Hybrid Machine Learning-based Intelligent Distance Protection and Control Schemes with Fault and Zonal Classification Capabilities for Grid-connected Wind Farms

This paper presents the analysis of several hybrid intelligent protection and control algorithms to improve the reliability of doubly-fed induction generator (DFIG)-based wind farms during faults, and other dynamic operating conditions. First, a decision tree (DT) classification algorithm is developed as a fault classifier for the purpose of distinguishing between different types of faults, as well as normal operation and grid disturbances. Next, a support vector machine (SVM) as a fault location estimator and zonal protection scheme is proposed to assist with the decision-making process of distance relay by detecting the location of any type of fault on the transmission line, and precise line zoning protection with a high reliability. Lastly, a combined direct PI control-based scheme is developed for both rotor and grid side converters of the DFIG based wind energy conversion system (WECS). This scheme avoids extra PI based current loop to achieve robust performance at the time of grid side voltage dip as well as normal operating condition. In this research, MATLAB and WEKA software are used for developing, training and testing the proposed machine learning algorithms and designing proposed control scheme, while ETAP and PSCAD software are used for design, modelling, fault analysis and data acquisition of the wind farm, as well as testing the operation of distance relays for various conditions. The analysis of the proposed intelligent protection and control schemes exhibits satisfactory results in improving the reliability and stability of grid-connected wind farms.