A Novel ZIP-Based NILM Method Design Robust to Undervoltage and Overvoltage Conditions

Abstract Non-intrusive load monitoring is focused on recognizing active loads in the power system. In such studies, generally, current, voltage, power, etc., which can be obtained from meter data, are employed as features because they are the only utility side data of the location. However, the voltage level of the grid is subjected to power quality problems such as undervoltage and overvoltage, and the success of the NILM methods can fail in such situations. In this scope, the study specifically aims to identify loads instantly without relying on past information and cumulative data, so a novel approach robust to voltage fluctuations is proposed. In the study, the negative effects of voltage fluctuations on load recognition performance are examined first. Subsequently, a new approach combined with a static load model with constant impedance ( Z ), constant current ( I ), and constant power ( P ), which is known as the ZIP model, is proposed to minimize these misleading effects. For the proposed approach, an embedded system is utilized to determine the demand side accurately. Experimental results demonstrate that the ZIP-based method effectively minimizes the misleading effects of voltage fluctuations, achieving a 17.12% increase in accuracy compared to not using the ZIP model.