Generative heliostat field layout optimization and application based on an improved Harris Hawk Optimization algorithm

This study proposes a pattern-free layout method for continuous generation and optimization of heliostat positions based on optical efficiency. Using annual weighted cosine efficiency as the objective function, the initial layout is generated through a continuous search from the optimal point of the entire field. This is followed by a secondary optimization and selection of all heliostats based on annual weighted optical efficiency. Points with efficiency above the threshold are retained, while low-efficiency points are discarded and re-entered into the search optimization until all eligible heliostat positions are identified, ultimately resulting in a highly efficient heliostat field. The efficiency improvement from the generative pattern-free layout optimization is primarily attributed to the initial search using annual weighted cosine efficiency, which enhances the annual weighted cosine efficiency of the field, and the secondary screening optimization, which improves shading and blocking efficiency. Compared to the original PS10 heliostat field, the annual weighted efficiency increased by 1.77%. Finally, two clustering-based pattern-free layout methods are proposed. The total distance from all points within a K-means cluster to the reference center is 58,238.1974 m, a reduction of 71.18% compared to the untreated original heliostat field. The method based on the Gaussian Mixture Model reduces the distance to the reference center by 66.90%. Classification based on optical efficiency reflects the overall distribution structure of the heliostat field's optical efficiency, reducing the layout difficulty of pattern-free heliostat fields and providing feasibility for transforming theoretical research into practical engineering applications.