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).

Please note that all times are shown in the time zone of the conference. The current conference time is: 21st May 2022, 16:44:41 CEST

 
 
Session Overview
Session
Session F3.3: Buildings paving the way for the energy transition
Time:
Friday, 03/Sept/2021:
13:30 - 15:00

Session Chair: Elli Nikolaidou, University of Bath
Session Chair: Roel Loonen, Eindhoven University of Technology
Location: Cityhall (Belfry) - Room 3

External Resource: Click here to join the livestream. Only registered participants have received the access code for the livestream.
Show help for 'Increase or decrease the abstract text size'
Presentations
13:30 - 13:48

Data-driven estimation of parametric uncertainty of reduced order RC models for building climate control

Anke Uytterhoeven1,2, Ina De Jaeger1,3, Kenneth Bruninx1,2, Dirk Saelens1,3, Lieve Helsen1,2

1EnergyVille, Genk, Belgium; 2KU Leuven, Mechanical Engineering Department, Leuven, Belgium; 3KU Leuven, Civil Engineering Department, Leuven, Belgium

Aim and Approach

(max 200 words)

Current model predictive control (MPC) applications for residential space heating typically rely upon accurate building models, obtained via extensive data acquisition and/or experts’ knowledge. Here, the feedback mechanism of the receding horizon implementation of MPC offers sufficient robustness against the small remaining uncertainties. In contrast, in the context of older, existing residential buildings, one needs to rely on sparse, publicly available data to construct controller models. In that case, the parametric uncertainty of the controller model can become non-negligible, and additional measures might be necessary (Bengea et al. (2011); Oldewurtel et al. (2010,2012); Ioannou and Itard (2015)). However, precisely because of the lack of information, a scientifically sound characterization of the building model and associated parametric uncertainty, as well as an assessment of the impact of the uncertainty, is a challenging task. Therefore, the aim of this paper is to come up with an estimate of the parametric uncertainty of a building controller model in case neither detailed information about the building thermal properties nor experts’ knowledge is available. In addition, the impact of this uncertainty on the resulting optimal space heating strategy is investigated.

Scientific Innovation and Relevance

(max 200 words)

This paper determines a substantiated estimate of the parametric uncertainty for building simulation or controller models generated with the help of a statistical, data-driven building characterization method leveraging publicly available data (De Jaeger et al. (2021)). Since neither detailed information about the building thermal properties nor experts’ knowledge is incorporated, this can be seen as a worst-case assessment. In addition, this paper investigates the impact of the obtained variation in model parameters on the energy demand profile determined by optimal control, and on the resulting yearly energy use, via a Monte Carlo analysis. This is a first step in assessing the importance of building model parametric uncertainty regarding MPC performance.

Preliminary Results and Conclusions

(max 200 words)

The results show that the considered approach gives rise to a rather large uncertainty. The obtained variation in model parameters is shown to markedly affect the optimal space heating control, both in terms of dynamic effects and yearly energy use, thereby indicating the need for improved data acquisition and/or dedicated control strategies that operate robustly under uncertainty.

Main References

(max 200 words)

Bengea, S., V. Adetola, K. Kang, M. J. Liba, D. Vrabie, R. Bitmead, and S. Narayanan (2011). Parameter estimation of a building system model and impact of estimation error on closed-loop performance. In 50th IEEE Conference on Decision and Control and European Control Conference, 2011, Orlando, FL, USA, pp. 5137-5143. IEEE.

Oldewurtel, F., D. Gyalistras, M. Gwerder, C. Jones, A. Parisio, V. Stauch, B. Lehmann, and M. Morari (2010). Increasing energy efficiency in building climate control using weather forecasts and model predictive control. In Clima RHEVA World Congress, 2010, Antalya, Turkey.

Oldewurtel, F., A. Parisio, C. N. Jones, D. Gyalistras, M. Gwerder, V. Stauch, B. Lehmann, and M. Morari (2012). Use of model predictive control and weather forecasts for energy efficient building climate control. Energy and Buildings 45, 15-27.

Ioannou, A. and L. C. Itard (2015). Energy performance and comfort in residential buildings: Sensitivity for building parameters and occupancy. Energy and Buildings 92, 216-233.

De Jaeger, I., J. Lago, and D. Saelens (2021). A probabilistic building characterization method for district energy simulations. Energy and Buildings 230, 110566.



