Development of an Improved Heuristic Algorithm for Inverse Lighting Design Based on Efficacy Evaluation

The new trend of lighting-building design integration necessitates the multiple lighting design goals of quality improvement，atmosphere forming，and energy saving. This study introduces generative methods，a mathematical model for inverse lighting design，and an improved heuristic algorithm to meet these goals. This model adjusts the spatial luminance coefficient(Feu)and the illuminance on the working surface to improve the visual quality and ambiance，as well as optimize energy saving via controlling lighting power density(LPD). The heuristic algorithm optimizes 5-dimensional decision variables that describe the luminous flux，lateral/longitudinal distance，and the average wall/floor reflectivity to obtain the optimal solution. To determine the best algorithm，the performance of four algorithms，namely，genetic algorithm(GA)，differential evolution(DE)，particle swarm optimization(PSO)，and artificial fish swarm algorithm(AFSA)，was compared for inverse design optimization via the Python library scikit-opt. Then，one inverse design and four mathematical test functions were set as benchmarks to compare the accuracy，stability，and computational efficiency of the algorithms. The inverse design multi-strategy algorithm (IDMSA)was created based on these results. Its features include：①The choice between DE or AFSA operations is determined by the stage and the state of optimization；②The clustering and tail-chasing behavior of AFSA occurs in the von Neumann neighborhood；③The step parameter is adaptive. The efficacy of IDMSA was verified based on benchmarks that were modified to be more difficult to optimize. The results demonstrate that IDMSA achieves the best accuracy and stability and is faster than AFSA and DE. The joint optimization of the illuminance-Feu-LPD can be realized via the proposed mathematical model and IDMSA. The generated scheme conforms to the unified glare rating and uniformity limit of the lighting standard. The deviation of illuminance and Feu from the respective reference values were less than 5%. Furthermore，the LPD obtained by IDMSA was 2.6% and 6.1% lower than that obtained by DE and PSO，respectively，demonstrating better energy efficiency. © 2023 Tianjin University. All rights reserved.