by Roland Jackson

Optimising power control

1 May 2000



Using software to optimise control patterns during the recent design of an operating cycle in a plant has boosted a utility’s confidence that a proposed operating plan could be achieved.


Determining a new power control pattern would normally take an engineer weeks spent running nuclear physics and thermal hydraulic simulation software. GE engineers have used simulation software to create a design exploration engine that, in one case, drove it to an optimum pattern in only eight hours of unattended operation.

Reactor control parameters

The control rods in a reactor’s core comprise a material that absorbs neutrons to dampen the reaction. The number of rods that are inserted at any time, and the depth to which they are pushed, is called the control pattern and determines the operating level of the reactor. The control pattern is varied over the life cycle of the core, typically at 10 to 20 or more state points, in order to fully utilise each different section of the core. Later in the life of the core, fewer rods are inserted because the core is generating fewer neutrons that need to be absorbed.

At every stage in a reactor’s life, the core must operate within safety margins that are tightly controlled by the Nuclear Regulatory Commission. The goal of the plant operator is usually to maximise the electrical output of the reactor while staying within the operating margins at every phase of the core’s life. This provides a very challenging optimisation problem that involves determining core flow rate and a position for each set of control rods at every state point. In a typical reactor design, there are seven sets of control rods, each of which can be set to 24 possible positions.

Close to margins

GE engineers originally developed an efficient control pattern for a particular nuclear power plant. This plan was predicated on an assumed operation plan in the previous cycle. Circumstances caused the operator to deviate from the original prior cycle operating plan, which meant that the original control plan for the design cycle no longer provided the expected operating margin. The plant operator found that the reduced margins would reduce the cycle operating flexibility and increase the risk of de-rate to an unacceptable level.

GE engineers were asked to develop an efficient control pattern to fit the new operating plan, so that the necessary design margin could be regained within a tight deadline. The traditional approach would have been for the engineers to sit down with a proprietary nuclear physics and thermal hydraulics simulation program developed by GE, called Panacea, and investigate alternative control patterns. The problem is that the number of possible control patterns is so large that it takes an engineer a considerable amount of time just to get a feel for the sensitivities involved. It could have taken one or more days, depending on experience, for an engineer to develop a reasonable solution for a single state point. Even then, there would be no guarantee that the solution was optimised or even close to the optimum.

Exploring a better alternative

In order to overcome this problem, GE engineers had already begun experimenting with a design exploration software package called iSIGHT. The package was originally developed at General Electric Corporate Research and Development, but is now being developed, marketed and supported by Engineous Software in North Carolina, US. This package is a general-purpose optimisation framework that can be coupled to virtually any simulation model and computer–aided engineering software package. It selects input parameters for the analysis package, runs the analysis, reads the results, maps them onto a design space and then determines the best solution for the problem.

Key features of the software package, although not applicable to this case, are its ability to integrate the simulation processes, use approximation methods to substantially reduce the number of iterations required for computer intensive analysis packages, and to evaluate the results of multiple analysis disciplines.

GE engineers coupled iSIGHT to their existing proprietary analysis code. This involved the development of script files that helped the optimisation software create the right format input file and read the output file from Panacea. They then set constraints for each of the design parameters and defined variables for optimisation. The initial process of setting up the control pattern optimisation problem took about two weeks. Additional effort to modify this model for other reactor control pattern problems would take about four hours. Then, using iSIGHT software they performed a series of analyses by controlling the input parameters that are fed into the analysis code.

iSIGHT allowed the engineer to apply one of several optimisation tools, depending on the stage of optimisation. Early progress was made by applying rules to establish constraints and a genetic algorithm to search a broad solution space. As the solution converged iSIGHT constructed a multidimensional design space by varying each of the design parameters within their constrained range.

Iteration to a solution

During the initial stages of the design exploration process, the genetic algorithm looked for patterns of inputs that yielded progressively more effective solutions.

The primary purpose of this first phase was to establish a feasible solution space. That is a set of inputs that resulted in a solution that met all of the constraints but was not necessarily the best. Once this series of analyses was completed, the software further investigated the area of the design space that it had identified by performing a new series of analyses using a more efficient local search procedure. This second series of analyses fixed the optimum design to a much higher level of accuracy.

The program proceeded through several more stages of analyses with progressively smaller input variations until it reached the true optimum, a combination of design parameters that ensures maximum performance while satisfying other design constraints.

Many of the state points were not close to an operating margin and could therefore be patterned by hand. On the other hand, eight were near limits and were optimised using the software engine. It took an hour to optimise each of these points without any manual intervention on the part of GE engineers. The state points were optimised one at a time over a period of several weeks in order to communicate with and obtain approval of each point by the client. In general, it is entirely possible to batch the optimisation runs for all of the state points together and solve them overnight.

The utility was pleased with the quality of the results and the speed with which they were created. The design margins achieved with the new control patterns increased the utility’s confidence that the new control plan would result in adequate operating margins, eliminating any significant probability of de-rating and any associated loss in revenue.

The result of using the design exploration techniques within iSIGHT has been a substantial productivity increase – from one or more days of engineering time per state point to the time required to set up and execute an optimisation run which is, in most cases, less than one hour.

At the same time, the move to optimisation provides the certainty that the design solution is the best that can be determined.

In the past, when an engineer had difficulty finding a pattern that generated reasonable amounts of energy and met operating margins, it was difficult to determine whether the fuel design or rod patterns were to blame. Now, when the optimisation software cannot find a satisfactory solution, the fuel design is clearly the culprit.

All in all, the streamlined optimisation has dramatically improved the design process.




Privacy Policy
We have updated our privacy policy. In the latest update it explains what cookies are and how we use them on our site. To learn more about cookies and their benefits, please view our privacy policy. Please be aware that parts of this site will not function correctly if you disable cookies. By continuing to use this site, you consent to our use of cookies in accordance with our privacy policy unless you have disabled them.