| |
 |
UNIVERSITY OF BUCHAREST
FACULTY OF PHYSICS Guest
2026-06-11 23:58 |
 |
|
|
|
Conference: Bucharest University Faculty of Physics 2026 Meeting
Section: Physics and Technology of Renewable and Alternative Energy Sources
Title:
Machine-Learning-Based Optimization of Renewable Energy Distribution and Storage for Domestic Use
Authors:
Tiberiu Adrian SALAORU(1), Marius Cristian POP(1), Vlad TUDORACHE PROHNITCHI(1,2), Adina Diana DOBRIN(3)
*
Affiliation:
1) National Institute for Aerospace Research “Elie Carafoli”, B-dul Iuliu Maniu 220, Bucharest 061126, Romania
2) Reykjavik University, Menntavegur 1, Reykjavik, Iceland
3) Faculty of Physics, University of Bucharest, Str. Atomistilor nr. 405, Magurele, Romania
E-mail
salaoru.tiberiu@incas.ro, pop.marius@incas.ro, vlad.tudorache.prohnitchi@gmail.com, dadinadiana1999@yahoo.com
Keywords:
solar and wind turbine power distribution, optimization of electrical power distribution, health monitoring system
Abstract:
For renewable energy systems, such as photovoltaic and wind turbine installations, efficiency represents a critical performance parameter. Improving overall efficiency is primarily pursued through two main directions: advancements in hardware and the development of intelligent control strategies. On the hardware side, a key objective is the optimization of electrical power distribution in order to minimize energy and excess generated power losses. To achieve this, power generation units must be dynamically connected or disconnected from loads and energy storage systems, allowing for flexible energy management. In addition, the power flow along each pathway must be continuously adjusted through automated control mechanisms to ensure optimal utilization of the generated energy. Such coordinated management enables the system to operate closer to its maximum efficiency, enhancing both energy yield and system reliability. In addition to the system-level optimization strategy, the paper proposes the use of a time-series neural-network-based energy optimization model for domestic renewable energy systems. The proposed model is intended to exploit the temporal evolution of relevant PV parameters and household energy-use patterns in order to support improved decision-making under dynamic operating conditions [1,2]. By learning sequential dependencies and short-term variations in renewable energy production and consumption, the time-series neural architecture can contribute to more adaptive energy harvesting, improved battery-charging and discharging decisions, and better coordination between renewable generation, storage, local consumption, and grid exchange [3].
References:
[1] M. Pop, M. Tudose, D. Visan, M. Bocioaga, M. Botan, C. Banu, T. Salaoru, A Machine Learning-DrivenWireless System for Structural Health Monitoring, INCAS BULLETIN, (print) ISSN 2066–8201, (online)ISSN 2247–4528, ISSN–L 2066–8201, vol 16, pp 77-93, https://doi.org/10.13111/2066-8201.2024.16.3.8,2024.
[2] Elman, J. L. (1990). Finding structure in time. Cognitive Science, 14(2), 179–211. https://doi.org/10.1207/s15516709cog1402_1
[3] LeCun, Y. (2022). A path towards autonomous machine intelligence. OpenReview. https://openreview.net/forum?id=BZ5a1r-kVsf
|
|
|
|