Robust 24 Hours ahead Forecast in a Microgrid: A Real Case Study
, Marco Mussetta 1,†, Emanuele Ogliari 1,†, Sonia Leva 1,†, Luis Fernández-RamĂrez 2 and Pablo GarcĂa-Triviño 2Round 1
Reviewer 1 Report
The authors have to update the acronyms It is still not clear why ANN is used instead of regression, and the algorithm looks to me that does not update with time. if the weather forecast, it is also not clear why a rule base or fuzzy algorithm is not used instead of ANN. the other two methods can be more effective. the paper needs to consider the above to strengthen the needs of ANN, otherwise the novelty seems not high. the error of 24.57% seems very high, need more elaboration.Author Response
Reviewer 1
R1.1 The authors have to update the acronyms
According to this reviewer’s suggestion, acronyms have been checked and fixed through the manuscript, as highlighted in the revised version uploaded
R1.2 It is still not clear why ANN is used instead of regression, and the algorithm looks to me that does not update with time. if the weather forecast, it is also not clear why a rule base or fuzzy algorithm is not used instead of ANN. the other two methods can be more effective. the paper needs to consider the above to strengthen the needs of ANN, otherwise the novelty seems not high.
Following this reviewer’s suggestion, an additional explanation has been included in the Introduction (Section 1) to motivate the choice of the proposed Hybrid ANN approach and its ability to update with time running (also highlighted in the revised manuscript):
Short term PV power prediction based on weather forecast can be obtained by means of simple rule base or fuzzy logic algorithms: for instance in [9] a fuzzy logic model is presented for short term PV forecasting using the measured solar irradiance data; in [10] a model employing fuzzy logic is proposed to forecast global solar energy using the dew-point as the main variable among many other meteorological parameters for different sky-conditions. In [11] a hybrid forecasting algorithm is proposed, based on ANN and fuzzy logic pre-processing, in order to increase forecast accuracy. In particular, the robustness of ANN approach for day-ahead PV forecasting is also assessed in [12].
The accuracy of the prediction is nowadays very good and the error is quite similar to the one of the weather prediction [13]. Moreover, as already mentioned and demonstrated in [14] a machine learning technique is natively able to update with time, after an updated training on recently measured data.
And:
In this work, hybrid ANN model is used, since its effectiveness and robustness have been assessed in [20], where the hybrid methodology for the PV forecasting exploiting clear-sky models and ANN ensembles, based on day-ahead weather forecast, is validated on a real PV plant as a robust and accurate procedure.
R1.3 The error of 24.57% seems very high, need more elaboration.
This value refers to the newly proposed error indicator EMAE, and it was reported in order to show the significant improvement with respect to previous approach (35.12%). The amended text is highlighted in the revised manuscript, before Figure 7:
Similar considerations can be drawn for every indicator reported in Table 2. Since for the recently proposed EMAE and OMAE [14] the availability of results in literature is limited, thus, for the sake of clarity, we chose to refer to indicators usually employed (NMAE, nRMSE, etc…), whose average values are generally better than those reported in literature [13].
Author Response File: 
Author Response.pdf
Reviewer 2 Report
A forecasting method based on neural network with inputs of weather forecasts and clear sky solar irradiance. However, the manuscript can be improved from following aspects:
1) The recent publications on PV power forecasting (robust forecasting) as well on solar irradiance forecasting could be further reviewed.
2) The comparison against the state-of-the-art methods is missing.
Author Response
Reviewer 2
A forecasting method based on neural network with inputs of weather forecasts and clear sky solar irradiance. However, the manuscript can be improved from following aspects:
R2.1 The recent publications on PV power forecasting (robust forecasting) as well on solar irradiance forecasting could be further reviewed.
Literature review has been increased following this reviewer’s suggestion (and also R1.2), as highlighted in the revised manuscript (section 1. Introduction):
Short term PV power prediction based on weather forecast can be obtained by means of simple rule base or fuzzy logic algorithms: for instance in [9] a fuzzy logic model is presented for short term PV forecasting using the measured solar irradiance data; in [10] a model employing fuzzy logic is proposed to forecast global solar energy using the dew-point as the main variable among many other meteorological parameters for different sky-conditions. In [11] a hybrid forecasting algorithm is proposed, based on ANN and fuzzy logic pre-processing, in order to increase forecast accuracy. In particular, the robustness of ANN approach for day-ahead PV forecasting is also assessed in [12].
