Carbon dioxide emissions prediction using meta-heuristic methods for renewable and non-renewable sources' applications

The Shuffled Frog-Leaping Algorithm (SFLA) is a nature-inspired and swarm-based metaheuristic algorithm proposed in 2003. SFLA is an optimization algorithm that deals with the movements of frogs in a given population to reach the maximum amount of food with minimum movement. In this thesis, as a first phase, a hybrid method that increases the speed of reaching the optimum solution of the SFLA and minimizes the possibility of getting stuck in the local minimum is proposed. In the proposed hybrid method, the search capability of the SFLA algorithm is increased at first time with the Levy flight function. Then, the developed method (ILSFLAFA) is performed in a hybrid structure by using the Firefly Algorithm (FA) and the success of this proposed method is proven with the Benchmark functions. As a second phase, a new estimation method, that involves a multilayer feed forward artificial neural network, is proposed. The artificial neural network is trained with the ILSFLAFA. This newly developed hybrid-swarm-based artificial neural network (HSBNN) estimation method has been used to estimate the CO2 emission amount of Türkiye. The developed method, which is proposed in the literature as a first time, is compared with the estimation study conducted in 2016 on this issue, and the success and effect of the proposed method are demonstrated. Then, a new estimation method has been developed for Türkiye's CO2 emission estimation problem using the proposed HSBNN. At this stage, in order to find the best model, four different forecasting models are developed and the results are examined. The input variables of year, R&D investments, renewable energy ratio of total final energy consumption, population, urbanization, number of motor vehicles, energy consumption and GDP variables are used determining optimal parameters of artificial neural network (ANN). This data set for Türkiye is created by collecting data from different institutions. In the first prediction model, one layer is used in the hidden layer. The number of neurons in the layers of this method is studied as 8-12-1. In the other three estimation methods, two layers are used in the hidden layer. The number of neurons belonging to the second prediction model is 8-6-4-1. The neuron numbers of the artificial neural network used in the third and fourth prediction models are 8-6-6-1 and 8-8-4-1. In this study, which is conducted with the data of Türkiye's 1990-2018 years, the data set of the years 1990-2012 is used to create the model, while the data set of the years 2013-2018 is used in the testing phase of the model. The results obtained at this stage of the study are given in detail and the best estimation model is selected and the next step is taken. In the last stage of the thesis, Türkiye's future CO2 emission estimation is examined in three different scenarios. These scenarios developed; the estimates published by official institutions, the increase rates suggested in the literature studies and the increase rates in Türkiye in recent years have been created. Türkiye's future CO2 emission estimation is made towards 2030. The results obtained at the end of the study are discussed in detail and the effect of renewable energy use on the amount of CO2 emissions is emphasized. In addition, the obtained results show that the proposed estimation method can be successfully applied in areas requiring future estimation. Keywords: renewable energy, CO2 emission, neural networks, shuffled frog-leaping algorithm, Levy flight, firefly algorithm, hybrid structure, sustainability, green technology