Forecasting methane spot price by combining artificial networks and geostatistics

With the increasing importance of methane gas in the context of the energy transition and climate change mitigation, accurate forecasting of the spot price of natural gas plays a crucial role in commercial policies and strategies, as well as in cost-effective market opportunities. Recently, many advanced methodologies have been used to model the gas spot price, such as several artificial intelligence algorithms, especially artificial neural networks (ANNs). Their combination with traditional time series approaches, such as auto-regressive moving average (ARMA) models, was implemented to improve forecasting accuracy. In this context, a new hybrid model, which combines a geostatistical tool, i.e., the ordinary Kriging, and ANNs, in particular the long short-term memory and the gated recurrent unit neural networks, is provided. In addition, the supremacy of the proposed hybrid model is assessed through a comparison with respect to the pure ANNs and a well-known hybrid model available in the literature. Methodologies are applied to the spot price of natural gas in Italy using data from the Virtual Exchange Point (VEP).