Representative Scenarios to Capture Renewable Generation Stochasticity and Cross-Correlations

TL;DR

This paper introduces a new method for generating representative power system scenarios that accurately capture renewable generation variability and cross-correlations, improving planning accuracy.

Contribution

It proposes a systematic approach and new metrics for scenario generation that include both seasonal and extreme days, enhancing diversity and relevance.

Findings

Generated scenarios include a balanced mix of seasonal and atypical days

New validation metrics effectively evaluate scenario quality

Trade-off analysis for optimal number of scenarios for specific applications

Abstract

Generating representative scenarios for power system planning in which the stochasticity of renewable generation and cross-correlations between renewables and load are fully captured, is a challenging problem. Traditional methods for scenario generation often fail to generate diverse scenarios that include both seasonal (frequently occurring) and atypical (extreme) days required for planning purposes. This paper presents a methodical approach to generate representative scenarios. It also proposes new metrics that are more relevant for evaluating the generated scenarios from an applications perspective. When applied to historical data from a power utility, the proposed approach resulted in scenarios that included a good mix of seasonal and atypical days. The results also demonstrated pertinence of the proposed cluster validation metrics. Finally, the paper presents a trade-off for determining optimal number of scenarios for a given application.