Prediction of Greenhouse Indoor Air Temperature Using Artificial Intelligence (AI) Combined with Sensitivity Analysis

Round 1

Reviewer 1 Report

This article uses various machine learning techniques to model temperature in greenhouses based on outside variables, such as outside temperature, relative humidity, wind speed and solar radiation. The results showed that the machine learning techniques can simulate inside temperature fairly well, with modeling using Radial Basis Function (RBF) being the best among the techniques that were studied. I support the publication of this paper once the following comments are addressed by the authors:

1.)     I think the authors spent too much text on the mathematical basis for the machine learning techniques used in the paper, especially for the RBF model. I think the authors can simply that;

2.)     The authors need to add more descriptions of the modeling datasets. How many records are there? Do you model the inside temperature that is at the same hour as predictors (i.e., outside meteorological conditions) or do you model inside temperature a few hours later than the predictors? In the abstract, it says “predicting temperature in the next few hours”, but in the text, it does not say that. I think this needs clarification.

Author Response

This article uses various machine learning techniques to model temperature in greenhouses based on outside variables, such as outside temperature, relative humidity, wind speed and solar radiation. The results showed that the machine learning techniques can simulate inside temperature fairly well, with modeling using Radial Basis Function (RBF) being the best among the techniques that were studied. I support the publication of this paper once the following comments are addressed by the authors. 

Thank you for your comments. We have carefully reviewed your feedback and made revisions to the paper based on your valuable suggestions. Your input has been incredibly helpful in improving the quality of our work, and we appreciate the time and effort you took to provide it.

 

I think the authors spent too much text on the mathematical basis for the machine learning techniques used in the paper, especially for the RBF model. I think the authors can simply that.

Thank you so much for your comment. Sections 2.2. and 2.3 were revised based on your comments.

 

The authors need to add more descriptions of the modeling datasets. How many records are there? Do you model the inside temperature that is at the same hour as predictors (i.e., outside meteorological conditions) or do you model inside temperature a few hours later than the predictors? In the abstract, it says “predicting temperature in the next few hours”, but in the text, it does not say that. I think this needs clarification

Thank you so much for your comment. Neural network models typically utilize a sequence of data as training data to estimate the subsequent series of data, which often pertains to future time periods. Similarly, this study employed a comparable approach by utilizing a training dataset to train the neural network, which was then utilized to estimate future data points. Also the following text was added to the methodology section.

Also, the data were collected with a 5-minute interval for 10 days in September-October 2022.

Reviewer 2 Report

This manuscript employed several machine learning methods to predict indoor air temperature in an even-span Mediterranean greenhouse. The manuscript is of technical sound. Before possible publication, several concerns must be addressed.

1. The Abstract is too long and tedious. The authors are suggested to rewrite the Abstract, which should clearly show the problems of current studies, motivation of the paper, main steps of the method, and the results of this study.

2. In Introduction part, the authors need to not only review previous studies, but also analyze the shortcoming of these studies. In addition, the motivation the paper is weak,. At last, the contributions of the paper should be concluded and shown in a point-to-point format.

3. The font size in figures are incongruous. For example, the font size in Figure 4 is too small to see. In addition, the content of Figure 5 is very large, but the font size in Figure 5 is small.

4. Generally, Equations (7)-(10) is known to most readers. It is suggested to remove these Equations.

5. All figure and table captions are too simple and hard for readers to understand the table/figure independently to the context. So, the reviewer suggest the authors strengthen the captions.

6. Conclusions should not only summarize the main conclusions, but also point out the deficiencies of the proposed method. This could involve providing more insight into the implications of the findings and suggesting avenues for future research based on the results. This would help to demonstrate the significance of the research and its potential impact in the field.

Author Response

This manuscript employed several machine learning methods to predict indoor air temperature in an even-span Mediterranean greenhouse. The manuscript is of technical sound. Before possible publication, several concerns must be addressed

Thank you for your comments. We have carefully reviewed your feedback and made revisions to the paper based on your valuable suggestions. Your input has been incredibly helpful in improving the quality of our work, and we appreciate the time and effort you took to provide it.

 

The Abstract is too long and tedious. The authors are suggested to rewrite the Abstract, which should clearly show the problems of current studies, motivation of the paper, main steps of the method, and the results of this study

Thank you so much for your comment. The abstract section was edited and we removed some excess texts.

 

In Introduction part, the authors need to not only review previous studies, but also analyze the shortcoming of these studies. In addition, the motivation the paper is weak,. At last, the contributions of the paper should be concluded and shown in a point-to-point format

Thank you so much for your comment. The introduction section was edited based on your comments and at the end of it, the contributions of the paper was concluded and shown in a point-to-point format.

