Drilling fluid contamination determination for downhole fluid sampling tool

Abstract

A method and system for determining fluid contamination. The method may comprise monitoring a fluid sample, wherein the fluid sample comprises a reservoir fluid contaminated with a well fluid, and obtaining input parameters, wherein the input parameters comprise fluid properties obtained from measurement of the fluid sample and mud filtrate composition. The method may further comprise representing a mud composition as a Gaussian distribution, selecting a plurality of input data during a pumpout, determining calculated fluid properties of the reservoir fluid using an equation of state filtrate analysis, and further obtaining updated values of iterative parameters for use in a mole fraction distribution function. The system may comprise a downhole fluid sampling tool operable to obtain fluid samples of a reservoir fluid contaminated with a well fluid while the downhole fluid sampling tool is disposed in a wellbore, and a processor

What is claimed is: 
     
         1 . A spectral data reconstruction method, comprising:
 despiking an original dataset of optical spectral data;   applying a moving average to the original dataset to obtain a despiked dataset; performing a robust principal component analysis on the despiked dataset to identify a first set of outliers;   excluding the first set of outliers from a principal component analysis score and a principal component analysis loading;   performing robustfit regression on the principal component analysis score to obtain a second set of outliers;   excluding the second set of outliers from the principal component analysis score and the principal component analysis loading to obtain a spectral data;   reconstructing the spectral data;   inverting the principal component analysis score;   dividing the principal component analysis score for a predefined number of data points;   identifying a location of a clean reservoir fluid in a subterranean formation penetrated by a wellbore;   removing a contaminated reservoir fluid; and   performing a pumpout to remove the clean reservoir fluid from the wellbore.   
     
     
         2 . The method of  claim 1 , wherein the despiking of the original dataset is performed by a Hampel filter. 
     
     
         3 . The method of  claim 1 , wherein the robusfit regression is an inverse of accumulated volume of a formation fluid pumped out raised to a power density decay rate. 
     
     
         4 . The method of  claim 1 , wherein the reconstructing the spectral data is performed by an algorithm X=U*V T , wherein X is reconstructed spectral data, U is the principal component analysis score, and V is the principal component analysis loading. 
     
     
         5 . The method of  claim 1 , wherein inverting the principal component analysis is performed when accumulated formation volume is at infinity. 
     
     
         6 . The method of  claim 1 , further comprising picking a corresponding position in the reconstructed data. 
     
     
         7 . The method of  claim 1 , wherein the contaminated reservoir fluid is a reservoir fluid contaminated by drilling fluid. 
     
     
         8 . The method of  claim 1  using a neural network to predict fluid contamination, wherein fluid contamination is the amount of mud filtrate in the contaminated reservoir fluid. 
     
     
         9 . The method of  claim 1  using an asymptotic value of the principal component analysis score for estimating a clean formation fluid composition. 
     
     
         10 . The method of  claim 9 , further automating a multipoint Cubic Equation of State (EOS) contamination position using the principal component analysis score.