Smart grid precision: Evaluating machine learning models for forecasting of energy consumption from a smart grid

Abstract

The widespread adoption of smart home technologies has led to a significant increase in the generation of high-frequency energy consumption data from smart grids. Accurate forecasting of energy consumption in smart homes is crucial for optimizing resource utilization and promoting energy efficiency. This research work investigates the precision of energy consumption forecasting within a smart grid environment, employing machine learning algorithms such as convolutional neural network (CNN), recurrent neural network (RNN), long short-term memory (LSTM), temporal fusion transformer (TFT) and Prophet. The CNN model extracts spatial features, while RNN and LSTM capture temporal dependencies in time series data. Prophet, recognized for handling seasonality and holidays, is included for comparative analysis. Utilizing a dataset from Pecan Street, Texas, performance metrics like mean absolute error (MAE), root mean squared error (RMSE), and mean absolute percentage error (MAPE) assess each model’s accuracy. This work aids in improving energy management systems, contributing to sustainable and efficient energy use in residential environments.