Energy Consumption Prediction for Home Appliances with Recurrent Neural Networks

Efficient energy management requires a thorough understanding of individual appliance energy consumption in both residential and business environments. Because of the growing number of Internet of Things (loT) devices in smart homes, household appliances are a significant source of total energy consumption. Commonly used equipment such as heating systems, washing machines, and vacuum cleaners have a substantial impact on daily energy consumption. Significant energy savings can only be realized by careful management and forecasting of these appliances' energy consumption. This research employs Recurrent Neural Network (RNN) models for energy forecasting of housing appliances, with Long Short-Term Memory (LSTM) and a hybrid model known as Bi-LSTM-Attention Network. A Kaggle dataset with 28 features is used in this study, which is then reduced to 17 using a correlation plot. The processed data is used in both the training and testing phases of the RNN models. Many metrics are used to evaluate the performance of the models, including R-squared (R2), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). The Bi- LSTM-Attention Network achieves excellent performance, with training and testing R2 values of 0.9978 and 0.9835, respectively. In addition, lower MSE and RMSE values demonstrate that Bi-LSTM-Attention Network constantly outperforms LSTM. The findings show that the proposed Bi-LSTM-Attention Network is capable of accurately predicting the amount of energy household appliances will consume.