|
چکیده
|
This study introduces a method for fault detection, classification, and location (FDCL) in transmission lines (TLs) equipped with a Unified Interphase Power Controller (UIPC), utilizing the Random Forest (RF) algorithm. The distinctive characteristics of RF in classification and regression tasks are employed to develop a comprehensive protection algorithm. The algorithm uses the current and voltage of the three phases at a terminal on one side of the line. The method operates without threshold values or initial settings and incorporates a self-healing capability in the network expansion mode. The data required for model training comprises the outcomes of 3,860 scenarios executed in MATLAB/Simulink. The final model of each algorithm is implemented in Python using the gathered data. The proposed technique for fault detection and classification has been examined and assessed against three different methods: Artificial Neural Network (ANN), Support Vector Machine (SVM), and k-Nearest Neighbors (k-NN), using the following metrics: accuracy, precision, recall, and F1-score. The proposed method for estimating the distance of fault location is compared and assessed against six other methods: gradient boosting regression, linear regression, Bayesian regression, Ridge, Lasso, and elastic net, using the following metrics: Mean Squared Error (MSE), Mean Absolute Error (MAE), and R2. The outcomes of implementing the proposed procedure validate the efficacy of the algorithm.
|