Assignment No. 1
To Study an Application of Genetic Algorithms
The process of genetic optimization can be divided into the following steps:
Generation of the initial population.
Evaluation of the fitness of each individual in the populatio
Ranking of individuals based on their fitness.
Selecting those individuals to produce the next generation based on their fitness.
Using genetic operations, such as crossover, inversion and mutation, to generate a
Continue the process by
going back to step 2 until the problem’s objectives are
Petroleum and Gas
A Typical Application of
have found the following applications
in petroleum and natural gas
eservoir characterization a
istribution of gas
etrophysics and petroleum geology
ell test analysis
ydraulic fracturing design
Genetic algorithms found their fir
optimum hydraulic fractures in a
gas storage field
s & Issues
Virtual intelligence techniques were utilized to design optimum hydraulic fractures. In
order to maintain and/or enhance deliverability of gas storage, an a
restimulation program was used that
calls for as many as twenty hydraulic fractur
and refractures per year.
Lack of engineering data for hydraulic fracture design and evaluation had, therefore,
made use of 2D or 3D hydraulic fracture simulators impractical.
prior designs of
dependent on engineers’ intu
ition about the formation
sometimes pure guesswork.
The data set used in this study was collected using well files that included the design
of the hydraulic fractures. The following parameters were extracted from the well files
for each hydraulic frac
ture treatment: the year the well was drilled, total number of
fractures performed on the well, number of years since the last fracture, fracture fluid,
amount of fluid, amount of sand used as proppant, sand concentration, acid volume,
nitrogen volume, ave
rage pumping rate, and the service company performing the job.
Use of Genetic Algorithms
The first step in this study was to develop a set of neural network models of the hydraulic
Using the data mentioned above, t
hese models were capa
predicting post fracture deliverability
Once the neural network model’s accuracy was
established, it was used as the fitness function for the genetic algorithm process to form the
hybrid intelligent system. The input data to the neural network
divided into three
Basic well information
Well production history
Hydraulic fracture design parameters
sand concentration, rate of injection, sand
mesh size, fluid type, etc
Of the above, o
nly the third (hydraulic fracture design paramete
rs) are among the
controllable parameters. In other words, these are the parameters that can be modified for
each well to achieve a better hydraulic fracture design. A two
stage process was developed
to produce the optimum hydraulic fracture design for eac
In the first stage, a
neural network for the first stage is designed and trained to
perform a rapid screening of the wells.
The second stage of the process is the genetic optimization routine. This stage is
performed on one well at a time.
econd stage process (the genetic optimization
routine) starts by generating 100 random solutions.
Each solution is defined as a
combination of hydraulic fracture design parameters. Genetic operations such as
crossover, inversion and mutations are performed
, and a new generation of solutions
is generated. This process is continued until a convergence criterion is reached.
objective of this stage is to search among all the possible combinations of design
parameters and identify the combination of the hydr
aulic fracture parameters for a
specific well that results in the highest incremental post fracture deliverability.
This process is repeated for all the wells. The wells with highest potential for post fracture
deliverability enhancement are selected as th
e candidate wells. The combination of the
design parameters identified for each well is also provided to the operator to be used as the
guideline for achieving the well’s potential.
Goldberg, D. E.,
Computer Aided Gas Pipeline Operation Using Ge
and Rule Learning
, Ph.D. dissertation, University of Michigan, Ann Arbor, Michigan.
Sen, M.K. et. al.,
Stochastic Reservoir Modeling Using Simulated Annealing and
, SPE 24754, SPE 67th Annual Technical Conference and
of the Society of Petroleum Engineers held in Washington, DC, October 4
Martinez, E.R. et. al.,
Application of Genetic Algorithm on the Distribution of Gas
, SPE 26993, SPE 69th Annual Technical Conference and Exhibition
New Orleans, LA, U.S.A., 25
28 September, 1994.
Fang, J.H., et. al.
Genetic Algorithm and Its Application to Petrophysics
, SPE 26208,
Yin, Hongjun, Zhai, Yunfang,
An Optimum Method of Early
Time Well Test Analysis
, SPE 50905, International Oil & Gas Conference and Exhibition in
China held in Beijing, China, 2
6 November, 1998
13. Mohaghegh, S., Platon, V., and Ameri, S.,
Candidate Selection for Stimulation of
Gas Storage Wells Using Available Data With Neural
Networks and Genetic
, SPE 51080, SPE Eastern Regional Meeting held in Pittsburgh, PA, 9