基于解析模型的太阳能塔式系统实时高精度辐射通量分布仿真

Abstract Analytical models are commonly used for simulating radiative flux density distributions (RFDD) on the receiver in the design, optimization and operation of solar power tower systems. However, existing analytical model simulation methods typically accumulate the simulation results of individual heliostats sequentially, leading to inefficiencies, high computational costs and loss of the efficiency advantage of the analytical model, especially for large-scale heliostat fields. In this paper, a novel real-time and accurate RFDD simulation method for large-scale fields based on an analytical model is proposed, namely fast-NEG (Neural Elliptical Gaussian). The proposed simulation method re-frames the slow sequential accumulation process into highly parallelized computation on a graphics processing unit (GPU), where each thread computes the radiative flux density mapping from a receiver pixel to a visible heliostat. The efficient parallel simulation process is applicable to all Gaussian or elliptical Gaussian analytical models. The accurate NEG model is adopted and accelerated by tensor product decomposition and precomputation to describe the RFDD of the individual heliostat. The Fast-NEG simulation is compared with the state-of-the-art method Quasi-Monte Carlo Ray Tracing (QMCRT) on a large-scale field with 6282 heliostats at different times of one year, resulting in simulation speed improvement by two orders of magnitude, and a mean relative error of total energy, peak value and Root Mean Square Error (RMSE) are 0.40%, 0.25% and 0.0068%, respectively. Compared with the prevalent analytical model simulations, the Fast-NEG approach significantly enhances the simulation accuracy and boosts the efficiency by 1-7 orders of magnitude.