Abstrakt: |
The operating conditions of modern electric power systems and their dynamic changes require a fast and accurate power security assessment system. Conventional power system security assessment depends on the operator's expertise and experience in conducting contingency simulations and data analysis involving many variables and parameters. It is a time-consuming process and has the potential for human error. Machine learning techniques have been proven capable of classifying and predicting various kinds of problems fast and accurately. In this paper, a random forest-based static safety assessment technique is proposed to reduce the gap in static safety assessments in modern electric power systems. The power flow simulation was carried out under normal topology conditions and N-1 contingency. The optimal power flow simulation is used to determine the real power generation settings on the generator to approach the real operating conditions of the electric power system. The static security assessment based on random forest model is trained and tested using different datasets such as IEEE 9 bus, 14 bus, and 39 bus to measure model accuracy with the smallest root means square error (RMSE) results of 0.0207, 0.0291, and 0.0731, respectively. The highest static security assessment (SSA) status classification accuracy reaches 98.48%, 98.42%, and 97.82% tested on 9 bus, 14 bus, and 39 bus test systems, respectively. Static security indirect assessment time processing requires a total of 0.4, 0.3, and 0.6 milliseconds for the 9 bus, 14 bus, and 39 bus test systems, respectively which are faster than the PMU time processing. From the model test simulation results, it is verified that the proposed SSA model can be used as an efficient electric power system security assessment with a good level of accuracy. [ABSTRACT FROM AUTHOR] |