Gerenciamento centralizado de microrredes de energia elétrica com operação em modo ilhado

Abstract

The use of electric power microgrids for the integration of distributed generation, energy storage systems and loads as a single unit has brought several benefits with respect to their management from the point of view of distribution networks and are essential parts of the new intelligent electric power systems. In a Microgrid (MG) it is necessary to have a management system to coordinate the operation of its components, with the MG running in grid-connected or islanded mode. In the islanded mode the MG management system must provide the set points for the Distributed Energy Resources (DER) in order to meet operational constraints, such as economic dispatch, use of renewable sources and load prioritization. The resolution of a mathematical model with the use of mixed integer linear programming seeks to meet these constraints in the optimal management of a MG. However, complex MG configurations, with the insertion of many DERs and loads, may impact the time required to solve this optimization problem. This work proposes a real-time Microgrid Central Controller (MGCC) with the use of mixed integer linear modeling to solve the energy and load management problem. To meet the response time requirements of its execution, the concept of Dynamic Intervals (DI) is applied to the mathematical model, reducing the size of data sets in the model for resolution and consequently the computational cost. Additionally, the model presents a load prioritization method based on the definition of classes and penalties. These penalties are related to the classes of loads and are applied when there is a need for shedding, seeking to supply energy to the most priority classes. Among the possible scenarios used in the tests, some had focus on cases in which the power generation by the DERs is lower than the demand of the existing loads, in which case the tests seeks to evaluate the MGCC response to the issues of load prioritization and also the use of DI. The results presented shows an expressive gain in the computational cost of the model resolution with the use of DI aligned with results that has very close values concerned with the load prioritization and shedding.