Developing load-predicting tool for day-ahead schedule planning

OData support
Supervisor:
Dr. Divényi Dániel Péter
Department of Electric Power Engineering

Load forecasting has been an area of research for decades now and thousands of scientific articles has been published about this topic. Despite the estimation of electricity consumption has such old traditions, it still has its new challenges. Currently the energy sector is going through drastic changes due to the boom of the renewable energy sources, distributed energy production and also the deregulation of the electricity market. These factors greatly complicate the forecast of the consumer demand.

Appropriate forecasting has an outstanding importance as without it safe and economical operation of the power grid cannot be maintained. A precise load forecast is essential for planning the production of power plants, running the transmission network and the timing of the system development. Apart from this, an efficient system also has significant economic advantage, with a prompt forecast it is possible to reduce costs and in the meantime reduce pollution.

In my thesis, first I explain the concept of balancing energy and its implemented settlement method in Hungary. After that I will analyse the gained data. With different mathematical methods, I am aiming to find patterns, similarities and coherence within the measured data. I will go in more details about the potential of load estimation and its field of utilization. Later I will describe the functioning of the neural network and similar days’ approach. I will also explain the structure of the neural network, the necessary steps to train them and my own decisions during the implementation. I will examine which of the implemented solutions provide the best results and I will compare it with the forecast of the electricity trader. For closing I am proposing ideas to further improve the methods.

Downloads

Please sign in to download the files of this thesis.