over 4% with a 7-fold increase in expected unserved energy, failing to meet planned reliability and renewable design targets. It is possible for a single year of coincident load-wind-solar data to reasonably approximate system characteristics; however, the best year changes with renewable penetration.