Evaluation of Urban-Scale Building Energy-Use Models and Tools—Application for the City of Fribourg, Switzerland

Reviewer comment

Author response

Major changes:

For the lines 118-122 of the paper (anonymized monitoring data):

All the investigation is based on ‘calibrated space heating consumption data’ -not on real data- because of privacy concerns (totally understandable); however, no mention is done about the original description of the data (i.e. monthly and at house level), nor the calibration process.

We thank you for your comment. An explanation of the calibration methodology has been added on lines 169-176. The sentences added are the following:

“The calibration was based on annual heat demand data, for which measurements were available on a per-building basis. The shares of this energy used for space heating and domestic hot water were estimated with the methodology contained in the Swiss norms, which consider the number of occupants. Buildings were grouped into clusters according to their normalized space heating demand and occupancy type, and a search algorithm was used to find the optimal value of the unintended air infiltration rate (ach) within each cluster, with which the buildings were finally simulated.”

For the lines 210-213 of the paper (ach in the different models) and Table 6 (ach and ach*):

The air infiltration rate is considered as function of construction period for the EN model, calibrated with measured data for the CS tool (reference) and the ML model refers to the calibrated ones. In which way this consideration of ach benefits to ML model (same values than reference) with respect to EN model (different values than reference)? This situation can be described with Table 6, where Building ID 4397 has a double ach value for ML model than for EN model; for building ID 761 this difference is of three times! They are very high!.

Thank you for pointing this out. We have changed the sentence on lines 231-234 and added a further description of the situation on lines 415-419.

For the lines 259-261 of the paper (training features for ML):

“The model was trained on hourly data using a combination of building features and climate data with a lag of 3 hours, for a total of 29 inputs”. What building features and climate data are used?

Thank you for your observation. We have added the list of input features used in the ML model in lines 284-287.

Table 3, Uwall (column 5) for period ‘From 2010’ (last row):

The value of 1.35 is completely out of context and this one is a crucial input.

Thank you for your comment. An incorrect U_wall value was inserted in Table 3 and it has been updated.

For the lines 301-302 (no night interruption in EN model) and the lines 407-408 of the paper (high consumption at night):

‘The heating system is always on’ (line 301), which is far away from reality.

‘In the colder months the heating system was always switched on, with high consumption between 11 pm and 6 am’. ¿This situation is accordance with real data?

Thank you for pointing this out. The Authors confirm that for this case study the heating system is always on, and the hourly profile described in this work is in accordance with the real data. A clarification was added in lines 328-331.

In global:

Sections numbering is wrong.

Sections numbering was updated.

Minor changes:

For lines 139-143 of the paper (brief description of ML model):

Reference 38: “XGBoost (Extreme Gradient Boost) and LSTM (Long Short Term Memory) provided the most accurate load prediction in the shallow and deep learning category”. LightGBM, which is the gradient boosting machine used by the authors, is not mention in this paper (Reference 38). It is true that both algorithms are gradient boosting machines; however, the former is a top performer while the performance of the last one is unknown (according to Reference 38).

Thank you for this suggestion. The reference of a work where LightGBM resulted among the top performing algorithms for energy related predictions  was added.

For line 177 of the paper (buildings classification):

The selected buildings are classified in nine classes attending their building shape and period of construction, but nine construction periods are considered, so… ¿building shape is taken in account for the buildings classification or not? (if buildings shape refers to compact condominiums, detached houses and row-houses, there should be 27 classes).

Thank you for pointing this out. In this work, only compact condominiums were used for the energy simulation. The sentence was updated as follows in order to clarify it: “…taking into consideration: (i) the building shape, only compact condominiums, which represent the most common building typology of Fribourg, were selected for the energy simulation; (ii) and the construction periods, the buildings were classified into nine classes. For each zone a cluster of compact condominiums with different construction periods was selected.”(lines 190-193)

For lines 391-394 of the paper (MAPE vs Tae):

MAPE is analyzed versus the external air temperature and these lines remit to Table 7, where MAE is shown but not MAPE.

Thank you for your comment. Table 7 shows the MAE, therefore the text was updated.

Table 7 (Tae and Tae,avg column):

There are two external air temperatures; ¿what is the difference between them?

Thank you for your comment. A sentence in lines 435-436 was added to better describe the values in Table 7: “In Table 7 the external air temperature (T_ae) refers to the selected day and the average values of T_ae,avg and I_sol,avg refer to the month.”

Errata:

Line 102: The number (‘2.’) seems to be wrong (there is no title for this section) and it is supposed to be the cause of the wrong sections numbering.

Sections numbering was updated.

Line 128: ‘follows’ instead of ‘follow’.

Thank you, the word was updated.

Line 176: ‘nine’ instead of ‘eight’; if there are ten zones and only one is discarded (zone 3), then should be nine zones left.

Thank you for pointing this out. The following sentence was added in lines 189-196 to clarify it: “… about 300 of them were selected among the nine of ten zones … . A second selection was made to discard the anomalous data in which the geometric characteristics of buildings elaborated in GIS did not correspond with the CitySim database; consequently the selected buildings used have become 200 located in eight zones (the buildings in zone 2 did not meet the requirement).”

Line 410: ‘midday’ instead of ’12 am’ (midday and midnight are preferred as they avoid the confusion with 12 am or 12 pm).

Thank you for this suggestion, we have changed the time references.

Suggestions:

Line 31: Building sector is made up of residential subsector and non-residential subsector (also commonly referred as commercial and public service sector or tertiary sector). The term ‘civil sector’ is not usual in energy statistics.

Thank you, the term has been corrected.

Line 85: ‘(Swiss Federal Institute of Technology Lausanne)’ after the EPFL acronym (I suppose to be this Institution).

