هیدرات تتراهیدروفوران، سینتیک رشد، مسیر ترمودینامیکی طبیعی، تمایل شیمیایی، نمکهای سدیم هالید

Many recent studies have forecasted that the carbon dioxide emission rates will rise dramatically in near future due to several issues such as production of high carbon content heavy oils. Such large increases in carbon dioxide emission may lead to grave climate changes. Empirical modeling such as artificial neural networks with adequate training has been proved to be powerful tools to forecast the future trends of CO2 emission rates. A large collection of data including CO2 emission rates, annual GDPs and populations for 162 countries during 1980-2016 are gathered and used for training various back-propagation networks. A novel stabilized MLP network is introduced and its recall, validation and generalization performances are compared with two conventional networks, namely, our in-house regular (unstabilized) network and the optimal back-propagation network of MATLAB toolbox library. It is clearly shown that the optimal stabilization level is essential for filtering the noise and providing faithful generalizations and forecasting. Visual method is used to choose the optimal value of stabilization parameter. The stabilized ML P network predicts that the amount of CO2 emission remains almost constant during 2030-2050 for extremely populated countries and may dramatically reduce for low population nations.

کلیدواژه ها:

Carbon dioxide ، emission rate ، GDP ، population ، MLP.