Excitation Current Forecasting for Reactive Power Compensation in Synchronous Motors: A Data Mining Approach

Abstract

Excitation current of a synchronous motor has a key role in reactive power compensation. For this purpose, the k-nearest neighbor (k-NN) classifier designed in this paper predicts the excitation current parameter using n-tupled inputs. Load current, power factor, power factor error and the change of excitation current parameters were utilized in n-tupled inputs. Moreover, Euclidean, Manhattan and Minkowski distance metrics were employed for measuring the closeness among the observations and the nearest neighbor number k was assigned as 1, 2, 3, 4 and 5, respectively. The forecasting results have shown that the k-NN classifier which uses power factor and the change of excitation current parameters achieved the best forecasting accuracy for k=1 in Minkowski distance metric. However, the k-NN classifier which uses load current, power factor and power factor error parameters gave the worst forecasting accuracy for k=5 in Minkowski distance metric.