This paper presents a detailed comparative analysis of two data-driven probabilistic forecasting methods - a densitybased Empirical Copula (EC) and a non-parametric Quantile Regression (QR) - in the context of load forecasting for day-ahead market applications. In modern power systems with increasing stochastic renewable production, Distribution System Operators (DSOs) face the critical task of accurately forecasting production and consumption to strategically place bids in the day-ahead market. Any deviation from their dispatch plan can lead to the costly activation of reserve power. For DSOs with a small number of customers, or those requiring more detailed forecasts, traditional statistical methods may be inadequate due to their reliance on parametric statistical distributions of forecasted variables. This paper evaluates two probabilistic load forecasting methods that effectively operate without the need for parametric distribution assumptions: the EC and the QR. The associated forecasting models are identified with suitable optimization strategies and their performance is evaluated using standard probabilistic metrics on a real-world electricity consumption dataset. The analysis focuses on day-ahead 24-hour forecasts, and compares the proposed methods against a persistence baseline method. The results demonstrate the superiority of both EC and QR over the baseline, with the EC excelling in forecast accuracy, while QR showing dominance in prediction intervals.

A Comparative Analysis of Empirical Copula and Quantile Regression Methods for Probabilistic Load Forecasting

Anna Rita Di Fazio;Sara Perna
;
2024-01-01

Abstract

This paper presents a detailed comparative analysis of two data-driven probabilistic forecasting methods - a densitybased Empirical Copula (EC) and a non-parametric Quantile Regression (QR) - in the context of load forecasting for day-ahead market applications. In modern power systems with increasing stochastic renewable production, Distribution System Operators (DSOs) face the critical task of accurately forecasting production and consumption to strategically place bids in the day-ahead market. Any deviation from their dispatch plan can lead to the costly activation of reserve power. For DSOs with a small number of customers, or those requiring more detailed forecasts, traditional statistical methods may be inadequate due to their reliance on parametric statistical distributions of forecasted variables. This paper evaluates two probabilistic load forecasting methods that effectively operate without the need for parametric distribution assumptions: the EC and the QR. The associated forecasting models are identified with suitable optimization strategies and their performance is evaluated using standard probabilistic metrics on a real-world electricity consumption dataset. The analysis focuses on day-ahead 24-hour forecasts, and compares the proposed methods against a persistence baseline method. The results demonstrate the superiority of both EC and QR over the baseline, with the EC excelling in forecast accuracy, while QR showing dominance in prediction intervals.
File in questo prodotto:
File Dimensione Formato  
submitted.pdf

solo utenti autorizzati

Licenza: Copyright dell'editore
Dimensione 1.62 MB
Formato Adobe PDF
1.62 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11580/107428
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
social impact