Using artificial neural network models in Turkey electricity consumption forecast for coming period

One of the most needed types of energy worldwide is electric energy. We can see the need for electrical energy at every point of our daily lives. In many areas such as transportation, medicine, communication, industry and technology, electrical energy is an indispensable need for people. According to data from the Ministry of Energy and Natural Resources, electricity consumption in Turkey has increased to 254 billion kWh in 2018, while 249 billion kWh in 2017. The need for electrical enegy needs to increase continuously, the electricity cannot be stored and the electricity needs to be consumed at the moment it is produced. Because of these reasons supply-demand balance needs to be maintained. Failure to supply with this continuously increasing demand creates problems for countries in economic terms. This demand should be supported by an effective demand forecasting method. Strategic goals, planning and tactical decisions required for electricity supply to meet demand play an important role in the electricity market. It is very important to estimate electricity consumption in order to increase important factors such as economy, welfare level, production, technology and industry of our country. If the energy supply is lower than the demand, the welfare level will decrease, the risk of economic crises will increase, problems in production and other areas will arise. In addition to being very low, too high will also cause waste of electrical energy. Therefore, it is important to find the demand forecast value close to the actual value. Recently, with the development of machine learning, traditional demand forecasting methods have been replaced by artificial neural networks. In the study, the amount of electricity consumption for the next period was tried to be estimated by artificial neural network method which is one of the quantitative estimation methods. Properties, structure, models, advantages and disadvantages of artificial neural networks have been examined in detail. According to previous studies, population, production, number of transformers, Gross Domestic Product (GDP) and investment in electricity were determined as factors affecting electricity consumption. These factors were used as variables of artificial neural network. Firstly, the variables of the year to be estimated were tried to be estimated by Holt method. Excel program was used for Holt iv method. Then, the growth rate of the data was found and the environment was prepared for the network to learn more easily. The data obtained were normalized by D_Min_Max normalization method and a data set consisting of 0.1-0.9 range was obtained. Artificial neural network method was applied by using MATLAB R2018a program. In this study, multilayer perceptron (MLP) model, which is the most used model in artificial neural networks, is used. In this structure, delta learning rule was used as the learning rule, sigmoid function was used as the transfer function, MSE was the mean of mean squared error is used as the performance measurement. As a result of this study, artificial neural networks produced estimation value close to real value. It has been observed that the predicted value is so close to the real value with an accuracy of 0,999308677. It is observed that it is a suitable model for the estimation of electricity consumption for the future periods and for the estimation of similar studies.