Design of an Efficient Maximum Power Point Tracker Based on ANFIS Using an Experimental Photovoltaic System Data

Round 1

Reviewer 1 Report

My comments are:

1.    Some writing errors and small typos were found in the manuscript, please correct and check all the probable errors and mistakes.

2.    Clarify what is the paper contribution and what parts of the paper are original.

3.    Please discuss in detail how the authors choose fuzzy rules.

4.    Please provide the required reference for Figure 2.

5.    Please provide the required reference for equation (1).

Author Response

1.     Some writing errors and small typos were found in the manuscript, please correct and check all the probable errors and mistakes.

The manuscript is revised and checked all the probable errors and mistakes.  

2.     Clarify what is the paper contribution and what parts of the paper are original.

The novelty of this work is that an experimental training data is collected during one year from experimental tests of a PV array installed at Brunel University London, London, United Kingdom, and then, they are analysed and optimized using Curve Fitting technique to design an efficient maximum power point tracking technique for photovoltaic system. Please, see page 5 (190-193). The original parts of paper are 5,6,7,8, and 9.

3.     Please discuss in detail how the authors choose fuzzy rules.

Thank you for this suggestion. The fuzzy rules of ANFIS technique are generated based on the training data sets. We are just training and tuning the ANFIS model. Please, see section 7.

4.     Please provide the required reference for Figure 2.

               The required reference of Figure 2 is added.

5.    Please provide the required reference for equation (1).

The required reference of Eq. (1) is added.

Reviewer 2 Report

The authors use an ANFIS to achieve the greatest possible power in a photovoltaic system. The system is trained with data obtained for one year by the authors themselves. This Maximum power point tracking technique is compared to others to check its efficiency. This paper is well written and it is easy to read. The presentation of ideas is very clear and in a very well structured way. The justification of the paper is well explained, but the document can be improved by solving the issues that appear below.

Q1 Line 48. Fig 1 appears very far from this reference.

Q2 Line 55. Some reference to the conventional hill climbing technique must be added, for example:

Xiao, W., & Dunford, W. G. (2004, June). A modified adaptive hill climbing MPPT method for photovoltaic power systems. In 2004 IEEE 35th annual power electronics specialists conference (IEEE Cat. No. 04CH37551) (Vol. 3, pp. 1957-1963). Ieee.

Q3 First lines of section 2 are partially repeated.

Q4 Line 162. The acronym SEPIC is not defined.

Q5 Section 4. Add an introduction on the uses of ANFIS in other nonlinear systems approach applications.

Q6 Line 271. Can any reference be included as an example?

Q7 Line 300. The Internet 'i' must appear in uppercase.

Q8 Section 5 could include more information as a table with the total data that has been obtained, how much data is obtained per day, ..

Q9 The expression "whilst" is repeated very often in the text.

Q10 Section 7. It is indicated that an adjustment of a large number of data is made. It is understood that one year data is available and 40 days are chosen, 10 for each season. How was this selection produced?

Q11 Section 7. It is not indicated how many data have been used.

Q12 Table 2. What unemployment criteria have been chosen to determine the number of epochs?

Q13.Table 3. The difference between trimf, gbellmf and gaussmf is very small. Why?

Q14 Table 3. The functions have not been explained, I suppose they are the same as those used in MATLAB, right?

Q15 Section 5. Include an exact definition of convergence time.

Q16 Table 4, Convergence times are very similar. The gain of the output power is very small. The three types of MPPT have similar benefits. Give some reason to choose ANFIS.

Q17 It would be interesting to provide the data to reproduce the proposal.

Q18 The training of the data has not been explained or a whole set has been used. It would be necessary to differentiate the data set into three parts: training, test and validation. An example is found in:

Buitrago, J., & Asfour, S. (2017). Short-term forecasting of electric loads using nonlinear autoregressive artificial neural networks with exogenous vector inputs. Energies, 10(1), 40.

Author Response

Q1 Line 48. Fig 1 appears very far from this reference.

The referred location of Fig 1 is moved as advised. Please, see page (2) line (92)

Q2 Line 55. Some reference to the conventional hill climbing technique must be added, for example:

Xiao, W., & Dunford, W. G. (2004, June). A modified adaptive hill climbing MPPT method for photovoltaic power systems. In 2004 IEEE 35th annual power electronics specialists conference (IEEE Cat. No. 04CH37551) (Vol. 3, pp. 1957-1963). IEEE.

