Search

Energy flexible buildings in combination with demand response will play a key role in the future smart grid. To implement control strategies, which enable demand response, like model predictive control, thermal building models are necessary. Therefore, three lumped capacitance models, are compared with a k-Nearest Neighbor regression model. All models show accurate prediction results, if the operating condition of the building is similar during parameter identification or rather during training and the validation period. Parameter identification of lumped capacitance models is a time-consuming task. Especially for complex lumped capacitance models, the search space for certain parameters has to be reduced to avoid local minima. The investigated k-Nearest Neighbor algorithm has the advantage of easy implementation, very fast training and minimal effort for parameter identification in combination with accurate predictions. But its seasonal dependency is very strong, which can be easily overcome with periodically data update, as it is an instance-based learning algorithm.

Gas turbines are internal combustion engines widely used in industry as main source of power for aircrafts, turbo-generators, turbo-pumps and turbo-compressors. Modelling these engines can help to improve their design and manufacturing processes, as well as to facilitate their operability and maintenance. These eventually lead to manufacturing of gas turbines with lower costs and higher efficiency at the same time. The models may also be employed to unfold nonlinear dynamics of these systems. The aim of this study is to predict the dynamic behavior of a single shaft gas turbine by using open-loop and closed-loop NARX models, which are subsets of artificial neural networks. To set up these models, datasets of significant variables of the gas turbine are used for training, test and validation processes. For this purpose, a comprehensive code is developed in MATLAB programming environment. In addition to the open-loop model, a closed-loop model is set up for multi-step prediction. The results of this study demonstrate the capability of the NARX models in reliable prediction of gas turbines’ dynamic behaviors over different operational ranges.

International audience; Explanations for algorithmic decision systems can take different forms, they can target different types of users with different goals. One of the main challenges in this area is therefore to devise explanation methods that can accommodate this variety of situations. A first step to address this challenge is to allow explainees to express their needs in the most convenient way, depending on their level of expertise and motivation. In this paper, we present a solution to this problem based on a multi-layered approach allowing users to express their requests for explanations at different levels of abstraction. We illustrate the approach with the application of a proof-of-concept system called IBEX to two case studies.

This is a pre-print of an article published in the International Workshop on Fair and Interpretable Learning Algorithms; International audience; Beyond their differences, most black-box explanation methods share a number of features and can be framed in a common structure. We identify these features and propose a generic and parameterized framework which makes it possible to combine them in different ways. This framework has been implemented in a proof of concept system called IBEX (for “Interactive Black-box EXplanation system”). IBEX makes it possible to address a variety of needs of different types of explainees (e.g. local or global explanations, detailed or simple explanations, explanations in the form of counterfactuals, rules, plots, etc.). We illustrate the benefit of the approach in terms of versatility through several case studies corresponding to different types of explainees.

A black-box model for arc-based current interruption, originally developed by the same authors for single-phase short-circuit tests, is here challenged with multipolar circuit breakers in three-phase interruptions. Multipolar devices are built by composition of elementary, single pole components. The three phases adopt three replicas of the same single-phase model, its parameters having been identified in single pole tests. Splitter-less neutral poles adopt a simplified model, deduced by specialization of the original model after suitably dropping splitter plate features. Various breakers and test cases are considered. Comparisons with experimental results in real short-circuit tests show very good agreement, in line with motivating industrial, system-level applications. Indeed, model parameters are obviously single pole design dependent, but they are invariant to different operating conditions and no fine tuning is required to match empirical evidence.

The black-box arc model is known to be a useful technique to simulate dynamic arc-circuit interaction by reflecting arc characteristics. Existing researches have shown that the black-box model has been widely used to analyse the ac arc characteristics of SF(6) and air circuit breakers. Due to the enormous energy and steep rise rate (di/dt) during dc fault, it is important to consider dc arc characteristics. However, there are no examples of black-box models for dc circuit breakers utilised in railway systems and dc microgrid. In this study, the applicability of the black-box model, which was applied to the existing ac arc analysis, was verified for the dc arc analysis. Black-box modelling is applied to datasheet of industrial low-voltage circuit breakers and parametric sweep method was used to select Schwarz model parameters considering the tendency of dc arc current waveform for the dc pole-to-pole fault. The authors also applied the Levenberg–Marquardt algorithm, which is the most extensively used for the optimisation of functional parameters, to the Schwarz model for accurate and reliable arc modelling. As a result, the dc arc feasibility of the black-box model was analysed through simulation results, and a model optimisation technique was proposed.

Plasma Physics and Technology Journal, 6 (2), ISSN:2336-2634

Published in IFAC-PapersOnLine 51(15):688-693, 2018; International audience; The CONtinuous-Time System IDentification (CONTSID) toolbox provides Matlab functions for estimating continuous-time black-box models of dynamical systems from measured data without having to fully characterize the mathematics governing the system behavior. The toolbox includes tools for standard identification of linear continuous-time models such as simple process, transfer functions and state-space models. The toolbox also provides algorithms for more advanced identification such as errors-in-variable (EIV) and closed-loop model estimation or to capture nonlinear system dynamics. This paper presents an overview of the main features of the latest release of the CONTSID toolbox and outlines some recent developments for on-line parameter and time-delay system estimation.

We present a black-box model of miniature circuit breakers having a realistic physical background. In compliance with basic physics, engineering know-how and magneto-hydro-dynamics arc simulations, two sets of differential-algebraic equations are derived to model air plasma chemical dynamics and its interaction with breaker's physical environment (chamber shape, splitter plates, arc traveling). A numerically efficient and accurate electrical model at the breaker terminals is proposed. Its behavior is shown by comparing simulations with short-circuit experiments. Applications are at system level, including simulation of networks with several breakers, such as multiphase switching transients, selectivity, and coordination with fuses, other breakers and surge protection devices.

[EN] We address the novel question of determining which kind of machine learning model is behind the predictions when we interact with a black-box model. This may allow us to identify families of techniques whose models exhibit similar vulnerabilities and strengths. In our method, we first consider how an adversary can systematically query a given black-box model (oracle) to label an artificially-generated dataset. This labelled dataset is then used for training different surrogate models (each one trying to imitate the oracle¿s behaviour). The method has two different approaches. First, we assume that the family of the surrogate model that achieves the maximum Kappa metric against the oracle labels corresponds to the family of the oracle model. The other approach, based on machine learning, consists in learning a meta-model that is able to predict the model family of a new black-box model. We compare these two approaches experimentally, giving us insight about how explanatory and predictable our concept of family is., This material is based upon work supported by the Air Force Office of Scientific Research under award number FA9550-17-1-0287, the EU (FEDER), and the Spanish MINECO under grant TIN 2015-69175-C4-1-R, the Generalitat Valenciana PROMETEOII/2015/013. F. Martinez-Plumed was also supported by INCIBE under grant INCIBEI-2015-27345 (Ayudas para la excelencia de los equipos de investigacion avanzada en ciberseguridad). J. H-Orallo also received a Salvador de Madariaga grant (PRX17/00467) from the Spanish MECD for a research stay at the CFI, Cambridge, and a BEST grant (BEST/2017/045) from the GVA for another research stay at the CFI.