Research on power dynamic data sample generation technology based on brain-like computation and its efficient computation methods
, , , , , and
Feb 03, 2025
About this article
Published Online: Feb 03, 2025
Received: Sep 01, 2024
Accepted: Dec 20, 2024
DOI: https://doi.org/10.2478/amns-2025-0023
Keywords
© 2025 Peng Zhang et al., published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.
Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.
Figure 6.
Figure 7.
Figure 8.
Figure 9.
Figure 10.
The quality of the model generated samples
| Evaluation criteria for generating samples | Generated sample quality | Remark |
|---|---|---|
| Degree of Completeness | 0.97 | The proportion of the number of feature dimensions contained in each generated fault sample to the total number of features |
| Coverage rate | 0.90 | Whether the device fault sample database contains all typical discharge defect types |
| Degree of Overlap | 0.95 | The degree of overlap between the value range of each parameter in the generated sample and the actual sample |
| Identification quality | 1.0 | The clustering distribution of good samples and damaged samples in the generated samples |
Change in uniformity of the sample set before and after adding virtual input samples
| Original training sample + virtual input sample size | No fixed L2 differences |
|---|---|
| 280+0 | 5.28E-02 |
| 280+100 | 1.51E-02 |
| 280+200 | 1.37E-02 |
| 280+300 | 1.47E-02 |
| 280+400 | 1.29E-02 |
| 280+500 | 1.34E-02 |
| 280+600 | 1.56E-02 |
Parameters of GPNN on Three-dimensional standard functions
| Hyperparameter description | Set value |
|---|---|
| G Optimizer used | Adam |
| D Optimizer used | Adam |
| G The learning rate used | 0.001 |
| D The learning rate used | 0.005 |
| Generator and recognizer | 30 Neurons |
| Lot size | 200 |
| Noise allocation | Noise with a 5-dimensional Gaussian distribution |
| Number of iterations | 2000 |