I am running a time series simulation on an electricity power grid simulation package and I want to use this data to train an LSTM to predict the stability of the grid over a given time interval.
My time interval will be 10 seconds, with 1 second time steps. My data will have n features.
What I want to do is create thousands of simulations where I tweak some parameters with each simulation, outputting data which would have shape time_steps * n_features * num_simulations. This is the data I want to use to train my model.
My question related to the shape of data required by the LSTM. How does the third dimension of my data, ie, the many thousands of distinct simulated data fit into the batch size for the input to my model? I see this as a single sample of my data batch but might be wrong in my thinking... can I feed each sample into my model in a sequential fashion?