Thank you. The acronym has been specified: EPFL (Ecole Polytechnique Fédérale de Lausanne).

Line 190: ‘aspect ratio’ instead of ‘urban canyon effect’; the former is the right term for H/W, while the last term is a complex phenomenon (i.e. “the result of building and street architecture on airflow in the street” [Farrell, W.J.et al. 2015, https://doi.org/10.1016/j.buildenv.2015.05.004]).

Thank you for pointing this out. The text was updated and the reference was inserted in the paper.

Lines 217-218 of the paper: the ML and EN models use the SVF and the H/W ratio. ¿Why the MOS (Main Orientation of the Streets) is not considered instead if SVF? (The buildings surface-to-volume ratio (S/V), the aspect ratio (H/W), and the main orientation of the streets (MOS) express the compactness of the built environment and the type of the surrounding open spaces” [Xu et al. 2019, https://doi.org/10.3390/su11133683]).

In previous works [1], the Authors designed the EN model using: (i) the MOS and the H/H_avg to take into account the influence of the incident solar irradiance; (ii) the SVF to describe the solar exposition and the thermal radiation lost to the sky from the built environment.

After this first version, the EN model was optimized by replacing MOS and H/H_avg with solar height and H/W. In particular, the incident solar irradiance on walls was assessed considering the hourly variation in the shadow percentage for each building as a function of the solar height and the aspect ratio H/W [2]. The Authors found that this new version of the model was more accurate.

[1]         Mutani G, Todeschi V. Building energy modeling at neighborhood scale. Energy Effic 2020. https://doi.org/10.1007/s12053-020-09882-4.

[2]         Mutani G, Todeschi V, Beltramino S. Energy Consumption Models at Urban Scale to Measure Energy Resilience. Sustain 2020. https://doi.org/10.3390/su12145678.

Reviewer comment

Author response

The article is very well written and expresses the contents very clearly.

However, the authors are encouraged to consider the following suggestions:

The Authors would like to thank the Reviewer for his/her appreciation. The manuscript has been improved according to your comments and suggestions.

Line 85: specify the initials EPFL: Ecole polytechnique fédérale de Lausanne

Thank you for your comment. The acronym was specified: EPFL (Ecole Polytechnique Fédérale de Lausanne).

Line 102: 2… put the paragraph title

The position of section 2 was wrong, this number was removed.

Lines 128-132 can be anticipated and placed at the end of the introduction

Since the position of section 2 was wrong, the paragraph in lines 128-132 is at the end of the introduction (section 1).

Line 165 why was zone 3 excluded from the simulations?

Thank you for pointing this out. The following sentence was added in lines 178-179 to clarify it. “The monitoring data was available for every zone except zone 3, which was therefore excluded from the simulations (not having the measured data it would not have been possible to evaluate the precision of the tested models).”

Figures 3 a and b improve the definition

The definition of Figures 3a and 3b was improved.

Figure 5a improve the quality

The quality of Figures 5a was improved.

Figure 5b put the color legend

The legend in Figures 5b was added.

Improve the quality of figures 9 (a, b, c, d) and figures 10 (a, b, c, d)

The quality of Figures 9 (a, b, c, d) and Figures 10 (a, b, c, d) was improved.

Reviewer comment

Author response

This paper is related to the investigation of two existing urban simulation models (machine learning and GIS based) for predicting the hourly space heating consumption of residential buildings. The accuracies of these models were carried out against anonymized monitoring data derived from CitySIM for the city of Fribourg. Further, a sensitivity analysis using the Morris method was carried out on the GIS based model for assessing the impact of input variables on space heating consumption and to identify possible optimization opportunities of the existing model.

The research theme of urban scale energy efficiency is, as such, of great importance, also in the framework of current European EPBD standards related to the designing of Nearly Zero Energy Buildings and Communities. The paper presents a clear research methodology and useful results for the scientific community. Paper is well structured and well written.

The Authors would like to thank the Reviewer for his/her appreciation. The manuscript has been improved according to your comments and suggestions.

Some minor/specific comments are as follows:

·        I suggest making the title more appropriate, for example ‘Evaluation of Urban scale building energy-use models and tools – Application for the City of Fribourg, Switzerland’

Thank you for pointing this out. The title was updated as “Evaluation of Urban Scale Building Energy-Use Models and Tools – Application for the City of Fribourg, Switzerland”.

·        Authors are suggested to extend the literature review part on the urban scale energy modelling, for example see the following paper (or else) for cross references:

Ferrari et al. (2019), Energy, Volume 175, 15 May 2019, Pages 1130-1137.

Thank you for this suggestion. The reference Ferrari et al. (2019) was added in the text, furthermore other useful references (indicated below) to extend the literature review were inserted (lines 59-62).

-        Akbari, K.; Jolai, F.; Ghaderi, S.F. Optimal design of distributed energy system in a neighborhood under uncertainty. Energy 2016, 116, 567–582.

-        Al-Shammari, E.T.; Keivani, A.; Shamshirband, S.; Mostafaeipour, A.; Yee, P.L.; Petković, D.; Ch, S. Prediction of heat load in district heating systems by Support Vector Machine with Firefly searching algorithm. Energy 2016, 95, 266–273.

·        It is not clear how the space heating and DHW systems are treated in the studied models since the whole paper is focused on heating energy consumption (rather than heating energy demand). I suggest including some information regarding this issue.

Thank you for pointing this out. In this work the space heating energy demand and consumption were simulated (the models are not considering DHW). To clarify this concept a paragraph was added in lines 128-134. The following sentence was also added in line 233 “the systems’ efficiency for the space heating was assumed equal to 0.90”.

·        Corrections are required for the decimal points which are not uniform. I mean sometimes these are expressed as comma not the dot.

Thank you for your comment. The decimal points in the text were updated (using dot).