ISSN:
1573-2932
Keywords:
Artificial Neural Network
;
chlorine concentration
;
groundwater simulation
Source:
Springer Online Journal Archives 1860-2000
Topics:
Energy, Environment Protection, Nuclear Power Engineering
Notes:
Abstract From hydrocarbon reservoirs, beside of oil and natural gas, thebrine is also produced as a waste material, which may bedischarged at the surface or re-injected into the ground. Whenthe wastewater is injected into the ground, it may be mixed withfresh water source due to to several reasons. Forecastingthe pollutant concentrations by knowing the historical data atseveral locations on a field has a great importance to take thenecessary precautions before the undesired situations arehappened.The aim of this study is to describe Artificial Neural Network(ANN) approach that can be used to forecast the future pollutantconcentrations and hydraulic heads of a groundwater source. Inorder to check the validity of the approach, a hypotheticalfield data as a case study were produced by using groundwatersimulator (MOC). Hydraulic heads and chlorine concentrationswere obtained from groundwater simulations. ANN was trained byusing the historical data of last two years. The future chlorineconcentrations and hydraulic heads were estimated by applyingboth the long-term and the short-term ANN predictions. Anapproach to overcome the effects of using the data of a singlewell was proposed by favouring the use of data set for aneighbour well. The higher errors for the long-term ANNpredictions were obtained at the observation wells, which wereaway from an injection well. In order to minimise the differencebetween the results of long-term ANN approach and flowsimulation runs; the short-term prediction was applied. The useof short-term prediction for the wells away from an injectionwell was found to give highly acceptable results when thelong-term prediction fails. The average absolute error obtainedfrom the shortterm forecasting study was 3.5% when compared to18.5% for the long-term forecasting.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1023/A:1005165315197
Permalink
