Comparative Analysis of Time Series Forecasting of a Household Load Consumption with LSTM Neural Networks and XGBoost Model

Forecasting electrical power usage is seen as a crucial step in a process to increase energy efficiency, manage consumption, and save expenses. Demand response strategies, the setup of distributed generating systems, and their operations all depend on forecasting residential energy usage as essential elements of smart grid energy management. Furthermore, it is getting harder for service providers to satisfy everyone's needs because of the rising demand and the expanding economy. To maintain the stability of the system, forecasting the load consumption and observing the behavior of changing in load of a household is very important. This paper provides two methods: Long Short-Term Memory (LSTM) and Extreme Gradient Boosting (XGboost) machine learning models for forecasting the load consumption of a household for the period of one year per minute dataset, not only this but this study does a comparative analysis of two models based on accuracy, computational burden, and Performance Metrics. Simulation results show that the both models are reliable and efficient for forecasting the load consumption.