1,721,424 research outputs found
Daylighting driven design: Optimizing Kaleidocycle façade for hot arid climate
No full textFacade design has significant impact on daylight. This paper presents a facade based on origami: kaleidocycle rings that can be morphed enhancing daylight performance in residential spaces, which complies with both LEED V4 and Daylight availability. Daylighting analysis was integrated using Grasshopper, Diva and Genetic optimization for a south-oriented living room facade in Cairo, Egypt, through two phases. First phase dealt with base cases of specific typology. Second phase was conducted using parametric optimization process. Results demonstrate that Kaleidocycle rings of 30 cm size and 64 rotation’s angle reached results that exceed LEED v4 requirements while passing Daylight availability standards
Case Study for Energy Efficiency Measures of Buildings on an Urban Scale
Get PDFThe energy efficiency of existing buildings is one of the challenges launched by the EPBD recast. The RWTH Aachen University accepted this challenge and started the project EnEff: Campus - Roadmap aiming at reducing the specific primary energy consumption of the university campus building stock (about 300 buildings) by 50 % until 2025. For the estimation of refurbishments for this kind of big data, data mining techniques can be used like the CART method (Classification and Regression Tree). In this investigation, the method applied on the RWTH Aachen buildings stock and the estimated results will be compared to results from a simple data mining technique, called visual method. The comparison is performed by using low-order dynamic building model (LOM) performance simulation through the Modelica AixLib. The determined results of the recommendation of the CART method will be discussed and evaluated in this paper
A gap-filling method for room temperature data based on autoencoder neural networks
Get PDFThis study explores the applicability of a deep learning-based approach for reconstructing missing room temperature data from different domains where relatively few training samples are available. For that purpose, the existing convolutional, long short-term memory (LSTM) and feed-forward autoencoders were combined with a suitable domain adaptation procedure. Eventually, the developed models were evaluated on data collected in four buildings with significant differences in thermal mass, design and location. The findings pointed out that the domain adaptation can be conducted effciently by using a small data sample from the target domain. Additionally, the results showed that the proposed model can reconstruct up to 80 % of the missing daily room temperature inputs with RMSE accuracy of 0.6 °C
User Interactions with Environmental Control Systems in Buildings
No full textThe design and operation of energy-efficient systems for indoor environment control (heating, cooling, ventilation, lighting) can benefit from reliable (empirically grounded) information on occupants' actions to bring about changes in the status of building control systems. Specifically, the computational modeling of occupants' control-oriented control actions in building performance simulation applications can be significantly improved based on such empirical information on user behavior. The present paper concerns the type and number of user control actions as related to building systems in two office buildings in Vienn
Computational Derivation of incident irradiance on building facades based on measured global horizontal irradiance data
No full textReliable simulation of buildings' energy performance requires, amongst other things, the availability of detailed information on the magnitudes of incident solar radiation on building facades. In this paper we compare three methods to compute incident vertical irradiance values based on measured global horizontal irradiance values
Development of generative adversarial networks (GANs) for the reconstruction of missing energy data time-series
No full textLAUREA MAGISTRALEQuesto studio esplora l’applicazione di un modello di deep learning, basato sulla tecnologia
Generative Adversarial Network, chiamato Deep Convolutional GAN, nella ricostruzione
di serie temporali di dati energetici mancanti, e, in particolare, di dati di consumo elettrico
degli edifici. Data l’importanza di dati energetici accurati e completi per migliorare
l’efficienza degli edifici e mitigare l’impatto dei cambiamenti climatici, è fondamentale
colmare le lacune nei dati causate da malfunzionamenti dei sensori e anomalie nei sistemi
HVAC. Questo lavoro analizza in dettaglio la relazione tra qualità e quantità dei
dati di input, complessità del modello e approccio di machine learning, fondamentale per
l’efficacia dei metodi data-driven nella ricostruzione dei dati energetici.
Dalla ricerca è emerso che l’ottimizzazione del modello, attraverso il tuning degli iperparametri,
è di notevole importanza per un modello robusto e affidabile, con particolare
attenzione all’equilibrio delle capacità di apprendimento del generatore e del discriminatore,
alla calibrazione dell’ottimizzatore e all’uso di regolatori negli strati convoluzionali.
Le prestazioni del modello sono state analizzate in relazione a variabili quali la dimensione
del set di addestramento e il tasso di corruzione del set e l’uso dell’aumento dei dati nei
set di addestramento. I risultati hanno evidenziato prestazioni promettenti sull’edificio
campione, con un’errore quadratico medio (RMSE) inferiore del 7,84% rispetto al riferimento
comparativo. L’aumento dei dati ha prodotto risultati migliori con una riduzione
dell’8,42% rispetto alla stessa Baseline. Tuttavia, la sua efficacia diminuisce con l’aumento
dei tassi di corruzione, evidenziando le molteplici influenze della complessità e della
disponibilità dei dati.
