Sintonia e alocação de controladores de sistemas de potência considerando o critério N-1 utilizando um algoritmo NSGA2 paralelizado

Abstract

Electric power systems are frequently subjected to disturbances caused by, for example, a sudden increase in load or a short circuit in a transmission line. Such disturbances can generate oscillations in the system and consequently cause undesired operation. To reduce these oscillations, power system controllers can be used. Among power controllers used in industry, for damping oscillations, there is the PSS (Power System Stabilizer), device capable of increasing the stability margin of an EPS (Electric Power System) adding a modulated signal through the excitation control system. However, only their addition does not ensure the desired stability of an electrical system, requiring an appropriate tuning in order to make the system robust enough ahead changes in their operating points. This proposal presents an alternative for allocating and tuning PSS's using a multi-objective algorithm based on NSGA-II algorithm and considering the N-1 criterion, established by the ONS (National System Operator). To test the efficiency of the proposed algorithm, the benchmark system New test England / New York was used, which has 68 bars and 16 generators. To broadly consider the N-1 criterion, each line was removed, one at a time, making different operation points through system topology changes. For each operation point, the system characteristics were changed, a load increase and a load decrease, so, 3 new operation points were created for each original point. The proposed algorithm was aimed to find the locations of controllers and their tunings providing a Pareto frontier for the proposed objectives. For this task, the algorithm searched solutions with the number of controllers between 6 and 16, that is, a discretized search space with more than 20 trillion possible solutions. For the tuning, the algorithm searched appropriate parameters for all controllers used, tuning them for all operation points considered. In addition, a parallel algorithm structure was proposed to reduce the computational time. The results were analyzed using the hypervolume metric and validated through linear and nonlinear simulations. It was observed that the performance obtained was satisfactory for both purposes, indicating the effectiveness of the proposed technique. Thus, a significant gain in productivity can be achieved if the engineer responsible for the project only supervise the algorithm execution, instead of perform all calculations.