TY - GEN
T1 - Intrusion Detection Tool for Residential Consumers Equipped with Smart Meters
AU - Sharma, Akshat
AU - Saini, Vikash Kumar
AU - Kumar, Rajesh
AU - Al-Sumaiti, Ameena S.
AU - Heydarian-Forushani, Ehsan
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - Residential load forecasting holds significant importance in the supervision of contemporary power grids, aiding utilities in efficiently allocating resources to meet the surging electricity demand. However, the proliferation of smart meters and IoT devices introduces new challenges to the power grid, particularly concerning the security and integrity of load forecasting systems. Intrusion attacks targeting load signature data can distort predictions, potentially leading to power supply disruptions and substantial economic repercussions. This research introduces a robust residential load forecasting tool based on a secure deep learning algorithm. The system demonstrates a swift ability to detect attacks and differentiate between manipulated data from cyber-attackers and authentic data. The initial step involves identifying intrusions within the dataset, utilizing a decision tree classifier that relies on features extracted through explainable artificial intelligence. This process achieves a commendable 96.6% accuracy in intrusion classification. Subsequently, the identified intrusions are counteracted via a meticulously designed clipper-clamper data filter, effectively neutralizing threats and restoring the original data. The restoration's efficacy is assessed using the KS test, validating the data's recovery. The regenerated data is then utilized as input for deep-learning models in load forecasting, with the CNN-GRU-ATTENTION model emerging as the most effective performer compared to other alternatives.
AB - Residential load forecasting holds significant importance in the supervision of contemporary power grids, aiding utilities in efficiently allocating resources to meet the surging electricity demand. However, the proliferation of smart meters and IoT devices introduces new challenges to the power grid, particularly concerning the security and integrity of load forecasting systems. Intrusion attacks targeting load signature data can distort predictions, potentially leading to power supply disruptions and substantial economic repercussions. This research introduces a robust residential load forecasting tool based on a secure deep learning algorithm. The system demonstrates a swift ability to detect attacks and differentiate between manipulated data from cyber-attackers and authentic data. The initial step involves identifying intrusions within the dataset, utilizing a decision tree classifier that relies on features extracted through explainable artificial intelligence. This process achieves a commendable 96.6% accuracy in intrusion classification. Subsequently, the identified intrusions are counteracted via a meticulously designed clipper-clamper data filter, effectively neutralizing threats and restoring the original data. The restoration's efficacy is assessed using the KS test, validating the data's recovery. The regenerated data is then utilized as input for deep-learning models in load forecasting, with the CNN-GRU-ATTENTION model emerging as the most effective performer compared to other alternatives.
KW - Attention mechanism
KW - Cyber-security
KW - Deep learning
KW - False data injection
KW - Residential load forecasting
UR - http://www.scopus.com/inward/record.url?scp=85185790218&partnerID=8YFLogxK
U2 - 10.1109/ETFG55873.2023.10408544
DO - 10.1109/ETFG55873.2023.10408544
M3 - Conference contribution
AN - SCOPUS:85185790218
T3 - 2023 IEEE International Conference on Energy Technologies for Future Grids, ETFG 2023
BT - 2023 IEEE International Conference on Energy Technologies for Future Grids, ETFG 2023
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2023 IEEE International Conference on Energy Technologies for Future Grids, ETFG 2023
Y2 - 3 December 2023 through 6 December 2023
ER -