Tariff classification with multi-dimensional deep learning using time series load data

In this research, it is aimed to develop a novel model that predicts the tariff group of consumers, based on electricity consumption data. After becoming widespread of Smart Grids, network elements need to be analyzed correctly in balancing energy supply and demand, ensuring cost efficiency and managing consumer behavior. These analyzes is very important for the correct operation of smart grids components such as Demand Side Management (DSM). In the initial chapters of the thesis, a general framework of energy consumption is drawn and its importance in the economic growth and sustainable development of modern societies is emphasized. In this study, a novel Conventional Neural Networks (CNN) and Long-Short Term Memory (LSTM) model is proposed to classify electricity consumption data with deep learning methods. Relational features have been increased by CNN with 2D perspective on time series data and temporal features have been extracted by LSTM talent that is learning long term independence. Complex time series data is analyzed by combining two methods. This hybrid model can classify the tariff in electricity consumption more precisely and accurately. The data set used in this study includes annual consumption data of 5 different tariff groups of Afyonkarahisar province and includes active, inductive and capacitive measurements. During the data processing, statistical methods such as z-score were applied to eliminate anomalies and determining missing values and time series are set on a same size. When the proposed model was compared with other deep learning models, it was observed that the proposed model has better results with %87,83 training accuracy and %82,37 test accuracy. As a result, the findings of the thesis revealed that the classifications obtained from electricity consumption data provide high accuracy in the analysis of user behavior and tariff prediction. It has been stated that, this model can be used in strategic fields such as consumers segmentation, demand forecasting and demand side management of energy distribution companies and can contribute smart grids applications. The thesis is also considered an important tool in sustainable management of energy planning and supply-demand balance.