13:48 - 14:06

Formulation and implementation of a model predictive control (MPC) strategy for a PCM-driven building ventilation cooling system

Tao Yang1, Konstantin Filonenko1, Jonathan Dallaire1, Viktor Bue Ljungdahl1, Muhyiddine Jradi1, Esther Kieseritzky2, Felix Pawelz2, Christian Veje1

1Center for Energy Informatics, University of Southern Denmark, Denmark; 2Rubitherm Technologies GmbH, Berlin, Germany

Aim and Approach

(max 200 words)

This paper aims to integrate phase change material (PCM) based ventilation cooling system with a building model and studies feasibility and overall performance of MPC controllers in different context and conditions.

To achieve this, a surrogate model of a PCM ventilation unit is built in Dymola, the model is validated based on results from an actual experimental setup. The performance of the PCM based energy storage system is simulated and its cooling capacity is assessed. Subsequently, the PCM model is integrated with a calibrated gray-box office zone model. The simulation results of the whole system are performed under real occupancy and climate data.

An MPC controller controlling supply airflow rate of ventilation cooling is implemented by running a sequential quadratic programming optimization algorithm on the system functional mock-up unit (FMU). Performance of the entire system is compared with a conventional system employing rule-based control approach in terms of thermal comfort violation and electrical power consumption. The surrogate PCM model serve both as emulation model and control model in this setup.

Sensitivity with respect to control and zone model parameters affecting the MPC performance is inspected including model complexity, optimization horizon, objective function and weather conditions.

Scientific Innovation and Relevance

(max 200 words)

Phase change material, as a storage medium in latent thermal energy storage systems, has received considerable attention in recent research studies and investigations as it is able to reduce overall energy consumption and shift peak loads [1]. Meanwhile, advanced control strategies such as MPC have been proven to facilitate better performance in buildings as compared to conventional control strategies. There is a trend of applying PCM-based energy storage systems and MPC in buildings aiming for improving the performance. Previous research studies of building MPC mainly focus on traditional heating, ventilation, and air conditioning (HVAC) systems such as controlling radiators, cooling and heating coils in ventilation. There are some studies reporting MPC on PCM-based energy storage system, but they either focus on PCM for heating or PCM embedded in building envelope [2-4]. Few studies have been found in literature regarding PCM units for separate ventilation cooling in summer, which forms the basis of this paper’s novelty.

Preliminary Results and Conclusions

(max 200 words)

The surrogate model of the PCM energy storage system yields considerable accuracy of predicting the thermal dynamics as compared with experimental setup measurements. It enables passive cooling and shows capability of regulating indoor thermal comfort when integrated in an office room in summer. Results show that implementing the MPC approach enables enhanced cooling capability of a PCM energy storage system as compared to conventional control. The complexity of the control model has a direct impact on the overall MPC performance in terms of results accuracy and simulation time, while optimization horizon, objective function and weather conditions allow for an appropriate tuning depending on a specific archetype data.

Main References

(max 200 words)

[1] E. Osterman, V. Butala, and U. Stritih, “PCM thermal storage system for ‘free’ heating and cooling of buildings,” Energy Build., vol. 106, pp. 125–133, Nov. 2015

[2] A. C. Papachristou, C. A. Vallianos, V. Dermardiros, A. K. Athienitis, and J. A. Candanedo, “A numerical and experimental study of a simple model-based predictive control strategy in a perimeter zone with phase change material,” Sci. Technol. Built Environ., vol. 24, no. 9, pp. 933–944, Oct. 2018.

[3] G. Serale, M. Fiorentini, A. Capozzoli, P. Cooper, and M. Perino, “Formulation of a model predictive control algorithm to enhance the performance of a latent heat solar thermal system,” Energy Convers. Manag., vol. 173, pp. 438–449, Oct. 2018.

[4] M. Fiorentini, P. Cooper, Z. Ma, and D. A. Robinson, “Hybrid model predictive control of a residential HVAC system with PVT energy generation and PCM thermal storage,” in Energy Procedia, 2015, vol. 83, pp. 21–30.



