LSTM for multiple time series

I am working with FE simulation data of 30 inputs and 30 outputs of various lengths. My input is the temperature cycle over time and I want to predict accumulation of plastic strains over time.

My data just an example

x-- 273,300,325,385,310.....

y-- 0.0176,0.01764,0.0177,0.0181,0.0183...

30 such curves with 400-500 points each. I have resampled the data to have each time stamps of 60secs.

I used taken differences of the output as had an increasing trend. My prediction data now in the order of y-- 1e-10,1e-7,1e-6,0,0,1e-11 really small and most of them are zeros.It is right skewed. I am trying to predict plastic strain difference at time 't' taking into consideration the previous inputs and outputs.

I have used NARX from MATLAB by using a window size of 7. Predict y(t) using x(t-7),x(t-6),x(t-5),x(t-4),x(t-3),x(t-2),x(t-1) y(t-7),y(t-6),y(t-5),y(t-4),y(t-3),y(t-2),y(t-1). I have also scaled the y values by 1e6.

I get good result using NARX and Levenberg-Marquardt and Bayesian regularisation solvers

Since I have multiple time series, how do i implement the same using lstm?