Mars Express Orbiter Power Consumption Prediction Based on Bionic Hierarchical Learning Network.

Area of Science:

• Aerospace Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Accurate power consumption prediction is crucial for the Mars Express (MEX) mission's longevity and operational efficiency.
• The complex Martian environment and solar cycles hinder traditional methods in capturing power consumption's intraperiodic and interperiodic features.

Purpose of the Study:

• To develop an advanced model for enhancing power consumption predictions for the Mars Express mission.
• To address the limitations of existing methods in handling complex temporal patterns in power data.

Main Methods:

• Introduction of the bionic hierarchical learning network (BHL-Net), utilizing 2-D frequency preprocessing and brain visual modeling.
• Incorporation of temporal oscillation activation, stripe intensity attention, and multihead attention adaptive aggregation modules.
• Mimicking prefrontal cortex (PFC) natural image encoding for improved predictive performance.

Main Results:

• BHL-Net demonstrated superior performance compared to existing transformer-based models in MEX power consumption prediction.
• Ablation studies confirmed the efficacy of the FFT-based 2-D transformation and the bionic attention framework.
• The model successfully captured variations within and between complex power consumption cycles.

Conclusions:

• BHL-Net offers a competitive solution for time series prediction, particularly for industrial applications with complex cyclical patterns.
• Emulating human brain response coding mechanisms enhances the model's ability to interpret intricate temporal data.
• The developed framework provides a robust approach for optimizing space mission power management.