الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 169 |  |  |
| | | from | 169 | | |
Abstract
South Dabaa (DAPETCO) field is a mature gas lift oilfield consisting of multiple reservoirs,
those reservoirs share the same surface processing facilities. In such context, gas lift optimization
is crucial to ensure maximum oil production within facility constraints. Nodal analysis, and Gas
lift Optimization Allocation model (GOAL) are among the tools applied to meet this objective.
The scope of the present thesis is to assess the gas lift and production performance of Dapetco
field, to find out whether it works at its optimum behavior or if there is any margin for
improvement.
In this thesis, each well model was built using Pipesim software. Nodal analysis for each well
were performed and compared against actual well tests. The results of nodal analysis were reliable
to conduct further investigation on wells parameters.
The thesis presented artificial neural network (ANN) in two different approaches: global ANN
model and well by well ANN to predict the oil rate and gas lift rate of Dapetco field wells and
compared against actual data while applying different scenarios including but not limited to;
forward and backward propagation, radial base neural network, one hidden layer with different
numbers of neurons and two hidden layers with different numbers of neurons as well. In addition,
applied the generated data from ANN model to generate an equation to predict the production rate
with optimum gas injection rate.
The applied methods were compared and statistically analyzed. Analysis showed that Global
ANN model and well by Well ANN models produce more accurate results than Pipesim models.
After that applied a comprehensive study using Garson law to detect the importance and effect of
input production system parameters on outputs.
A comparison of ANN models with previous studies were conducted and showed that this study
presents more accurate results because a larger accurate data set of more input parameters were
incorporated. After applying test data from the different fields found the fastest training and testing
error were achieved with BP and FP neural network which showed in statistical analysis compared
with Pipesim models with accuracy up to 91%.
The experimental results indicate that a strong matching between model predictions and observed
values, since MSE is 0.0012. When performance results are compared, it was concluded that
RBFNN-based model is a more reliable predictor, with MSE value of 0.003 and ARPE of 8.2.
Therefore, the smallest MSE value indicates a creditable method for accuracy, while RBF finding
illustrates best proposed model to analyze the output.
At last, the thesis went through some ideas and solutions to prove that depending on such
artificial intelligence in calculating the production data or downhole data could be achieved and
reliable to avoid the risk of well intervention operations by downhole gauges and to get more data
for the wells which suffer from remote area and could not get the production data for it easily also
applied simple study on the available wells and data to overcome the problem of limited
compression capacity. Technical, logistical, and economic analysis were performed for each
method.
Well intervention operations effect was studied in the thesis, such as: water shut-off, well
stimulation and gas lift valves change. These remedial actions helped improve wells productivity.
The thesis investigated to the surface gas lift lines and found out it is recommended to upsize
one main gas lift line to allow the gas to reach the far wells which was suffering from low injection
pressure.