At the 2025 CASE conference on Advancements in Sustainable Engineering, the team led by M. A. Qureshi and N. Christofides will unveil their unified machine‑learning framework for energy‑profile forecasting in Cypriot neighborhoods. Key highlights include:
- Leveraging Long Short‑Term Memory (LSTM) networks to predict photovoltaic (PV) output and household electricity consumption at the individual‑prosumer level.
- Training on a year‑long dataset of minute‑resolution smart‑meter readings from over 200 homes across urban, suburban and peri‑urban districts of Nicosia.
- Achieving mean absolute percentage errors below 7 % for PV forecasts and under 10 % for load predictions, outperforming baseline gradient‑boosted models.
- Demonstrating how accurate, near‑real‑time forecasts can improve demand‑response scheduling, reduce imbalance penalties and inform dynamic pricing schemes within peer‑to‑peer energy communities.
This work promises to equip community managers and aggregators in Cyprus with robust forecasting tools, enabling more reliable integration of distributed renewables and smarter local energy markets.
M. A. Qureshi, N. Christofides, T. Leontiou and M. Lestas, “Towards a general machine learning model for energy profile predictions in smart energy communities in Cyprus,” Conference on the Advancements in Sustainable Engineering (CASE 2025), Cyprus, 2025.