Meaning of several peaks of information curves in GRM I fitted my data to GRM, got the information curves and try to interprete the pictures.

I'm new to IRT and still don't get a lot of things. Here I can observe that the items are pretty good and measure what we actually meant. The best item is Q15_3, but what do the peaks mean on this green curve, why is there a lot of them?
 A: The items that you are analysing likely consist of more than 2 response categories, falling into the topic of polytomous IRT models (as opposed to dichotomous, which IRT is commonly presented for). 
The reason you are seeing multiple peaks is because each response category is providing a unique piece of information about the underlying ability based upon where the points of maximum inflection occur in the probability functions. For dichotmous models, which have a single inflection point (e.g., the 1-4PL models), the information function has only a single peak, while for non-monotonic probability response models and polytomous models multiple inflection points can occur, resulting in information functions which rise and fall with respect to $\theta$. The effect becomes even more pronounced when the so-called difficulty or intercept parameters in polytomous models are farther away from each other. 
For a visualisation of this effect, try manipulating the GUI that ships with the mirt package by passing mirt::itemplot(shiny = TRUE), changing the IRT class to 'graded', selecting the output plot as 'info', and moving the intercept/slope parameters manually with the supplied sliders. You should also see the relationship with the classical parameterization in the output as well, though that can be manipulated directly too. 
