Low dimension active power load data using autoencoder

Dimensionality Reduction (DR) is key machine learning technique used to convert data from higher dimensional space to lower dimensionality space in order to build a predictive machine learning models with less number of model parameters. Original active power load dataset is prepared by collecting the data from 33/11KV substation near Godishala village in Telangana state, India. It consists total 12 features like L(T-1), L(T-2), L(T-3), L(T-4), L(T-24), L(T-48), L(T-72), L(T-96), Temperature, Humidity, Season and Day. This 12 features data is reconstructed in to 10 features using autoencoder with a training loss of 0.0061 and validation loss of 0.0062.