14:06 - 14:24

Data-driven prediction of flow parameters in a ventilated cavity using high-fidelity CFD simulations

Nina Morozova, Fransesc Xavier Trias Miquel, Roser Capdevila Paramio, Asensio Oliva Llena

​Heat and Mass Transfer Technological Center (CTTC), Universitat Politècnica de Catalunya (UPC), Spain

Aim and Approach

(max 200 words)

In this work, we elaborate data-driven models for predicting the flow parameters in a three-dimensional ventilated cavity with a heated floor. It was first studied experimentally by Blay et al. [1]. Cold air enters the cavity through the inlet at the top of the left wall. The air is discharged at the bottom of the right wall. The bottom wall is hot, while other walls are cold. This configuration is typical for many indoor environmental applications. This study aims to develop an affordable data-driven model, which accurately predicts the airflow parameters. The predictions of the model are based on the results of high-fidelity CFD simulations. The elaborated model could be considered a cheaper alternative for CFD in indoor environmental design and control. The input parameters for the model are the results of CFD simulations with different geometrical aspect ratios, boundary, and initial conditions. The CFD simulations are conducted using an in-house code [2] with symmetry-preserving finite volume discretization on staggered grids. This numerical configuration has provided the best trade-off between the computational cost and accuracy in our previous studies [3, 4]. We compare artificial neural networks (ANN) with support vector regression (SVR) and gradient boosting regression (GBR) using open-source libraries.

Scientific Innovation and Relevance

(max 200 words)

CFD is a reliable tool for indoor environmental applications. However, accurate CFD simulations require large computational resources, whereas significant cost reduction can lead to unreliable results. The high cost prevents CFD from becoming the primary tool for indoor environmental simulations. Our previous findings [3, 4] suggest that the growth of computational resources in the near future would not be enough to make CFD available for routine use in building applications. This means more work is required on developing better models and numerical methods, to reduce the computational cost of the simulations while maintaining accuracy. In our study, we develop machine learning algorithms based on data from previous CFD simulations. The algorithms aim to accurately predict the airflow parameters while having lower than CFD computational cost. The main focus of our research is on investigating the capabilities and limitations of machine learning algorithms as a cheaper alternative to CFD simulations. We compare the computational cost and accuracy of different machine learning algorithms for the prediction of flow parameters in a ventilated cavity.

Preliminary Results and Conclusions

(max 200 words)

We consider a characteristic test case of turbulent(Ra= 2.4×109)mixed convection in a ventilated square cavity. In the previous work [3, 4], we analyzed this test case using different turbulence models, discretization techniques, and mesh resolutions. This test case is difficult to be solved accurately due to the small aspect ratios of the inlet and outlet openings. The LES simulation with staggered symmetry-preserving discretization provided the best trade-off between computational cost and accuracy, while the accuracy of the RANS turbulence model appeared to be insufficient. Reliable, high-fidelity CFD simulations require significant computational resources, thus alternative methods of obtaining high-fidelity simulation results with lower computational cost should be explored. We use the obtained results as the guidance for the turbulence model and the discretization method choice for the CFD simulations, which form the basis of our data-driven model. Moreover, the performed analysis allows investigating the requirements for the minimal set of simulations to build up a reliable data-driven model, such as the quality of CFD (high-fidelity or coarse-grid), the number of datasets, and the principal data components, which affect the prediction the most.

Main References

(max 200 words)

[1] D. Blay, S. Mergui, J. L. Tuhault, and F. Penot. Experimental turbulent mixed convection created by confined buoyant wall jets. In Proceedings of the First European Heat Transfer Conference, UK, pages 821–828, 1992.[2] A. Gorobets, F. X. Trias, M. Soria, and A. Oliva. A scalable parallel Poisson solver for three-dimensional problems with one periodic direction. Computers & Fluids, 39(3):525–538, 2010.

[3] N. Morozova, F. X. Trias, R. Capdevila, C. D. Pérez-Segarra, and A. Oliva. On the feasibility of affordable high-fidelity CFD simulations for indoor environment design and control. Building and Environment,(accepted for publication), 2020.

[4] N. Morozova, R. Capdevila, F. X. Trias, and A. Oliva. On the feasibility of CFD for transient airflow simulations in buildings. In Proceedings of Building Simulation 2019: 16th Conference of IBPSA, September2-4, 2019.



