{Rudolfovo – Znanstveno in tehnološko središče Novo mesto vabi, da se udeležite znanstvenega seminarja z naslovom »Robotina - Energy Consumption Prediction with LSTMs and Hybrid models«, ki ga bo predstavila dr. Jelena Joksimović.
Seminar bo v četrtek, 26. septembra 2024, ob 10. uri v dvorani K01 Razvojnega centra Novo mesto.
Povzetek:
»In the context of hybrid energy systems that integrate multiple energy sources, including solar power, supercapacitors, UPS batteries, generators, hydrogen cells, and the grid, accurate energy consumption prediction is essential for efficient energy management. This work presents the application of Long Short-Term Memory (LSTM) neural networks and hybrid models, developed at our Laboratory for Energy Management (LabE), and applied to two additional Robotina buildings to forecast energy usage. Instead of increasing the prediction step for LSTMs, we adopted an interval-based approach, where 96 one-step LSTM predictions were made, each corresponding to 15-minute intervals. These predictions were then combined and weighted against the centroid of the behavioral cluster to which the day belonged, producing an accurate one-day energy forecast. Our experimental results demonstrate that this hybrid method—by leveraging both temporal predictions and daily behavioral patterns—effectively balances computational load and prediction accuracy, making it well-suited for short-term energy forecasting in hybrid energy systems.«
Vljudno vabljeni! Seminar bo potekal v angleškem jeziku.
Dodatne informacije: dr. Jelena Joksimović, jelena.joksimovic@rudolfovo.eu, Katja Sinur, katja.sinur@rudolfovo.eu.
Comments