Edge computing for improving energy management in smart homes

With the use of smart meters and data-driven solutions, home energy management has become more feasible and accurate in recent years. The main aspect of energy management is energy consumption forecasting of a household. However, it is very critical given the computational, data transmission, and privacy constraints by the end users. Thus, energy consumption forecasting demands an efficient and privacy-preserving scheme. In this context, this work presents federated learning-based scheme for forecasting the energy consumption of households while considering user's constraints. We utilized the well-established Federated Averaging (FedAvg) algorithm for achieving faster and accurate forecasting results. The proposed approach considers a real dataset of energy consumption for 51 households for a period of 3 years. The performance of the proposed approach showcases the significance of utilizing the federated learning framework in forecasting problems.