N.C. Cruz

A parallel Teaching---Learning-Based Optimization procedure for automatic heliostat aiming

The flux distribution generated by the heliostat field of solar central receiver system (SCRS) over the receiver needs to be carefully controlled. It is necessary to avoid dangerous radiation peaks and temperature distributions to maximize the efficiency and keep the system in a safe state. These tasks imply both selecting the subset of heliostats to be activated and assigning each one to a certain aiming point at the receiver. The heliostat field is usually under human control and supervision, what is a potential limiting factor. Thus, there is an active research line to define automatic aiming procedures. In fact, a general and autonomous methodology is being developed by the authors of this work. However, the mathematical modeling leads to face a complex large-scale optimization problem. In this work, applying Teaching–Learning-Based Optimization (TLBO), a population-based large-scale optimizer, is considered. It is intended to serve to perform large explorations of the search-space to finally deploy further local optimizers over the most promising results. Considering the computational cost of the objective function, a parallel version of TLBO has been developed. It significantly accelerates the procedure, and the possibility of being included in a more complex process remains viable. Additionally, the parallel version of TLBO is also linked as a generic open-source library.

High performance computing for the heliostat field layout evaluation

In Solar Central Receiver Systems (SCRS), the heliostat field is generally the most important subsystem in terms of initial investment and energy losses. Therefore, heliostat field layout needs to be carefully designed and optimized when deploying this kind of power facilities. This optimization procedure can be focused on multiple and heterogeneous criteria depending on particular factors that lead to define different optimization problems based on specific objective functions. However, objective functions defined for this problem are, in general terms, computationally very expensive. This fact may make an exhaustive optimization process infeasible, specially depending on the available resources, and forces particular simplifications at some steps of the process. Fortunately, some of the objective functions defined can benefit from parallelization, even though this idea is not usually pointed out or discussed, and then, become affordable in better conditions. In this paper, the heliostat field optical efficiency, which is a common objective function in this area, is analyzed to be parallelized by three different approaches.

A two-layered solution for automatic heliostat aiming

Abstract The efficiency and safety of a solar central receiver system depend on the flux distribution reflected by the heliostat field on its receiver. Thus, the field must be carefully controlled to avoid dangerous radiation peaks and temperature gradients while also maximizing the efficiency of the system. Control tasks include deciding which heliostats to activate and where to aim them. The field is usually under direct human supervision, which is a potential limitation, and automatic aiming procedures are of great interest. This work proposes a general aiming methodology for flat-plate receivers. It intends to cover heliostat selection and aim point assignation to replicate any given reference flux distribution on the receiver. The methodology, which addresses this situation as a large-scale optimization problem, defines two consecutive stages. The first one handles heliostat selection by applying a specific genetic algorithm. The second one, based on a local gradient descent, assigns a final aim point to every active heliostat. The proposed methodology, in contrast to other existing methods in the literature, is not limited to achieve any specific target distribution. It exploits the analytical characterization of the considered field to minimize the accumulated squared error between any reference flux distribution and the achieved one. The results show very good replication quality and, considering its execution time, this method is suitable for preliminary and high-resolution field configuration.

Design of a parallel genetic algorithm for continuous and pattern-free heliostat field optimization

The heliostat field of solar power tower plants can suppose up to 50% of investment costs and 40% of energy loss. Unfortunately, obtaining an optimal field requires facing a complex non-convex, continuous, large-scale, and constrained optimization problem. Although pattern-based layouts and iterative deployment are popular heuristics to simplify the problem, they limit flexibility and might be suboptimal. This work describes a new genetic algorithm for continuous and pattern-free heliostat field optimization. Considering the potential computational cost of the objective function and the necessity of broad explorations, it has been adapted to run in parallel on shared-memory environments. It relies on elitism, uniform crossover, static penalization of infeasibility, and tournament selection. Interesting experimental results show an optimization speedup up to 15

A New Methodology for Building-Up a Robust Model for Heliostat Field Flux Characterization

\times

On building-up a yearly characterization of a heliostat field: A new methodology and an application example

× with 16 threads. It could approximately reduce a one year runtime, at complete optimization, to a month only. The optimizer has also been made available as a generic C++ library.