Evaluation of Ambient and Soil Temperature for Natural Gas Consumption Prediction Using ANFIS
Year: 2026
Type: article
Abstract: Natural gas consumption (NGC) is an important consideration in energy planning, supply management, and the efficient operation of natural gas distribution systems. Among the meteorological variables affecting NGC, temperature is one of the most widely used predictors because of its close relationship with heating-related consumption. In this study, average NGC per customer in Konya, Türkiye, was predicted using an Adaptive Neuro-Fuzzy Inference System (ANFIS) under different input configurations based on Ambient Temperature (AT) and Soil Temperature (ST). Three ANFIS models were developed using ST alone (ANFIS-NGC1), AT alone (ANFIS-NGC2), and the combination of ST and AT (ANFIS-NGC3). The models were trained using data from 2018 to 2021 and tested using data from 2022. The testing R² values were 0.906, 0.980, and 0.995 for ANFIS-NGC1, ANFIS-NGC2, and ANFIS-NGC3, respectively, while the corresponding Mean Absolute Percentage Error (MAPE) values were 30.350%, 11.420%, and 9.710%. The results showed that AT provided better prediction performance than ST when used individually, whereas the highest accuracy was obtained when AT and ST were used together. These findings indicate that ST can provide additional predictive information when combined with AT and that the combined temperature input configuration can improve monthly NGC prediction performance.
Institution: Konya Technical University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7233583323
