Say, I select 50 generation for Genetic Algorithm. I choose multi-objective (Power loss and voltage)
Step 1:
(a) To obtain minimum and maximum value after 50 generation, run the algorithm for single objective (Power loss).
(b) Obtain minimum and maximum value after 50 generation, Run the algorithm for single objective (Voltage).
Step 2: say, Power loss range [1100 kW to 3500 kW] and the voltage range [15 A to 100 A]. Both are not in same scale. So, normalization is required. And multiply the weighting factor (W1 and W2, W1+W2 = 1).
So, for multi-objective
F_obj = W1 * [(Power loss_each generation – 1100)/ (3500 – 1100)]
+ W2 * [(Voltage_each generation – 15)/ (100 -15)]
But how to choose suitable value of W1 and W2?
It is also advised that vary W1 from 0 to 1 while W1+W2 = 1. If I do so, I need a fixed value of Power loss and voltage whereas power loss and voltage values are changing at each generation.
The weighting factors W1 and W2 are your choice (between 0 and 1). It depends on which parameter's error (deviation from set point) is of higher concern to you. Let's say you want to keep your voltage level more nearer to its set point (than the power loss level to its set point). In this case, you should choose W2>>W1 (say W2= 0.8, W1=0.2) to severely penalize the objective function once there is higher deviation in voltage level. If the case is reverse, then set W2
You can also try using another optimization strategy to optimize this weighting parameters.
Ashik Ahmed Thanks for your reply. Here I have attached a picture. I wish to produce such graph. By varrying weighting factor.... For this reason, I need fixed value of two control variables. How to get that fixed two values ?
Md RABIUL Islam , The last query is unclear to me as you have not provided any relation between this plot, your objective function and the weight factors.
However, What I understand is that you are trying to find out the relationship between the Operation cost and Emission trend of different harmful substances for a certain power loss and voltage level. If this is the case, set a certain value of weight factors (W1 and W2 as mentioned in the earlier answer) and do any of the following:
i) If you are interested in the values of each generation, then after the end of each generation, for a certain set of weight factors, power loss plus voltage level will be calculated. Use those data to find out your desired graph. But this is not the optimal result.
ii) If you are interested in the optimal outcome, then let GA finish its search for the optimal result (after the stopping criteria has been satisfied), get the final values of your power loss and voltage level and calculate your desired plots.
You may experiment the above steps for any combination of weight factors.
- Badges
- Science topic
- Similar topics
- Engineering
- Materials Engineering
- Powders