Residential Energy Consumer Occupancy Prediction Based on Support Vector Machine

The occupancy of residential energy consumers is an important subject to be studied account for the changes on load curve shape caused by paradigm shifts consumer-centric markets or significant demand variations due pandemics, such as COVID-19. For non-intrusive analysis, multiple types sensors can installed collect data based which consumer learned. However, overall system cost will increased a result. Therefore, this research proposes cheap and lightweight machine learning approach predict solely their electricity consumption data. proposed employs support vector (SVM), in different kernels are used compared, including positive semi-definite conditionally definite kernels. Efficiency depicted performance indexes calculated simulation results with realistic, publicly available dataset. Among SVM models kernels, those Gaussian (rbf) sigmoid have highest indexes, hence they may most suitable prediction.