This paper proposes a novel framework for fault detection, classification, and localization in power transmission lines (TLs) using optimized machine learning algorithms and meta-heuristic methods. The framework integrates fuzzy logic, adaptive fuzzy neural networks, and machine learning techniques to enhance the efficiency and accuracy of fault management in smart grids. Key contributions include: 1. **Fuzzy Thresholding for Fault Detection**: A fuzzy logic system is used to estimate the time of fault occurrence based on line voltage data, achieving detection within 1.2 clock cycles. 2. **Machine Learning for Classification**: Decision tree and random forest algorithms are optimized using the Wild Horse (WHO) meta-heuristic algorithm to classify fault types from frequency signals. The WHO algorithm improves the accuracy and robustness of these models. 3. **Adaptive Neural Fuzzy Inference System (ANFIS) for Localization**: ANFIS is used to locate the exact position of faults along the transmission line, with a placement error of less than 153.6 meters. The proposed methods were tested using a simulated power transmission system in MATLAB, demonstrating high accuracy and efficiency. The mean square error (MSE) for the decision tree algorithm was 2.34e-4 with an accuracy of 98.1%, and for the random forest algorithm, it was 3.54e-6 with an accuracy of 100%. The WHO-RF model outperformed other state-of-the-art methods in terms of classification accuracy, especially in noisy environments. The study highlights the effectiveness of combining fuzzy logic, machine learning, and meta-heuristic algorithms for advanced fault management in smart grids, offering a robust solution for improving grid reliability and resilience.This paper proposes a novel framework for fault detection, classification, and localization in power transmission lines (TLs) using optimized machine learning algorithms and meta-heuristic methods. The framework integrates fuzzy logic, adaptive fuzzy neural networks, and machine learning techniques to enhance the efficiency and accuracy of fault management in smart grids. Key contributions include: 1. **Fuzzy Thresholding for Fault Detection**: A fuzzy logic system is used to estimate the time of fault occurrence based on line voltage data, achieving detection within 1.2 clock cycles. 2. **Machine Learning for Classification**: Decision tree and random forest algorithms are optimized using the Wild Horse (WHO) meta-heuristic algorithm to classify fault types from frequency signals. The WHO algorithm improves the accuracy and robustness of these models. 3. **Adaptive Neural Fuzzy Inference System (ANFIS) for Localization**: ANFIS is used to locate the exact position of faults along the transmission line, with a placement error of less than 153.6 meters. The proposed methods were tested using a simulated power transmission system in MATLAB, demonstrating high accuracy and efficiency. The mean square error (MSE) for the decision tree algorithm was 2.34e-4 with an accuracy of 98.1%, and for the random forest algorithm, it was 3.54e-6 with an accuracy of 100%. The WHO-RF model outperformed other state-of-the-art methods in terms of classification accuracy, especially in noisy environments. The study highlights the effectiveness of combining fuzzy logic, machine learning, and meta-heuristic algorithms for advanced fault management in smart grids, offering a robust solution for improving grid reliability and resilience.