Research of natural gas consumption profile in sinop and modeling with artificial neural networks

In this thesis, natural gas consumption is estimated for Sinop city by using artificial neural networks and multiple regression analysis methods. As factors affecting natural gas consumption, the temperature, number of users, unit price were taken into consideration and the monthly consumption amount was tried to be estimated. In artificial neural networks, the number of neurons and activation function values were changed and the most suitable method was found. R2 values were compared in the prediction studies. Also, the effect of input parameters on natural gas consumption was statistically investigated by performing analysis of variance (ANOVA). As a result, it was observed that the artificial neural networks method gave successful results as compared to the multiple regression analysis method. In artificial neural networks, the most successful results were obtained from 3-10-1 as the network architecture, LOGSIS as a transfer function and Levenberg-Maquardt as a training function. In this network structure, the R2 value was calculated as 0.99346. It has been seen that artificial neural networks can be used in natural gas consumption estimation for Sinop province.