تقييم أداء خوارزمية شجرة القرار في التنبؤ بتوتر الانهيار للعوازل الملوثة
Keywords:
: Power transmission lines, Contaminated insulators, Flashover voltage, Decision Tree.Abstract
Electrical insulators play a vital role in protecting power transmission lines. However, their exposure to environmental pollutants such as industrial salts, desert sand, and atmospheric humidity leads to a reduction in surface resistance and an increased likelihood of electrical flashover. This study aims to evaluate the efficiency of the Decision Tree Regression algorithm in predicting the flashover voltage of contaminated insulators in power transmission lines. A dataset derived from laboratory experiments and mathematical models was used, where published experimental values were combined with values computed using a mathematical model. The entire dataset was then randomly split into training and testing subsets. The dataset included variables such as salt deposit density, insulator shape factor, creepage distance, and insulator height, while the target variable was the flashover voltage. The results obtained on the training data showed a high agreement between predicted and actual values, achieving a coefficient of determination (R²) of 0.995 and an RMSE of 0.335, indicating the model's ability to accurately capture the nonlinear relationships among the variables. When applied to the test data, the model maintained good performance with an R² of 0.966 and an RMSE of 1.017, confirming its capability to generalize beyond the training data. This study highlights that the Decision Tree algorithm provides an effective tool for predicting flashover voltage, thereby reducing reliance on costly laboratory experiments and enabling early maintenance decisions.