Abstract
Heliostats in tower solar plants, typically mounted on columnar supports in open terrains. During operation, the mirror panel is subject to significant wind pressure, resulting in considerable forces on the support structure. This paper explores the probabilistic characteristics of forces in the heliostat support structure under various conditions. Through time history curves, histograms of probability density distributions, skewness coefficients, and kurtosis coefficients, it assesses whether the forces in the heliostat support structure follow a Gaussian distribution model across all conditions and summarizes corresponding criteria. Neural network models trained using the DBO and BP algorithm are employed to calculate force coefficients. Dung beetle optimization (DBO) algorithm is a metaheuristic algorithm mimicking dung beetle ball-rolling behavior, while Back propagation (BP) neural network is a feedforward artificial neural network trained via error backpropagation to adjust parameters. Relevant indicators are used to compare the models' performance. Based on the calculation results and summarized criteria, the conditions are classified as either following Gaussian or non-Gaussian distributions. The main reasons for the Gaussian and non-Gaussian characteristics of forces in the heliostat support structure under certain conditions are explained, providing a reference for the wind-resistant design of heliostat structures.