Abstract
This paper offers an efficient and robust concept for a decentralized bad data processing, able to supply in real-time a power system state estimator with a repaired measurement set. Corrupted measurement vectors are funneled through a denoising auto-associative neural network in order to project the biased vector back to the data manifold learned during an offline training process. In order to improve accuracy, a maximum similarity with the solution manifold, measured with Correntropy, is searched for by a meta-heuristic. The extreme robustness and scalability of the process is demonstrated in multiple characteristic case studies.

Abstract

Abstract

Abstract

Abstract
During many years, battery models have been proposed with different levels of accuracy and complexity. In some cases, simple low-order aggregated battery pack models may be more appropriate and feasible than complex physic-chemical or high-order multi-cell battery pack models. For example: in early stages of the system design process, in non-focused battery applications, or whenever low configuration effort or low computational complexity is a requirement. The latter may be the case of Electrical Equivalent Circuit Models (EECM) suitable for energy optimization purposes at a system level in the context of energy management or sizing problem of energy storage systems. In this paper, an improved parametrization method for Li-ion linear static EECMs based on the so called concept of direct current resistance (DCR) is presented. By drawing on a DCR-based parametrization, the influence of both diffusion polarization effects and changing of Open-Circuit Voltage (OCV) are virtually excluded on the estimation of the battery's inner resistance. This results in a parametrization that only accounts for pure ohmic and charge transfer effects, which may be beneficial, since these effects dominate the battery dynamic power response in the range of interest of many applications, including electro-mobility. Model validation and performance evaluation is achieved in simulations by comparison with other low and high order EECM battery models over a dynamic driving profile. Significant improvements in terms of terminal voltage and power losses estimation may be achieved by a DCR-based parametrization, which in its simplest form may only require one short pulse characterization test within a relatively wide range of SoCs and currents. Experimental data from a 53 Ah Li-ion pouch cell produced by Kokam (Type SLPB 120216216) with Nickel Manganese Cobalt oxide (NMC) cathode material is used.

Abstract
With the rising research in emotionally believable agents, several advances in agent technology have been made, ranging from interactive virtual agents to emotional mechanism simulations and emotional agent architectures. However, creating an emotionally believable agent capable of emotional thought is still largely out of reach. It has been proposed that being able to accurately model human emotion would allow agents to mimic human behaviour while these models are studied to create more accurate theoretical models. In light of these challenges, we present a general method for human emotional state modelling in interactive environments. The proposed method employs a three-layered classification process to model the arousal and valence (i.e., hedonic) emotional components, based on four selected psychophysiological metrics. Additionally, we also developed a simplified version of our system for use in real-time systems and low-fidelity applications. The modelled emotional states by both approaches compared favourably with a manual approach following the current best practices reported in the literature while also improving on its predictive ability. The obtained results indicate we are able to accurately predict human emotional states, both in offline and online scenarios with varying levels of granularity; thus, providing a transversal method for modelling and reproducing human emotional profiles.