And by effectiveness I mean performance;
When you have a simple control loop that minimize the error is simple, but what happens when you have a hierarchical system with multiple simple control loops.
How can you say if the whole system performed well or not?
Dear Andreea lon,
I allowed myself to reformulate the theme "Driving performance quantified for a hierarchical system with multiple control loops" and from my personal experience in the control of complex systems. here are some files attached on this rather important issue from my point of view.
Best regards
The criteria of controller performance are IAE( intergral absolute error), ISE ( intergral square error), and ITAE( intergral time-absolute error).
Thank you for your answer. I know these criterion for controller performance. My question is about the situation when you have a system that contains multiples controllers and you have to evaluate somehow if the entire system has a good performance or not.
I look at this problem specifically for demand response aggregators in smart grids. The way I approach the issue is that I establish an overall goal for the system (in this case the aggregator) and quantify that goal. In this case I'm not quantifying the performance of each control loop within the aggregator architecture, but just evaluating how well the system achieves it's final objective. I discuss it in one of my publications.
Conference Paper Performance Assessment of Aggregation Control Services for D...
Thank you Daniel for your answer. Can you give me a link or pdf with this publication where you discussed this problem?
Hi Andreea,
Here you go:
https://www.researchgate.net/publication/272743815_Performance_Assessment_of_Aggregation_Control_Services_for_Demand_Response
Hope it gives you a bit of inspiration for your specific topic.
Conference Paper Performance Assessment of Aggregation Control Services for D...
Dear Andreea,
I think the subject is topical and he will have to clear a lot of bibliographic references here are someone hoping you find the right address.
Best regards
Controller effectiveness can be demonstrated
one is robustness and other one is sensitivity. But both can be measured through any performance measures, the common way using error quantifiers.
There are several quantifiers for example
error norm, integral square error, sum of absolute error, etc.
You can search more with these key words
ISE, IATE, IAE, etc.
Yes, in your example, both can be measured individually, but how can you measure them both while keeping the logic and the relevance of the whole?
For single input single output. The performance measure is easy and direct, like steady state error, for example which is a scalar quantity. When the system becomes Multi-input-Multi-output, you can use a vector where you list all the errors or the performance indexes you use for single loop case, then take one of the vector norms, such as the L1, L2 or L_infinity norms of vectors, i,e [y1^2+y2^2]. Other ways to measure the performance is through measuring the energy which is usually done by using a functional. See optimal control literature for more info.
The term effectiveness really has no meaning unless weighed against some requirements and it's measuring the gap between performance and requirements that determines effectiveness. So you control how effective a system is, not only by design of the system, but how you define the requirements. If you fail to define requirements then you may fall into the trap of hopelessly trying to determine if the design is finished. There are no standard requirements. Commitees, organizations try to standardize in different fields, but most systems are unique.
Hi Andreea - I think Tariq has provided a very good answer. Just to expand a bit: as others have said, you need to define what is important, and these items may include actuator efforts or power as well as tracking/regulating measures - for example if you have a limited energy source onboard the controlled system. The LQG approach provides a model for combining actuation costs as well as error costs. However, as Tariq points out, you may be more concerned about maximum values rather than rms values (infinity-norm as opposed to 2-norm).
If you are asking the practical way to measure performance of a control system, it is straight forward. In any control system, there is a target expected performance as per design. Control system must meet this expected performance. There is always a permissible range of target. If the result of control system is well with in the permissible limits, then performance is considered good.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems