Geographical information systems supported statistical analysis of instant natural gas consumption data

This study is about machine learning, regression analysis, decision tree and Geographic Information Systems (GIS) supported analysis of the factors affecting the natural gas consumption collected on an hourly basis and the creation of consumption profiles with 17 data loggers in Isparta city center. Natural gas, 99% supplied externally, used for heating must me used efficiently due to the changes in geopolitical conditions. Data-based management strategies should be developed to increase efficiency. Hourly analysis of the factors affecting consumption and detailed profiles make the study important. Significant results were found in multiple linear regression model used analyze meteorological and consumption data which is clustered into groups by decision trees using area, direction, floor and height of flat, between them. Three different scenarios were studied in calculating the most appropriate number of data collector devices to model the whole city. D2LogTAY GIS module developed to analyze and determine buildings where the devices will be installed. Four sensitivity scenarios studied to reduce the natural gas usage by reporting the average consumption in similar flats (area, direction, floor and height) to increase awareness and decrease greenhouse gas emission amounts. Detailed data supplied %19 to %32 differences and decision tree 23% difference from regular averages in scenarios. Consumption on weekdays is higher than the weekend, Wednesdays and Thursdays higher than others, and the consumption of school breaks is more than other holidays. It has been observed that hourly consumptions in residences are mostly similar and those that differ can be detected.