Evaluation and modelling of particle collection efficiency of electrostatic precipitation

Abstract

Many South-African power plants use wire-plate electrostatic precipitators (ESPs); however, these ESPs, in practice, have yet to meet the expected high particle collection efficiencies and present relatively low collection rates under the current operating conditions. The characteristically high resistivity of South-African coal fly ashes contributes to the low collection efficiencies of the ESPs. Therefore, the operational configuration of these ESPs needs to be refined to improve the unit performance. Experimental research was conducted on a laboratory-scale version of the industrial ESPs. An in-depth examination of the structural components and most significant contributing factors of ESP performance was carried out. The investigation was undertaken on a fundamental level to gain a detailed understanding of the basic working principles of the ESP unit and its response to different operating variables. This information was translated into several models (statistical, computational fluid dynamic, and empirical) that could be used to improve ESP collection efficiencies. The ESP experimental data were first fitted to a quadratic response surface model (RSM). The RSM model was used to predict how – and to what extent – different variables affect the ESP collection efficiency. An analysis of variance (ANOVA) on the RSM model revealed that electrode geometry, collection plate spacing, and ash resistivity were the most significant factors affecting ESP efficiency. The number of discharge electrodes showed less remarkable yet notable effects on the collection efficiency, while the electrode spacing effects were almost negligible for non-shielding conditions. Furthermore, a computational fluid dynamic (CFD) model was developed to describe the particle collection efficiency of an ESP with various electrode configurations and ash properties. The coupled air flow profiles, electrostatic field, and particle trajectories were modelled using the relevant physics relations in STAR-CCM+. Three-dimensional geometries of the ESP section were created in NX-12, and served as the simulation domains for the CFD model in STAR-CCM+. The CFD model exhibited a good correlation with the experimental data. Two South-African coal fly ashes with -75 μm and -150+75 μm particle size fractions were used as feeds to the ESP for the experiments. Particle size distribution, particle density, X-ray fluorescence, and resistivity analyses were performed to study the effects of the related ash properties on ESP efficiency. Both ashes delivered low resistivity readings (1010 – 1011 Ω.cm) at atmospheric conditions. The higher resistivity ash showed shorter voltage ranges but better particle collection rates than the lower resistivity ash. The ESP performance was also evaluated for various ash feed loadings and deteriorated with decreased particle concentrations in the air. This was correlated to larger interdependencies between the fluid dynamic, particle dynamic, and electrostatic fields, making it more difficult to collect ash particles from low particle concentrations in the ESP, as opposed to bulk particle loads.

The CFD model was used to interpret the results from the ESP experiments. G-spike, sawtooth, and threaded rod geometries were considered for discharge electrodes in the ESP. The G-spike and sawtooth electrode designs produced high current densities, yielding up to 99% collection efficiencies. The threaded rod electrodes presented smaller current densities, with lower efficiencies (< 93%) than the other electrode geometries. The current and field distributions were unique for each electrode geometry and showed strong correlations with the particle flow patterns in the ESP. Shielding effects were observed for several ESP configurations, especially those involving G-spike electrodes. The shielding degree increased for smaller electrode spacings and larger plate spacings, notably reducing current densities and particle collection rates. The particle collection efficiencies were also computed using the empirical Deutsch-Anderson and Matts-Öhnfeldt equations. The Deutsch-Anderson model assumes ideal ESP conditions, yielding modelled efficiencies 10 – 20% higher than the measured values. The Matts-Öhnfeldt model, on the other hand, produced accurate efficiency predictions with k-values ranging from 0.2 to 0.9 for various ESP operating conditions.