Sequential design strategies for mean response surface metamodeling via stochastic kriging with adaptive exploration and exploitation

Chen, Xi; Qiang, Zexuan

doi:10.1016/j.ejor.2017.03.042

Cited by 26 publications

(20 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The standard approach of allocating a uniform number of replicates leaves plenty of room for improvement. One exception is Chen andZhou (2014, 2017) who proposed several criteria to explore the replication/exploration trade-off, but only for a finite set of candidate designs.…”

Section: Introductionmentioning

confidence: 99%

Replication or Exploration? Sequential Design for Stochastic Simulation Experiments

et al. 2018

View full text Add to dashboard Cite

We investigate the merits of replication, and provide methods for optimal design (including replicates), with the goal of obtaining globally accurate emulation of noisy computer simulation experiments. We first show that replication can be beneficial from both design and computational perspectives, in the context of Gaussian process surrogate modeling. We then develop a lookahead based sequential design scheme that can determine if a new run should be at an existing input location (i.e., replicate) or at a new one (explore). When paired with a newly developed heteroskedastic Gaussian process model, our dynamic design scheme facilitates learning of signal and noise relationships which can vary throughout the input space. We show that it does so efficiently, on both computational and statistical grounds. In addition to illustrative synthetic examples, we demonstrate performance on two challenging real-data simulation experiments, from inventory management and epidemiology.

show abstract

Section: Introductionmentioning

confidence: 99%

Replication or Exploration? Sequential Design for Stochastic Simulation Experiments

et al. 2018

View full text Add to dashboard Cite

show abstract

“…Focusing on the uncertainty of the latent GP is a common strategy in sequential design (especially when the data likelihood is Gaussian), see for example Ankenman et al (2010); Chen and Zhou (2017). Another common measure for constructing sequential designs is the posterior predictive entropy (Kapoor et al, 2007) which is the preferred uncertainty measure in the active learning framework.…”

Section: Adaptive Batching Using the Posterior Gp Variancementioning

confidence: 99%

Probabilistic bisection with spatial metamodels

Rodríguez

Ludkovski

2020

European Journal of Operational Research

View full text Add to dashboard Cite

“…For continuous outputs, such as the minimum voltage, peak current, peak load, aggregated demand and others, a typical approach is to metamodel the mean response [65][66][67]. This may be done by using all replications as independent observations, or simply the calculated mean for each design point, such as U min in Figure 4.…”

Section: Selection Of Output Variablesmentioning

confidence: 99%

“…In many cases, however, additional properties of the distribution are of interest, such as the variance, a quantile, or the probability of achieving some threshold. For instance, there exist applications of Kriging in metamodeling jointly the mean and variance response, or quantiles of stochastic simulation outputs [67]. In the grid impact context, voltage violations would preferably be assessed in terms of the probability of a violation occurring.…”

Section: Selection Of Output Variablesmentioning

confidence: 99%

Towards metamodeling the neighborhood-level grid impact of low-carbon technologies

Protopapadaki

Sælens

2019

Energy and Buildings

View full text Add to dashboard Cite

Heating electrification and distributed renewable generation in the residential sector are among prominent solutions advocated for energy saving and carbon emission reduction. However, research shows these low-carbon technologies may create issues at the low-voltage (LV) distribution grid. High-level policy assessment currently lacks the support to take into account such local grid restrictions. To achieve this, we propose the use of a probabilistic simulation framework in combination with metamodeling, that allows to assess the potential LV grid impact for a wide range of cases. The probabilistic framework is first presented, which is developed for Belgian residential neighborhoods with air-source heat pumps and rooftop PV, based on previous work. Given the complexity and computational requirements of this approach, the paper furthermore proposes metamodeling as a technique to obtain inexpensive evaluation of low-voltage grid impact indicators, suitable for high-level assessments. Although metamodeling is extensively used in various engineering domains, no application in district-level grid-related indicators is available. Consequently, this paper's focus lies on discussing the various steps and options of the metamodeling procedure, while emphasizing problemspecific challenges. Lastly, the proposed metamodeling methodology is used for training simple metamodels for voltage indicators in neighborhood-level LV grids. Linear regression performed fairly well in predicting the minimum voltage levels, though less accurately close to the lower voltage limit, while logistic regression effectively detected feeders with violations.

show abstract

Sequential design strategies for mean response surface metamodeling via stochastic kriging with adaptive exploration and exploitation

Cited by 26 publications

References 46 publications

Replication or Exploration? Sequential Design for Stochastic Simulation Experiments

Replication or Exploration? Sequential Design for Stochastic Simulation Experiments

Probabilistic bisection with spatial metamodels

Towards metamodeling the neighborhood-level grid impact of low-carbon technologies

Contact Info

Product

Resources

About