14:24 - 14:42

Evaluating data-driven building stock heat demand forecasting models for energy optimization

Tohid Jafarinejad, Ina De Jaeger, Arash Erfani, Dirk Saelens

Katholieke Universiteit Leuven, Belgium

Aim and Approach

(max 200 words)

The building sector is one of the most energy intensive sectors. In order to decarbonize society, the focus on reducing the energy use of dwellings is important. This reduction can be achieved by looking at individual building level, but many researchers also made assessments on the district and city scale by developing innovative techniques to improve district energy planning, develop optimal controllers for network operation, and demand side management. For these analyses researchers use district models to simulate the energy demand under varying weather conditions and end use decisions. In many cases, the energy management in district level involves non-convex optimization schemes. Predicting these energy needs requires accurate and fast responsive models. In this study, various data-driven models, which forecast the aggregate energy demand of the building stock, are developed based on a residential district data generated by the Modelica simulation environment. To see how well and fast they predict the Modelica generated results, their suitability for the aforementioned applications is assessed based on several Key Performance Indicators (KPIs), namely the coefficient of determination (R2), Normalized Root Mean Square Error (NRMSE), and run time, which ascertains the agility of the developed data-driven model for optimization over a specific time horizon.

Scientific Innovation and Relevance

(max 200 words)

Conventional and prevalent high order white-box models that simulate and forecast the district energy demand are difficult to be deployed for optimization purposes since they are time intensive. In many cases, using surrogate simplified models is proved to be an interesting alternative. Nowadays with the development of building energy management tools and having a huge load of data associated with buildings in a neighborhood at hand, deriving surrogate models for districts is facilitated. These data-driven ersatz models, however a bit less accurate than the conventional white-box models, are precise enough to represent a building stock, while taking a much shorter time to simulate the district demand at each run, and carry out the optimization task. The two main criteria that evaluate the appropriateness of a black-box model for district level optimization applications are the forecast accuracy and the response time (computational load) of the model. In this study, the competence of most frequent black-box models including ANN (Artificial Neural Network) models, SVM (Support Vector Machine) models and etc. are investigated. In contrast to previous studies, this study also takes into account the influence of the district size on the developed models in terms of the aforementioned criteria.

Preliminary Results and Conclusions

(max 200 words)

The black-box techniques for modeling a building stock are divided into two main categories; statistic based and machine learning based models. Although on the one hand machine learning based models are more accurate than statistic based models, they are prone to be over-fit and unable to generalize on new datasets. In this study also the generalization of black-box models is evaluated on test datasets as well. The training dataset is 4380 hourly data points (half a year) of simulation data in a year, and the test dataset is 4380 hourly data points (half a year) of simulation data. Preliminary results demonstrate that through opting a suitable regularization mechanism, lack of generalization could be avoided while maintaining the expedient precision. The outcomes signify that machine learning based models have more edge as compared to statistic based ones. Moreover, they perform more rapidly in comparison to the white-box models, which makes them adept for the intended optimization purposes. Last but not least, in districts with a large number of buildings, Support Vector Machine identification technique shows a remarkable performance in terms of agility (response time) and accuracy both combined together.

Main References

(max 200 words)

Idowu, S., Saguna, S., Åhlund, C., & Schelén, O. (2016). Applied machine learning: Forecasting heat load in district heating system. Energy and Buildings, 133, 478–488. https://doi.org/10.1016/j.enbuild.2016.09.068

Swan, L. G., & Ugursal, V. I. (2009). Modeling of end-use energy consumption in the residential sector: A review of modeling techniques. Renewable and Sustainable Energy Reviews, 13(8), 1819–1835. https://doi.org/10.1016/j.rser.2008.09.033

Xie, J., Li, H., Ma, Z., Sun, Q., Wallin, F., Si, Z., & Guo, J. (2017). Analysis of Key Factors in Heat Demand Prediction with Neural Networks. Energy Procedia, 105, 2965–2970. https://doi.org/10.1016/j.egypro.2017.03.704

Zhao, F., Lee, S. H., & Augenbroe, G. (2016). Reconstructing building stock to replicate energy consumption data. Energy and Buildings, 117, 301–312. https://doi.org/10.1016/j.enbuild.2015.10.001

Zygmunt, M., & Gawin, D. (2019). Application of ANN for analysing a neighbourhood of single-family houses constituting an Energy Cluster. MATEC Web of Conferences, 282(2019), 02072. https://doi.org/10.1051/matecconf/201928202072



 
Contact and Legal Notice · Contact Address:
Privacy Statement · Conference: Building Simulation 2021
Conference Software - ConfTool Pro 2.6.143+TC
© 2001–2022 by Dr. H. Weinreich, Hamburg, Germany