Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

Session Overview
Poster introductions 19: Energy consumption & optimization
Thursday, 26/Aug/2021:
11:35am - 12:00pm

Session Chair: Dr. Rongling Li, DTU Byg
Location: Room 5 - Room 019, Building: 116

Show help for 'Increase or decrease the abstract text size'
11:35am - 11:38am

Energy performance assessment of passive buildings in future climatic scenarios: the case of study of the childcare centre in Putignano (Bari, Italy).

Ludovica Maria Campagna1, Francesco Carlucci1, Piero Russo2, Francesco Fiorito1

1Department of Civil, Environmental, Land, Building Engineering and Chemistry (DICATECh), Polytechnic University of Bari, via Orabona 4, 70125 Bari, Italy; 2Studio Tecnico “Piero Russo – Ingegneria ed Architettura”, Putignano (BA)

The building sector should be a primary target for GreenHouse Gas (GHG) emissions mitigation efforts, as it accounts for 36% of final energy use and 39% of energy and process-related emissions. The most diffuse and effective mitigation strategies include the construction of nearly zero energy buildings (nZEBs) and the energy retrofit of the existing building stock. Among existing buildings, particular attention should be paid to school buildings, for several reasons. Educational buildings are among the most diffuse public buildings in Europe, most of them built decades ago, resulting in high potential in terms of refurbishment effectiveness. Moreover, schools cover a social function and require high levels of indoor environmental quality. In this field, the research activity is intense, but retrofit strategies are still conceived considering historical weather data, which could not represent correctly present and future climate patterns, thus reducing the retrofit effectiveness.
In this work, an energy retrofit to “Passivhaus standard” of a childcare centre located in the Mediterranean area is analysed through dynamic simulations. A post-retrofit building model is simulated using typical weather conditions (Typical Meteorological Year – TMY) and then compared with the same model simulated in future weather scenarios, created using the morphing method proposed by Belcher et al.
The analyses aim to assess if the technical solutions currently adopted on the basis of the TMY will lead to acceptable energy performance and indoor comfort conditions in future decades. The results highlight that global warming will reduce the acceptability of standard solutions, requiring the installation of active cooling systems or the adoption of passive cooling strategies, in particular over a long-term horizon.
Furthermore, a sensitivity analysis of different design solutions is performed showing that high-performance design solutions, in future weather conditions, will have a lower efficacy on the reduction of building energy consumptions.

11:38am - 11:41am

Quantile regression using gradient boosted decision trees for daily residential energy load disaggregation

Benoit Delcroix, Simon Sansregret, Gilbert Larochelle Martin, Ahmed Daoud

Hydro-Quebec Research Institute, Canada

The building sector (residential, commercial and institutional) is responsible for approximately one-third of the total energy consumption, worldwide. This sector is undergoing a major digital transformation, buildings being more and more equipped with connected devices such as smart meters and IoT devices. This transformation offers the opportunity to better monitor and optimize building operations. In the province of Quebec (Canada), the vast majority of buildings are equipped with smart meters providing electricity usage data every 15 minutes. A current major challenge is to disaggregate the different energy use (heating, domestic hot water, appliances, etc.) from smart meter data, a discipline called non-intrusive load monitoring in literature. In this work, the objective is to develop and validate a model identifying the daily share of each energy use using building information (number of rooms, number of people, appliances, etc.), weather data and smart meter data. Input features are selected and ordered based on a composite index including the correlation coefficient, the feature importance given by a decision tree, and the predictive power score. Two modelling approaches based on quantile regression are tested: linear regression and gradient boosted decision trees (non-linear). Compared to traditional ordinary least squares regression, quantile methods inherently provide more robustness (median vs. mean) and confidence intervals (i.e., quantiles). Both models are trained and validated using separate datasets collected in 8 houses in Canada where metering and sub-metering were performed during a whole year. Preliminary results on the validation dataset indicate that the maximum absolute error between the actual and predicted values in 95 % of the cases is around 10 % (non-linear method) and 14 % (linear method). Further research will focus on improving the modelling methodology, better understanding the model’s behaviour (winter vs. summer, workday vs. weekend), and highlighting the potential applications such as fault detection and diagnosis.

11:41am - 11:44am

Study of power demand forecasting of a hospital by ensemble machine learning

Mayuka Nakai1, Ryozo Ooka1, Shintaro Ikeda2

1University of Tokyo, Japan; 2Tokyo University of Science

To save energy in existing buildings, there are hardware measures such as replacement of equipment and software measures such as improving the operational efficiency of equipment. Since hardware measures cost a lot of money, energy saving by optimizing operations is becoming more popular. We can save energy in buildings by forecasting power demand because equipment can be operated more efficiently like utilizing heat storage to lower the peak. Many attempts to predict building power consumption by machine learning so far have used simulation values ​​in virtual buildings with no measurement errors or defects in the data. These models tend to have higher accuracy scores but have the risk of overfitting and possibly malfunction for missing data or outliers.

To avoid the above problems, this study proposes an ensemble machine learning algorithm to forecast power demand for a hospital building in Japan.

Using the power consumption data of almost four and a half years, initial predictions were made by using algorithms such as Deep Neural Network (DNN), Random Forest (RF) and Long Short-Term Memory (LSTM). Each algorithm was combined to create ensemble models that take the weighted average of the predicted values. As a result, we overcome the issues of each individual method, such as the delay in responding to sudden changes in numerical values ​​with DNN or long calculation time with LSTM and achieved higher prediction accuracies. We then selected the appropriate method for forecasting the power demand of real buildings based on accuracy.

In future studies, we will apply the same methodology to predict steam, hot water, heating, and cooling load and employ transfer learning algorithms to forecast demand for other buildings.

Contact and Legal Notice · Contact Address:
Privacy Statement · Conference: IBPC 2021
Conference Software - ConfTool Pro 2.6.142+TC
© 2001 - 2021 by Dr. H. Weinreich, Hamburg, Germany