Uncertainty calibration of building energy models by combining approximate Bayesian computation and machine learning algorithms

2020 ◽  
Vol 268 ◽  
pp. 115025
Author(s):  
Chuanqi Zhu ◽  
Wei Tian ◽  
Baoquan Yin ◽  
Zhanyong Li ◽  
Jiaxin Shi
Keyword(s):  
Machine Learning ◽  
Approximate Bayesian Computation ◽  
Learning Algorithms ◽  
Building Energy ◽  
Machine Learning Algorithms ◽  
Bayesian Computation ◽  
Energy Models ◽  
Approximate Bayesian

scholarly journals Forecasting Algorithms and Optimization Strategies for Building Energy Management & Demand Response

Proceedings ◽  
2018 ◽  
Vol 2 (15) ◽  
pp. 1133 ◽  
Author(s):  
Fanlin Meng ◽  
Kui Weng ◽  
Balsam Shallal ◽  
Xiangping Chen ◽  
Monjur Mourshed
Keyword(s):  
Machine Learning ◽  
Energy Management ◽  
Demand Response ◽  
Learning Algorithms ◽  
Building Energy ◽  
Machine Learning Algorithms ◽  
Future Research ◽  
Research Directions ◽  
Building Energy Management ◽  
Future Research Directions

In this paper, we look at the key forecasting algorithms and optimization strategies for the building energy management and demand response management. By conducting a combined and critical review of forecast learning algorithms and optimization models/algorithms, current research gaps and future research directions and potential technical routes are identified. To be more specific, ensemble/hybrid machine learning algorithms and deep machine learning algorithms are promising in solving challenging energy forecasting problems while large-scale and distributed optimization algorithms are the future research directions for energy optimization in the context of smart buildings and smart grids.

scholarly journals Constraining the recent star formation history of galaxies: an approximate Bayesian computation approach

2020 ◽  
Vol 635 ◽  
pp. A136 ◽  
Author(s):  
G. Aufort ◽  
L. Ciesla ◽  
P. Pudlo ◽  
V. Buat
Keyword(s):  
Star Formation ◽  
Error Rate ◽  
Approximate Bayesian Computation ◽  
Spectral Energy Distribution ◽  
Stellar Mass ◽  
Machine Learning Algorithms ◽  
Star Formation History ◽  
High Signal ◽  
Bayesian Computation ◽  
Approximate Bayesian

Although galaxies are found to follow a tight relation between their star formation rate and stellar mass, they are expected to exhibit complex star formation histories (SFH) with short-term fluctuations. The goal of this pilot study is to present a method that identifies galaxies that undergo strong variation in star formation activity in the last ten to some hundred million years. In other words, the proposed method determines whether a variation in the last few hundred million years of the SFH is needed to properly model the spectral energy distribution (SED) rather than a smooth normal SFH. To do so, we analyzed a sample of COSMOS galaxies with 0.5 <  z <  1 and log M* >  8.5 using high signal-to-noise ratio broadband photometry. We applied approximate Bayesian computation, a custom statistical method for performing model choice, which is associated with machine-learning algorithms to provide the probability that a flexible SFH is preferred based on the observed flux density ratios of galaxies. We present the method and test it on a sample of simulated SEDs. The input information fed to the algorithm is a set of broadband UV to NIR (rest-frame) flux ratios for each galaxy. The choice of using colors is made to remove any difficulty linked to normalization when classification algorithms are used. The method has an error rate of 21% in recovering the correct SFH and is sensitive to SFR variations larger than 1 dex. A more traditional SED-fitting method using CIGALE is tested to achieve the same goal, based on fit comparisons through the Bayesian information criterion, but the best error rate we obtained is higher, 28%. We applied our new method to the COSMOS galaxies sample. The stellar mass distribution of galaxies with a strong to decisive evidence against the smooth delayed-τ SFH peaks at lower M* than for galaxies where the smooth delayed-τ SFH is preferred. We discuss the fact that this result does not come from any bias due to our training. Finally, we argue that flexible SFHs are needed to be able to cover the largest possible SFR-M* parameter space.

Automated data-driven modeling of building energy systems via machine learning algorithms

2019 ◽  
Vol 202 ◽  
pp. 109384 ◽  
Author(s):  
Martin Rätz ◽  
Amir Pasha Javadi ◽  
Marc Baranski ◽  
Konstantin Finkbeiner ◽  
Dirk Müller
Keyword(s):  
Machine Learning ◽  
Learning Algorithms ◽  
Energy Systems ◽  
Building Energy ◽  
Machine Learning Algorithms ◽  
Data Driven ◽  
Building Energy Systems ◽  
Data Driven Modeling

scholarly journals PipeMaster: inferring population divergence and demographic history with approximate Bayesian computation and supervised machine-learning in R

2020 ◽  
Author(s):  
Marcelo Gehara ◽  
Guilherme G. Mazzochinni ◽  
Frank Burbrink
Keyword(s):  
Machine Learning ◽  
Approximate Bayesian Computation ◽  
Demographic History ◽  
Divergence Time ◽  
R Package ◽  
Supervised Machine Learning ◽  
Population Divergence ◽  
Bayesian Computation ◽  
Approximate Bayesian ◽  
Recent Bottleneck

AbstractUnderstanding population divergence involves testing diversification scenarios and estimating historical parameters, such as divergence time, population size and migration rate. There is, however, an immense space of possible highly parameterized scenarios that are difsficult or impossible to solve analytically. To overcome this problem researchers have used alternative simulation-based approaches, such as approximate Bayesian computation (ABC) and supervised machine learning (SML), to approximate posterior probabilities of hypotheses. In this study we demonstrate the utility of our newly developed R-package to simulate summary statistics to perform ABC and SML inferences. We compare the power of both ABC and SML methods and the influence of the number of loci in the accuracy of inferences; and we show three empirical examples: (i) the Muller’s termite frog genomic data from Southamerica; (ii) the cottonmouth and (iii) and the copperhead snakes sanger data from Northamerica. We found that SML is more efficient than ABC. It is generally more accurate and needs fewer simulations to perform an inference. We found support for a divergence model without migration, with a recent bottleneck for one of the populations of the southamerican frog. For the cottonmouth we found support for divergence with migration and recent expansion and for the copperhead we found support for a model of divergence with migration and recent bottleneck. Interestingly, by using an SML method it was possible to achieve high accuracy in model selection even when several models were compared in a single inference. We also found a higher accuracy when inferring parameters with SML.

Sign in / Sign up

Export Citation Format

Share Document