The energy sector is no exception to being subject to the rapid development of technology in today’s world, as disruptive smart solutions and analytical mechanisms continuously flood the industry. The ever-growing vast amount of publicly available energy-related datasets highly attribute to the stimulation of the increasing role of data analysis in optimization. In the field of energy consumption, predicting the future is of paramount importance, thus the majority use of data mining techniques nowadays in forecasting.
In my thesis, I demonstrate how I created a forecasting environment to predict the system load measures. In order to successfully estimate future values, it is critical to find, which parameters affect energy consumption and to run these variables in the chosen regression algorithm, resulting in the model, which will do the predicting. In the following section, I focused on the importance of implementing quantile regression, which gives a more detailed overview of the independent variables’ conditional distribution. I exhibit the mathematical and implementational aspects of the regression, and how I wired it in the forecasting environment. Furthermore, I had created a new evaluation and post-processing system, which aimed to optimize the estimations. As a final step, I compared the techniques I tried and summed up the results achieved.