Study On Forcasting Model Of Land Subsidence And Its Application

land subsidence is one of the main geological hazard in a city, and its relatedenviomental geology problems seriously affect the estabilishment of baseinfrastructure, andrestrict the sustainable development of economy. Existing research shows thatexcessivemining of groundwater is the mainly reason. Taking the subsidence of Xian district as thestudy object, this paper aims to establish proper land subsidence prediction model, auxiliaryguides the mining of groundwater, to maintain the sustainable utilization of groundwaterresources and effectively control land subsidence.Based on collecting, sorting and analyzing the historical observation data in xi’an, thispaper takes the amountment of regional groundwater mining and groundwater level as inputvariables, sedimentation rate of the monitoring as model output variables, attempts to modeland predict the subsidence trend, in order to provide reference basis for reasonableexploitation of groundwater. The research achievements are as follows:1. The land subsidence single prediction modelis established. Using the typicalobservation sequence in xi ’an region,respectively establishesthree kinds of prediction model,includingthe BP neural network, regression analysis and grey theory. On the basis ofexperimental test, the applicability of theabove modelsisverified.2. The land subsidence combination prediction model is established. Experimentalresearch shows that combined prediction model is the effective measure to improve the landsubsidence prediction accuracy, and it can provide the reference frame for designinggroundwater exploitation plan.3. Set the exploitation reference scheme of groundwater. Withthe analysis of modelapplication, Under the ideal mining patterns, when hypothetic yield of groundwater meets192.62million m3/a,674.17million m3/a,268.72million m3/a, the correspondingestimated average land subsidence rate is13.2427mm,30.1565mm and73.4636mmrespectively, compared with the previous studies prediction results, predicted results tend tobe conservative.