Fire hawks optimized radial basis function neural network based feature extraction and ON/OFF detection of household appliances

Abstract

Accurate ON/OFF detection of household appliances is essential for smart energy monitoring, lowering costs, and improving energy efficiency in smart homes. However, existing ON/OFF detection methods have several challenges, such as high computational complexity, overfitting and overlapping power usage patterns, which lead to false classifications and reduced performance. This study proposes a novel hybrid method combining a Fire Hawks optimized radial basis function neural network (FH_RBFNN) in order to extract and detect ON/OFF status at the source end of a residential building. The Fire Hawks Optimization Algorithm (FHO) is employed to fine-tune Radial Basis Function Neural Network (RBFNN) layer parameters, which ensures effective feature extraction by reducing redundancy. Subsequently, the Xtreme Gradient Boosting (XGBoost) technique is employed to classify the extracted features in order to identify the ON/OFF stage of house appliances. The proposed FH_RBFNN+ XGBoost model achieves high detection performance in terms of accuracy of 0.995, Precision of 0.99324, Recall of 0.99606, F1-Score of 0.99465, and Specificity of 0.99067, respectively.