Improved analytical expression for calculating total harmonic distortion of PC clusters

Being widely used, personal computers (PCs) are, in relation to harmonics emission, very important within the category of non-linear low power loads (IL ≤ 16 A). A simultaneous operation of a group of PCs causes the distortion of the current waveform and negatively affects the supply of voltage. Therefore, the exact determination of current harmonics quantified with total harmonic distortion, THDI, is of great importance. The existing models and expressions for the THDI are not adequate in cases of large PC clusters. The aim of this paper is to present an improved analytical expression for the determination of the THDI dependence on the PC cluster size, suitable for practical engineering application. Based on the data available in references, the results of power quality measurements carried out in a computer centre on a group of 163 PCs, and the results of computer simulations, the THDI function on PC cluster size proposed in the paper is simple and easy to implement. The expression is suitable for engineering application for PC clusters up to 200 PCs per phase and for the most probable network connection of the cluster. It allows quick assessment of the current distortion, without the use of a simulation software. To improve the applicability, the influence of the stiffness of the grid is considered separately and then included in the enlarged expression. At the end, the effects of AC/DC power conversion, grid stiffness and cluster size on model application is further discussed. A direction for future research is also given.

Research highlights
► The existing models and expressions for the THDI calculations are not adequate in cases of large PC clusters.
► We proposed an improved analytical expression for the determination of the THDI dependence on the PC cluster size.
► The proposed model is suitable for practical engineering application for PC clusters up to 200 PCs per phase.
► The effects of AC/DC power conversion, grid stiffness and cluster size on model application are discussed, also.