Distribution System Planning with Distributed Generation: Optimal versus Heuristic Approach

Abstract

Distribution system design and planning is facing a major change in paradigm because of deregulation of the power industry and with rapid penetration of distributed generation (DG) sources. Distribution system design and planning are key features for determining the best expansion strategies to provide reliable and economic services to the customer. In classical planning, the load growth is typically met by adding a new substation or upgrading the existing substation capacity along with their feeders. Today, rapid advances in DG technology and their numerous benefits have made them an attractive option to the distribution companies, power system planners and operators, energy policy makers and regulators, as well as developers. This thesis first presents a comprehensive planning framework for the distribution system from the distribution company perspective. It incorporates DG units as an option for local distribution companies (LDCs) and determines the sizing, placement and upgrade plans for feeders and substations. Thereafter, a new heuristic approach to multi-year distribution system planning is proposed which is based on a back-propagation algorithm starting from the terminal year and arriving at the first year. It is based on cost-benefit analysis, which incorporates various energy supply options for LDCs such as DG, substations and feeders and determines the size, placement and upgrade plan. The proposed heuristic approach combines a bi-level procedure in which Level-1 selects the optimal size and location of distribution system component upgrades and Level-2 determines the optimal period of commissioning for the selected upgrades in Level-1. The proposed heuristic is applied to a 32-bus radial distribution system. The first level of the distribution system planning framework is formulated as a mixed integer linear programming (MILP) problem while the second level is a linear programming (LP) model. The results demonstrate that the proposed approach can achieve better performance than a full optimization for the same distribution system.