Abstract:

The reference works describe numerous forecasting methods used directly in power demand forecasting process. Most of them are universal and multi-sector which results in numerous advantages, including availability, understandability as well as easy and popular use. However, the methods may turn out to be of low effectiveness if the data to be used for the analysis is complex and mass and if the specific needs of a given company are considered. This is why this article captures the problem of power demand forecasting from the microperspective in the context of the unique forecasting requirements typical of power trading companies. The major objective of this article is to diagnose the reasons, advantages and directions of power demand forecasting in such entities. To implement the goal established in this way, the article authors use the case study (to diagnose the reasons, needs and advantages) of a trading company as well as reference works (to assess and pre-select) the forecasting methods, paying particular attention to forecasting models used in the power sector. According to the analyses carried out, the best tool to forecast power needs in trading companies could be deep machine learning using neural networks.