This paper proposes a novel solution technique using a matheuristic approach to address the expansion planning of modern distribution systems. The planning problem can be accurately modeled as a non-convex mixed-integer non-linear programming (MINLP) model, but this formulation is extremely difficult to solve. In this context, traditional solution techniques often rely on convex relaxations or approximations. However, these approaches often result in infeasible solutions for the original non-convex MINLP model (especially in the context of modern distribution systems), and can also be computationally expensive for large-scale problems. The proposed technique overcomes these challenges with an iterative local search process that improves solution feasibility (regarding the original MINLP model) at each step until a high-quality local optimal solution is achieved. To do this, the problem is reformulated as a mixed-integer quadratic programming (MIQP) model that penalize the feasibility errors. The planning actions considered include the installation of distributed generation (DG) units, electrical energy storage (EES) systems, fixed and switchable capacitor banks (CBs) and line reconductoring. System variability in demand and energy resources is captured through representative days, preserving EES temporal transition. Furthermore, the model guarantees CO2 reduction in order to be on track to limit global warming. The effective performance of the proposed approach in terms of solution quality and computational time has been illustrated with two case studies based on a 33-node test system. Scalability is further evaluated using a real 135-node distribution system.