The accuracy of the prediction is nowadays very good and the error is quite similar to the one of the weather prediction [13]. Moreover, as already mentioned and demonstrated in [14] a machine learning technique is natively able to update with time, after an updated training on recently measured data.
R2.2 The comparison against the state-of-the-art methods is missing.
The aim of this work is to focus on PV power forecasting for microgrids: to the best of authors’ knowledge, previous literature on this specific application is very limited: thus, in this paper the aim is to assess an effective and robust on-grid method with a microgrid implementation, considering in particular the RPP condition and how it affects measurements and data used to train the forecasting model. This has been clarified in Section 1, as highlighted in the revised manuscript.
Author Response File: Author Response.pdf
Reviewer 3 Report
The abstract need to be more informative
The English must be improved
The figures are not clear
Is RPP algorithm is implemented better?
The flow chart is ambiguous
The paper should be gone through plagiarism checker
Author Response
Reviewer 3
R3.1 The abstract need to be more informative
According to this reviewer’s request, the abstract has been reformulated in order to better highlight the main results achieved by the methodology proposed in this paper.
R3.2 The English must be improved
According to this reviewer’s suggestion, the text has been checked and fixed through the whole manuscript, as highlighted in the revised version uploaded.
R3.3 The figures are not clear
According to this reviewer’s request, we have improved all the figures for the sake of legibility and clarity, as highlighted in the revised manuscript.
R3.4 Is RPP algorithm is implemented better?
The RPP condition here considered is requested by the microgrid EMS implementation. This affects the measurements and data used to train the forecasting model. The procedure proposed in the paper was meant to be a simple and effective approach to implement a robust forecasting technique.
R3.5 The flow chart is ambiguous
The flow charts reported in figures 1 and 4 have been improved for legibility and clarity, as highlighted in the revised manuscript.
R3.6 The paper should be gone through plagiarism checker
According to this reviewer’s suggestion, the manuscript has been checked against plagiarism using Turnitin: the crosscheck result is 6%.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The paper can be accepted for publication.
Author Response
Thank you for your time and consideration.
Reviewer 3 Report
The findings are good, but few things need to be improve:
1.Abstract should be more specific.
2.The figures are not represented well.
3.English writtings skills need to be improve.
4.Express the uniqueness of this work.
Author Response
R3.1: Abstract should be more specific.
According to this reviewer’s suggestion, the abstract was improved adding the following sentence:
“The proposed approach has to properly validate measured data, through an effective algorithm, and further refine the power forecast when newer data are available.”
R3.2:The figures are not represented well.
According to this reviewer’s request, we have further improved the figures 1 and 2. In order to improve readability, figures captions have been updated. For the sake of legibility and clarity, the improvements are highlighted in the revised manuscript
R3.3.English writtings skills need to be improve
English was improved
R3.4: Express the uniqueness of this work.
According to the reviewer suggestion, the abstract and introductions has been revised. For the sake of legibility and clarity, the improvements are highlighted in the revised manuscript and here reported:
“In [13], a hybrid ANN model for the PV power forecasting exploiting clear-sky models and ANN ensembles, based on day-ahead weather forecast, is validated on a real PV plant as a robust and accurate procedure.”
“Among the aforementioned methods, it was proven ANN based hybrid methods provide the best solution in terms of accuracy of prediction [16].”
“In addition to the aforementioned drawbacks, new problems arises forecasting the PV power production in microgrid applications. Firstly, data processing and validation is extremely critical due to the need of working in Reduction Power Point (RPP) during several hours of the day to properly follow the load profile.”
“The goal of this paper is to provide a robust methodology to forecast 24 hours in advance the PV power production in a microgrid environment. The issues related to the validation and management of real data through a fast and effective algorithm are addressed. Moreover, particular attention is paid to working conditions such as the microgrid operating state, PV inverters working status and RPP tracking algorithm.”
