Enhancing Smart City Efficiency by Mitigating Electricity Theft in Smart Grids Using Lightweight DNN and SMO

Smart cities require smart grid integration, industrial IoT, energy communities, renewable energy, smart healthcare systems, and the 6G network. Smart grids enable two-way power and data communication between various IoT devices and apps. However, increasing cyber layers and IoT connection threaten grid stability. Miscreants use these loopholes to illegally cut their power costs. The recommended approach includes model training, feature extraction, and preprocessing. Fixing outliers, standardizing units, and filling missing data are preprocessing steps. Feature extraction emphasizes majority class samples while considering minority class samples to account for the power consumption dataset's class imbalance. During model training, a lightweight deep neural network (DNN) addresses these difficulties. The Lightweight DNN model outperforms DNN and CNN for power theft detection. It improved smart grid power theft defenses to 94.28% accuracy. Smart grid electricity theft hinders smart city growth, but this study indicates that the Lightweight DNN model can handle it. The method ensures detection accuracy and grid stability, helping smart city infrastructures integrate current technology sustainably.