Three-Level Active Power Filter Based on Model Predictive Control
Round 1
Reviewer 1 Report
The authors present a novel method for controlling APF. The beginning of the article, which presents the theoretical background, is very detailed. The presentation of the improved MPC is perfunctory (e.g. description of Figure 8).
My comments and suggestions:
- In equation (14) the prediction for the k+1 instants of the alfa and beta current components are given. Equation (21) also presents the prediction of alfa and beta current components in the k+1 instants. It is not clearly described why it is important to use the Lagrangian interpolation. Which equation is used during the predictions (based on the flowchart equation 21)?
- It is not clearly described how the reference currents are calculated.
- Figure 8 presents the schematic diagram of the overall voltage control, but it is not given what is C and C32 (most likely C referers to the Clarke transformation). It is also not mentioned what is the role of the low-pass filter.
- According to Figure 8 the improved model predictive control algorithm outputs a specific small current vector delta ip through a PI controller. Am I right in that, you removed by the novel method one weighting factor lambda_2 but you added a PI controller with two gains (proportional and integral gain)? How the gains of the PI controller should be determined?
- It is stated that in the paper the optimization should be elaborated for 27 voltage vector for normal case. However, there are some redundant voltage vectors, which is not taken into account.
- In Table 5 the fs is called as switching frequency. I think it should be sampling frequency, as the method is Finite-Set MPC having not constant switching frequency.
- Please provide what is the average switching frequency in steady state for the studied working points.
- One well-known disadvantage of real-time digital implementation
of MPC techniques is that due to the the extensive number of calculations there is a delay between the measurements and the
actuation, which can deteriorate the performance of the system
if it is not considered and not compensated. What delay compensation method is used? - I suggest to use a table for the text on page 12 lines 310-315.
- The method has the clear advantage to have a shorter turnaround time than classic method. It is good to have a comparison between the improved and classic method, but the conclusion which states that the improved version has better performance is misleading. The performance clearly depends on the selection of the weighting factor lambda_2 for the classic method and on the PI gains for the improved method. If the comparsion is made by using a non properly chosen weighting factor than it is not difficult to achieve better performance. The improved version uses a reduced control set, it means that there is a chance that the selected optimal voltage vector is not the global optimal. Theoretically, this means that we can only approach the behavior of the method which considers all voltage vectors.
Author Response
Thank you for your comment and we have responded according to the points you raised. Please refer to the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper proposed a three-level Active Power Filter based on MPC. The paper is well organized and written. I feel like the overall quality is ready for publication in its current form. However, a few more discussions on the computational performances and real-world implementations will further enhance the contribution of the paper. For example, is it possible to apply your algorithm in electric autonomous vehicles? I recently read a few interesting papers in the Journal of Intelligent and Connected Vehicles (https://www.emeraldgrouppublishing.com/journal/jicv ). Feel free to relate to some papers from there to strengthen your research.
Author Response
Thank you for your suggestions. We will study and consider further expanding the application field of this algorithm
Round 2
Reviewer 1 Report
Thank you for your answers to my queries. Based on yoru answers I understand more clearly the introduced method. I can accept the paper in it current form.
