Abstract: During eclipse periods, the spacecraft relies on electricity, which its solar arrays produce and store in batteries. Forecasting a solar array’s power output employed during space missions is of significant importance. Varying space environments and satellite loads, which are characterized by significant randomness and uncertainty, affect the generated power of the spacecraft’s solar array. These challenges pose difficulties in power prediction. To address these issues and achieve a more accurate estimation of the solar array’s generated power during a space mission, this study develops a metabolic model termed TDGM(1, 1, r) that incorporates an enhanced accumulating fractional-order, optimizing it within the discrete grey TDGM(1,1) model’s framework with three parameters. The optimization model’s objective function is defined as the mean absolute percentage error (MAPE) within the modeling context. In order to minimize MAPE, the differential equation’s order and accumulation number are determined using a particle swarm approach. The TDGM(1, 1, r) demonstrates superior forecasting performance in comparison to the classical GM(1,1) and Grey–Markov models. These findings indi-cate the superiority of TDGM(1,1,r) over GM(1,1) and Grey–Markov, with improvements of 84.2% and 81.2% for MAPE (from 1.83% to 0.29% and from 1.54% to 0.29%). The metabolic TDGM(1,1,r) employing the particle swarm algorithm (PSO) is better suited for short-term predictions. Finally, relevant suggestions for future development of the prediction model are proposed.  

Key words: Solar array, Prediction, TDGM(1,1,r), Spacecraft, Particle swarm optimization