TY - GEN
T1 - Applying an Agent-based Distributed AI Framework to Forecast Power for the Mini-Grid Stability
AU - Ioannou, Iacovos I.
AU - Javaid, Saher
AU - Vassiliou, Vasos
AU - Pitsillides, Andreas
AU - Tan, Yasuo
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Renewable energy sources, expected to form about 70% of power systems by 2050, bring challenges like fluctuating outputs and grid instability. Advanced power monitoring systems, crucial in environments like research facilities and hospitals, must navigate these dynamic scenarios. Traditional power management, especially Uninterruptible Power Supply (UPS) systems, often needs to catch up due to high costs and limited response to varied power demands, focusing mainly on constant power supply without differentiating between constant and fluctuating loads. In response, Artificial intelligence (AI) techniques are becoming indispensable for real-time power prediction and control. A distributed AI framework forecasts power needs, considering renewable sources, loads, and storage. This is key to ensuring smooth mini-grid operations, balancing operational demands with environmental considerations, and advancing intelligent energy management. Such systems are essential in optimizing energy usage, aligning it with available power to enhance efficiency and reduce waste. This is particularly important for mini-grids, with or without UPS systems, where predictive monitoring can substantially cut operational costs and extend lifespan.The paper focuses on providing consistent, constant, and fluctuating power by predicting mini-grid power needs hourly from the previous day's data. We use a Temporal Convolutional Network (TCN) for time series prediction, integrated within the BDIx agent's belief system through TensorFlow Lite. This approach accurately predicts upcoming power needs, ensures smooth operation, and prevents power outages. The TCN model's predictive capabilities highlight a significant stride in combining AI with energy management to address the complexities of modern power systems.
AB - Renewable energy sources, expected to form about 70% of power systems by 2050, bring challenges like fluctuating outputs and grid instability. Advanced power monitoring systems, crucial in environments like research facilities and hospitals, must navigate these dynamic scenarios. Traditional power management, especially Uninterruptible Power Supply (UPS) systems, often needs to catch up due to high costs and limited response to varied power demands, focusing mainly on constant power supply without differentiating between constant and fluctuating loads. In response, Artificial intelligence (AI) techniques are becoming indispensable for real-time power prediction and control. A distributed AI framework forecasts power needs, considering renewable sources, loads, and storage. This is key to ensuring smooth mini-grid operations, balancing operational demands with environmental considerations, and advancing intelligent energy management. Such systems are essential in optimizing energy usage, aligning it with available power to enhance efficiency and reduce waste. This is particularly important for mini-grids, with or without UPS systems, where predictive monitoring can substantially cut operational costs and extend lifespan.The paper focuses on providing consistent, constant, and fluctuating power by predicting mini-grid power needs hourly from the previous day's data. We use a Temporal Convolutional Network (TCN) for time series prediction, integrated within the BDIx agent's belief system through TensorFlow Lite. This approach accurately predicts upcoming power needs, ensures smooth operation, and prevents power outages. The TCN model's predictive capabilities highlight a significant stride in combining AI with energy management to address the complexities of modern power systems.
KW - Distributed AI
KW - constant power load prediction
KW - distributed power loads
KW - distributed power sources
KW - fluctuating power prediction
KW - power control
KW - power fluctuations
UR - http://www.scopus.com/inward/record.url?scp=85205832098&partnerID=8YFLogxK
U2 - 10.1109/ICCE-Taiwan62264.2024.10674231
DO - 10.1109/ICCE-Taiwan62264.2024.10674231
M3 - Conference contribution
AN - SCOPUS:85205832098
T3 - 11th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2024
SP - 255
EP - 256
BT - 11th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 11th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2024
Y2 - 9 July 2024 through 11 July 2024
ER -