Autoencoder using MLP for anomaly detection in multivariate timeseries

Question

I am developing an autoencoder using an MLP to detect anomalies in a multivariate time series. To simplify the problem, I started using only one series variable.