Short-term household electricity load forecasting based on nonintrusive load monitoring using long-short term memory

Load forecasting is an essential part of the energy field. The excellent performance of the prediction algorithm can directly reduce the operation and maintenance cost of the equipment. Therefore, making load forecasting more accurate is a concern to energy experts. In the types of load forecasting, short-term load forecasting (STLF) is usually used to test the accuracy of new algorithms. In the previous papers, amounts authors provided various methods to improve load forecasting accuracy. In recent years, STLF based on the neural network has been proposed by researchers. But few papers mentioned combined energy disaggregation technique and neural networks to improve the forecasting model. Therefore, this paper proposed a new method that combined energy disaggregation and long-short term memory (LSTM) to enhance the accuracy of traditional STLF. In this article, the author will introduce a non-intrusive load monitoring tool kit (NILMTK) for getting each household appliance consumption from the total consumption and then predicting the future load value of each appliance based on its load data. Finally, sum the predicted values to get the result. Through experiments, the results of our proposed new method reduce the root mean squared error (RMSE) of traditional forecasting models by 7% and the mean absolute error (MAE) of it by 15%, respectively.