Saving/Reading variables based on ga optimization results

I have a question with regards to the ga solver in which I want to get the output of specific intermediate variables within the problem I am trying to solve. I only want to see the results corresponding to the optimal values.

In my objective function, I am trying to minimize the electricity cost of a building by optimizing the resistance of the insulation materials of the walls. The optimization works fine and I can get the x values and the cost of electricity by default. The script is long but lets say I have the zone indoor temperature (ZoneTemp) and the electricity end use (ElecUse) as intermediate variables that the final electricity cost depends on, how can I get these 2 variables as an output from the ga solver? I am sure there is a way to store these values to be read after the optimization is done. I tried to look at the "Custom Output Function for Genetic Algorithm" documentation but I am somehow lost as I am new to MATLAB.

I will just give a simple example of what I am trying to get

%This is my objective function in which I am using ga to minimize the toal cost of electricity

%To calculate the TotalCost there are too many variables and equations, these include ElecUse and ZoneTemp.

TotalCost = ElecUse * ElecRate

ElecUse = LightLoad + HVACLoad + PplLoad ... etc...

HVACLoad = ZoneTemp*...... etc...

I hope this is clear...

Any hints or examples would be appreciated. Thank you!

[Ameer Hamza]

Does the value of ZoneTemp and ElecUse also depend on the value of x (resistance)?

[Amde]

Yes they are all dependant. I care about the values of ZoneTemp and ElecUse based on the optimal values i.e. x!

[Ameer Hamza]

So if you get the value of x at each iteration, you can then use those values to find the values of ZoneTemp and ElecUse at each iteration. Is that correct?

[Amde]

Yes, ZoneTemp depends directly on resistance (x) and then ZoneTemp is used to calculate ElecUse which is eventually used to calculate the final cost. So basically, the ga solver has already done the calculations for all variables to get x, I just need to store the values of ZoneTemp and ElecUse for the best iteration.

Answer 1

[Ameer Hamza]

See my code here: https://www.mathworks.com/matlabcentral/answers/514594-plot-the-function-value-according-to-the-value-of-the-optimized-parameter-by-genetic-algorithm to see how to extract the best values of x for all iterations. The once extracted, you can use those to the values of x to calculate ZoneTemp and ElecUse according to their formula.

[Amde]

Thank you Ameer. I already saw your code before but let me try to explain agian. What you mentioned is actually what I used to do, I get the x values for the best iteration and then use it to rerun the code given the x as input. But What I am looking for now, is to to get the values of ElecUse and ZoneTemp from the optimization run. Based on my understanding, the ga solver has already done the calculations for all variables to get the best set of resistnces (x). I would assume that there is a way to ask MATLAB to store the values of ZoneTemp and ElecUse when the best x values are used (i.e. x that provides the minimum cost). Is there a way to do that?

[Ameer Hamza]

Isn't the value of ZoneTemp and ElecUse you are trying to get, same as calculated using the best value of 'x' you get after running the ga()?

For example, consider in 1st iteration, you have x_best=1. If you also look at the value of ZoneTemp for that iteration, isn't it equivalent to finding the value of ZoneTemp using its formula after the optimization is complete?

I am emphasizing this because MATLAB optimization function treats objective function as black-box; they are only aware of the input 'x' to the optimization function and its output value. They are not aware of what is happening inside the objective function itself. As far as ga() is concerned, it does not care about the logic and algorithm used to calculate the output value; it just cares about what the value is. Therefore, there is no elegant way to get the intermediate states out of the objective function. The only practical way is to complete the optimization process. Record the best solution at each iteration, and use that solution to calculate the value of intermediate states again.

For example, you define the objective function like this.

function y = objFun(x)
    % some code
    ZoneTemp = f1(x) % some function of x
    ElecUse = f2(x) % some function of x
    % more code
    y = % value of objective function
end

Now you run ga() and get the best value x_best at each iteration. You can just find the values of ZoneTemp and ElecUse on those iterations using.

ElecUse_best = f1(x_best)
ElecUse = f2(x_best)

[Amde]

Yes this is exactly What I was doing, I thought there is a way to get the intemediate values from the optimization run. I also came across this Custom Output documentation, not sure if it would be helpful. I guess I will continue doing what I am doing. Thank you agian, Ameer!

[Ameer Hamza]

The outputFcn is only used to analyze the states of the optimizer itself. The code in my other answer is also a custom output function. They are still not aware of the internal mechanism of the objective function. They are used to analyze states of the optimizer, e.g., population size, current population members, their fitness values, etc.