Progressive Short-Term Household Load Demand Forecast: A Dual-Model Approach
Silva, J. S.
;
Bartolomeu, P.
Progressive Short-Term Household Load Demand Forecast: A Dual-Model Approach, Proc IEEE World Forum on Internet of Things WF-IoT, Yokohama, Japan, Vol. , pp. 1 - 7, October, 2022.
Digital Object Identifier:
Abstract
The rapid evolution in the most recent years has raised power consumption in today’s world. Alternative energy sources are gaining increasing importance and, with that, so are the studies on how to better utilize renewable energy. This work was conducted within the scope of a project that aims at developing an intelligent energy management system for residential buildings. One of the requirements for said system is to accurately predict the overall household energy consumption. The objective of this study is to utilize machine learning techniques to develop a model that can learn the consumption habits and patterns of a household in order to forecast its future electrical load. Firstly, a study comparing Support Vector Regression (SVR) and LSTM (Long Short Term Memory) neural network is employed. It was found that SVR slightly outperforms LSTM when the training and prediction is performed utilizing only one house. Secondly, utilizing a dataset with multiple households and LSTM, a generic model is built. The generic model will be utilized in the early stages of the final system, while enough information about the specific house is being collected, in order to train a specific model with SVR.