Let \(f:V \rightarrow \{0,1,2\}\) be a function, \(G=(V,E)\) be a graph with a vertex set V and a set of edges E and let the weight of the vertex \(u \in V\) be defined by f(u). A vertex u with property \(f(u)=0\) is considered to be defended with respect to the function f if it is adjacent to a vertex with positive weight. Further, the function f is called a weak Roman dominating function (WRDF) if for every vertex u with property \(f(u)=0\) there exists at least one adjacent vertex v with positive weight such that the function \(f':V \rightarrow \{0,1,2\}\) defined by \(f'(u)=1\), \(f'(v)=f(v)-1\) and \(f'(w)=f(w)\), \(w \in V \setminus \{u,v\}\) has no undefended vertices. In this paper, an optimization problem of finding the WRDF f such that \(\sum _{u \in V}{f(u)}\) is minimal, known as the weak Roman domination problem (WRDP), is considered. Therefore, a new integer linear programing (ILP) formulation is proposed and compared with the one known from the literature. Comparison between the new and the existing formulation is made through computational experiments on a grid, planar, net and randomly generated graphs known from the literature and up to 600 vertices. Tests were run using standard CPLEX and Gurobi optimization solvers. The obtained results demonstrate that the proposed new ILP formulation clearly outperforms the existing formulation in the sense of solutions’ quality and running times.