Research On Production Logging Datum Interpretation Method On Fuzzy Neural Networks

It is important for multiphase flow measurement to the exploitation of oil field,which can provide many dynamic data for geology analysis,make a diagnosis for abnormity of oil well, confirm the production state of oil well, supervise exploitation area by the numbers,and research exploitation status of each oil layer so as to take integrative way , examine effect of different ways,and make production increasing in the end.This thesis originates in the Heilongjiang Province natural sciences fund subsidization project item"The research on testing interpretation of three-phase flow measurement based on the RBF fuzzy neural network".In the paper,on the basis of experimenting on the multiphase flow loop with the Impedance Watercut Meter(IWM), experiment results are compared and analysed. Testing interpretation of three-phase flow measurement applied for IWM is studied successfully.At first,IWM and pressure drop densimeter are experimented on the multiphase flow loop. Experiment is carried out with two-phase flow and three-phase flow.The way of experiment is decided by the interpretation method.Experiment results of flowrate measurement and watercut measurement are analysed. Analysis is performed to flowrate measurement and watercut measurement of two-phase flow and three-phase flow separately. The relationship of flowrate measurement and watercut measurement with actual flowrate and watercut and gascut is analysed.The rule of apparatus response is compared to two-phase flow and three-phase flow. And, applies drift mode as interpretation model of three-phase flow to predic effect of model is checked up by data.Using powerful approaching ability of the artifical neural networks to non-linear mapping, prediction model of flowrate and watercut is established by BP neural networks. The complicated modeling process is avoided. The characteristic of BP neural networks is analysed. Neural networks structure is designed. Neural networks is trained and simulated subsequently. Flowrate and watercut is predicted after neural networks is fixed.The results predicted by neural networks are compared with those calculated by flowrate plate.Next,using powerful approaching ability of the artifical neural networks to non-linear mapping, prediction model of flowrate and watercut is established by the traditional BP neural networks. After analyzes in the BP algorithm the parameter nature, relates and the algorithm structure complexity mutually to design network architecture and carry on the training and the simulation.Finally, using the function equivalence of the RBF neural network and the fuzzy reasoning system to make the two organic synthesis, the mold which obtains the fuzzy RBF neural network interpretation model, after and various models forecast obtains the flowrate and watercut have carried on the comparison.