The following repository contains severall notebooks to work with ASHRAE's data base for the Great Energy Predictor III competition on Kaggle. We compare three different models with slightly different problem frames between the first and the other two models. They differ on the nature of the network, being the first two LSTM RNN models and the third one, a simple FCNN model.
The results are then compared and some observations regarding the conveniences of using LSTM on forecasting problems are then assessed. The data with which the models were trained and evaluated corresponds to the electrical load of a resiential building, as well as hourly meteorological data from the nearest station being given.
You can find the original data and data description here.