Dynamic energy prices for residential users based on Deep Learning prediction models of consumption and renewable generation

Abstract

New demand-side management models have emerged as a result of rising energy prices, the development of artificial intelligence, and the rise of prosumers. The purpose of this research is to use deep learning techniques to predict the energy production and demand of a prosumer network to determine dynamic prices for the local market. Gated Recurrent Units (GRU) and Long Short-Term Memory (LSTM) were two methods that were taken into consideration for forecasting consumer demand and wind and solar energy generation. Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) were used to compare the various approaches. The results demonstrated that GRU, with 0.0273, 0.0158, and 49.8 in RMSE, MAE, and MAPE respectively, is the best method for predicting energy generation and consumption in our datasets. Demand management system dynamic prices were calculated on an hourly basis using input from energy generation and demand forecasts. Finally, an optimization method was developed for establishing the energy planning.