Bayesian optimization

Bayesian optimization is a sequential design strategy for global optimization of black-box functions that does not assume any functional forms. It is usually employed to optimize expensive-to-evaluate functions. History The term is generally attributed to Jonas Mockus and is coined in his work from a series of publications on global optimization in the 1970s and 1980s. Strategy Bayesian optimization is typically used on problems of the form \max_{x \in A} f(x), where A is a set of points, x, which rely upon less than 20 dimensions (\mathbb{R}^d, d \le 20), and whose membership can easily be evaluated. Bayesian optimization is particularly advantageous for problems where f(x) is difficult to evaluate due to its computational cost. The objective function, f, is cont

Iterative simulation and optimization approach for energy performance evaluation of ground source heat pump systems

Sarah Bourgarel

The energy transition in Switzerland specifies reduction of the country’s primary energy consumption and greenhouse gas emissions. In the building sector, one way to achieve this is through adoption of efficient and sustainable technologies. With this aim, the building energy systems are gaining complexity. Modelling tools are becoming increasingly important in the design and evaluation of such systems. Two modelling methods are commonly used, simulation and optimization, which substantially differ in their approach and purpose. Simulation is a descriptive tool used to virtually represent systems’ behaviour under given conditions and operation strategies. Optimization approaches explore the possible scenarios under given system limits; they determine the best solution by optimizing an objective function described in mathematical form. In this project, the interactions and complementary use of these methods are investigated through evaluation of the newly installed energy systems of multi-familiy houses built in Zurich. The systems include renewable sources from solar and geothermal energy. The main components are a heat pump coupled with a borehole heat exchanger (BHE) and solar photovoltaic panels (PV) or hybrid photovoltaic panels (PV/T) for electricity production. Two system variants including different rates of borehole regeneration resulting from free cooling of the houses or injection of heat produced by PV/T are evaluated. The systems are divided into interlinked blocks representing the main components of the system: the borehole heat exchanger and surrounding ground, the heat pump, the storage tanks, the piping systems and the photovoltaic installations. The blocks are simulated with an hourly time scale. Energy consumption, self-consumption of on site produced electricity as well as performance degradation due to the long term operation are some of the results obtained from simulation. Monitoring data are used for calibration and validation of the simulation model. Separate optimization models of the evaluated energy systems are then developed. They are improved based on system characteristics obtained from the simulation model. In this way, optimal operation strategies which take into account specific operational limits are identified. Different levels of precision of parameters integration from the simulation results (constant over the year and hourly defined) are implemented and the influences on the optimization results are investigated. The results of the optimization are subsequently implemented in the simulation model. The results of the simulation found that the boreholes are sustainably exploited due the conservative design of the system according the simulated operation conditions. The influence of the regeneration is noticeable, however the simulation of the long term operation of a hypothetical variant without regeneration induces only a limited efficiency decrease and can also be considered as sustainable. The optimization results showed that the self consumption of electricity produced on site can be significantly improved by adapting the heat pump production profile with the electricity production; this was achieved through judicious management of the storage tanks. The analysis of the different level of parameter integration in the optimization model revealed that the optimization results are sensitive to the level of precision of the parameters. The iterative process allowed to combine the strengths of both modelling methods: the simulation was used to precisely describe the systems behaviour. Operational profiles from optimization were subsequently used in the simulation. The outcome of this process helped to better understand the system, identify optimal operating strategies while considering system limitations, as well as presenting

2017

Dynamic Probabilistic Pruning: A General Framework for Hardware-Constrained Pruning at Different Granularities

Lizeth Gonzalez Carabarin, Alexandre Schmid

Unstructured neural network pruning algorithms have achieved impressive compression ratios. However, the resulting-typically irregular-sparse matrices hamper efficient hardware implementations, leading to additional memory usage and complex control logic that diminishes the benefits of unstructured pruning. This has spurred structured coarse-grained pruning solutions that prune entire feature maps or even layers, enabling efficient implementation at the expense of reduced flexibility. Here, we propose a flexible new pruning mechanism that facilitates pruning at different granularities (weights, kernels, and feature maps) while retaining efficient memory organization (e.g., pruning exactly k-out-of- n weights for every output neuron or pruning exactly k-out-of-n kernels for every feature map). We refer to this algorithm as dynamic probabilistic pruning (DPP). DPP leverages the Gumbel-softmax relaxation for differentiable k-out-of-n sampling, facilitating end-to-end optimization. We show that DPP achieves competitive compression ratios and classification accuracy when pruning common deep learning models trained on different benchmark datasets for image classification. Relevantly, the dynamic masking of DPP facilitates for joint optimization of pruning and weight quantization in order to even further compress the network, which we show as well. Finally, we propose novel information-theoretic metrics that show the confidence and pruning diversity of pruning masks within a layer.

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC2022

Iterative simulation and optimization approach for energy performance evaluation of ground source heat pump systems

Sarah Bourgarel

2017

Contaminant source localization via Bayesian global optimization

Iterative simulation and optimization approach for energy performance evaluation of ground source heat pump systems

Dynamic Probabilistic Pruning: A General Framework for Hardware-Constrained Pruning at Different Granularities

Iterative simulation and optimization approach for energy performance evaluation of ground source heat pump systems

Dynamic Probabilistic Pruning: A General Framework for Hardware-Constrained Pruning at Different Granularities

Contaminant source localization via Bayesian global optimization