Find Your Solution with AiSara

Updated: Mar 19, 2019

Optimize with no equation


#optimization #maximization #minimization #oilandgasexample #solvingcomplexproblem


Do you know that you can easily find an optimize solution with AiSara? The trick is to utilize AiSara in combination with the Excel Solver. But before that let us discuss why you want to optimize.


In oil and gas, you want to maximize production, reverse engineer what your reservoir looks like (via history matching) so you can make better prediction, design parameters that give you the maximum production, field development planning, and many more.


Application in other industry include sales, what to sell, when and where, minimum house pricing based on desired criteria and so much more, There is also the resource allocation problem, company and government policy, environmental assessment, plant optimization, engineering design, transportation, scheduling, etc. These are all some form of optimization problems.


Now, how do we do this with AiSara? For those who can’t be bothered to read this whole article and find all this very easy,, here is how.


Use the PREDICT function to get to a solution, any solution, by providing a first pass input set. Then you can use the Solver to find the combination of input data that yield either the minimum, maximum or a certain target value.


Now here is a walk-through step by step.


I am using the oil and gas History Matching 6 variables from the Sample Data-set as the objective there is to find a minimum target value which represents the minimum error.

Why minimum error? Because the minimum error represents the best solution possible in this case.

Using the Sample, download the history match data-set.



Click Sample and Retrieve this Dataset

Then you let AiSara learn the relationship between the input variables and the output variables using the LEARN function.


You almost always start with LEARN.
LEARN is usually one of the first things you do

In case you want to understand the source of the data, it is coming from hours of simulation to produce the results for each row. This time consuming process to locate the minimum is exacty why predicting the minimum results based on limited data will cut the time. So yes, you always need data to use AiSara, but you may need less than usual to reach your answer.


Ok, here we are going to take a quick look at the SPIDER. The SPIDER will provide results for combination of data, In this article, we will not dwell too much on SPIDER.


But what SPIDER will provide here quickly is the minimum and maximum values for each input. You can then use this to constrain the Solver later.

Typically you want to PREDICT within the range of your data.

Next, time to do the SPIDER. This is not a necessary step, but it will conveniently provide the min and max later for the Solver.


So there you go. The SPIDER shows the min and max of the input range

To find your solution, in the case the minimum, you first need to create a prediction. The input values are not too important here. Solver will then vary the input and watch the PREDICTion change. It will iterate to find the minimum.

You can now do the PREDICT, first variable is the LEARNed cell, the input data range (any reasonable numbers to start with), and which output column, in this case 1, there is only 1 output column here anyway..

See below how you constraints the Solver.