TY - GEN
T1 - Adaptive Energy Management Systems for Smart Grids
T2 - 13th IEEE International Conference on Smart Grid, icSmartGrid 2025
AU - Khan, Baseem
AU - Ramakrishna, N. S.S.
AU - Palakurthi, Shashank
AU - Tunga, Neves Binza
AU - Hushein, R.
AU - Mandadi, Ranadheer Reddy
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - An Adaptive Energy Management System (AEMS) is proposed for smart grids, which integrates machine learning techniques with internet of things connectivity so as to improve predictive control ability, response times and even the efficacy of operations. The system effectively manages residential energy sources and distributed energy assets by means of load forecasting based on Long Short-Term Memory (LSTM), dynamic demand response, controlling assets with big data. This concept was tested out on a mock-up and showed that the LSTM model had the highest accuracy: a Mean Absolute Percentage Error (MAPE) of just 2.14 percent. This allowed for optimal load scheduling, cutting household energy costs by 28.17 percent. The AEMS also improved utilization of renewable energy, with 89.2% solar and 72% less dependence on the grid. Grid stability was improved when deviation in frequency and voltage dropped by more than 70 percent. In addition, the system is capable of triggering 92 demand response events every month. When tested for scalability, it was found that the latency was still acceptable (<250 ms) and packet loss low (<2%) even under 200-node configurations; this shows that it can be used at scales large enough to be meaningful. In conclusion, this paper introduces a flexible and intelligent platform for improving the performance of smart grids. The proposed solution can integrate renewable energy more efficiently into grid systems, generate savings for users at the same time as strengthening cyber electrical reliability.
AB - An Adaptive Energy Management System (AEMS) is proposed for smart grids, which integrates machine learning techniques with internet of things connectivity so as to improve predictive control ability, response times and even the efficacy of operations. The system effectively manages residential energy sources and distributed energy assets by means of load forecasting based on Long Short-Term Memory (LSTM), dynamic demand response, controlling assets with big data. This concept was tested out on a mock-up and showed that the LSTM model had the highest accuracy: a Mean Absolute Percentage Error (MAPE) of just 2.14 percent. This allowed for optimal load scheduling, cutting household energy costs by 28.17 percent. The AEMS also improved utilization of renewable energy, with 89.2% solar and 72% less dependence on the grid. Grid stability was improved when deviation in frequency and voltage dropped by more than 70 percent. In addition, the system is capable of triggering 92 demand response events every month. When tested for scalability, it was found that the latency was still acceptable (<250 ms) and packet loss low (<2%) even under 200-node configurations; this shows that it can be used at scales large enough to be meaningful. In conclusion, this paper introduces a flexible and intelligent platform for improving the performance of smart grids. The proposed solution can integrate renewable energy more efficiently into grid systems, generate savings for users at the same time as strengthening cyber electrical reliability.
KW - Adaptive Energy Management System (AEMS)
KW - Internet of Things (IoT)
KW - Load Forecasting
KW - Machine Learning
KW - Smart Grids
UR - https://www.scopus.com/pages/publications/105013473926
U2 - 10.1109/ICSMARTGRID66138.2025.11071808
DO - 10.1109/ICSMARTGRID66138.2025.11071808
M3 - Conference contribution
AN - SCOPUS:105013473926
T3 - 13th IEEE International Conference on Smart Grid, icSmartGrid 2025
SP - 434
EP - 439
BT - 13th IEEE International Conference on Smart Grid, icSmartGrid 2025
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 27 May 2025 through 29 May 2025
ER -
