Short-Term Electricity Price Forecasting with Recurrent Regimes and Structural Breaks

Abstract

This paper develops a new approach to short-term electricity forecasting by focusing upon the dynamic specification of an appropriate calibration dataset prior to model specification. It challenges the conventional forecasting principles which argue that adaptive methods should place most emphasis upon recent data and that regime-switching should likewise model transitions from the latest regime. The approach in this paper recognises that the most relevant dataset in the episodic, recurrent nature of electricity dynamics may not be the most recent. This methodology provides a dynamic calibration dataset approach that is based on cluster analysis applied to fundamental market regime indicators, as well as structural time series breakpoint analyses. Forecasting is based upon applying a hybrid fundamental optimisation model with a neural network to the appropriate calibration data. The results outperform other benchmark models in backtesting on data from the Iberian electricity market of 2017, which presents a considerable number of market structural breaks and evolving market price drivers.

Calibration window selection based on change-point detection for forecasting electricity prices

“…However, the automation of the forecasting process may make the trade-off worthwhile. If this is the case for more complex models than the autoregressive ones considered here or the shallow neural network in [2], e.g., LASSO-estimated AR (LEAR) and deep neural networks [6], is left for future work.…”

“…The key feature is its focus on the smallest local sections of the data on which the existence of a change-point is suspected. A change-point is said to occur when the behavior of the series changes significantly [2]. See www.changepoint.info for an excellent review site and software repository on this topic.…”

“…This work is inspired by a recent article [2], which utilized a relatively simple change-point detection method [9] to split the time series into segments with the 'same' price level, and an ICCS 2021 paper [7], which employed the k-nearest neighbors (k-NN) algorithm to select the calibration sample based on similarity over a subset of explanatory variables. Here, we utilize a recently proposed change-point detection algorithm -the Narrowest-Over-Threshold (NOT) Fig.…”

Forecasting Electricity Prices: Autoregressive Hybrid Nearest Neighbors (ARHNN) Method

“…However, in these cases, the calibration sample consists of the most recent and successive observations. Methods that rely on machine learning employ a variety of techniques such as artificial neural networks [24], support vector machines [27], clustering algorithms [22] or a combination of them [17]. It is worthy to mention that a hybrid approach that employs statistical and machine learning techniques has been also proposed in the literature.…”

Redes Neurais Atencionais aplicadas a modelagem e previsão de preços no Mercado de Eletricidade Brasileiro

“…Outra forma de abordar a questão é treinar os modelos em janelas passadas similares. Estas janelas podem ser identificadas por meio de algoritmos de clusterização (MARCOS et al, 2020). Do ponto de vista da volatilidade como um termo adicional ao modelo de predição, em econometria é difundido o uso de GARCH.…”