Estimation of voltage instability inception time by employing k‐nearest neighbour learning algorithm

Agent‐based situational awareness system for severity in closeness of voltage instability occurrence

“…To achieve a rich data base for training k‐NN classifier, we created 300 scenarios leading to long‐term voltage instability (among many others which were stable) by time‐domain simulations. The details of these cases have been given in [28]. However, the following disturbances were applied: Small/load disturbances, i.e.…”

“…Although extensive tests on numerous datasets, with different learning techniques, have shown that 10 is about the appropriate number of folds in cross‐validation technique, to get the best estimate of the error; arguments are by no means conclusive, and debate continues to find which is the best [28]. In this regard, the evaluation results for the five different fold selection (10, 20, 50, 100, and 300) have been shown in Table 5.…”

“…Another attracted family of measurement‐based methods that learn off‐line and produce accurate results on‐line are machine learning (ML)‐based techniques, i.e. artificial neural networks [20–22], support vector machine [22–25], decision tree (DT) [22, 26], k ‐nearest neighbour ( k ‐NN) [27, 28]; the main concern for these methods is difficulties in interpreting proposed model.…”

“…Really, the former contribution is the most important improvement of this work rather than our previous paper [28] and creating SA platform for voltage instability analysis with reduction of feature space to 2D from n D space (‘ n ’ is the number of generators) is novel in the literature.…”