The Application Of Constrained Sparse Spike Inversion In The Coal Bed Identification

Impedance is an important formation parameter, and it is a direct reflection of the reservoir,it is also a key parameter of the distribution of space and reservoir conditions forecast,so impedance inversion is developing rapidly recent years.Today impedance inversion method has become a large system,which has develop dozens of inversion method, so how to well choose these inversion method has become an urgent problem need to solve for the geophysical workers.Based on the assessment of several commonly used impedance inversion methods,I summarize the following conclusions: trace integration inversion and recursive inversion have lower requirement on the exploration degreee, which are used in early exploration,but the accuracy of inversion results is low,so these results are generally regarded as other inversion methods'initial impedance; band-constrained inversion, constrained sparse spike inversion and generalized linear inversion don't have very high requirement on the exploration degree,so we can use them at various stages of exploration and development,but their result accuracy is in the middle in many inversion methods,we can use these inversion results only when the requirement of the result accuracy is not high; stochastic simulation methods and stochastic inversion inversion method are very high for the requirement on exploration degreee,which are only used in the late stage of exploration and development, and the inversion result accuracy is higher ,so they can be used as the final value of impedance inversion.Then I choose the recursive inversion to inverse.The reason I chosed this method is that it is a method based on seismic data,and its inversion result match well with the seismic data ,so the basic shape of ground interface can be completly inversed.Moreover the certainty of the inversion result is high,and it does not appear the phenomenon of multiple solutions,so it can well reflect the formation physical property.And I did error analysis on the inversion result,obtaining the following conclusions:1,For the seismic records without noise,the recursive inversion result is accurate,and it can accurately describe the formation interface, the error is very small; For the seismic records with the standard normal distribution noise and uniform distribution noise, the recursive inversion's requirement is high,when the noise intensity is respectively 10?8 and 10?7 ,the inversion result is accurate, while the describtion for the formation interface is also very good. And the seismic records with uniform distribution noise show the phenomenon of error accumulation, but the seismic records with standard normal distribution noise doesn't obviously demonstrate this phenomenon.Therefore,for the seismic records with standard normal distribution noise,we should both take error accumulation and inversion accuracy into account to determine the degree of noise suppression;and the seismic records with uniform distribution noise is more suitable for shallow layer as the error accumulation, the deep layer inversion result accuracy will decrease as the depth increase, and we should enhance the suppression of uniform distribution noise to improve the inversion accuracy.For the seismic records with the combination of standard normal distribution noise and uniform distribution noise, its error variation followed similar laws with the seismic records with the standard normal distribution noise. This phenomenon indicates that normally distributed noise plays a major role, so in the data processing, its treatment base is similar to the treatment base of the standard normal distribution noise.2,We use the seismic record with Yu wavelet to synthesis and the the seismic record with ricker wavelet to synthesis to do recursive inversion and compare their inversion results. In general they are very accurate,and they well describe the formation interface, but they both show the phenomenon of error accumulation,and in the same peak frequency, it can be found that in shallow layer the former's error is greater than the latter's error,but in deep layer it is contrary.Thus the seismic record with ricker wavelet to synthesis is more suitable for shallow layer inversion,and the seismic record with ricker wavelet to synthesis is more suitable for for deep layer inversion, This is very important for how we choose the seismic wave to inverse.In summary, for a variety of geological model,models'inversion results all have error accumulation besides the model with standard normal noise, The reason is that impedance is obtained from the downward through the recursion in the recursive inversion leading to the accumulation of errors from top to bottom, which may make shallow layer's error much larger than the error of deep layer. This is very negative for deep layer seismic research. Finally, I studied in the application of constrained sparse spike inversion in the coal bed identification.I use this method to inverse the actual seimic data,and analysis the inversion results obtaining the following conclusions:1,The method of constrained sparse spike inversion have a unique ability for identifying the coal bed,in the inversion result the coal impance is significantly lower than the surrounding formation's impedance,which make we can clearly find the coal bed. In general seimic profile, we only can see a coal bed, but in the wave impedance inversion profile we can clearly see the two coal beds.This provides a new clue and exploration direction to expand the scope of searching for coal.2,The inversion result more clearly reflect the contact relationship of the formation,which is showed by the color difference for the different formation impedance.And the inversion result matches well with the seismic data,but for the minor changes of the reservoir it did not reflect very well.3,A key link of the constrained sparse spike inversion is the choice of the parameterλ,in theory, the value is inversely proportional to the sparse level of reflection coefficient, that is, the smaller the value, the higher the reflection coefficient's sparse level, the worse the detail description; and the greater the value, the lower the reflection coefficient's sparse level, the well the detail description,but some seismic noise may be introduced,which can affect the accuracy of inversion result.Therefore,it requires tha we make a comprehensiv analysis of the signal to noise ratio of seismic records,seismic records and well logs'correlation and seismic records and well logs'standard deviation,, we select the equilibrium pointλof the three curves by trial and error.This is the only way to get an inversion result with higher resolution and signal to noise ratio.