In this study, we focused on predicting transmission losses with the aim of reducing errors and optimizing the costs of their purchase. To achieve this goal, we decided to develop and test a completely new approach that would enable predictions that are more accurate and thus contribute to reducing the cost of purchasing transmission losses.
Our approach involved dividing Slovenia into two parts: the northwestern region (Severna Primorska) and the rest of the country. In each part, we carried out separate predictions of transmission losses to determine if this approach could improve the accuracy of the predictions.
As part of our research, we also made predictions for the entire country, allowing a direct comparison between the two approaches. The results showed that the method that predicts transmission losses for the entire country was more accurate. This finding highlights the importance of a holistic approach to predicting transmission losses, as this method is better able to capture various influences and patterns throughout the entire network.
We also delved into the reasons for poor predictions in the northwestern region and found that the key factor was the inaccurate prediction of the operation of the Avče pump hydroelectric power plant. This power plant represents a significant part of the transmission losses in the northwestern loop, and obtaining accurate insights into its operation is crucial for improving predictions in this region.
During the testing period, we evaluated the created method that utilizes the XGBoost algorithm and found that it predicted transmission losses more accurately than the method that is currently used. This observation indicates promising results for the application of machine learning to enhance predictive models.
|
