人工神经网络在油田注采量描述中的应用

Application of Artificial Neural Netw ork in the Description of the Relationship Between Fluid Withdrawal Rate and Injected Water Volum e in Water Injection Oilfields

考虑油田注水开发期采液量是注水参数和地层参数等的连续函数，应用人工神经网络技术，探讨油田注水开发期见水后的各个单井产状描述和产液量预测的新方法．采用某油田某区块内５口采油井配置１０口注水井，在１２个月的注采数据来训练所建立的注采系统人工神经网络模型．结果表明，人工神经网络可很好地用来定量描述油田注水期采出液量与注入液量之间的复杂函数关系，可用于油田水驱产状的描述．滚动预测的结果表明，预测的累计采油量随月份的增长变化和实测的符合良好．

Fluid w ithdraw alrate is a continuous function ofw aterinjection param eters, form ation param e- tersand otherparam etersin thew aterinjection developm entperiod. Forthisreason, The description ofindivid- ual-w ellattitude and the prediction ofindividual-w ellfluid w ithdraw alrate by artificialneuralnetw ork are dis- cussed. The injection-production system ofa tractin som eoilfield is com posed of5 oilw ellsand 10 w aterinjec- tion w ells. The artificialneuralnetw ork m odelestablished in this paper is exercised by the injection-productiondata in 12 m onths. The resultshow sthatthe com plicated functionalrelationship betw een fluid w ithdraw alrate and injected w ater volum e can be quantitatively described by artificialneuralnetw ork. The m ethod can be used for the description ofw aterflooding attitude. Theresultofrolling prediction show sthatthepredicted accum ula- tive oilproduction is identicalto m easured result.