In "Natural duality, Modality, and Coalgebras", in his thesis Meaning and Duality - From Categorical Logic to Quantum Physics, and elsewhere Yoshihiro Maruyama talks about $\mathrm{ISP}$, $\mathrm{ISP}$(M), how it is composed of $\mathrm{I}$, $\mathrm{S}$ and $\mathrm{P}$, but I can't figure out what these stand for.
It must be a well known concept as it seems to be assumed what it means.
$\mathbb{I}$, $\mathbb{S}$, and $\mathbb{P}$ are each operators on classes of structures:
$\mathbb{I}(\mathcal{K})$ is the class of structures isomorphic to a structure in $\mathcal{K}$.
$\mathbb{S}(\mathcal{K})$ is the class of substructures of structures of $\mathcal{K}$.
$\mathbb{P}(\mathcal{K})$ is the set of all products of structures in $\mathcal{K}$.
In particular, $\mathbb{ISP}(\mathcal{K})$ is the class of structures isomorphic to a substructure of a product of structures in $\mathcal{K}$. A bit of abuse of notation occurs here - an individual structure $X$ is conflated with its singleton $\{X\}$ and composition is given by juxtaposition, so that e.g. "$\mathbb{ISP}(L)$" is written for "$\mathbb{I(S(P}(\{L\})))$."
The paper's "$\mathbb{P}_M$" is then a modification of the operator $\mathbb{P}$, gotten by replacing "power" with "modal power."
A quick aside:
Another operator of this type which is quite important is $\mathbb{H}$, taking a class of structures to the class of homomorphic images of structures in the original class. Birkhoff's theorem says that the classes of structures described by equational theories are exactly those which are closed under $\mathbb{H}$, $\mathbb{S}$, and $\mathbb{P}$, and more generally that $\mathbb{HSP}(\mathcal{K})$ is always the class of structures satisfying all equations true of every element of $\mathcal{K}$.
But here, for whatever reason, $\mathbb{H}$ does not seem relevant.