Predictive Accuracy of a Hybrid Generalized Long Memory Model for Short Term Electricity Price Forecasting

TL;DR

This paper introduces a hybrid model combining advanced time series and neural network techniques to improve short-term electricity price forecasting accuracy, addressing the challenges posed by the non-storable nature of electricity.

Contribution

The study develops and evaluates a novel hybrid forecasting model integrating GARMA, G-GARCH, wavelet decomposition, and LLWNN, optimized with BP and PSO algorithms, demonstrating superior performance.

Findings

The proposed model outperforms existing parametric and non-parametric methods.

Wavelet decomposition enhances the model's ability to capture price dynamics.

Optimization with PSO improves the neural network's forecasting accuracy.

Abstract

Accurate electricity price forecasting is the main management goal for market participants since it represents the fundamental basis to maximize the profits for market players. However, electricity is a non-storable commodity and the electricity prices are affected by some social and natural factors that make the price forecasting a challenging task. This study investigates the predictive performance of a new hybrid model based on the Generalized long memory autoregressive model (k-factor GARMA), the Gegenbauer Generalized Autoregressive Conditional Heteroscedasticity(G-GARCH) process, Wavelet decomposition, and Local Linear Wavelet Neural Network (LLWNN) optimized using two different learning algorithms; the Backpropagation algorithm (BP) and the Particle Swarm optimization algorithm (PSO). The performance of the proposed model is evaluated using data from Nord Pool Electricity markets. Moreover, it is compared with some other parametric and non-parametric models in order to prove its robustness. The empirical results prove that the proposed method performs well than other competing techniques.