The Research And Application Of Data-drive Law In Predicting Reservoir Parameters

This paper is chiefly to set up a platform of multi-parameter nonlinear study and multi-parameter estimate to reservoir by means of the technology of multi-attribute transformation and neutral network combined with multi-attribute analysis of seismic parameters, reservoir inversion and reconstruction of reservoir geophysical characteristics on the foundation of large numbers researches and data-drive law in prediction ofreservoir so as to provide more accurate geology conclusion and exploration deployment scheme in practical application.Basing on the mentioned objective above, this paper has analyzed characteristics of many methods of reservoir prediction in the petroleum prospecting and their shortcomings in the practical geology problem of resolution. This paper has set up a platform of multi-parameter nonlinear study and multi-parameter estimate to reservoir by means of the technology of multi-attribute transformation and neutral network combined with multi-attribute analysis of seismic parameters, reservoir inversion and reconstruction of reservoir geophysical characteristics on the foundation of large numbers researches and data-drive law in prediction of reservoir so as to provide more accurate geology conclusion and exploration deployment scheme in practical application . Besides , concrete example analysis has been made on this technology aiming at different types of oil-bearing reservoir prediction. Summing up the characteristic of this technology, this paper point out its further direction in development."Utilization of multi-attribute transformation in predicting well logging parameter" has transcended many traditional methods of reservoir research in many aspects, and possessing many outstanding technology superiority, which are showed below:(1) It takes new technology thought - "date-driven law" as the guidance, and inheritsand synthesizes forefathers successful technology formed in many years. (2)'It directly calculates the well logging parameter by way of the multiple attribute transformation, rather than by way of the sound impedance, like the porosity, while the way before is to make further estimates of the sound impedance from the seismic inversion result so that the result suffers the influence of many factors.(3) The usefulness of the seismic attribute is drawn from the seismic data, rather than the traditional poststack seismic data after nonlinear a transformation.(4) It does not Like the traditional method to suppose a special designated model concerns between well logging and seismic data. It gains a statistics relation from a series of data training and analysis, which_ is linear relation gained from multivariate regression or nonlinear relation gained from neural net training.(5) Its most important characteristic is using the thought "alternation check" to evaluate the reliability of prediction, and can be used in the optimization of seismic attribute series.(6) This technology has gained big success guiding in the practical drilling application and develop in the research of reservoir prediction in domain of petroleum exploration gradually.