Research On PNN Construction Method And Application Technology For Oilfield Dynamic Information Modeling

With the development of science and technology, information processing methods such as system simulation, data modeling and so on are also expanding. The identification accuracy and modeling performance of complex nonlinear dynamic systems are also placed greater demands on the research of the system simulation model and the mechanism of signal transformation. This thesis is a typical problem of dynamic diagnosis, process simulation and forecast analysis for oilfield exploration and development, researching on information processing mechanism, data modeling and application technology based on process neuron network.This paper according to the summary and analysis of the current information processing and intelligent data modeling methods of oilfield, it can be summarized as the typical application of dynamic diagnosis, process simulation and forecast analysis, the process neural network method is used for modeling and processing. Through the analysis of information aggregation and time effect accumulation model of dynamic system in oilfield development, for different purposes, put forward the time aggregation operator, the excitation function and the construction and selection of network model appropriate to different applications, and its theoretical properties were analyzed. For network structure optimization and learning problem, proposed two hybrid optimization algorithms, which are the combination of LMS algorithm and PSO algorithm, and the combination of LM algorithm and QGA algorithm, the optimization of network structure and learning property of process neuron can be realized simultaneously. In the application technology research, two typical problems for the process simulation and dynamic pattern recognition of oilfield development system, the method and technology of the process simulation and dynamic diagnosis analysis of the process neural network are established.The subject is based on the research results of the method and application of the process neuron network construction for dynamic information modeling of Oilfield, and data processing for the two typical problems of Simulation of injection production process in oilfield development wells and Superiority seepage field discriminant, Made Good results. The research on the methods of spatial and temporal information processing has great theoretical and practical value.