 

The font size in figures are incongruous. For example, the font size in Figure 4 is too small to see. In addition, the content of Figure 5 is very large, but the font size in Figure 5 is small

Thank you for your comment. We have made some changes to the paper based on the editor's suggestion. Figure 5 has been removed. Regarding Figure 4, we tried to reduce the font size, but the text is quite long, so we improved the image quality instead. This should allow readers to zoom in and read the text more easily. We appreciate your feedback and hope these changes address your concerns.

 

Generally, Equations (7)-(10) is known to most readers. It is suggested to remove these Equations

Thank you so much for your comment. Section 2.5 was edited based on the above comment.

 

All figure and table captions are too simple and hard for readers to understand the table/figure independently to the context. So, the reviewer suggest the authors strengthen the captions

Thank you so much for your comment. All the captions of tables and figures were edited.

 

Conclusions should not only summarize the main conclusions, but also point out the deficiencies of the proposed method. This could involve providing more insight into the implications of the findings and suggesting avenues for future research based on the results. This would help to demonstrate the significance of the research and its potential impact in the field

Thank you so much for your comment. The conclusion section was edited and some extra texts were removed.

Reviewer 3 Report

In this paper, RBF, GPR, and SVM models were used to predict the indoor air temperature in a greenhouse near the Mediterranean Sea. Since the RBF model is more stable than the GPR and SVM models, it is concluded that energy and cost reduction is possible when applied to the greenhouse control system.

The summary evaluation of this paper is as follows.

In the introduction section, the importance of temperature maintenance in the greenhouse was mentioned, and several AI-based data processing technologies were reviewed accordingly. The review was specific, and several papers were reviewed to secure the validity of the paper.

In the material and method section, data from the greenhouse actually installed was collected and compared with the AI model.

As for the prediction model, the accuracy of the indoor air temperature prediction model was confirmed by securing the adequacy of the three models using RBF (radial basis function), GPR (Gaussian process regression), and Support Vector Machine (SVM).

In the conclusion and discussion, the reasons and results for selecting the RBF model that exhibits the most optimal performance and climatic conditions were confirmed. In addition, input parameters were secured through RMSE, MAPE, TSSE, and EF to predict accuracy, resulting in reliable results.

In conclusion, the applicability and usefulness of the RBF model, which can be excellently used for the important factors in selecting a machine learning model and the temperature setting of the greenhouse, were demonstrated. It is judged that this can be of great help in automatic temperature control in greenhouses.

It is judged that the reference form needs to be modified to the MDPI form.

According to the above review, this paper applied three models for greenhouse temperature prediction, selected an appropriate model, and verified it. Therefore, it is judged that there is no big problem to be published in Horiculturae Journal.

Author Response

In this paper, RBF, GPR, and SVM models were used to predict the indoor air temperature in a greenhouse near the Mediterranean Sea. Since the RBF model is more stable than the GPR and SVM models, it is concluded that energy and cost reduction is possible when applied to the greenhouse control system.

The summary evaluation of this paper is as follows.

In the introduction section, the importance of temperature maintenance in the greenhouse was mentioned, and several AI-based data processing technologies were reviewed accordingly. The review was specific, and several papers were reviewed to secure the validity of the paper.

In the material and method section, data from the greenhouse actually installed was collected and compared with the AI model.

As for the prediction model, the accuracy of the indoor air temperature prediction model was confirmed by securing the adequacy of the three models using RBF (radial basis function), GPR (Gaussian process regression), and Support Vector Machine (SVM).

In the conclusion and discussion, the reasons and results for selecting the RBF model that exhibits the most optimal performance and climatic conditions were confirmed. In addition, input parameters were secured through RMSE, MAPE, TSSE, and EF to predict accuracy, resulting in reliable results.

In conclusion, the applicability and usefulness of the RBF model, which can be excellently used for the important factors in selecting a machine learning model and the temperature setting of the greenhouse, were demonstrated. It is judged that this can be of great help in automatic temperature control in greenhouses.

It is judged that the reference form needs to be modified to the MDPI form.

According to the above review, this paper applied three models for greenhouse temperature prediction, selected an appropriate model, and verified it. Therefore, it is judged that there is no big problem to be published in Horiculturae Journal.

Thank you very much for taking the time to provide your valuable comments. We have carefully reviewed your feedback and made revisions to the paper and reference section based on the standards of MDPI publication. Your contributions have been incredibly helpful in improving the quality of our work, and we sincerely appreciate your efforts. Thank you once again for your thoughtful suggestions.