For as sophisticated as our energy infrastructure may seem to us, it is still a challenge to make the best trade-off between the effort required to produce our energy and the costs to the stakeholders.
Now, a group of international researchers from TU Delft and Lund University have proposed an efficient method for considering both types of costs in a distributed control problem.
Up until now, decision-makers have been using complex calculations, called algorithms, to optimize the energy use and economic costs, but arriving at an optimal mix of the two has always been limited by the calculation time. This new method has an accelerated algorithm that guarantees reaching a solution significantly faster than commercially available software.
Algorithms provide input to many of the systems that pervade our lives. Let’s say that you are trying to organize a dinner party with all of your friends, but your roommate’s girlfriend is a vegetarian, your best friend always eats twice as much as everyone else, and your girlfriend can’t stand sitting next to your best friend. With so many preferences, or constraints, to consider on top of the money that you will spend on food, deciding exactly how to execute the get-together becomes problematic. But you could formulate the problem so that an algorithm could churn out an indication of the seating arrangement and how much to buy of what, the variables, with the hit of a button. The field of algorithm research studies much more complicated cases, such as how to create an investment portfolio that minimizes risk and maximizes return or how a hospital manager decides how to balance the available space she has for patients and her available staff.
Many theoretical cases have been proposed by researchers for decades, and corresponding algorithms have been created, but calculation time has always been a concern. For example, when variables and their constraints are related to one another, the computer that is executing the algorithm needs to consider a large body of information, both of which increases the calculation time. The Delft-Lund team has now demonstrated that their algorithm can achieve a solution to such a coupled-constraint problem at its fastest theoretical convergence rate. Practically, their method could help optimize energy inputs, or so-called control efforts, and electricity profits in power-producing water networks, which could be represented as a large-scale optimization problem with coupled variables. Their algorithm has recently been publishedin the journal Automatica.
The team used a method called dual decomposition to break up the problem into manageable pieces. Therefore, the calculation can be distributed among several computers who work in parallel to achieve the solution. When comparing the convergence rates to those of software that is available on the market, the researchers found they could outperform them. A mentor in the study, Dr. Tamas Keviczky, who is an assistant professor at the Delft Center for Systems and Control, said, “For these types of problems, this is the fastest approach possible.”
Comments are closed.