New And Very Powerful Microseismic Inversion Algorithm

Greetings!

Please take a look at the recent progress of my microseismic research, which is a significant improvement of state of the art.

All the parameters used are very objective. The only thing that has changed is the use of new inversion algorithm instead of algorithm in the literature. For the following inversion results, the unit of time is 0.0109 s and the unit of distance is 51.5958 m. There are four layers in the formation. The first 20 parameters are VTI parameters ordered in C11, C13, C33, C55, and C66, the 21-23 parameters are the formation depth, and the last 24 parameters are the x, z coordinates, and origin times. Noisy traveltimes are obtained by ray tracing with 3% Gaussian noise to simulate real traveltimes. As following figures show, in close to 90% cases, inverting noise-free traveltimes obtain exact true parameters. In close to 90% cases, inverting noisy traveltimes obtain true parameters plus noise-level perturbations. I also used the new algorithm to invert other input data, and the probability of successfully inverting the real parameters input into the model is close to 90%. The other 10% is not so good. I am still implementing another new algorithm with a lot of progress, the near future success rate maybe close to 99%. The new algorithm could replace seismic AVO and reflection moveout inversion with similar acquisition configuration and geological structure.

Compare with recent peer-reviewed publication with similar configuration (2019, Journal of Petroleum Science and Engineering), the new algorithm show great improvements. This publication is available upon request. My algorithm requires acquisition geometry reasonably well-imposed.

Welcome to test the new algorithm! The testing procedure is quite simple: you do forward modeling with 10 set of random input parameters and noise without telling me, and I will do the inversion. The artificial noise should not be too big. Then we compare. I believe this process usually gives you a lot of confidence in my new algorithm. I may provide my unique anisotropic ray tracing code.

Rui Zhou, PhD

E-mail: zhourui2058@gmail.com, zhourui2058@126.com

Phone: 86-13665556361 (same as WeChat).

Linkedin: https://www.linkedin.com/in/zhourui2058/

Skype: zhourui2058

The Matlab code for input parameters and acquisition geometry:

ms=[80 151;90 155;95 161;100 165;150 161;160 165;165 171;170 175];

ms(:,3)=0;

C_all_L=[20.01105;7.18102335942054;16.4025;5.588496;7.15327488;22.821084;7.94939882314651;18.4041;6.4009;7.937116;18.4781646;5.60906442483534;13.198689;5.1984;7.277760;23.79951964;8.45932219941062;19.193161;7.198489;8.35024724];

C_all_L=C_all_L*10^6;

number_rec_1=20;

rec=zeros(20,2);

for i=1:number_rec_1; rec(i,2)=40+i*10; end

for ir=number_rec_1+1:30

rec(ir,1)=110;

rec(ir,2)=10*ir-130;

end

top=[0 95 135 195 500000];