Power Router Energy Optimization Sensitivity Analysis Results

Junyan Lyu's sensitivity analysis dataset contains experimental results from a reinforcement learning study on multi-agent energy management for power routers. The 5.5 KB XLS file details performance metrics like round rewards and voltage fluctuations from 420 training episodes. The dataset was uploaded to figshare in April 2026.

Use Cases

• Analyze the relationship between training episodes (420) and the stabilization of the average round reward (approximately -410) to assess algorithm convergence.
• Compare model performance using specific metrics: DC bus voltage fluctuation range (728V-732V), electricity cost (3846.36 yuan), and total runtime (53.32 seconds).
• Study the stabilization point of the strategy loss function (after 500 times) relative to other training dynamics.
• Validate the effectiveness of a double-buffer mechanism and Generative Adversarial Imitation Learning for improving data utilization and multi-agent collaboration.

Strengths

• Provides specific, quantitative results from 420 training episodes.
• Includes multiple performance metrics: reward, loss function, voltage range, cost, and runtime for direct comparison.

Limitations

• Extremely small scale at 5.5 KB, indicating a summary of results rather than raw training or simulation data.
• Unknown row count and lack of column definitions limit reproducibility and in-depth analysis.

Provenance

Source

Research by Junyan Lyu, published on figshare.

Collection Method

Results generated from simulation experiments using a Proximal Policy Optimization multi-agent framework.

Time Range

Experimental timeframe not specified; dataset reflects a published study.

Freshness

Dataset reflects research current as of its 2026 publication date.

Geography

Geographic scope of the power system model is not specified.