La capacità di generalizzare il modello su altri edifici, caratterizzati da dati con distribuzione
di densità simile a quella dell’edificio campione, è ancora una sfida aperta, al
fine di ottenere un modello di deep learning adattabile e affidabile nel settore dell’efficienza
energetica degli edifici.This study explores the application of a deep learning model, based on Generative Adversarial
Network technology, called Deep Convolutional GAN, in reconstruction of missing
energy data time series, and, in particular, electricity consumption data of buildings.
Given the importance of accurate and complete energy data to improve building efficiency
and mitigate the impact of climate change, filling data gaps caused by sensor
malfunctions and anomalies in HVAC systems is crucial. This work analyzes in detail the
relationship between quality and quantity of input data, model complexity and machine
learning approach, which is fundamental for the effectiveness of data-driven methods in
reconstructing energy data.
From the research, it emerged that model optimization, through hyperparameter tuning,
is of significant importance for a robust and reliable model, with particular attention to
the balance of the learning capabilities of the generator and discriminator, to the optimizer’s
calibration and to the use of regulators in the convolutional layers.
The performance of the model was analyzed in relation to variables such as the size of the
training set and the corruption rate of the set and the use of data augmentation in the
training data. The results highlighted promising performances on the building sample,
with a root mean squared error (RMSE) 7.84% lower than the comparative baseline. The
data augmentation produced increased results with a reduction of 8.42% compared to the
baseline. However, its effectiveness decreases as corruption rates increase, highlighting
the multiple influences of data complexity and availability.
The ability to generalize of the model on other buildings, characterized by data with
similar density distribution to the sample building, is still an open challenge, in order to
obtain an effectively adaptable deep learning model in building energy efficiency sector
Indoor environment data time-series reconstruction using autoencoder neural networks
No full textLAUREA MAGISTRALELa recente diffusione di sistemi di monitoraggio ed automazione degli edifici, ha reso
disponibili ampie serie temporali di dati che possono essere usate come input in modelli
di analisi data-driven per diversi scopi (e.g. valutazione delle prestazioni, modellazione
del comportamento degli occupanti, fault detection). Questi modelli riescono, di
solito, a fornire previsioni più accurate rispetto a modelli di tipo diretto, poiché si
basano su dati prestazionali reali. Tuttavia, i dati temporali provenienti dai sistemi
edificio-impianto sono spesso caratterizzati da errori o valori mancanti. I metodi
che sono stati usati fino ad ora per gestire queste anomalie hanno spesso portato
ad una diminuzione delle prestazioni dei modelli di analisi data-driven applicati. La
presente tesi propone una soluzione a questo problema, attraverso l’uso e l’analisi di tre
diverse reti neurali di tipo "autoencoder" per ricostruire dati mancanti di temperatura
dell’aria interna provenienti da un edificio-ufficio localizzato ad Aquisgrana, Germania.
I risultati hanno dimostrato la superiorità di questi modelli rispetto ai classici approcci
di tipo numerico. Gli stessi modelli hanno raggiunto prestazioni elevate anche nella
stima diretta di dati futuri di temperatura interna dell’aria.As the number of installed meters in buildings increases, there is a growing number of
data time-series that could be used as input variables to data-driven models. These
models usually capture more accurate as-built system performance than classical
forward approaches, since they rely on real building temporal data. However, building
data sets are often characterized by errors and missing values that could hinder
further energy analysis. Existing research addressed this topic by using methods
that often resulted in poor performance of later applied data-driven models. In this
thesis, three different autoencoder neural networks are trained to reconstruct missing
indoor air temperature time-series in a data set collected in an office building in
Aachen, Germany. The results proved that the proposed models outperform classic
numerical approaches. The same models experienced high performance also in the
direct forecasting of future indoor air temperature data time-series
A computational inquiry into the effectiveness of passive cooling measures in buildings
No full textThis paper applies parametric simulation studies to examine the effectiveness of various passive cooling measures (external and internal shading, natural ventilation, phase change materials) toward the reduction of overheating magnitudes in buildings in the middle-European and Mediterranean context
Generation of detailed sky luminance maps via calibrated digital imaging
No full textReliable prediction of daylight availability in indoor environments via computational simulation requires detailed and accurate sky luminance models. Sky luminance mapping via digital imaging can provide an alternative to high-end research-level sky scanners and thus support the provision of information on sky luminance distribution patterns on a more pervasive basis. In this paper, we compare veriously calibrated sky luminance data derived from real-time digital sky images to photometric measurements. Subsequently we compare the application of a digitally derived sky model with other sky models toward the prediciton of illuminance levels in a room
Calibration of schedules in building simulation using runtime monitoring
No full textnone - see english versionUsing runtime measurements from a monitoring
system, building performance models can be
automatically calibrated on a recurrent basis to
decrease deviations from real world conditions. This
paper describes algorithms to perform such an
automated calibration process. The proposed layerbased
structure of the calibration framework allows
simple component exchangeability and expandability
for implementation of different simulation and
optimization tools. The currently proposed approach
allows calibration of both non-time dependent model
variables (e.g., U-value of a window) and time
dependent variables (e.g., schedule of window
status)
- …