The reference of a conventional hill climbing technique is added. Please see ref. 13

Q3 First lines of section 2 are partially repeated.

The First lines of section 2 are revised. See page (3) line (106)

Q4 Line 162. The acronym SEPIC is not defined.

The acronym SEPIC is defined. Please see page (4) line (157)

Q5 Section 4. Add an introduction on the uses of ANFIS in other nonlinear systems approach applications.

The introduction on the uses of ANFIS in other nonlinear systems approach applications is added . please see page( 7)  line( 246-249)

Q6 Line 271. Can any reference be included as an example?

The example reference based on the traditional ANFIS-MPPT method is given. Please, see page (8) line(268)

Q7 Line 300. The Internet 'i' must appear in uppercase.

The uppercase letter in word ‘’ internet’’ is revised as advised.  Please, see page (9) line (298)

Q8 Section 5 could include more information as a table with the total data that has been obtained, how much data is obtained per day.

About 48,500 collected data sets are measured during one year. The sunny boy controller pulse reads the data each 5 minutes intervals in the day and switches off in the night then wakes up every 15 minutes to check whether it can reach the system otherwise it returns to power-save mode. That means, the the number of data per day is change regarding to the sunny day. In addition, we gave a sample data in fig 6. It is revised and referred. Please, see page (9) line (299-303)

Q9 The expression "whilst" is repeated very often in the text.

The expression "whilst" is revised, and alternative phrase has been used.

Q10 Section 7. It is indicated that an adjustment of a large number of data is made. It is understood that one year data is available and 40 days are chosen, 10 for each season. How was this selection produced?

Thank you for this deep equation. The answer is already given at the top of page 12 line (361-363). (( The estimation accurate PV power depends on the training dataset; therefore, it is very important to select the training data with wide variations of the solar irradiance and the operating temperature[38])). That means, we are selected the days whose have wide variations of weather conditions.

Q11 Section 7. It is not indicated how many data have been used.

About 6200 data have been used to training the proposed ANFIS model. It is revised and referred. See page 12 line (364-365).  

Q12 Table 2. What unemployment criteria have been chosen to determine the number of epochs?

The number of epochs are not chosen. They are gotten after training end of optimised ANFIS model.

Q13.Table 3. The difference between trimf, gbellmf and gaussmf is very small. Why?

Thank you for this equation. Table 3 explains the training error of ANFIS model based on different membership function and this number error automatically determined by the training of ANFIS network. Although this difference between trimf, gbellmf and gaussmf is very small, but its very effective on the output predicting function of ANFIS model.

Q14 Table 3. The functions have not been explained, I suppose they are the same as those used in MATLAB, right?

Yes, the membership functions are the principle work of ANFIS model and they are explained in Matlab tools.

Q15 Section 5. Include an exact definition of convergence time.

The definition of convergence time is explained. Please, see page 14 line 429

Q16 Table 4, Convergence times are very similar. The gain of the output power is very small. The three types of MPPT have similar benefits. Give some reason to choose ANFIS.

Thank you for this tricky equation. The main reasons to choose the ANFIS technique to designed MPPT controller are that it is able to avoid the drift problem associated with fast changing irradiation as well as its low oscillation at steady state conditions, as result, it achieving the highest efficiency on semi-cloudy days.

Q17 It would be interesting to provide the data to reproduce the proposal.

It is good idea but the collected data are very large about 48500. However, it is our pleasure to send the data who is requested.   

Q18 The training of the data has not been explained or a whole set has been used. It would be necessary to differentiate the data set into three parts: training, test and validation. An example is found in:

Buitrago, J., & Asfour, S. (2017). Short-term forecasting of electric loads using nonlinear autoregressive artificial neural networks with exogenous vector inputs. Energies, 10(1), 40.

Thank you for this note. The divided training data in to training, test and validation is very suitable for a training of artificial neural network (ANN) technique, as explained in this reference.     

Round 2

Reviewer 1 Report

 I recommend it for acceptance.

Reviewer 2 Report

The authors have responded adequately to all the suggested changes.