A new 5D19P digitization-fused inference to analyze the motion characteristics of output power for floating photovoltaic systems

Floating photovoltaic systems represent complex electromechanically coupled systems whose power output is significantly influenced by external dynamic conditions, yet effective methods for quantifying these motion-induced effects remain lacking. To tackle these challenges, the paper introduces a new 5-dimensional 19-parameter digitization-fused inference method to analyze the motion characteristics of floating photovoltaic systems' output power. By constructing a dynamic model of floating photovoltaic modules and a 5-dimensional 19-parameter digitization-fused model of output power, a complete characteristic dataset of motion and output power is achieved, which effectively integrates simulation, experimental and actual measurement data. To quantify the feature importance relationship between the motion information and the output power, a fusion inference method based on gradient boosting decision tree is established. Finally, the proposed inference method is evaluated using simulation, experimental, and measured data. The results demonstrate that the digitization-fused model achieves an accuracy of over 99.2%, with a determination coefficient of 0.985 for the inference method. The motion of Z-axis direction is the primary factor, accounting for 70.9%. In conclusion, the new digitization-fused inference method proposed in the paper has strong generalization ability and interpretability.