Analysing energy poverty in warm climate zones in Spain through artificial intelligence

Abstract

Using automated tools to detect energy poverty (EP) is a developing field. Artificial intelligence and data mining could be used to provide solutions to reduce EP cases. As for Spain, there is no study addressing this characterization that could be significant in warmer zones of the country (i.e., the most exposed zones to climate change). Simulated energy consumption data were used with data of energy prices and family units' incomes based on the public income indicator of multiple effects (IPREM in Spanish). In addition, the high share of energy expenditure in income (2 M) was used to assess EP. A total of 36,230,400 cases were simulated to train and test 312 prediction models, 104 by each algorithm. The algorithms were multilayer perceptron (MLP), random forest (RF), and M5P. The results showed that these three algorithms were appropriate, with tree-type models obtaining better estimates. For greater effectiveness, prediction models should also be used for the income threshold considered in their development. The results also showed the utility of artificial intelligence in the prediction of EP without performing an energy analysis in detail, thus optimizing energy managers and social workers' work. In addition, prediction tools could be used to estimate monthly family units’ EP situation.