Static Security Assessment using Random Forest Based on Digsilent-Python Simulation

An electric power system is expected to continuously operate in safe conditions at all times and in all configurations. Modern electric power systems have dynamic characteristics with varying generation compositions, load types and network topologies. The dynamic changes of the electric power system require a fast, accurate, and efficient power system operation assessment. Conventional power system assessment mechanism is carried out by simulating contingency analysis with sampling testing on the network topology and generation and load profiles. Its processes also depend on the operator's expertise and experience in analyzing simulation data results, involving many variables and parameters. It is a time-consuming process and has a potential of human error. This gap can be reduced with random forest approach in power system assessment. Digsilent Power factory and python software are used for efficiently generating normal operation and N-1 contingency load flow simulation data with different loads: 50%-150% from the baseload as the datasets. The datasets are then used for training models based on a random forest algorithm for assessing the power system operation status. Finally, the proposed method will be tested on IEEE 14 and 39 test systems with various types of datasets training data to measure the level of accuracy and error proposed model. The simulations have shown that the proposed method has an MSE value of 0.0007 and 0.0005 for static security index prediction on IEEE 14 and IEEE 39, respectively and has an accuracy value of 100.0% and 99.97% for static security operation status classification on IEEE 14 and IEEE 39, respectively. With this level of accuracy, the proposed model may serve as a useful and efficient tool for power system assessment.