Support Vector Machine-Based Islanding and Grid Fault Detection in Active Distribution Networks

Many techniques used and still in usage for solving the problem of islanding detection are intrinsically passive, active, or hybrid of both. Each one of them has its own benefits and drawbacks. In this paper, we propose a method, which takes the advantage of a machine learning (ML)-based algorithm, namely, support vector machine (SVM), in order to produce the results more efficiently. The results of the simulations based on the model and experimentally measured parameters of a real-life practical photovoltaic (PV) plant give much better output than the traditional reported methods. During the tests and simulations, an additional problem, namely, grid fault, emerged, posing new challenges for the proposed method. Occurrences of islanding and grid fault are grouped together with the same kernel dimension and no custom hyperplane bordering. Discrimination between islanding and grid fault events is an essential dilemma, which is handled by the proposed SVM-based algorithm to achieve more precision in islanding detection and simultaneously detect the grid faults authentically. Nondetection zones (NDZs) and detection time (DT) are tested using two dimensions, namely, the generated active energy from PV plant (0%–110% of $P_{n}$ ) and distribution network voltage levels (±10% of $U_{n}$ ). Simulations based on the model and parameters of a real-life practical PV power plant are performed in MATLAB/Simulink environment, and several tests are executed for several scenarios. Finally, comparisons with previously reported techniques prove the effectiveness, authenticity, selectivity, accuracy, and precision of the proposed islanding and grid fault detection strategy with allowable impact on power quality according to UL1741 and its superiority over other methods.