The structural configuration of bioretention plays an important role in the consumption and purification of nitrogen pollutants in rainfall runoff. Three layered bioretention tanks – 7#, 9#, and 10# – with artificial packing layers of fly ash mixing sand, blast furnace slag, and planting soil, respectively, were selected for intermittent and continuous operational tests. All load-reduction rates of nitrogen pollutants for intermittent running test exceeded 40% in three tanks, and tank 7# showed >70%. Moreover, the effluent pollutant concentration of 7# increased with time, whereas those of 9# and 10# fluctuated and then decreased slowly. The correlation model between TN removal and its influencing factors was established using the partial least regression method. Modeling analysis suggested that the filler type was the most important factor affecting TN removal. TN removal was positively correlated with packing factor and submerged zone height, while it was negatively correlated with antecedent dry time and influent loading. Soil pollutant original content and texture classification were detected before the continuous running test. The percentages of NO3-N and NH3-N accumulating in three facilities accounted for a total influent load of approximately 77% (7#), 61% (9#), and 43% (10#) when the exhaustion point was reached, demonstrating the relatively poor performance of planting soil.