The disjoint union of $n$ copies of $K_5$ (or $K_{3,3}$) is a minimal forbidden minor for the graphs of genus $n-1$; the same is true for a graph in which some of these copies share a single vertex, so that the blocks of the graph are $K_5$ or $K_{3,3}$. This follows from results in J. Battle, F. Harary, Y. Kodama and J.W.T. Youngs, "Additivity of the genus of a graph", Bull. Amer. Math. Soc. 68 (1962) 565–568, and is already enough to show that there are at least exponentially many forbidden minors.
Bojan Mohar, "An obstruction to embedding graphs in surfaces", Discrete Math. 78 (1989) 135–142, lists the graph formed from $K_8$ by removing a 4-cycle as having genus 2. Since $K_7$ is toroidal, this means that either $K_8\setminus C_4$ or one of its spanning subgraphs is an obstruction to torus embedding, and that graphs that have $n$ copies of this graph as their blocks have genus $2n$.
Mohar also shows that the graph formed from a $(2k+2)$-cycle by connecting vertex 0 to all the even vertices and vertex 1 to all the odd vertices has "relative genus" at least $\lceil k/2\rceil$. The graph is planar, but I think relative genus means that the cycle has to be a face; or you could add another vertex to the graph, connected to all of the cycle vertices, to effectively force it to be a face. Maybe this is closer to the sort of thing you want. But I don't think he shows that these graphs are minimal forbidden minors.