Monte Carlo Ray Tracing In Parallel For Flux Density Distribution Reflected By Heliostats

The solar power tower system can generate electricity power and has advantages in good thermal storage,low cost and environmental friendliness,etc.To improve the efficiency of power generation in the system,the heliostat filed layout and focusing strategy should be optimized.The simulation of the flux density distribution on the receiver surface,contributed by all heliostats in the field,is the foundation of these optirmization tasks.Monte Carlo ray tracing is one of the widely used methods in flux density distribution simulation.Traditional ray tracing algorithm samples and traces tens of millions or even more rays to obtain a high precision result for a single heliostat.It leads to high time complexity and storage consumptions.In addition,the maximum flux density varies with the simulations due to the discretization and randomness in Monte Carlo ray tracing method.To this end,Monte Carlo ray tracing in parallel for flux density distribution reflected by heliostats is investigated carefully in this thesis.First,an efficient and parallel ray tracing simulation method is proposed,namely Quasi Monte Carlo Ray Tracing(QMCRT in short).For time complexity,QMCRT utilizes the high throughput and tremendous parallel computing power of GPU to accelerate the ray tracing calculations.Meanwhile,QMCRT takes the advantage of bidirectional ray tracing method to improve the intersection efficiency between rays and heliostats.For space complexity,QMCRT pre-computes and stores two lookup tables in advance about sun shape and normal distribution of heliostat surface.By randomly combining the start indices of extracted entrances of two lookup tables,the randomness in the algorithm is well guaranteed while the demands for time complexity and memory consumption on GPU drop significantly.To address the problem of unstable maximum flux density value,QMCRT performs a trimmed mean smoothing operation to filter the flux density distribution on the receiver,while keeping the total energy as unchanged as possible.As a result,a stable and accurate maximum flux density in the simulation result will be obtained.Compared with two GPU-based ray tracing algorithms,simulation softwares,i.e.,SolTrace and Tonatiuh,as well as the captured data,QMCRT is more accurate,efficient,and the generated maximum flux density is more stable.A simulation system for a large scale heliostat field based on QMCRT is proposed in the paper as well.For the heliostat scene processing,class encapsulation is used to increase its scalability and maintainability;for parallel ray tracing calculation,the system uses traditional process-oriented method to improve the computational efficiency.The system could be extended to other new type of heliostats flexibly by inheriting the existing base class without modification of main part of codes.Experimental results demonstrate that the simulation system is scalable,maintainable and reusable.