Study On Automatic Fitting Algorithm For Unstable Well Testing Well

Well test as an important means to determine the reservoir characteristics and parameters of reservoir geological and wells,has become an important part of the oil and gas exploration and development.Fitting speed and accuracy of conventional well test analysis are low.To overcome the bias caused by human factors,improving the efficiency of the fitting,the computer automatic matching has been introduced into welltest interpretation.The results depend on the measured data,well test model and optimization algorithm.And that the algorithm is all important for automatic matching,hence,it is not only need to study the data de-noising method but also to research more on algorithm.In this paper,the wavelet transform data de-noising and transient well testing automatic fitting algorithm have been studied.(1)first of all,this article describes the theory of wavelet transform de-noising.The type of wavelet,wavelet denoising threshold and the layer number of wavelet decomposition for denoising effect have been analyzed.the measured pressure buildup test data were de-noising and the effect is good.(2)then,the article studies the LM algorithm,genetic algorithm and particle swarm optimization algorithm.Through the study found that:①genetic algorithm and particle swarm algorithm are groups search and global convergence methods,but their local search capability is very poor.②Particle swarm optimization algorithm is faster than the speed of convergence of genetic.③LM algorithm is highly dependent on the initial value,and only when the iterative initial value close to the optimal solution,fast local search capabilities is reflected.Therefore,this article combines particle swarm optimization algorithm and LM algorithm. Particle swarm algorithm is used to search Approximate optimal solution in the global solution space,at the same time by using LM algorithm in particle swarm optimization near the approximate optimal solution to detailed search the optimal solution.Such not only can overcome the LM algorithm for initial value dependencies,but also can make full use of particle swarm algorithm of global convergence and fast search speed.(3)Finally,At last,we programmed the algorithm with Matlab language to complete wavelet denoising and automatic fitting of well test analysis,and to validate denoising effect and automatic fitting algorithm with examples.