Mixed-Integer Linear Programming Solvers for Local Grid Capacity Planning Problems

The algorithms and software tools for solving large-scale NP-hard MILP (mixed-integer linear programming) problems are analyzed, specifically focusing on the modeling of electric power systems arising in the planning of the development of generating equipment. A generalized algorithm for solving such problems is defined, based on the possibility of their decomposition: investment costs for the construction of generating units and the number of such units are reflected in the master MILP subproblem, while the system operator's costs for the operation of generating units, as well as the technological indicators of their load for a set of electricity consumption graphs, are reflected in a set of independent secondary MILP subproblems. It is noted that for Local Grids with large shares of electricity production from renewable energy sources, the sets of independent secondary MILP subproblems are quite substantial. Therefore, the speed of solving the secondary MILP subproblems determines the speed of solving the original MILP problems of electric power system modeling. Based on the results of experimental studies using the SCIP, HiGHS, Cbc, and GLPK solvers to solve a test MILP problem of optimizing the generating capacity structure of a Local Grid, it was established that the SCIP and HiGHS solvers are appropriate for solving the secondary MILP subproblems.