Grid-Constrained Data Cleansing Method for Enhanced Bus Load Forecasting
Zhenghui Li, Jiaming Liu, Yuzhang Lin, Fei Wang
Abstract
A number of measurement data cleansing methods have been proposed aiming at improving the accuracy of bus load forecasting (BLF). However, almost all these methods only exploit the temporal characteristics of a single measurement time series of bus load data and neglect the information that can be extracted from the power grid. As a matter of fact, the power grid model derived from physical circuit laws can provide an entirely different dimension of information facilitating the cleansing of bus load data. In view of this gap, a comprehensive data cleansing framework is proposed for the sake of enhancing BLF accuracy. Bus load datasets are cleansed based on their consistencies with both the temporal statistics and the physical power grid model. Simulation results on the IEEE 30-bus system verify that the proposed method remarkably improves BLF accuracy under various types of measurement data corruption conditions, including random noise, temporary gross errors, permanent biases, and cyberattacks.