Forecasting Occupants’ Presence and Photovoltaic Power using an Enhanced Stacking Algorithm for Demand/Supply Energy Management

Abstract

This paper proposes an improved stacking algorithm based on an ANN (S-ANN) model to predict occupants’ presence probability and solar photovoltaic (PV) power generation over a 24-hour horizon for demand and supply energy management. The proposed S-ANN improves on traditional stacking by training each base model on various datasets instead of the same dataset. Python is employed as the simulation software for deploying the S-ANN model and conducting the analyses. The model’s performance is evaluated using the Normalized Root Mean Square Error (NRMSE) and Normalized Mean Absolute Error (NMAE) metrics. The results establish the effectiveness of the proposed model with an NRMSE of 2.91% and NMAE of 2.20% on a cloudy day (prediction of occupant presence in the kitchen reported as an example).