Hi, Elton. Could you help us out with a reproducible example, called a reprex?
PLSR is a sort of unholy alliance between principal component analysis and linear regression. Instead of minimizing the variance on the cartesian plane, some varieties minimize it on the orthagonal plane.
You generally see it used with high-dimensional data, with more predictors than observations.
The principal R package seems to be pls and I think I'll be able to help you with cross validation and diagnostics. But to do that, of course, require some representative data and a fitted model.