Electrical Modeling and Control of a Synchronous Non-Ideal Step-Down Converter Using a Proportional–Integral–Derivative Neural Network Controller
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper proposes a novel online tuning method for PID-controllers of Buck converters using neural networks to ensure convergence.
My comments are:
- Please include an overview of model reference adaptive control methods for power electronics and how to place your research within it.
- Specify how your method is different from [14].
- Explain all references individually and remove those that are not needed.
- Add all voltages and currents in the circuit diagrams as well as in the block diagram.
- Please discuss whether Theorem 1 is only for Buck converters or can be used for PID controllers in general.
- Please give more details on the neural network. What were the inputs and how many inputs did you have? Which activation function did you use?
- How were the parameters chosen? Is there a physical explanation behind it?
- How are the parameters delta_1 and delta_2 chosen?
- Please give some explanation about which parameters are known to the system. For example, if the network is used on an ageing system with changing parameter, do you assume the algorithm knows those parameters? Do you use any estimation?
- How do you calculate the derivatives in equation (17)?
- In the results, please show a scenario where your method can shine. For example a system that changes and where the alternative methods are failing.
- Please also show and discuss the limitations of your method.
Comments on the Quality of English LanguageThe English is generally understandable. However, there are some formulations that are not appropriate and need to be changed.
Examples:
"activity in order to learn something"
" the uncertain parameters of the passive elements"
Author Response
Dear Reviewer,
First and foremost, we sincerely thank you for your comments, which have undoubtedly contributed to significantly improving both the article and the proposed algorithm. We have carefully addressed each of your suggestions, placing particular emphasis on creating comparative tables, algorithm summary tables, and incorporating other enhancements as recommended.
Likewise, we were diligent in reviewing and refining the English throughout the article to ensure clarity and quality. We hope you find the revised manuscript satisfactory. For details, please see the attachment document.
Thank you once again for your valuable feedback. Wishing you happy holidays and a wonderful New Year!
Best regards,
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe article presents an innovative approach to control a DC-DC converter based on a self-learning PID controller using neural networks. The main emphasis is on the integration of Lyapunov theory to ensure stability and optimization of the controller parameters. The study is technically sound and includes a comparison with traditional control methods, which reinforces the significance of the proposed approach. Unfortunately, the authors have not been able to develop their idea in an appropriate way, expanding and deepening their research. My important remarks and comments are as follows:
- the manuscript needs a complete review and editing. For example, on page 2 line 30 it is not clear which literature is cited. There is also a list of definitions and abbreviations used;
- the literature review is quite superficial. There are many publications on this topic with significant contributions and in this regard, the purpose and contributions of the study are not well motivated;
- The study is based only on simulations. Adding real experimental data would increase the credibility of the results.
- Potential limitations in the implementation of the proposed architecture, such as computational complexity and hardware resource requirements, are not addressed.
- Although the approach has significant potential, the article does not provide examples of real-world applications or industrial integration scenarios.
- Some graphs (e.g. Bode plots) are not clear enough and improvement of their visualization could be considered.
- It would be useful to add a table summarizing the results of the comparisons for the different control strategies.
Author Response
Dear Reviewer,
First and foremost, we sincerely thank you for your comments, which have undoubtedly contributed to significantly improving both the article and the proposed algorithm. We have carefully addressed each of your suggestions, placing particular emphasis on creating comparative tables, algorithm summary tables, and incorporating other enhancements as recommended.
Likewise, we were diligent in reviewing and refining the English throughout the article to ensure clarity and quality. We hope you find the revised manuscript satisfactory. For details, please see the attachment document.
Thank you once again for your valuable feedback. Wishing you happy holidays and a wonderful New Year!
Best regards,
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors- How the method can be extended to dc-dc units in parallel?. Please, add a real application where this parallel or interleave connection is required.
- Ron, Roff, Vdiode and dead-time distortion must be included in the modeling of Synchronous non-ideal Buck converter in Fig. 3. These parameters are relevant during low duty cycle operation.
- In FIg. 11 to 15, please add a delay before to apply the reference step change to see the steady-state error before and after the step reference.
- With the proposed method, the time response is a quasi-Dirac pulse. How to avoid exciting natural resonances with this method?
- Please comment about the computational load used with methods PID-NN, ZG-PID, Type II, Type III and others
- Experimental results must be added to validate the proposed design. This is mandatory in electroncis field.
- Final contribution of the proposed method must be included along the text. A state of art table with similar solutions is missing at introduction to see the real innovation in this approach.
Author Response
Dear Reviewer,
First and foremost, we sincerely thank you for your comments, which have undoubtedly contributed to significantly improving both the article and the proposed algorithm. We have carefully addressed each of your suggestions, placing particular emphasis on creating comparative tables, algorithm summary tables, and incorporating other enhancements as recommended.
Likewise, we were diligent in reviewing and refining the English throughout the article to ensure clarity and quality. We hope you find the revised manuscript satisfactory. For details, please see the attachment document.
Thank you once again for your valuable feedback. Wishing you happy holidays and a wonderful New Year!
Best regards,
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsI have no further comments.
Author Response
Dear Reviewer,
Thank you for your thoughtful comments and insightful feedback on our article. Your suggestions have been invaluable in helping us improve and refine our work.
We truly appreciate the time and effort you dedicated to reviewing our manuscript.
Kind regards,
Reviewer 3 Report
Comments and Suggestions for Authors1) All the time response must be tested with a reference signal such as u(t-1). The responses are unclear by using u(t) step reference.
2) Fig. 16 must be presented in a different display, for example Method vs time.
3) Space utilization in Fig. 3 to 5 must be improved.
4) Why RC < RL ?. Do you have any reference or argument about that?
Author Response
Dear Reviewer,
We would like to begin by extending our warm regards and expressing our heartfelt appreciation for your comments. Below, we have provided detailed responses to your observations and suggestions:
-
Testing Time Response with a Reference Signal
We acknowledge your suggestion to test the time response with a reference signal such as u(t−1). However, based on several state-of-the-art works that incorporate adaptive gain models in controllers, such as:- Optimization of Neural Network-Based Self-Tuning PID Controllers for Second Order Mechanical Systems,
- Comparative Studies of PID Controller Tuning Methods on a DC-DC Boost Converter, and
- Self-Tuning Neural Network PID With Dynamic Response Control,
among others, we believe that maintaining the simulations without the mentioned delay aligns with the approach we aim to convey. Although certainly, there are other methodologies where what you mention is done and with good reason, for this particular study, we believe that the current representation is appropriate, as it effectively highlights the varying delays and the comparison of control laws. We hope this explanation clarifies our reasoning and aligns with your understanding of the approach we are discussing.
-
Improvement of Figure 16
As per your recommendation, Figure 16 has been redesigned to present the results in a more informative format, such as Method vs. Time. Additionally, we have also improved Figure 13, which shares a similar format, for consistency and clarity. -
Space Utilization in Figures 3 to 5
We have optimized the layout and space utilization in Figures 3 to 5 to enhance their readability and visual appeal. -
In the given example, the parameters were selected for illustrative purposes. The RL622 series from Bourns was used as a reference for the inductor, while the GXC series from Nichicon was chosen for the capacitor. In the selected series, the inductor exhibits a higher active resistance compared to the internal resistance of the capacitor. However, this selection was purely for demonstration and does not account for factors such as impedance at specific frequencies, currents, or other operational conditions. In future experimental models, all these parameters will be carefully considered.
Once again, thank you for your valuable feedback. Please let us know if further adjustments are required.
Best regards,
Jesús